1; RUN: opt %loadPolly -polly-codegen-ppcg -polly-acc-dump-code \ 2; RUN: -disable-output < %s | \ 3; RUN: FileCheck -check-prefix=CODE %s 4 5; RUN: opt %loadPolly -polly-codegen-ppcg -S < %s | \ 6; RUN: FileCheck %s -check-prefix=IR 7; 8; REQUIRES: pollyacc 9 10; CODE: cudaCheckReturn(cudaMemcpy(dev_MemRef_B, MemRef_B, (16) * sizeof(float), cudaMemcpyHostToDevice)); 11; CODE-NEXT: cudaCheckReturn(cudaMemcpy(dev_MemRef_A, MemRef_A, (8) * sizeof(float), cudaMemcpyHostToDevice)); 12 13; CODE: dim3 k0_dimBlock(8); 14; CODE-NEXT: dim3 k0_dimGrid(1); 15; CODE-NEXT: kernel0 <<<k0_dimGrid, k0_dimBlock>>> (dev_MemRef_A); 16; CODE-NEXT: cudaCheckKernel(); 17; CODE-NEXT: } 18 19; CODE: { 20; CODE-NEXT: dim3 k1_dimBlock(8); 21; CODE-NEXT: dim3 k1_dimGrid(1); 22; CODE-NEXT: kernel1 <<<k1_dimGrid, k1_dimBlock>>> (dev_MemRef_B); 23; CODE-NEXT: cudaCheckKernel(); 24; CODE-NEXT: } 25 26; CODE: cudaCheckReturn(cudaMemcpy(MemRef_B, dev_MemRef_B, (16) * sizeof(float), cudaMemcpyDeviceToHost)); 27; CODE-NEXT: cudaCheckReturn(cudaMemcpy(MemRef_A, dev_MemRef_A, (8) * sizeof(float), cudaMemcpyDeviceToHost)); 28 29; CODE: # kernel0 30; CODE-NEXT: Stmt_bb11(t0); 31 32; CODE: # kernel1 33; CODE-NEXT: Stmt_bb3(t0); 34 35; IR: %p_dev_array_MemRef_B = call i8* @polly_allocateMemoryForDevice(i64 32) 36; IR-NEXT: %p_dev_array_MemRef_A = call i8* @polly_allocateMemoryForDevice(i64 32) 37; IR-NEXT: [[REG0:%.+]] = getelementptr float, float* %B, i64 8 38; IR-NEXT: [[REG1:%.+]] = bitcast float* [[REG0]] to i8* 39; IR-NEXT: call void @polly_copyFromHostToDevice(i8* [[REG1]], i8* %p_dev_array_MemRef_B, i64 32) 40 41; IR: [[REGA:%.+]] = call i8* @polly_getDevicePtr(i8* %p_dev_array_MemRef_B) 42; IR-NEXT: [[REGB:%.+]] = bitcast i8* [[REGA]] to float* 43; IR-NEXT: [[REGC:%.+]] = getelementptr float, float* [[REGB]], i64 -8 44; IR-NEXT: [[REGD:%.+]] = bitcast float* [[REGC]] to i8* 45 46; void foo(float A[], float B[]) { 47; for (long i = 0; i < 8; i++) 48; B[i + 8] *= 4; 49; 50; for (long i = 0; i < 8; i++) 51; A[i] *= 12; 52; } 53; 54; #ifdef OUTPUT 55; int main() { 56; float A[16]; 57; 58; for (long i = 0; i < 16; i++) { 59; __sync_synchronize(); 60; A[i] = i; 61; } 62; 63; foo(A, A); 64; 65; float sum = 0; 66; for (long i = 0; i < 16; i++) { 67; __sync_synchronize(); 68; sum += A[i]; 69; } 70; 71; printf("%f\n", sum); 72; } 73; #endif 74; 75target datalayout = "e-m:e-i64:64-f80:128-n8:16:32:64-S128" 76 77define void @foo(float* %A, float* %B) { 78bb: 79 br label %bb2 80 81bb2: ; preds = %bb7, %bb 82 %i.0 = phi i64 [ 0, %bb ], [ %tmp8, %bb7 ] 83 %exitcond1 = icmp ne i64 %i.0, 8 84 br i1 %exitcond1, label %bb3, label %bb9 85 86bb3: ; preds = %bb2 87 %tmp = add nuw nsw i64 %i.0, 8 88 %tmp4 = getelementptr inbounds float, float* %B, i64 %tmp 89 %tmp5 = load float, float* %tmp4, align 4 90 %tmp6 = fmul float %tmp5, 4.000000e+00 91 store float %tmp6, float* %tmp4, align 4 92 br label %bb7 93 94bb7: ; preds = %bb3 95 %tmp8 = add nuw nsw i64 %i.0, 1 96 br label %bb2 97 98bb9: ; preds = %bb2 99 br label %bb10 100 101bb10: ; preds = %bb15, %bb9 102 %i1.0 = phi i64 [ 0, %bb9 ], [ %tmp16, %bb15 ] 103 %exitcond = icmp ne i64 %i1.0, 8 104 br i1 %exitcond, label %bb11, label %bb17 105 106bb11: ; preds = %bb10 107 %tmp12 = getelementptr inbounds float, float* %A, i64 %i1.0 108 %tmp13 = load float, float* %tmp12, align 4 109 %tmp14 = fmul float %tmp13, 1.200000e+01 110 store float %tmp14, float* %tmp12, align 4 111 br label %bb15 112 113bb15: ; preds = %bb11 114 %tmp16 = add nuw nsw i64 %i1.0, 1 115 br label %bb10 116 117bb17: ; preds = %bb10 118 ret void 119} 120