File: vectorize_i16.ll

package info (click to toggle)
llvm-toolchain-20 1%3A20.1.6-1~exp1
  • links: PTS, VCS
  • area: main
  • in suites: experimental
  • size: 2,111,304 kB
  • sloc: cpp: 7,438,677; ansic: 1,393,822; asm: 1,012,926; python: 241,650; f90: 86,635; objc: 75,479; lisp: 42,144; pascal: 17,286; sh: 10,027; ml: 5,082; perl: 4,730; awk: 3,523; makefile: 3,349; javascript: 2,251; xml: 892; fortran: 672
file content (17 lines) | stat: -rw-r--r-- 473 bytes parent folder | download | duplicates (12) 
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
 
; RUN: opt -mtriple=nvptx64-nvidia-cuda -passes=load-store-vectorizer -S -o - %s | FileCheck %s

; CHECK-LABEL: @int16x2
; CHECK: load <2 x i16>
; CHECK: store <2 x i16>
define void @int16x2(ptr nocapture align 4 %ptr) {
  %ptr0 = getelementptr i16, ptr %ptr, i64 0
  %ptr1 = getelementptr i16, ptr %ptr, i64 1

  %l0 = load i16, ptr %ptr0, align 4
  %l1 = load i16, ptr %ptr1, align 2

  store i16 %l1, ptr %ptr0, align 4
  store i16 %l0, ptr %ptr1, align 2

  ret void
}