Merging r258901:

------------------------------------------------------------------------
r258901 | Matthew.Arsenault | 2016-01-26 18:17:49 -0800 (Tue, 26 Jan 2016) | 17 lines

AMDGPU: Fix default device handling

When no device name is specified, default to kaveri
for HSA since SI is not supported and it woud fail.

Default to "tahiti" instead of "SI" since these are
effectively the same, and tahiti is an actual device.

Move default device handling to the TargetMachine
rather than the AMDGPUSubtarget. The module ISA version
is computed from the device name provided with the target
machine, so the attributes printed by the AsmPrinter were
inconsistent with those computed in the subtarget.

Also remove DevName field from subtarget since it's redundant
with getCPU() in the superclass.

------------------------------------------------------------------------

git-svn-id: https://llvm.org/svn/llvm-project/llvm/branches/release_38@271588 91177308-0d34-0410-b5e6-96231b3b80d8

diff --git a/lib/Target/AMDGPU/AMDGPUSubtarget.cpp b/lib/Target/AMDGPU/AMDGPUSubtarget.cpp
index 1f51b554bd5dbe5c46dce0031c4807ae11ae8b9c..9dd86a25b2d6dd1bdaef07bc2dc70fc542e4f4f3 100644 (file)
--- a/lib/Target/AMDGPU/AMDGPUSubtarget.cpp
+++ b/lib/Target/AMDGPU/AMDGPUSubtarget.cpp
@@ -49,9 +49,6 @@ AMDGPUSubtarget::initializeSubtargetDependencies(const Triple &TT,
     FullFS += "+flat-for-global,";
   FullFS += FS;
 
-  if (GPU == "" && TT.getArch() == Triple::amdgcn)
-    GPU = "SI";
-
   ParseSubtargetFeatures(GPU, FullFS);
 
   // FIXME: I don't think think Evergreen has any useful support for
@@ -66,7 +63,7 @@ AMDGPUSubtarget::initializeSubtargetDependencies(const Triple &TT,
 
 AMDGPUSubtarget::AMDGPUSubtarget(const Triple &TT, StringRef GPU, StringRef FS,
                                  TargetMachine &TM)
-    : AMDGPUGenSubtargetInfo(TT, GPU, FS), DevName(GPU),
+    : AMDGPUGenSubtargetInfo(TT, GPU, FS),
       DumpCode(false), R600ALUInst(false), HasVertexCache(false),
       TexVTXClauseSize(0), Gen(AMDGPUSubtarget::R600), FP64(false),
       FP64Denormals(false), FP32Denormals(false), FastFMAF32(false),

diff --git a/lib/Target/AMDGPU/AMDGPUSubtarget.h b/lib/Target/AMDGPU/AMDGPUSubtarget.h
index 50510c268be42823fb1a64f75ef985c9a6039c4c..441bfab40899c1ddbf7acb48d8ea2b5d3f69e854 100644 (file)
--- a/lib/Target/AMDGPU/AMDGPUSubtarget.h
+++ b/lib/Target/AMDGPU/AMDGPUSubtarget.h
@@ -58,7 +58,6 @@ public:
   };
 
 private:
-  std::string DevName;
   bool DumpCode;
   bool R600ALUInst;
   bool HasVertexCache;
@@ -269,10 +268,6 @@ public:
     return false;
   }
 
-  StringRef getDeviceName() const {
-    return DevName;
-  }
-
   bool enableHugeScratchBuffer() const {
     return EnableHugeScratchBuffer;
   }

diff --git a/lib/Target/AMDGPU/AMDGPUTargetMachine.cpp b/lib/Target/AMDGPU/AMDGPUTargetMachine.cpp
index 7da7fad543d2fca35dae17277cc711a2f9181946..5a91be442771be81d89e66985d721db7414fabd9 100644 (file)
--- a/lib/Target/AMDGPU/AMDGPUTargetMachine.cpp
+++ b/lib/Target/AMDGPU/AMDGPUTargetMachine.cpp
@@ -88,14 +88,28 @@ static std::string computeDataLayout(const Triple &TT) {
   return Ret;
 }
 
+LLVM_READNONE
+static StringRef getGPUOrDefault(const Triple &TT, StringRef GPU) {
+  if (!GPU.empty())
+    return GPU;
+
+  // HSA only supports CI+, so change the default GPU to a CI for HSA.
+  if (TT.getArch() == Triple::amdgcn)
+    return (TT.getOS() == Triple::AMDHSA) ? "kaveri" : "tahiti";
+
+  return "";
+}
+
 AMDGPUTargetMachine::AMDGPUTargetMachine(const Target &T, const Triple &TT,
                                          StringRef CPU, StringRef FS,
                                          TargetOptions Options, Reloc::Model RM,
                                          CodeModel::Model CM,
                                          CodeGenOpt::Level OptLevel)
-    : LLVMTargetMachine(T, computeDataLayout(TT), TT, CPU, FS, Options, RM, CM,
+    : LLVMTargetMachine(T, computeDataLayout(TT), TT,
+                        getGPUOrDefault(TT, CPU), FS, Options, RM, CM,
                         OptLevel),
-      TLOF(createTLOF(getTargetTriple())), Subtarget(TT, CPU, FS, *this),
+      TLOF(createTLOF(getTargetTriple())),
+      Subtarget(TT, getTargetCPU(), FS, *this),
       IntrinsicInfo() {
   setRequiresStructuredCFG(true);
   initAsmInfo();

diff --git a/test/CodeGen/AMDGPU/hsa-default-device.ll b/test/CodeGen/AMDGPU/hsa-default-device.ll
new file mode 100644 (file)
index 0000000..631d6de
--- /dev/null
+++ b/test/CodeGen/AMDGPU/hsa-default-device.ll
@@ -0,0 +1,11 @@
+; RUN: llc -march=amdgcn -mtriple=amdgcn-unknown-amdhsa < %s | FileCheck %s
+
+; Make sure that with an HSA triple, we don't default to an
+; unsupported device.
+
+; CHECK: .hsa_code_object_isa 7,0,0,"AMD","AMDGPU"
+define void @test_kernel(float addrspace(1)* %out0, double addrspace(1)* %out1) nounwind {
+  store float 0.0, float addrspace(1)* %out0
+  ret void
+}
+