nuget: bump XamlNameReferenceGenerator from 1.4.2 to 1.5.1 (#4026)

Bumps [XamlNameReferenceGenerator](https://github.com/avaloniaui/Avalonia.NameGenerator) from 1.4.2 to 1.5.1.
- [Release notes](https://github.com/avaloniaui/Avalonia.NameGenerator/releases)
- [Commits](https://github.com/avaloniaui/Avalonia.NameGenerator/compare/1.4.2...1.5.1)

---
updated-dependencies:
- dependency-name: XamlNameReferenceGenerator
  dependency-type: direct:production
  update-type: version-update:semver-minor
...

Signed-off-by: dependabot[bot] <support@github.com>

Signed-off-by: dependabot[bot] <support@github.com>
Co-authored-by: dependabot[bot] <49699333+dependabot[bot]@users.noreply.github.com>

.editorconfig

@@ -89,6 +89,7 @@ csharp_style_conditional_delegate_call = true:suggestion
 # Modifier preferences
 csharp_prefer_static_local_function = true:suggestion
 csharp_preferred_modifier_order = public,private,protected,internal,static,extern,new,virtual,abstract,sealed,override,readonly,unsafe,volatile,async:silent
+csharp_style_prefer_readonly_struct = true
 
 # Code-block preferences
 csharp_prefer_braces = true:silent

.github/ISSUE_TEMPLATE/bug_report.md

@@ -1,6 +1,6 @@
 ---
 name: Bug Report
-about: Something doesn't work correctly in Ryujinx. Note that game-specific issues should be instead posted on the Game Compatibility List at https://github.com/Ryujinx/Ryujinx-Games-List, unless it is a provable regression.
+about: Something doesn't work correctly in Ryujinx. Game-specific issues should be posted at https://github.com/Ryujinx/Ryujinx-Games-List instead, unless it is a provable regression.
 #assignees:
 ---
 
@@ -34,7 +34,7 @@ about: Something doesn't work correctly in Ryujinx. Note that game-specific issu
   - OS: *(e.g. Windows 10)*
   - CPU: *(e.g. i7-6700)*
   - GPU: *(e.g. NVIDIA RTX 2070)*
-  - RAM: *(e.g. 16GB)*
+  - RAM: *(e.g. 16GiB)*
 - Applied Mods : [ Yes (Which ones) / No ]
 
 ### Additional context?

.github/dependabot.yml

@@ -0,0 +1,24 @@
+version: 2
+updates:
+  - package-ecosystem: "github-actions"
+    directory: "/"
+    schedule:
+      interval: weekly
+    labels:
+      - "infra"
+    reviewers:
+      - marysaka
+    commit-message:
+      prefix: "ci"
+
+  - package-ecosystem: nuget
+    directory: /
+    open-pull-requests-limit: 5
+    schedule:
+      interval: daily
+    labels:
+      - "infra"
+    reviewers:
+      - marysaka
+    commit-message:
+      prefix: nuget

.github/workflows/build.yml

@@ -48,44 +48,41 @@ jobs:
       DOTNET_CLI_TELEMETRY_OPTOUT: 1
       RYUJINX_BASE_VERSION: "1.1.0"
     steps:
-      - uses: actions/checkout@v2
-      - uses: actions/setup-dotnet@v1
+      - uses: actions/checkout@v3
+      - uses: actions/setup-dotnet@v3
         with:
-          dotnet-version: 6.0.x
-      - name: Ensure NuGet Source
-        uses: fabriciomurta/ensure-nuget-source@v1
+          dotnet-version: 7.0.x
       - name: Get git short hash
         id: git_short_hash
-        run: echo "::set-output name=result::$(git rev-parse --short "${{ github.sha }}")"
-      - name: Clear
-        run: dotnet clean && dotnet nuget locals all --clear
+        run: echo "result=$(git rev-parse --short "${{ github.sha }}")" >> $GITHUB_OUTPUT
+        shell: bash
       - name: Build
-        run: dotnet build -c "${{ matrix.configuration }}" /p:Version="${{ env.RYUJINX_BASE_VERSION }}" /p:SourceRevisionId="${{ steps.git_short_hash.outputs.result }}" /p:ExtraDefineConstants=DISABLE_UPDATER
+        run: dotnet build -c "${{ matrix.configuration }}" -p:Version="${{ env.RYUJINX_BASE_VERSION }}" -p:SourceRevisionId="${{ steps.git_short_hash.outputs.result }}" -p:ExtraDefineConstants=DISABLE_UPDATER
       - name: Test
-        run: dotnet test -c "${{ matrix.configuration }}"
+        run: dotnet test --no-build -c "${{ matrix.configuration }}"
       - name: Publish Ryujinx
-        run: dotnet publish -c "${{ matrix.configuration }}" -r "${{ matrix.DOTNET_RUNTIME_IDENTIFIER }}" -o ./publish /p:Version="${{ env.RYUJINX_BASE_VERSION }}" /p:DebugType=embedded /p:SourceRevisionId="${{ steps.git_short_hash.outputs.result }}" /p:ExtraDefineConstants=DISABLE_UPDATER Ryujinx --self-contained
+        run: dotnet publish -c "${{ matrix.configuration }}" -r "${{ matrix.DOTNET_RUNTIME_IDENTIFIER }}" -o ./publish -p:Version="${{ env.RYUJINX_BASE_VERSION }}" -p:DebugType=embedded -p:SourceRevisionId="${{ steps.git_short_hash.outputs.result }}" -p:ExtraDefineConstants=DISABLE_UPDATER Ryujinx --self-contained true
         if: github.event_name == 'pull_request'
       - name: Publish Ryujinx.Headless.SDL2
-        run: dotnet publish -c "${{ matrix.configuration }}" -r "${{ matrix.DOTNET_RUNTIME_IDENTIFIER }}" -o ./publish_sdl2_headless /p:Version="${{ env.RYUJINX_BASE_VERSION }}" /p:DebugType=embedded /p:SourceRevisionId="${{ steps.git_short_hash.outputs.result }}" /p:ExtraDefineConstants=DISABLE_UPDATER Ryujinx.Headless.SDL2 --self-contained
+        run: dotnet publish -c "${{ matrix.configuration }}" -r "${{ matrix.DOTNET_RUNTIME_IDENTIFIER }}" -o ./publish_sdl2_headless -p:Version="${{ env.RYUJINX_BASE_VERSION }}" -p:DebugType=embedded -p:SourceRevisionId="${{ steps.git_short_hash.outputs.result }}" -p:ExtraDefineConstants=DISABLE_UPDATER Ryujinx.Headless.SDL2 --self-contained true
         if: github.event_name == 'pull_request'
       - name: Publish Ryujinx.Ava
-        run: dotnet publish -c "${{ matrix.configuration }}" -r "${{ matrix.DOTNET_RUNTIME_IDENTIFIER }}" -o ./publish_ava /p:Version="1.0.0" /p:DebugType=embedded /p:SourceRevisionId="${{ steps.git_short_hash.outputs.result }}" /p:ExtraDefineConstants=DISABLE_UPDATER Ryujinx.Ava
+        run: dotnet publish -c "${{ matrix.configuration }}" -r "${{ matrix.DOTNET_RUNTIME_IDENTIFIER }}" -o ./publish_ava -p:Version="${{ env.RYUJINX_BASE_VERSION }}" -p:DebugType=embedded -p:SourceRevisionId="${{ steps.git_short_hash.outputs.result }}" -p:ExtraDefineConstants=DISABLE_UPDATER Ryujinx.Ava --self-contained true
         if: github.event_name == 'pull_request'
       - name: Upload Ryujinx artifact
-        uses: actions/upload-artifact@v2
+        uses: actions/upload-artifact@v3
         with:
           name: ryujinx-${{ matrix.configuration }}-${{ env.RYUJINX_BASE_VERSION }}+${{ steps.git_short_hash.outputs.result }}-${{ matrix.RELEASE_ZIP_OS_NAME }}
           path: publish
         if: github.event_name == 'pull_request'
       - name: Upload Ryujinx.Headless.SDL2 artifact
-        uses: actions/upload-artifact@v2
+        uses: actions/upload-artifact@v3
         with:
           name: sdl2-ryujinx-headless-${{ matrix.configuration }}-${{ env.RYUJINX_BASE_VERSION }}+${{ steps.git_short_hash.outputs.result }}-${{ matrix.RELEASE_ZIP_OS_NAME }}
           path: publish_sdl2_headless
         if: github.event_name == 'pull_request'
       - name: Upload Ryujinx.Ava artifact
-        uses: actions/upload-artifact@v2
+        uses: actions/upload-artifact@v3
         with:
           name: ava-ryujinx-${{ matrix.configuration }}-${{ env.RYUJINX_BASE_VERSION }}+${{ steps.git_short_hash.outputs.result }}-${{ matrix.RELEASE_ZIP_OS_NAME }}
           path: publish_ava

.github/workflows/nightly_pr_comment.yml

@@ -8,7 +8,7 @@ jobs:
     if: github.event.workflow_run.event == 'pull_request' && github.event.workflow_run.conclusion == 'success'
     runs-on: ubuntu-latest
     steps:
-      - uses: actions/github-script@v3
+      - uses: actions/github-script@v6
         with:
           script: |
             const {owner, repo} = context.repo;
@@ -16,7 +16,7 @@ jobs:
             const pull_head_sha = '${{github.event.workflow_run.head_sha}}';
 
             const issue_number = await (async () => {
-              const pulls = await github.pulls.list({owner, repo});
+              const pulls = await github.rest.pulls.list({owner, repo});
               for await (const {data} of github.paginate.iterator(pulls)) {
                 for (const pull of data) {
                   if (pull.head.sha === pull_head_sha) {
@@ -31,7 +31,7 @@ jobs:
               return core.error(`No matching pull request found`);
             }
 
-            const {data: {artifacts}} = await github.actions.listWorkflowRunArtifacts({owner, repo, run_id});
+            const {data: {artifacts}} = await github.rest.actions.listWorkflowRunArtifacts({owner, repo, run_id});
             if (!artifacts.length) {
               return core.error(`No artifacts found`);
             }
@@ -57,12 +57,12 @@ jobs:
             body += hidden_headless_artifacts;
             body += hidden_debug_artifacts;
 
-            const {data: comments} = await github.issues.listComments({repo, owner, issue_number});
+            const {data: comments} = await github.rest.issues.listComments({repo, owner, issue_number});
             const existing_comment = comments.find((c) => c.user.login === 'github-actions[bot]');
             if (existing_comment) {
               core.info(`Updating comment ${existing_comment.id}`);
-              await github.issues.updateComment({repo, owner, comment_id: existing_comment.id, body});
+              await github.rest.issues.updateComment({repo, owner, comment_id: existing_comment.id, body});
             } else {
               core.info(`Creating a comment`);
-              await github.issues.createComment({repo, owner, issue_number, body});
+              await github.rest.issues.createComment({repo, owner, issue_number, body});
             }

.github/workflows/release.yml

@@ -25,19 +25,15 @@ jobs:
       RYUJINX_TARGET_RELEASE_CHANNEL_REPO: "release-channel-master"
 
     steps:
-      - uses: actions/checkout@v2
-      - uses: actions/setup-dotnet@v1
+      - uses: actions/checkout@v3
+      - uses: actions/setup-dotnet@v3
         with:
-          dotnet-version: 6.0.x
-      - name: Ensure NuGet Source
-        uses: fabriciomurta/ensure-nuget-source@v1
-      - name: Clear
-        run: dotnet clean && dotnet nuget locals all --clear
+          dotnet-version: 7.0.x
       - name: Get version info
         id: version_info
         run: |
-          echo "::set-output name=build_version::${{ env.RYUJINX_BASE_VERSION }}.${{ github.run_number }}"
-          echo "::set-output name=git_short_hash::$(git rev-parse --short "${{ github.sha }}")"
+          echo "build_version=${{ env.RYUJINX_BASE_VERSION }}.${{ github.run_number }}" >> $GITHUB_OUTPUT
+          echo "git_short_hash=$(git rev-parse --short "${{ github.sha }}")" >> $GITHUB_OUTPUT
         shell: bash
       - name: Configure for release
         run: |
@@ -51,9 +47,9 @@ jobs:
         run: "mkdir release_output"
       - name: Publish Windows
         run: |
-          dotnet publish -c Release -r win10-x64 -o ./publish_windows/publish /p:Version="${{ steps.version_info.outputs.build_version }}" /p:SourceRevisionId="${{ steps.version_info.outputs.git_short_hash }}" /p:DebugType=embedded Ryujinx --self-contained
-          dotnet publish -c Release -r win10-x64 -o ./publish_windows_sdl2_headless/publish /p:Version="${{ steps.version_info.outputs.build_version }}" /p:SourceRevisionId="${{ steps.version_info.outputs.git_short_hash }}" /p:DebugType=embedded Ryujinx.Headless.SDL2 --self-contained
-          dotnet publish -c Release -r win10-x64 -o ./publish_windows_ava/publish /p:Version="${{ steps.version_info.outputs.build_version }}" /p:SourceRevisionId="${{ steps.version_info.outputs.git_short_hash }}" /p:DebugType=embedded Ryujinx.Ava --self-contained
+          dotnet publish -c Release -r win10-x64 -o ./publish_windows/publish -p:Version="${{ steps.version_info.outputs.build_version }}" -p:SourceRevisionId="${{ steps.version_info.outputs.git_short_hash }}" -p:DebugType=embedded Ryujinx --self-contained true
+          dotnet publish -c Release -r win10-x64 -o ./publish_windows_sdl2_headless/publish -p:Version="${{ steps.version_info.outputs.build_version }}" -p:SourceRevisionId="${{ steps.version_info.outputs.git_short_hash }}" -p:DebugType=embedded Ryujinx.Headless.SDL2 --self-contained true
+          dotnet publish -c Release -r win10-x64 -o ./publish_windows_ava/publish -p:Version="${{ steps.version_info.outputs.build_version }}" -p:SourceRevisionId="${{ steps.version_info.outputs.git_short_hash }}" -p:DebugType=embedded Ryujinx.Ava --self-contained true
       - name: Packing Windows builds
         run: |
           pushd publish_windows
@@ -71,9 +67,9 @@ jobs:
 
       - name: Publish Linux
         run: |
-          dotnet publish -c Release -r linux-x64 -o ./publish_linux/publish /p:Version="${{ steps.version_info.outputs.build_version }}" /p:SourceRevisionId="${{ steps.version_info.outputs.git_short_hash }}" /p:DebugType=embedded Ryujinx --self-contained
-          dotnet publish -c Release -r linux-x64 -o ./publish_linux_sdl2_headless/publish /p:Version="${{ steps.version_info.outputs.build_version }}" /p:SourceRevisionId="${{ steps.version_info.outputs.git_short_hash }}" /p:DebugType=embedded Ryujinx.Headless.SDL2 --self-contained
-          dotnet publish -c Release -r linux-x64 -o ./publish_linux_ava/publish /p:Version="${{ steps.version_info.outputs.build_version }}" /p:SourceRevisionId="${{ steps.version_info.outputs.git_short_hash }}" /p:DebugType=embedded Ryujinx.Ava --self-contained
+          dotnet publish -c Release -r linux-x64 -o ./publish_linux/publish -p:Version="${{ steps.version_info.outputs.build_version }}" -p:SourceRevisionId="${{ steps.version_info.outputs.git_short_hash }}" -p:DebugType=embedded Ryujinx --self-contained true
+          dotnet publish -c Release -r linux-x64 -o ./publish_linux_sdl2_headless/publish -p:Version="${{ steps.version_info.outputs.build_version }}" -p:SourceRevisionId="${{ steps.version_info.outputs.git_short_hash }}" -p:DebugType=embedded Ryujinx.Headless.SDL2 --self-contained true
+          dotnet publish -c Release -r linux-x64 -o ./publish_linux_ava/publish -p:Version="${{ steps.version_info.outputs.build_version }}" -p:SourceRevisionId="${{ steps.version_info.outputs.git_short_hash }}" -p:DebugType=embedded Ryujinx.Ava --self-contained true
 
       - name: Packing Linux builds
         run: |

ARMeilleure/ARMeilleure.csproj

@@ -1,7 +1,7 @@
 <Project Sdk="Microsoft.NET.Sdk">
 
   <PropertyGroup>
-    <TargetFramework>net6.0</TargetFramework>
+    <TargetFramework>net7.0</TargetFramework>
     <AllowUnsafeBlocks>true</AllowUnsafeBlocks>
   </PropertyGroup>
 

ARMeilleure/CodeGen/RegisterAllocators/AllocationResult.cs

@@ -1,6 +1,6 @@
 namespace ARMeilleure.CodeGen.RegisterAllocators
 {
-    struct AllocationResult
+    readonly struct AllocationResult
     {
         public int IntUsedRegisters { get; }
         public int VecUsedRegisters { get; }

ARMeilleure/CodeGen/RegisterAllocators/CopyResolver.cs

@@ -11,7 +11,7 @@ namespace ARMeilleure.CodeGen.RegisterAllocators
     {
         private class ParallelCopy
         {
-            private struct Copy
+            private readonly struct Copy
             {
                 public Register Dest   { get; }
                 public Register Source { get; }

ARMeilleure/CodeGen/RegisterAllocators/HybridAllocator.cs

@@ -11,7 +11,7 @@ namespace ARMeilleure.CodeGen.RegisterAllocators
 {
     class HybridAllocator : IRegisterAllocator
     {
-        private struct BlockInfo
+        private readonly struct BlockInfo
         {
             public bool HasCall { get; }
 

ARMeilleure/CodeGen/RegisterAllocators/RegisterMasks.cs

@@ -3,7 +3,7 @@ using System;
 
 namespace ARMeilleure.CodeGen.RegisterAllocators
 {
-    struct RegisterMasks
+    readonly struct RegisterMasks
     {
         public int IntAvailableRegisters   { get; }
         public int VecAvailableRegisters   { get; }

ARMeilleure/CodeGen/X86/AssemblerTable.cs

@@ -62,224 +62,225 @@ namespace ARMeilleure.CodeGen.X86
             _instTable = new InstructionInfo[(int)X86Instruction.Count];
 
