b0045f5595
D100244 missed a check on the ResNo of the extract's operand 0 when finding a pair of extracts to combine into a VMOVRRD (extract(x, n); extract(x, n+1) -> VMOVRRD(extract x, n/2)). As a result, it can incorrectly pair an extract(x, n) with another extract(x:3, n+1) for example. This patch fixes the bug by adding the proper check on ResNo. Reviewed By: dmgreen Differential Revision: https://reviews.llvm.org/D111188
28 lines
1.1 KiB
LLVM
28 lines
1.1 KiB
LLVM
; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py
|
|
; RUN: llc -mtriple=armv7a-none-eabi -mattr=+neon %s -o - | FileCheck %s
|
|
|
|
; Check that the two extracts are not combined into a vmov.
|
|
|
|
%struct.__neon_int32x4x4_t = type { <4 x i32>, <4 x i32>, <4 x i32>, <4 x i32> }
|
|
|
|
define i32 @vld4Qi32(i32* %A) nounwind {
|
|
; CHECK-LABEL: vld4Qi32:
|
|
; CHECK: @ %bb.0:
|
|
; CHECK-NEXT: vld4.32 {d16, d18, d20, d22}, [r0]!
|
|
; CHECK-NEXT: vld4.32 {d17, d19, d21, d23}, [r0]
|
|
; CHECK-NEXT: vmov.32 r0, d18[1]
|
|
; CHECK-NEXT: vmov.32 r1, d16[0]
|
|
; CHECK-NEXT: add r0, r1, r0
|
|
; CHECK-NEXT: bx lr
|
|
%tmp0 = bitcast i32* %A to i8*
|
|
%tmp1 = call %struct.__neon_int32x4x4_t @llvm.arm.neon.vld4.v4i32.p0i8(i8* %tmp0, i32 1)
|
|
%tmp2 = extractvalue %struct.__neon_int32x4x4_t %tmp1, 0
|
|
%tmp3 = extractelement <4 x i32> %tmp2, i32 0
|
|
%tmp4 = extractvalue %struct.__neon_int32x4x4_t %tmp1, 1
|
|
%tmp5 = extractelement <4 x i32> %tmp4, i32 1
|
|
%tmp6 = add i32 %tmp3, %tmp5
|
|
ret i32 %tmp6
|
|
}
|
|
|
|
declare %struct.__neon_int32x4x4_t @llvm.arm.neon.vld4.v4i32.p0i8(i8*, i32) nounwind readonly
|