[clang] Change AMX macros to match names from GCC

[jloser]

Summary

The current behavior for AMX macros is:

gcc -march=native -dM -E - < /dev/null | grep TILE
#define __AMX_TILE__ 1

clang -march=native -dM -E - < /dev/null | grep TILE
#define __AMXTILE__ 1

which is not ideal. Change __AMXTILE__ and friends to __AMX_TILE__ (i.e.
have an underscore in them). This makes GCC and Clang agree on the naming of
these AMX macros to simplify downstream user code.

Fix this for __AMXTILE__, __AMX_INT8__, __AMX_BF16__, and __AMX_FP16__.

[jloser]

Update commit message to properly display the #define

[LuoYuanke]

LGTM, thanks

[jloser]

In D143094#4098208, @LuoYuanke wrote:

LGTM, thanks

Thanks for the review. Do I need to wait for another approver before landing this? I mostly work on libc++, so not too familiar with expectations on Clang reviews, or whether one approver who's familiar/knowledgable in the area is sufficient.

I'm hoping to land this soon and request to cherry-pick it into LLVM 16 to avoid the macro value changing across 16/17 releases.

[LuoYuanke]

@pengfei, would you take a look to double confirm?

[pengfei]

The change makes sense. LGTM.

clang/lib/Basic/Targets/X86.cpp

if (HasPTWRITE)
     Builder.defineMacro("__PTWRITE__");
   if (HasINVPCID)
     Builder.defineMacro("__INVPCID__");
   if (HasENQCMD)
     Builder.defineMacro("__ENQCMD__");
   if (HasHRESET)
     Builder.defineMacro("__HRESET__");
   if (HasAMXTILE)
-    Builder.defineMacro("__AMXTILE__");
+    Builder.defineMacro("__AMX_TILE__");
   if (HasAMXINT8)
-    Builder.defineMacro("__AMXINT8__");
+    Builder.defineMacro("__AMX_INT8__");
   if (HasAMXBF16)
-    Builder.defineMacro("__AMXBF16__");
+    Builder.defineMacro("__AMX_BF16__");
   if (HasAMXFP16)
-    Builder.defineMacro("__AMXFP16__");
+    Builder.defineMacro("__AMX_FP16__");
   if (HasCMPCCXADD)
     Builder.defineMacro("__CMPCCXADD__");
   if (HasRAOINT)
     Builder.defineMacro("__RAOINT__");
   if (HasAVXIFMA)
     Builder.defineMacro("__AVXIFMA__");
   if (HasAVXNECONVERT)
     Builder.defineMacro("__AVXNECONVERT__");

clang/lib/Headers/immintrin.h

 #include <ptwriteintrin.h>
 #endif
 
 #if !(defined(_MSC_VER) || defined(__SCE__)) || __has_feature(modules) ||      \
     defined(__INVPCID__)
 #include <invpcidintrin.h>
 #endif
 #if !(defined(_MSC_VER) || defined(__SCE__)) || __has_feature(modules) ||      \
-    defined(__AMXFP16__)
+    defined(__AMX_FP16__)
 #include <amxfp16intrin.h>
 #endif
 
 #if !(defined(_MSC_VER) || defined(__SCE__)) || __has_feature(modules) ||      \
     defined(__KL__) || defined(__WIDEKL__)
 #include <keylockerintrin.h>
 #endif
 
 #if !(defined(_MSC_VER) || defined(__SCE__)) || __has_feature(modules) ||      \
-    defined(__AMXTILE__) || defined(__AMXINT8__) || defined(__AMXBF16__)
+    defined(__AMX_TILE__) || defined(__AMX_INT8__) || defined(__AMX_BF16__)
 #include <amxintrin.h>
 #endif
 