             //  Name                                             RM/R        RM/I8       RM/I32      R/I64       R/RM        Flags
-            Add(X86Instruction.Add,          new InstructionInfo(0x00000001, 0x00000083, 0x00000081, BadOp,      0x00000003, InstructionFlags.None));
-            Add(X86Instruction.Addpd,        new InstructionInfo(BadOp,      BadOp,      BadOp,      BadOp,      0x00000f58, InstructionFlags.Vex | InstructionFlags.Prefix66));
-            Add(X86Instruction.Addps,        new InstructionInfo(BadOp,      BadOp,      BadOp,      BadOp,      0x00000f58, InstructionFlags.Vex));
-            Add(X86Instruction.Addsd,        new InstructionInfo(BadOp,      BadOp,      BadOp,      BadOp,      0x00000f58, InstructionFlags.Vex | InstructionFlags.PrefixF2));
-            Add(X86Instruction.Addss,        new InstructionInfo(BadOp,      BadOp,      BadOp,      BadOp,      0x00000f58, InstructionFlags.Vex | InstructionFlags.PrefixF3));
-            Add(X86Instruction.Aesdec,       new InstructionInfo(BadOp,      BadOp,      BadOp,      BadOp,      0x000f38de, InstructionFlags.Vex | InstructionFlags.Prefix66));
-            Add(X86Instruction.Aesdeclast,   new InstructionInfo(BadOp,      BadOp,      BadOp,      BadOp,      0x000f38df, InstructionFlags.Vex | InstructionFlags.Prefix66));
-            Add(X86Instruction.Aesenc,       new InstructionInfo(BadOp,      BadOp,      BadOp,      BadOp,      0x000f38dc, InstructionFlags.Vex | InstructionFlags.Prefix66));
-            Add(X86Instruction.Aesenclast,   new InstructionInfo(BadOp,      BadOp,      BadOp,      BadOp,      0x000f38dd, InstructionFlags.Vex | InstructionFlags.Prefix66));
-            Add(X86Instruction.Aesimc,       new InstructionInfo(BadOp,      BadOp,      BadOp,      BadOp,      0x000f38db, InstructionFlags.Vex | InstructionFlags.Prefix66));
-            Add(X86Instruction.And,          new InstructionInfo(0x00000021, 0x04000083, 0x04000081, BadOp,      0x00000023, InstructionFlags.None));
-            Add(X86Instruction.Andnpd,       new InstructionInfo(BadOp,      BadOp,      BadOp,      BadOp,      0x00000f55, InstructionFlags.Vex | InstructionFlags.Prefix66));
-            Add(X86Instruction.Andnps,       new InstructionInfo(BadOp,      BadOp,      BadOp,      BadOp,      0x00000f55, InstructionFlags.Vex));
-            Add(X86Instruction.Andpd,        new InstructionInfo(BadOp,      BadOp,      BadOp,      BadOp,      0x00000f54, InstructionFlags.Vex | InstructionFlags.Prefix66));
-            Add(X86Instruction.Andps,        new InstructionInfo(BadOp,      BadOp,      BadOp,      BadOp,      0x00000f54, InstructionFlags.Vex));
-            Add(X86Instruction.Blendvpd,     new InstructionInfo(BadOp,      BadOp,      BadOp,      BadOp,      0x000f3815, InstructionFlags.Prefix66));
-            Add(X86Instruction.Blendvps,     new InstructionInfo(BadOp,      BadOp,      BadOp,      BadOp,      0x000f3814, InstructionFlags.Prefix66));
-            Add(X86Instruction.Bsr,          new InstructionInfo(BadOp,      BadOp,      BadOp,      BadOp,      0x00000fbd, InstructionFlags.None));
-            Add(X86Instruction.Bswap,        new InstructionInfo(0x00000fc8, BadOp,      BadOp,      BadOp,      BadOp,      InstructionFlags.RegOnly));
-            Add(X86Instruction.Call,         new InstructionInfo(0x020000ff, BadOp,      BadOp,      BadOp,      BadOp,      InstructionFlags.None));
-            Add(X86Instruction.Cmovcc,       new InstructionInfo(BadOp,      BadOp,      BadOp,      BadOp,      0x00000f40, InstructionFlags.None));
-            Add(X86Instruction.Cmp,          new InstructionInfo(0x00000039, 0x07000083, 0x07000081, BadOp,      0x0000003b, InstructionFlags.None));
-            Add(X86Instruction.Cmppd,        new InstructionInfo(BadOp,      BadOp,      BadOp,      BadOp,      0x00000fc2, InstructionFlags.Vex | InstructionFlags.Prefix66));
-            Add(X86Instruction.Cmpps,        new InstructionInfo(BadOp,      BadOp,      BadOp,      BadOp,      0x00000fc2, InstructionFlags.Vex));
-            Add(X86Instruction.Cmpsd,        new InstructionInfo(BadOp,      BadOp,      BadOp,      BadOp,      0x00000fc2, InstructionFlags.Vex | InstructionFlags.PrefixF2));
-            Add(X86Instruction.Cmpss,        new InstructionInfo(BadOp,      BadOp,      BadOp,      BadOp,      0x00000fc2, InstructionFlags.Vex | InstructionFlags.PrefixF3));
-            Add(X86Instruction.Cmpxchg,      new InstructionInfo(0x00000fb1, BadOp,      BadOp,      BadOp,      BadOp,      InstructionFlags.None));
-            Add(X86Instruction.Cmpxchg16b,   new InstructionInfo(0x01000fc7, BadOp,      BadOp,      BadOp,      BadOp,      InstructionFlags.RexW));
-            Add(X86Instruction.Cmpxchg8,     new InstructionInfo(0x00000fb0, BadOp,      BadOp,      BadOp,      BadOp,      InstructionFlags.Reg8Src));
-            Add(X86Instruction.Comisd,       new InstructionInfo(BadOp,      BadOp,      BadOp,      BadOp,      0x00000f2f, InstructionFlags.Vex | InstructionFlags.Prefix66));
-            Add(X86Instruction.Comiss,       new InstructionInfo(BadOp,      BadOp,      BadOp,      BadOp,      0x00000f2f, InstructionFlags.Vex));
-            Add(X86Instruction.Crc32,        new InstructionInfo(BadOp,      BadOp,      BadOp,      BadOp,      0x000f38f1, InstructionFlags.PrefixF2));
-            Add(X86Instruction.Crc32_16,     new InstructionInfo(BadOp,      BadOp,      BadOp,      BadOp,      0x000f38f1, InstructionFlags.PrefixF2 | InstructionFlags.Prefix66));
-            Add(X86Instruction.Crc32_8,      new InstructionInfo(BadOp,      BadOp,      BadOp,      BadOp,      0x000f38f0, InstructionFlags.PrefixF2 | InstructionFlags.Reg8Src));
-            Add(X86Instruction.Cvtdq2pd,     new InstructionInfo(BadOp,      BadOp,      BadOp,      BadOp,      0x00000fe6, InstructionFlags.Vex | InstructionFlags.PrefixF3));
-            Add(X86Instruction.Cvtdq2ps,     new InstructionInfo(BadOp,      BadOp,      BadOp,      BadOp,      0x00000f5b, InstructionFlags.Vex));
-            Add(X86Instruction.Cvtpd2dq,     new InstructionInfo(BadOp,      BadOp,      BadOp,      BadOp,      0x00000fe6, InstructionFlags.Vex | InstructionFlags.PrefixF2));
-            Add(X86Instruction.Cvtpd2ps,     new InstructionInfo(BadOp,      BadOp,      BadOp,      BadOp,      0x00000f5a, InstructionFlags.Vex | InstructionFlags.Prefix66));
-            Add(X86Instruction.Cvtps2dq,     new InstructionInfo(BadOp,      BadOp,      BadOp,      BadOp,      0x00000f5b, InstructionFlags.Vex | InstructionFlags.Prefix66));
-            Add(X86Instruction.Cvtps2pd,     new InstructionInfo(BadOp,      BadOp,      BadOp,      BadOp,      0x00000f5a, InstructionFlags.Vex));
-            Add(X86Instruction.Cvtsd2si,     new InstructionInfo(BadOp,      BadOp,      BadOp,      BadOp,      0x00000f2d, InstructionFlags.Vex | InstructionFlags.PrefixF2));
-            Add(X86Instruction.Cvtsd2ss,     new InstructionInfo(BadOp,      BadOp,      BadOp,      BadOp,      0x00000f5a, InstructionFlags.Vex | InstructionFlags.PrefixF2));
-            Add(X86Instruction.Cvtsi2sd,     new InstructionInfo(BadOp,      BadOp,      BadOp,      BadOp,      0x00000f2a, InstructionFlags.Vex | InstructionFlags.PrefixF2));
-            Add(X86Instruction.Cvtsi2ss,     new InstructionInfo(BadOp,      BadOp,      BadOp,      BadOp,      0x00000f2a, InstructionFlags.Vex | InstructionFlags.PrefixF3));
-            Add(X86Instruction.Cvtss2sd,     new InstructionInfo(BadOp,      BadOp,      BadOp,      BadOp,      0x00000f5a, InstructionFlags.Vex | InstructionFlags.PrefixF3));
-            Add(X86Instruction.Cvtss2si,     new InstructionInfo(BadOp,      BadOp,      BadOp,      BadOp,      0x00000f2d, InstructionFlags.Vex | InstructionFlags.PrefixF3));
-            Add(X86Instruction.Div,          new InstructionInfo(BadOp,      BadOp,      BadOp,      BadOp,      0x060000f7, InstructionFlags.None));
-            Add(X86Instruction.Divpd,        new InstructionInfo(BadOp,      BadOp,      BadOp,      BadOp,      0x00000f5e, InstructionFlags.Vex | InstructionFlags.Prefix66));
-            Add(X86Instruction.Divps,        new InstructionInfo(BadOp,      BadOp,      BadOp,      BadOp,      0x00000f5e, InstructionFlags.Vex));
-            Add(X86Instruction.Divsd,        new InstructionInfo(BadOp,      BadOp,      BadOp,      BadOp,      0x00000f5e, InstructionFlags.Vex | InstructionFlags.PrefixF2));
-            Add(X86Instruction.Divss,        new InstructionInfo(BadOp,      BadOp,      BadOp,      BadOp,      0x00000f5e, InstructionFlags.Vex | InstructionFlags.PrefixF3));
-            Add(X86Instruction.Haddpd,       new InstructionInfo(BadOp,      BadOp,      BadOp,      BadOp,      0x00000f7c, InstructionFlags.Vex | InstructionFlags.Prefix66));
-            Add(X86Instruction.Haddps,       new InstructionInfo(BadOp,      BadOp,      BadOp,      BadOp,      0x00000f7c, InstructionFlags.Vex | InstructionFlags.PrefixF2));
-            Add(X86Instruction.Idiv,         new InstructionInfo(BadOp,      BadOp,      BadOp,      BadOp,      0x070000f7, InstructionFlags.None));
-            Add(X86Instruction.Imul,         new InstructionInfo(BadOp,      0x0000006b, 0x00000069, BadOp,      0x00000faf, InstructionFlags.None));
-            Add(X86Instruction.Imul128,      new InstructionInfo(BadOp,      BadOp,      BadOp,      BadOp,      0x050000f7, InstructionFlags.None));
-            Add(X86Instruction.Insertps,     new InstructionInfo(BadOp,      BadOp,      BadOp,      BadOp,      0x000f3a21, InstructionFlags.Vex | InstructionFlags.Prefix66));
-            Add(X86Instruction.Jmp,          new InstructionInfo(0x040000ff, BadOp,      BadOp,      BadOp,      BadOp,      InstructionFlags.None));
-            Add(X86Instruction.Ldmxcsr,      new InstructionInfo(0x02000fae, BadOp,      BadOp,      BadOp,      BadOp,      InstructionFlags.Vex));
-            Add(X86Instruction.Lea,          new InstructionInfo(BadOp,      BadOp,      BadOp,      BadOp,      0x0000008d, InstructionFlags.None));
-            Add(X86Instruction.Maxpd,        new InstructionInfo(BadOp,      BadOp,      BadOp,      BadOp,      0x00000f5f, InstructionFlags.Vex | InstructionFlags.Prefix66));
-            Add(X86Instruction.Maxps,        new InstructionInfo(BadOp,      BadOp,      BadOp,      BadOp,      0x00000f5f, InstructionFlags.Vex));
-            Add(X86Instruction.Maxsd,        new InstructionInfo(BadOp,      BadOp,      BadOp,      BadOp,      0x00000f5f, InstructionFlags.Vex | InstructionFlags.PrefixF2));
-            Add(X86Instruction.Maxss,        new InstructionInfo(BadOp,      BadOp,      BadOp,      BadOp,      0x00000f5f, InstructionFlags.Vex | InstructionFlags.PrefixF3));
-            Add(X86Instruction.Minpd,        new InstructionInfo(BadOp,      BadOp,      BadOp,      BadOp,      0x00000f5d, InstructionFlags.Vex | InstructionFlags.Prefix66));
-            Add(X86Instruction.Minps,        new InstructionInfo(BadOp,      BadOp,      BadOp,      BadOp,      0x00000f5d, InstructionFlags.Vex));
-            Add(X86Instruction.Minsd,        new InstructionInfo(BadOp,      BadOp,      BadOp,      BadOp,      0x00000f5d, InstructionFlags.Vex | InstructionFlags.PrefixF2));
-            Add(X86Instruction.Minss,        new InstructionInfo(BadOp,      BadOp,      BadOp,      BadOp,      0x00000f5d, InstructionFlags.Vex | InstructionFlags.PrefixF3));
-            Add(X86Instruction.Mov,          new InstructionInfo(0x00000089, BadOp,      0x000000c7, 0x000000b8, 0x0000008b, InstructionFlags.None));
-            Add(X86Instruction.Mov16,        new InstructionInfo(0x00000089, BadOp,      0x000000c7, BadOp,      0x0000008b, InstructionFlags.Prefix66));
-            Add(X86Instruction.Mov8,         new InstructionInfo(0x00000088, 0x000000c6, BadOp,      BadOp,      0x0000008a, InstructionFlags.Reg8Src | InstructionFlags.Reg8Dest));
-            Add(X86Instruction.Movd,         new InstructionInfo(0x00000f7e, BadOp,      BadOp,      BadOp,      0x00000f6e, InstructionFlags.Vex | InstructionFlags.Prefix66));
-            Add(X86Instruction.Movdqu,       new InstructionInfo(0x00000f7f, BadOp,      BadOp,      BadOp,      0x00000f6f, InstructionFlags.Vex | InstructionFlags.PrefixF3));
-            Add(X86Instruction.Movhlps,      new InstructionInfo(BadOp,      BadOp,      BadOp,      BadOp,      0x00000f12, InstructionFlags.Vex));
-            Add(X86Instruction.Movlhps,      new InstructionInfo(BadOp,      BadOp,      BadOp,      BadOp,      0x00000f16, InstructionFlags.Vex));
-            Add(X86Instruction.Movq,         new InstructionInfo(BadOp,      BadOp,      BadOp,      BadOp,      0x00000f7e, InstructionFlags.Vex | InstructionFlags.PrefixF3));
-            Add(X86Instruction.Movsd,        new InstructionInfo(0x00000f11, BadOp,      BadOp,      BadOp,      0x00000f10, InstructionFlags.Vex | InstructionFlags.PrefixF2));
-            Add(X86Instruction.Movss,        new InstructionInfo(0x00000f11, BadOp,      BadOp,      BadOp,      0x00000f10, InstructionFlags.Vex | InstructionFlags.PrefixF3));
-            Add(X86Instruction.Movsx16,      new InstructionInfo(BadOp,      BadOp,      BadOp,      BadOp,      0x00000fbf, InstructionFlags.None));
-            Add(X86Instruction.Movsx32,      new InstructionInfo(BadOp,      BadOp,      BadOp,      BadOp,      0x00000063, InstructionFlags.None));
-            Add(X86Instruction.Movsx8,       new InstructionInfo(BadOp,      BadOp,      BadOp,      BadOp,      0x00000fbe, InstructionFlags.Reg8Src));
-            Add(X86Instruction.Movzx16,      new InstructionInfo(BadOp,      BadOp,      BadOp,      BadOp,      0x00000fb7, InstructionFlags.None));
-            Add(X86Instruction.Movzx8,       new InstructionInfo(BadOp,      BadOp,      BadOp,      BadOp,      0x00000fb6, InstructionFlags.Reg8Src));
-            Add(X86Instruction.Mul128,       new InstructionInfo(BadOp,      BadOp,      BadOp,      BadOp,      0x040000f7, InstructionFlags.None));
-            Add(X86Instruction.Mulpd,        new InstructionInfo(BadOp,      BadOp,      BadOp,      BadOp,      0x00000f59, InstructionFlags.Vex | InstructionFlags.Prefix66));
-            Add(X86Instruction.Mulps,        new InstructionInfo(BadOp,      BadOp,      BadOp,      BadOp,      0x00000f59, InstructionFlags.Vex));
-            Add(X86Instruction.Mulsd,        new InstructionInfo(BadOp,      BadOp,      BadOp,      BadOp,      0x00000f59, InstructionFlags.Vex | InstructionFlags.PrefixF2));
-            Add(X86Instruction.Mulss,        new InstructionInfo(BadOp,      BadOp,      BadOp,      BadOp,      0x00000f59, InstructionFlags.Vex | InstructionFlags.PrefixF3));
-            Add(X86Instruction.Neg,          new InstructionInfo(0x030000f7, BadOp,      BadOp,      BadOp,      BadOp,      InstructionFlags.None));
-            Add(X86Instruction.Not,          new InstructionInfo(0x020000f7, BadOp,      BadOp,      BadOp,      BadOp,      InstructionFlags.None));
-            Add(X86Instruction.Or,           new InstructionInfo(0x00000009, 0x01000083, 0x01000081, BadOp,      0x0000000b, InstructionFlags.None));
-            Add(X86Instruction.Paddb,        new InstructionInfo(BadOp,      BadOp,      BadOp,      BadOp,      0x00000ffc, InstructionFlags.Vex | InstructionFlags.Prefix66));
-            Add(X86Instruction.Paddd,        new InstructionInfo(BadOp,      BadOp,      BadOp,      BadOp,      0x00000ffe, InstructionFlags.Vex | InstructionFlags.Prefix66));
-            Add(X86Instruction.Paddq,        new InstructionInfo(BadOp,      BadOp,      BadOp,      BadOp,      0x00000fd4, InstructionFlags.Vex | InstructionFlags.Prefix66));
-            Add(X86Instruction.Paddw,        new InstructionInfo(BadOp,      BadOp,      BadOp,      BadOp,      0x00000ffd, InstructionFlags.Vex | InstructionFlags.Prefix66));
-            Add(X86Instruction.Palignr,      new InstructionInfo(BadOp,      BadOp,      BadOp,      BadOp,      0x000f3a0f, InstructionFlags.Vex | InstructionFlags.Prefix66));
-            Add(X86Instruction.Pand,         new InstructionInfo(BadOp,      BadOp,      BadOp,      BadOp,      0x00000fdb, InstructionFlags.Vex | InstructionFlags.Prefix66));
-            Add(X86Instruction.Pandn,        new InstructionInfo(BadOp,      BadOp,      BadOp,      BadOp,      0x00000fdf, InstructionFlags.Vex | InstructionFlags.Prefix66));
-            Add(X86Instruction.Pavgb,        new InstructionInfo(BadOp,      BadOp,      BadOp,      BadOp,      0x00000fe0, InstructionFlags.Vex | InstructionFlags.Prefix66));
-            Add(X86Instruction.Pavgw,        new InstructionInfo(BadOp,      BadOp,      BadOp,      BadOp,      0x00000fe3, InstructionFlags.Vex | InstructionFlags.Prefix66));
-            Add(X86Instruction.Pblendvb,     new InstructionInfo(BadOp,      BadOp,      BadOp,      BadOp,      0x000f3810, InstructionFlags.Prefix66));
-            Add(X86Instruction.Pclmulqdq,    new InstructionInfo(BadOp,      BadOp,      BadOp,      BadOp,      0x000f3a44, InstructionFlags.Vex | InstructionFlags.Prefix66));
-            Add(X86Instruction.Pcmpeqb,      new InstructionInfo(BadOp,      BadOp,      BadOp,      BadOp,      0x00000f74, InstructionFlags.Vex | InstructionFlags.Prefix66));
-            Add(X86Instruction.Pcmpeqd,      new InstructionInfo(BadOp,      BadOp,      BadOp,      BadOp,      0x00000f76, InstructionFlags.Vex | InstructionFlags.Prefix66));
-            Add(X86Instruction.Pcmpeqq,      new InstructionInfo(BadOp,      BadOp,      BadOp,      BadOp,      0x000f3829, InstructionFlags.Vex | InstructionFlags.Prefix66));
-            Add(X86Instruction.Pcmpeqw,      new InstructionInfo(BadOp,      BadOp,      BadOp,      BadOp,      0x00000f75, InstructionFlags.Vex | InstructionFlags.Prefix66));
-            Add(X86Instruction.Pcmpgtb,      new InstructionInfo(BadOp,      BadOp,      BadOp,      BadOp,      0x00000f64, InstructionFlags.Vex | InstructionFlags.Prefix66));
-            Add(X86Instruction.Pcmpgtd,      new InstructionInfo(BadOp,      BadOp,      BadOp,      BadOp,      0x00000f66, InstructionFlags.Vex | InstructionFlags.Prefix66));
-            Add(X86Instruction.Pcmpgtq,      new InstructionInfo(BadOp,      BadOp,      BadOp,      BadOp,      0x000f3837, InstructionFlags.Vex | InstructionFlags.Prefix66));
-            Add(X86Instruction.Pcmpgtw,      new InstructionInfo(BadOp,      BadOp,      BadOp,      BadOp,      0x00000f65, InstructionFlags.Vex | InstructionFlags.Prefix66));
-            Add(X86Instruction.Pextrb,       new InstructionInfo(0x000f3a14, BadOp,      BadOp,      BadOp,      BadOp,      InstructionFlags.Vex | InstructionFlags.Prefix66));
-            Add(X86Instruction.Pextrd,       new InstructionInfo(0x000f3a16, BadOp,      BadOp,      BadOp,      BadOp,      InstructionFlags.Vex | InstructionFlags.Prefix66));
-            Add(X86Instruction.Pextrq,       new InstructionInfo(0x000f3a16, BadOp,      BadOp,      BadOp,      BadOp,      InstructionFlags.Vex | InstructionFlags.RexW | InstructionFlags.Prefix66));
-            Add(X86Instruction.Pextrw,       new InstructionInfo(BadOp,      BadOp,      BadOp,      BadOp,      0x00000fc5, InstructionFlags.Vex | InstructionFlags.Prefix66));
-            Add(X86Instruction.Pinsrb,       new InstructionInfo(BadOp,      BadOp,      BadOp,      BadOp,      0x000f3a20, InstructionFlags.Vex | InstructionFlags.Prefix66));
-            Add(X86Instruction.Pinsrd,       new InstructionInfo(BadOp,      BadOp,      BadOp,      BadOp,      0x000f3a22, InstructionFlags.Vex | InstructionFlags.Prefix66));
-            Add(X86Instruction.Pinsrq,       new InstructionInfo(BadOp,      BadOp,      BadOp,      BadOp,      0x000f3a22, InstructionFlags.Vex | InstructionFlags.RexW | InstructionFlags.Prefix66));
-            Add(X86Instruction.Pinsrw,       new InstructionInfo(BadOp,      BadOp,      BadOp,      BadOp,      0x00000fc4, InstructionFlags.Vex | InstructionFlags.Prefix66));
-            Add(X86Instruction.Pmaxsb,       new InstructionInfo(BadOp,      BadOp,      BadOp,      BadOp,      0x000f383c, InstructionFlags.Vex | InstructionFlags.Prefix66));
-            Add(X86Instruction.Pmaxsd,       new InstructionInfo(BadOp,      BadOp,      BadOp,      BadOp,      0x000f383d, InstructionFlags.Vex | InstructionFlags.Prefix66));
-            Add(X86Instruction.Pmaxsw,       new InstructionInfo(BadOp,      BadOp,      BadOp,      BadOp,      0x00000fee, InstructionFlags.Vex | InstructionFlags.Prefix66));
-            Add(X86Instruction.Pmaxub,       new InstructionInfo(BadOp,      BadOp,      BadOp,      BadOp,      0x00000fde, InstructionFlags.Vex | InstructionFlags.Prefix66));
-            Add(X86Instruction.Pmaxud,       new InstructionInfo(BadOp,      BadOp,      BadOp,      BadOp,      0x000f383f, InstructionFlags.Vex | InstructionFlags.Prefix66));
-            Add(X86Instruction.Pmaxuw,       new InstructionInfo(BadOp,      BadOp,      BadOp,      BadOp,      0x000f383e, InstructionFlags.Vex | InstructionFlags.Prefix66));
-            Add(X86Instruction.Pminsb,       new InstructionInfo(BadOp,      BadOp,      BadOp,      BadOp,      0x000f3838, InstructionFlags.Vex | InstructionFlags.Prefix66));
-            Add(X86Instruction.Pminsd,       new InstructionInfo(BadOp,      BadOp,      BadOp,      BadOp,      0x000f3839, InstructionFlags.Vex | InstructionFlags.Prefix66));
-            Add(X86Instruction.Pminsw,       new InstructionInfo(BadOp,      BadOp,      BadOp,      BadOp,      0x00000fea, InstructionFlags.Vex | InstructionFlags.Prefix66));
-            Add(X86Instruction.Pminub,       new InstructionInfo(BadOp,      BadOp,      BadOp,      BadOp,      0x00000fda, InstructionFlags.Vex | InstructionFlags.Prefix66));
-            Add(X86Instruction.Pminud,       new InstructionInfo(BadOp,      BadOp,      BadOp,      BadOp,      0x000f383b, InstructionFlags.Vex | InstructionFlags.Prefix66));
-            Add(X86Instruction.Pminuw,       new InstructionInfo(BadOp,      BadOp,      BadOp,      BadOp,      0x000f383a, InstructionFlags.Vex | InstructionFlags.Prefix66));
-            Add(X86Instruction.Pmovsxbw,     new InstructionInfo(BadOp,      BadOp,      BadOp,      BadOp,      0x000f3820, InstructionFlags.Vex | InstructionFlags.Prefix66));
-            Add(X86Instruction.Pmovsxdq,     new InstructionInfo(BadOp,      BadOp,      BadOp,      BadOp,      0x000f3825, InstructionFlags.Vex | InstructionFlags.Prefix66));
-            Add(X86Instruction.Pmovsxwd,     new InstructionInfo(BadOp,      BadOp,      BadOp,      BadOp,      0x000f3823, InstructionFlags.Vex | InstructionFlags.Prefix66));
-            Add(X86Instruction.Pmovzxbw,     new InstructionInfo(BadOp,      BadOp,      BadOp,      BadOp,      0x000f3830, InstructionFlags.Vex | InstructionFlags.Prefix66));
-            Add(X86Instruction.Pmovzxdq,     new InstructionInfo(BadOp,      BadOp,      BadOp,      BadOp,      0x000f3835, InstructionFlags.Vex | InstructionFlags.Prefix66));
-            Add(X86Instruction.Pmovzxwd,     new InstructionInfo(BadOp,      BadOp,      BadOp,      BadOp,      0x000f3833, InstructionFlags.Vex | InstructionFlags.Prefix66));
-            Add(X86Instruction.Pmulld,       new InstructionInfo(BadOp,      BadOp,      BadOp,      BadOp,      0x000f3840, InstructionFlags.Vex | InstructionFlags.Prefix66));
-            Add(X86Instruction.Pmullw,       new InstructionInfo(BadOp,      BadOp,      BadOp,      BadOp,      0x00000fd5, InstructionFlags.Vex | InstructionFlags.Prefix66));
-            Add(X86Instruction.Pop,          new InstructionInfo(0x0000008f, BadOp,      BadOp,      BadOp,      BadOp,      InstructionFlags.None));
-            Add(X86Instruction.Popcnt,       new InstructionInfo(BadOp,      BadOp,      BadOp,      BadOp,      0x00000fb8, InstructionFlags.PrefixF3));
-            Add(X86Instruction.Por,          new InstructionInfo(BadOp,      BadOp,      BadOp,      BadOp,      0x00000feb, InstructionFlags.Vex | InstructionFlags.Prefix66));
-            Add(X86Instruction.Pshufb,       new InstructionInfo(BadOp,      BadOp,      BadOp,      BadOp,      0x000f3800, InstructionFlags.Vex | InstructionFlags.Prefix66));
-            Add(X86Instruction.Pshufd,       new InstructionInfo(BadOp,      BadOp,      BadOp,      BadOp,      0x00000f70, InstructionFlags.Vex | InstructionFlags.Prefix66));
-            Add(X86Instruction.Pslld,        new InstructionInfo(BadOp,      0x06000f72, BadOp,      BadOp,      0x00000ff2, InstructionFlags.Vex | InstructionFlags.Prefix66));
-            Add(X86Instruction.Pslldq,       new InstructionInfo(BadOp,      0x07000f73, BadOp,      BadOp,      BadOp,      InstructionFlags.Vex | InstructionFlags.Prefix66));
-            Add(X86Instruction.Psllq,        new InstructionInfo(BadOp,      0x06000f73, BadOp,      BadOp,      0x00000ff3, InstructionFlags.Vex | InstructionFlags.Prefix66));
-            Add(X86Instruction.Psllw,        new InstructionInfo(BadOp,      0x06000f71, BadOp,      BadOp,      0x00000ff1, InstructionFlags.Vex | InstructionFlags.Prefix66));
-            Add(X86Instruction.Psrad,        new InstructionInfo(BadOp,      0x04000f72, BadOp,      BadOp,      0x00000fe2, InstructionFlags.Vex | InstructionFlags.Prefix66));
-            Add(X86Instruction.Psraw,        new InstructionInfo(BadOp,      0x04000f71, BadOp,      BadOp,      0x00000fe1, InstructionFlags.Vex | InstructionFlags.Prefix66));
-            Add(X86Instruction.Psrld,        new InstructionInfo(BadOp,      0x02000f72, BadOp,      BadOp,      0x00000fd2, InstructionFlags.Vex | InstructionFlags.Prefix66));
-            Add(X86Instruction.Psrlq,        new InstructionInfo(BadOp,      0x02000f73, BadOp,      BadOp,      0x00000fd3, InstructionFlags.Vex | InstructionFlags.Prefix66));
-            Add(X86Instruction.Psrldq,       new InstructionInfo(BadOp,      0x03000f73, BadOp,      BadOp,      BadOp,      InstructionFlags.Vex | InstructionFlags.Prefix66));
-            Add(X86Instruction.Psrlw,        new InstructionInfo(BadOp,      0x02000f71, BadOp,      BadOp,      0x00000fd1, InstructionFlags.Vex | InstructionFlags.Prefix66));
-            Add(X86Instruction.Psubb,        new InstructionInfo(BadOp,      BadOp,      BadOp,      BadOp,      0x00000ff8, InstructionFlags.Vex | InstructionFlags.Prefix66));
-            Add(X86Instruction.Psubd,        new InstructionInfo(BadOp,      BadOp,      BadOp,      BadOp,      0x00000ffa, InstructionFlags.Vex | InstructionFlags.Prefix66));
-            Add(X86Instruction.Psubq,        new InstructionInfo(BadOp,      BadOp,      BadOp,      BadOp,      0x00000ffb, InstructionFlags.Vex | InstructionFlags.Prefix66));
-            Add(X86Instruction.Psubw,        new InstructionInfo(BadOp,      BadOp,      BadOp,      BadOp,      0x00000ff9, InstructionFlags.Vex | InstructionFlags.Prefix66));
-            Add(X86Instruction.Punpckhbw,    new InstructionInfo(BadOp,      BadOp,      BadOp,      BadOp,      0x00000f68, InstructionFlags.Vex | InstructionFlags.Prefix66));
-            Add(X86Instruction.Punpckhdq,    new InstructionInfo(BadOp,      BadOp,      BadOp,      BadOp,      0x00000f6a, InstructionFlags.Vex | InstructionFlags.Prefix66));
-            Add(X86Instruction.Punpckhqdq,   new InstructionInfo(BadOp,      BadOp,      BadOp,      BadOp,      0x00000f6d, InstructionFlags.Vex | InstructionFlags.Prefix66));
-            Add(X86Instruction.Punpckhwd,    new InstructionInfo(BadOp,      BadOp,      BadOp,      BadOp,      0x00000f69, InstructionFlags.Vex | InstructionFlags.Prefix66));
-            Add(X86Instruction.Punpcklbw,    new InstructionInfo(BadOp,      BadOp,      BadOp,      BadOp,      0x00000f60, InstructionFlags.Vex | InstructionFlags.Prefix66));
-            Add(X86Instruction.Punpckldq,    new InstructionInfo(BadOp,      BadOp,      BadOp,      BadOp,      0x00000f62, InstructionFlags.Vex | InstructionFlags.Prefix66));
-            Add(X86Instruction.Punpcklqdq,   new InstructionInfo(BadOp,      BadOp,      BadOp,      BadOp,      0x00000f6c, InstructionFlags.Vex | InstructionFlags.Prefix66));
-            Add(X86Instruction.Punpcklwd,    new InstructionInfo(BadOp,      BadOp,      BadOp,      BadOp,      0x00000f61, InstructionFlags.Vex | InstructionFlags.Prefix66));
-            Add(X86Instruction.Push,         new InstructionInfo(BadOp,      0x0000006a, 0x00000068, BadOp,      0x060000ff, InstructionFlags.None));
-            Add(X86Instruction.Pxor,         new InstructionInfo(BadOp,      BadOp,      BadOp,      BadOp,      0x00000fef, InstructionFlags.Vex | InstructionFlags.Prefix66));
-            Add(X86Instruction.Rcpps,        new InstructionInfo(BadOp,      BadOp,      BadOp,      BadOp,      0x00000f53, InstructionFlags.Vex));
-            Add(X86Instruction.Rcpss,        new InstructionInfo(BadOp,      BadOp,      BadOp,      BadOp,      0x00000f53, InstructionFlags.Vex | InstructionFlags.PrefixF3));
-            Add(X86Instruction.Ror,          new InstructionInfo(0x010000d3, 0x010000c1, BadOp,      BadOp,      BadOp,      InstructionFlags.None));
-            Add(X86Instruction.Roundpd,      new InstructionInfo(BadOp,      BadOp,      BadOp,      BadOp,      0x000f3a09, InstructionFlags.Vex | InstructionFlags.Prefix66));
-            Add(X86Instruction.Roundps,      new InstructionInfo(BadOp,      BadOp,      BadOp,      BadOp,      0x000f3a08, InstructionFlags.Vex | InstructionFlags.Prefix66));
-            Add(X86Instruction.Roundsd,      new InstructionInfo(BadOp,      BadOp,      BadOp,      BadOp,      0x000f3a0b, InstructionFlags.Vex | InstructionFlags.Prefix66));
-            Add(X86Instruction.Roundss,      new InstructionInfo(BadOp,      BadOp,      BadOp,      BadOp,      0x000f3a0a, InstructionFlags.Vex | InstructionFlags.Prefix66));
-            Add(X86Instruction.Rsqrtps,      new InstructionInfo(BadOp,      BadOp,      BadOp,      BadOp,      0x00000f52, InstructionFlags.Vex));
-            Add(X86Instruction.Rsqrtss,      new InstructionInfo(BadOp,      BadOp,      BadOp,      BadOp,      0x00000f52, InstructionFlags.Vex | InstructionFlags.PrefixF3));
-            Add(X86Instruction.Sar,          new InstructionInfo(0x070000d3, 0x070000c1, BadOp,      BadOp,      BadOp,      InstructionFlags.None));
-            Add(X86Instruction.Setcc,        new InstructionInfo(BadOp,      BadOp,      BadOp,      BadOp,      0x00000f90, InstructionFlags.Reg8Dest));
-            Add(X86Instruction.Sha256Msg1,   new InstructionInfo(BadOp,      BadOp,      BadOp,      BadOp,      0x000f38cc, InstructionFlags.None));
-            Add(X86Instruction.Sha256Msg2,   new InstructionInfo(BadOp,      BadOp,      BadOp,      BadOp,      0x000f38cd, InstructionFlags.None));
-            Add(X86Instruction.Sha256Rnds2,  new InstructionInfo(BadOp,      BadOp,      BadOp,      BadOp,      0x000f38cb, InstructionFlags.None));
-            Add(X86Instruction.Shl,          new InstructionInfo(0x040000d3, 0x040000c1, BadOp,      BadOp,      BadOp,      InstructionFlags.None));
-            Add(X86Instruction.Shr,          new InstructionInfo(0x050000d3, 0x050000c1, BadOp,      BadOp,      BadOp,      InstructionFlags.None));
-            Add(X86Instruction.Shufpd,       new InstructionInfo(BadOp,      BadOp,      BadOp,      BadOp,      0x00000fc6, InstructionFlags.Vex | InstructionFlags.Prefix66));
-            Add(X86Instruction.Shufps,       new InstructionInfo(BadOp,      BadOp,      BadOp,      BadOp,      0x00000fc6, InstructionFlags.Vex));
-            Add(X86Instruction.Sqrtpd,       new InstructionInfo(BadOp,      BadOp,      BadOp,      BadOp,      0x00000f51, InstructionFlags.Vex | InstructionFlags.Prefix66));
-            Add(X86Instruction.Sqrtps,       new InstructionInfo(BadOp,      BadOp,      BadOp,      BadOp,      0x00000f51, InstructionFlags.Vex));
-            Add(X86Instruction.Sqrtsd,       new InstructionInfo(BadOp,      BadOp,      BadOp,      BadOp,      0x00000f51, InstructionFlags.Vex | InstructionFlags.PrefixF2));
-            Add(X86Instruction.Sqrtss,       new InstructionInfo(BadOp,      BadOp,      BadOp,      BadOp,      0x00000f51, InstructionFlags.Vex | InstructionFlags.PrefixF3));
-            Add(X86Instruction.Stmxcsr,      new InstructionInfo(0x03000fae, BadOp,      BadOp,      BadOp,      BadOp,      InstructionFlags.Vex));
-            Add(X86Instruction.Sub,          new InstructionInfo(0x00000029, 0x05000083, 0x05000081, BadOp,      0x0000002b, InstructionFlags.None));
-            Add(X86Instruction.Subpd,        new InstructionInfo(BadOp,      BadOp,      BadOp,      BadOp,      0x00000f5c, InstructionFlags.Vex | InstructionFlags.Prefix66));
-            Add(X86Instruction.Subps,        new InstructionInfo(BadOp,      BadOp,      BadOp,      BadOp,      0x00000f5c, InstructionFlags.Vex));
-            Add(X86Instruction.Subsd,        new InstructionInfo(BadOp,      BadOp,      BadOp,      BadOp,      0x00000f5c, InstructionFlags.Vex | InstructionFlags.PrefixF2));
-            Add(X86Instruction.Subss,        new InstructionInfo(BadOp,      BadOp,      BadOp,      BadOp,      0x00000f5c, InstructionFlags.Vex | InstructionFlags.PrefixF3));
-            Add(X86Instruction.Test,         new InstructionInfo(0x00000085, BadOp,      0x000000f7, BadOp,      BadOp,      InstructionFlags.None));
-            Add(X86Instruction.Unpckhpd,     new InstructionInfo(BadOp,      BadOp,      BadOp,      BadOp,      0x00000f15, InstructionFlags.Vex | InstructionFlags.Prefix66));
-            Add(X86Instruction.Unpckhps,     new InstructionInfo(BadOp,      BadOp,      BadOp,      BadOp,      0x00000f15, InstructionFlags.Vex));
-            Add(X86Instruction.Unpcklpd,     new InstructionInfo(BadOp,      BadOp,      BadOp,      BadOp,      0x00000f14, InstructionFlags.Vex | InstructionFlags.Prefix66));
-            Add(X86Instruction.Unpcklps,     new InstructionInfo(BadOp,      BadOp,      BadOp,      BadOp,      0x00000f14, InstructionFlags.Vex));
-            Add(X86Instruction.Vblendvpd,    new InstructionInfo(BadOp,      BadOp,      BadOp,      BadOp,      0x000f3a4b, InstructionFlags.Vex | InstructionFlags.Prefix66));
-            Add(X86Instruction.Vblendvps,    new InstructionInfo(BadOp,      BadOp,      BadOp,      BadOp,      0x000f3a4a, InstructionFlags.Vex | InstructionFlags.Prefix66));
-            Add(X86Instruction.Vcvtph2ps,    new InstructionInfo(BadOp,      BadOp,      BadOp,      BadOp,      0x000f3813, InstructionFlags.Vex | InstructionFlags.Prefix66));
-            Add(X86Instruction.Vcvtps2ph,    new InstructionInfo(0x000f3a1d, BadOp,      BadOp,      BadOp,      BadOp,      InstructionFlags.Vex | InstructionFlags.Prefix66));
-            Add(X86Instruction.Vfmadd231ps,  new InstructionInfo(BadOp,      BadOp,      BadOp,      BadOp,      0x000f38b8, InstructionFlags.Vex | InstructionFlags.Prefix66));
-            Add(X86Instruction.Vfmadd231sd,  new InstructionInfo(BadOp,      BadOp,      BadOp,      BadOp,      0x000f38b9, InstructionFlags.Vex | InstructionFlags.Prefix66 | InstructionFlags.RexW));
-            Add(X86Instruction.Vfmadd231ss,  new InstructionInfo(BadOp,      BadOp,      BadOp,      BadOp,      0x000f38b9, InstructionFlags.Vex | InstructionFlags.Prefix66));
-            Add(X86Instruction.Vfmsub231sd,  new InstructionInfo(BadOp,      BadOp,      BadOp,      BadOp,      0x000f38bb, InstructionFlags.Vex | InstructionFlags.Prefix66 | InstructionFlags.RexW));
-            Add(X86Instruction.Vfmsub231ss,  new InstructionInfo(BadOp,      BadOp,      BadOp,      BadOp,      0x000f38bb, InstructionFlags.Vex | InstructionFlags.Prefix66));
-            Add(X86Instruction.Vfnmadd231ps, new InstructionInfo(BadOp,      BadOp,      BadOp,      BadOp,      0x000f38bc, InstructionFlags.Vex | InstructionFlags.Prefix66));
-            Add(X86Instruction.Vfnmadd231sd, new InstructionInfo(BadOp,      BadOp,      BadOp,      BadOp,      0x000f38bd, InstructionFlags.Vex | InstructionFlags.Prefix66 | InstructionFlags.RexW));
-            Add(X86Instruction.Vfnmadd231ss, new InstructionInfo(BadOp,      BadOp,      BadOp,      BadOp,      0x000f38bd, InstructionFlags.Vex | InstructionFlags.Prefix66));
-            Add(X86Instruction.Vfnmsub231sd, new InstructionInfo(BadOp,      BadOp,      BadOp,      BadOp,      0x000f38bf, InstructionFlags.Vex | InstructionFlags.Prefix66 | InstructionFlags.RexW));
-            Add(X86Instruction.Vfnmsub231ss, new InstructionInfo(BadOp,      BadOp,      BadOp,      BadOp,      0x000f38bf, InstructionFlags.Vex | InstructionFlags.Prefix66));
-            Add(X86Instruction.Vpblendvb,    new InstructionInfo(BadOp,      BadOp,      BadOp,      BadOp,      0x000f3a4c, InstructionFlags.Vex | InstructionFlags.Prefix66));
-            Add(X86Instruction.Xor,          new InstructionInfo(0x00000031, 0x06000083, 0x06000081, BadOp,      0x00000033, InstructionFlags.None));
-            Add(X86Instruction.Xorpd,        new InstructionInfo(BadOp,      BadOp,      BadOp,      BadOp,      0x00000f57, InstructionFlags.Vex | InstructionFlags.Prefix66));
-            Add(X86Instruction.Xorps,        new InstructionInfo(BadOp,      BadOp,      BadOp,      BadOp,      0x00000f57, InstructionFlags.Vex));
+            Add(X86Instruction.Add,           new InstructionInfo(0x00000001, 0x00000083, 0x00000081, BadOp,      0x00000003, InstructionFlags.None));
+            Add(X86Instruction.Addpd,         new InstructionInfo(BadOp,      BadOp,      BadOp,      BadOp,      0x00000f58, InstructionFlags.Vex | InstructionFlags.Prefix66));
+            Add(X86Instruction.Addps,         new InstructionInfo(BadOp,      BadOp,      BadOp,      BadOp,      0x00000f58, InstructionFlags.Vex));
+            Add(X86Instruction.Addsd,         new InstructionInfo(BadOp,      BadOp,      BadOp,      BadOp,      0x00000f58, InstructionFlags.Vex | InstructionFlags.PrefixF2));
+            Add(X86Instruction.Addss,         new InstructionInfo(BadOp,      BadOp,      BadOp,      BadOp,      0x00000f58, InstructionFlags.Vex | InstructionFlags.PrefixF3));
+            Add(X86Instruction.Aesdec,        new InstructionInfo(BadOp,      BadOp,      BadOp,      BadOp,      0x000f38de, InstructionFlags.Vex | InstructionFlags.Prefix66));
+            Add(X86Instruction.Aesdeclast,    new InstructionInfo(BadOp,      BadOp,      BadOp,      BadOp,      0x000f38df, InstructionFlags.Vex | InstructionFlags.Prefix66));
+            Add(X86Instruction.Aesenc,        new InstructionInfo(BadOp,      BadOp,      BadOp,      BadOp,      0x000f38dc, InstructionFlags.Vex | InstructionFlags.Prefix66));
+            Add(X86Instruction.Aesenclast,    new InstructionInfo(BadOp,      BadOp,      BadOp,      BadOp,      0x000f38dd, InstructionFlags.Vex | InstructionFlags.Prefix66));
+            Add(X86Instruction.Aesimc,        new InstructionInfo(BadOp,      BadOp,      BadOp,      BadOp,      0x000f38db, InstructionFlags.Vex | InstructionFlags.Prefix66));
+            Add(X86Instruction.And,           new InstructionInfo(0x00000021, 0x04000083, 0x04000081, BadOp,      0x00000023, InstructionFlags.None));
+            Add(X86Instruction.Andnpd,        new InstructionInfo(BadOp,      BadOp,      BadOp,      BadOp,      0x00000f55, InstructionFlags.Vex | InstructionFlags.Prefix66));
+            Add(X86Instruction.Andnps,        new InstructionInfo(BadOp,      BadOp,      BadOp,      BadOp,      0x00000f55, InstructionFlags.Vex));
+            Add(X86Instruction.Andpd,         new InstructionInfo(BadOp,      BadOp,      BadOp,      BadOp,      0x00000f54, InstructionFlags.Vex | InstructionFlags.Prefix66));
+            Add(X86Instruction.Andps,         new InstructionInfo(BadOp,      BadOp,      BadOp,      BadOp,      0x00000f54, InstructionFlags.Vex));
+            Add(X86Instruction.Blendvpd,      new InstructionInfo(BadOp,      BadOp,      BadOp,      BadOp,      0x000f3815, InstructionFlags.Prefix66));
+            Add(X86Instruction.Blendvps,      new InstructionInfo(BadOp,      BadOp,      BadOp,      BadOp,      0x000f3814, InstructionFlags.Prefix66));
+            Add(X86Instruction.Bsr,           new InstructionInfo(BadOp,      BadOp,      BadOp,      BadOp,      0x00000fbd, InstructionFlags.None));
+            Add(X86Instruction.Bswap,         new InstructionInfo(0x00000fc8, BadOp,      BadOp,      BadOp,      BadOp,      InstructionFlags.RegOnly));
+            Add(X86Instruction.Call,          new InstructionInfo(0x020000ff, BadOp,      BadOp,      BadOp,      BadOp,      InstructionFlags.None));
+            Add(X86Instruction.Cmovcc,        new InstructionInfo(BadOp,      BadOp,      BadOp,      BadOp,      0x00000f40, InstructionFlags.None));
+            Add(X86Instruction.Cmp,           new InstructionInfo(0x00000039, 0x07000083, 0x07000081, BadOp,      0x0000003b, InstructionFlags.None));
+            Add(X86Instruction.Cmppd,         new InstructionInfo(BadOp,      BadOp,      BadOp,      BadOp,      0x00000fc2, InstructionFlags.Vex | InstructionFlags.Prefix66));
+            Add(X86Instruction.Cmpps,         new InstructionInfo(BadOp,      BadOp,      BadOp,      BadOp,      0x00000fc2, InstructionFlags.Vex));
+            Add(X86Instruction.Cmpsd,         new InstructionInfo(BadOp,      BadOp,      BadOp,      BadOp,      0x00000fc2, InstructionFlags.Vex | InstructionFlags.PrefixF2));
+            Add(X86Instruction.Cmpss,         new InstructionInfo(BadOp,      BadOp,      BadOp,      BadOp,      0x00000fc2, InstructionFlags.Vex | InstructionFlags.PrefixF3));
+            Add(X86Instruction.Cmpxchg,       new InstructionInfo(0x00000fb1, BadOp,      BadOp,      BadOp,      BadOp,      InstructionFlags.None));
+            Add(X86Instruction.Cmpxchg16b,    new InstructionInfo(0x01000fc7, BadOp,      BadOp,      BadOp,      BadOp,      InstructionFlags.RexW));
+            Add(X86Instruction.Cmpxchg8,      new InstructionInfo(0x00000fb0, BadOp,      BadOp,      BadOp,      BadOp,      InstructionFlags.Reg8Src));
+            Add(X86Instruction.Comisd,        new InstructionInfo(BadOp,      BadOp,      BadOp,      BadOp,      0x00000f2f, InstructionFlags.Vex | InstructionFlags.Prefix66));
+            Add(X86Instruction.Comiss,        new InstructionInfo(BadOp,      BadOp,      BadOp,      BadOp,      0x00000f2f, InstructionFlags.Vex));
+            Add(X86Instruction.Crc32,         new InstructionInfo(BadOp,      BadOp,      BadOp,      BadOp,      0x000f38f1, InstructionFlags.PrefixF2));
+            Add(X86Instruction.Crc32_16,      new InstructionInfo(BadOp,      BadOp,      BadOp,      BadOp,      0x000f38f1, InstructionFlags.PrefixF2 | InstructionFlags.Prefix66));
+            Add(X86Instruction.Crc32_8,       new InstructionInfo(BadOp,      BadOp,      BadOp,      BadOp,      0x000f38f0, InstructionFlags.PrefixF2 | InstructionFlags.Reg8Src));
+            Add(X86Instruction.Cvtdq2pd,      new InstructionInfo(BadOp,      BadOp,      BadOp,      BadOp,      0x00000fe6, InstructionFlags.Vex | InstructionFlags.PrefixF3));
+            Add(X86Instruction.Cvtdq2ps,      new InstructionInfo(BadOp,      BadOp,      BadOp,      BadOp,      0x00000f5b, InstructionFlags.Vex));
+            Add(X86Instruction.Cvtpd2dq,      new InstructionInfo(BadOp,      BadOp,      BadOp,      BadOp,      0x00000fe6, InstructionFlags.Vex | InstructionFlags.PrefixF2));
+            Add(X86Instruction.Cvtpd2ps,      new InstructionInfo(BadOp,      BadOp,      BadOp,      BadOp,      0x00000f5a, InstructionFlags.Vex | InstructionFlags.Prefix66));
+            Add(X86Instruction.Cvtps2dq,      new InstructionInfo(BadOp,      BadOp,      BadOp,      BadOp,      0x00000f5b, InstructionFlags.Vex | InstructionFlags.Prefix66));
+            Add(X86Instruction.Cvtps2pd,      new InstructionInfo(BadOp,      BadOp,      BadOp,      BadOp,      0x00000f5a, InstructionFlags.Vex));
+            Add(X86Instruction.Cvtsd2si,      new InstructionInfo(BadOp,      BadOp,      BadOp,      BadOp,      0x00000f2d, InstructionFlags.Vex | InstructionFlags.PrefixF2));
+            Add(X86Instruction.Cvtsd2ss,      new InstructionInfo(BadOp,      BadOp,      BadOp,      BadOp,      0x00000f5a, InstructionFlags.Vex | InstructionFlags.PrefixF2));
+            Add(X86Instruction.Cvtsi2sd,      new InstructionInfo(BadOp,      BadOp,      BadOp,      BadOp,      0x00000f2a, InstructionFlags.Vex | InstructionFlags.PrefixF2));
+            Add(X86Instruction.Cvtsi2ss,      new InstructionInfo(BadOp,      BadOp,      BadOp,      BadOp,      0x00000f2a, InstructionFlags.Vex | InstructionFlags.PrefixF3));
+            Add(X86Instruction.Cvtss2sd,      new InstructionInfo(BadOp,      BadOp,      BadOp,      BadOp,      0x00000f5a, InstructionFlags.Vex | InstructionFlags.PrefixF3));
+            Add(X86Instruction.Cvtss2si,      new InstructionInfo(BadOp,      BadOp,      BadOp,      BadOp,      0x00000f2d, InstructionFlags.Vex | InstructionFlags.PrefixF3));
+            Add(X86Instruction.Div,           new InstructionInfo(BadOp,      BadOp,      BadOp,      BadOp,      0x060000f7, InstructionFlags.None));
+            Add(X86Instruction.Divpd,         new InstructionInfo(BadOp,      BadOp,      BadOp,      BadOp,      0x00000f5e, InstructionFlags.Vex | InstructionFlags.Prefix66));
+            Add(X86Instruction.Divps,         new InstructionInfo(BadOp,      BadOp,      BadOp,      BadOp,      0x00000f5e, InstructionFlags.Vex));
+            Add(X86Instruction.Divsd,         new InstructionInfo(BadOp,      BadOp,      BadOp,      BadOp,      0x00000f5e, InstructionFlags.Vex | InstructionFlags.PrefixF2));
+            Add(X86Instruction.Divss,         new InstructionInfo(BadOp,      BadOp,      BadOp,      BadOp,      0x00000f5e, InstructionFlags.Vex | InstructionFlags.PrefixF3));
+            Add(X86Instruction.Gf2p8affineqb, new InstructionInfo(BadOp,      BadOp,      BadOp,      BadOp,      0x000f3ace, InstructionFlags.Prefix66));
+            Add(X86Instruction.Haddpd,        new InstructionInfo(BadOp,      BadOp,      BadOp,      BadOp,      0x00000f7c, InstructionFlags.Vex | InstructionFlags.Prefix66));
+            Add(X86Instruction.Haddps,        new InstructionInfo(BadOp,      BadOp,      BadOp,      BadOp,      0x00000f7c, InstructionFlags.Vex | InstructionFlags.PrefixF2));
+            Add(X86Instruction.Idiv,          new InstructionInfo(BadOp,      BadOp,      BadOp,      BadOp,      0x070000f7, InstructionFlags.None));
+            Add(X86Instruction.Imul,          new InstructionInfo(BadOp,      0x0000006b, 0x00000069, BadOp,      0x00000faf, InstructionFlags.None));
+            Add(X86Instruction.Imul128,       new InstructionInfo(BadOp,      BadOp,      BadOp,      BadOp,      0x050000f7, InstructionFlags.None));
+            Add(X86Instruction.Insertps,      new InstructionInfo(BadOp,      BadOp,      BadOp,      BadOp,      0x000f3a21, InstructionFlags.Vex | InstructionFlags.Prefix66));
+            Add(X86Instruction.Jmp,           new InstructionInfo(0x040000ff, BadOp,      BadOp,      BadOp,      BadOp,      InstructionFlags.None));
+            Add(X86Instruction.Ldmxcsr,       new InstructionInfo(0x02000fae, BadOp,      BadOp,      BadOp,      BadOp,      InstructionFlags.Vex));
+            Add(X86Instruction.Lea,           new InstructionInfo(BadOp,      BadOp,      BadOp,      BadOp,      0x0000008d, InstructionFlags.None));
+            Add(X86Instruction.Maxpd,         new InstructionInfo(BadOp,      BadOp,      BadOp,      BadOp,      0x00000f5f, InstructionFlags.Vex | InstructionFlags.Prefix66));
+            Add(X86Instruction.Maxps,         new InstructionInfo(BadOp,      BadOp,      BadOp,      BadOp,      0x00000f5f, InstructionFlags.Vex));
+            Add(X86Instruction.Maxsd,         new InstructionInfo(BadOp,      BadOp,      BadOp,      BadOp,      0x00000f5f, InstructionFlags.Vex | InstructionFlags.PrefixF2));
+            Add(X86Instruction.Maxss,         new InstructionInfo(BadOp,      BadOp,      BadOp,      BadOp,      0x00000f5f, InstructionFlags.Vex | InstructionFlags.PrefixF3));
+            Add(X86Instruction.Minpd,         new InstructionInfo(BadOp,      BadOp,      BadOp,      BadOp,      0x00000f5d, InstructionFlags.Vex | InstructionFlags.Prefix66));
+            Add(X86Instruction.Minps,         new InstructionInfo(BadOp,      BadOp,      BadOp,      BadOp,      0x00000f5d, InstructionFlags.Vex));
+            Add(X86Instruction.Minsd,         new InstructionInfo(BadOp,      BadOp,      BadOp,      BadOp,      0x00000f5d, InstructionFlags.Vex | InstructionFlags.PrefixF2));
+            Add(X86Instruction.Minss,         new InstructionInfo(BadOp,      BadOp,      BadOp,      BadOp,      0x00000f5d, InstructionFlags.Vex | InstructionFlags.PrefixF3));
+            Add(X86Instruction.Mov,           new InstructionInfo(0x00000089, BadOp,      0x000000c7, 0x000000b8, 0x0000008b, InstructionFlags.None));
+            Add(X86Instruction.Mov16,         new InstructionInfo(0x00000089, BadOp,      0x000000c7, BadOp,      0x0000008b, InstructionFlags.Prefix66));
+            Add(X86Instruction.Mov8,          new InstructionInfo(0x00000088, 0x000000c6, BadOp,      BadOp,      0x0000008a, InstructionFlags.Reg8Src | InstructionFlags.Reg8Dest));
+            Add(X86Instruction.Movd,          new InstructionInfo(0x00000f7e, BadOp,      BadOp,      BadOp,      0x00000f6e, InstructionFlags.Vex | InstructionFlags.Prefix66));
+            Add(X86Instruction.Movdqu,        new InstructionInfo(0x00000f7f, BadOp,      BadOp,      BadOp,      0x00000f6f, InstructionFlags.Vex | InstructionFlags.PrefixF3));
+            Add(X86Instruction.Movhlps,       new InstructionInfo(BadOp,      BadOp,      BadOp,      BadOp,      0x00000f12, InstructionFlags.Vex));
+            Add(X86Instruction.Movlhps,       new InstructionInfo(BadOp,      BadOp,      BadOp,      BadOp,      0x00000f16, InstructionFlags.Vex));
+            Add(X86Instruction.Movq,          new InstructionInfo(BadOp,      BadOp,      BadOp,      BadOp,      0x00000f7e, InstructionFlags.Vex | InstructionFlags.PrefixF3));
+            Add(X86Instruction.Movsd,         new InstructionInfo(0x00000f11, BadOp,      BadOp,      BadOp,      0x00000f10, InstructionFlags.Vex | InstructionFlags.PrefixF2));
+            Add(X86Instruction.Movss,         new InstructionInfo(0x00000f11, BadOp,      BadOp,      BadOp,      0x00000f10, InstructionFlags.Vex | InstructionFlags.PrefixF3));
+            Add(X86Instruction.Movsx16,       new InstructionInfo(BadOp,      BadOp,      BadOp,      BadOp,      0x00000fbf, InstructionFlags.None));
+            Add(X86Instruction.Movsx32,       new InstructionInfo(BadOp,      BadOp,      BadOp,      BadOp,      0x00000063, InstructionFlags.None));
+            Add(X86Instruction.Movsx8,        new InstructionInfo(BadOp,      BadOp,      BadOp,      BadOp,      0x00000fbe, InstructionFlags.Reg8Src));
+            Add(X86Instruction.Movzx16,       new InstructionInfo(BadOp,      BadOp,      BadOp,      BadOp,      0x00000fb7, InstructionFlags.None));
+            Add(X86Instruction.Movzx8,        new InstructionInfo(BadOp,      BadOp,      BadOp,      BadOp,      0x00000fb6, InstructionFlags.Reg8Src));
+            Add(X86Instruction.Mul128,        new InstructionInfo(BadOp,      BadOp,      BadOp,      BadOp,      0x040000f7, InstructionFlags.None));
+            Add(X86Instruction.Mulpd,         new InstructionInfo(BadOp,      BadOp,      BadOp,      BadOp,      0x00000f59, InstructionFlags.Vex | InstructionFlags.Prefix66));
+            Add(X86Instruction.Mulps,         new InstructionInfo(BadOp,      BadOp,      BadOp,      BadOp,      0x00000f59, InstructionFlags.Vex));
+            Add(X86Instruction.Mulsd,         new InstructionInfo(BadOp,      BadOp,      BadOp,      BadOp,      0x00000f59, InstructionFlags.Vex | InstructionFlags.PrefixF2));
+            Add(X86Instruction.Mulss,         new InstructionInfo(BadOp,      BadOp,      BadOp,      BadOp,      0x00000f59, InstructionFlags.Vex | InstructionFlags.PrefixF3));
+            Add(X86Instruction.Neg,           new InstructionInfo(0x030000f7, BadOp,      BadOp,      BadOp,      BadOp,      InstructionFlags.None));
+            Add(X86Instruction.Not,           new InstructionInfo(0x020000f7, BadOp,      BadOp,      BadOp,      BadOp,      InstructionFlags.None));
+            Add(X86Instruction.Or,            new InstructionInfo(0x00000009, 0x01000083, 0x01000081, BadOp,      0x0000000b, InstructionFlags.None));
+            Add(X86Instruction.Paddb,         new InstructionInfo(BadOp,      BadOp,      BadOp,      BadOp,      0x00000ffc, InstructionFlags.Vex | InstructionFlags.Prefix66));
+            Add(X86Instruction.Paddd,         new InstructionInfo(BadOp,      BadOp,      BadOp,      BadOp,      0x00000ffe, InstructionFlags.Vex | InstructionFlags.Prefix66));
+            Add(X86Instruction.Paddq,         new InstructionInfo(BadOp,      BadOp,      BadOp,      BadOp,      0x00000fd4, InstructionFlags.Vex | InstructionFlags.Prefix66));
+            Add(X86Instruction.Paddw,         new InstructionInfo(BadOp,      BadOp,      BadOp,      BadOp,      0x00000ffd, InstructionFlags.Vex | InstructionFlags.Prefix66));
+            Add(X86Instruction.Palignr,       new InstructionInfo(BadOp,      BadOp,      BadOp,      BadOp,      0x000f3a0f, InstructionFlags.Vex | InstructionFlags.Prefix66));
+            Add(X86Instruction.Pand,          new InstructionInfo(BadOp,      BadOp,      BadOp,      BadOp,      0x00000fdb, InstructionFlags.Vex | InstructionFlags.Prefix66));
+            Add(X86Instruction.Pandn,         new InstructionInfo(BadOp,      BadOp,      BadOp,      BadOp,      0x00000fdf, InstructionFlags.Vex | InstructionFlags.Prefix66));
+            Add(X86Instruction.Pavgb,         new InstructionInfo(BadOp,      BadOp,      BadOp,      BadOp,      0x00000fe0, InstructionFlags.Vex | InstructionFlags.Prefix66));
+            Add(X86Instruction.Pavgw,         new InstructionInfo(BadOp,      BadOp,      BadOp,      BadOp,      0x00000fe3, InstructionFlags.Vex | InstructionFlags.Prefix66));
+            Add(X86Instruction.Pblendvb,      new InstructionInfo(BadOp,      BadOp,      BadOp,      BadOp,      0x000f3810, InstructionFlags.Prefix66));
+            Add(X86Instruction.Pclmulqdq,     new InstructionInfo(BadOp,      BadOp,      BadOp,      BadOp,      0x000f3a44, InstructionFlags.Vex | InstructionFlags.Prefix66));
+            Add(X86Instruction.Pcmpeqb,       new InstructionInfo(BadOp,      BadOp,      BadOp,      BadOp,      0x00000f74, InstructionFlags.Vex | InstructionFlags.Prefix66));
+            Add(X86Instruction.Pcmpeqd,       new InstructionInfo(BadOp,      BadOp,      BadOp,      BadOp,      0x00000f76, InstructionFlags.Vex | InstructionFlags.Prefix66));
+            Add(X86Instruction.Pcmpeqq,       new InstructionInfo(BadOp,      BadOp,      BadOp,      BadOp,      0x000f3829, InstructionFlags.Vex | InstructionFlags.Prefix66));
+            Add(X86Instruction.Pcmpeqw,       new InstructionInfo(BadOp,      BadOp,      BadOp,      BadOp,      0x00000f75, InstructionFlags.Vex | InstructionFlags.Prefix66));
+            Add(X86Instruction.Pcmpgtb,       new InstructionInfo(BadOp,      BadOp,      BadOp,      BadOp,      0x00000f64, InstructionFlags.Vex | InstructionFlags.Prefix66));
+            Add(X86Instruction.Pcmpgtd,       new InstructionInfo(BadOp,      BadOp,      BadOp,      BadOp,      0x00000f66, InstructionFlags.Vex | InstructionFlags.Prefix66));
+            Add(X86Instruction.Pcmpgtq,       new InstructionInfo(BadOp,      BadOp,      BadOp,      BadOp,      0x000f3837, InstructionFlags.Vex | InstructionFlags.Prefix66));
+            Add(X86Instruction.Pcmpgtw,       new InstructionInfo(BadOp,      BadOp,      BadOp,      BadOp,      0x00000f65, InstructionFlags.Vex | InstructionFlags.Prefix66));
+            Add(X86Instruction.Pextrb,        new InstructionInfo(0x000f3a14, BadOp,      BadOp,      BadOp,      BadOp,      InstructionFlags.Vex | InstructionFlags.Prefix66));
+            Add(X86Instruction.Pextrd,        new InstructionInfo(0x000f3a16, BadOp,      BadOp,      BadOp,      BadOp,      InstructionFlags.Vex | InstructionFlags.Prefix66));
+            Add(X86Instruction.Pextrq,        new InstructionInfo(0x000f3a16, BadOp,      BadOp,      BadOp,      BadOp,      InstructionFlags.Vex | InstructionFlags.RexW | InstructionFlags.Prefix66));
+            Add(X86Instruction.Pextrw,        new InstructionInfo(BadOp,      BadOp,      BadOp,      BadOp,      0x00000fc5, InstructionFlags.Vex | InstructionFlags.Prefix66));
+            Add(X86Instruction.Pinsrb,        new InstructionInfo(BadOp,      BadOp,      BadOp,      BadOp,      0x000f3a20, InstructionFlags.Vex | InstructionFlags.Prefix66));
+            Add(X86Instruction.Pinsrd,        new InstructionInfo(BadOp,      BadOp,      BadOp,      BadOp,      0x000f3a22, InstructionFlags.Vex | InstructionFlags.Prefix66));
+            Add(X86Instruction.Pinsrq,        new InstructionInfo(BadOp,      BadOp,      BadOp,      BadOp,      0x000f3a22, InstructionFlags.Vex | InstructionFlags.RexW | InstructionFlags.Prefix66));
+            Add(X86Instruction.Pinsrw,        new InstructionInfo(BadOp,      BadOp,      BadOp,      BadOp,      0x00000fc4, InstructionFlags.Vex | InstructionFlags.Prefix66));
+            Add(X86Instruction.Pmaxsb,        new InstructionInfo(BadOp,      BadOp,      BadOp,      BadOp,      0x000f383c, InstructionFlags.Vex | InstructionFlags.Prefix66));
+            Add(X86Instruction.Pmaxsd,        new InstructionInfo(BadOp,      BadOp,      BadOp,      BadOp,      0x000f383d, InstructionFlags.Vex | InstructionFlags.Prefix66));
+            Add(X86Instruction.Pmaxsw,        new InstructionInfo(BadOp,      BadOp,      BadOp,      BadOp,      0x00000fee, InstructionFlags.Vex | InstructionFlags.Prefix66));
+            Add(X86Instruction.Pmaxub,        new InstructionInfo(BadOp,      BadOp,      BadOp,      BadOp,      0x00000fde, InstructionFlags.Vex | InstructionFlags.Prefix66));
+            Add(X86Instruction.Pmaxud,        new InstructionInfo(BadOp,      BadOp,      BadOp,      BadOp,      0x000f383f, InstructionFlags.Vex | InstructionFlags.Prefix66));
+            Add(X86Instruction.Pmaxuw,        new InstructionInfo(BadOp,      BadOp,      BadOp,      BadOp,      0x000f383e, InstructionFlags.Vex | InstructionFlags.Prefix66));
+            Add(X86Instruction.Pminsb,        new InstructionInfo(BadOp,      BadOp,      BadOp,      BadOp,      0x000f3838, InstructionFlags.Vex | InstructionFlags.Prefix66));
+            Add(X86Instruction.Pminsd,        new InstructionInfo(BadOp,      BadOp,      BadOp,      BadOp,      0x000f3839, InstructionFlags.Vex | InstructionFlags.Prefix66));
+            Add(X86Instruction.Pminsw,        new InstructionInfo(BadOp,      BadOp,      BadOp,      BadOp,      0x00000fea, InstructionFlags.Vex | InstructionFlags.Prefix66));
+            Add(X86Instruction.Pminub,        new InstructionInfo(BadOp,      BadOp,      BadOp,      BadOp,      0x00000fda, InstructionFlags.Vex | InstructionFlags.Prefix66));
+            Add(X86Instruction.Pminud,        new InstructionInfo(BadOp,      BadOp,      BadOp,      BadOp,      0x000f383b, InstructionFlags.Vex | InstructionFlags.Prefix66));
+            Add(X86Instruction.Pminuw,        new InstructionInfo(BadOp,      BadOp,      BadOp,      BadOp,      0x000f383a, InstructionFlags.Vex | InstructionFlags.Prefix66));
+            Add(X86Instruction.Pmovsxbw,      new InstructionInfo(BadOp,      BadOp,      BadOp,      BadOp,      0x000f3820, InstructionFlags.Vex | InstructionFlags.Prefix66));
+            Add(X86Instruction.Pmovsxdq,      new InstructionInfo(BadOp,      BadOp,      BadOp,      BadOp,      0x000f3825, InstructionFlags.Vex | InstructionFlags.Prefix66));
+            Add(X86Instruction.Pmovsxwd,      new InstructionInfo(BadOp,      BadOp,      BadOp,      BadOp,      0x000f3823, InstructionFlags.Vex | InstructionFlags.Prefix66));
+            Add(X86Instruction.Pmovzxbw,      new InstructionInfo(BadOp,      BadOp,      BadOp,      BadOp,      0x000f3830, InstructionFlags.Vex | InstructionFlags.Prefix66));
+            Add(X86Instruction.Pmovzxdq,      new InstructionInfo(BadOp,      BadOp,      BadOp,      BadOp,      0x000f3835, InstructionFlags.Vex | InstructionFlags.Prefix66));
+            Add(X86Instruction.Pmovzxwd,      new InstructionInfo(BadOp,      BadOp,      BadOp,      BadOp,      0x000f3833, InstructionFlags.Vex | InstructionFlags.Prefix66));
+            Add(X86Instruction.Pmulld,        new InstructionInfo(BadOp,      BadOp,      BadOp,      BadOp,      0x000f3840, InstructionFlags.Vex | InstructionFlags.Prefix66));
+            Add(X86Instruction.Pmullw,        new InstructionInfo(BadOp,      BadOp,      BadOp,      BadOp,      0x00000fd5, InstructionFlags.Vex | InstructionFlags.Prefix66));
+            Add(X86Instruction.Pop,           new InstructionInfo(0x0000008f, BadOp,      BadOp,      BadOp,      BadOp,      InstructionFlags.None));
+            Add(X86Instruction.Popcnt,        new InstructionInfo(BadOp,      BadOp,      BadOp,      BadOp,      0x00000fb8, InstructionFlags.PrefixF3));
+            Add(X86Instruction.Por,           new InstructionInfo(BadOp,      BadOp,      BadOp,      BadOp,      0x00000feb, InstructionFlags.Vex | InstructionFlags.Prefix66));
+            Add(X86Instruction.Pshufb,        new InstructionInfo(BadOp,      BadOp,      BadOp,      BadOp,      0x000f3800, InstructionFlags.Vex | InstructionFlags.Prefix66));
+            Add(X86Instruction.Pshufd,        new InstructionInfo(BadOp,      BadOp,      BadOp,      BadOp,      0x00000f70, InstructionFlags.Vex | InstructionFlags.Prefix66));
+            Add(X86Instruction.Pslld,         new InstructionInfo(BadOp,      0x06000f72, BadOp,      BadOp,      0x00000ff2, InstructionFlags.Vex | InstructionFlags.Prefix66));
+            Add(X86Instruction.Pslldq,        new InstructionInfo(BadOp,      0x07000f73, BadOp,      BadOp,      BadOp,      InstructionFlags.Vex | InstructionFlags.Prefix66));
+            Add(X86Instruction.Psllq,         new InstructionInfo(BadOp,      0x06000f73, BadOp,      BadOp,      0x00000ff3, InstructionFlags.Vex | InstructionFlags.Prefix66));
+            Add(X86Instruction.Psllw,         new InstructionInfo(BadOp,      0x06000f71, BadOp,      BadOp,      0x00000ff1, InstructionFlags.Vex | InstructionFlags.Prefix66));
+            Add(X86Instruction.Psrad,         new InstructionInfo(BadOp,      0x04000f72, BadOp,      BadOp,      0x00000fe2, InstructionFlags.Vex | InstructionFlags.Prefix66));
+            Add(X86Instruction.Psraw,         new InstructionInfo(BadOp,      0x04000f71, BadOp,      BadOp,      0x00000fe1, InstructionFlags.Vex | InstructionFlags.Prefix66));
+            Add(X86Instruction.Psrld,         new InstructionInfo(BadOp,      0x02000f72, BadOp,      BadOp,      0x00000fd2, InstructionFlags.Vex | InstructionFlags.Prefix66));
+            Add(X86Instruction.Psrlq,         new InstructionInfo(BadOp,      0x02000f73, BadOp,      BadOp,      0x00000fd3, InstructionFlags.Vex | InstructionFlags.Prefix66));
+            Add(X86Instruction.Psrldq,        new InstructionInfo(BadOp,      0x03000f73, BadOp,      BadOp,      BadOp,      InstructionFlags.Vex | InstructionFlags.Prefix66));
+            Add(X86Instruction.Psrlw,         new InstructionInfo(BadOp,      0x02000f71, BadOp,      BadOp,      0x00000fd1, InstructionFlags.Vex | InstructionFlags.Prefix66));
+            Add(X86Instruction.Psubb,         new InstructionInfo(BadOp,      BadOp,      BadOp,      BadOp,      0x00000ff8, InstructionFlags.Vex | InstructionFlags.Prefix66));
+            Add(X86Instruction.Psubd,         new InstructionInfo(BadOp,      BadOp,      BadOp,      BadOp,      0x00000ffa, InstructionFlags.Vex | InstructionFlags.Prefix66));
+            Add(X86Instruction.Psubq,         new InstructionInfo(BadOp,      BadOp,      BadOp,      BadOp,      0x00000ffb, InstructionFlags.Vex | InstructionFlags.Prefix66));
+            Add(X86Instruction.Psubw,         new InstructionInfo(BadOp,      BadOp,      BadOp,      BadOp,      0x00000ff9, InstructionFlags.Vex | InstructionFlags.Prefix66));
+            Add(X86Instruction.Punpckhbw,     new InstructionInfo(BadOp,      BadOp,      BadOp,      BadOp,      0x00000f68, InstructionFlags.Vex | InstructionFlags.Prefix66));
+            Add(X86Instruction.Punpckhdq,     new InstructionInfo(BadOp,      BadOp,      BadOp,      BadOp,      0x00000f6a, InstructionFlags.Vex | InstructionFlags.Prefix66));
+            Add(X86Instruction.Punpckhqdq,    new InstructionInfo(BadOp,      BadOp,      BadOp,      BadOp,      0x00000f6d, InstructionFlags.Vex | InstructionFlags.Prefix66));
+            Add(X86Instruction.Punpckhwd,     new InstructionInfo(BadOp,      BadOp,      BadOp,      BadOp,      0x00000f69, InstructionFlags.Vex | InstructionFlags.Prefix66));
+            Add(X86Instruction.Punpcklbw,     new InstructionInfo(BadOp,      BadOp,      BadOp,      BadOp,      0x00000f60, InstructionFlags.Vex | InstructionFlags.Prefix66));
+            Add(X86Instruction.Punpckldq,     new InstructionInfo(BadOp,      BadOp,      BadOp,      BadOp,      0x00000f62, InstructionFlags.Vex | InstructionFlags.Prefix66));
+            Add(X86Instruction.Punpcklqdq,    new InstructionInfo(BadOp,      BadOp,      BadOp,      BadOp,      0x00000f6c, InstructionFlags.Vex | InstructionFlags.Prefix66));
+            Add(X86Instruction.Punpcklwd,     new InstructionInfo(BadOp,      BadOp,      BadOp,      BadOp,      0x00000f61, InstructionFlags.Vex | InstructionFlags.Prefix66));
+            Add(X86Instruction.Push,          new InstructionInfo(BadOp,      0x0000006a, 0x00000068, BadOp,      0x060000ff, InstructionFlags.None));
+            Add(X86Instruction.Pxor,          new InstructionInfo(BadOp,      BadOp,      BadOp,      BadOp,      0x00000fef, InstructionFlags.Vex | InstructionFlags.Prefix66));
+            Add(X86Instruction.Rcpps,         new InstructionInfo(BadOp,      BadOp,      BadOp,      BadOp,      0x00000f53, InstructionFlags.Vex));
+            Add(X86Instruction.Rcpss,         new InstructionInfo(BadOp,      BadOp,      BadOp,      BadOp,      0x00000f53, InstructionFlags.Vex | InstructionFlags.PrefixF3));
+            Add(X86Instruction.Ror,           new InstructionInfo(0x010000d3, 0x010000c1, BadOp,      BadOp,      BadOp,      InstructionFlags.None));
+            Add(X86Instruction.Roundpd,       new InstructionInfo(BadOp,      BadOp,      BadOp,      BadOp,      0x000f3a09, InstructionFlags.Vex | InstructionFlags.Prefix66));
+            Add(X86Instruction.Roundps,       new InstructionInfo(BadOp,      BadOp,      BadOp,      BadOp,      0x000f3a08, InstructionFlags.Vex | InstructionFlags.Prefix66));
+            Add(X86Instruction.Roundsd,       new InstructionInfo(BadOp,      BadOp,      BadOp,      BadOp,      0x000f3a0b, InstructionFlags.Vex | InstructionFlags.Prefix66));
+            Add(X86Instruction.Roundss,       new InstructionInfo(BadOp,      BadOp,      BadOp,      BadOp,      0x000f3a0a, InstructionFlags.Vex | InstructionFlags.Prefix66));
+            Add(X86Instruction.Rsqrtps,       new InstructionInfo(BadOp,      BadOp,      BadOp,      BadOp,      0x00000f52, InstructionFlags.Vex));
+            Add(X86Instruction.Rsqrtss,       new InstructionInfo(BadOp,      BadOp,      BadOp,      BadOp,      0x00000f52, InstructionFlags.Vex | InstructionFlags.PrefixF3));
+            Add(X86Instruction.Sar,           new InstructionInfo(0x070000d3, 0x070000c1, BadOp,      BadOp,      BadOp,      InstructionFlags.None));
+            Add(X86Instruction.Setcc,         new InstructionInfo(BadOp,      BadOp,      BadOp,      BadOp,      0x00000f90, InstructionFlags.Reg8Dest));
+            Add(X86Instruction.Sha256Msg1,    new InstructionInfo(BadOp,      BadOp,      BadOp,      BadOp,      0x000f38cc, InstructionFlags.None));
+            Add(X86Instruction.Sha256Msg2,    new InstructionInfo(BadOp,      BadOp,      BadOp,      BadOp,      0x000f38cd, InstructionFlags.None));
+            Add(X86Instruction.Sha256Rnds2,   new InstructionInfo(BadOp,      BadOp,      BadOp,      BadOp,      0x000f38cb, InstructionFlags.None));
+            Add(X86Instruction.Shl,           new InstructionInfo(0x040000d3, 0x040000c1, BadOp,      BadOp,      BadOp,      InstructionFlags.None));
+            Add(X86Instruction.Shr,           new InstructionInfo(0x050000d3, 0x050000c1, BadOp,      BadOp,      BadOp,      InstructionFlags.None));
+            Add(X86Instruction.Shufpd,        new InstructionInfo(BadOp,      BadOp,      BadOp,      BadOp,      0x00000fc6, InstructionFlags.Vex | InstructionFlags.Prefix66));
+            Add(X86Instruction.Shufps,        new InstructionInfo(BadOp,      BadOp,      BadOp,      BadOp,      0x00000fc6, InstructionFlags.Vex));
+            Add(X86Instruction.Sqrtpd,        new InstructionInfo(BadOp,      BadOp,      BadOp,      BadOp,      0x00000f51, InstructionFlags.Vex | InstructionFlags.Prefix66));
+            Add(X86Instruction.Sqrtps,        new InstructionInfo(BadOp,      BadOp,      BadOp,      BadOp,      0x00000f51, InstructionFlags.Vex));
+            Add(X86Instruction.Sqrtsd,        new InstructionInfo(BadOp,      BadOp,      BadOp,      BadOp,      0x00000f51, InstructionFlags.Vex | InstructionFlags.PrefixF2));
+            Add(X86Instruction.Sqrtss,        new InstructionInfo(BadOp,      BadOp,      BadOp,      BadOp,      0x00000f51, InstructionFlags.Vex | InstructionFlags.PrefixF3));
+            Add(X86Instruction.Stmxcsr,       new InstructionInfo(0x03000fae, BadOp,      BadOp,      BadOp,      BadOp,      InstructionFlags.Vex));
+            Add(X86Instruction.Sub,           new InstructionInfo(0x00000029, 0x05000083, 0x05000081, BadOp,      0x0000002b, InstructionFlags.None));
+            Add(X86Instruction.Subpd,         new InstructionInfo(BadOp,      BadOp,      BadOp,      BadOp,      0x00000f5c, InstructionFlags.Vex | InstructionFlags.Prefix66));
+            Add(X86Instruction.Subps,         new InstructionInfo(BadOp,      BadOp,      BadOp,      BadOp,      0x00000f5c, InstructionFlags.Vex));
+            Add(X86Instruction.Subsd,         new InstructionInfo(BadOp,      BadOp,      BadOp,      BadOp,      0x00000f5c, InstructionFlags.Vex | InstructionFlags.PrefixF2));
+            Add(X86Instruction.Subss,         new InstructionInfo(BadOp,      BadOp,      BadOp,      BadOp,      0x00000f5c, InstructionFlags.Vex | InstructionFlags.PrefixF3));
+            Add(X86Instruction.Test,          new InstructionInfo(0x00000085, BadOp,      0x000000f7, BadOp,      BadOp,      InstructionFlags.None));
+            Add(X86Instruction.Unpckhpd,      new InstructionInfo(BadOp,      BadOp,      BadOp,      BadOp,      0x00000f15, InstructionFlags.Vex | InstructionFlags.Prefix66));
+            Add(X86Instruction.Unpckhps,      new InstructionInfo(BadOp,      BadOp,      BadOp,      BadOp,      0x00000f15, InstructionFlags.Vex));
+            Add(X86Instruction.Unpcklpd,      new InstructionInfo(BadOp,      BadOp,      BadOp,      BadOp,      0x00000f14, InstructionFlags.Vex | InstructionFlags.Prefix66));
+            Add(X86Instruction.Unpcklps,      new InstructionInfo(BadOp,      BadOp,      BadOp,      BadOp,      0x00000f14, InstructionFlags.Vex));
+            Add(X86Instruction.Vblendvpd,     new InstructionInfo(BadOp,      BadOp,      BadOp,      BadOp,      0x000f3a4b, InstructionFlags.Vex | InstructionFlags.Prefix66));
+            Add(X86Instruction.Vblendvps,     new InstructionInfo(BadOp,      BadOp,      BadOp,      BadOp,      0x000f3a4a, InstructionFlags.Vex | InstructionFlags.Prefix66));
+            Add(X86Instruction.Vcvtph2ps,     new InstructionInfo(BadOp,      BadOp,      BadOp,      BadOp,      0x000f3813, InstructionFlags.Vex | InstructionFlags.Prefix66));
+            Add(X86Instruction.Vcvtps2ph,     new InstructionInfo(0x000f3a1d, BadOp,      BadOp,      BadOp,      BadOp,      InstructionFlags.Vex | InstructionFlags.Prefix66));
+            Add(X86Instruction.Vfmadd231ps,   new InstructionInfo(BadOp,      BadOp,      BadOp,      BadOp,      0x000f38b8, InstructionFlags.Vex | InstructionFlags.Prefix66));
+            Add(X86Instruction.Vfmadd231sd,   new InstructionInfo(BadOp,      BadOp,      BadOp,      BadOp,      0x000f38b9, InstructionFlags.Vex | InstructionFlags.Prefix66 | InstructionFlags.RexW));
+            Add(X86Instruction.Vfmadd231ss,   new InstructionInfo(BadOp,      BadOp,      BadOp,      BadOp,      0x000f38b9, InstructionFlags.Vex | InstructionFlags.Prefix66));
+            Add(X86Instruction.Vfmsub231sd,   new InstructionInfo(BadOp,      BadOp,      BadOp,      BadOp,      0x000f38bb, InstructionFlags.Vex | InstructionFlags.Prefix66 | InstructionFlags.RexW));
+            Add(X86Instruction.Vfmsub231ss,   new InstructionInfo(BadOp,      BadOp,      BadOp,      BadOp,      0x000f38bb, InstructionFlags.Vex | InstructionFlags.Prefix66));
+            Add(X86Instruction.Vfnmadd231ps,  new InstructionInfo(BadOp,      BadOp,      BadOp,      BadOp,      0x000f38bc, InstructionFlags.Vex | InstructionFlags.Prefix66));
+            Add(X86Instruction.Vfnmadd231sd,  new InstructionInfo(BadOp,      BadOp,      BadOp,      BadOp,      0x000f38bd, InstructionFlags.Vex | InstructionFlags.Prefix66 | InstructionFlags.RexW));
+            Add(X86Instruction.Vfnmadd231ss,  new InstructionInfo(BadOp,      BadOp,      BadOp,      BadOp,      0x000f38bd, InstructionFlags.Vex | InstructionFlags.Prefix66));
+            Add(X86Instruction.Vfnmsub231sd,  new InstructionInfo(BadOp,      BadOp,      BadOp,      BadOp,      0x000f38bf, InstructionFlags.Vex | InstructionFlags.Prefix66 | InstructionFlags.RexW));
+            Add(X86Instruction.Vfnmsub231ss,  new InstructionInfo(BadOp,      BadOp,      BadOp,      BadOp,      0x000f38bf, InstructionFlags.Vex | InstructionFlags.Prefix66));
+            Add(X86Instruction.Vpblendvb,     new InstructionInfo(BadOp,      BadOp,      BadOp,      BadOp,      0x000f3a4c, InstructionFlags.Vex | InstructionFlags.Prefix66));
+            Add(X86Instruction.Xor,           new InstructionInfo(0x00000031, 0x06000083, 0x06000081, BadOp,      0x00000033, InstructionFlags.None));
+            Add(X86Instruction.Xorpd,         new InstructionInfo(BadOp,      BadOp,      BadOp,      BadOp,      0x00000f57, InstructionFlags.Vex | InstructionFlags.Prefix66));
+            Add(X86Instruction.Xorps,         new InstructionInfo(BadOp,      BadOp,      BadOp,      BadOp,      0x00000f57, InstructionFlags.Vex));
 
