[clang][openmp][NFC] Remove arch-specific CGOpenMPRuntimeGPU files

The existing CGOpenMPRuntimeAMDGCN and CGOpenMPRuntimeNVPTX classes are
just code bloat. By removing them, the codebase gets a bit cleaner.

Reviewed By: jdoerfert, JonChesterfield, tianshilei1992

Differential Revision: https://reviews.llvm.org/D113421

Author: atmnpatel

clang/lib/CodeGen/CGOpenMPRuntimeAMDGCN.cpp

@@ -12,7 +12,6 @@
 //===----------------------------------------------------------------------===//
 
 #include "CGOpenMPRuntimeGPU.h"
-#include "CGOpenMPRuntimeNVPTX.h"
 #include "CodeGenFunction.h"
 #include "clang/AST/Attr.h"
 #include "clang/AST/DeclOpenMP.h"
@@ -21,7 +20,6 @@
 #include "clang/Basic/Cuda.h"
 #include "llvm/ADT/SmallPtrSet.h"
 #include "llvm/Frontend/OpenMP/OMPGridValues.h"
-#include "llvm/IR/IntrinsicsNVPTX.h"
 #include "llvm/Support/MathExtras.h"
 
 using namespace clang;
@@ -1197,7 +1195,7 @@ unsigned CGOpenMPRuntimeGPU::getDefaultLocationReserved2Flags() const {
 CGOpenMPRuntimeGPU::CGOpenMPRuntimeGPU(CodeGenModule &CGM)
     : CGOpenMPRuntime(CGM, "_", "$") {
   if (!CGM.getLangOpts().OpenMPIsDevice)
-    llvm_unreachable("OpenMP NVPTX can only handle device code.");
+    llvm_unreachable("OpenMP can only handle device code.");
 
   llvm::OpenMPIRBuilder &OMPBuilder = getOMPBuilder();
   if (CGM.getLangOpts().OpenMPTargetNewRuntime) {
@@ -3960,3 +3958,18 @@ llvm::Value *CGOpenMPRuntimeGPU::getGPUNumThreads(CodeGenFunction &CGF) {
   }
   return Bld.CreateCall(F, llvm::None, "nvptx_num_threads");
 }
+
+llvm::Value *CGOpenMPRuntimeGPU::getGPUThreadID(CodeGenFunction &CGF) {
+  ArrayRef<llvm::Value *> Args{};
+  return CGF.EmitRuntimeCall(
+      OMPBuilder.getOrCreateRuntimeFunction(
+          CGM.getModule(), OMPRTL___kmpc_get_hardware_thread_id_in_block),
+      Args);
+}
+
+llvm::Value *CGOpenMPRuntimeGPU::getGPUWarpSize(CodeGenFunction &CGF) {
+  ArrayRef<llvm::Value *> Args{};
+  return CGF.EmitRuntimeCall(OMPBuilder.getOrCreateRuntimeFunction(
+                                 CGM.getModule(), OMPRTL___kmpc_get_warp_size),
+                             Args);
+}

clang/lib/CodeGen/CGOpenMPRuntimeAMDGCN.h

@@ -176,10 +176,10 @@ class CGOpenMPRuntimeGPU : public CGOpenMPRuntime {
   /// and NVPTX.
 
   /// Get the GPU warp size.
-  virtual llvm::Value *getGPUWarpSize(CodeGenFunction &CGF) = 0;
+  llvm::Value *getGPUWarpSize(CodeGenFunction &CGF);
 
   /// Get the id of the current thread on the GPU.
-  virtual llvm::Value *getGPUThreadID(CodeGenFunction &CGF) = 0;
+  llvm::Value *getGPUThreadID(CodeGenFunction &CGF);
 
   /// Get the maximum number of threads in a block of the GPU.
   llvm::Value *getGPUNumThreads(CodeGenFunction &CGF);

clang/lib/CodeGen/CGOpenMPRuntimeGPU.cpp

@@ -59,9 +59,7 @@ add_clang_library(clangCodeGen
   CGObjCRuntime.cpp
   CGOpenCLRuntime.cpp
   CGOpenMPRuntime.cpp
-  CGOpenMPRuntimeAMDGCN.cpp
   CGOpenMPRuntimeGPU.cpp
-  CGOpenMPRuntimeNVPTX.cpp
   CGRecordLayoutBuilder.cpp
   CGStmt.cpp
   CGStmtOpenMP.cpp