 #if !(defined(_MSC_VER) || defined(__SCE__)) || __has_feature(modules) ||      \
     defined(__AVX512VP2INTERSECT__)
 #include <avx512vp2intersectintrin.h>
 #endif
 

clang/test/Preprocessor/predefined-arch-macros.c

 
 // RUN: %clang -march=sapphirerapids -m32 -E -dM %s -o - 2>&1 \
 // RUN:     -target i386-unknown-linux \
 // RUN:   | FileCheck -match-full-lines %s -check-prefix=CHECK_SPR_M32
 // RUN: %clang -march=emeraldrapids -m32 -E -dM %s -o - 2>&1 \
 // RUN:     -target i386-unknown-linux \
 // RUN:   | FileCheck -match-full-lines %s -check-prefix=CHECK_SPR_M32
 // CHECK_SPR_M32: #define __AES__ 1
-// CHECK_SPR_M32: #define __AMXBF16__ 1
-// CHECK_SPR_M32: #define __AMXINT8__ 1
-// CHECK_SPR_M32: #define __AMXTILE__ 1
+// CHECK_SPR_M32: #define __AMX_BF16__ 1
+// CHECK_SPR_M32: #define __AMX_INT8__ 1
+// CHECK_SPR_M32: #define __AMX_TILE__ 1
 // CHECK_SPR_M32: #define __AVX2__ 1
 // CHECK_SPR_M32: #define __AVX512BF16__ 1
 // CHECK_SPR_M32: #define __AVX512BITALG__ 1
 // CHECK_SPR_M32: #define __AVX512BW__ 1
 // CHECK_SPR_M32: #define __AVX512CD__ 1
 // CHECK_SPR_M32: #define __AVX512DQ__ 1
 // CHECK_SPR_M32: #define __AVX512FP16__ 1
 // CHECK_SPR_M32: #define __AVX512F__ 1

 
 // RUN: %clang -march=sapphirerapids -m64 -E -dM %s -o - 2>&1 \
 // RUN:     -target i386-unknown-linux \
 // RUN:   | FileCheck -match-full-lines %s -check-prefix=CHECK_SPR_M64
 // RUN: %clang -march=emeraldrapids -m64 -E -dM %s -o - 2>&1 \
 // RUN:     -target i386-unknown-linux \
 // RUN:   | FileCheck -match-full-lines %s -check-prefix=CHECK_SPR_M64
 // CHECK_SPR_M64: #define __AES__ 1
-// CHECK_SPR_M64: #define __AMXBF16__ 1
-// CHECK_SPR_M64: #define __AMXINT8__ 1
-// CHECK_SPR_M64: #define __AMXTILE__ 1
+// CHECK_SPR_M64: #define __AMX_BF16__ 1
+// CHECK_SPR_M64: #define __AMX_INT8__ 1
+// CHECK_SPR_M64: #define __AMX_TILE__ 1
 // CHECK_SPR_M64: #define __AVX2__ 1
 // CHECK_SPR_M64: #define __AVX512BF16__ 1
 // CHECK_SPR_M64: #define __AVX512BITALG__ 1
 // CHECK_SPR_M64: #define __AVX512BW__ 1
 // CHECK_SPR_M64: #define __AVX512CD__ 1
 // CHECK_SPR_M64: #define __AVX512DQ__ 1
 // CHECK_SPR_M64: #define __AVX512FP16__ 1
 // CHECK_SPR_M64: #define __AVX512F__ 1

 // CHECK_SPR_M64: #define __tune_corei7__ 1
 // CHECK_SPR_M64: #define __x86_64 1
 // CHECK_SPR_M64: #define __x86_64__ 1
 