             static void Add(X86Instruction inst, in InstructionInfo info)
             {

ARMeilleure/CodeGen/X86/HardwareCapabilities.cs

@@ -20,8 +20,9 @@ namespace ARMeilleure.CodeGen.X86
 
             if (maxNum >= 7)
             {
-                (_, int ebx7, _, _) = X86Base.CpuId(0x00000007, 0x00000000);
+                (_, int ebx7, int ecx7, _) = X86Base.CpuId(0x00000007, 0x00000000);
                 FeatureInfo7Ebx = (FeatureFlags7Ebx)ebx7;
+                FeatureInfo7Ecx = (FeatureFlags7Ecx)ecx7;
             }
         }
 
@@ -54,9 +55,16 @@ namespace ARMeilleure.CodeGen.X86
             Sha = 1 << 29
         }
 
+        [Flags]
+        public enum FeatureFlags7Ecx
+        {
+            Gfni = 1 << 8,
+        }
+
         public static FeatureFlags1Edx FeatureInfo1Edx { get; }
         public static FeatureFlags1Ecx FeatureInfo1Ecx { get; }
         public static FeatureFlags7Ebx FeatureInfo7Ebx { get; } = 0;
+        public static FeatureFlags7Ecx FeatureInfo7Ecx { get; } = 0;
 
         public static bool SupportsSse => FeatureInfo1Edx.HasFlag(FeatureFlags1Edx.Sse);
         public static bool SupportsSse2 => FeatureInfo1Edx.HasFlag(FeatureFlags1Edx.Sse2);
@@ -72,6 +80,7 @@ namespace ARMeilleure.CodeGen.X86
         public static bool SupportsAvx2 => FeatureInfo7Ebx.HasFlag(FeatureFlags7Ebx.Avx2) && SupportsAvx;
         public static bool SupportsF16c => FeatureInfo1Ecx.HasFlag(FeatureFlags1Ecx.F16c);
         public static bool SupportsSha => FeatureInfo7Ebx.HasFlag(FeatureFlags7Ebx.Sha);
+        public static bool SupportsGfni => FeatureInfo7Ecx.HasFlag(FeatureFlags7Ecx.Gfni);
 
         public static bool ForceLegacySse { get; set; }
 

ARMeilleure/CodeGen/X86/IntrinsicInfo.cs

@@ -1,6 +1,6 @@
 namespace ARMeilleure.CodeGen.X86
 {
-    struct IntrinsicInfo
+    readonly struct IntrinsicInfo
     {
         public X86Instruction Inst { get; }
         public IntrinsicType  Type { get; }

ARMeilleure/CodeGen/X86/IntrinsicTable.cs

@@ -13,176 +13,177 @@ namespace ARMeilleure.CodeGen.X86
         {
             _intrinTable = new IntrinsicInfo[EnumUtils.GetCount(typeof(Intrinsic))];
 
