File: ymm-ordering.ll

package info (click to toggle)
llvm-toolchain-14 1%3A14.0.6-12
  • links: PTS, VCS
  • area: main
  • in suites: bookworm
  • size: 1,496,180 kB
  • sloc: cpp: 5,593,972; ansic: 986,872; asm: 585,869; python: 184,223; objc: 72,530; lisp: 31,119; f90: 27,793; javascript: 9,780; pascal: 9,762; sh: 9,482; perl: 7,468; ml: 5,432; awk: 3,523; makefile: 2,538; xml: 953; cs: 573; fortran: 567
file content (21 lines) | stat: -rw-r--r-- 767 bytes parent folder | download | duplicates (6) 
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
 
; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py
; RUN: llc < %s -mtriple=x86_64-linux -mattr=+avx | FileCheck %s

; TODO: We If we stored ymm0 before ymm1, then we could execute 2nd load and 1st store in
; parallel.
define void @test_01(i8* %src, i8* %dest) {
; CHECK-LABEL: test_01:
; CHECK:       # %bb.0: # %entry
; CHECK-NEXT:    vmovups (%rdi), %ymm0
; CHECK-NEXT:    vmovups 32(%rdi), %ymm1
; CHECK-NEXT:    vmovups %ymm1, 32(%rsi)
; CHECK-NEXT:    vmovups %ymm0, (%rsi)
; CHECK-NEXT:    vzeroupper
; CHECK-NEXT:    retq
entry:
  %read = bitcast i8* %src to <64 x i8>*
  %value = load <64 x i8>, <64 x i8>* %read, align 1
  %write = bitcast i8* %dest to <64 x i8>*
  store <64 x i8> %value, <64 x i8>* %write, align 1
  ret void
}