 // RUN: %clang -march=graniterapids -m32 -E -dM %s -o - 2>&1 \
 // RUN:     --target=i386 \
 // RUN:   | FileCheck -match-full-lines %s -check-prefix=CHECK_GNR_M32
 // CHECK_GNR_M32: #define __AES__ 1
-// CHECK_GNR_M32: #define __AMXBF16__ 1
-// CHECK_GNR_M32: #define __AMXFP16__ 1
-// CHECK_GNR_M32: #define __AMXINT8__ 1
-// CHECK_GNR_M32: #define __AMXTILE__ 1
+// CHECK_GNR_M32: #define __AMX_BF16__ 1
+// CHECK_GNR_M32: #define __AMX_FP16__ 1
+// CHECK_GNR_M32: #define __AMX_INT8__ 1
+// CHECK_GNR_M32: #define __AMX_TILE__ 1
 // CHECK_GNR_M32: #define __AVX2__ 1
 // CHECK_GNR_M32: #define __AVX512BF16__ 1
 // CHECK_GNR_M32: #define __AVX512BITALG__ 1
 // CHECK_GNR_M32: #define __AVX512BW__ 1
 // CHECK_GNR_M32: #define __AVX512CD__ 1
 // CHECK_GNR_M32: #define __AVX512DQ__ 1
 // CHECK_GNR_M32: #define __AVX512FP16__ 1
 // CHECK_GNR_M32: #define __AVX512F__ 1

 // CHECK_GNR_M32: #define __i386__ 1
 // CHECK_GNR_M32: #define __tune_corei7__ 1
 // CHECK_GNR_M32: #define i386 1
 
 // RUN: %clang -march=graniterapids -m64 -E -dM %s -o - 2>&1 \
 // RUN:     --target=x86_64 \
 // RUN:   | FileCheck -match-full-lines %s -check-prefix=CHECK_GNR_M64
 // CHECK_GNR_M64: #define __AES__ 1
-// CHECK_GNR_M64: #define __AMXBF16__ 1
-// CHECK_GNR_M64: #define __AMXFP16__ 1
-// CHECK_GNR_M64: #define __AMXINT8__ 1
-// CHECK_GNR_M64: #define __AMXTILE__ 1
+// CHECK_GNR_M64: #define __AMX_BF16__ 1
+// CHECK_GNR_M64: #define __AMX_FP16__ 1
+// CHECK_GNR_M64: #define __AMX_INT8__ 1
+// CHECK_GNR_M64: #define __AMX_TILE__ 1
 // CHECK_GNR_M64: #define __AVX2__ 1
 // CHECK_GNR_M64: #define __AVX512BF16__ 1
 // CHECK_GNR_M64: #define __AVX512BITALG__ 1
 // CHECK_GNR_M64: #define __AVX512BW__ 1
 // CHECK_GNR_M64: #define __AVX512CD__ 1
 // CHECK_GNR_M64: #define __AVX512DQ__ 1
 // CHECK_GNR_M64: #define __AVX512FP16__ 1
 // CHECK_GNR_M64: #define __AVX512F__ 1

clang/test/Preprocessor/x86_amx_target_features.c

 // RUN: %clang -target i386-unknown-linux-gnu -march=i386 -mamx-tile -x c -E -dM -o - %s | FileCheck  -check-prefix=AMX-TILE %s
 
-// AMX-TILE: #define __AMXTILE__ 1
+// AMX-TILE: #define __AMX_TILE__ 1
 
 // RUN: %clang -target i386-unknown-linux-gnu -march=i386 -mamx-bf16 -x c -E -dM -o - %s | FileCheck -check-prefix=AMX-BF16 %s
 
-// AMX-BF16: #define __AMXBF16__ 1
-// AMX-BF16: #define __AMXTILE__ 1
+// AMX-BF16: #define __AMX_BF16__ 1
+// AMX-BF16: #define __AMX_TILE__ 1
 
 // RUN: %clang -target i386-unknown-linux-gnu -march=i386 -mamx-int8 -x c -E -dM -o - %s | FileCheck -check-prefix=AMX-INT8 %s
 
-// AMX-INT8: #define __AMXINT8__ 1
-// AMX-INT8: #define __AMXTILE__ 1
+// AMX-INT8: #define __AMX_INT8__ 1
+// AMX-INT8: #define __AMX_TILE__ 1
 
 // RUN: %clang -target i386-unknown-linux-gnu -march=i386 -mno-amx-tile -x c -E -dM -o - %s | FileCheck  -check-prefix=NOAMX-TILE %s
 
