[polly-acc] change cl_get_* return types to 32/64bit

[alaindenzler]

Summary

This patch changes the return types for ocl_get_* functions during SPIR code generation. Because these functions return size_t types, the return type needs to be changed to the actual size of size_t on the device.

Based on work by Michal Babej and Pekka Jääskeläinen

[philip.pfaffe]

LGTM, thanks!

[grosser]

Also looks good from me except the minor changes. Please update the patch if you agree.

Then Philip or me can commit it.

lib/CodeGen/PPCGCodeGeneration.cpp

582

Can you document the new parameter.

test/GPGPU/spir-typesize.ll

91

Maybe add a newline here.

[alaindenzler]

Updated documentation and newline at the end of the test file.

lib/CodeGen/PPCGCodeGeneration.cpp

   /// Insert function calls to retrieve the SPIR group/local ids.
   ///
   /// @param The kernel to generate the function calls for.
-  void insertKernelCallsSPIR(ppcg_kernel *Kernel);
+  void insertKernelCallsSPIR(ppcg_kernel *Kernel, bool SizeTypeIs64bit);
 
   /// Setup the creation of functions referenced by the GPU kernel.
   ///

   }
 }
 
-void GPUNodeBuilder::insertKernelCallsSPIR(ppcg_kernel *Kernel) {
+void GPUNodeBuilder::insertKernelCallsSPIR(ppcg_kernel *Kernel,
+                                           bool SizeTypeIs64bit) {
   const char *GroupName[3] = {"__gen_ocl_get_group_id0",
                               "__gen_ocl_get_group_id1",
                               "__gen_ocl_get_group_id2"};

   const char *LocalName[3] = {"__gen_ocl_get_local_id0",
                               "__gen_ocl_get_local_id1",
                               "__gen_ocl_get_local_id2"};
+  IntegerType *SizeT =
+      SizeTypeIs64bit ? Builder.getInt64Ty() : Builder.getInt32Ty();
 
-  auto createFunc = [this](const char *Name, __isl_take isl_id *Id) mutable {
+  auto createFunc = [this](const char *Name, __isl_take isl_id *Id,
+                           IntegerType *SizeT) mutable {
     Module *M = Builder.GetInsertBlock()->getParent()->getParent();
     Function *FN = M->getFunction(Name);
 

     if (!FN) {
       GlobalValue::LinkageTypes Linkage = Function::ExternalLinkage;
       std::vector<Type *> Args;
-      FunctionType *Ty = FunctionType::get(Builder.getInt32Ty(), Args, false);
+      FunctionType *Ty = FunctionType::get(SizeT, Args, false);
       FN = Function::Create(Ty, Linkage, Name, M);
       FN->setCallingConv(CallingConv::SPIR_FUNC);
     }
 
     Value *Val = Builder.CreateCall(FN, {});
-    Val = Builder.CreateIntCast(Val, Builder.getInt64Ty(), false, Name);
+    if (SizeT == Builder.getInt32Ty())
+      Val = Builder.CreateIntCast(Val, Builder.getInt64Ty(), false, Name);
     IDToValue[Id] = Val;
     KernelIDs.insert(std::unique_ptr<isl_id, IslIdDeleter>(Id));
   };
 
   for (int i = 0; i < Kernel->n_grid; ++i)
-    createFunc(GroupName[i], isl_id_list_get_id(Kernel->block_ids, i));
+    createFunc(GroupName[i], isl_id_list_get_id(Kernel->block_ids, i), SizeT);
 
   for (int i = 0; i < Kernel->n_block; ++i)
-    createFunc(LocalName[i], isl_id_list_get_id(Kernel->thread_ids, i));
+    createFunc(LocalName[i], isl_id_list_get_id(Kernel->thread_ids, i), SizeT);
 }
 
 void GPUNodeBuilder::prepareKernelArguments(ppcg_kernel *Kernel, Function *FN) {

     insertKernelIntrinsics(Kernel);
     break;
   case GPUArch::SPIR32:
+    insertKernelCallsSPIR(Kernel, false);
+    break;
   case GPUArch::SPIR64:
-    insertKernelCallsSPIR(Kernel);
+    insertKernelCallsSPIR(Kernel, true);
     break;
   }
 }