-            Add(Intrinsic.X86Addpd,        new IntrinsicInfo(X86Instruction.Addpd,        IntrinsicType.Binary));
-            Add(Intrinsic.X86Addps,        new IntrinsicInfo(X86Instruction.Addps,        IntrinsicType.Binary));
-            Add(Intrinsic.X86Addsd,        new IntrinsicInfo(X86Instruction.Addsd,        IntrinsicType.Binary));
-            Add(Intrinsic.X86Addss,        new IntrinsicInfo(X86Instruction.Addss,        IntrinsicType.Binary));
-            Add(Intrinsic.X86Aesdec,       new IntrinsicInfo(X86Instruction.Aesdec,       IntrinsicType.Binary));
-            Add(Intrinsic.X86Aesdeclast,   new IntrinsicInfo(X86Instruction.Aesdeclast,   IntrinsicType.Binary));
-            Add(Intrinsic.X86Aesenc,       new IntrinsicInfo(X86Instruction.Aesenc,       IntrinsicType.Binary));
-            Add(Intrinsic.X86Aesenclast,   new IntrinsicInfo(X86Instruction.Aesenclast,   IntrinsicType.Binary));
-            Add(Intrinsic.X86Aesimc,       new IntrinsicInfo(X86Instruction.Aesimc,       IntrinsicType.Unary));
-            Add(Intrinsic.X86Andnpd,       new IntrinsicInfo(X86Instruction.Andnpd,       IntrinsicType.Binary));
-            Add(Intrinsic.X86Andnps,       new IntrinsicInfo(X86Instruction.Andnps,       IntrinsicType.Binary));
-            Add(Intrinsic.X86Andpd,        new IntrinsicInfo(X86Instruction.Andpd,        IntrinsicType.Binary));
-            Add(Intrinsic.X86Andps,        new IntrinsicInfo(X86Instruction.Andps,        IntrinsicType.Binary));
-            Add(Intrinsic.X86Blendvpd,     new IntrinsicInfo(X86Instruction.Blendvpd,     IntrinsicType.Ternary));
-            Add(Intrinsic.X86Blendvps,     new IntrinsicInfo(X86Instruction.Blendvps,     IntrinsicType.Ternary));
-            Add(Intrinsic.X86Cmppd,        new IntrinsicInfo(X86Instruction.Cmppd,        IntrinsicType.TernaryImm));
-            Add(Intrinsic.X86Cmpps,        new IntrinsicInfo(X86Instruction.Cmpps,        IntrinsicType.TernaryImm));
-            Add(Intrinsic.X86Cmpsd,        new IntrinsicInfo(X86Instruction.Cmpsd,        IntrinsicType.TernaryImm));
-            Add(Intrinsic.X86Cmpss,        new IntrinsicInfo(X86Instruction.Cmpss,        IntrinsicType.TernaryImm));
-            Add(Intrinsic.X86Comisdeq,     new IntrinsicInfo(X86Instruction.Comisd,       IntrinsicType.Comis_));
-            Add(Intrinsic.X86Comisdge,     new IntrinsicInfo(X86Instruction.Comisd,       IntrinsicType.Comis_));
-            Add(Intrinsic.X86Comisdlt,     new IntrinsicInfo(X86Instruction.Comisd,       IntrinsicType.Comis_));
-            Add(Intrinsic.X86Comisseq,     new IntrinsicInfo(X86Instruction.Comiss,       IntrinsicType.Comis_));
-            Add(Intrinsic.X86Comissge,     new IntrinsicInfo(X86Instruction.Comiss,       IntrinsicType.Comis_));
-            Add(Intrinsic.X86Comisslt,     new IntrinsicInfo(X86Instruction.Comiss,       IntrinsicType.Comis_));
-            Add(Intrinsic.X86Crc32,        new IntrinsicInfo(X86Instruction.Crc32,        IntrinsicType.Crc32));
-            Add(Intrinsic.X86Crc32_16,     new IntrinsicInfo(X86Instruction.Crc32_16,     IntrinsicType.Crc32));
-            Add(Intrinsic.X86Crc32_8,      new IntrinsicInfo(X86Instruction.Crc32_8,      IntrinsicType.Crc32));
-            Add(Intrinsic.X86Cvtdq2pd,     new IntrinsicInfo(X86Instruction.Cvtdq2pd,     IntrinsicType.Unary));
-            Add(Intrinsic.X86Cvtdq2ps,     new IntrinsicInfo(X86Instruction.Cvtdq2ps,     IntrinsicType.Unary));
-            Add(Intrinsic.X86Cvtpd2dq,     new IntrinsicInfo(X86Instruction.Cvtpd2dq,     IntrinsicType.Unary));
-            Add(Intrinsic.X86Cvtpd2ps,     new IntrinsicInfo(X86Instruction.Cvtpd2ps,     IntrinsicType.Unary));
-            Add(Intrinsic.X86Cvtps2dq,     new IntrinsicInfo(X86Instruction.Cvtps2dq,     IntrinsicType.Unary));
-            Add(Intrinsic.X86Cvtps2pd,     new IntrinsicInfo(X86Instruction.Cvtps2pd,     IntrinsicType.Unary));
-            Add(Intrinsic.X86Cvtsd2si,     new IntrinsicInfo(X86Instruction.Cvtsd2si,     IntrinsicType.UnaryToGpr));
-            Add(Intrinsic.X86Cvtsd2ss,     new IntrinsicInfo(X86Instruction.Cvtsd2ss,     IntrinsicType.Binary));
-            Add(Intrinsic.X86Cvtsi2sd,     new IntrinsicInfo(X86Instruction.Cvtsi2sd,     IntrinsicType.BinaryGpr));
-            Add(Intrinsic.X86Cvtsi2si,     new IntrinsicInfo(X86Instruction.Movd,         IntrinsicType.UnaryToGpr));
-            Add(Intrinsic.X86Cvtsi2ss,     new IntrinsicInfo(X86Instruction.Cvtsi2ss,     IntrinsicType.BinaryGpr));
-            Add(Intrinsic.X86Cvtss2sd,     new IntrinsicInfo(X86Instruction.Cvtss2sd,     IntrinsicType.Binary));
-            Add(Intrinsic.X86Cvtss2si,     new IntrinsicInfo(X86Instruction.Cvtss2si,     IntrinsicType.UnaryToGpr));
-            Add(Intrinsic.X86Divpd,        new IntrinsicInfo(X86Instruction.Divpd,        IntrinsicType.Binary));
-            Add(Intrinsic.X86Divps,        new IntrinsicInfo(X86Instruction.Divps,        IntrinsicType.Binary));
-            Add(Intrinsic.X86Divsd,        new IntrinsicInfo(X86Instruction.Divsd,        IntrinsicType.Binary));
-            Add(Intrinsic.X86Divss,        new IntrinsicInfo(X86Instruction.Divss,        IntrinsicType.Binary));
-            Add(Intrinsic.X86Haddpd,       new IntrinsicInfo(X86Instruction.Haddpd,       IntrinsicType.Binary));
-            Add(Intrinsic.X86Haddps,       new IntrinsicInfo(X86Instruction.Haddps,       IntrinsicType.Binary));
-            Add(Intrinsic.X86Insertps,     new IntrinsicInfo(X86Instruction.Insertps,     IntrinsicType.TernaryImm));
-            Add(Intrinsic.X86Maxpd,        new IntrinsicInfo(X86Instruction.Maxpd,        IntrinsicType.Binary));
-            Add(Intrinsic.X86Maxps,        new IntrinsicInfo(X86Instruction.Maxps,        IntrinsicType.Binary));
-            Add(Intrinsic.X86Maxsd,        new IntrinsicInfo(X86Instruction.Maxsd,        IntrinsicType.Binary));
-            Add(Intrinsic.X86Maxss,        new IntrinsicInfo(X86Instruction.Maxss,        IntrinsicType.Binary));
-            Add(Intrinsic.X86Minpd,        new IntrinsicInfo(X86Instruction.Minpd,        IntrinsicType.Binary));
-            Add(Intrinsic.X86Minps,        new IntrinsicInfo(X86Instruction.Minps,        IntrinsicType.Binary));
-            Add(Intrinsic.X86Minsd,        new IntrinsicInfo(X86Instruction.Minsd,        IntrinsicType.Binary));
-            Add(Intrinsic.X86Minss,        new IntrinsicInfo(X86Instruction.Minss,        IntrinsicType.Binary));
-            Add(Intrinsic.X86Movhlps,      new IntrinsicInfo(X86Instruction.Movhlps,      IntrinsicType.Binary));
-            Add(Intrinsic.X86Movlhps,      new IntrinsicInfo(X86Instruction.Movlhps,      IntrinsicType.Binary));
-            Add(Intrinsic.X86Movss,        new IntrinsicInfo(X86Instruction.Movss,        IntrinsicType.Binary));
-            Add(Intrinsic.X86Mulpd,        new IntrinsicInfo(X86Instruction.Mulpd,        IntrinsicType.Binary));
-            Add(Intrinsic.X86Mulps,        new IntrinsicInfo(X86Instruction.Mulps,        IntrinsicType.Binary));
-            Add(Intrinsic.X86Mulsd,        new IntrinsicInfo(X86Instruction.Mulsd,        IntrinsicType.Binary));
-            Add(Intrinsic.X86Mulss,        new IntrinsicInfo(X86Instruction.Mulss,        IntrinsicType.Binary));
-            Add(Intrinsic.X86Mxcsrmb,      new IntrinsicInfo(X86Instruction.None,         IntrinsicType.Mxcsr)); // Mask bits.
-            Add(Intrinsic.X86Mxcsrub,      new IntrinsicInfo(X86Instruction.None,         IntrinsicType.Mxcsr)); // Unmask bits.
-            Add(Intrinsic.X86Paddb,        new IntrinsicInfo(X86Instruction.Paddb,        IntrinsicType.Binary));
-            Add(Intrinsic.X86Paddd,        new IntrinsicInfo(X86Instruction.Paddd,        IntrinsicType.Binary));
-            Add(Intrinsic.X86Paddq,        new IntrinsicInfo(X86Instruction.Paddq,        IntrinsicType.Binary));
-            Add(Intrinsic.X86Paddw,        new IntrinsicInfo(X86Instruction.Paddw,        IntrinsicType.Binary));
-            Add(Intrinsic.X86Palignr,      new IntrinsicInfo(X86Instruction.Palignr,      IntrinsicType.TernaryImm));
-            Add(Intrinsic.X86Pand,         new IntrinsicInfo(X86Instruction.Pand,         IntrinsicType.Binary));
-            Add(Intrinsic.X86Pandn,        new IntrinsicInfo(X86Instruction.Pandn,        IntrinsicType.Binary));
-            Add(Intrinsic.X86Pavgb,        new IntrinsicInfo(X86Instruction.Pavgb,        IntrinsicType.Binary));
-            Add(Intrinsic.X86Pavgw,        new IntrinsicInfo(X86Instruction.Pavgw,        IntrinsicType.Binary));
-            Add(Intrinsic.X86Pblendvb,     new IntrinsicInfo(X86Instruction.Pblendvb,     IntrinsicType.Ternary));
-            Add(Intrinsic.X86Pclmulqdq,    new IntrinsicInfo(X86Instruction.Pclmulqdq,    IntrinsicType.TernaryImm));
-            Add(Intrinsic.X86Pcmpeqb,      new IntrinsicInfo(X86Instruction.Pcmpeqb,      IntrinsicType.Binary));
-            Add(Intrinsic.X86Pcmpeqd,      new IntrinsicInfo(X86Instruction.Pcmpeqd,      IntrinsicType.Binary));
-            Add(Intrinsic.X86Pcmpeqq,      new IntrinsicInfo(X86Instruction.Pcmpeqq,      IntrinsicType.Binary));
-            Add(Intrinsic.X86Pcmpeqw,      new IntrinsicInfo(X86Instruction.Pcmpeqw,      IntrinsicType.Binary));
-            Add(Intrinsic.X86Pcmpgtb,      new IntrinsicInfo(X86Instruction.Pcmpgtb,      IntrinsicType.Binary));
-            Add(Intrinsic.X86Pcmpgtd,      new IntrinsicInfo(X86Instruction.Pcmpgtd,      IntrinsicType.Binary));
-            Add(Intrinsic.X86Pcmpgtq,      new IntrinsicInfo(X86Instruction.Pcmpgtq,      IntrinsicType.Binary));
-            Add(Intrinsic.X86Pcmpgtw,      new IntrinsicInfo(X86Instruction.Pcmpgtw,      IntrinsicType.Binary));
-            Add(Intrinsic.X86Pmaxsb,       new IntrinsicInfo(X86Instruction.Pmaxsb,       IntrinsicType.Binary));
-            Add(Intrinsic.X86Pmaxsd,       new IntrinsicInfo(X86Instruction.Pmaxsd,       IntrinsicType.Binary));
-            Add(Intrinsic.X86Pmaxsw,       new IntrinsicInfo(X86Instruction.Pmaxsw,       IntrinsicType.Binary));
-            Add(Intrinsic.X86Pmaxub,       new IntrinsicInfo(X86Instruction.Pmaxub,       IntrinsicType.Binary));
-            Add(Intrinsic.X86Pmaxud,       new IntrinsicInfo(X86Instruction.Pmaxud,       IntrinsicType.Binary));
-            Add(Intrinsic.X86Pmaxuw,       new IntrinsicInfo(X86Instruction.Pmaxuw,       IntrinsicType.Binary));
-            Add(Intrinsic.X86Pminsb,       new IntrinsicInfo(X86Instruction.Pminsb,       IntrinsicType.Binary));
-            Add(Intrinsic.X86Pminsd,       new IntrinsicInfo(X86Instruction.Pminsd,       IntrinsicType.Binary));
-            Add(Intrinsic.X86Pminsw,       new IntrinsicInfo(X86Instruction.Pminsw,       IntrinsicType.Binary));
-            Add(Intrinsic.X86Pminub,       new IntrinsicInfo(X86Instruction.Pminub,       IntrinsicType.Binary));
-            Add(Intrinsic.X86Pminud,       new IntrinsicInfo(X86Instruction.Pminud,       IntrinsicType.Binary));
-            Add(Intrinsic.X86Pminuw,       new IntrinsicInfo(X86Instruction.Pminuw,       IntrinsicType.Binary));
-            Add(Intrinsic.X86Pmovsxbw,     new IntrinsicInfo(X86Instruction.Pmovsxbw,     IntrinsicType.Unary));
-            Add(Intrinsic.X86Pmovsxdq,     new IntrinsicInfo(X86Instruction.Pmovsxdq,     IntrinsicType.Unary));
-            Add(Intrinsic.X86Pmovsxwd,     new IntrinsicInfo(X86Instruction.Pmovsxwd,     IntrinsicType.Unary));
-            Add(Intrinsic.X86Pmovzxbw,     new IntrinsicInfo(X86Instruction.Pmovzxbw,     IntrinsicType.Unary));
-            Add(Intrinsic.X86Pmovzxdq,     new IntrinsicInfo(X86Instruction.Pmovzxdq,     IntrinsicType.Unary));
-            Add(Intrinsic.X86Pmovzxwd,     new IntrinsicInfo(X86Instruction.Pmovzxwd,     IntrinsicType.Unary));
-            Add(Intrinsic.X86Pmulld,       new IntrinsicInfo(X86Instruction.Pmulld,       IntrinsicType.Binary));
-            Add(Intrinsic.X86Pmullw,       new IntrinsicInfo(X86Instruction.Pmullw,       IntrinsicType.Binary));
-            Add(Intrinsic.X86Popcnt,       new IntrinsicInfo(X86Instruction.Popcnt,       IntrinsicType.PopCount));
-            Add(Intrinsic.X86Por,          new IntrinsicInfo(X86Instruction.Por,          IntrinsicType.Binary));
-            Add(Intrinsic.X86Pshufb,       new IntrinsicInfo(X86Instruction.Pshufb,       IntrinsicType.Binary));
-            Add(Intrinsic.X86Pshufd,       new IntrinsicInfo(X86Instruction.Pshufd,       IntrinsicType.BinaryImm));
-            Add(Intrinsic.X86Pslld,        new IntrinsicInfo(X86Instruction.Pslld,        IntrinsicType.Binary));
-            Add(Intrinsic.X86Pslldq,       new IntrinsicInfo(X86Instruction.Pslldq,       IntrinsicType.Binary));
-            Add(Intrinsic.X86Psllq,        new IntrinsicInfo(X86Instruction.Psllq,        IntrinsicType.Binary));
-            Add(Intrinsic.X86Psllw,        new IntrinsicInfo(X86Instruction.Psllw,        IntrinsicType.Binary));
-            Add(Intrinsic.X86Psrad,        new IntrinsicInfo(X86Instruction.Psrad,        IntrinsicType.Binary));
-            Add(Intrinsic.X86Psraw,        new IntrinsicInfo(X86Instruction.Psraw,        IntrinsicType.Binary));
-            Add(Intrinsic.X86Psrld,        new IntrinsicInfo(X86Instruction.Psrld,        IntrinsicType.Binary));
-            Add(Intrinsic.X86Psrlq,        new IntrinsicInfo(X86Instruction.Psrlq,        IntrinsicType.Binary));
-            Add(Intrinsic.X86Psrldq,       new IntrinsicInfo(X86Instruction.Psrldq,       IntrinsicType.Binary));
-            Add(Intrinsic.X86Psrlw,        new IntrinsicInfo(X86Instruction.Psrlw,        IntrinsicType.Binary));
-            Add(Intrinsic.X86Psubb,        new IntrinsicInfo(X86Instruction.Psubb,        IntrinsicType.Binary));
-            Add(Intrinsic.X86Psubd,        new IntrinsicInfo(X86Instruction.Psubd,        IntrinsicType.Binary));
-            Add(Intrinsic.X86Psubq,        new IntrinsicInfo(X86Instruction.Psubq,        IntrinsicType.Binary));
-            Add(Intrinsic.X86Psubw,        new IntrinsicInfo(X86Instruction.Psubw,        IntrinsicType.Binary));
-            Add(Intrinsic.X86Punpckhbw,    new IntrinsicInfo(X86Instruction.Punpckhbw,    IntrinsicType.Binary));
-            Add(Intrinsic.X86Punpckhdq,    new IntrinsicInfo(X86Instruction.Punpckhdq,    IntrinsicType.Binary));
-            Add(Intrinsic.X86Punpckhqdq,   new IntrinsicInfo(X86Instruction.Punpckhqdq,   IntrinsicType.Binary));
-            Add(Intrinsic.X86Punpckhwd,    new IntrinsicInfo(X86Instruction.Punpckhwd,    IntrinsicType.Binary));
-            Add(Intrinsic.X86Punpcklbw,    new IntrinsicInfo(X86Instruction.Punpcklbw,    IntrinsicType.Binary));
-            Add(Intrinsic.X86Punpckldq,    new IntrinsicInfo(X86Instruction.Punpckldq,    IntrinsicType.Binary));
-            Add(Intrinsic.X86Punpcklqdq,   new IntrinsicInfo(X86Instruction.Punpcklqdq,   IntrinsicType.Binary));
-            Add(Intrinsic.X86Punpcklwd,    new IntrinsicInfo(X86Instruction.Punpcklwd,    IntrinsicType.Binary));
-            Add(Intrinsic.X86Pxor,         new IntrinsicInfo(X86Instruction.Pxor,         IntrinsicType.Binary));
-            Add(Intrinsic.X86Rcpps,        new IntrinsicInfo(X86Instruction.Rcpps,        IntrinsicType.Unary));
-            Add(Intrinsic.X86Rcpss,        new IntrinsicInfo(X86Instruction.Rcpss,        IntrinsicType.Unary));
-            Add(Intrinsic.X86Roundpd,      new IntrinsicInfo(X86Instruction.Roundpd,      IntrinsicType.BinaryImm));
-            Add(Intrinsic.X86Roundps,      new IntrinsicInfo(X86Instruction.Roundps,      IntrinsicType.BinaryImm));
-            Add(Intrinsic.X86Roundsd,      new IntrinsicInfo(X86Instruction.Roundsd,      IntrinsicType.BinaryImm));
-            Add(Intrinsic.X86Roundss,      new IntrinsicInfo(X86Instruction.Roundss,      IntrinsicType.BinaryImm));
-            Add(Intrinsic.X86Rsqrtps,      new IntrinsicInfo(X86Instruction.Rsqrtps,      IntrinsicType.Unary));
-            Add(Intrinsic.X86Rsqrtss,      new IntrinsicInfo(X86Instruction.Rsqrtss,      IntrinsicType.Unary));
-            Add(Intrinsic.X86Sha256Msg1,   new IntrinsicInfo(X86Instruction.Sha256Msg1,   IntrinsicType.Binary));
-            Add(Intrinsic.X86Sha256Msg2,   new IntrinsicInfo(X86Instruction.Sha256Msg2,   IntrinsicType.Binary));
-            Add(Intrinsic.X86Sha256Rnds2,  new IntrinsicInfo(X86Instruction.Sha256Rnds2,  IntrinsicType.Ternary));
-            Add(Intrinsic.X86Shufpd,       new IntrinsicInfo(X86Instruction.Shufpd,       IntrinsicType.TernaryImm));
-            Add(Intrinsic.X86Shufps,       new IntrinsicInfo(X86Instruction.Shufps,       IntrinsicType.TernaryImm));
-            Add(Intrinsic.X86Sqrtpd,       new IntrinsicInfo(X86Instruction.Sqrtpd,       IntrinsicType.Unary));
-            Add(Intrinsic.X86Sqrtps,       new IntrinsicInfo(X86Instruction.Sqrtps,       IntrinsicType.Unary));
-            Add(Intrinsic.X86Sqrtsd,       new IntrinsicInfo(X86Instruction.Sqrtsd,       IntrinsicType.Unary));
-            Add(Intrinsic.X86Sqrtss,       new IntrinsicInfo(X86Instruction.Sqrtss,       IntrinsicType.Unary));
-            Add(Intrinsic.X86Subpd,        new IntrinsicInfo(X86Instruction.Subpd,        IntrinsicType.Binary));
-            Add(Intrinsic.X86Subps,        new IntrinsicInfo(X86Instruction.Subps,        IntrinsicType.Binary));
-            Add(Intrinsic.X86Subsd,        new IntrinsicInfo(X86Instruction.Subsd,        IntrinsicType.Binary));
-            Add(Intrinsic.X86Subss,        new IntrinsicInfo(X86Instruction.Subss,        IntrinsicType.Binary));
-            Add(Intrinsic.X86Unpckhpd,     new IntrinsicInfo(X86Instruction.Unpckhpd,     IntrinsicType.Binary));
-            Add(Intrinsic.X86Unpckhps,     new IntrinsicInfo(X86Instruction.Unpckhps,     IntrinsicType.Binary));
-            Add(Intrinsic.X86Unpcklpd,     new IntrinsicInfo(X86Instruction.Unpcklpd,     IntrinsicType.Binary));
-            Add(Intrinsic.X86Unpcklps,     new IntrinsicInfo(X86Instruction.Unpcklps,     IntrinsicType.Binary));
-            Add(Intrinsic.X86Vcvtph2ps,    new IntrinsicInfo(X86Instruction.Vcvtph2ps,    IntrinsicType.Unary));
-            Add(Intrinsic.X86Vcvtps2ph,    new IntrinsicInfo(X86Instruction.Vcvtps2ph,    IntrinsicType.BinaryImm));
-            Add(Intrinsic.X86Vfmadd231ps,  new IntrinsicInfo(X86Instruction.Vfmadd231ps,  IntrinsicType.Fma));
-            Add(Intrinsic.X86Vfmadd231sd,  new IntrinsicInfo(X86Instruction.Vfmadd231sd,  IntrinsicType.Fma));
-            Add(Intrinsic.X86Vfmadd231ss,  new IntrinsicInfo(X86Instruction.Vfmadd231ss,  IntrinsicType.Fma));
-            Add(Intrinsic.X86Vfmsub231sd,  new IntrinsicInfo(X86Instruction.Vfmsub231sd,  IntrinsicType.Fma));
-            Add(Intrinsic.X86Vfmsub231ss,  new IntrinsicInfo(X86Instruction.Vfmsub231ss,  IntrinsicType.Fma));
-            Add(Intrinsic.X86Vfnmadd231ps, new IntrinsicInfo(X86Instruction.Vfnmadd231ps, IntrinsicType.Fma));
-            Add(Intrinsic.X86Vfnmadd231sd, new IntrinsicInfo(X86Instruction.Vfnmadd231sd, IntrinsicType.Fma));
-            Add(Intrinsic.X86Vfnmadd231ss, new IntrinsicInfo(X86Instruction.Vfnmadd231ss, IntrinsicType.Fma));
-            Add(Intrinsic.X86Vfnmsub231sd, new IntrinsicInfo(X86Instruction.Vfnmsub231sd, IntrinsicType.Fma));
-            Add(Intrinsic.X86Vfnmsub231ss, new IntrinsicInfo(X86Instruction.Vfnmsub231ss, IntrinsicType.Fma));
-            Add(Intrinsic.X86Xorpd,        new IntrinsicInfo(X86Instruction.Xorpd,        IntrinsicType.Binary));
-            Add(Intrinsic.X86Xorps,        new IntrinsicInfo(X86Instruction.Xorps,        IntrinsicType.Binary));
+            Add(Intrinsic.X86Addpd,         new IntrinsicInfo(X86Instruction.Addpd,         IntrinsicType.Binary));
+            Add(Intrinsic.X86Addps,         new IntrinsicInfo(X86Instruction.Addps,         IntrinsicType.Binary));
+            Add(Intrinsic.X86Addsd,         new IntrinsicInfo(X86Instruction.Addsd,         IntrinsicType.Binary));
+            Add(Intrinsic.X86Addss,         new IntrinsicInfo(X86Instruction.Addss,         IntrinsicType.Binary));
+            Add(Intrinsic.X86Aesdec,        new IntrinsicInfo(X86Instruction.Aesdec,        IntrinsicType.Binary));
+            Add(Intrinsic.X86Aesdeclast,    new IntrinsicInfo(X86Instruction.Aesdeclast,    IntrinsicType.Binary));
+            Add(Intrinsic.X86Aesenc,        new IntrinsicInfo(X86Instruction.Aesenc,        IntrinsicType.Binary));
+            Add(Intrinsic.X86Aesenclast,    new IntrinsicInfo(X86Instruction.Aesenclast,    IntrinsicType.Binary));
+            Add(Intrinsic.X86Aesimc,        new IntrinsicInfo(X86Instruction.Aesimc,        IntrinsicType.Unary));
+            Add(Intrinsic.X86Andnpd,        new IntrinsicInfo(X86Instruction.Andnpd,        IntrinsicType.Binary));
+            Add(Intrinsic.X86Andnps,        new IntrinsicInfo(X86Instruction.Andnps,        IntrinsicType.Binary));
+            Add(Intrinsic.X86Andpd,         new IntrinsicInfo(X86Instruction.Andpd,         IntrinsicType.Binary));
+            Add(Intrinsic.X86Andps,         new IntrinsicInfo(X86Instruction.Andps,         IntrinsicType.Binary));
+            Add(Intrinsic.X86Blendvpd,      new IntrinsicInfo(X86Instruction.Blendvpd,      IntrinsicType.Ternary));
+            Add(Intrinsic.X86Blendvps,      new IntrinsicInfo(X86Instruction.Blendvps,      IntrinsicType.Ternary));
+            Add(Intrinsic.X86Cmppd,         new IntrinsicInfo(X86Instruction.Cmppd,         IntrinsicType.TernaryImm));
+            Add(Intrinsic.X86Cmpps,         new IntrinsicInfo(X86Instruction.Cmpps,         IntrinsicType.TernaryImm));
+            Add(Intrinsic.X86Cmpsd,         new IntrinsicInfo(X86Instruction.Cmpsd,         IntrinsicType.TernaryImm));
+            Add(Intrinsic.X86Cmpss,         new IntrinsicInfo(X86Instruction.Cmpss,         IntrinsicType.TernaryImm));
+            Add(Intrinsic.X86Comisdeq,      new IntrinsicInfo(X86Instruction.Comisd,        IntrinsicType.Comis_));
+            Add(Intrinsic.X86Comisdge,      new IntrinsicInfo(X86Instruction.Comisd,        IntrinsicType.Comis_));
+            Add(Intrinsic.X86Comisdlt,      new IntrinsicInfo(X86Instruction.Comisd,        IntrinsicType.Comis_));
+            Add(Intrinsic.X86Comisseq,      new IntrinsicInfo(X86Instruction.Comiss,        IntrinsicType.Comis_));
+            Add(Intrinsic.X86Comissge,      new IntrinsicInfo(X86Instruction.Comiss,        IntrinsicType.Comis_));
+            Add(Intrinsic.X86Comisslt,      new IntrinsicInfo(X86Instruction.Comiss,        IntrinsicType.Comis_));
+            Add(Intrinsic.X86Crc32,         new IntrinsicInfo(X86Instruction.Crc32,         IntrinsicType.Crc32));
+            Add(Intrinsic.X86Crc32_16,      new IntrinsicInfo(X86Instruction.Crc32_16,      IntrinsicType.Crc32));
+            Add(Intrinsic.X86Crc32_8,       new IntrinsicInfo(X86Instruction.Crc32_8,       IntrinsicType.Crc32));
+            Add(Intrinsic.X86Cvtdq2pd,      new IntrinsicInfo(X86Instruction.Cvtdq2pd,      IntrinsicType.Unary));
+            Add(Intrinsic.X86Cvtdq2ps,      new IntrinsicInfo(X86Instruction.Cvtdq2ps,      IntrinsicType.Unary));
+            Add(Intrinsic.X86Cvtpd2dq,      new IntrinsicInfo(X86Instruction.Cvtpd2dq,      IntrinsicType.Unary));
+            Add(Intrinsic.X86Cvtpd2ps,      new IntrinsicInfo(X86Instruction.Cvtpd2ps,      IntrinsicType.Unary));
+            Add(Intrinsic.X86Cvtps2dq,      new IntrinsicInfo(X86Instruction.Cvtps2dq,      IntrinsicType.Unary));
+            Add(Intrinsic.X86Cvtps2pd,      new IntrinsicInfo(X86Instruction.Cvtps2pd,      IntrinsicType.Unary));
+            Add(Intrinsic.X86Cvtsd2si,      new IntrinsicInfo(X86Instruction.Cvtsd2si,      IntrinsicType.UnaryToGpr));
+            Add(Intrinsic.X86Cvtsd2ss,      new IntrinsicInfo(X86Instruction.Cvtsd2ss,      IntrinsicType.Binary));
+            Add(Intrinsic.X86Cvtsi2sd,      new IntrinsicInfo(X86Instruction.Cvtsi2sd,      IntrinsicType.BinaryGpr));
+            Add(Intrinsic.X86Cvtsi2si,      new IntrinsicInfo(X86Instruction.Movd,          IntrinsicType.UnaryToGpr));
+            Add(Intrinsic.X86Cvtsi2ss,      new IntrinsicInfo(X86Instruction.Cvtsi2ss,      IntrinsicType.BinaryGpr));
+            Add(Intrinsic.X86Cvtss2sd,      new IntrinsicInfo(X86Instruction.Cvtss2sd,      IntrinsicType.Binary));
+            Add(Intrinsic.X86Cvtss2si,      new IntrinsicInfo(X86Instruction.Cvtss2si,      IntrinsicType.UnaryToGpr));
+            Add(Intrinsic.X86Divpd,         new IntrinsicInfo(X86Instruction.Divpd,         IntrinsicType.Binary));
+            Add(Intrinsic.X86Divps,         new IntrinsicInfo(X86Instruction.Divps,         IntrinsicType.Binary));
+            Add(Intrinsic.X86Divsd,         new IntrinsicInfo(X86Instruction.Divsd,         IntrinsicType.Binary));
+            Add(Intrinsic.X86Divss,         new IntrinsicInfo(X86Instruction.Divss,         IntrinsicType.Binary));
+            Add(Intrinsic.X86Gf2p8affineqb, new IntrinsicInfo(X86Instruction.Gf2p8affineqb, IntrinsicType.TernaryImm));
+            Add(Intrinsic.X86Haddpd,        new IntrinsicInfo(X86Instruction.Haddpd,        IntrinsicType.Binary));
+            Add(Intrinsic.X86Haddps,        new IntrinsicInfo(X86Instruction.Haddps,        IntrinsicType.Binary));
+            Add(Intrinsic.X86Insertps,      new IntrinsicInfo(X86Instruction.Insertps,      IntrinsicType.TernaryImm));
+            Add(Intrinsic.X86Maxpd,         new IntrinsicInfo(X86Instruction.Maxpd,         IntrinsicType.Binary));
+            Add(Intrinsic.X86Maxps,         new IntrinsicInfo(X86Instruction.Maxps,         IntrinsicType.Binary));
+            Add(Intrinsic.X86Maxsd,         new IntrinsicInfo(X86Instruction.Maxsd,         IntrinsicType.Binary));
+            Add(Intrinsic.X86Maxss,         new IntrinsicInfo(X86Instruction.Maxss,         IntrinsicType.Binary));
+            Add(Intrinsic.X86Minpd,         new IntrinsicInfo(X86Instruction.Minpd,         IntrinsicType.Binary));
+            Add(Intrinsic.X86Minps,         new IntrinsicInfo(X86Instruction.Minps,         IntrinsicType.Binary));
+            Add(Intrinsic.X86Minsd,         new IntrinsicInfo(X86Instruction.Minsd,         IntrinsicType.Binary));
+            Add(Intrinsic.X86Minss,         new IntrinsicInfo(X86Instruction.Minss,         IntrinsicType.Binary));
+            Add(Intrinsic.X86Movhlps,       new IntrinsicInfo(X86Instruction.Movhlps,       IntrinsicType.Binary));
+            Add(Intrinsic.X86Movlhps,       new IntrinsicInfo(X86Instruction.Movlhps,       IntrinsicType.Binary));
+            Add(Intrinsic.X86Movss,         new IntrinsicInfo(X86Instruction.Movss,         IntrinsicType.Binary));
+            Add(Intrinsic.X86Mulpd,         new IntrinsicInfo(X86Instruction.Mulpd,         IntrinsicType.Binary));
+            Add(Intrinsic.X86Mulps,         new IntrinsicInfo(X86Instruction.Mulps,         IntrinsicType.Binary));
+            Add(Intrinsic.X86Mulsd,         new IntrinsicInfo(X86Instruction.Mulsd,         IntrinsicType.Binary));
+            Add(Intrinsic.X86Mulss,         new IntrinsicInfo(X86Instruction.Mulss,         IntrinsicType.Binary));
+            Add(Intrinsic.X86Mxcsrmb,       new IntrinsicInfo(X86Instruction.None,          IntrinsicType.Mxcsr)); // Mask bits.
+            Add(Intrinsic.X86Mxcsrub,       new IntrinsicInfo(X86Instruction.None,          IntrinsicType.Mxcsr)); // Unmask bits.
+            Add(Intrinsic.X86Paddb,         new IntrinsicInfo(X86Instruction.Paddb,         IntrinsicType.Binary));
+            Add(Intrinsic.X86Paddd,         new IntrinsicInfo(X86Instruction.Paddd,         IntrinsicType.Binary));
+            Add(Intrinsic.X86Paddq,         new IntrinsicInfo(X86Instruction.Paddq,         IntrinsicType.Binary));
+            Add(Intrinsic.X86Paddw,         new IntrinsicInfo(X86Instruction.Paddw,         IntrinsicType.Binary));
+            Add(Intrinsic.X86Palignr,       new IntrinsicInfo(X86Instruction.Palignr,       IntrinsicType.TernaryImm));
+            Add(Intrinsic.X86Pand,          new IntrinsicInfo(X86Instruction.Pand,          IntrinsicType.Binary));
+            Add(Intrinsic.X86Pandn,         new IntrinsicInfo(X86Instruction.Pandn,         IntrinsicType.Binary));
+            Add(Intrinsic.X86Pavgb,         new IntrinsicInfo(X86Instruction.Pavgb,         IntrinsicType.Binary));
+            Add(Intrinsic.X86Pavgw,         new IntrinsicInfo(X86Instruction.Pavgw,         IntrinsicType.Binary));
+            Add(Intrinsic.X86Pblendvb,      new IntrinsicInfo(X86Instruction.Pblendvb,      IntrinsicType.Ternary));
+            Add(Intrinsic.X86Pclmulqdq,     new IntrinsicInfo(X86Instruction.Pclmulqdq,     IntrinsicType.TernaryImm));
+            Add(Intrinsic.X86Pcmpeqb,       new IntrinsicInfo(X86Instruction.Pcmpeqb,       IntrinsicType.Binary));
+            Add(Intrinsic.X86Pcmpeqd,       new IntrinsicInfo(X86Instruction.Pcmpeqd,       IntrinsicType.Binary));
+            Add(Intrinsic.X86Pcmpeqq,       new IntrinsicInfo(X86Instruction.Pcmpeqq,       IntrinsicType.Binary));
+            Add(Intrinsic.X86Pcmpeqw,       new IntrinsicInfo(X86Instruction.Pcmpeqw,       IntrinsicType.Binary));
+            Add(Intrinsic.X86Pcmpgtb,       new IntrinsicInfo(X86Instruction.Pcmpgtb,       IntrinsicType.Binary));
+            Add(Intrinsic.X86Pcmpgtd,       new IntrinsicInfo(X86Instruction.Pcmpgtd,       IntrinsicType.Binary));
+            Add(Intrinsic.X86Pcmpgtq,       new IntrinsicInfo(X86Instruction.Pcmpgtq,       IntrinsicType.Binary));
+            Add(Intrinsic.X86Pcmpgtw,       new IntrinsicInfo(X86Instruction.Pcmpgtw,       IntrinsicType.Binary));
+            Add(Intrinsic.X86Pmaxsb,        new IntrinsicInfo(X86Instruction.Pmaxsb,        IntrinsicType.Binary));
+            Add(Intrinsic.X86Pmaxsd,        new IntrinsicInfo(X86Instruction.Pmaxsd,        IntrinsicType.Binary));
+            Add(Intrinsic.X86Pmaxsw,        new IntrinsicInfo(X86Instruction.Pmaxsw,        IntrinsicType.Binary));
+            Add(Intrinsic.X86Pmaxub,        new IntrinsicInfo(X86Instruction.Pmaxub,        IntrinsicType.Binary));
+            Add(Intrinsic.X86Pmaxud,        new IntrinsicInfo(X86Instruction.Pmaxud,        IntrinsicType.Binary));
+            Add(Intrinsic.X86Pmaxuw,        new IntrinsicInfo(X86Instruction.Pmaxuw,        IntrinsicType.Binary));
+            Add(Intrinsic.X86Pminsb,        new IntrinsicInfo(X86Instruction.Pminsb,        IntrinsicType.Binary));
+            Add(Intrinsic.X86Pminsd,        new IntrinsicInfo(X86Instruction.Pminsd,        IntrinsicType.Binary));
+            Add(Intrinsic.X86Pminsw,        new IntrinsicInfo(X86Instruction.Pminsw,        IntrinsicType.Binary));
+            Add(Intrinsic.X86Pminub,        new IntrinsicInfo(X86Instruction.Pminub,        IntrinsicType.Binary));
+            Add(Intrinsic.X86Pminud,        new IntrinsicInfo(X86Instruction.Pminud,        IntrinsicType.Binary));
+            Add(Intrinsic.X86Pminuw,        new IntrinsicInfo(X86Instruction.Pminuw,        IntrinsicType.Binary));
+            Add(Intrinsic.X86Pmovsxbw,      new IntrinsicInfo(X86Instruction.Pmovsxbw,      IntrinsicType.Unary));
+            Add(Intrinsic.X86Pmovsxdq,      new IntrinsicInfo(X86Instruction.Pmovsxdq,      IntrinsicType.Unary));
+            Add(Intrinsic.X86Pmovsxwd,      new IntrinsicInfo(X86Instruction.Pmovsxwd,      IntrinsicType.Unary));
+            Add(Intrinsic.X86Pmovzxbw,      new IntrinsicInfo(X86Instruction.Pmovzxbw,      IntrinsicType.Unary));
+            Add(Intrinsic.X86Pmovzxdq,      new IntrinsicInfo(X86Instruction.Pmovzxdq,      IntrinsicType.Unary));
+            Add(Intrinsic.X86Pmovzxwd,      new IntrinsicInfo(X86Instruction.Pmovzxwd,      IntrinsicType.Unary));
+            Add(Intrinsic.X86Pmulld,        new IntrinsicInfo(X86Instruction.Pmulld,        IntrinsicType.Binary));
+            Add(Intrinsic.X86Pmullw,        new IntrinsicInfo(X86Instruction.Pmullw,        IntrinsicType.Binary));
+            Add(Intrinsic.X86Popcnt,        new IntrinsicInfo(X86Instruction.Popcnt,        IntrinsicType.PopCount));
+            Add(Intrinsic.X86Por,           new IntrinsicInfo(X86Instruction.Por,           IntrinsicType.Binary));
+            Add(Intrinsic.X86Pshufb,        new IntrinsicInfo(X86Instruction.Pshufb,        IntrinsicType.Binary));
+            Add(Intrinsic.X86Pshufd,        new IntrinsicInfo(X86Instruction.Pshufd,        IntrinsicType.BinaryImm));
+            Add(Intrinsic.X86Pslld,         new IntrinsicInfo(X86Instruction.Pslld,         IntrinsicType.Binary));
+            Add(Intrinsic.X86Pslldq,        new IntrinsicInfo(X86Instruction.Pslldq,        IntrinsicType.Binary));
+            Add(Intrinsic.X86Psllq,         new IntrinsicInfo(X86Instruction.Psllq,         IntrinsicType.Binary));
+            Add(Intrinsic.X86Psllw,         new IntrinsicInfo(X86Instruction.Psllw,         IntrinsicType.Binary));
+            Add(Intrinsic.X86Psrad,         new IntrinsicInfo(X86Instruction.Psrad,         IntrinsicType.Binary));
+            Add(Intrinsic.X86Psraw,         new IntrinsicInfo(X86Instruction.Psraw,         IntrinsicType.Binary));
+            Add(Intrinsic.X86Psrld,         new IntrinsicInfo(X86Instruction.Psrld,         IntrinsicType.Binary));
+            Add(Intrinsic.X86Psrlq,         new IntrinsicInfo(X86Instruction.Psrlq,         IntrinsicType.Binary));
+            Add(Intrinsic.X86Psrldq,        new IntrinsicInfo(X86Instruction.Psrldq,        IntrinsicType.Binary));
+            Add(Intrinsic.X86Psrlw,         new IntrinsicInfo(X86Instruction.Psrlw,         IntrinsicType.Binary));
+            Add(Intrinsic.X86Psubb,         new IntrinsicInfo(X86Instruction.Psubb,         IntrinsicType.Binary));
+            Add(Intrinsic.X86Psubd,         new IntrinsicInfo(X86Instruction.Psubd,         IntrinsicType.Binary));
+            Add(Intrinsic.X86Psubq,         new IntrinsicInfo(X86Instruction.Psubq,         IntrinsicType.Binary));
+            Add(Intrinsic.X86Psubw,         new IntrinsicInfo(X86Instruction.Psubw,         IntrinsicType.Binary));
+            Add(Intrinsic.X86Punpckhbw,     new IntrinsicInfo(X86Instruction.Punpckhbw,     IntrinsicType.Binary));
+            Add(Intrinsic.X86Punpckhdq,     new IntrinsicInfo(X86Instruction.Punpckhdq,     IntrinsicType.Binary));
+            Add(Intrinsic.X86Punpckhqdq,    new IntrinsicInfo(X86Instruction.Punpckhqdq,    IntrinsicType.Binary));
+            Add(Intrinsic.X86Punpckhwd,     new IntrinsicInfo(X86Instruction.Punpckhwd,     IntrinsicType.Binary));
+            Add(Intrinsic.X86Punpcklbw,     new IntrinsicInfo(X86Instruction.Punpcklbw,     IntrinsicType.Binary));
+            Add(Intrinsic.X86Punpckldq,     new IntrinsicInfo(X86Instruction.Punpckldq,     IntrinsicType.Binary));
+            Add(Intrinsic.X86Punpcklqdq,    new IntrinsicInfo(X86Instruction.Punpcklqdq,    IntrinsicType.Binary));
+            Add(Intrinsic.X86Punpcklwd,     new IntrinsicInfo(X86Instruction.Punpcklwd,     IntrinsicType.Binary));
+            Add(Intrinsic.X86Pxor,          new IntrinsicInfo(X86Instruction.Pxor,          IntrinsicType.Binary));
+            Add(Intrinsic.X86Rcpps,         new IntrinsicInfo(X86Instruction.Rcpps,         IntrinsicType.Unary));
+            Add(Intrinsic.X86Rcpss,         new IntrinsicInfo(X86Instruction.Rcpss,         IntrinsicType.Unary));
+            Add(Intrinsic.X86Roundpd,       new IntrinsicInfo(X86Instruction.Roundpd,       IntrinsicType.BinaryImm));
+            Add(Intrinsic.X86Roundps,       new IntrinsicInfo(X86Instruction.Roundps,       IntrinsicType.BinaryImm));
+            Add(Intrinsic.X86Roundsd,       new IntrinsicInfo(X86Instruction.Roundsd,       IntrinsicType.BinaryImm));
+            Add(Intrinsic.X86Roundss,       new IntrinsicInfo(X86Instruction.Roundss,       IntrinsicType.BinaryImm));
+            Add(Intrinsic.X86Rsqrtps,       new IntrinsicInfo(X86Instruction.Rsqrtps,       IntrinsicType.Unary));
+            Add(Intrinsic.X86Rsqrtss,       new IntrinsicInfo(X86Instruction.Rsqrtss,       IntrinsicType.Unary));
+            Add(Intrinsic.X86Sha256Msg1,    new IntrinsicInfo(X86Instruction.Sha256Msg1,    IntrinsicType.Binary));
+            Add(Intrinsic.X86Sha256Msg2,    new IntrinsicInfo(X86Instruction.Sha256Msg2,    IntrinsicType.Binary));
+            Add(Intrinsic.X86Sha256Rnds2,   new IntrinsicInfo(X86Instruction.Sha256Rnds2,   IntrinsicType.Ternary));
+            Add(Intrinsic.X86Shufpd,        new IntrinsicInfo(X86Instruction.Shufpd,        IntrinsicType.TernaryImm));
+            Add(Intrinsic.X86Shufps,        new IntrinsicInfo(X86Instruction.Shufps,        IntrinsicType.TernaryImm));
+            Add(Intrinsic.X86Sqrtpd,        new IntrinsicInfo(X86Instruction.Sqrtpd,        IntrinsicType.Unary));
+            Add(Intrinsic.X86Sqrtps,        new IntrinsicInfo(X86Instruction.Sqrtps,        IntrinsicType.Unary));
+            Add(Intrinsic.X86Sqrtsd,        new IntrinsicInfo(X86Instruction.Sqrtsd,        IntrinsicType.Unary));
+            Add(Intrinsic.X86Sqrtss,        new IntrinsicInfo(X86Instruction.Sqrtss,        IntrinsicType.Unary));
+            Add(Intrinsic.X86Subpd,         new IntrinsicInfo(X86Instruction.Subpd,         IntrinsicType.Binary));
+            Add(Intrinsic.X86Subps,         new IntrinsicInfo(X86Instruction.Subps,         IntrinsicType.Binary));
+            Add(Intrinsic.X86Subsd,         new IntrinsicInfo(X86Instruction.Subsd,         IntrinsicType.Binary));
+            Add(Intrinsic.X86Subss,         new IntrinsicInfo(X86Instruction.Subss,         IntrinsicType.Binary));
+            Add(Intrinsic.X86Unpckhpd,      new IntrinsicInfo(X86Instruction.Unpckhpd,      IntrinsicType.Binary));
+            Add(Intrinsic.X86Unpckhps,      new IntrinsicInfo(X86Instruction.Unpckhps,      IntrinsicType.Binary));
+            Add(Intrinsic.X86Unpcklpd,      new IntrinsicInfo(X86Instruction.Unpcklpd,      IntrinsicType.Binary));
+            Add(Intrinsic.X86Unpcklps,      new IntrinsicInfo(X86Instruction.Unpcklps,      IntrinsicType.Binary));
+            Add(Intrinsic.X86Vcvtph2ps,     new IntrinsicInfo(X86Instruction.Vcvtph2ps,     IntrinsicType.Unary));
+            Add(Intrinsic.X86Vcvtps2ph,     new IntrinsicInfo(X86Instruction.Vcvtps2ph,     IntrinsicType.BinaryImm));
+            Add(Intrinsic.X86Vfmadd231ps,   new IntrinsicInfo(X86Instruction.Vfmadd231ps,   IntrinsicType.Fma));
+            Add(Intrinsic.X86Vfmadd231sd,   new IntrinsicInfo(X86Instruction.Vfmadd231sd,   IntrinsicType.Fma));
+            Add(Intrinsic.X86Vfmadd231ss,   new IntrinsicInfo(X86Instruction.Vfmadd231ss,   IntrinsicType.Fma));
+            Add(Intrinsic.X86Vfmsub231sd,   new IntrinsicInfo(X86Instruction.Vfmsub231sd,   IntrinsicType.Fma));
+            Add(Intrinsic.X86Vfmsub231ss,   new IntrinsicInfo(X86Instruction.Vfmsub231ss,   IntrinsicType.Fma));
+            Add(Intrinsic.X86Vfnmadd231ps,  new IntrinsicInfo(X86Instruction.Vfnmadd231ps,  IntrinsicType.Fma));
+            Add(Intrinsic.X86Vfnmadd231sd,  new IntrinsicInfo(X86Instruction.Vfnmadd231sd,  IntrinsicType.Fma));
+            Add(Intrinsic.X86Vfnmadd231ss,  new IntrinsicInfo(X86Instruction.Vfnmadd231ss,  IntrinsicType.Fma));
+            Add(Intrinsic.X86Vfnmsub231sd,  new IntrinsicInfo(X86Instruction.Vfnmsub231sd,  IntrinsicType.Fma));
+            Add(Intrinsic.X86Vfnmsub231ss,  new IntrinsicInfo(X86Instruction.Vfnmsub231ss,  IntrinsicType.Fma));
+            Add(Intrinsic.X86Xorpd,         new IntrinsicInfo(X86Instruction.Xorpd,         IntrinsicType.Binary));
+            Add(Intrinsic.X86Xorps,         new IntrinsicInfo(X86Instruction.Xorps,         IntrinsicType.Binary));
         }
 
         private static void Add(Intrinsic intrin, IntrinsicInfo info)

