1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
|
; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py
; RUN: llc < %s -mtriple=x86_64-apple-macosx10.15 | FileCheck %s
%0 = type <{ half }>
%1 = type <{ <16 x half> }>
define void @foo(ptr %0, ptr %1, ptr %2) #0 {
; CHECK-LABEL: foo:
; CHECK: ## %bb.0:
; CHECK-NEXT: pushq %rbp
; CHECK-NEXT: movq %rsp, %rbp
; CHECK-NEXT: andq $-32, %rsp
; CHECK-NEXT: subq $64, %rsp
; CHECK-NEXT: movl (%rdx), %eax
; CHECK-NEXT: andl $15, %eax
; CHECK-NEXT: vpinsrw $0, (%rdi), %xmm0, %xmm0
; CHECK-NEXT: vmovups (%rsi), %ymm1
; CHECK-NEXT: vmovaps %ymm1, (%rsp)
; CHECK-NEXT: vpextrw $0, %xmm0, (%rsp,%rax,2)
; CHECK-NEXT: vmovaps (%rsp), %ymm0
; CHECK-NEXT: vmovups %ymm0, (%rsi)
; CHECK-NEXT: movq %rbp, %rsp
; CHECK-NEXT: popq %rbp
; CHECK-NEXT: vzeroupper
; CHECK-NEXT: retq
%4 = bitcast ptr %2 to ptr
%5 = load i64, ptr %4, align 8
%6 = getelementptr inbounds %0, ptr %0, i64 0, i32 0
%7 = load half, ptr %6, align 2
%8 = getelementptr inbounds %1, ptr %1, i64 0, i32 0
%9 = load <16 x half>, ptr %8, align 16
%10 = trunc i64 %5 to i32
%11 = insertelement <16 x half> %9, half %7, i32 %10
store <16 x half> %11, ptr %8, align 16
ret void
}
attributes #0 = { nounwind "target-features"="+f16c" }
|