clang/lib/CodeGen/CGOpenMPRuntimeGPU.h

@@ -19,8 +19,7 @@
 #include "CGObjCRuntime.h"
 #include "CGOpenCLRuntime.h"
 #include "CGOpenMPRuntime.h"
-#include "CGOpenMPRuntimeAMDGCN.h"
-#include "CGOpenMPRuntimeNVPTX.h"
+#include "CGOpenMPRuntimeGPU.h"
 #include "CodeGenFunction.h"
 #include "CodeGenPGO.h"
 #include "ConstantEmitter.h"
@@ -244,14 +243,10 @@ void CodeGenModule::createOpenMPRuntime() {
   switch (getTriple().getArch()) {
   case llvm::Triple::nvptx:
   case llvm::Triple::nvptx64:
-    assert(getLangOpts().OpenMPIsDevice &&
-           "OpenMP NVPTX is only prepared to deal with device code.");
-    OpenMPRuntime.reset(new CGOpenMPRuntimeNVPTX(*this));
-    break;
   case llvm::Triple::amdgcn:
     assert(getLangOpts().OpenMPIsDevice &&
-           "OpenMP AMDGCN is only prepared to deal with device code.");
-    OpenMPRuntime.reset(new CGOpenMPRuntimeAMDGCN(*this));
+           "OpenMP AMDGPU/NVPTX is only prepared to deal with device code.");
+    OpenMPRuntime.reset(new CGOpenMPRuntimeGPU(*this));
     break;
   default:
     if (LangOpts.OpenMPSimd)