ARMeilleure/CodeGen/X86/X86Instruction.cs

@@ -54,6 +54,7 @@ namespace ARMeilleure.CodeGen.X86
         Divps,
         Divsd,
         Divss,
+        Gf2p8affineqb,
         Haddpd,
         Haddps,
         Idiv,

ARMeilleure/Common/BitUtils.cs

@@ -1,15 +1,11 @@
+using System;
 using System.Numerics;
 
 namespace ARMeilleure.Common
 {
     static class BitUtils
     {
-        private static readonly sbyte[] HbsNibbleLut;
-
-        static BitUtils()
-        {
-            HbsNibbleLut = new sbyte[] { -1, 0, 1, 1, 2, 2, 2, 2, 3, 3, 3, 3, 3, 3, 3, 3 };
-        }
+        private static ReadOnlySpan<sbyte> HbsNibbleLut => new sbyte[] { -1, 0, 1, 1, 2, 2, 2, 2, 3, 3, 3, 3, 3, 3, 3, 3 };
 
         public static long FillWithOnes(int bits)
         {

ARMeilleure/Decoders/InstDescriptor.cs

@@ -2,7 +2,7 @@ using ARMeilleure.Instructions;
 
 namespace ARMeilleure.Decoders
 {
-    struct InstDescriptor
+    readonly struct InstDescriptor
     {
         public static InstDescriptor Undefined => new InstDescriptor(InstName.Und, InstEmit.Und);
 

ARMeilleure/Decoders/OpCode32SimdCvtTB.cs

@@ -0,0 +1,44 @@
+namespace ARMeilleure.Decoders
+{
+    class OpCode32SimdCvtTB : OpCode32, IOpCode32Simd
+    {
+        public int Vd { get; }
+        public int Vm { get; }
+        public bool Op { get; } // Convert to Half / Convert from Half
+        public bool T { get; } // Top / Bottom
+        public int Size { get; } // Double / Single
+
+        public new static OpCode Create(InstDescriptor inst, ulong address, int opCode) => new OpCode32SimdCvtTB(inst, address, opCode, false);
+        public static OpCode CreateT32(InstDescriptor inst, ulong address, int opCode) => new OpCode32SimdCvtTB(inst, address, opCode, true);
+
+        public OpCode32SimdCvtTB(InstDescriptor inst, ulong address, int opCode, bool isThumb) : base(inst, address, opCode)
+        {
+            IsThumb = isThumb;
+
+            Op   = ((opCode >> 16) & 0x1) != 0;
+            T    = ((opCode >> 7) & 0x1) != 0;
+            Size = ((opCode >> 8) & 0x1);
+
+            RegisterSize = Size == 1 ? RegisterSize.Int64 : RegisterSize.Int32;
+
+            if (Size == 1)
+            {
+                if (Op)
+                {
+                    Vm = ((opCode >> 1) & 0x10) | ((opCode >> 0) & 0xf);
+                    Vd = ((opCode >> 22) & 0x1) | ((opCode >> 11) & 0x1e);
+                }
+                else
+                {
+                    Vm = ((opCode >> 5) & 0x1) | ((opCode << 1) & 0x1e);
+                    Vd = ((opCode >> 18) & 0x10) | ((opCode >> 12) & 0xf);
+                }
+            }
+            else
+            {
+                Vm = ((opCode >> 5) & 0x1) | ((opCode << 1) & 0x1e);
+                Vd = ((opCode >> 22) & 0x1) | ((opCode >> 11) & 0x1e);
+            }
+        }
+    }
+}

ARMeilleure/Decoders/OpCodeTable.cs

@@ -11,7 +11,7 @@ namespace ARMeilleure.Decoders
 
         private const int FastLookupSize = 0x1000;
 
-        private struct InstInfo
+        private readonly struct InstInfo
         {
             public int Mask  { get; }
             public int Value { get; }
@@ -828,6 +828,7 @@ namespace ARMeilleure.Decoders
             SetVfp("<<<<11101x11110xxxxx101x11x0xxxx", InstName.Vcvt,   InstEmit32.Vcvt_FI,  OpCode32SimdCvtFI.Create,       OpCode32SimdCvtFI.CreateT32); // FP32 to int.
             SetVfp("<<<<11101x111000xxxx101xx1x0xxxx", InstName.Vcvt,   InstEmit32.Vcvt_FI,  OpCode32SimdCvtFI.Create,       OpCode32SimdCvtFI.CreateT32); // Int to FP32.
             SetVfp("111111101x1111xxxxxx101xx1x0xxxx", InstName.Vcvt,   InstEmit32.Vcvt_RM,  OpCode32SimdCvtFI.Create,       OpCode32SimdCvtFI.CreateT32); // The many FP32 to int encodings (fp).
+            SetVfp("<<<<11101x11001xxxxx101xx1x0xxxx", InstName.Vcvt,   InstEmit32.Vcvt_TB,  OpCode32SimdCvtTB.Create,       OpCode32SimdCvtTB.CreateT32);
             SetVfp("<<<<11101x00xxxxxxxx101xx0x0xxxx", InstName.Vdiv,   InstEmit32.Vdiv_S,   OpCode32SimdRegS.Create,        OpCode32SimdRegS.CreateT32);
             SetVfp("<<<<11101xx0xxxxxxxx1011x0x10000", InstName.Vdup,   InstEmit32.Vdup,     OpCode32SimdDupGP.Create,       OpCode32SimdDupGP.CreateT32);
             SetVfp("<<<<11101x10xxxxxxxx101xx0x0xxxx", InstName.Vfma,   InstEmit32.Vfma_S,   OpCode32SimdRegS.Create,        OpCode32SimdRegS.CreateT32);

ARMeilleure/Diagnostics/Symbols.cs

@@ -6,7 +6,7 @@ namespace ARMeilleure.Diagnostics
 {
     static class Symbols
     {
-        private struct RangedSymbol
+        private readonly struct RangedSymbol
         {
             public readonly ulong Start;
             public readonly ulong End;

ARMeilleure/Diagnostics/TranslatorEventSource.cs

@@ -25,13 +25,13 @@ namespace ARMeilleure.Diagnostics
             _funcTabSizeCounter = new PollingCounter("addr-tab-alloc", this, () => _funcTabSize / 1024d / 1024d)
             {
                 DisplayName = "AddressTable Total Bytes Allocated",
-                DisplayUnits = "MB"
+                DisplayUnits = "MiB"
             };
 
             _funcTabLeafSizeCounter = new PollingCounter("addr-tab-leaf-alloc", this, () => _funcTabLeafSize / 1024d / 1024d)
             {
                 DisplayName = "AddressTable Total Leaf Bytes Allocated",
-                DisplayUnits = "MB"
+                DisplayUnits = "MiB"
             };
         }
 

ARMeilleure/Instructions/InstEmitSimdArithmetic.cs

@@ -726,7 +726,7 @@ namespace ARMeilleure.Instructions
             {
                 EmitVectorAcrossVectorOpF(context, (op1, op2) =>
                 {
-                    return context.Call(typeof(SoftFloat32).GetMethod(nameof(SoftFloat32.FPMaxNum)), op1, op2);
+                    return EmitSoftFloatCall(context, nameof(SoftFloat32.FPMaxNum), op1, op2);
                 });
             }
         }
@@ -774,7 +774,7 @@ namespace ARMeilleure.Instructions
             {
                 EmitVectorAcrossVectorOpF(context, (op1, op2) =>
                 {
-                    return context.Call(typeof(SoftFloat32).GetMethod(nameof(SoftFloat32.FPMax)), op1, op2);
+                    return EmitSoftFloatCall(context, nameof(SoftFloat32.FPMax), op1, op2);
                 });
             }
         }
@@ -900,7 +900,7 @@ namespace ARMeilleure.Instructions
             {
                 EmitVectorAcrossVectorOpF(context, (op1, op2) =>
                 {
-                    return context.Call(typeof(SoftFloat32).GetMethod(nameof(SoftFloat32.FPMinNum)), op1, op2);
+                    return EmitSoftFloatCall(context, nameof(SoftFloat32.FPMinNum), op1, op2);
                 });
             }
         }
@@ -948,7 +948,7 @@ namespace ARMeilleure.Instructions
             {
                 EmitVectorAcrossVectorOpF(context, (op1, op2) =>
                 {
-                    return context.Call(typeof(SoftFloat32).GetMethod(nameof(SoftFloat32.FPMin)), op1, op2);
+                    return EmitSoftFloatCall(context, nameof(SoftFloat32.FPMin), op1, op2);
                 });
             }
         }
@@ -1617,53 +1617,47 @@ namespace ARMeilleure.Instructions
 
         public static void Frinta_S(ArmEmitterContext context)
         {
-            EmitScalarUnaryOpF(context, (op1) =>
+            if (Optimizations.UseSse41)
             {
-                return EmitRoundMathCall(context, MidpointRounding.AwayFromZero, op1);
-            });
+                EmitSse41ScalarRoundOpF(context, FPRoundingMode.ToNearestAway);
+            }
+            else
+            {
+                EmitScalarUnaryOpF(context, (op1) =>
+                {
+                    return EmitRoundMathCall(context, MidpointRounding.AwayFromZero, op1);
+                });
+            }
         }
 
         public static void Frinta_V(ArmEmitterContext context)
         {
-            EmitVectorUnaryOpF(context, (op1) =>
+            if (Optimizations.UseSse41)
             {
-                return EmitRoundMathCall(context, MidpointRounding.AwayFromZero, op1);
-            });
+                EmitSse41VectorRoundOpF(context, FPRoundingMode.ToNearestAway);
+            }
+            else
+            {
+                EmitVectorUnaryOpF(context, (op1) =>
+                {
+                    return EmitRoundMathCall(context, MidpointRounding.AwayFromZero, op1);
+                });
+            }
         }
 
         public static void Frinti_S(ArmEmitterContext context)
         {
-            OpCodeSimd op = (OpCodeSimd)context.CurrOp;
-
             EmitScalarUnaryOpF(context, (op1) =>
             {
-                if (op.Size == 0)
-                {
-                    return context.Call(typeof(SoftFallback).GetMethod(nameof(SoftFallback.RoundF)), op1);
-                }
-                else /* if (op.Size == 1) */
-                {
-                    return context.Call(typeof(SoftFallback).GetMethod(nameof(SoftFallback.Round)), op1);
-                }
+                return EmitRoundByRMode(context, op1);
             });
         }
 
         public static void Frinti_V(ArmEmitterContext context)
         {
-            OpCodeSimd op = (OpCodeSimd)context.CurrOp;
-
-            int sizeF = op.Size & 1;
-
             EmitVectorUnaryOpF(context, (op1) =>
             {
-                if (sizeF == 0)
-                {
-                    return context.Call(typeof(SoftFallback).GetMethod(nameof(SoftFallback.RoundF)), op1);
-                }
-                else /* if (sizeF == 1) */
-                {
-                    return context.Call(typeof(SoftFallback).GetMethod(nameof(SoftFallback.Round)), op1);
-                }
+                return EmitRoundByRMode(context, op1);
             });
         }
 
@@ -1759,37 +1753,17 @@ namespace ARMeilleure.Instructions
 
         public static void Frintx_S(ArmEmitterContext context)
         {
-            OpCodeSimd op = (OpCodeSimd)context.CurrOp;
-
             EmitScalarUnaryOpF(context, (op1) =>
             {
-                if (op.Size == 0)
-                {
-                    return context.Call(typeof(SoftFallback).GetMethod(nameof(SoftFallback.RoundF)), op1);
-                }
-                else /* if (op.Size == 1) */
-                {
-                    return context.Call(typeof(SoftFallback).GetMethod(nameof(SoftFallback.Round)), op1);
-                }
+                return EmitRoundByRMode(context, op1);
             });
         }
 
         public static void Frintx_V(ArmEmitterContext context)
         {
-            OpCodeSimd op = (OpCodeSimd)context.CurrOp;
-
-            int sizeF = op.Size & 1;
-
             EmitVectorUnaryOpF(context, (op1) =>
             {
-                if (sizeF == 0)
-                {
-                    return context.Call(typeof(SoftFallback).GetMethod(nameof(SoftFallback.RoundF)), op1);
-                }
-                else /* if (sizeF == 1) */
-                {
-                    return context.Call(typeof(SoftFallback).GetMethod(nameof(SoftFallback.Round)), op1);
-                }
+                return EmitRoundByRMode(context, op1);
             });
         }
 
@@ -3556,9 +3530,18 @@ namespace ARMeilleure.Instructions
 
             Operand n = GetVec(op.Rn);
 
-            Intrinsic inst = (op.Size & 1) != 0 ? Intrinsic.X86Roundsd : Intrinsic.X86Roundss;
+            Operand res;
 
-            Operand res = context.AddIntrinsic(inst, n, Const(X86GetRoundControl(roundMode)));
+            if (roundMode != FPRoundingMode.ToNearestAway)
+            {
+                Intrinsic inst = (op.Size & 1) != 0 ? Intrinsic.X86Roundsd : Intrinsic.X86Roundss;
+
+                res = context.AddIntrinsic(inst, n, Const(X86GetRoundControl(roundMode)));
+            }
+            else
+            {
+                res = EmitSse41RoundToNearestWithTiesToAwayOpF(context, n, scalar: true);
+            }
 
             if ((op.Size & 1) != 0)
             {
@@ -3578,9 +3561,18 @@ namespace ARMeilleure.Instructions
 
             Operand n = GetVec(op.Rn);
 
-            Intrinsic inst = (op.Size & 1) != 0 ? Intrinsic.X86Roundpd : Intrinsic.X86Roundps;
+            Operand res;
 
-            Operand res = context.AddIntrinsic(inst, n, Const(X86GetRoundControl(roundMode)));
+            if (roundMode != FPRoundingMode.ToNearestAway)
+            {
+                Intrinsic inst = (op.Size & 1) != 0 ? Intrinsic.X86Roundpd : Intrinsic.X86Roundps;
+
+                res = context.AddIntrinsic(inst, n, Const(X86GetRoundControl(roundMode)));
+            }
+            else
+            {
+                res = EmitSse41RoundToNearestWithTiesToAwayOpF(context, n, scalar: false);
+            }
 
             if (op.RegisterSize == RegisterSize.Simd64)
             {

ARMeilleure/Instructions/InstEmitSimdCmp32.cs

@@ -3,7 +3,6 @@ using ARMeilleure.IntermediateRepresentation;
 using ARMeilleure.State;
 using ARMeilleure.Translation;
 using System;
-using System.Reflection;
 
 using static ARMeilleure.Instructions.InstEmitHelper;
 using static ARMeilleure.Instructions.InstEmitSimdHelper;
@@ -178,37 +177,20 @@ namespace ARMeilleure.Instructions
 
         private static void EmitCmpOpF32(ArmEmitterContext context, string name, bool zero)
         {
-            Operand one = Const(1);
             if (zero)
             {
                 EmitVectorUnaryOpF32(context, (m) =>
                 {
-                    OperandType type = m.Type;
+                    Operand zeroOp = m.Type == OperandType.FP64 ? ConstF(0.0d) : ConstF(0.0f);
 
-                    if (type == OperandType.FP64)
-                    {
-                        return context.Call(typeof(SoftFloat64).GetMethod(name), m, ConstF(0.0d), one);
-                    }
-                    else
-                    {
-                        return context.Call(typeof(SoftFloat32).GetMethod(name), m, ConstF(0.0f), one);
-                    }
+                    return EmitSoftFloatCallDefaultFpscr(context, name, m, zeroOp);
                 });
             }
             else
             {
                 EmitVectorBinaryOpF32(context, (n, m) =>
                 {
-                    OperandType type = n.Type;
-
-                    if (type == OperandType.FP64)
-                    {
-                        return context.Call(typeof(SoftFloat64).GetMethod(name), n, m, one);
-                    }
-                    else
-                    {
-                        return context.Call(typeof(SoftFloat32).GetMethod(name), n, m, one);
-                    }
+                    return EmitSoftFloatCallDefaultFpscr(context, name, n, m);
                 });
             }
         }
@@ -357,11 +339,7 @@ namespace ARMeilleure.Instructions
                     me = ExtractScalar(context, type, op.Vm);
                 }
 
-                MethodInfo info = sizeF != 0
-                    ? typeof(SoftFloat64).GetMethod(nameof(SoftFloat64.FPCompare))
-                    : typeof(SoftFloat32).GetMethod(nameof(SoftFloat32.FPCompare));
-
-                Operand nzcv = context.Call(info, ne, me, Const(signalNaNs));
+                Operand nzcv = EmitSoftFloatCall(context, nameof(SoftFloat32.FPCompare), ne, me, Const(signalNaNs));
 
                 EmitSetFpscrNzcv(context, nzcv);
             }

ARMeilleure/Instructions/InstEmitSimdCvt.cs

@@ -76,7 +76,9 @@ namespace ARMeilleure.Instructions
                 {
                     Operand ne = context.VectorExtract(OperandType.FP32, GetVec(op.Rn), 0);
 
+                    context.StoreToContext();
                     Operand res = context.Call(typeof(SoftFloat32_16).GetMethod(nameof(SoftFloat32_16.FPConvert)), ne);
+                    context.LoadFromContext();
 
                     res = context.ZeroExtend16(OperandType.I64, res);
 
@@ -98,7 +100,9 @@ namespace ARMeilleure.Instructions
                 {
                     Operand ne = EmitVectorExtractZx(context, op.Rn, 0, 1);
 
+                    context.StoreToContext();
                     Operand res = context.Call(typeof(SoftFloat16_32).GetMethod(nameof(SoftFloat16_32.FPConvert)), ne);
+                    context.LoadFromContext();
 
                     context.Copy(GetVec(op.Rd), context.VectorInsert(context.VectorZero(), res, 0));
                 }
@@ -120,7 +124,9 @@ namespace ARMeilleure.Instructions
                 {
                     Operand ne = context.VectorExtract(OperandType.FP64, GetVec(op.Rn), 0);
 
+                    context.StoreToContext();
                     Operand res = context.Call(typeof(SoftFloat64_16).GetMethod(nameof(SoftFloat64_16.FPConvert)), ne);
+                    context.LoadFromContext();
 
                     res = context.ZeroExtend16(OperandType.I64, res);
 
@@ -143,7 +149,9 @@ namespace ARMeilleure.Instructions
                 {
                     Operand ne = EmitVectorExtractZx(context, op.Rn, 0, 1);
 
+                    context.StoreToContext();
                     Operand res = context.Call(typeof(SoftFloat16_64).GetMethod(nameof(SoftFloat16_64.FPConvert)), ne);
+                    context.LoadFromContext();
 
                     context.Copy(GetVec(op.Rd), context.VectorInsert(context.VectorZero(), res, 0));
                 }
@@ -156,32 +164,74 @@ namespace ARMeilleure.Instructions
 
         public static void Fcvtas_Gp(ArmEmitterContext context)
         {
-            EmitFcvt_s_Gp(context, (op1) => EmitRoundMathCall(context, MidpointRounding.AwayFromZero, op1));
+            if (Optimizations.UseSse41)
+            {
+                EmitSse41Fcvts_Gp(context, FPRoundingMode.ToNearestAway, isFixed: false);
+            }
+            else
+            {
+                EmitFcvt_s_Gp(context, (op1) => EmitRoundMathCall(context, MidpointRounding.AwayFromZero, op1));
+            }
         }
 
         public static void Fcvtas_S(ArmEmitterContext context)
         {
-            EmitFcvt(context, (op1) => EmitRoundMathCall(context, MidpointRounding.AwayFromZero, op1), signed: true, scalar: true);
+            if (Optimizations.UseSse41)
+            {
+                EmitSse41FcvtsOpF(context, FPRoundingMode.ToNearestAway, scalar: true);
+            }
+            else
+            {
+                EmitFcvt(context, (op1) => EmitRoundMathCall(context, MidpointRounding.AwayFromZero, op1), signed: true, scalar: true);
+            }
         }
 
         public static void Fcvtas_V(ArmEmitterContext context)
         {
-            EmitFcvt(context, (op1) => EmitRoundMathCall(context, MidpointRounding.AwayFromZero, op1), signed: true, scalar: false);
+            if (Optimizations.UseSse41)
+            {
+                EmitSse41FcvtsOpF(context, FPRoundingMode.ToNearestAway, scalar: false);
+            }
+            else
+            {
+                EmitFcvt(context, (op1) => EmitRoundMathCall(context, MidpointRounding.AwayFromZero, op1), signed: true, scalar: false);
+            }
         }
 
         public static void Fcvtau_Gp(ArmEmitterContext context)
         {
-            EmitFcvt_u_Gp(context, (op1) => EmitRoundMathCall(context, MidpointRounding.AwayFromZero, op1));
+            if (Optimizations.UseSse41)
+            {
+                EmitSse41Fcvtu_Gp(context, FPRoundingMode.ToNearestAway, isFixed: false);
+            }
+            else
+            {
+                EmitFcvt_u_Gp(context, (op1) => EmitRoundMathCall(context, MidpointRounding.AwayFromZero, op1));
+            }
         }
 
         public static void Fcvtau_S(ArmEmitterContext context)
         {
-            EmitFcvt(context, (op1) => EmitRoundMathCall(context, MidpointRounding.AwayFromZero, op1), signed: false, scalar: true);
+            if (Optimizations.UseSse41)
+            {
+                EmitSse41FcvtuOpF(context, FPRoundingMode.ToNearestAway, scalar: true);
+            }
+            else
+            {
+                EmitFcvt(context, (op1) => EmitRoundMathCall(context, MidpointRounding.AwayFromZero, op1), signed: false, scalar: true);
+            }
         }
 
         public static void Fcvtau_V(ArmEmitterContext context)
         {
-            EmitFcvt(context, (op1) => EmitRoundMathCall(context, MidpointRounding.AwayFromZero, op1), signed: false, scalar: false);
+            if (Optimizations.UseSse41)
+            {
+                EmitSse41FcvtuOpF(context, FPRoundingMode.ToNearestAway, scalar: false);
+            }
+            else
+            {
+                EmitFcvt(context, (op1) => EmitRoundMathCall(context, MidpointRounding.AwayFromZero, op1), signed: false, scalar: false);
+            }
         }
 
         public static void Fcvtl_V(ArmEmitterContext context)
@@ -224,7 +274,9 @@ namespace ARMeilleure.Instructions
                     {
                         Operand ne = EmitVectorExtractZx(context, op.Rn, part + index, 1);
 
+                        context.StoreToContext();
                         Operand e = context.Call(typeof(SoftFloat16_32).GetMethod(nameof(SoftFloat16_32.FPConvert)), ne);
+                        context.LoadFromContext();
 
                         res = context.VectorInsert(res, e, index);
                     }
@@ -333,7 +385,9 @@ namespace ARMeilleure.Instructions
 
                     if (sizeF == 0)
                     {
+                        context.StoreToContext();
                         Operand e = context.Call(typeof(SoftFloat32_16).GetMethod(nameof(SoftFloat32_16.FPConvert)), ne);
+                        context.LoadFromContext();
 
                         e = context.ZeroExtend16(OperandType.I64, e);
 
@@ -1211,7 +1265,14 @@ namespace ARMeilleure.Instructions
                     nRes = context.AddIntrinsic(Intrinsic.X86Mulps, nRes, fpScaledMask);
                 }
 
-                nRes = context.AddIntrinsic(Intrinsic.X86Roundps, nRes, Const(X86GetRoundControl(roundMode)));
+                if (roundMode != FPRoundingMode.ToNearestAway)
+                {
+                    nRes = context.AddIntrinsic(Intrinsic.X86Roundps, nRes, Const(X86GetRoundControl(roundMode)));
+                }
+                else
+                {
+                    nRes = EmitSse41RoundToNearestWithTiesToAwayOpF(context, nRes, scalar);
+                }
 
                 Operand nInt = context.AddIntrinsic(Intrinsic.X86Cvtps2dq, nRes);
 
@@ -1253,7 +1314,14 @@ namespace ARMeilleure.Instructions
                     nRes = context.AddIntrinsic(Intrinsic.X86Mulpd, nRes, fpScaledMask);
                 }
 
-                nRes = context.AddIntrinsic(Intrinsic.X86Roundpd, nRes, Const(X86GetRoundControl(roundMode)));
+                if (roundMode != FPRoundingMode.ToNearestAway)
+                {
+                    nRes = context.AddIntrinsic(Intrinsic.X86Roundpd, nRes, Const(X86GetRoundControl(roundMode)));
+                }
+                else
+                {
+                    nRes = EmitSse41RoundToNearestWithTiesToAwayOpF(context, nRes, scalar);
+                }
 
                 Operand nLong = EmitSse2CvtDoubleToInt64OpF(context, nRes, scalar);
 
@@ -1302,7 +1370,14 @@ namespace ARMeilleure.Instructions
                     nRes = context.AddIntrinsic(Intrinsic.X86Mulps, nRes, fpScaledMask);
                 }
 
-                nRes = context.AddIntrinsic(Intrinsic.X86Roundps, nRes, Const(X86GetRoundControl(roundMode)));
+                if (roundMode != FPRoundingMode.ToNearestAway)
+                {
+                    nRes = context.AddIntrinsic(Intrinsic.X86Roundps, nRes, Const(X86GetRoundControl(roundMode)));
+                }
+                else
+                {
+                    nRes = EmitSse41RoundToNearestWithTiesToAwayOpF(context, nRes, scalar);
+                }
 
                 Operand zero = context.VectorZero();
 
@@ -1357,7 +1432,14 @@ namespace ARMeilleure.Instructions
                     nRes = context.AddIntrinsic(Intrinsic.X86Mulpd, nRes, fpScaledMask);
                 }
 
-                nRes = context.AddIntrinsic(Intrinsic.X86Roundpd, nRes, Const(X86GetRoundControl(roundMode)));
+                if (roundMode != FPRoundingMode.ToNearestAway)
+                {
+                    nRes = context.AddIntrinsic(Intrinsic.X86Roundpd, nRes, Const(X86GetRoundControl(roundMode)));
+                }
+                else
+                {
+                    nRes = EmitSse41RoundToNearestWithTiesToAwayOpF(context, nRes, scalar);
+                }
 
                 Operand zero = context.VectorZero();
 
@@ -1412,7 +1494,14 @@ namespace ARMeilleure.Instructions
                     nRes = context.AddIntrinsic(Intrinsic.X86Mulss, nRes, fpScaledMask);
                 }
 
-                nRes = context.AddIntrinsic(Intrinsic.X86Roundss, nRes, Const(X86GetRoundControl(roundMode)));
+                if (roundMode != FPRoundingMode.ToNearestAway)
+                {
+                    nRes = context.AddIntrinsic(Intrinsic.X86Roundss, nRes, Const(X86GetRoundControl(roundMode)));
+                }
+                else
+                {
+                    nRes = EmitSse41RoundToNearestWithTiesToAwayOpF(context, nRes, scalar: true);
+                }
 
                 Operand nIntOrLong = op.RegisterSize == RegisterSize.Int32
                     ? context.AddIntrinsicInt (Intrinsic.X86Cvtss2si, nRes)
@@ -1452,7 +1541,14 @@ namespace ARMeilleure.Instructions
                     nRes = context.AddIntrinsic(Intrinsic.X86Mulsd, nRes, fpScaledMask);
                 }
 
-                nRes = context.AddIntrinsic(Intrinsic.X86Roundsd, nRes, Const(X86GetRoundControl(roundMode)));
+                if (roundMode != FPRoundingMode.ToNearestAway)
+                {
+                    nRes = context.AddIntrinsic(Intrinsic.X86Roundsd, nRes, Const(X86GetRoundControl(roundMode)));
+                }
+                else
+                {
+                    nRes = EmitSse41RoundToNearestWithTiesToAwayOpF(context, nRes, scalar: true);
+                }
 
                 Operand nIntOrLong = op.RegisterSize == RegisterSize.Int32
                     ? context.AddIntrinsicInt (Intrinsic.X86Cvtsd2si, nRes)
@@ -1500,7 +1596,14 @@ namespace ARMeilleure.Instructions
                     nRes = context.AddIntrinsic(Intrinsic.X86Mulss, nRes, fpScaledMask);
                 }
 
-                nRes = context.AddIntrinsic(Intrinsic.X86Roundss, nRes, Const(X86GetRoundControl(roundMode)));
+                if (roundMode != FPRoundingMode.ToNearestAway)
+                {
+                    nRes = context.AddIntrinsic(Intrinsic.X86Roundss, nRes, Const(X86GetRoundControl(roundMode)));
+                }
+                else
+                {
+                    nRes = EmitSse41RoundToNearestWithTiesToAwayOpF(context, nRes, scalar: true);
+                }
 
                 Operand zero = context.VectorZero();
 
@@ -1555,7 +1658,14 @@ namespace ARMeilleure.Instructions
                     nRes = context.AddIntrinsic(Intrinsic.X86Mulsd, nRes, fpScaledMask);
                 }
 
-                nRes = context.AddIntrinsic(Intrinsic.X86Roundsd, nRes, Const(X86GetRoundControl(roundMode)));
+                if (roundMode != FPRoundingMode.ToNearestAway)
+                {
+                    nRes = context.AddIntrinsic(Intrinsic.X86Roundsd, nRes, Const(X86GetRoundControl(roundMode)));
+                }
+                else
+                {
+                    nRes = EmitSse41RoundToNearestWithTiesToAwayOpF(context, nRes, scalar: true);
+                }
 
                 Operand zero = context.VectorZero();
 

ARMeilleure/Instructions/InstEmitSimdCvt32.cs

@@ -161,34 +161,15 @@ namespace ARMeilleure.Instructions
                 {
                     Operand toConvert = ExtractScalar(context, floatSize, op.Vm);
 
-                    Operand asInteger;
-
                     // TODO: Fast Path.
                     if (roundWithFpscr)
                     {
-                        MethodInfo info;
-
-                        if (floatSize == OperandType.FP64)
-                        {
-                            info = unsigned
-                                ? typeof(SoftFallback).GetMethod(nameof(SoftFallback.DoubleToUInt32))
-                                : typeof(SoftFallback).GetMethod(nameof(SoftFallback.DoubleToInt32));
-                        }
-                        else
-                        {
-                            info = unsigned
-                                ? typeof(SoftFallback).GetMethod(nameof(SoftFallback.FloatToUInt32))
-                                : typeof(SoftFallback).GetMethod(nameof(SoftFallback.FloatToInt32));
-                        }
-
-                        asInteger = context.Call(info, toConvert);
-                    }
-                    else
-                    {
-                        // Round towards zero.
-                        asInteger = EmitSaturateFloatToInt(context, toConvert, unsigned);
+                        toConvert = EmitRoundByRMode(context, toConvert);
                     }
 
+                    // Round towards zero.
+                    Operand asInteger = EmitSaturateFloatToInt(context, toConvert, unsigned);
+
                     InsertScalar(context, op.Vd, asInteger);
                 }
             }
@@ -222,6 +203,9 @@ namespace ARMeilleure.Instructions
             FPRoundingMode roundMode;
             switch (rm)
             {
+                case 0b00:
+                    roundMode = FPRoundingMode.ToNearestAway;
+                    break;
                 case 0b01:
                     roundMode = FPRoundingMode.ToNearest;
                     break;
@@ -247,7 +231,7 @@ namespace ARMeilleure.Instructions
             bool unsigned = op.Opc == 0;
             int rm = op.Opc2 & 3;
 
-            if (Optimizations.UseSse41 && rm != 0b00)
+            if (Optimizations.UseSse41)
             {
                 EmitSse41ConvertInt32(context, RMToRoundMode(rm), !unsigned);
             }
@@ -271,14 +255,80 @@ namespace ARMeilleure.Instructions
                         break;
                 }
 
-                Operand asInteger;
-
-                asInteger = EmitSaturateFloatToInt(context, toConvert, unsigned);
+                Operand asInteger = EmitSaturateFloatToInt(context, toConvert, unsigned);
 
                 InsertScalar(context, op.Vd, asInteger);
             }
         }
 
+        public static void Vcvt_TB(ArmEmitterContext context)
+        {
+            OpCode32SimdCvtTB op = (OpCode32SimdCvtTB)context.CurrOp;
+
+            if (Optimizations.UseF16c)
+            {
+                Debug.Assert(!Optimizations.ForceLegacySse);
+
+                if (op.Op)
+                {
+                    Operand res = ExtractScalar(context, op.Size == 1 ? OperandType.FP64 : OperandType.FP32, op.Vm);
+                    if (op.Size == 1)
+                    {
+                        res = context.AddIntrinsic(Intrinsic.X86Cvtsd2ss, context.VectorZero(), res);
+                    }
+                    res = context.AddIntrinsic(Intrinsic.X86Vcvtps2ph, res, Const(X86GetRoundControl(FPRoundingMode.ToNearest)));
+                    res = context.VectorExtract16(res, 0);
+                    InsertScalar16(context, op.Vd, op.T, res);
+                }
+                else
+                {
+                    Operand res = context.VectorCreateScalar(ExtractScalar16(context, op.Vm, op.T));
+                    res = context.AddIntrinsic(Intrinsic.X86Vcvtph2ps, res);
+                    if (op.Size == 1)
+                    {
+                        res = context.AddIntrinsic(Intrinsic.X86Cvtss2sd, context.VectorZero(), res);
+                    }
+                    res = context.VectorExtract(op.Size == 1 ? OperandType.I64 : OperandType.I32, res, 0);
+                    InsertScalar(context, op.Vd, res);
+                }
+            }
+            else
+            {
+                if (op.Op)
+                {
+                    // Convert to half.
+
+                    Operand src = ExtractScalar(context, op.Size == 1 ? OperandType.FP64 : OperandType.FP32, op.Vm);
+
+                    MethodInfo method = op.Size == 1
+                        ? typeof(SoftFloat64_16).GetMethod(nameof(SoftFloat64_16.FPConvert))
+                        : typeof(SoftFloat32_16).GetMethod(nameof(SoftFloat32_16.FPConvert));
+
+                    context.StoreToContext();
+                    Operand res = context.Call(method, src);
+                    context.LoadFromContext();
+
+                    InsertScalar16(context, op.Vd, op.T, res);
+                }
+                else
+                {
+                    // Convert from half.
+
+                    Operand src = ExtractScalar16(context, op.Vm, op.T);
+
+                    MethodInfo method = op.Size == 1
+                        ? typeof(SoftFloat16_64).GetMethod(nameof(SoftFloat16_64.FPConvert))
+                        : typeof(SoftFloat16_32).GetMethod(nameof(SoftFloat16_32.FPConvert));
+
+                    context.StoreToContext();
+                    Operand res = context.Call(method, src);
+                    context.LoadFromContext();
+
+                    InsertScalar(context, op.Vd, res);
+                }
+            }
+        }
+
         // VRINTA/M/N/P (floating-point).
         public static void Vrint_RM(ArmEmitterContext context)
         {
@@ -288,15 +338,21 @@ namespace ARMeilleure.Instructions
 
             int rm = op.Opc2 & 3;
 
-            if (Optimizations.UseSse2 && rm != 0b00)
+            if (Optimizations.UseSse41)
             {
                 EmitScalarUnaryOpSimd32(context, (m) =>
                 {
-                    Intrinsic inst = (op.Size & 1) == 0 ? Intrinsic.X86Roundss : Intrinsic.X86Roundsd;
-
                     FPRoundingMode roundMode = RMToRoundMode(rm);
 
