xref: /llvm-project/llvm/test/Transforms/SLPVectorizer/X86/cross_block_slp.ll (revision 580210a0c938531ef9fd79f9ffedb93eeb2e66c2)
1; NOTE: Assertions have been autogenerated by utils/update_test_checks.py
2; RUN: opt < %s -passes=slp-vectorizer,dce -S -mtriple=x86_64-apple-macosx10.8.0 -mcpu=corei7-avx | FileCheck %s
3
4target datalayout = "e-p:64:64:64-i1:8:8-i8:8:8-i16:16:16-i32:32:32-i64:64:64-f32:32:32-f64:64:64-v64:64:64-v128:128:128-a0:0:64-s0:64:64-f80:128:128-n8:16:32:64-S128"
5target triple = "x86_64-apple-macosx10.8.0"
6
7; int foo(ptr A, ptr B, int g) {
8;   float B0 = B[0];
9;   float B1 = B[1]; <----- BasicBlock #1
10;   B0 += 5;
11;   B1 += 8;
12;
13;   if (g) bar();
14;
15;   A[0] += B0;     <------- BasicBlock #3
16;   A[1] += B1;
17; }
18
19
20define i32 @foo(ptr nocapture %A, ptr nocapture %B, i32 %g) {
21; CHECK-LABEL: @foo(
22; CHECK-NEXT:  entry:
23; CHECK-NEXT:    [[TMP1:%.*]] = load <2 x float>, ptr [[B:%.*]], align 4
24; CHECK-NEXT:    [[TMP2:%.*]] = fadd <2 x float> [[TMP1]], <float 5.000000e+00, float 8.000000e+00>
25; CHECK-NEXT:    [[TOBOOL:%.*]] = icmp eq i32 [[G:%.*]], 0
26; CHECK-NEXT:    br i1 [[TOBOOL]], label [[IF_END:%.*]], label [[IF_THEN:%.*]]
27; CHECK:       if.then:
28; CHECK-NEXT:    [[CALL:%.*]] = tail call i32 (...) @bar()
29; CHECK-NEXT:    br label [[IF_END]]
30; CHECK:       if.end:
31; CHECK-NEXT:    [[TMP3:%.*]] = fpext <2 x float> [[TMP2]] to <2 x double>
32; CHECK-NEXT:    [[TMP5:%.*]] = load <2 x double>, ptr [[A:%.*]], align 8
33; CHECK-NEXT:    [[TMP6:%.*]] = fadd <2 x double> [[TMP3]], [[TMP5]]
34; CHECK-NEXT:    store <2 x double> [[TMP6]], ptr [[A]], align 8
35; CHECK-NEXT:    ret i32 undef
36;
37entry:
38  %0 = load float, ptr %B, align 4
39  %arrayidx1 = getelementptr inbounds float, ptr %B, i64 1
40  %1 = load float, ptr %arrayidx1, align 4
41  %add = fadd float %0, 5.000000e+00
42  %add2 = fadd float %1, 8.000000e+00
43  %tobool = icmp eq i32 %g, 0
44  br i1 %tobool, label %if.end, label %if.then
45
46if.then:
47  %call = tail call i32 (...) @bar()
48  br label %if.end
49
50if.end:
51  %conv = fpext float %add to double
52  %2 = load double, ptr %A, align 8
53  %add4 = fadd double %conv, %2
54  store double %add4, ptr %A, align 8
55  %conv5 = fpext float %add2 to double
56  %arrayidx6 = getelementptr inbounds double, ptr %A, i64 1
57  %3 = load double, ptr %arrayidx6, align 8
58  %add7 = fadd double %conv5, %3
59  store double %add7, ptr %arrayidx6, align 8
60  ret i32 undef
61}
62
63declare i32 @bar(...)
64