Fix input size flag on vpermd. (#106054)

* Fix input size flag on `vpermd`. * Update src/tests/JIT/Regression/JitBlue/Runtime_105824/Runtime_105824.cs Co-authored-by: Tanner Gooding <[email protected]> * Update src/tests/JIT/Regression/JitBlue/Runtime_105824/Runtime_105824.cs Co-authored-by: Tanner Gooding <[email protected]> --------- Co-authored-by: Tanner Gooding <[email protected]>
dotnet · Aug 8, 2024 · 019d758 · 019d758
1 parent a562a9f
commit 019d758
Show file tree

Hide file tree

Showing 3 changed files with 51 additions and 1 deletion.
diff --git a/src/coreclr/jit/instrsxarch.h b/src/coreclr/jit/instrsxarch.h
@@ -494,7 +494,7 @@ INST3(vpbroadcastd,     "pbroadcastd",      IUM_WR, BAD_CODE,     BAD_CODE,
 INST3(vpbroadcastq,     "pbroadcastq",      IUM_WR, BAD_CODE,     BAD_CODE,     SSE38(0x59),                             INS_TT_TUPLE1_SCALAR,                Input_64Bit    | REX_WX       | Encoding_VEX  | Encoding_EVEX)                                                                                                                                  // Broadcast int64 value from reg/memory to entire ymm register
 INST3(vpbroadcastw,     "pbroadcastw",      IUM_WR, BAD_CODE,     BAD_CODE,     SSE38(0x79),                             INS_TT_TUPLE1_SCALAR,                Input_16Bit    | REX_W0       | Encoding_VEX  | Encoding_EVEX)                                                                                                                                  // Broadcast int16 value from reg/memory to entire ymm register
 INST3(vperm2i128,       "perm2i128",        IUM_WR, BAD_CODE,     BAD_CODE,     SSE3A(0x46),                             INS_TT_NONE,                                          REX_W0       | Encoding_VEX                   | INS_Flags_IsDstDstSrcAVXInstruction)                                                                                           // Permute 128-bit halves of input register
-INST3(vpermd,           "permd",            IUM_WR, BAD_CODE,     BAD_CODE,     SSE38(0x36),                             INS_TT_FULL,                         Input_64Bit    | REX_W0       | Encoding_VEX  | Encoding_EVEX  | INS_Flags_IsDstDstSrcAVXInstruction | INS_Flags_EmbeddedBroadcastSupported)                                                    // Permute Packed Doublewords Elements
+INST3(vpermd,           "permd",            IUM_WR, BAD_CODE,     BAD_CODE,     SSE38(0x36),                             INS_TT_FULL,                         Input_32Bit    | REX_W0       | Encoding_VEX  | Encoding_EVEX  | INS_Flags_IsDstDstSrcAVXInstruction | INS_Flags_EmbeddedBroadcastSupported)                                                    // Permute Packed Doublewords Elements
 INST3(vpermpd,          "permpd",           IUM_WR, BAD_CODE,     BAD_CODE,     SSE3A(0x01),                             INS_TT_FULL,                         Input_64Bit    | REX_W1       | Encoding_VEX  | Encoding_EVEX                                        | INS_Flags_EmbeddedBroadcastSupported)                                                    // Permute Double-Precision Floating-Point Values
 INST3(vpermps,          "permps",           IUM_WR, BAD_CODE,     BAD_CODE,     SSE38(0x16),                             INS_TT_FULL,                         Input_32Bit    | REX_W0       | Encoding_VEX  | Encoding_EVEX  | INS_Flags_IsDstDstSrcAVXInstruction | INS_Flags_EmbeddedBroadcastSupported)                                                    // Permute Single-Precision Floating-Point Elements
 INST3(vpermq,           "permq",            IUM_WR, BAD_CODE,     BAD_CODE,     SSE3A(0x00),                             INS_TT_FULL,                         Input_64Bit    | REX_W1       | Encoding_VEX  | Encoding_EVEX                                        | INS_Flags_EmbeddedBroadcastSupported)                                                    // Permute 64-bit of input register

diff --git a/src/tests/JIT/Regression/JitBlue/Runtime_105824/Runtime_105824.cs b/src/tests/JIT/Regression/JitBlue/Runtime_105824/Runtime_105824.cs
@@ -0,0 +1,42 @@
+// Licensed to the .NET Foundation under one or more agreements.
+// The .NET Foundation licenses this file to you under the MIT license.
+
+using System;
+using System.Runtime.CompilerServices;
+using System.Numerics;
+using System.Runtime.Intrinsics;
+using System.Runtime.Intrinsics.X86;
+using Xunit;
+
+// Generated by Fuzzlyn v2.2 on 2024-08-01 15:01:31
+// Run on X64 Windows
+// Seed: 8577405969008594479-vectort,vector128,vector256,x86aes,x86avx,x86avx2,x86avx512bw,x86avx512bwvl,x86avx512cd,x86avx512cdvl,x86avx512dq,x86avx512dqvl,x86avx512f,x86avx512fvl,x86avx512fx64,x86bmi1,x86bmi1x64,x86bmi2,x86bmi2x64,x86fma,x86lzcnt,x86lzcntx64,x86pclmulqdq,x86popcnt,x86popcntx64,x86sse,x86ssex64,x86sse2,x86sse2x64,x86sse3,x86sse41,x86sse41x64,x86sse42,x86sse42x64,x86ssse3,x86x86base
+// Reduced from 43.4 KiB to 0.7 KiB in 00:01:17
+// Debug: Outputs -1
+// Release: Outputs 0
+
+public class Runtime_105824
+{
+    [Fact]
+    public static void TestEntryPoint()
+    {
+        if (Avx2.IsSupported)
+        {
+            uint[][] vr6 = new uint[][]
+            {
+                new uint[]
+                {
+                    0
+                }
+            };
+            var vr7 = Vector256.Create<uint>(0);
+            var vr8 = vr6[0][0];
+            var vr9 = Vector256.Create<uint>(vr8);
+            var vr10 = Vector256.CreateScalar(0U);
+            var vr11 = Avx2.PermuteVar8x32(vr9, vr10);
+            var vr12 = Avx2.CompareEqual(vr7, vr11);
+            short vr13 = (short)Avx2.ConvertToUInt32(vr12);
+            Assert.Equal(-1, vr13);
+        }
+    }
+}
diff --git a/src/tests/JIT/Regression/JitBlue/Runtime_105824/Runtime_105824.csproj b/src/tests/JIT/Regression/JitBlue/Runtime_105824/Runtime_105824.csproj
@@ -0,0 +1,8 @@
+<Project Sdk="Microsoft.NET.Sdk">
+  <PropertyGroup>
+    <Optimize>True</Optimize>
+  </PropertyGroup>
+  <ItemGroup>
+    <Compile Include="$(MSBuildProjectName).cs" />
+  </ItemGroup>
+</Project>