-                    return context.AddIntrinsic(inst, m, Const(X86GetRoundControl(roundMode)));
+                    if (roundMode != FPRoundingMode.ToNearestAway)
+                    {
+                        Intrinsic inst = (op.Size & 1) == 0 ? Intrinsic.X86Roundss : Intrinsic.X86Roundsd;
+                        return context.AddIntrinsic(inst, m, Const(X86GetRoundControl(roundMode)));
+                    }
+                    else
+                    {
+                        return EmitSse41RoundToNearestWithTiesToAwayOpF(context, m, scalar: true);
+                    }
                 });
             }
             else
@@ -326,7 +382,17 @@ namespace ARMeilleure.Instructions
         // VRINTA (vector).
         public static void Vrinta_V(ArmEmitterContext context)
         {
-            EmitVectorUnaryOpF32(context, (m) => EmitRoundMathCall(context, MidpointRounding.AwayFromZero, m));
+            if (Optimizations.UseSse41)
+            {
+                EmitVectorUnaryOpSimd32(context, (m) =>
+                {
+                    return EmitSse41RoundToNearestWithTiesToAwayOpF(context, m, scalar: false);
+                });
+            }
+            else
+            {
+                EmitVectorUnaryOpF32(context, (m) => EmitRoundMathCall(context, MidpointRounding.AwayFromZero, m));
+            }
         }
 
         // VRINTM (vector).
@@ -399,15 +465,9 @@ namespace ARMeilleure.Instructions
         // VRINTX (floating-point).
         public static void Vrintx_S(ArmEmitterContext context)
         {
-            OpCode32SimdS op = (OpCode32SimdS)context.CurrOp;
-
-            bool doubleSize = (op.Size & 1) == 1;
-            string methodName = doubleSize ? nameof(SoftFallback.Round) : nameof(SoftFallback.RoundF);
-
             EmitScalarUnaryOpF32(context, (op1) =>
             {
-                MethodInfo info = typeof(SoftFallback).GetMethod(methodName);
-                return context.Call(info, op1);
+                return EmitRoundByRMode(context, op1);
             });
         }
 
@@ -440,7 +500,14 @@ namespace ARMeilleure.Instructions
                 Operand nRes = context.AddIntrinsic(Intrinsic.X86Cmpss, n, n, Const((int)CmpCondition.OrderedQ));
                 nRes = context.AddIntrinsic(Intrinsic.X86Pand, nRes, n);
 
-                nRes = context.AddIntrinsic(Intrinsic.X86Roundss, nRes, Const(X86GetRoundControl(roundMode)));
+                if (roundMode != FPRoundingMode.ToNearestAway)
+                {
+                    nRes = context.AddIntrinsic(Intrinsic.X86Roundss, nRes, Const(X86GetRoundControl(roundMode)));
+                }
+                else
+                {
+                    nRes = EmitSse41RoundToNearestWithTiesToAwayOpF(context, nRes, scalar: true);
+                }
 
                 Operand zero = context.VectorZero();
 
@@ -491,7 +558,14 @@ namespace ARMeilleure.Instructions
                 Operand nRes = context.AddIntrinsic(Intrinsic.X86Cmpsd, n, n, Const((int)CmpCondition.OrderedQ));
                 nRes = context.AddIntrinsic(Intrinsic.X86Pand, nRes, n);
 
-                nRes = context.AddIntrinsic(Intrinsic.X86Roundsd, nRes, Const(X86GetRoundControl(roundMode)));
+                if (roundMode != FPRoundingMode.ToNearestAway)
+                {
+                    nRes = context.AddIntrinsic(Intrinsic.X86Roundsd, nRes, Const(X86GetRoundControl(roundMode)));
+                }
+                else
+                {
+                    nRes = EmitSse41RoundToNearestWithTiesToAwayOpF(context, nRes, scalar: true);
+                }
 
                 Operand zero = context.VectorZero();
 

ARMeilleure/Instructions/InstEmitSimdHelper.cs

@@ -33,6 +33,14 @@ namespace ARMeilleure.Instructions
         };
 
         public static readonly long ZeroMask = 128L << 56 | 128L << 48 | 128L << 40 | 128L << 32 | 128L << 24 | 128L << 16 | 128L << 8 | 128L << 0;
+
+        public static ulong X86GetGf2p8LogicalShiftLeft(int shift)
+        {
+            ulong identity = (0b00000001UL << 56) | (0b00000010UL << 48) | (0b00000100UL << 40) | (0b00001000UL << 32) |
+                             (0b00010000UL << 24) | (0b00100000UL << 16) | (0b01000000UL <<  8) | (0b10000000UL <<  0);
+
+            return shift >= 0 ? identity >> (shift * 8) : identity << (-shift * 8);
+        }
 #endregion
 
 #region "X86 SSE Intrinsics"
@@ -243,6 +251,46 @@ namespace ARMeilleure.Instructions
             throw new ArgumentException($"Invalid rounding mode \"{roundMode}\".");
         }
 
+        public static Operand EmitSse41RoundToNearestWithTiesToAwayOpF(ArmEmitterContext context, Operand n, bool scalar)
+        {
+            Debug.Assert(n.Type == OperandType.V128);
+
+            Operand nCopy = context.Copy(n);
+
+            Operand rC = Const(X86GetRoundControl(FPRoundingMode.TowardsZero));
+
+            IOpCodeSimd op = (IOpCodeSimd)context.CurrOp;
+
+            if ((op.Size & 1) == 0)
+            {
+                Operand signMask = scalar ? X86GetScalar(context, int.MinValue) : X86GetAllElements(context, int.MinValue);
+                        signMask = context.AddIntrinsic(Intrinsic.X86Pand, signMask, nCopy);
+
+                // 0x3EFFFFFF == BitConverter.SingleToInt32Bits(0.5f) - 1
+                Operand valueMask = scalar ? X86GetScalar(context, 0x3EFFFFFF) : X86GetAllElements(context, 0x3EFFFFFF);
+                        valueMask = context.AddIntrinsic(Intrinsic.X86Por, valueMask, signMask);
+
+                nCopy = context.AddIntrinsic(scalar ? Intrinsic.X86Addss : Intrinsic.X86Addps, nCopy, valueMask);
+
+                nCopy = context.AddIntrinsic(scalar ? Intrinsic.X86Roundss : Intrinsic.X86Roundps, nCopy, rC);
+            }
+            else
+            {
+                Operand signMask = scalar ? X86GetScalar(context, long.MinValue) : X86GetAllElements(context, long.MinValue);
+                        signMask = context.AddIntrinsic(Intrinsic.X86Pand, signMask, nCopy);
+
+                // 0x3FDFFFFFFFFFFFFFL == BitConverter.DoubleToInt64Bits(0.5d) - 1L
+                Operand valueMask = scalar ? X86GetScalar(context, 0x3FDFFFFFFFFFFFFFL) : X86GetAllElements(context, 0x3FDFFFFFFFFFFFFFL);
+                        valueMask = context.AddIntrinsic(Intrinsic.X86Por, valueMask, signMask);
+
+                nCopy = context.AddIntrinsic(scalar ? Intrinsic.X86Addsd : Intrinsic.X86Addpd, nCopy, valueMask);
+
+                nCopy = context.AddIntrinsic(scalar ? Intrinsic.X86Roundsd : Intrinsic.X86Roundpd, nCopy, rC);
+            }
+
+            return nCopy;
+        }
+
         public static Operand EmitCountSetBits8(ArmEmitterContext context, Operand op) // "size" is 8 (SIMD&FP Inst.).
         {
             Debug.Assert(op.Type == OperandType.I32 || op.Type == OperandType.I64);
@@ -361,6 +409,54 @@ namespace ARMeilleure.Instructions
             return context.Call(info, n, Const((int)roundMode));
         }
 
+        public static Operand EmitGetRoundingMode(ArmEmitterContext context)
+        {
+            Operand rMode = context.ShiftLeft(GetFpFlag(FPState.RMode1Flag), Const(1));
+                    rMode = context.BitwiseOr(rMode, GetFpFlag(FPState.RMode0Flag));
+
+            return rMode;
+        }
+
+        public static Operand EmitRoundByRMode(ArmEmitterContext context, Operand op)
+        {
+            Debug.Assert(op.Type == OperandType.FP32 || op.Type == OperandType.FP64);
+
+            Operand lbl1 = Label();
+            Operand lbl2 = Label();
+            Operand lbl3 = Label();
+            Operand lblEnd = Label();
+
+            Operand rN = Const((int)FPRoundingMode.ToNearest);
+            Operand rP = Const((int)FPRoundingMode.TowardsPlusInfinity);
+            Operand rM = Const((int)FPRoundingMode.TowardsMinusInfinity);
+
+            Operand res = context.AllocateLocal(op.Type);
+
+            Operand rMode = EmitGetRoundingMode(context);
+
+            context.BranchIf(lbl1, rMode, rN, Comparison.NotEqual);
+            context.Copy(res, EmitRoundMathCall(context, MidpointRounding.ToEven, op));
+            context.Branch(lblEnd);
+
+            context.MarkLabel(lbl1);
+            context.BranchIf(lbl2, rMode, rP, Comparison.NotEqual);
+            context.Copy(res, EmitUnaryMathCall(context, nameof(Math.Ceiling), op));
+            context.Branch(lblEnd);
+
+            context.MarkLabel(lbl2);
+            context.BranchIf(lbl3, rMode, rM, Comparison.NotEqual);
+            context.Copy(res, EmitUnaryMathCall(context, nameof(Math.Floor), op));
+            context.Branch(lblEnd);
+
+            context.MarkLabel(lbl3);
+            context.Copy(res, EmitUnaryMathCall(context, nameof(Math.Truncate), op));
+            context.Branch(lblEnd);
+
+            context.MarkLabel(lblEnd);
+
+            return res;
+        }
+
         public static Operand EmitSoftFloatCall(ArmEmitterContext context, string name, params Operand[] callArgs)
         {
             IOpCodeSimd op = (IOpCodeSimd)context.CurrOp;
@@ -369,7 +465,11 @@ namespace ARMeilleure.Instructions
                 ? typeof(SoftFloat32).GetMethod(name)
                 : typeof(SoftFloat64).GetMethod(name);
 
-            return context.Call(info, callArgs);
+            context.StoreToContext();
+            Operand res = context.Call(info, callArgs);
+            context.LoadFromContext();
+
+            return res;
         }
 
         public static void EmitScalarBinaryOpByElemF(ArmEmitterContext context, Func2I emit)
@@ -1269,7 +1369,7 @@ namespace ARMeilleure.Instructions
 
         public static void EmitSseOrAvxEnterFtzAndDazModesOpF(ArmEmitterContext context, out Operand isTrue)
         {
-            isTrue = context.Call(typeof(NativeInterface).GetMethod(nameof(NativeInterface.GetFpcrFz)));
+            isTrue = GetFpFlag(FPState.FzFlag);
 
             Operand lblTrue = Label();
             context.BranchIfFalse(lblTrue, isTrue);
@@ -1281,9 +1381,7 @@ namespace ARMeilleure.Instructions
 
         public static void EmitSseOrAvxExitFtzAndDazModesOpF(ArmEmitterContext context, Operand isTrue = default)
         {
-            isTrue = isTrue == default
-                ? context.Call(typeof(NativeInterface).GetMethod(nameof(NativeInterface.GetFpcrFz)))
-                : isTrue;
+            isTrue = isTrue == default ? GetFpFlag(FPState.FzFlag) : isTrue;
 
             Operand lblTrue = Label();
             context.BranchIfFalse(lblTrue, isTrue);
@@ -1552,13 +1650,13 @@ namespace ARMeilleure.Instructions
 
             context.BranchIf(lbl1, op, zeroL, Comparison.LessOrEqual);
             context.Copy(res, maxT);
-            context.Call(typeof(NativeInterface).GetMethod(nameof(NativeInterface.SetFpsrQc)));
+            SetFpFlag(context, FPState.QcFlag, Const(1));
             context.Branch(lblEnd);
 
             context.MarkLabel(lbl1);
             context.BranchIf(lblEnd, op, zeroL, Comparison.GreaterOrEqual);
             context.Copy(res, minT);
-            context.Call(typeof(NativeInterface).GetMethod(nameof(NativeInterface.SetFpsrQc)));
+            SetFpFlag(context, FPState.QcFlag, Const(1));
             context.Branch(lblEnd);
 
             context.MarkLabel(lblEnd);
@@ -1583,7 +1681,7 @@ namespace ARMeilleure.Instructions
 
             context.BranchIf(lblEnd, op, zeroUL, Comparison.LessOrEqualUI);
             context.Copy(res, maxT);
-            context.Call(typeof(NativeInterface).GetMethod(nameof(NativeInterface.SetFpsrQc)));
+            SetFpFlag(context, FPState.QcFlag, Const(1));
             context.Branch(lblEnd);
 
             context.MarkLabel(lblEnd);
@@ -1610,13 +1708,13 @@ namespace ARMeilleure.Instructions
 
             context.BranchIf(lbl1, op, maxT, Comparison.LessOrEqual);
             context.Copy(res, maxT);
-            context.Call(typeof(NativeInterface).GetMethod(nameof(NativeInterface.SetFpsrQc)));
+            SetFpFlag(context, FPState.QcFlag, Const(1));
             context.Branch(lblEnd);
 
             context.MarkLabel(lbl1);
             context.BranchIf(lblEnd, op, minT, Comparison.GreaterOrEqual);
             context.Copy(res, minT);
-            context.Call(typeof(NativeInterface).GetMethod(nameof(NativeInterface.SetFpsrQc)));
+            SetFpFlag(context, FPState.QcFlag, Const(1));
             context.Branch(lblEnd);
 
             context.MarkLabel(lblEnd);
@@ -1641,7 +1739,7 @@ namespace ARMeilleure.Instructions
 
             context.BranchIf(lblEnd, op, maxT, Comparison.LessOrEqualUI);
             context.Copy(res, maxT);
-            context.Call(typeof(NativeInterface).GetMethod(nameof(NativeInterface.SetFpsrQc)));
+            SetFpFlag(context, FPState.QcFlag, Const(1));
             context.Branch(lblEnd);
 
             context.MarkLabel(lblEnd);
@@ -1663,7 +1761,7 @@ namespace ARMeilleure.Instructions
 
             context.BranchIf(lblEnd, op, minL, Comparison.NotEqual);
             context.Copy(res, maxL);
-            context.Call(typeof(NativeInterface).GetMethod(nameof(NativeInterface.SetFpsrQc)));
+            SetFpFlag(context, FPState.QcFlag, Const(1));
             context.Branch(lblEnd);
 
             context.MarkLabel(lblEnd);
@@ -1691,7 +1789,7 @@ namespace ARMeilleure.Instructions
 
             Operand isPositive = context.ICompareGreaterOrEqual(op1, zeroL);
             context.Copy(res, context.ConditionalSelect(isPositive, maxL, minL));
-            context.Call(typeof(NativeInterface).GetMethod(nameof(NativeInterface.SetFpsrQc)));
+            SetFpFlag(context, FPState.QcFlag, Const(1));
             context.Branch(lblEnd);
 
             context.MarkLabel(lblEnd);
@@ -1713,7 +1811,7 @@ namespace ARMeilleure.Instructions
 
             context.BranchIf(lblEnd, add, op1, Comparison.GreaterOrEqualUI);
             context.Copy(res, maxUL);
-            context.Call(typeof(NativeInterface).GetMethod(nameof(NativeInterface.SetFpsrQc)));
+            SetFpFlag(context, FPState.QcFlag, Const(1));
             context.Branch(lblEnd);
 
             context.MarkLabel(lblEnd);
@@ -1741,7 +1839,7 @@ namespace ARMeilleure.Instructions
 
             Operand isPositive = context.ICompareGreaterOrEqual(op1, zeroL);
             context.Copy(res, context.ConditionalSelect(isPositive, maxL, minL));
-            context.Call(typeof(NativeInterface).GetMethod(nameof(NativeInterface.SetFpsrQc)));
+            SetFpFlag(context, FPState.QcFlag, Const(1));
             context.Branch(lblEnd);
 
             context.MarkLabel(lblEnd);
@@ -1763,7 +1861,7 @@ namespace ARMeilleure.Instructions
 
             context.BranchIf(lblEnd, op1, op2, Comparison.GreaterOrEqualUI);
             context.Copy(res, zeroL);
-            context.Call(typeof(NativeInterface).GetMethod(nameof(NativeInterface.SetFpsrQc)));
+            SetFpFlag(context, FPState.QcFlag, Const(1));
             context.Branch(lblEnd);
 
             context.MarkLabel(lblEnd);
@@ -1790,19 +1888,19 @@ namespace ARMeilleure.Instructions
             Operand notOp2AndRes = context.BitwiseAnd(context.BitwiseNot(op2), add);
             context.BranchIf(lblEnd, notOp2AndRes, zeroL, Comparison.GreaterOrEqual);
             context.Copy(res, maxL);
-            context.Call(typeof(NativeInterface).GetMethod(nameof(NativeInterface.SetFpsrQc)));
+            SetFpFlag(context, FPState.QcFlag, Const(1));
             context.Branch(lblEnd);
 
             context.MarkLabel(lbl1);
             context.BranchIf(lbl2, op2, zeroL, Comparison.Less);
             context.Copy(res, maxL);
-            context.Call(typeof(NativeInterface).GetMethod(nameof(NativeInterface.SetFpsrQc)));
+            SetFpFlag(context, FPState.QcFlag, Const(1));
             context.Branch(lblEnd);
 
             context.MarkLabel(lbl2);
             context.BranchIf(lblEnd, add, maxL, Comparison.LessOrEqualUI);
             context.Copy(res, maxL);
-            context.Call(typeof(NativeInterface).GetMethod(nameof(NativeInterface.SetFpsrQc)));
+            SetFpFlag(context, FPState.QcFlag, Const(1));
             context.Branch(lblEnd);
 
             context.MarkLabel(lblEnd);
@@ -1828,14 +1926,14 @@ namespace ARMeilleure.Instructions
             context.BranchIf(lbl1, op1, zeroL, Comparison.Less);
             context.BranchIf(lblEnd, add, op1, Comparison.GreaterOrEqualUI);
             context.Copy(res, maxUL);
-            context.Call(typeof(NativeInterface).GetMethod(nameof(NativeInterface.SetFpsrQc)));
+            SetFpFlag(context, FPState.QcFlag, Const(1));
             context.Branch(lblEnd);
 
             context.MarkLabel(lbl1);
             context.BranchIf(lblEnd, op2, maxL, Comparison.GreaterUI);
             context.BranchIf(lblEnd, add, zeroL, Comparison.GreaterOrEqual);
             context.Copy(res, zeroL);
-            context.Call(typeof(NativeInterface).GetMethod(nameof(NativeInterface.SetFpsrQc)));
+            SetFpFlag(context, FPState.QcFlag, Const(1));
             context.Branch(lblEnd);
 
             context.MarkLabel(lblEnd);

ARMeilleure/Instructions/InstEmitSimdHelper32.cs

@@ -70,6 +70,22 @@ namespace ARMeilleure.Instructions
             context.Copy(vec, insert);
         }
 
+        public static Operand ExtractScalar16(ArmEmitterContext context, int reg, bool top)
+        {
+            return context.VectorExtract16(GetVecA32(reg >> 2), ((reg & 3) << 1) | (top ? 1 : 0));
+        }
+
+        public static void InsertScalar16(ArmEmitterContext context, int reg, bool top, Operand value)
+        {
+            Debug.Assert(value.Type == OperandType.FP32 || value.Type == OperandType.I32);
+
+            Operand vec, insert;
+            vec = GetVecA32(reg >> 2);
+            insert = context.VectorInsert16(vec, value, ((reg & 3) << 1) | (top ? 1 : 0));
+
+            context.Copy(vec, insert);
+        }
+
         public static Operand ExtractElement(ArmEmitterContext context, int reg, int size, bool signed)
         {
             return EmitVectorExtract32(context, reg >> (4 - size), reg & ((16 >> size) - 1), size, signed);
@@ -1181,7 +1197,11 @@ namespace ARMeilleure.Instructions
             Array.Resize(ref callArgs, callArgs.Length + 1);
             callArgs[callArgs.Length - 1] = Const(1);
 
-            return context.Call(info, callArgs);
+            context.StoreToContext();
+            Operand res = context.Call(info, callArgs);
+            context.LoadFromContext();
+
+            return res;
         }
 
         public static Operand EmitVectorExtractSx32(ArmEmitterContext context, int reg, int index, int size)

ARMeilleure/Instructions/InstEmitSimdLogical.cs

@@ -336,20 +336,45 @@ namespace ARMeilleure.Instructions
         {
             OpCodeSimd op = (OpCodeSimd)context.CurrOp;
 
-            Operand res = context.VectorZero();
-
-            int elems = op.RegisterSize == RegisterSize.Simd128 ? 16 : 8;
-
-            for (int index = 0; index < elems; index++)
+            if (Optimizations.UseGfni)
             {
-                Operand ne = EmitVectorExtractZx(context, op.Rn, index, 0);
+                const long bitMatrix =
+                    (0b10000000L << 56) |
+                    (0b01000000L << 48) |
+                    (0b00100000L << 40) |
+                    (0b00010000L << 32) |
+                    (0b00001000L << 24) |
+                    (0b00000100L << 16) |
+                    (0b00000010L <<  8) |
+                    (0b00000001L <<  0);
 
-                Operand de = EmitReverseBits8Op(context, ne);
+                Operand vBitMatrix = X86GetAllElements(context, bitMatrix);
 
-                res = EmitVectorInsert(context, res, de, index, 0);
+                Operand res = context.AddIntrinsic(Intrinsic.X86Gf2p8affineqb, GetVec(op.Rn), vBitMatrix, Const(0));
+
+                if (op.RegisterSize == RegisterSize.Simd64)
+                {
+                    res = context.VectorZeroUpper64(res);
+                }
+
+                context.Copy(GetVec(op.Rd), res);
             }
+            else
+            {
+                Operand res = context.VectorZero();
+                int elems = op.RegisterSize == RegisterSize.Simd128 ? 16 : 8;
 
-            context.Copy(GetVec(op.Rd), res);
+                for (int index = 0; index < elems; index++)
+                {
+                    Operand ne = EmitVectorExtractZx(context, op.Rn, index, 0);
+
+                    Operand de = EmitReverseBits8Op(context, ne);
+
+                    res = EmitVectorInsert(context, res, de, index, 0);
+                }
+
+                context.Copy(GetVec(op.Rd), res);
+            }
         }
 
         private static Operand EmitReverseBits8Op(ArmEmitterContext context, Operand op)

ARMeilleure/Instructions/InstEmitSimdShift.cs

@@ -88,8 +88,35 @@ namespace ARMeilleure.Instructions
             OpCodeSimdShImm op = (OpCodeSimdShImm)context.CurrOp;
 
             int shift = GetImmShl(op);
+            int eSize = 8 << op.Size;
 
-            if (Optimizations.UseSse2 && op.Size > 0)
+            if (shift >= eSize)
+            {
+                if ((op.RegisterSize == RegisterSize.Simd64))
+                {
+                    Operand res = context.VectorZeroUpper64(GetVec(op.Rd));
+
+                    context.Copy(GetVec(op.Rd), res);
+                }
+            }
+            else if (Optimizations.UseGfni && op.Size == 0)
+            {
+                Operand n = GetVec(op.Rn);
+
+                ulong bitMatrix = X86GetGf2p8LogicalShiftLeft(shift);
+
+                Operand vBitMatrix = X86GetElements(context, bitMatrix, bitMatrix);
+
+                Operand res = context.AddIntrinsic(Intrinsic.X86Gf2p8affineqb, n, vBitMatrix, Const(0));
+
+                if (op.RegisterSize == RegisterSize.Simd64)
+                {
+                    res = context.VectorZeroUpper64(res);
+                }
+
+                context.Copy(GetVec(op.Rd), res);
+            }
+            else if (Optimizations.UseSse2 && op.Size > 0)
             {
                 Operand n = GetVec(op.Rn);
 
@@ -396,10 +423,40 @@ namespace ARMeilleure.Instructions
         {
             OpCodeSimdShImm op = (OpCodeSimdShImm)context.CurrOp;
 
-            if (Optimizations.UseSse2 && op.Size > 0 && op.Size < 3)
-            {
-                int shift = GetImmShr(op);
+            int shift = GetImmShr(op);
 
+            if (Optimizations.UseGfni && op.Size == 0)
+            {
+                Operand n = GetVec(op.Rn);
+
+                ulong bitMatrix;
+
+                if (shift < 8)
+                {
+                    bitMatrix = X86GetGf2p8LogicalShiftLeft(-shift);
+
+                    // Extend sign-bit
+                    bitMatrix |= 0x8080808080808080UL >> (64 - shift * 8);
+                }
+                else
+                {
+                    // Replicate sign-bit into all bits
+                    bitMatrix = 0x8080808080808080UL;
+                }
+
+                Operand vBitMatrix = X86GetElements(context, bitMatrix, bitMatrix);
+
+                Operand res = context.AddIntrinsic(Intrinsic.X86Gf2p8affineqb, n, vBitMatrix, Const(0));
+
+                if (op.RegisterSize == RegisterSize.Simd64)
+                {
+                    res = context.VectorZeroUpper64(res);
+                }
+
+                context.Copy(GetVec(op.Rd), res);
+            }
+            else if (Optimizations.UseSse2 && op.Size > 0 && op.Size < 3)
+            {
                 Operand n = GetVec(op.Rn);
 
                 Intrinsic sraInst = X86PsraInstruction[op.Size];
@@ -929,10 +986,44 @@ namespace ARMeilleure.Instructions
             OpCodeSimdShImm op = (OpCodeSimdShImm)context.CurrOp;
 
             int shift = GetImmShl(op);
+            int eSize = 8 << op.Size;
 
             ulong mask = shift != 0 ? ulong.MaxValue >> (64 - shift) : 0UL;
 
-            if (Optimizations.UseSse2 && op.Size > 0)
+            if (shift >= eSize)
+            {
+                if ((op.RegisterSize == RegisterSize.Simd64) || scalar)
+                {
+                    Operand res = context.VectorZeroUpper64(GetVec(op.Rd));
+
+                    context.Copy(GetVec(op.Rd), res);
+                }
+            }
+            else if (Optimizations.UseGfni && op.Size == 0)
+            {
+                Operand d = GetVec(op.Rd);
+                Operand n = GetVec(op.Rn);
+
+                ulong bitMatrix = X86GetGf2p8LogicalShiftLeft(shift);
+
+                Operand vBitMatrix = X86GetElements(context, bitMatrix, bitMatrix);
+
+                Operand nShifted = context.AddIntrinsic(Intrinsic.X86Gf2p8affineqb, n, vBitMatrix, Const(0));
+
+                Operand dMask = X86GetAllElements(context, (long)mask * _masks_SliSri[op.Size]);
+
+                Operand dMasked = context.AddIntrinsic(Intrinsic.X86Pand, d, dMask);
+
+                Operand res = context.AddIntrinsic(Intrinsic.X86Por, nShifted, dMasked);
+
+                if ((op.RegisterSize == RegisterSize.Simd64) || scalar)
+                {
+                    res = context.VectorZeroUpper64(res);
+                }
+
+                context.Copy(d, res);
+            }
+            else if (Optimizations.UseSse2 && op.Size > 0)
             {
                 Operand d = GetVec(op.Rd);
                 Operand n = GetVec(op.Rn);
@@ -988,7 +1079,40 @@ namespace ARMeilleure.Instructions
 
             ulong mask = (ulong.MaxValue << (eSize - shift)) & (ulong.MaxValue >> (64 - eSize));
 
-            if (Optimizations.UseSse2 && op.Size > 0)
+            if (shift >= eSize)
+            {
+                if ((op.RegisterSize == RegisterSize.Simd64) || scalar)
+                {
+                    Operand res = context.VectorZeroUpper64(GetVec(op.Rd));
+
+                    context.Copy(GetVec(op.Rd), res);
+                }
+            }
+            else if (Optimizations.UseGfni && op.Size == 0)
+            {
+                Operand d = GetVec(op.Rd);
+                Operand n = GetVec(op.Rn);
+
+                ulong bitMatrix = X86GetGf2p8LogicalShiftLeft(-shift);
+
+                Operand vBitMatrix = X86GetElements(context, bitMatrix, bitMatrix);
+
+                Operand nShifted = context.AddIntrinsic(Intrinsic.X86Gf2p8affineqb, n, vBitMatrix, Const(0));
+
+                Operand dMask = X86GetAllElements(context, (long)mask * _masks_SliSri[op.Size]);
+
+                Operand dMasked = context.AddIntrinsic(Intrinsic.X86Pand, d, dMask);
+
+                Operand res = context.AddIntrinsic(Intrinsic.X86Por, nShifted, dMasked);
+
+                if ((op.RegisterSize == RegisterSize.Simd64) || scalar)
+                {
+                    res = context.VectorZeroUpper64(res);
+                }
+
+                context.Copy(d, res);
+            }
+            else if (Optimizations.UseSse2 && op.Size > 0)
             {
                 Operand d = GetVec(op.Rd);
                 Operand n = GetVec(op.Rn);

ARMeilleure/Instructions/InstEmitSimdShift32.cs

@@ -1,5 +1,6 @@
 using ARMeilleure.Decoders;
 using ARMeilleure.IntermediateRepresentation;
+using ARMeilleure.State;
 using ARMeilleure.Translation;
 using System;
 using System.Diagnostics;
@@ -378,7 +379,7 @@ namespace ARMeilleure.Instructions
 
             context.BranchIfFalse(lblNoSat, context.BitwiseOr(gt, lt));
 
-            context.Call(typeof(NativeInterface).GetMethod(nameof(NativeInterface.SetFpsrQc)));
+            SetFpFlag(context, FPState.QcFlag, Const(1));
 
             context.MarkLabel(lblNoSat);
 

ARMeilleure/Instructions/InstEmitSystem.cs

@@ -31,8 +31,8 @@ namespace ARMeilleure.Instructions
                 case 0b11_011_0000_0000_001: info = typeof(NativeInterface).GetMethod(nameof(NativeInterface.GetCtrEl0)); break;
                 case 0b11_011_0000_0000_111: info = typeof(NativeInterface).GetMethod(nameof(NativeInterface.GetDczidEl0)); break;
                 case 0b11_011_0100_0010_000: EmitGetNzcv(context); return;
-                case 0b11_011_0100_0100_000: info = typeof(NativeInterface).GetMethod(nameof(NativeInterface.GetFpcr)); break;
-                case 0b11_011_0100_0100_001: info = typeof(NativeInterface).GetMethod(nameof(NativeInterface.GetFpsr)); break;
+                case 0b11_011_0100_0100_000: EmitGetFpcr(context); return;
+                case 0b11_011_0100_0100_001: EmitGetFpsr(context); return;
                 case 0b11_011_1101_0000_010: info = typeof(NativeInterface).GetMethod(nameof(NativeInterface.GetTpidrEl0)); break;
                 case 0b11_011_1101_0000_011: info = typeof(NativeInterface).GetMethod(nameof(NativeInterface.GetTpidrroEl0)); break;
                 case 0b11_011_1110_0000_000: info = typeof(NativeInterface).GetMethod(nameof(NativeInterface.GetCntfrqEl0)); break;
@@ -53,9 +53,9 @@ namespace ARMeilleure.Instructions
 
             switch (GetPackedId(op))
             {
-                case 0b11_011_0100_0010_000: EmitSetNzcv(context);                                                          return;
-                case 0b11_011_0100_0100_000: info = typeof(NativeInterface).GetMethod(nameof(NativeInterface.SetFpcr));     break;
-                case 0b11_011_0100_0100_001: info = typeof(NativeInterface).GetMethod(nameof(NativeInterface.SetFpsr));     break;
+                case 0b11_011_0100_0010_000: EmitSetNzcv(context); return;
+                case 0b11_011_0100_0100_000: EmitSetFpcr(context); return;
+                case 0b11_011_0100_0100_001: EmitSetFpsr(context); return;
                 case 0b11_011_1101_0000_010: info = typeof(NativeInterface).GetMethod(nameof(NativeInterface.SetTpidrEl0)); break;
 
                 default: throw new NotImplementedException($"Unknown MSR 0x{op.RawOpCode:X8} at 0x{op.Address:X16}.");
@@ -121,39 +121,91 @@ namespace ARMeilleure.Instructions
         {
             OpCodeSystem op = (OpCodeSystem)context.CurrOp;
 
-            Operand vSh = context.ShiftLeft(GetFlag(PState.VFlag), Const((int)PState.VFlag));
-            Operand cSh = context.ShiftLeft(GetFlag(PState.CFlag), Const((int)PState.CFlag));
-            Operand zSh = context.ShiftLeft(GetFlag(PState.ZFlag), Const((int)PState.ZFlag));
-            Operand nSh = context.ShiftLeft(GetFlag(PState.NFlag), Const((int)PState.NFlag));
+            Operand nzcv = context.ShiftLeft(GetFlag(PState.VFlag), Const((int)PState.VFlag));
+            nzcv = context.BitwiseOr(nzcv, context.ShiftLeft(GetFlag(PState.CFlag), Const((int)PState.CFlag)));
+            nzcv = context.BitwiseOr(nzcv, context.ShiftLeft(GetFlag(PState.ZFlag), Const((int)PState.ZFlag)));
+            nzcv = context.BitwiseOr(nzcv, context.ShiftLeft(GetFlag(PState.NFlag), Const((int)PState.NFlag)));
 
-            Operand nzcvSh = context.BitwiseOr(context.BitwiseOr(nSh, zSh), context.BitwiseOr(cSh, vSh));
+            SetIntOrZR(context, op.Rt, nzcv);
+        }
 
-            SetIntOrZR(context, op.Rt, nzcvSh);
+        private static void EmitGetFpcr(ArmEmitterContext context)
+        {
+            OpCodeSystem op = (OpCodeSystem)context.CurrOp;
+
+            Operand fpcr = Const(0);
+
+            for (int flag = 0; flag < RegisterConsts.FpFlagsCount; flag++)
+            {
+                if (FPCR.Mask.HasFlag((FPCR)(1u << flag)))
+                {
+                    fpcr = context.BitwiseOr(fpcr, context.ShiftLeft(GetFpFlag((FPState)flag), Const(flag)));
+                }
+            }
+
+            SetIntOrZR(context, op.Rt, fpcr);
+        }
+
+        private static void EmitGetFpsr(ArmEmitterContext context)
+        {
+            OpCodeSystem op = (OpCodeSystem)context.CurrOp;
+
+            Operand fpsr = Const(0);
+
+            for (int flag = 0; flag < RegisterConsts.FpFlagsCount; flag++)
+            {
+                if (FPSR.Mask.HasFlag((FPSR)(1u << flag)))
+                {
+                    fpsr = context.BitwiseOr(fpsr, context.ShiftLeft(GetFpFlag((FPState)flag), Const(flag)));
+                }
+            }
+
+            SetIntOrZR(context, op.Rt, fpsr);
         }
 
         private static void EmitSetNzcv(ArmEmitterContext context)
         {
             OpCodeSystem op = (OpCodeSystem)context.CurrOp;
 
-            Operand t = GetIntOrZR(context, op.Rt);
-                    t = context.ConvertI64ToI32(t);
+            Operand nzcv = GetIntOrZR(context, op.Rt);
+                    nzcv = context.ConvertI64ToI32(nzcv);
 
-            Operand v = context.ShiftRightUI(t, Const((int)PState.VFlag));
-                    v = context.BitwiseAnd  (v, Const(1));
+            SetFlag(context, PState.VFlag, context.BitwiseAnd(context.ShiftRightUI(nzcv, Const((int)PState.VFlag)), Const(1)));
+            SetFlag(context, PState.CFlag, context.BitwiseAnd(context.ShiftRightUI(nzcv, Const((int)PState.CFlag)), Const(1)));
+            SetFlag(context, PState.ZFlag, context.BitwiseAnd(context.ShiftRightUI(nzcv, Const((int)PState.ZFlag)), Const(1)));
+            SetFlag(context, PState.NFlag, context.BitwiseAnd(context.ShiftRightUI(nzcv, Const((int)PState.NFlag)), Const(1)));
+        }
 
-            Operand c = context.ShiftRightUI(t, Const((int)PState.CFlag));
-                    c = context.BitwiseAnd  (c, Const(1));
+        private static void EmitSetFpcr(ArmEmitterContext context)
+        {
+            OpCodeSystem op = (OpCodeSystem)context.CurrOp;
 
-            Operand z = context.ShiftRightUI(t, Const((int)PState.ZFlag));
-                    z = context.BitwiseAnd  (z, Const(1));
+            Operand fpcr = GetIntOrZR(context, op.Rt);
+                    fpcr = context.ConvertI64ToI32(fpcr);
 
-            Operand n = context.ShiftRightUI(t, Const((int)PState.NFlag));
-                    n = context.BitwiseAnd  (n, Const(1));
+            for (int flag = 0; flag < RegisterConsts.FpFlagsCount; flag++)
+            {
+                if (FPCR.Mask.HasFlag((FPCR)(1u << flag)))
+                {
+                    SetFpFlag(context, (FPState)flag, context.BitwiseAnd(context.ShiftRightUI(fpcr, Const(flag)), Const(1)));
+                }
+            }
+        }
 
-            SetFlag(context, PState.VFlag, v);
-            SetFlag(context, PState.CFlag, c);
-            SetFlag(context, PState.ZFlag, z);
-            SetFlag(context, PState.NFlag, n);
+        private static void EmitSetFpsr(ArmEmitterContext context)
+        {
+            OpCodeSystem op = (OpCodeSystem)context.CurrOp;
+
+            Operand fpsr = GetIntOrZR(context, op.Rt);
+                    fpsr = context.ConvertI64ToI32(fpsr);
+
+            for (int flag = 0; flag < RegisterConsts.FpFlagsCount; flag++)
+            {
+                if (FPSR.Mask.HasFlag((FPSR)(1u << flag)))
+                {
+                    SetFpFlag(context, (FPState)flag, context.BitwiseAnd(context.ShiftRightUI(fpsr, Const(flag)), Const(1)));
+                }
+            }
         }
     }
 }

ARMeilleure/Instructions/InstEmitSystem32.cs

@@ -169,14 +169,11 @@ namespace ARMeilleure.Instructions
             }
             else
             {
-                Operand vSh = context.ShiftLeft(GetFlag(PState.VFlag), Const((int)PState.VFlag));
-                Operand cSh = context.ShiftLeft(GetFlag(PState.CFlag), Const((int)PState.CFlag));
-                Operand zSh = context.ShiftLeft(GetFlag(PState.ZFlag), Const((int)PState.ZFlag));
-                Operand nSh = context.ShiftLeft(GetFlag(PState.NFlag), Const((int)PState.NFlag));
-                Operand qSh = context.ShiftLeft(GetFlag(PState.QFlag), Const((int)PState.QFlag));
-
-                Operand spsr = context.BitwiseOr(context.BitwiseOr(nSh, zSh), context.BitwiseOr(cSh, vSh));
-                spsr = context.BitwiseOr(spsr, qSh);
+                Operand spsr = context.ShiftLeft(GetFlag(PState.VFlag), Const((int)PState.VFlag));
+                spsr = context.BitwiseOr(spsr, context.ShiftLeft(GetFlag(PState.CFlag), Const((int)PState.CFlag)));
+                spsr = context.BitwiseOr(spsr, context.ShiftLeft(GetFlag(PState.ZFlag), Const((int)PState.ZFlag)));
+                spsr = context.BitwiseOr(spsr, context.ShiftLeft(GetFlag(PState.NFlag), Const((int)PState.NFlag)));
+                spsr = context.BitwiseOr(spsr, context.ShiftLeft(GetFlag(PState.QFlag), Const((int)PState.QFlag)));
 
                 // TODO: Remaining flags.
 
