[X86] Add support for avx512bf16 for __builtin_cpu_supports and compiler-rt's cpu...

llvm-svn: 370915

diff --git a/clang/test/CodeGen/target-builtin-noerror.c b/clang/test/CodeGen/target-builtin-noerror.c
index 364eae7..37e9b50 100644 (file)
--- a/clang/test/CodeGen/target-builtin-noerror.c
+++ b/clang/test/CodeGen/target-builtin-noerror.c
@@ -80,6 +80,7 @@ void verifyfeaturestrings() {
   (void)__builtin_cpu_supports("vpclmulqdq");
   (void)__builtin_cpu_supports("avx512vnni");
   (void)__builtin_cpu_supports("avx512bitalg");
+  (void)__builtin_cpu_supports("avx512bf16");
 }
 
 void verifycpustrings() {

diff --git a/compiler-rt/lib/builtins/cpu_model.c b/compiler-rt/lib/builtins/cpu_model.c
index b37fdba..cdeb037 100644 (file)
--- a/compiler-rt/lib/builtins/cpu_model.c
+++ b/compiler-rt/lib/builtins/cpu_model.c
@@ -121,7 +121,8 @@ enum ProcessorFeatures {
   FEATURE_GFNI,
   FEATURE_VPCLMULQDQ,
   FEATURE_AVX512VNNI,
-  FEATURE_AVX512BITALG
+  FEATURE_AVX512BITALG,
+  FEATURE_AVX512BF16
 };
 
 // The check below for i386 was copied from clang's cpuid.h (__get_cpuid_max).
@@ -582,6 +583,11 @@ static void getAvailableFeatures(unsigned ECX, unsigned EDX, unsigned MaxLeaf,
   if (HasLeaf7 && ((EDX >> 3) & 1) && HasAVX512Save)
     setFeature(FEATURE_AVX5124FMAPS);
 
+  bool HasLeaf7Subleaf1 =
+      MaxLeaf >= 0x7 && !getX86CpuIDAndInfoEx(0x7, 0x1, &EAX, &EBX, &ECX, &EDX);
+  if (HasLeaf7Subleaf1 && ((EAX >> 5) & 1) && HasAVX512Save)
+    setFeature(FEATURE_AVX512BF16);
+
   unsigned MaxExtLevel;
   getX86CpuIDAndInfo(0x80000000, &MaxExtLevel, &EBX, &ECX, &EDX);
 

diff --git a/llvm/include/llvm/Support/X86TargetParser.def b/llvm/include/llvm/Support/X86TargetParser.def
index a4ed8ee..4ebf2d7 100644 (file)
--- a/llvm/include/llvm/Support/X86TargetParser.def
+++ b/llvm/include/llvm/Support/X86TargetParser.def
@@ -161,13 +161,13 @@ X86_FEATURE_COMPAT(32, FEATURE_GFNI,            "gfni")
 X86_FEATURE_COMPAT(33, FEATURE_VPCLMULQDQ,      "vpclmulqdq")
 X86_FEATURE_COMPAT(34, FEATURE_AVX512VNNI,      "avx512vnni")
 X86_FEATURE_COMPAT(35, FEATURE_AVX512BITALG,    "avx512bitalg")
+X86_FEATURE_COMPAT(36, FEATURE_AVX512BF16,      "avx512bf16")
 // Features below here are not in libgcc/compiler-rt.
 X86_FEATURE       (64, FEATURE_MOVBE)
 X86_FEATURE       (65, FEATURE_ADX)
 X86_FEATURE       (66, FEATURE_EM64T)
 X86_FEATURE       (67, FEATURE_CLFLUSHOPT)
 X86_FEATURE       (68, FEATURE_SHA)
-X86_FEATURE       (69, FEATURE_AVX512BF16)
-X86_FEATURE       (70, FEATURE_AVX512VP2INTERSECT)
+X86_FEATURE       (69, FEATURE_AVX512VP2INTERSECT)
 #undef X86_FEATURE_COMPAT
 #undef X86_FEATURE

diff --git a/llvm/lib/Support/Host.cpp b/llvm/lib/Support/Host.cpp
index ef10b96..0d6fab5 100644 (file)
--- a/llvm/lib/Support/Host.cpp
+++ b/llvm/lib/Support/Host.cpp
@@ -680,7 +680,7 @@ getIntelProcessorTypeAndSubtype(unsigned Family, unsigned Model,
     // Skylake Xeon:
     case 0x55:
       *Type = X86::INTEL_COREI7;
-      if (Features3 & (1 << (X86::FEATURE_AVX512BF16 - 64)))
+      if (Features2 & (1 << (X86::FEATURE_AVX512BF16 - 32)))
         *Subtype = X86::INTEL_COREI7_COOPERLAKE; // "cooperlake"
       else if (Features2 & (1 << (X86::FEATURE_AVX512VNNI - 32)))
         *Subtype = X86::INTEL_COREI7_CASCADELAKE; // "cascadelake"
@@ -765,7 +765,7 @@ getIntelProcessorTypeAndSubtype(unsigned Family, unsigned Model,
         break;
       }
 
-      if (Features3 & (1 << (X86::FEATURE_AVX512BF16 - 64))) {
+      if (Features2 & (1 << (X86::FEATURE_AVX512BF16 - 32))) {
         *Type = X86::INTEL_COREI7;
         *Subtype = X86::INTEL_COREI7_COOPERLAKE;
         break;
@@ -1088,6 +1088,11 @@ static void getAvailableFeatures(unsigned ECX, unsigned EDX, unsigned MaxLeaf,
   if (HasLeaf7 && ((EDX >> 8) & 1) && HasAVX512Save)
     setFeature(X86::FEATURE_AVX512VP2INTERSECT);
 
+  bool HasLeaf7Subleaf1 =
+      MaxLeaf >= 7 && !getX86CpuIDAndInfoEx(0x7, 0x1, &EAX, &EBX, &ECX, &EDX);
+  if (HasLeaf7Subleaf1 && ((EAX >> 5) & 1) && HasAVX512Save)
+    setFeature(X86::FEATURE_AVX512BF16);
+
   unsigned MaxExtLevel;
   getX86CpuIDAndInfo(0x80000000, &MaxExtLevel, &EBX, &ECX, &EDX);