-// NOAMX-TILE-NOT: #define __AMXTILE__ 1
+// NOAMX-TILE-NOT: #define __AMX_TILE__ 1
 
 // RUN: %clang -target i386-unknown-linux-gnu -march=i386 -mno-amx-bf16 -x c -E -dM -o - %s | FileCheck  -check-prefix=NOAMX-BF16 %s
 
-// NOAMX-BF16-NOT: #define __AMXBF16__ 1
+// NOAMX-BF16-NOT: #define __AMX_BF16__ 1
 
 // RUN: %clang -target i386-unknown-linux-gnu -march=i386 -mamx-bf16 -mno-amx-tile -x c -E -dM -o - %s | FileCheck  -check-prefix=NOAMX-BF16 %s
 
-// NOAMX-BF16-NOT: #define __AMXTILE__ 1
-// NOAMX-BF16-NOT: #define __AMXBF16__ 1
+// NOAMX-BF16-NOT: #define __AMX_TILE__ 1
+// NOAMX-BF16-NOT: #define __AMX_BF16__ 1
 
 // RUN: %clang -target i386-unknown-linux-gnu -march=i386 -mno-amx-int8 -x c -E -dM -o - %s | FileCheck  -check-prefix=NOAMX-INT8 %s
 
-// NOAMX-INT8-NOT: #define __AMXINT8__ 1
+// NOAMX-INT8-NOT: #define __AMX_INT8__ 1
 
 // RUN: %clang -target i386-unknown-linux-gnu -march=i386 -mamx-int8 -mno-amx-tile -x c -E -dM -o - %s | FileCheck  -check-prefix=NOAMX-INT8 %s
 
-// NOAMX-INT8-NOT: #define __AMXTILE__ 1
-// NOAMX-INT8-NOT: #define __AMXINT8__ 1
+// NOAMX-INT8-NOT: #define __AMX_TILE__ 1
+// NOAMX-INT8-NOT: #define __AMX_INT8__ 1

clang/test/Preprocessor/x86_target_features.c

 
 // RUN: %clang -target i386-unknown-linux-gnu -march=i386 -mno-uintr -x c -E -dM -o - %s | FileCheck -check-prefix=NOUINTR %s
 
 // NOUINTR-NOT: #define __UINTR__ 1
 
 // RUN: %clang -target x86_64-unknown-linux-gnu -march=atom -mamx-fp16 -x c \
 // RUN: -E -dM -o - %s | FileCheck  -check-prefix=AMX-FP16 %s
 
-// AMX-FP16: #define __AMXFP16__ 1
-// AMX-FP16: #define __AMXTILE__ 1
+// AMX-FP16: #define __AMX_FP16__ 1
+// AMX-FP16: #define __AMX_TILE__ 1
 
 // RUN: %clang -target x86_64-unknown-linux-gnu -march=atom -mno-amx-fp16 \
 // RUN: -x c -E -dM -o - %s | FileCheck  -check-prefix=NO-AMX-FP16 %s
 // RUN: %clang -target x86_64-unknown-linux-gnu -march=atom -mamx-fp16 \
 // RUN: -mno-amx-tile -x c -E -dM -o - %s | FileCheck  -check-prefix=NO-AMX-FP16 %s
 
-// NO-AMX-FP16-NOT: #define __AMXFP16__ 1
-// NO-AMX-FP16-NOT: #define __AMXTILE__ 1
+// NO-AMX-FP16-NOT: #define __AMX_FP16__ 1
+// NO-AMX-FP16-NOT: #define __AMX_TILE__ 1
 
 // RUN: %clang -target i386-unknown-unknown -march=atom -mavxvnni -x c -E -dM -o - %s | FileCheck -match-full-lines --check-prefix=AVXVNNI %s
 
 // AVXVNNI: #define __AVX2__ 1
 // AVXVNNI: #define __AVXVNNI__ 1
 
 // RUN: %clang -target i386-unknown-unknown -march=atom -mno-avxvnni -x c -E -dM -o - %s | FileCheck -match-full-lines --check-prefix=NOAVXVNNI %s