@@ -200,8 +197,7 @@ namespace ARMeilleure.Instructions
 
                     EmitSetNzcv(context, value);
 
-                    Operand q = context.ShiftRightUI(value, Const((int)PState.QFlag));
-                    q = context.BitwiseAnd(q, Const(1));
+                    Operand q = context.BitwiseAnd(context.ShiftRightUI(value, Const((int)PState.QFlag)), Const(1));
 
                     SetFlag(context, PState.QFlag, q);
                 }
@@ -284,17 +280,10 @@ namespace ARMeilleure.Instructions
 
         private static void EmitSetNzcv(ArmEmitterContext context, Operand t)
         {
-            Operand v = context.ShiftRightUI(t, Const((int)PState.VFlag));
-            v = context.BitwiseAnd(v, Const(1));
-
-            Operand c = context.ShiftRightUI(t, Const((int)PState.CFlag));
-            c = context.BitwiseAnd(c, Const(1));
-
-            Operand z = context.ShiftRightUI(t, Const((int)PState.ZFlag));
-            z = context.BitwiseAnd(z, Const(1));
-
-            Operand n = context.ShiftRightUI(t, Const((int)PState.NFlag));
-            n = context.BitwiseAnd(n, Const(1));
+            Operand v = context.BitwiseAnd(context.ShiftRightUI(t, Const((int)PState.VFlag)), Const(1));
+            Operand c = context.BitwiseAnd(context.ShiftRightUI(t, Const((int)PState.CFlag)), Const(1));
+            Operand z = context.BitwiseAnd(context.ShiftRightUI(t, Const((int)PState.ZFlag)), Const(1));
+            Operand n = context.BitwiseAnd(context.ShiftRightUI(t, Const((int)PState.NFlag)), Const(1));
 
             SetFlag(context, PState.VFlag, v);
             SetFlag(context, PState.CFlag, c);
@@ -306,42 +295,32 @@ namespace ARMeilleure.Instructions
         {
             OpCode32SimdSpecial op = (OpCode32SimdSpecial)context.CurrOp;
 
-            Operand vSh = context.ShiftLeft(GetFpFlag(FPState.VFlag), Const((int)FPState.VFlag));
-            Operand cSh = context.ShiftLeft(GetFpFlag(FPState.CFlag), Const((int)FPState.CFlag));
-            Operand zSh = context.ShiftLeft(GetFpFlag(FPState.ZFlag), Const((int)FPState.ZFlag));
-            Operand nSh = context.ShiftLeft(GetFpFlag(FPState.NFlag), Const((int)FPState.NFlag));
+            Operand fpscr = Const(0);
 
-            Operand nzcvSh = context.BitwiseOr(context.BitwiseOr(nSh, zSh), context.BitwiseOr(cSh, vSh));
+            for (int flag = 0; flag < RegisterConsts.FpFlagsCount; flag++)
+            {
+                if (FPSCR.Mask.HasFlag((FPSCR)(1u << flag)))
+                {
+                    fpscr = context.BitwiseOr(fpscr, context.ShiftLeft(GetFpFlag((FPState)flag), Const(flag)));
+                }
+            }
 
-            Operand fpscr = context.Call(typeof(NativeInterface).GetMethod(nameof(NativeInterface.GetFpscr)));
-
-            SetIntA32(context, op.Rt, context.BitwiseOr(nzcvSh, fpscr));
+            SetIntA32(context, op.Rt, fpscr);
         }
 
         private static void EmitSetFpscr(ArmEmitterContext context)
         {
             OpCode32SimdSpecial op = (OpCode32SimdSpecial)context.CurrOp;
 
-            Operand t = GetIntA32(context, op.Rt);
+            Operand fpscr = GetIntA32(context, op.Rt);
 
-            Operand v = context.ShiftRightUI(t, Const((int)FPState.VFlag));
-            v = context.BitwiseAnd(v, Const(1));
-
-            Operand c = context.ShiftRightUI(t, Const((int)FPState.CFlag));
-            c = context.BitwiseAnd(c, Const(1));
-
-            Operand z = context.ShiftRightUI(t, Const((int)FPState.ZFlag));
-            z = context.BitwiseAnd(z, Const(1));
-
-            Operand n = context.ShiftRightUI(t, Const((int)FPState.NFlag));
-            n = context.BitwiseAnd(n, Const(1));
-
-            SetFpFlag(context, FPState.VFlag, v);
-            SetFpFlag(context, FPState.CFlag, c);
-            SetFpFlag(context, FPState.ZFlag, z);
-            SetFpFlag(context, FPState.NFlag, n);
-
-            context.Call(typeof(NativeInterface).GetMethod(nameof(NativeInterface.SetFpscr)), t);
+            for (int flag = 0; flag < RegisterConsts.FpFlagsCount; flag++)
+            {
+                if (FPSCR.Mask.HasFlag((FPSCR)(1u << flag)))
+                {
+                    SetFpFlag(context, (FPState)flag, context.BitwiseAnd(context.ShiftRightUI(fpscr, Const(flag)), Const(1)));
+                }
+            }
         }
     }
 }

ARMeilleure/Instructions/NativeInterface.cs

@@ -72,29 +72,6 @@ namespace ARMeilleure.Instructions
             return (ulong)GetContext().DczidEl0;
         }
 
-        public static ulong GetFpcr()
-        {
-            return (ulong)GetContext().Fpcr;
-        }
-
-        public static bool GetFpcrFz()
-        {
-            return (GetContext().Fpcr & FPCR.Fz) != 0;
-        }
-
-        public static ulong GetFpsr()
-        {
-            return (ulong)GetContext().Fpsr;
-        }
-
-        public static uint GetFpscr()
-        {
-            ExecutionContext context = GetContext();
-
-            return (uint)(context.Fpsr & FPSR.A32Mask & ~FPSR.Nzcv) |
-                   (uint)(context.Fpcr & FPCR.A32Mask);
-        }
-
         public static ulong GetTpidrEl0()
         {
             return (ulong)GetContext().TpidrEl0;
@@ -130,29 +107,6 @@ namespace ARMeilleure.Instructions
             return GetContext().CntvctEl0;
         }
 
-        public static void SetFpcr(ulong value)
-        {
-            GetContext().Fpcr = (FPCR)value;
-        }
-
-        public static void SetFpsr(ulong value)
-        {
-            GetContext().Fpsr = (FPSR)value;
-        }
-
-        public static void SetFpsrQc()
-        {
-            GetContext().Fpsr |= FPSR.Qc;
-        }
-
-        public static void SetFpscr(uint fpscr)
-        {
-            ExecutionContext context = GetContext();
-
-            context.Fpsr = FPSR.A32Mask & (FPSR)fpscr;
-            context.Fpcr = FPCR.A32Mask & (FPCR)fpscr;
-        }
-
         public static void SetTpidrEl0(ulong value)
         {
             GetContext().TpidrEl0 = (long)value;

ARMeilleure/Instructions/SoftFallback.cs

@@ -91,76 +91,6 @@ namespace ARMeilleure.Instructions
         }
 #endregion
 
-#region "Rounding"
-        public static double Round(double value)
-        {
-            ExecutionContext context = NativeInterface.GetContext();
-
-            FPRoundingMode roundMode = context.Fpcr.GetRoundingMode();
-
-            if (roundMode == FPRoundingMode.ToNearest)
-            {
-                return Math.Round(value); // even
-            }
-            else if (roundMode == FPRoundingMode.TowardsPlusInfinity)
-            {
-                return Math.Ceiling(value);
-            }
-            else if (roundMode == FPRoundingMode.TowardsMinusInfinity)
-            {
-                return Math.Floor(value);
-            }
-            else /* if (roundMode == FPRoundingMode.TowardsZero) */
-            {
-                return Math.Truncate(value);
-            }
-        }
-
-        public static float RoundF(float value)
-        {
-            ExecutionContext context = NativeInterface.GetContext();
-
-            FPRoundingMode roundMode = context.Fpcr.GetRoundingMode();
-
-            if (roundMode == FPRoundingMode.ToNearest)
-            {
-                return MathF.Round(value); // even
-            }
-            else if (roundMode == FPRoundingMode.TowardsPlusInfinity)
-            {
-                return MathF.Ceiling(value);
-            }
-            else if (roundMode == FPRoundingMode.TowardsMinusInfinity)
-            {
-                return MathF.Floor(value);
-            }
-            else /* if (roundMode == FPRoundingMode.TowardsZero) */
-            {
-                return MathF.Truncate(value);
-            }
-        }
-
-        public static int FloatToInt32(float value)
-        {
-            return SatF32ToS32(RoundF(value));
-        }
-
-        public static int DoubleToInt32(double value)
-        {
-            return SatF64ToS32(Round(value));
-        }
-
-        public static uint FloatToUInt32(float value)
-        {
-            return SatF32ToU32(RoundF(value));
-        }
-
-        public static uint DoubleToUInt32(double value)
-        {
-            return SatF64ToU32(Round(value));
-        }
-#endregion
-
 #region "Saturation"
         public static int SatF32ToS32(float value)
         {

ARMeilleure/Instructions/SoftFloat.cs

@@ -12,8 +12,8 @@ namespace ARMeilleure.Instructions
             RecipSqrtEstimateTable = BuildRecipSqrtEstimateTable();
         }
 
-        internal static readonly byte[] RecipEstimateTable;
-        internal static readonly byte[] RecipSqrtEstimateTable;
+        public static readonly byte[] RecipEstimateTable;
+        public static readonly byte[] RecipSqrtEstimateTable;
 
         private static byte[] BuildRecipEstimateTable()
         {
@@ -94,6 +94,13 @@ namespace ARMeilleure.Instructions
                 context.Fpsr |= (FPSR)(1 << (int)exc);
             }
         }
+
+        public static FPRoundingMode GetRoundingMode(this FPCR fpcr)
+        {
+            const int RModeShift = 22;
+
+            return (FPRoundingMode)(((uint)fpcr >> RModeShift) & 3u);
+        }
     }
 
     static class SoftFloat16

ARMeilleure/IntermediateRepresentation/Intrinsic.cs

@@ -47,6 +47,7 @@ namespace ARMeilleure.IntermediateRepresentation
         X86Divps,
         X86Divsd,
         X86Divss,
+        X86Gf2p8affineqb,
         X86Haddpd,
         X86Haddps,
         X86Insertps,

ARMeilleure/IntermediateRepresentation/PhiOperation.cs

@@ -3,7 +3,7 @@ using static ARMeilleure.IntermediateRepresentation.Operand.Factory;
 
 namespace ARMeilleure.IntermediateRepresentation
 {
-    struct PhiOperation
+    readonly struct PhiOperation
     {
         private readonly Operation _operation;
 

ARMeilleure/IntermediateRepresentation/Register.cs

@@ -2,7 +2,7 @@ using System;
 
 namespace ARMeilleure.IntermediateRepresentation
 {
-    struct Register : IEquatable<Register>
+    readonly struct Register : IEquatable<Register>
     {
         public int Index { get; }
 

ARMeilleure/Optimizations.cs

@@ -22,6 +22,7 @@ namespace ARMeilleure
         public static bool UseAesniIfAvailable     { get; set; } = true;
         public static bool UsePclmulqdqIfAvailable { get; set; } = true;
         public static bool UseShaIfAvailable       { get; set; } = true;
+        public static bool UseGfniIfAvailable      { get; set; } = true;
 
         public static bool ForceLegacySse
         {
@@ -42,5 +43,6 @@ namespace ARMeilleure
         internal static bool UseAesni     => UseAesniIfAvailable     && HardwareCapabilities.SupportsAesni;
         internal static bool UsePclmulqdq => UsePclmulqdqIfAvailable && HardwareCapabilities.SupportsPclmulqdq;
         internal static bool UseSha       => UseShaIfAvailable       && HardwareCapabilities.SupportsSha;
+        internal static bool UseGfni      => UseGfniIfAvailable      && HardwareCapabilities.SupportsGfni;
     }
 }

ARMeilleure/State/ExecutionContext.cs

@@ -36,10 +36,25 @@ namespace ARMeilleure.State
             set => _nativeContext.SetPstate(value);
         }
 
-        public FPCR Fpcr { get; set; }
-        public FPSR Fpsr { get; set; }
+        public FPSR Fpsr
+        {
+            get => (FPSR)_nativeContext.GetFPState((uint)FPSR.Mask);
+            set => _nativeContext.SetFPState((uint)value, (uint)FPSR.Mask);
+        }
+
+        public FPCR Fpcr
+        {
+            get => (FPCR)_nativeContext.GetFPState((uint)FPCR.Mask);
+            set => _nativeContext.SetFPState((uint)value, (uint)FPCR.Mask);
+        }
         public FPCR StandardFpcrValue => (Fpcr & (FPCR.Ahp)) | FPCR.Dn | FPCR.Fz;
 
+        public FPSCR Fpscr
+        {
+            get => (FPSCR)_nativeContext.GetFPState((uint)FPSCR.Mask);
+            set => _nativeContext.SetFPState((uint)value, (uint)FPSCR.Mask);
+        }
+
         public bool IsAarch32 { get; set; }
 
         internal ExecutionMode ExecutionMode

ARMeilleure/State/FPCR.cs

@@ -5,21 +5,18 @@ namespace ARMeilleure.State
     [Flags]
     public enum FPCR : uint
     {
+        Ioe = 1u << 8,
+        Dze = 1u << 9,
+        Ofe = 1u << 10,
         Ufe = 1u << 11,
+        Ixe = 1u << 12,
+        Ide = 1u << 15,
+        RMode0 = 1u << 22,
+        RMode1 = 1u << 23,
         Fz  = 1u << 24,
         Dn  = 1u << 25,
         Ahp = 1u << 26,
 
-        A32Mask = 0x07FF9F00u
-    }
-
-    public static class FPCRExtensions
-    {
-        private const int RModeShift = 22;
-
-        public static FPRoundingMode GetRoundingMode(this FPCR fpcr)
-        {
-            return (FPRoundingMode)(((int)fpcr >> RModeShift) & 3);
-        }
+        Mask = Ahp | Dn | Fz | RMode1 | RMode0 | Ide | Ixe | Ufe | Ofe | Dze | Ioe // 0x07C09F00u
     }
 }

ARMeilleure/State/FPRoundingMode.cs

@@ -2,9 +2,10 @@ namespace ARMeilleure.State
 {
     public enum FPRoundingMode
     {
-        ToNearest            = 0,
+        ToNearest            = 0, // With ties to even.
         TowardsPlusInfinity  = 1,
         TowardsMinusInfinity = 2,
-        TowardsZero          = 3
+        TowardsZero          = 3,
+        ToNearestAway        = 4 // With ties to away.
     }
 }

ARMeilleure/State/FPSCR.cs

@@ -0,0 +1,15 @@
+using System;
+
+namespace ARMeilleure.State
+{
+    [Flags]
+    public enum FPSCR : uint
+    {
+        V = 1u << 28,
+        C = 1u << 29,
+        Z = 1u << 30,
+        N = 1u << 31,
+
+        Mask = N | Z | C | V | FPSR.Mask | FPCR.Mask // 0xFFC09F9Fu
+    }
+}

ARMeilleure/State/FPSR.cs

@@ -5,11 +5,14 @@ namespace ARMeilleure.State
     [Flags]
     public enum FPSR : uint
     {
+        Ioc = 1u << 0,
+        Dzc = 1u << 1,
+        Ofc = 1u << 2,
         Ufc = 1u << 3,
-        Qc  = 1u << 27,
+        Ixc = 1u << 4,
+        Idc = 1u << 7,
+        Qc = 1u << 27,
 
-        Nzcv = (1u << 31) | (1u << 30) | (1u << 29) | (1u << 28),
-
-        A32Mask = 0xF800009Fu
+        Mask = Qc | Idc | Ixc | Ufc | Ofc | Dzc | Ioc // 0x0800009Fu
     }
 }

ARMeilleure/State/FPState.cs

@@ -2,9 +2,30 @@
 {
     public enum FPState
     {
+        // FPSR Flags.
+        IocFlag = 0,
+        DzcFlag = 1,
+        OfcFlag = 2,
+        UfcFlag = 3,
+        IxcFlag = 4,
+        IdcFlag = 7,
+        QcFlag = 27,
         VFlag = 28,
         CFlag = 29,
         ZFlag = 30,
-        NFlag = 31
+        NFlag = 31,
+
+        // FPCR Flags.
+        IoeFlag = 8,
+        DzeFlag = 9,
+        OfeFlag = 10,
+        UfeFlag = 11,
+        IxeFlag = 12,
+        IdeFlag = 15,
+        RMode0Flag = 22,
+        RMode1Flag = 23,
+        FzFlag = 24,
+        DnFlag = 25,
+        AhpFlag = 26
     }
 }

ARMeilleure/State/NativeContext.cs

@@ -140,6 +140,34 @@ namespace ARMeilleure.State
             GetStorage().FpFlags[(int)flag] = value ? 1u : 0u;
         }
 
+        public unsafe uint GetFPState(uint mask = uint.MaxValue)
+        {
+            uint value = 0;
+            for (int flag = 0; flag < RegisterConsts.FpFlagsCount; flag++)
+            {
+                uint bit = 1u << flag;
+
+                if ((mask & bit) == bit)
+                {
+                    value |= GetStorage().FpFlags[flag] != 0 ? bit : 0u;
+                }
+            }
+            return value;
+        }
+
+        public unsafe void SetFPState(uint value, uint mask = uint.MaxValue)
+        {
+            for (int flag = 0; flag < RegisterConsts.FpFlagsCount; flag++)
+            {
+                uint bit = 1u << flag;
+
+                if ((mask & bit) == bit)
+                {
+                    GetStorage().FpFlags[flag] = (value & bit) == bit ? 1u : 0u;
+                }
+            }
+        }
+
         public int GetCounter() => GetStorage().Counter;
         public void SetCounter(int value) => GetStorage().Counter = value;
 

ARMeilleure/Translation/Cache/CacheEntry.cs

@@ -4,7 +4,7 @@ using System.Diagnostics.CodeAnalysis;
 
 namespace ARMeilleure.Translation.Cache
 {
-    struct CacheEntry : IComparable<CacheEntry>
+    readonly struct CacheEntry : IComparable<CacheEntry>
     {
         public int Offset { get; }
         public int Size   { get; }

ARMeilleure/Translation/Cache/CacheMemoryAllocator.cs

@@ -6,7 +6,7 @@ namespace ARMeilleure.Translation.Cache
 {
     class CacheMemoryAllocator
     {
-        private struct MemoryBlock : IComparable<MemoryBlock>
+        private readonly struct MemoryBlock : IComparable<MemoryBlock>
         {
             public int Offset { get; }
             public int Size { get; }

ARMeilleure/Translation/CompilerContext.cs

@@ -2,7 +2,7 @@ using ARMeilleure.IntermediateRepresentation;
 
 namespace ARMeilleure.Translation
 {
-    struct CompilerContext
+    readonly struct CompilerContext
     {
         public ControlFlowGraph Cfg { get; }
 

ARMeilleure/Translation/Delegates.cs

@@ -109,10 +109,6 @@ namespace ARMeilleure.Translation
             SetDelegateInfo(typeof(NativeInterface).GetMethod(nameof(NativeInterface.GetCntvctEl0)));
             SetDelegateInfo(typeof(NativeInterface).GetMethod(nameof(NativeInterface.GetCtrEl0)));
             SetDelegateInfo(typeof(NativeInterface).GetMethod(nameof(NativeInterface.GetDczidEl0)));
-            SetDelegateInfo(typeof(NativeInterface).GetMethod(nameof(NativeInterface.GetFpcr)));
-            SetDelegateInfo(typeof(NativeInterface).GetMethod(nameof(NativeInterface.GetFpcrFz)));
-            SetDelegateInfo(typeof(NativeInterface).GetMethod(nameof(NativeInterface.GetFpscr))); // A32 only.
-            SetDelegateInfo(typeof(NativeInterface).GetMethod(nameof(NativeInterface.GetFpsr)));
             SetDelegateInfo(typeof(NativeInterface).GetMethod(nameof(NativeInterface.GetFunctionAddress)));
             SetDelegateInfo(typeof(NativeInterface).GetMethod(nameof(NativeInterface.InvalidateCacheLine)));
             SetDelegateInfo(typeof(NativeInterface).GetMethod(nameof(NativeInterface.GetTpidrroEl0)));
@@ -124,10 +120,6 @@ namespace ARMeilleure.Translation
             SetDelegateInfo(typeof(NativeInterface).GetMethod(nameof(NativeInterface.ReadUInt32)));
             SetDelegateInfo(typeof(NativeInterface).GetMethod(nameof(NativeInterface.ReadUInt64)));
             SetDelegateInfo(typeof(NativeInterface).GetMethod(nameof(NativeInterface.ReadVector128)));
-            SetDelegateInfo(typeof(NativeInterface).GetMethod(nameof(NativeInterface.SetFpcr)));
-            SetDelegateInfo(typeof(NativeInterface).GetMethod(nameof(NativeInterface.SetFpscr))); // A32 only.
-            SetDelegateInfo(typeof(NativeInterface).GetMethod(nameof(NativeInterface.SetFpsr)));
-            SetDelegateInfo(typeof(NativeInterface).GetMethod(nameof(NativeInterface.SetFpsrQc)));
             SetDelegateInfo(typeof(NativeInterface).GetMethod(nameof(NativeInterface.SetTpidrEl0)));
             SetDelegateInfo(typeof(NativeInterface).GetMethod(nameof(NativeInterface.SetTpidrEl032))); // A32 only.
             SetDelegateInfo(typeof(NativeInterface).GetMethod(nameof(NativeInterface.SignalMemoryTracking)));
@@ -151,12 +143,8 @@ namespace ARMeilleure.Translation
             SetDelegateInfo(typeof(SoftFallback).GetMethod(nameof(SoftFallback.Crc32w)));
             SetDelegateInfo(typeof(SoftFallback).GetMethod(nameof(SoftFallback.Crc32x)));
             SetDelegateInfo(typeof(SoftFallback).GetMethod(nameof(SoftFallback.Decrypt)));
-            SetDelegateInfo(typeof(SoftFallback).GetMethod(nameof(SoftFallback.DoubleToInt32))); // A32 only.
-            SetDelegateInfo(typeof(SoftFallback).GetMethod(nameof(SoftFallback.DoubleToUInt32))); // A32 only.
             SetDelegateInfo(typeof(SoftFallback).GetMethod(nameof(SoftFallback.Encrypt)));
             SetDelegateInfo(typeof(SoftFallback).GetMethod(nameof(SoftFallback.FixedRotate)));
-            SetDelegateInfo(typeof(SoftFallback).GetMethod(nameof(SoftFallback.FloatToInt32))); // A32 only.
-            SetDelegateInfo(typeof(SoftFallback).GetMethod(nameof(SoftFallback.FloatToUInt32))); // A32 only.
             SetDelegateInfo(typeof(SoftFallback).GetMethod(nameof(SoftFallback.HashChoose)));
             SetDelegateInfo(typeof(SoftFallback).GetMethod(nameof(SoftFallback.HashLower)));
             SetDelegateInfo(typeof(SoftFallback).GetMethod(nameof(SoftFallback.HashMajority)));
@@ -165,8 +153,6 @@ namespace ARMeilleure.Translation
             SetDelegateInfo(typeof(SoftFallback).GetMethod(nameof(SoftFallback.InverseMixColumns)));
             SetDelegateInfo(typeof(SoftFallback).GetMethod(nameof(SoftFallback.MixColumns)));
             SetDelegateInfo(typeof(SoftFallback).GetMethod(nameof(SoftFallback.PolynomialMult64_128)));
-            SetDelegateInfo(typeof(SoftFallback).GetMethod(nameof(SoftFallback.Round)));
-            SetDelegateInfo(typeof(SoftFallback).GetMethod(nameof(SoftFallback.RoundF)));
             SetDelegateInfo(typeof(SoftFallback).GetMethod(nameof(SoftFallback.SatF32ToS32)));
             SetDelegateInfo(typeof(SoftFallback).GetMethod(nameof(SoftFallback.SatF32ToS64)));
             SetDelegateInfo(typeof(SoftFallback).GetMethod(nameof(SoftFallback.SatF32ToU32)));

ARMeilleure/Translation/PTC/Ptc.cs

@@ -27,7 +27,7 @@ namespace ARMeilleure.Translation.PTC
         private const string OuterHeaderMagicString = "PTCohd\0\0";
         private const string InnerHeaderMagicString = "PTCihd\0\0";
 
-        private const uint InternalVersion = 3700; //! To be incremented manually for each change to the ARMeilleure project.
+        private const uint InternalVersion = 3713; //! To be incremented manually for each change to the ARMeilleure project.
 
         private const string ActualDir = "0";
         private const string BackupDir = "1";
@@ -951,7 +951,8 @@ namespace ARMeilleure.Translation.PTC
             return new FeatureInfo(
                 (uint)HardwareCapabilities.FeatureInfo1Ecx,
                 (uint)HardwareCapabilities.FeatureInfo1Edx,
-                (uint)HardwareCapabilities.FeatureInfo7Ebx);
+                (uint)HardwareCapabilities.FeatureInfo7Ebx,
+                (uint)HardwareCapabilities.FeatureInfo7Ecx);
         }
 
         private static byte GetMemoryManagerMode()
@@ -971,7 +972,7 @@ namespace ARMeilleure.Translation.PTC
             return osPlatform;
         }
 
-        [StructLayout(LayoutKind.Sequential, Pack = 1/*, Size = 54*/)]
+        [StructLayout(LayoutKind.Sequential, Pack = 1/*, Size = 58*/)]
         private struct OuterHeader
         {
             public ulong Magic;
@@ -1002,8 +1003,8 @@ namespace ARMeilleure.Translation.PTC
             }
         }
 
-        [StructLayout(LayoutKind.Sequential, Pack = 1/*, Size = 12*/)]
-        private record struct FeatureInfo(uint FeatureInfo0, uint FeatureInfo1, uint FeatureInfo2);
+        [StructLayout(LayoutKind.Sequential, Pack = 1/*, Size = 16*/)]
+        private record struct FeatureInfo(uint FeatureInfo0, uint FeatureInfo1, uint FeatureInfo2, uint FeatureInfo3);
 
         [StructLayout(LayoutKind.Sequential, Pack = 1/*, Size = 128*/)]
         private struct InnerHeader

ARMeilleure/Translation/RegisterUsage.cs

@@ -14,7 +14,7 @@ namespace ARMeilleure.Translation
         private const int RegsCount = 32;
         private const int RegsMask  = RegsCount - 1;
 
-        private struct RegisterMask : IEquatable<RegisterMask>
+        private readonly struct RegisterMask : IEquatable<RegisterMask>
         {
             public long IntMask => Mask.GetElement(0);
             public long VecMask => Mask.GetElement(1);

ARMeilleure/Translation/Translator.cs

@@ -293,7 +293,7 @@ namespace ARMeilleure.Translation
             }
         }
 
-        private struct Range
+        private readonly struct Range
         {
             public ulong Start { get; }
             public ulong End { get; }
@@ -455,13 +455,16 @@ namespace ARMeilleure.Translation
 
         public void InvalidateJitCacheRegion(ulong address, ulong size)
         {
-            // If rejit is running, stop it as it may be trying to rejit a function on the invalidated region.
-            ClearRejitQueue(allowRequeue: true);
-
             ulong[] overlapAddresses = Array.Empty<ulong>();
 
             int overlapsCount = Functions.GetOverlaps(address, size, ref overlapAddresses);
 
+            if (overlapsCount != 0)
+            {
+                // If rejit is running, stop it as it may be trying to rejit a function on the invalidated region.
+                ClearRejitQueue(allowRequeue: true);
+            }
+
             for (int index = 0; index < overlapsCount; index++)
             {
                 ulong overlapAddress = overlapAddresses[index];

README.md

@@ -21,6 +21,10 @@
         <img src="https://github.com/Ryujinx/Ryujinx/actions/workflows/release.yml/badge.svg"
             alt="">
     </a>
+    <a href="https://crwd.in/ryujinx">
+        <img src="https://badges.crowdin.net/ryujinx/localized.svg"
+            alt="">
+    </a>
     <a href="https://discord.com/invite/VkQYXAZ">
         <img src="https://img.shields.io/discord/410208534861447168?color=5865F2&label=Ryujinx&logo=discord&logoColor=white"
             alt="Discord">
@@ -36,18 +40,20 @@
 
 ## Compatibility
 
-As of September 2022, Ryujinx has been tested on approximately 3,600 titles; over 3,400 boot past menus and into gameplay, with roughly 2,700 of those being considered playable. You can check out the compatibility list [here](https://github.com/Ryujinx/Ryujinx-Games-List/issues).  
-Anyone is free to submit a new game test or update an existing game test entry; simply follow the new issue template and testing guidelines, or post as a reply to the applicable game issue. Use the search function to see if a game has been tested already!
+As of November 2022, Ryujinx has been tested on approximately 3,800 titles; over 3,600 boot past menus and into gameplay, with roughly 3,200 of those being considered playable.
+You can check out the compatibility list [here](https://github.com/Ryujinx/Ryujinx-Games-List/issues). Anyone is free to submit a new game test or update an existing game test entry; simply follow the new issue template and testing guidelines, or post as a reply to the applicable game issue. Use the search function to see if a game has been tested already!
 
 ## Usage
 
-To run this emulator, your PC must be equipped with at least 8GB of RAM; failing to meet this requirement may result in a poor gameplay experience or unexpected crashes.
+To run this emulator, your PC must be equipped with at least 8GiB of RAM; failing to meet this requirement may result in a poor gameplay experience or unexpected crashes.
 
 See our [Setup & Configuration Guide](https://github.com/Ryujinx/Ryujinx/wiki/Ryujinx-Setup-&-Configuration-Guide) on how to set up the emulator.
 
 For our Local Wireless and LAN builds, see our [Multiplayer: Local Play/Local Wireless Guide
 ](https://github.com/Ryujinx/Ryujinx/wiki/Multiplayer-(LDN-Local-Wireless)-Guide).
 
+Avalonia UI comes with translations for various languages. See [Crowdin](https://crwd.in/ryujinx) for more information.
+
 ## Latest build
 
 These builds are compiled automatically for each commit on the master branch. While we strive to ensure optimal stability and performance prior to pushing an update, our automated builds **may be unstable or completely broken.**
@@ -62,7 +68,7 @@ The latest automatic build for Windows, macOS, and Linux can be found on the [Of
 If you wish to build the emulator yourself, follow these steps:
 
 ### Step 1
-Install the X64 version of [.NET 6.0 (or higher) SDK](https://dotnet.microsoft.com/download/dotnet/6.0).
+Install the X64 version of [.NET 7.0 (or higher) SDK](https://dotnet.microsoft.com/download/dotnet/7.0).
 
 ### Step 2
 Either use `git clone https://github.com/Ryujinx/Ryujinx` on the command line to clone the repository or use Code --> Download zip button to get the files.
@@ -90,7 +96,7 @@ Ryujinx system files are stored in the `Ryujinx` folder. This folder is located
 
 - **GPU**
 
-  The GPU emulator emulates the Switch's Maxwell GPU using the OpenGL API (version 4.5 minimum) through a custom build of OpenTK. There are currently four graphics enhancements available to the end user in Ryujinx: disk shader caching, resolution scaling, aspect ratio adjustment and anisotropic filtering. These enhancements can be adjusted or toggled as desired in the GUI.
+  The GPU emulator emulates the Switch's Maxwell GPU using either the OpenGL (version 4.5 minimum), Vulkan, or Metal (via MoltenVK) APIs through a custom build of OpenTK or Silk.NET respectively. There are currently four graphics enhancements available to the end user in Ryujinx: Disk Shader Caching, Resolution Scaling, Aspect Ratio Adjustment, and Anisotropic Filtering. These enhancements can be adjusted or toggled as desired in the GUI.
 
 - **Input**
 

Ryujinx.Audio.Backends.OpenAL/Ryujinx.Audio.Backends.OpenAL.csproj

@@ -1,11 +1,11 @@
 <Project Sdk="Microsoft.NET.Sdk">
 
   <PropertyGroup>
-    <TargetFramework>net6.0</TargetFramework>
+    <TargetFramework>net7.0</TargetFramework>
   </PropertyGroup>
 
   <ItemGroup>
-    <PackageReference Include="OpenTK.OpenAL" Version="4.7.2" />
+    <PackageReference Include="OpenTK.OpenAL" Version="4.7.5" />
   </ItemGroup>
 
   <ItemGroup>

Ryujinx.Audio.Backends.SDL2/Ryujinx.Audio.Backends.SDL2.csproj

@@ -1,7 +1,7 @@
 <Project Sdk="Microsoft.NET.Sdk">
 
   <PropertyGroup>
-    <TargetFramework>net6.0</TargetFramework>
+    <TargetFramework>net7.0</TargetFramework>
     <AllowUnsafeBlocks>true</AllowUnsafeBlocks>
   </PropertyGroup>
 

Ryujinx.Audio.Backends.SDL2/SDL2HardwareDeviceDriver.cs

@@ -1,5 +1,6 @@
 using Ryujinx.Audio.Common;
 using Ryujinx.Audio.Integration;
+using Ryujinx.Common.Logging;
 using Ryujinx.Memory;
 using Ryujinx.SDL2.Common;
 using System;
@@ -112,6 +113,9 @@ namespace Ryujinx.Audio.Backends.SDL2
 
             if (device == 0)
             {
+                Logger.Error?.Print(LogClass.Application, 
+                    $"SDL2 open audio device initialization failed with error \"{SDL_GetError()}\"");
+
                 return 0;
             }
 
@@ -119,6 +123,7 @@ namespace Ryujinx.Audio.Backends.SDL2
 
             if (!isValid)
             {
+                Logger.Error?.Print(LogClass.Application, "SDL2 open audio device is not valid");
                 SDL_CloseAudioDevice(device);
 
                 return 0;

Ryujinx.Audio.Backends.SoundIo/Native/libsoundio/SoundIOChannelLayout.cs

@@ -4,7 +4,7 @@ using System.Runtime.InteropServices;
 
 namespace SoundIOSharp
 {
-    public struct SoundIOChannelLayout
+    public readonly struct SoundIOChannelLayout
     {
         public static int BuiltInCount
         {

Ryujinx.Audio.Backends.SoundIo/Native/libsoundio/SoundIOSampleRateRange.cs

@@ -1,6 +1,6 @@
 namespace SoundIOSharp
 {
-    public struct SoundIOSampleRateRange
+    public readonly struct SoundIOSampleRateRange
     {
         internal SoundIOSampleRateRange(int min, int max)
         {

Ryujinx.Audio.Backends.SoundIo/Ryujinx.Audio.Backends.SoundIo.csproj

@@ -1,7 +1,7 @@
 <Project Sdk="Microsoft.NET.Sdk">
 
   <PropertyGroup>
-    <TargetFramework>net6.0</TargetFramework>
+    <TargetFramework>net7.0</TargetFramework>
     <AllowUnsafeBlocks>true</AllowUnsafeBlocks>
     <RuntimeIdentifiers>win10-x64;linux-x64;osx-x64</RuntimeIdentifiers>
   </PropertyGroup>

Ryujinx.Audio/Renderer/Common/EffectType.cs

@@ -48,6 +48,11 @@ namespace Ryujinx.Audio.Renderer.Common
         /// <summary>
         /// Effect to capture mixes (via auxiliary buffers).
         /// </summary>
-        CaptureBuffer
+        CaptureBuffer,
+
+        /// <summary>
+        /// Effect applying a compressor filter (DRC).
+        /// </summary>
+        Compressor,
     }
 }

Ryujinx.Audio/Renderer/Common/PerformanceDetailType.cs

@@ -14,6 +14,7 @@ namespace Ryujinx.Audio.Renderer.Common
         Reverb3d,
         PcmFloat,
         Limiter,
-        CaptureBuffer
+        CaptureBuffer,
+        Compressor
     }
 }

Ryujinx.Audio/Renderer/Dsp/AdpcmHelper.cs

@@ -1,4 +1,5 @@
 using Ryujinx.Audio.Renderer.Dsp.State;
+using Ryujinx.Common.Logging;
 using System;
 using System.Diagnostics;
 using System.Runtime.CompilerServices;
@@ -71,6 +72,19 @@ namespace Ryujinx.Audio.Renderer.Dsp
             return (short)value;
         }
 
+        [MethodImpl(MethodImplOptions.AggressiveInlining)]
+        private static short GetCoefficientAtIndex(ReadOnlySpan<short> coefficients, int index)
+        {
+            if ((uint)index > (uint)coefficients.Length)
+            {
+                Logger.Error?.Print(LogClass.AudioRenderer, $"Out of bound read for coefficient at index {index}");
+
+                return 0;
+            }
+
+            return coefficients[index];
+        }
+
         [MethodImpl(MethodImplOptions.AggressiveInlining)]
         public static int Decode(Span<short> output, ReadOnlySpan<byte> input, int startSampleOffset, int endSampleOffset, int offset, int count, ReadOnlySpan<short> coefficients, ref AdpcmLoopContext loopContext)
         {
@@ -84,8 +98,8 @@ namespace Ryujinx.Audio.Renderer.Dsp
             byte coefficientIndex = (byte)((predScale >> 4) & 0xF);
             short history0 = loopContext.History0;
             short history1 = loopContext.History1;
-            short coefficient0 = coefficients[coefficientIndex * 2 + 0];
-            short coefficient1 = coefficients[coefficientIndex * 2 + 1];
+            short coefficient0 = GetCoefficientAtIndex(coefficients, coefficientIndex * 2 + 0);
+            short coefficient1 = GetCoefficientAtIndex(coefficients, coefficientIndex * 2 + 1);
 
             int decodedCount = Math.Min(count, endSampleOffset - startSampleOffset - offset);
             int nibbles = GetNibblesFromSampleCount(offset + startSampleOffset);
@@ -109,8 +123,8 @@ namespace Ryujinx.Audio.Renderer.Dsp
 
                     coefficientIndex = (byte)((predScale >> 4) & 0xF);
 
