1 ; RUN: opt %loadPolly -polly-codegen-ppcg -polly-acc-dump-code \
2 ; RUN: -disable-output < %s | \
3 ; RUN: FileCheck -check-prefix=CODE %s
5 ; RUN: opt %loadPolly -polly-codegen-ppcg -S < %s | \
6 ; RUN: FileCheck %s -check-prefix=IR
10 ; void foo(float A[]) {
11 ; for (long i = 0; i < 128; i++)
14 ; for (long i = 0; i < 128; i++)
15 ; for (long j = 0; j < 128; j++)
23 ; CODE-NEXT: cudaCheckReturn(cudaMemcpy(dev_MemRef_A, MemRef_A, (128) * sizeof(float), cudaMemcpyHostToDevice));
25 ; CODE-NEXT: dim3 k0_dimBlock(32);
26 ; CODE-NEXT: dim3 k0_dimGrid(4);
27 ; CODE-NEXT: kernel0 <<<k0_dimGrid, k0_dimBlock>>> (dev_MemRef_A);
28 ; CODE-NEXT: cudaCheckKernel();
31 ; CODE: for (int c0 = 0; c0 <= 127; c0 += 1)
32 ; CODE-NEXT: for (int c1 = 0; c1 <= 127; c1 += 1)
34 ; CODE-NEXT: dim3 k1_dimBlock;
35 ; CODE-NEXT: dim3 k1_dimGrid;
36 ; CODE-NEXT: kernel1 <<<k1_dimGrid, k1_dimBlock>>> (dev_MemRef_A, c0, c1);
37 ; CODE-NEXT: cudaCheckKernel();
40 ; CODE: cudaCheckReturn(cudaMemcpy(MemRef_A, dev_MemRef_A, (128) * sizeof(float), cudaMemcpyDeviceToHost));
44 ; CODE-NEXT: Stmt_bb4(32 * b0 + t0);
47 ; CODE-NEXT: Stmt_bb14(c0, c1);
49 ; Verify that we identified this kernel as non-profitable.
50 ; IR: br i1 false, label %polly.start, label %bb3
52 target datalayout = "e-m:e-i64:64-f80:128-n8:16:32:64-S128"
54 define void @foo(float* %A) {
58 bb3: ; preds = %bb8, %bb
59 %i.0 = phi i64 [ 0, %bb ], [ %tmp9, %bb8 ]
60 %exitcond2 = icmp ne i64 %i.0, 128
61 br i1 %exitcond2, label %bb4, label %bb10
64 %tmp = sitofp i64 %i.0 to float
65 %tmp5 = getelementptr inbounds float, float* %A, i64 %i.0
66 %tmp6 = load float, float* %tmp5, align 4
67 %tmp7 = fadd float %tmp6, %tmp
68 store float %tmp7, float* %tmp5, align 4
72 %tmp9 = add nuw nsw i64 %i.0, 1
78 bb11: ; preds = %bb23, %bb10
79 %i1.0 = phi i64 [ 0, %bb10 ], [ %tmp24, %bb23 ]
80 %exitcond1 = icmp ne i64 %i1.0, 128
81 br i1 %exitcond1, label %bb12, label %bb25
86 bb13: ; preds = %bb20, %bb12
87 %j.0 = phi i64 [ 0, %bb12 ], [ %tmp21, %bb20 ]
88 %exitcond = icmp ne i64 %j.0, 128
89 br i1 %exitcond, label %bb14, label %bb22
92 %tmp15 = add nuw nsw i64 %i1.0, %j.0
93 %tmp16 = sitofp i64 %tmp15 to float
94 %tmp17 = getelementptr inbounds float, float* %A, i64 42
95 %tmp18 = load float, float* %tmp17, align 4
96 %tmp19 = fadd float %tmp18, %tmp16
97 store float %tmp19, float* %tmp17, align 4
100 bb20: ; preds = %bb14
101 %tmp21 = add nuw nsw i64 %j.0, 1
104 bb22: ; preds = %bb13
107 bb23: ; preds = %bb22
108 %tmp24 = add nuw nsw i64 %i1.0, 1
111 bb25: ; preds = %bb11