[AMDGPU] add macro __AMDGCN_CDNA_VERSION__

If a processor belongs to CDNA generation, pre-define macro
`__AMDGCN_CDNA_VERSION__` as an integer.

Fixes: ROCm#59

Author: yxsamliu

clang/docs/AMDGPUSupport.rst

@@ -45,6 +45,8 @@ Predefined Macros
      - Defined with the target ID as a string.
    * - ``__amdgcn_feature_<feature-name>__``
      - Defined for each supported target feature. The value is 1 if the feature is enabled and 0 if it is disabled. Allowed feature names are sramecc and xnack.
+   * - ``__AMDGCN_CDNA_VERSION__``
+     - Defined with the CDNA version as an integer if the processor belongs to the CDNA generation.
    * - ``__AMDGCN_CUMODE__``
      - Defined as 1 if the CU mode is enabled and 0 if the WGP mode is enabled.
    * - ``__AMDGCN_UNSAFE_FP_ATOMICS__``

clang/lib/Basic/Targets/AMDGPU.cpp

@@ -249,6 +249,22 @@ AMDGPUTargetInfo::AMDGPUTargetInfo(const llvm::Triple &Triple,
   for (auto F : {"image-insts", "gws"})
     ReadOnlyFeatures.insert(F);
   HalfArgsAndReturns = true;
+
+  switch (GPUKind) {
+  case llvm::AMDGPU::GK_GFX908:
+    CDNAVersion = 1;
+    break;
+  case llvm::AMDGPU::GK_GFX90A:
+    CDNAVersion = 2;
+    break;
+  case llvm::AMDGPU::GK_GFX940:
+  case llvm::AMDGPU::GK_GFX941:
+  case llvm::AMDGPU::GK_GFX942:
+    CDNAVersion = 3;
+    break;
+  default:
+    CDNAVersion = 0;
+  }
 }
 
 void AMDGPUTargetInfo::adjust(DiagnosticsEngine &Diags, LangOptions &Opts) {
@@ -299,6 +315,8 @@ void AMDGPUTargetInfo::getTargetDefines(const LangOptions &Opts,
     StringRef CanonFamilyName = getArchFamilyNameAMDGCN(GPUKind);
     Builder.defineMacro(Twine("__") + Twine(CanonFamilyName.upper()) +
                         Twine("__"));
+    if (CDNAVersion)
+      Builder.defineMacro("__AMDGCN_CDNA_VERSION__", Twine(CDNAVersion));
     Builder.defineMacro("__amdgcn_processor__",
                         Twine("\"") + Twine(CanonName) + Twine("\""));
     Builder.defineMacro("__amdgcn_target_id__",

clang/lib/Basic/Targets/AMDGPU.h

@@ -51,6 +51,8 @@ class LLVM_LIBRARY_VISIBILITY AMDGPUTargetInfo final : public TargetInfo {
   llvm::StringMap<bool> OffloadArchFeatures;
   std::string TargetID;
 
+  unsigned CDNAVersion = 0;
+
   bool hasFP64() const {
     return getTriple().getArch() == llvm::Triple::amdgcn ||
            !!(GPUFeatures & llvm::AMDGPU::FEATURE_FP64);

