llvm-project

1
; RUN: opt %loadNPMPolly -passes=polly-opt-isl -polly-pattern-matching-based-opts=true \
2
; RUN: -polly-tc-opt=true -debug -disable-output < %s 2>&1 | FileCheck %s
3
; REQUIRES: asserts
4
;
5
;      for (i = 0; i < 1024; i++)
6
;        for (j = 0; j < 1024; j++)
7
;          for (l = 0; l < 64; ++l)
8
;            for (w = 0; w < 64; ++w)
9
;              C[i][j] += A[i][l][w] * B[w][j][l];
10
;
11
; CHECK: The tensor contraction pattern was detected
12
;
13
target datalayout = "e-m:e-p270:32:32-p271:32:32-p272:64:64-i64:64-f80:128-n8:16:32:64-S128"
14
target triple = "x86_64-unknown-linux-gnu"
15

16
define internal void @kernel_tc(i32 %ni, i32 %nj, i32 %nl, i32 %nq, i32 %nw, double %alpha, double %beta, ptr %C, ptr %A, ptr %B) {
17
entry:
18
  br label %for.cond1.preheader
19

20
for.cond1.preheader:                              ; preds = %for.inc30, %entry
21
  %indvars.iv43 = phi i64 [ 0, %entry ], [ %indvars.iv.next44, %for.inc30 ]
22
  br label %for.cond4.preheader
23

24
for.cond4.preheader:                              ; preds = %for.inc27, %for.cond1.preheader
25
  %indvars.iv40 = phi i64 [ 0, %for.cond1.preheader ], [ %indvars.iv.next41, %for.inc27 ]
26
  br label %for.cond7.preheader
27

28
for.cond7.preheader:                              ; preds = %for.inc24, %for.cond4.preheader
29
  %indvars.iv37 = phi i64 [ 0, %for.cond4.preheader ], [ %indvars.iv.next38, %for.inc24 ]
30
  br label %for.body9
31

32
for.body9:                                        ; preds = %for.body9, %for.cond7.preheader
33
  %indvars.iv = phi i64 [ 0, %for.cond7.preheader ], [ %indvars.iv.next, %for.body9 ]
34
  %arrayidx13 = getelementptr inbounds [64 x [64 x double]], ptr %A, i64 %indvars.iv43, i64 %indvars.iv37, i64 %indvars.iv
35
  %i = load double, ptr %arrayidx13, align 8
36
  %arrayidx19 = getelementptr inbounds [1024 x [64 x double]], ptr %B, i64 %indvars.iv, i64 %indvars.iv40, i64 %indvars.iv37
37
  %i1 = load double, ptr %arrayidx19, align 8
38
  %mul = fmul fast double %i1, %i
39
  %arrayidx23 = getelementptr inbounds [1024 x double], ptr %C, i64 %indvars.iv43, i64 %indvars.iv40
40
  %i2 = load double, ptr %arrayidx23, align 8
41
  %add = fadd fast double %i2, %mul
42
  store double %add, ptr %arrayidx23, align 8
43
  %indvars.iv.next = add nuw nsw i64 %indvars.iv, 1
44
  %exitcond = icmp ne i64 %indvars.iv.next, 64
45
  br i1 %exitcond, label %for.body9, label %for.inc24
46

47
for.inc24:                                        ; preds = %for.body9
48
  %indvars.iv.next38 = add nuw nsw i64 %indvars.iv37, 1
49
  %exitcond39 = icmp ne i64 %indvars.iv.next38, 64
50
  br i1 %exitcond39, label %for.cond7.preheader, label %for.inc27
51

52
for.inc27:                                        ; preds = %for.inc24
53
  %indvars.iv.next41 = add nuw nsw i64 %indvars.iv40, 1
54
  %exitcond42 = icmp ne i64 %indvars.iv.next41, 1024
55
  br i1 %exitcond42, label %for.cond4.preheader, label %for.inc30
56

57
for.inc30:                                        ; preds = %for.inc27
58
  %indvars.iv.next44 = add nuw nsw i64 %indvars.iv43, 1
59
  %exitcond45 = icmp ne i64 %indvars.iv.next44, 1024
60
  br i1 %exitcond45, label %for.cond1.preheader, label %for.end32
61

62
for.end32:                                        ; preds = %for.inc30
63
  ret void
64
}
65