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; This test case ensures that we properly sign-extend the types we are using.
11
12; CODE:      if (arg >= 1 && arg1 == 0) {
13; CODE:        cudaCheckReturn(cudaMemcpy(dev_MemRef_arg3, MemRef_arg3, (arg) * sizeof(double), cudaMemcpyHostToDevice));
14; CODE-NEXT:   {
15; CODE-NEXT:     dim3 k0_dimBlock(32);
16; CODE-NEXT:     dim3 k0_dimGrid(arg >= 1048545 ? 32768 : (arg + 31) / 32);
17; CODE-NEXT:     kernel0 <<<k0_dimGrid, k0_dimBlock>>> (dev_MemRef_arg3, dev_MemRef_arg2, arg, arg1);
18; CODE-NEXT:     cudaCheckKernel();
19; CODE-NEXT:   }
20
21; CODE:   cudaCheckReturn(cudaMemcpy(MemRef_arg2, dev_MemRef_arg2, (arg) * sizeof(double), cudaMemcpyDeviceToHost));
22; CODE-NEXT  cudaCheckReturn(cudaFree(dev_MemRef_arg3));
23; CODE-NEXT  cudaCheckReturn(cudaFree(dev_MemRef_arg2));
24
25; CODE: # kernel0
26; CODE-NEXT: for (int c0 = 0; c0 <= (arg - 32 * b0 - 1) / 1048576; c0 += 1)
27; CODE-NEXT:   if (arg >= 32 * b0 + t0 + 1048576 * c0 + 1)
28; CODE-NEXT:     Stmt_bb6(0, 32 * b0 + t0 + 1048576 * c0);
29
30; IR-LABEL:  call i8* @polly_initContextCUDA()
31; IR:        sext i32 %arg to i64
32; IR-NEXT:   mul i64
33; IR-NEXT:   @polly_allocateMemoryForDevice
34
35target datalayout = "e-m:e-i64:64-f80:128-n8:16:32:64-S128"
36target triple = "x86_64-unknown-linux-gnu"
37
38define void @hoge(i32 %arg, i32 %arg1, [1000 x double]* %arg2, double* %arg3) {
39bb:
40  br label %bb4
41
42bb4:                                              ; preds = %bb13, %bb
43  br label %bb6
44
45bb5:                                              ; preds = %bb13
46  ret void
47
48bb6:                                              ; preds = %bb6, %bb4
49  %tmp = phi i64 [ 0, %bb4 ], [ %tmp10, %bb6 ]
50  %tmp7 = getelementptr inbounds double, double* %arg3, i64 %tmp
51  %tmp8 = load double, double* %tmp7, align 8
52  %tmp9 = getelementptr inbounds [1000 x double], [1000 x double]* %arg2, i64 0, i64 %tmp
53  store double %tmp8, double* %tmp9, align 8
54  %tmp10 = add nuw nsw i64 %tmp, 1
55  %tmp11 = zext i32 %arg to i64
56  %tmp12 = icmp ne i64 %tmp10, %tmp11
57  br i1 %tmp12, label %bb6, label %bb13
58
59bb13:                                             ; preds = %bb6
60  %tmp14 = zext i32 %arg1 to i64
61  %tmp15 = icmp ne i64 0, %tmp14
62  br i1 %tmp15, label %bb4, label %bb5
63}
64