clang/test/Driver/amdgpu-macros.cl

@@ -103,14 +103,14 @@
 // RUN: %clang -E -dM -target amdgcn -mcpu=gfx902 %s 2>&1 | FileCheck --check-prefixes=ARCH-GCN,FAST_FMAF %s -DWAVEFRONT_SIZE=64 -DCPU=gfx902 -DFAMILY=GFX9
 // RUN: %clang -E -dM -target amdgcn -mcpu=gfx904 %s 2>&1 | FileCheck --check-prefixes=ARCH-GCN,FAST_FMAF %s -DWAVEFRONT_SIZE=64 -DCPU=gfx904 -DFAMILY=GFX9
 // RUN: %clang -E -dM -target amdgcn -mcpu=gfx906 %s 2>&1 | FileCheck --check-prefixes=ARCH-GCN,FAST_FMAF %s -DWAVEFRONT_SIZE=64 -DCPU=gfx906 -DFAMILY=GFX9
-// RUN: %clang -E -dM -target amdgcn -mcpu=gfx908 %s 2>&1 | FileCheck --check-prefixes=ARCH-GCN,FAST_FMAF %s -DWAVEFRONT_SIZE=64 -DCPU=gfx908 -DFAMILY=GFX9
+// RUN: %clang -E -dM -target amdgcn -mcpu=gfx908 %s 2>&1 | FileCheck --check-prefixes=ARCH-GCN,CDNA,FAST_FMAF %s -DWAVEFRONT_SIZE=64 -DCPU=gfx908 -DFAMILY=GFX9 -DCDNA=1
 // RUN: %clang -E -dM -target amdgcn -mcpu=gfx908 -munsafe-fp-atomics %s 2>&1 | FileCheck --check-prefixes=ARCH-GCN,UNSAFEFPATOMIC %s -DWAVEFRONT_SIZE=64 -DCPU=gfx908 -DFAMILY=GFX9
 // RUN: %clang -E -dM -target amdgcn -mcpu=gfx909 %s 2>&1 | FileCheck --check-prefixes=ARCH-GCN,FAST_FMAF %s -DWAVEFRONT_SIZE=64 -DCPU=gfx909 -DFAMILY=GFX9
-// RUN: %clang -E -dM -target amdgcn -mcpu=gfx90a %s 2>&1 | FileCheck --check-prefixes=ARCH-GCN,FAST_FMAF %s -DWAVEFRONT_SIZE=64 -DCPU=gfx90a -DFAMILY=GFX9
+// RUN: %clang -E -dM -target amdgcn -mcpu=gfx90a %s 2>&1 | FileCheck --check-prefixes=ARCH-GCN,CDNA,FAST_FMAF %s -DWAVEFRONT_SIZE=64 -DCPU=gfx90a -DFAMILY=GFX9 -DCDNA=2
 // RUN: %clang -E -dM -target amdgcn -mcpu=gfx90c %s 2>&1 | FileCheck --check-prefixes=ARCH-GCN,FAST_FMAF %s -DWAVEFRONT_SIZE=64 -DCPU=gfx90c -DFAMILY=GFX9
-// RUN: %clang -E -dM -target amdgcn -mcpu=gfx940 %s 2>&1 | FileCheck --check-prefixes=ARCH-GCN,FAST_FMAF %s -DWAVEFRONT_SIZE=64 -DCPU=gfx940 -DFAMILY=GFX9
-// RUN: %clang -E -dM -target amdgcn -mcpu=gfx941 %s 2>&1 | FileCheck --check-prefixes=ARCH-GCN,FAST_FMAF %s -DWAVEFRONT_SIZE=64 -DCPU=gfx941 -DFAMILY=GFX9
-// RUN: %clang -E -dM -target amdgcn -mcpu=gfx942 %s 2>&1 | FileCheck --check-prefixes=ARCH-GCN,FAST_FMAF %s -DWAVEFRONT_SIZE=64 -DCPU=gfx942 -DFAMILY=GFX9
+// RUN: %clang -E -dM -target amdgcn -mcpu=gfx940 %s 2>&1 | FileCheck --check-prefixes=ARCH-GCN,CDNA,FAST_FMAF %s -DWAVEFRONT_SIZE=64 -DCPU=gfx940 -DFAMILY=GFX9 -DCDNA=3
+// RUN: %clang -E -dM -target amdgcn -mcpu=gfx941 %s 2>&1 | FileCheck --check-prefixes=ARCH-GCN,CDNA,FAST_FMAF %s -DWAVEFRONT_SIZE=64 -DCPU=gfx941 -DFAMILY=GFX9 -DCDNA=3
+// RUN: %clang -E -dM -target amdgcn -mcpu=gfx942 %s 2>&1 | FileCheck --check-prefixes=ARCH-GCN,CDNA,FAST_FMAF %s -DWAVEFRONT_SIZE=64 -DCPU=gfx942 -DFAMILY=GFX9 -DCDNA=3
 // RUN: %clang -E -dM -target amdgcn -mcpu=gfx1010 %s 2>&1 | FileCheck --check-prefixes=ARCH-GCN,FAST_FMAF %s -DWAVEFRONT_SIZE=32 -DCPU=gfx1010 -DFAMILY=GFX10
 // RUN: %clang -E -dM -target amdgcn -mcpu=gfx1011 %s 2>&1 | FileCheck --check-prefixes=ARCH-GCN,FAST_FMAF %s -DWAVEFRONT_SIZE=32 -DCPU=gfx1011 -DFAMILY=GFX10
 // RUN: %clang -E -dM -target amdgcn -mcpu=gfx1012 %s 2>&1 | FileCheck --check-prefixes=ARCH-GCN,FAST_FMAF %s -DWAVEFRONT_SIZE=32 -DCPU=gfx1012 -DFAMILY=GFX10
@@ -150,6 +150,7 @@
 // ARCH-GCN-DAG: #define __[[CPU]]__ 1
 // ARCH-GCN-DAG: #define __[[FAMILY]]__ 1
 // ARCH-GCN-DAG: #define __amdgcn_processor__ "[[CPU]]"
+// CDNA-DAG: #define __AMDGCN_CDNA_VERSION__ [[CDNA]]
 // ARCH-GCN-DAG: #define __AMDGCN_WAVEFRONT_SIZE [[WAVEFRONT_SIZE]]
 // UNSAFEFPATOMIC-DAG: #define __AMDGCN_UNSAFE_FP_ATOMICS__ 1