test/GPGPU/spir-typesize.ll

+; RUN: opt %loadPolly -O3 -polly -polly-target=gpu \
+; RUN: -polly-gpu-arch=spir64 \
+; RUN: -polly-acc-dump-kernel-ir -polly-process-unprofitable -disable-output < %s | \
+; RUN: FileCheck -check-prefix=I64 %s
+
+; RUN: opt %loadPolly -O3 -polly -polly-target=gpu \
+; RUN: -polly-gpu-arch=spir32 \
+; RUN: -polly-acc-dump-kernel-ir -polly-process-unprofitable -disable-output < %s | \
+; RUN: FileCheck -check-prefix=I32 %s
+
+; REQUIRES pollyacc
+
+; This test case checks whether the openCl runtime functions (get_local_id/get_group_id) return the right types for 32 and 64bit devices.
+
+; I32:      target datalayout = "e-p:32:32:32-i1:8:8-i8:8:8-i16:16:16-i32:32:32-i64:64:64-i128:128:128-f32:32:32-f64:64:64-v16:16:16-v24:32:32-v32:32:32-v48:64:64-v64:64:64-v96:128:128-v128:128:128-v192:256:256-v256:256:256-v512:512:512-v1024:1024:1024"
+; I32-NEXT: target triple = "spir-unknown-unknown"
+
+; I32-LABEL: define spir_kernel void @FUNC_double_parallel_loop_SCOP_0_KERNEL_0(i8 addrspace(1)* %MemRef_A) #0 !kernel_arg_addr_space !0 !kernel_arg_name !1 !kernel_arg_access_qual !1 !kernel_arg_type !1 !kernel_arg_type_qual !1 !kernel_arg_base_type !1 {
+; I32-NEXT: entry:
+; I32-NEXT:   %0 = call i32 @__gen_ocl_get_group_id0()
+; I32-NEXT:   %__gen_ocl_get_group_id0 = zext i32 %0 to i64
+; I32-NEXT:   %1 = call i32 @__gen_ocl_get_group_id1()
+; I32-NEXT:   %__gen_ocl_get_group_id1 = zext i32 %1 to i64
+; I32-NEXT:   %2 = call i32 @__gen_ocl_get_local_id0()
+; I32-NEXT:   %__gen_ocl_get_local_id0 = zext i32 %2 to i64
+; I32-NEXT:   %3 = call i32 @__gen_ocl_get_local_id1()
+; I32-NEXT:   %__gen_ocl_get_local_id1 = zext i32 %3 to i64
+; I32-NEXT:   br label %polly.loop_preheader
+
+; I64:       target datalayout = "e-p:64:64:64-i1:8:8-i8:8:8-i16:16:16-i32:32:32-i64:64:64-i128:128:128-f32:32:32-f64:64:64-v16:16:16-v24:32:32-v32:32:32-v48:64:64-v64:64:64-v96:128:128-v128:128:128-v192:256:256-v256:256:256-v512:512:512-v1024:1024:1024"
+; I64-next:  target triple = "spir64-unknown-unknown"
+
+; I64-LABEL: define spir_kernel void @FUNC_double_parallel_loop_SCOP_0_KERNEL_0(i8 addrspace(1)* %MemRef_A) #0 !kernel_arg_addr_space !0 !kernel_arg_name !1 !kernel_arg_access_qual !1 !kernel_arg_type !1 !kernel_arg_type_qual !1 !kernel_arg_base_type !1 {
+; I64-NEXT: entry:
+; I64-NEXT:   %0 = call i64 @__gen_ocl_get_group_id0()
+; I64-NEXT:   %1 = call i64 @__gen_ocl_get_group_id1()
+; I64-NEXT:   %2 = call i64 @__gen_ocl_get_local_id0()
+; I64-NEXT:   %3 = call i64 @__gen_ocl_get_local_id1()
+; I64-NEXT:   br label %polly.loop_preheader
+
+
+;    void double_parallel_loop(float A[][1024]) {
+;      for (long i = 0; i < 1024; i++)
+;        for (long j = 0; j < 1024; j++)
+;          A[i][j] += i * j;
+;    }
+;
+
+target datalayout = "e-m:e-i64:64-f80:128-n8:16:32:64-S128"
+
+define void @double_parallel_loop([1024 x float]* %A) {
+bb:
+  br label %bb2
+
+bb2:                                              ; preds = %bb13, %bb
+  %i.0 = phi i64 [ 0, %bb ], [ %tmp14, %bb13 ]
+  %exitcond1 = icmp ne i64 %i.0, 1024
+  br i1 %exitcond1, label %bb3, label %bb15
+
+bb3:                                              ; preds = %bb2
+  br label %bb4
+
+bb4:                                              ; preds = %bb10, %bb3
+  %j.0 = phi i64 [ 0, %bb3 ], [ %tmp11, %bb10 ]
+  %exitcond = icmp ne i64 %j.0, 1024
+  br i1 %exitcond, label %bb5, label %bb12
+
+bb5:                                              ; preds = %bb4
+  %tmp = mul nuw nsw i64 %i.0, %j.0
+  %tmp6 = sitofp i64 %tmp to float
+  %tmp7 = getelementptr inbounds [1024 x float], [1024 x float]* %A, i64 %i.0, i64 %j.0
+  %tmp8 = load float, float* %tmp7, align 4
+  %tmp9 = fadd float %tmp8, %tmp6
+  store float %tmp9, float* %tmp7, align 4
+  br label %bb10
+
+bb10:                                             ; preds = %bb5
+  %tmp11 = add nuw nsw i64 %j.0, 1
+  br label %bb4
+
+bb12:                                             ; preds = %bb4
+  br label %bb13
+
+bb13:                                             ; preds = %bb12
+  %tmp14 = add nuw nsw i64 %i.0, 1
+  br label %bb2
+
+bb15:                                             ; preds = %bb2
+  ret void
+}
 No newline at end of file