File: neon-intrinsics.ll

package info (click to toggle)
llvm-toolchain-11 1%3A11.0.1-2
  • links: PTS, VCS
  • area: main
  • in suites: bullseye
  • size: 995,808 kB
  • sloc: cpp: 4,767,656; ansic: 760,916; asm: 477,436; python: 170,940; objc: 69,804; lisp: 29,914; sh: 23,855; f90: 18,173; pascal: 7,551; perl: 7,471; ml: 5,603; awk: 3,489; makefile: 2,573; xml: 915; cs: 573; fortran: 503; javascript: 452
file content (25 lines) | stat: -rw-r--r-- 1,190 bytes parent folder | download | duplicates (13)
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
; RUN: opt < %s -instcombine -S | FileCheck %s

; The alignment arguments for NEON load/store intrinsics can be increased
; by instcombine.  Check for this.

; CHECK: vld4.v2i32.p0i8({{.*}}, i32 32)
; CHECK: vst4.p0i8.v2i32({{.*}}, i32 16)

@x = common global [8 x i32] zeroinitializer, align 32
@y = common global [8 x i32] zeroinitializer, align 16

%struct.__neon_int32x2x4_t = type { <2 x i32>, <2 x i32>, <2 x i32>, <2 x i32> }

define void @test() nounwind ssp {
  %tmp1 = call %struct.__neon_int32x2x4_t @llvm.arm.neon.vld4.v2i32.p0i8(i8* bitcast ([8 x i32]* @x to i8*), i32 1)
  %tmp2 = extractvalue %struct.__neon_int32x2x4_t %tmp1, 0
  %tmp3 = extractvalue %struct.__neon_int32x2x4_t %tmp1, 1
  %tmp4 = extractvalue %struct.__neon_int32x2x4_t %tmp1, 2
  %tmp5 = extractvalue %struct.__neon_int32x2x4_t %tmp1, 3
  call void @llvm.arm.neon.vst4.p0i8.v2i32(i8* bitcast ([8 x i32]* @y to i8*), <2 x i32> %tmp2, <2 x i32> %tmp3, <2 x i32> %tmp4, <2 x i32> %tmp5, i32 1)
  ret void
}

declare %struct.__neon_int32x2x4_t @llvm.arm.neon.vld4.v2i32.p0i8(i8*, i32) nounwind readonly
declare void @llvm.arm.neon.vst4.p0i8.v2i32(i8*, <2 x i32>, <2 x i32>, <2 x i32>, <2 x i32>, i32) nounwind