1; RUN: llc < %s -march=x86-64 -mcpu=bdver1 | FileCheck %s
2
3; clang -Oz -c test1.cpp -emit-llvm -S -o
4; Verify that we generate shld insruction when we are optimizing for size,
5; even for X86_64 processors that are known to have poor latency double
6; precision shift instructions.
7; uint64_t lshift10(uint64_t a, uint64_t b)
8; {
9;     return (a << 10) | (b >> 54);
10; }
11
12; Function Attrs: minsize nounwind readnone uwtable
13define i64 @_Z8lshift10mm(i64 %a, i64 %b) #0 {
14entry:
15; CHECK:   shldq   $10
16  %shl = shl i64 %a, 10
17  %shr = lshr i64 %b, 54
18  %or = or i64 %shr, %shl
19  ret i64 %or
20}
21
22attributes #0 = { minsize nounwind readnone uwtable "less-precise-fpmad"="false" "no-frame-pointer-elim"="false" "no-infs-fp-math"="false" "no-nans-fp-math"="false" "stack-protector-buffer-size"="8" "unsafe-fp-math"="false" "use-soft-float"="false" }
23
24
25; clang -Os -c test2.cpp -emit-llvm -S
26; Verify that we generate shld insruction when we are optimizing for size,
27; even for X86_64 processors that are known to have poor latency double
28; precision shift instructions.
29; uint64_t lshift11(uint64_t a, uint64_t b)
30; {
31;     return (a << 11) | (b >> 53);
32; }
33
34; Function Attrs: nounwind optsize readnone uwtable
35define i64 @_Z8lshift11mm(i64 %a, i64 %b) #1 {
36entry:
37; CHECK:   shldq   $11
38  %shl = shl i64 %a, 11
39  %shr = lshr i64 %b, 53
40  %or = or i64 %shr, %shl
41  ret i64 %or
42}
43
44attributes #1 = { nounwind optsize readnone uwtable "less-precise-fpmad"="false" "no-frame-pointer-elim"="false" "no-infs-fp-math"="false" "no-nans-fp-math"="false" "stack-protector-buffer-size"="8" "unsafe-fp-math"="false" "use-soft-float"="false" }
45
46; clang -O2 -c test2.cpp -emit-llvm -S
47; Verify that we do not generate shld insruction when we are not optimizing
48; for size for X86_64 processors that are known to have poor latency double
49; precision shift instructions.
50; uint64_t lshift12(uint64_t a, uint64_t b)
51; {
52;     return (a << 12) | (b >> 52);
53; }
54
55; Function Attrs: nounwind optsize readnone uwtable
56define i64 @_Z8lshift12mm(i64 %a, i64 %b) #2 {
57entry:
58; CHECK:       shlq    $12
59; CHECK-NEXT:  shrq    $52
60  %shl = shl i64 %a, 12
61  %shr = lshr i64 %b, 52
62  %or = or i64 %shr, %shl
63  ret i64 %or
64}
65
66attributes #2= { nounwind readnone uwtable "less-precise-fpmad"="false" "no-frame-pointer-elim"="false" "no-infs-fp-math"="false" "no-nans-fp-math"="false" "stack-protector-buffer-size"="8" "unsafe-fp-math"="false" "use-soft-float"="false" }
67
68