1; RUN: opt %loadPolly -polly-opt-isl -polly-pattern-matching-based-opts=false \
2; RUN: -debug < %s 2>&1| FileCheck %s
3; RUN: opt %loadPolly -polly-opt-isl -polly-pattern-matching-based-opts=true -debug < %s 2>&1| FileCheck %s --check-prefix=PATTERN-MATCHING-OPTS
4; RUN: opt %loadPolly -polly-opt-isl -polly-pattern-matching-based-opts=true -polly-ast-detect-parallel -polly-ast -analyze < %s | FileCheck %s --check-prefix=PARALLEL-AST
5; RUN: opt %loadPolly -polly-opt-isl -polly-pattern-matching-based-opts=true -stats -disable-output < %s 2>&1| FileCheck %s --check-prefix=STATS -match-full-lines
6; REQUIRES: asserts
7;
8;    /* C := alpha*A*B + beta*C */
9;    for (i = 0; i < _PB_NI; i++)
10;      for (j = 0; j < _PB_NJ; j++)
11;        {
12;	   C[i][j] *= beta;
13;	   for (k = 0; k < _PB_NK; ++k)
14;	     C[i][j] += alpha * A[i][k] * B[k][j];
15;        }
16;
17; CHECK-NOT: The matrix multiplication pattern was detected
18; PATTERN-MATCHING-OPTS: The matrix multiplication pattern was detected
19; PARALLEL-AST: #pragma known-parallel
20; PARALLEL-AST: #pragma known-parallel
21; STATS:  1 polly-opt-isl    - Number of matrix multiplication patterns detected and optimized
22;
23target datalayout = "e-m:e-i64:64-f80:128-n8:16:32:64-S128"
24target triple = "x86_64-unknown-unknown"
25
26define internal void @kernel_gemm(i32 %arg, i32 %arg1, i32 %arg2, double %arg3, double %arg4, [1056 x double]* %arg5, [1024 x double]* %arg6, [1056 x double]* %arg7) #0 {
27bb:
28  br label %bb8
29
30bb8:                                              ; preds = %bb29, %bb
31  %tmp = phi i64 [ 0, %bb ], [ %tmp30, %bb29 ]
32  br label %bb9
33
34bb9:                                              ; preds = %bb26, %bb8
35  %tmp10 = phi i64 [ 0, %bb8 ], [ %tmp27, %bb26 ]
36  %tmp11 = getelementptr inbounds [1056 x double], [1056 x double]* %arg5, i64 %tmp, i64 %tmp10
37  %tmp12 = load double, double* %tmp11, align 8
38  %tmp13 = fmul double %tmp12, %arg4
39  store double %tmp13, double* %tmp11, align 8
40  br label %Copy_0
41
42Copy_0:                                             ; preds = %Copy_0, %bb9
43  %tmp15 = phi i64 [ 0, %bb9 ], [ %tmp24, %Copy_0 ]
44  %tmp16 = getelementptr inbounds [1024 x double], [1024 x double]* %arg6, i64 %tmp, i64 %tmp15
45  %tmp17 = load double, double* %tmp16, align 8
46  %tmp18 = fmul double %tmp17, %arg3
47  %tmp19 = getelementptr inbounds [1056 x double], [1056 x double]* %arg7, i64 %tmp15, i64 %tmp10
48  %tmp20 = load double, double* %tmp19, align 8
49  %tmp21 = fmul double %tmp18, %tmp20
50  %tmp22 = load double, double* %tmp11, align 8
51  %tmp23 = fadd double %tmp22, %tmp21
52  store double %tmp23, double* %tmp11, align 8
53  %tmp24 = add nuw nsw i64 %tmp15, 1
54  %tmp25 = icmp ne i64 %tmp24, 1024
55  br i1 %tmp25, label %Copy_0, label %bb26
56
57bb26:                                             ; preds = %Copy_0
58  %tmp27 = add nuw nsw i64 %tmp10, 1
59  %tmp28 = icmp ne i64 %tmp27, 1056
60  br i1 %tmp28, label %bb9, label %bb29
61
62bb29:                                             ; preds = %bb26
63  %tmp30 = add nuw nsw i64 %tmp, 1
64  %tmp31 = icmp ne i64 %tmp30, 1056
65  br i1 %tmp31, label %bb8, label %bb32
66
67bb32:                                             ; preds = %bb29
68  ret void
69}
70