-                    coefficient0 = coefficients[coefficientIndex * 2 + 0];
-                    coefficient1 = coefficients[coefficientIndex * 2 + 1];
+                    coefficient0 = GetCoefficientAtIndex(coefficients, coefficientIndex * 2);
+                    coefficient1 = GetCoefficientAtIndex(coefficients, coefficientIndex * 2 + 1);
 
                     nibbles += 2;
 

Ryujinx.Audio/Renderer/Dsp/AudioProcessor.cs

@@ -116,6 +116,11 @@ namespace Ryujinx.Audio.Renderer.Dsp
             };
         }
 
+        public bool HasRemainingCommands(int sessionId)
+        {
+            return _sessionCommandList[sessionId] != null;
+        }
+
         public void Signal()
         {
             _mailbox.SendMessage(MailboxMessage.RenderStart);

Ryujinx.Audio/Renderer/Dsp/Command/AdpcmDataSourceCommandVersion1.cs

@@ -13,7 +13,7 @@ namespace Ryujinx.Audio.Renderer.Dsp.Command
 
         public CommandType CommandType => CommandType.AdpcmDataSourceVersion1;
 
-        public ulong EstimatedProcessingTime { get; set; }
+        public uint EstimatedProcessingTime { get; set; }
 
         public ushort OutputBufferIndex { get; }
         public uint SampleRate { get; }

Ryujinx.Audio/Renderer/Dsp/Command/AuxiliaryBufferCommand.cs

@@ -16,7 +16,7 @@ namespace Ryujinx.Audio.Renderer.Dsp.Command
 
         public CommandType CommandType => CommandType.AuxiliaryBuffer;
 
-        public ulong EstimatedProcessingTime { get; set; }
+        public uint EstimatedProcessingTime { get; set; }
 
         public uint InputBufferIndex { get; }
         public uint OutputBufferIndex { get; }

Ryujinx.Audio/Renderer/Dsp/Command/BiquadFilterCommand.cs

@@ -12,7 +12,7 @@ namespace Ryujinx.Audio.Renderer.Dsp.Command
 
         public CommandType CommandType => CommandType.BiquadFilter;
 
-        public ulong EstimatedProcessingTime { get; set; }
+        public uint EstimatedProcessingTime { get; set; }
 
         public Memory<BiquadFilterState> BiquadFilterState { get; }
         public int InputBufferIndex { get; }

Ryujinx.Audio/Renderer/Dsp/Command/CaptureBufferCommand.cs

@@ -16,7 +16,7 @@ namespace Ryujinx.Audio.Renderer.Dsp.Command
 
         public CommandType CommandType => CommandType.CaptureBuffer;
 
-        public ulong EstimatedProcessingTime { get; set; }
+        public uint EstimatedProcessingTime { get; set; }
 
         public uint InputBufferIndex { get; }
 

Ryujinx.Audio/Renderer/Dsp/Command/CircularBufferSinkCommand.cs

@@ -12,7 +12,7 @@ namespace Ryujinx.Audio.Renderer.Dsp.Command
 
         public CommandType CommandType => CommandType.CircularBufferSink;
 
-        public ulong EstimatedProcessingTime { get; set; }
+        public uint EstimatedProcessingTime { get; set; }
 
         public ushort[] Input { get; }
         public uint InputCount { get; }

Ryujinx.Audio/Renderer/Dsp/Command/ClearMixBufferCommand.cs

@@ -8,7 +8,7 @@ namespace Ryujinx.Audio.Renderer.Dsp.Command
 
         public CommandType CommandType => CommandType.ClearMixBuffer;
 
-        public ulong EstimatedProcessingTime { get; set; }
+        public uint EstimatedProcessingTime { get; set; }
 
         public ClearMixBufferCommand(int nodeId)
         {

Ryujinx.Audio/Renderer/Dsp/Command/CommandType.cs

@@ -31,6 +31,7 @@ namespace Ryujinx.Audio.Renderer.Dsp.Command
         LimiterVersion1,
         LimiterVersion2,
         GroupedBiquadFilter,
-        CaptureBuffer
+        CaptureBuffer,
+        Compressor
     }
 }

Ryujinx.Audio/Renderer/Dsp/Command/CompressorCommand.cs

@@ -0,0 +1,173 @@
+using System;
+using System.Diagnostics;
+using Ryujinx.Audio.Renderer.Dsp.Effect;
+using Ryujinx.Audio.Renderer.Dsp.State;
+using Ryujinx.Audio.Renderer.Parameter.Effect;
+
+namespace Ryujinx.Audio.Renderer.Dsp.Command
+{
+    public class CompressorCommand : ICommand
+    {
+        private const int FixedPointPrecision = 15;
+
+        public bool Enabled { get; set; }
+
+        public int NodeId { get; }
+
+        public CommandType CommandType => CommandType.Compressor;
+
+        public uint EstimatedProcessingTime { get; set; }
+
+        public CompressorParameter Parameter => _parameter;
+        public Memory<CompressorState> State { get; }
+        public ushort[] OutputBufferIndices { get; }
+        public ushort[] InputBufferIndices { get; }
+        public bool IsEffectEnabled { get; }
+
+        private CompressorParameter _parameter;
+
+        public CompressorCommand(uint bufferOffset, CompressorParameter parameter, Memory<CompressorState> state, bool isEnabled, int nodeId)
+        {
+            Enabled = true;
+            NodeId = nodeId;
+            _parameter = parameter;
+            State = state;
+
+            IsEffectEnabled = isEnabled;
+
+            InputBufferIndices = new ushort[Constants.VoiceChannelCountMax];
+            OutputBufferIndices = new ushort[Constants.VoiceChannelCountMax];
+
+            for (int i = 0; i < _parameter.ChannelCount; i++)
+            {
+                InputBufferIndices[i] = (ushort)(bufferOffset + _parameter.Input[i]);
+                OutputBufferIndices[i] = (ushort)(bufferOffset + _parameter.Output[i]);
+            }
+        }
+
+        public void Process(CommandList context)
+        {
+            ref CompressorState state = ref State.Span[0];
+
+            if (IsEffectEnabled)
+            {
+                if (_parameter.Status == Server.Effect.UsageState.Invalid)
+                {
+                    state = new CompressorState(ref _parameter);
+                }
+                else if (_parameter.Status == Server.Effect.UsageState.New)
+                {
+                    state.UpdateParameter(ref _parameter);
+                }
+            }
+
+            ProcessCompressor(context, ref state);
+        }
+
+        private unsafe void ProcessCompressor(CommandList context, ref CompressorState state)
+        {
+            Debug.Assert(_parameter.IsChannelCountValid());
+
+            if (IsEffectEnabled && _parameter.IsChannelCountValid())
+            {
+                Span<IntPtr> inputBuffers = stackalloc IntPtr[Parameter.ChannelCount];
+                Span<IntPtr> outputBuffers = stackalloc IntPtr[Parameter.ChannelCount];
+                Span<float> channelInput = stackalloc float[Parameter.ChannelCount];
+                ExponentialMovingAverage inputMovingAverage = state.InputMovingAverage;
+                float unknown4 = state.Unknown4;
+                ExponentialMovingAverage compressionGainAverage = state.CompressionGainAverage;
+                float previousCompressionEmaAlpha = state.PreviousCompressionEmaAlpha;
+
+                for (int i = 0; i < _parameter.ChannelCount; i++)
+                {
+                    inputBuffers[i] = context.GetBufferPointer(InputBufferIndices[i]);
+                    outputBuffers[i] = context.GetBufferPointer(OutputBufferIndices[i]);
+                }
+
+                for (int sampleIndex = 0; sampleIndex < context.SampleCount; sampleIndex++)
+                {
+                    for (int channelIndex = 0; channelIndex < _parameter.ChannelCount; channelIndex++)
+                    {
+                        channelInput[channelIndex] = *((float*)inputBuffers[channelIndex] + sampleIndex);
+                    }
+
+                    float newMean = inputMovingAverage.Update(FloatingPointHelper.MeanSquare(channelInput), _parameter.InputGain);
+                    float y = FloatingPointHelper.Log10(newMean) * 10.0f;
+                    float z = 0.0f;
+
+                    bool unknown10OutOfRange = false;
+
+                    if (newMean < 1.0e-10f)
+                    {
+                        z = 1.0f;
+
+                        unknown10OutOfRange = state.Unknown10 < -100.0f;
+                    }
+
+                    if (y >= state.Unknown10 || unknown10OutOfRange)
+                    {
+                        float tmpGain;
+
+                        if (y >= state.Unknown14)
+                        {
+                            tmpGain = ((1.0f / Parameter.Ratio) - 1.0f) * (y - Parameter.Threshold);
+                        }
+                        else
+                        {
+                            tmpGain = (y - state.Unknown10) * ((y - state.Unknown10) * -state.CompressorGainReduction);
+                        }
+
+                        z = FloatingPointHelper.DecibelToLinearExtended(tmpGain);
+                    }
+
+                    float unknown4New = z;
+                    float compressionEmaAlpha;
+
+                    if ((unknown4 - z) <= 0.08f)
+                    {
+                        compressionEmaAlpha = Parameter.ReleaseCoefficient;
+
+                        if ((unknown4 - z) >= -0.08f)
+                        {
+                            if (MathF.Abs(compressionGainAverage.Read() - z) >= 0.001f)
+                            {
+                                unknown4New = unknown4;
+                            }
+
+                            compressionEmaAlpha = previousCompressionEmaAlpha;
+                        }
+                    }
+                    else
+                    {
+                        compressionEmaAlpha = Parameter.AttackCoefficient;
+                    }
+
+                    float compressionGain = compressionGainAverage.Update(z, compressionEmaAlpha);
+
+                    for (int channelIndex = 0; channelIndex < Parameter.ChannelCount; channelIndex++)
+                    {
+                        *((float*)outputBuffers[channelIndex] + sampleIndex) = channelInput[channelIndex] * compressionGain * state.OutputGain;
+                    }
+
+                    unknown4 = unknown4New;
+                    previousCompressionEmaAlpha = compressionEmaAlpha;
+                }
+
+                state.InputMovingAverage = inputMovingAverage;
+                state.Unknown4 = unknown4;
+                state.CompressionGainAverage = compressionGainAverage;
+                state.PreviousCompressionEmaAlpha = previousCompressionEmaAlpha;
+            }
+            else
+            {
+                for (int i = 0; i < Parameter.ChannelCount; i++)
+                {
+                    if (InputBufferIndices[i] != OutputBufferIndices[i])
+                    {
+                        context.CopyBuffer(OutputBufferIndices[i], InputBufferIndices[i]);
+                    }
+                }
+            }
+        }
+    }
+}

Ryujinx.Audio/Renderer/Dsp/Command/CopyMixBufferCommand.cs

@@ -8,7 +8,7 @@ namespace Ryujinx.Audio.Renderer.Dsp.Command
 
         public CommandType CommandType => CommandType.CopyMixBuffer;
 
-        public ulong EstimatedProcessingTime { get; set; }
+        public uint EstimatedProcessingTime { get; set; }
 
         public ushort InputBufferIndex { get; }
         public ushort OutputBufferIndex { get; }

Ryujinx.Audio/Renderer/Dsp/Command/DataSourceVersion2Command.cs

@@ -13,7 +13,7 @@ namespace Ryujinx.Audio.Renderer.Dsp.Command
 
         public CommandType CommandType { get; }
 
-        public ulong EstimatedProcessingTime { get; set; }
+        public uint EstimatedProcessingTime { get; set; }
 
         public ushort OutputBufferIndex { get; }
         public uint SampleRate { get; }

Ryujinx.Audio/Renderer/Dsp/Command/DelayCommand.cs

@@ -17,7 +17,7 @@ namespace Ryujinx.Audio.Renderer.Dsp.Command
 
         public CommandType CommandType => CommandType.Delay;
 
-        public ulong EstimatedProcessingTime { get; set; }
+        public uint EstimatedProcessingTime { get; set; }
 
         public DelayParameter Parameter => _parameter;
         public Memory<DelayState> State { get; }
@@ -49,15 +49,15 @@ namespace Ryujinx.Audio.Renderer.Dsp.Command
                 OutputBufferIndices[i] = (ushort)(bufferOffset + Parameter.Output[i]);
             }
 
-            // NOTE: We do the opposite as Nintendo here for now to restore previous behaviour
-            // TODO: Update delay processing and remove this to use RemapLegacyChannelEffectMappingToChannelResourceMapping.
-            DataSourceHelper.RemapChannelResourceMappingToLegacy(newEffectChannelMappingSupported, InputBufferIndices);
-            DataSourceHelper.RemapChannelResourceMappingToLegacy(newEffectChannelMappingSupported, OutputBufferIndices);
+            DataSourceHelper.RemapLegacyChannelEffectMappingToChannelResourceMapping(newEffectChannelMappingSupported, InputBufferIndices);
+            DataSourceHelper.RemapLegacyChannelEffectMappingToChannelResourceMapping(newEffectChannelMappingSupported, OutputBufferIndices);
         }
 
         [MethodImpl(MethodImplOptions.AggressiveInlining | MethodImplOptions.AggressiveOptimization)]
         private unsafe void ProcessDelayMono(ref DelayState state, float* outputBuffer, float* inputBuffer, uint sampleCount)
         {
+            const ushort channelCount = 1;
+
             float feedbackGain = FixedPointHelper.ToFloat(Parameter.FeedbackGain, FixedPointPrecision);
             float inGain = FixedPointHelper.ToFloat(Parameter.InGain, FixedPointPrecision);
             float dryGain = FixedPointHelper.ToFloat(Parameter.DryGain, FixedPointPrecision);
@@ -70,7 +70,7 @@ namespace Ryujinx.Audio.Renderer.Dsp.Command
 
                 float temp = input * inGain + delayLineValue * feedbackGain;
 
-                state.UpdateLowPassFilter(ref temp, 1);
+                state.UpdateLowPassFilter(ref temp, channelCount);
 
                 outputBuffer[i] = (input * dryGain + delayLineValue * outGain) / 64;
             }
@@ -104,7 +104,7 @@ namespace Ryujinx.Audio.Renderer.Dsp.Command
                     Y = state.DelayLines[1].Read(),
                 };
 
-                Vector2 temp = MatrixHelper.Transform(ref channelInput, ref delayFeedback) + channelInput * inGain;
+                Vector2 temp = MatrixHelper.Transform(ref delayLineValues, ref delayFeedback) + channelInput * inGain;
 
                 state.UpdateLowPassFilter(ref Unsafe.As<Vector2, float>(ref temp), channelCount);
 
@@ -148,7 +148,7 @@ namespace Ryujinx.Audio.Renderer.Dsp.Command
                     W = state.DelayLines[3].Read()
                 };
 
-                Vector4 temp = MatrixHelper.Transform(ref channelInput, ref delayFeedback) + channelInput * inGain;
+                Vector4 temp = MatrixHelper.Transform(ref delayLineValues, ref delayFeedback) + channelInput * inGain;
 
                 state.UpdateLowPassFilter(ref Unsafe.As<Vector4, float>(ref temp), channelCount);
 
@@ -171,12 +171,12 @@ namespace Ryujinx.Audio.Renderer.Dsp.Command
             float dryGain = FixedPointHelper.ToFloat(Parameter.DryGain, FixedPointPrecision);
             float outGain = FixedPointHelper.ToFloat(Parameter.OutGain, FixedPointPrecision);
 
-            Matrix6x6 delayFeedback = new Matrix6x6(delayFeedbackBaseGain, 0.0f, 0.0f, 0.0f, delayFeedbackCrossGain, delayFeedbackCrossGain,
-                                                    0.0f, delayFeedbackBaseGain, 0.0f, delayFeedbackCrossGain, delayFeedbackCrossGain, 0.0f,
-                                                    delayFeedbackCrossGain, 0.0f, delayFeedbackBaseGain, delayFeedbackCrossGain, 0.0f, 0.0f,
-                                                    0.0f, delayFeedbackCrossGain, delayFeedbackCrossGain, delayFeedbackBaseGain, 0.0f, 0.0f,
-                                                    delayFeedbackCrossGain, delayFeedbackCrossGain, 0.0f, 0.0f, delayFeedbackBaseGain, 0.0f,
-                                                    0.0f, 0.0f, 0.0f, 0.0f, 0.0f, feedbackGain);
+            Matrix6x6 delayFeedback = new Matrix6x6(delayFeedbackBaseGain, 0.0f, delayFeedbackCrossGain, 0.0f, delayFeedbackCrossGain, 0.0f,
+                                                    0.0f, delayFeedbackBaseGain, delayFeedbackCrossGain, 0.0f, 0.0f, delayFeedbackCrossGain,
+                                                    delayFeedbackCrossGain, delayFeedbackCrossGain, delayFeedbackBaseGain, 0.0f, 0.0f, 0.0f,
+                                                    0.0f, 0.0f, 0.0f, feedbackGain, 0.0f, 0.0f,
+                                                    delayFeedbackCrossGain, 0.0f, 0.0f, 0.0f, delayFeedbackBaseGain, delayFeedbackCrossGain,
+                                                    0.0f, delayFeedbackCrossGain, 0.0f, 0.0f, delayFeedbackCrossGain, delayFeedbackBaseGain);
 
             for (int i = 0; i < sampleCount; i++)
             {
@@ -200,7 +200,7 @@ namespace Ryujinx.Audio.Renderer.Dsp.Command
                     U = state.DelayLines[5].Read()
                 };
 
-                Vector6 temp = MatrixHelper.Transform(ref channelInput, ref delayFeedback) + channelInput * inGain;
+                Vector6 temp = MatrixHelper.Transform(ref delayLineValues, ref delayFeedback) + channelInput * inGain;
 
                 state.UpdateLowPassFilter(ref Unsafe.As<Vector6, float>(ref temp), channelCount);
 

Ryujinx.Audio/Renderer/Dsp/Command/DepopForMixBuffersCommand.cs

@@ -11,7 +11,7 @@ namespace Ryujinx.Audio.Renderer.Dsp.Command
 
         public CommandType CommandType => CommandType.DepopForMixBuffers;
 
-        public ulong EstimatedProcessingTime { get; set; }
+        public uint EstimatedProcessingTime { get; set; }
 
         public uint MixBufferOffset { get; }
 

Ryujinx.Audio/Renderer/Dsp/Command/DepopPrepareCommand.cs

@@ -11,7 +11,7 @@ namespace Ryujinx.Audio.Renderer.Dsp.Command
 
         public CommandType CommandType => CommandType.DepopPrepare;
 
-        public ulong EstimatedProcessingTime { get; set; }
+        public uint EstimatedProcessingTime { get; set; }
 
         public uint MixBufferCount { get; }
 

Ryujinx.Audio/Renderer/Dsp/Command/DeviceSinkCommand.cs

@@ -14,7 +14,7 @@ namespace Ryujinx.Audio.Renderer.Dsp.Command
 
         public CommandType CommandType => CommandType.DeviceSink;
 
-        public ulong EstimatedProcessingTime { get; set; }
+        public uint EstimatedProcessingTime { get; set; }
 
         public string DeviceName { get; }
 

Ryujinx.Audio/Renderer/Dsp/Command/DownMixSurroundToStereoCommand.cs

@@ -11,7 +11,7 @@ namespace Ryujinx.Audio.Renderer.Dsp.Command
 
         public CommandType CommandType => CommandType.DownMixSurroundToStereo;
 
-        public ulong EstimatedProcessingTime { get; set; }
+        public uint EstimatedProcessingTime { get; set; }
 
         public ushort[] InputBufferIndices { get; }
         public ushort[] OutputBufferIndices { get; }

Ryujinx.Audio/Renderer/Dsp/Command/GroupedBiquadFilterCommand.cs

@@ -12,7 +12,7 @@ namespace Ryujinx.Audio.Renderer.Dsp.Command
 
         public CommandType CommandType => CommandType.GroupedBiquadFilter;
 
-        public ulong EstimatedProcessingTime { get; set; }
+        public uint EstimatedProcessingTime { get; set; }
 
         private BiquadFilterParameter[] _parameters;
         private Memory<BiquadFilterState> _biquadFilterStates;
@@ -47,9 +47,8 @@ namespace Ryujinx.Audio.Renderer.Dsp.Command
                 }
             }
 
-            // NOTE: Nintendo also implements a hot path for double biquad filters, but no generic path when the command definition suggests it could be done.
-            // As such we currently only implement a generic path for simplicity.
-            // TODO: Implement double biquad filters fast path.
+            // NOTE: Nintendo only implement single and double biquad filters but no generic path when the command definition suggests it could be done.
+            // As such we currently only implement a generic path for simplicity for double biquad.
             if (_parameters.Length == 1)
             {
                 BiquadFilterHelper.ProcessBiquadFilter(ref _parameters[0], ref states[0], outputBuffer, inputBuffer, context.SampleCount);

Ryujinx.Audio/Renderer/Dsp/Command/ICommand.cs

@@ -8,7 +8,7 @@ namespace Ryujinx.Audio.Renderer.Dsp.Command
 
         public CommandType CommandType { get; }
 
-        public ulong EstimatedProcessingTime { get; }
+        public uint EstimatedProcessingTime { get; }
 
         public void Process(CommandList context);
 

Ryujinx.Audio/Renderer/Dsp/Command/LimiterCommandVersion1.cs

@@ -13,7 +13,7 @@ namespace Ryujinx.Audio.Renderer.Dsp.Command
 
         public CommandType CommandType => CommandType.LimiterVersion1;
 
-        public ulong EstimatedProcessingTime { get; set; }
+        public uint EstimatedProcessingTime { get; set; }
 
         public LimiterParameter Parameter => _parameter;
         public Memory<LimiterState> State { get; }
@@ -90,32 +90,31 @@ namespace Ryujinx.Audio.Renderer.Dsp.Command
 
                         float inputCoefficient = Parameter.ReleaseCoefficient;
 
-                        if (sampleInputMax > state.DectectorAverage[channelIndex])
+                        if (sampleInputMax > state.DetectorAverage[channelIndex].Read())
                         {
                             inputCoefficient = Parameter.AttackCoefficient;
                         }
 
-                        state.DectectorAverage[channelIndex] += inputCoefficient * (sampleInputMax - state.DectectorAverage[channelIndex]);
-
+                        float detectorValue = state.DetectorAverage[channelIndex].Update(sampleInputMax, inputCoefficient);
                         float attenuation = 1.0f;
 
-                        if (state.DectectorAverage[channelIndex] > Parameter.Threshold)
+                        if (detectorValue > Parameter.Threshold)
                         {
-                            attenuation = Parameter.Threshold / state.DectectorAverage[channelIndex];
+                            attenuation = Parameter.Threshold / detectorValue;
                         }
 
                         float outputCoefficient = Parameter.ReleaseCoefficient;
 
-                        if (state.CompressionGain[channelIndex] > attenuation)
+                        if (state.CompressionGainAverage[channelIndex].Read() > attenuation)
                         {
                             outputCoefficient = Parameter.AttackCoefficient;
                         }
 
-                        state.CompressionGain[channelIndex] += outputCoefficient * (attenuation - state.CompressionGain[channelIndex]);
+                        float compressionGain = state.CompressionGainAverage[channelIndex].Update(attenuation, outputCoefficient);
 
                         ref float delayedSample = ref state.DelayedSampleBuffer[channelIndex * Parameter.DelayBufferSampleCountMax + state.DelayedSampleBufferPosition[channelIndex]];
 
-                        float outputSample = delayedSample * state.CompressionGain[channelIndex] * Parameter.OutputGain;
+                        float outputSample = delayedSample * compressionGain * Parameter.OutputGain;
 
                         *((float*)outputBuffers[channelIndex] + sampleIndex) = outputSample * short.MaxValue;
 

Ryujinx.Audio/Renderer/Dsp/Command/LimiterCommandVersion2.cs

@@ -15,7 +15,7 @@ namespace Ryujinx.Audio.Renderer.Dsp.Command
 
         public CommandType CommandType => CommandType.LimiterVersion2;
 
-        public ulong EstimatedProcessingTime { get; set; }
+        public uint EstimatedProcessingTime { get; set; }
 
         public LimiterParameter Parameter => _parameter;
         public Memory<LimiterState> State { get; }
@@ -101,32 +101,31 @@ namespace Ryujinx.Audio.Renderer.Dsp.Command
 
                         float inputCoefficient = Parameter.ReleaseCoefficient;
 
-                        if (sampleInputMax > state.DectectorAverage[channelIndex])
+                        if (sampleInputMax > state.DetectorAverage[channelIndex].Read())
                         {
                             inputCoefficient = Parameter.AttackCoefficient;
                         }
 
-                        state.DectectorAverage[channelIndex] += inputCoefficient * (sampleInputMax - state.DectectorAverage[channelIndex]);
-
+                        float detectorValue = state.DetectorAverage[channelIndex].Update(sampleInputMax, inputCoefficient);
                         float attenuation = 1.0f;
 
-                        if (state.DectectorAverage[channelIndex] > Parameter.Threshold)
+                        if (detectorValue > Parameter.Threshold)
                         {
-                            attenuation = Parameter.Threshold / state.DectectorAverage[channelIndex];
+                            attenuation = Parameter.Threshold / detectorValue;
                         }
 
                         float outputCoefficient = Parameter.ReleaseCoefficient;
 
-                        if (state.CompressionGain[channelIndex] > attenuation)
+                        if (state.CompressionGainAverage[channelIndex].Read() > attenuation)
                         {
                             outputCoefficient = Parameter.AttackCoefficient;
                         }
 
-                        state.CompressionGain[channelIndex] += outputCoefficient * (attenuation - state.CompressionGain[channelIndex]);
+                        float compressionGain = state.CompressionGainAverage[channelIndex].Update(attenuation, outputCoefficient);
 
                         ref float delayedSample = ref state.DelayedSampleBuffer[channelIndex * Parameter.DelayBufferSampleCountMax + state.DelayedSampleBufferPosition[channelIndex]];
 
-                        float outputSample = delayedSample * state.CompressionGain[channelIndex] * Parameter.OutputGain;
+                        float outputSample = delayedSample * compressionGain * Parameter.OutputGain;
 
                         *((float*)outputBuffers[channelIndex] + sampleIndex) = outputSample * short.MaxValue;
 
@@ -144,7 +143,7 @@ namespace Ryujinx.Audio.Renderer.Dsp.Command
                             ref LimiterStatistics statistics = ref MemoryMarshal.Cast<byte, LimiterStatistics>(ResultState.Span[0].SpecificData)[0];
 
                             statistics.InputMax[channelIndex] = Math.Max(statistics.InputMax[channelIndex], sampleInputMax);
-                            statistics.CompressionGainMin[channelIndex] = Math.Min(statistics.CompressionGainMin[channelIndex], state.CompressionGain[channelIndex]);
+                            statistics.CompressionGainMin[channelIndex] = Math.Min(statistics.CompressionGainMin[channelIndex], compressionGain);
                         }
                     }
                 }

Ryujinx.Audio/Renderer/Dsp/Command/MixCommand.cs

@@ -15,7 +15,7 @@ namespace Ryujinx.Audio.Renderer.Dsp.Command
 
         public CommandType CommandType => CommandType.Mix;
 
-        public ulong EstimatedProcessingTime { get; set; }
+        public uint EstimatedProcessingTime { get; set; }
 
         public ushort InputBufferIndex { get; }
         public ushort OutputBufferIndex { get; }

Ryujinx.Audio/Renderer/Dsp/Command/MixRampCommand.cs

@@ -12,7 +12,7 @@ namespace Ryujinx.Audio.Renderer.Dsp.Command
 
         public CommandType CommandType => CommandType.MixRamp;
 
-        public ulong EstimatedProcessingTime { get; set; }
+        public uint EstimatedProcessingTime { get; set; }
 
         public ushort InputBufferIndex { get; }
         public ushort OutputBufferIndex { get; }

Ryujinx.Audio/Renderer/Dsp/Command/MixRampGroupedCommand.cs

@@ -12,7 +12,7 @@ namespace Ryujinx.Audio.Renderer.Dsp.Command
 
         public CommandType CommandType => CommandType.MixRampGrouped;
 
-        public ulong EstimatedProcessingTime { get; set; }
+        public uint EstimatedProcessingTime { get; set; }
 
         public uint MixBufferCount { get; }
 

Ryujinx.Audio/Renderer/Dsp/Command/PcmFloatDataSourceCommandVersion1.cs

@@ -13,7 +13,7 @@ namespace Ryujinx.Audio.Renderer.Dsp.Command
 
         public CommandType CommandType => CommandType.PcmFloatDataSourceVersion1;
 
-        public ulong EstimatedProcessingTime { get; set; }
+        public uint EstimatedProcessingTime { get; set; }
 
         public ushort OutputBufferIndex { get; }
         public uint SampleRate { get; }

Ryujinx.Audio/Renderer/Dsp/Command/PcmInt16DataSourceCommandVersion1.cs

@@ -13,7 +13,7 @@ namespace Ryujinx.Audio.Renderer.Dsp.Command
 
         public CommandType CommandType => CommandType.PcmInt16DataSourceVersion1;
 
-        public ulong EstimatedProcessingTime { get; set; }
+        public uint EstimatedProcessingTime { get; set; }
 
         public ushort OutputBufferIndex { get; }
         public uint SampleRate { get; }

Ryujinx.Audio/Renderer/Dsp/Command/PerformanceCommand.cs

@@ -17,7 +17,7 @@ namespace Ryujinx.Audio.Renderer.Dsp.Command
 
         public CommandType CommandType => CommandType.Performance;
 
-        public ulong EstimatedProcessingTime { get; set; }
+        public uint EstimatedProcessingTime { get; set; }
 
         public PerformanceEntryAddresses PerformanceEntryAddresses { get; }
 

Ryujinx.Audio/Renderer/Dsp/Command/Reverb3dCommand.cs

@@ -31,7 +31,7 @@ namespace Ryujinx.Audio.Renderer.Dsp.Command
 
         public CommandType CommandType => CommandType.Reverb3d;
 
-        public ulong EstimatedProcessingTime { get; set; }
+        public uint EstimatedProcessingTime { get; set; }
 
         public ushort InputBufferIndex { get; }
         public ushort OutputBufferIndex { get; }

Ryujinx.Audio/Renderer/Dsp/Command/ReverbCommand.cs

@@ -34,7 +34,7 @@ namespace Ryujinx.Audio.Renderer.Dsp.Command
 
         public CommandType CommandType => CommandType.Reverb;
 
-        public ulong EstimatedProcessingTime { get; set; }
+        public uint EstimatedProcessingTime { get; set; }
 
         public ReverbParameter Parameter => _parameter;
         public Memory<ReverbState> State { get; }

Ryujinx.Audio/Renderer/Dsp/Command/UpsampleCommand.cs

@@ -11,7 +11,7 @@ namespace Ryujinx.Audio.Renderer.Dsp.Command
 
         public CommandType CommandType => CommandType.Upsample;
 
-        public ulong EstimatedProcessingTime { get; set; }
+        public uint EstimatedProcessingTime { get; set; }
 
         public uint BufferCount { get; }
         public uint InputBufferIndex { get; }

Ryujinx.Audio/Renderer/Dsp/Command/VolumeCommand.cs

@@ -15,7 +15,7 @@ namespace Ryujinx.Audio.Renderer.Dsp.Command
 
         public CommandType CommandType => CommandType.Volume;
 
-        public ulong EstimatedProcessingTime { get; set; }
+        public uint EstimatedProcessingTime { get; set; }
 
         public ushort InputBufferIndex { get; }
         public ushort OutputBufferIndex { get; }

Ryujinx.Audio/Renderer/Dsp/Command/VolumeRampCommand.cs

@@ -11,7 +11,7 @@ namespace Ryujinx.Audio.Renderer.Dsp.Command
 
         public CommandType CommandType => CommandType.VolumeRamp;
 
-        public ulong EstimatedProcessingTime { get; set; }
+        public uint EstimatedProcessingTime { get; set; }
 
         public ushort InputBufferIndex { get; }
         public ushort OutputBufferIndex { get; }

Ryujinx.Audio/Renderer/Dsp/Effect/ExponentialMovingAverage.cs

@@ -0,0 +1,26 @@
+using System.Runtime.CompilerServices;
+
+namespace Ryujinx.Audio.Renderer.Dsp.Effect
+{
+    public struct ExponentialMovingAverage
+    {
+        private float _mean;
+
+        public ExponentialMovingAverage(float mean)
+        {
+            _mean = mean;
+        }
+
+        public float Read()
+        {
+            return _mean;
+        }
+
+        public float Update(float value, float alpha)
+        {
+            _mean += alpha * (value - _mean);
+
+            return _mean;
+        }
+    }
+}

Ryujinx.Audio/Renderer/Dsp/FixedPointHelper.cs

@@ -16,6 +16,12 @@ namespace Ryujinx.Audio.Renderer.Dsp
             return (float)value / (1 << qBits);
         }
 
+        [MethodImpl(MethodImplOptions.AggressiveInlining)]
+        public static float ConvertFloat(float value, int qBits)
+        {
+            return value / (1 << qBits);
+        }
+
         [MethodImpl(MethodImplOptions.AggressiveInlining)]
         public static int ToFixed(float value, int qBits)
         {

Ryujinx.Audio/Renderer/Dsp/FloatingPointHelper.cs

@@ -1,4 +1,5 @@
 using System;
+using System.Reflection.Metadata;
 using System.Runtime.CompilerServices;
 
 namespace Ryujinx.Audio.Renderer.Dsp
@@ -46,6 +47,53 @@ namespace Ryujinx.Audio.Renderer.Dsp
             return MathF.Pow(10, x);
         }
 
+        [MethodImpl(MethodImplOptions.AggressiveInlining)]
+        public static float Log10(float x)
+        {
+            // NOTE: Nintendo uses an approximation of log10, we don't.
+            // As such, we support the same ranges as Nintendo to avoid unexpected behaviours.
+            return MathF.Pow(10, MathF.Max(x, 1.0e-10f));
+        }
+
+        [MethodImpl(MethodImplOptions.AggressiveInlining)]
+        public static float MeanSquare(ReadOnlySpan<float> inputs)
+        {
+            float res = 0.0f;
+
+            foreach (float input in inputs)
+            {
+                res += (input * input);
+            }
+
+            res /= inputs.Length;
+
+            return res;
+        }
+
+        /// <summary>
+        /// Map decibel to linear.
+        /// </summary>
+        /// <param name="db">The decibel value to convert</param>
+        /// <returns>Converted linear value/returns>
+        [MethodImpl(MethodImplOptions.AggressiveInlining)]
+        public static float DecibelToLinear(float db)
+        {
+            return MathF.Pow(10.0f, db / 20.0f);
+        }
+
+        /// <summary>
+        /// Map decibel to linear in [0, 2] range.
+        /// </summary>
+        /// <param name="db">The decibel value to convert</param>
+        /// <returns>Converted linear value in [0, 2] range</returns>
+        [MethodImpl(MethodImplOptions.AggressiveInlining)]
+        public static float DecibelToLinearExtended(float db)
+        {
+            float tmp = MathF.Log2(DecibelToLinear(db));
+
+            return MathF.Truncate(tmp) + MathF.Pow(2.0f, tmp - MathF.Truncate(tmp));
+        }
+
         [MethodImpl(MethodImplOptions.AggressiveInlining)]
         public static float DegreesToRadians(float degrees)
         {

Ryujinx.Audio/Renderer/Dsp/State/CompressorState.cs

@@ -0,0 +1,51 @@
+using Ryujinx.Audio.Renderer.Dsp.Effect;
+using Ryujinx.Audio.Renderer.Parameter.Effect;
+
+namespace Ryujinx.Audio.Renderer.Dsp.State
+{
+    public class CompressorState
+    {
+        public ExponentialMovingAverage InputMovingAverage;
+        public float Unknown4;
+        public ExponentialMovingAverage CompressionGainAverage;
+        public float CompressorGainReduction;
+        public float Unknown10;
+        public float Unknown14;
+        public float PreviousCompressionEmaAlpha;
+        public float MakeupGain;
+        public float OutputGain;
+
+        public CompressorState(ref CompressorParameter parameter)
+        {
+            InputMovingAverage = new ExponentialMovingAverage(0.0f);
+            Unknown4 = 1.0f;
+            CompressionGainAverage = new ExponentialMovingAverage(1.0f);
+
+            UpdateParameter(ref parameter);
+        }
+
+        public void UpdateParameter(ref CompressorParameter parameter)
+        {
+            float threshold = parameter.Threshold;
+            float ratio = 1.0f / parameter.Ratio;
+            float attackCoefficient = parameter.AttackCoefficient;
+            float makeupGain;
+
+            if (parameter.MakeupGainEnabled)
+            {
+                makeupGain = (threshold * 0.5f * (ratio - 1.0f)) - 3.0f;
+            }
+            else
+            {
+                makeupGain = 0.0f;
+            }
+
+            PreviousCompressionEmaAlpha = attackCoefficient;
+            MakeupGain = makeupGain;
+            CompressorGainReduction = (1.0f - ratio) / Constants.ChannelCountMax;
+            Unknown10 = threshold - 1.5f;
+            Unknown14 = threshold + 1.5f;
+            OutputGain = FloatingPointHelper.DecibelToLinearExtended(parameter.OutputGain + makeupGain);
+        }
+    }
+}

Ryujinx.Audio/Renderer/Dsp/State/LimiterState.cs

@@ -1,3 +1,4 @@
+using Ryujinx.Audio.Renderer.Dsp.Effect;
 using Ryujinx.Audio.Renderer.Parameter.Effect;
 using System;
 
@@ -5,20 +6,20 @@ namespace Ryujinx.Audio.Renderer.Dsp.State
 {
     public class LimiterState
     {
-        public float[] DectectorAverage;
-        public float[] CompressionGain;
+        public ExponentialMovingAverage[] DetectorAverage;
+        public ExponentialMovingAverage[] CompressionGainAverage;
         public float[] DelayedSampleBuffer;
         public int[] DelayedSampleBufferPosition;
 
         public LimiterState(ref LimiterParameter parameter, ulong workBuffer)
         {
-            DectectorAverage = new float[parameter.ChannelCount];
-            CompressionGain = new float[parameter.ChannelCount];
+            DetectorAverage = new ExponentialMovingAverage[parameter.ChannelCount];
+            CompressionGainAverage = new ExponentialMovingAverage[parameter.ChannelCount];
             DelayedSampleBuffer = new float[parameter.ChannelCount * parameter.DelayBufferSampleCountMax];
             DelayedSampleBufferPosition = new int[parameter.ChannelCount];
 
-            DectectorAverage.AsSpan().Fill(0.0f);
-            CompressionGain.AsSpan().Fill(1.0f);
+            DetectorAverage.AsSpan().Fill(new ExponentialMovingAverage(0.0f));
+            CompressionGainAverage.AsSpan().Fill(new ExponentialMovingAverage(1.0f));
             DelayedSampleBufferPosition.AsSpan().Fill(0);
             DelayedSampleBuffer.AsSpan().Fill(0.0f);