clang/lib/CodeGen/CGOpenMPRuntimeNVPTX.cpp

@@ -1664,31 +1664,31 @@ int bar(int n){
 // CHECK1-NEXT:    store i32* [[A]], i32** [[A_ADDR]], align 8
 // CHECK1-NEXT:    [[TMP0:%.*]] = load i32*, i32** [[A_ADDR]], align 8
 // CHECK1-NEXT:    [[TMP1:%.*]] = call i64 @__kmpc_warp_active_thread_mask()
-// CHECK1-NEXT:    [[NVPTX_TID:%.*]] = call i32 @llvm.nvvm.read.ptx.sreg.tid.x()
+// CHECK1-NEXT:    [[TMP2:%.*]] = call i32 @__kmpc_get_hardware_thread_id_in_block()
 // CHECK1-NEXT:    [[NVPTX_NUM_THREADS:%.*]] = call i32 @__kmpc_get_hardware_num_threads_in_block()
 // CHECK1-NEXT:    store i32 0, i32* [[CRITICAL_COUNTER]], align 4
 // CHECK1-NEXT:    br label [[OMP_CRITICAL_LOOP:%.*]]
 // CHECK1:       omp.critical.loop:
-// CHECK1-NEXT:    [[TMP2:%.*]] = load i32, i32* [[CRITICAL_COUNTER]], align 4
-// CHECK1-NEXT:    [[TMP3:%.*]] = icmp slt i32 [[TMP2]], [[NVPTX_NUM_THREADS]]
-// CHECK1-NEXT:    br i1 [[TMP3]], label [[OMP_CRITICAL_TEST:%.*]], label [[OMP_CRITICAL_EXIT:%.*]]
+// CHECK1-NEXT:    [[TMP3:%.*]] = load i32, i32* [[CRITICAL_COUNTER]], align 4
+// CHECK1-NEXT:    [[TMP4:%.*]] = icmp slt i32 [[TMP3]], [[NVPTX_NUM_THREADS]]
+// CHECK1-NEXT:    br i1 [[TMP4]], label [[OMP_CRITICAL_TEST:%.*]], label [[OMP_CRITICAL_EXIT:%.*]]
 // CHECK1:       omp.critical.test:
-// CHECK1-NEXT:    [[TMP4:%.*]] = load i32, i32* [[CRITICAL_COUNTER]], align 4
-// CHECK1-NEXT:    [[TMP5:%.*]] = icmp eq i32 [[NVPTX_TID]], [[TMP4]]
-// CHECK1-NEXT:    br i1 [[TMP5]], label [[OMP_CRITICAL_BODY:%.*]], label [[OMP_CRITICAL_SYNC:%.*]]
+// CHECK1-NEXT:    [[TMP5:%.*]] = load i32, i32* [[CRITICAL_COUNTER]], align 4
+// CHECK1-NEXT:    [[TMP6:%.*]] = icmp eq i32 [[TMP2]], [[TMP5]]
+// CHECK1-NEXT:    br i1 [[TMP6]], label [[OMP_CRITICAL_BODY:%.*]], label [[OMP_CRITICAL_SYNC:%.*]]
 // CHECK1:       omp.critical.body:
-// CHECK1-NEXT:    [[TMP6:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
-// CHECK1-NEXT:    [[TMP7:%.*]] = load i32, i32* [[TMP6]], align 4
-// CHECK1-NEXT:    call void @__kmpc_critical(%struct.ident_t* @[[GLOB1]], i32 [[TMP7]], [8 x i32]* @"_gomp_critical_user_$var")
-// CHECK1-NEXT:    [[TMP8:%.*]] = load i32, i32* [[TMP0]], align 4
-// CHECK1-NEXT:    [[INC:%.*]] = add nsw i32 [[TMP8]], 1
+// CHECK1-NEXT:    [[TMP7:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
+// CHECK1-NEXT:    [[TMP8:%.*]] = load i32, i32* [[TMP7]], align 4
+// CHECK1-NEXT:    call void @__kmpc_critical(%struct.ident_t* @[[GLOB1]], i32 [[TMP8]], [8 x i32]* @"_gomp_critical_user_$var")
+// CHECK1-NEXT:    [[TMP9:%.*]] = load i32, i32* [[TMP0]], align 4
+// CHECK1-NEXT:    [[INC:%.*]] = add nsw i32 [[TMP9]], 1
 // CHECK1-NEXT:    store i32 [[INC]], i32* [[TMP0]], align 4
-// CHECK1-NEXT:    call void @__kmpc_end_critical(%struct.ident_t* @[[GLOB1]], i32 [[TMP7]], [8 x i32]* @"_gomp_critical_user_$var")
+// CHECK1-NEXT:    call void @__kmpc_end_critical(%struct.ident_t* @[[GLOB1]], i32 [[TMP8]], [8 x i32]* @"_gomp_critical_user_$var")
 // CHECK1-NEXT:    br label [[OMP_CRITICAL_SYNC]]
 // CHECK1:       omp.critical.sync:
 // CHECK1-NEXT:    call void @__kmpc_syncwarp(i64 [[TMP1]])
-// CHECK1-NEXT:    [[TMP9:%.*]] = add nsw i32 [[TMP4]], 1
-// CHECK1-NEXT:    store i32 [[TMP9]], i32* [[CRITICAL_COUNTER]], align 4
+// CHECK1-NEXT:    [[TMP10:%.*]] = add nsw i32 [[TMP5]], 1
+// CHECK1-NEXT:    store i32 [[TMP10]], i32* [[CRITICAL_COUNTER]], align 4
 // CHECK1-NEXT:    br label [[OMP_CRITICAL_LOOP]]
 // CHECK1:       omp.critical.exit:
 // CHECK1-NEXT:    ret void
@@ -1936,31 +1936,31 @@ int bar(int n){
 // CHECK2-NEXT:    store i32* [[A]], i32** [[A_ADDR]], align 4
 // CHECK2-NEXT:    [[TMP0:%.*]] = load i32*, i32** [[A_ADDR]], align 4
 // CHECK2-NEXT:    [[TMP1:%.*]] = call i64 @__kmpc_warp_active_thread_mask()
-// CHECK2-NEXT:    [[NVPTX_TID:%.*]] = call i32 @llvm.nvvm.read.ptx.sreg.tid.x()
+// CHECK2-NEXT:    [[TMP2:%.*]] = call i32 @__kmpc_get_hardware_thread_id_in_block()
 // CHECK2-NEXT:    [[NVPTX_NUM_THREADS:%.*]] = call i32 @__kmpc_get_hardware_num_threads_in_block()
 // CHECK2-NEXT:    store i32 0, i32* [[CRITICAL_COUNTER]], align 4
 // CHECK2-NEXT:    br label [[OMP_CRITICAL_LOOP:%.*]]
 // CHECK2:       omp.critical.loop:
-// CHECK2-NEXT:    [[TMP2:%.*]] = load i32, i32* [[CRITICAL_COUNTER]], align 4
-// CHECK2-NEXT:    [[TMP3:%.*]] = icmp slt i32 [[TMP2]], [[NVPTX_NUM_THREADS]]
-// CHECK2-NEXT:    br i1 [[TMP3]], label [[OMP_CRITICAL_TEST:%.*]], label [[OMP_CRITICAL_EXIT:%.*]]
+// CHECK2-NEXT:    [[TMP3:%.*]] = load i32, i32* [[CRITICAL_COUNTER]], align 4
+// CHECK2-NEXT:    [[TMP4:%.*]] = icmp slt i32 [[TMP3]], [[NVPTX_NUM_THREADS]]
+// CHECK2-NEXT:    br i1 [[TMP4]], label [[OMP_CRITICAL_TEST:%.*]], label [[OMP_CRITICAL_EXIT:%.*]]
 // CHECK2:       omp.critical.test:
-// CHECK2-NEXT:    [[TMP4:%.*]] = load i32, i32* [[CRITICAL_COUNTER]], align 4
-// CHECK2-NEXT:    [[TMP5:%.*]] = icmp eq i32 [[NVPTX_TID]], [[TMP4]]
-// CHECK2-NEXT:    br i1 [[TMP5]], label [[OMP_CRITICAL_BODY:%.*]], label [[OMP_CRITICAL_SYNC:%.*]]
+// CHECK2-NEXT:    [[TMP5:%.*]] = load i32, i32* [[CRITICAL_COUNTER]], align 4
+// CHECK2-NEXT:    [[TMP6:%.*]] = icmp eq i32 [[TMP2]], [[TMP5]]
+// CHECK2-NEXT:    br i1 [[TMP6]], label [[OMP_CRITICAL_BODY:%.*]], label [[OMP_CRITICAL_SYNC:%.*]]
 // CHECK2:       omp.critical.body:
-// CHECK2-NEXT:    [[TMP6:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
-// CHECK2-NEXT:    [[TMP7:%.*]] = load i32, i32* [[TMP6]], align 4
-// CHECK2-NEXT:    call void @__kmpc_critical(%struct.ident_t* @[[GLOB1]], i32 [[TMP7]], [8 x i32]* @"_gomp_critical_user_$var")
-// CHECK2-NEXT:    [[TMP8:%.*]] = load i32, i32* [[TMP0]], align 4
-// CHECK2-NEXT:    [[INC:%.*]] = add nsw i32 [[TMP8]], 1
+// CHECK2-NEXT:    [[TMP7:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
+// CHECK2-NEXT:    [[TMP8:%.*]] = load i32, i32* [[TMP7]], align 4
+// CHECK2-NEXT:    call void @__kmpc_critical(%struct.ident_t* @[[GLOB1]], i32 [[TMP8]], [8 x i32]* @"_gomp_critical_user_$var")
+// CHECK2-NEXT:    [[TMP9:%.*]] = load i32, i32* [[TMP0]], align 4
+// CHECK2-NEXT:    [[INC:%.*]] = add nsw i32 [[TMP9]], 1
 // CHECK2-NEXT:    store i32 [[INC]], i32* [[TMP0]], align 4
-// CHECK2-NEXT:    call void @__kmpc_end_critical(%struct.ident_t* @[[GLOB1]], i32 [[TMP7]], [8 x i32]* @"_gomp_critical_user_$var")
+// CHECK2-NEXT:    call void @__kmpc_end_critical(%struct.ident_t* @[[GLOB1]], i32 [[TMP8]], [8 x i32]* @"_gomp_critical_user_$var")
 // CHECK2-NEXT:    br label [[OMP_CRITICAL_SYNC]]
 // CHECK2:       omp.critical.sync:
 // CHECK2-NEXT:    call void @__kmpc_syncwarp(i64 [[TMP1]])
-// CHECK2-NEXT:    [[TMP9:%.*]] = add nsw i32 [[TMP4]], 1
-// CHECK2-NEXT:    store i32 [[TMP9]], i32* [[CRITICAL_COUNTER]], align 4
+// CHECK2-NEXT:    [[TMP10:%.*]] = add nsw i32 [[TMP5]], 1
+// CHECK2-NEXT:    store i32 [[TMP10]], i32* [[CRITICAL_COUNTER]], align 4
 // CHECK2-NEXT:    br label [[OMP_CRITICAL_LOOP]]
 // CHECK2:       omp.critical.exit:
 // CHECK2-NEXT:    ret void