File: ymm-ordering.ll

package info (click to toggle)
llvm-toolchain-13 1%3A13.0.1-6~deb11u1
  • links: PTS, VCS
  • area: main
  • in suites: bullseye
  • size: 1,418,812 kB
  • sloc: cpp: 5,290,827; ansic: 996,570; asm: 544,593; python: 188,212; objc: 72,027; lisp: 30,291; f90: 25,395; sh: 24,900; javascript: 9,780; pascal: 9,398; perl: 7,484; ml: 5,432; awk: 3,523; makefile: 2,892; xml: 953; cs: 573; fortran: 539
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
}