87bc91d425
This custom DAG combine works on a shuffle where one source vector is a zero splat, which means we can adjust the shuffle indices to refer to any element of the splat -- as long as we stay in the same vector. In the case where an undef (-1) index into the non-splat vector was used, we ended up adjusting the splat index to -1+NumElements, which points into the wrong vector. Fix this by using the first element from the splat if the other one is undef. There are four cases this theoretically affects, but in practice I only managed to demonstrate a miscompile with one of them. I think two of theses are effectively dead due to the operand canonicalization at the start of the transform. Fixes https://github.com/llvm/llvm-project/issues/77748.
101 lines
2.9 KiB
LLVM
101 lines
2.9 KiB
LLVM
; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py UTC_ARGS: --version 4
|
|
; RUN: llc -mtriple=powerpc64le-unknown-linux-gnu -mattr=+altivec -ppc-asm-full-reg-names < %s | FileCheck %s --check-prefixes=LE
|
|
; RUN: llc -mtriple=powerpc64-unknown-linux-gnu -mattr=+altivec,+direct-move -ppc-asm-full-reg-names < %s | FileCheck %s --check-prefixes=BE
|
|
|
|
; LE-LABEL: .LCPI0_0:
|
|
; LE-NEXT: .byte 31
|
|
; LE-NEXT: .byte 15
|
|
; LE-NEXT: .byte 30
|
|
; LE-NEXT: .byte 14
|
|
; LE-NEXT: .byte 29
|
|
; LE-NEXT: .byte 13
|
|
; LE-NEXT: .byte 28
|
|
; LE-NEXT: .byte 12
|
|
; LE-NEXT: .byte 31
|
|
; LE-NEXT: .byte 15
|
|
; LE-NEXT: .byte 31
|
|
; LE-NEXT: .byte 15
|
|
; LE-NEXT: .byte 31
|
|
; LE-NEXT: .byte 15
|
|
; LE-NEXT: .byte 31
|
|
; LE-NEXT: .byte 15
|
|
|
|
; BE-LABEL: .LCPI0_0:
|
|
; BE-NEXT: .byte 0
|
|
; BE-NEXT: .byte 16
|
|
; BE-NEXT: .byte 1
|
|
; BE-NEXT: .byte 16
|
|
; BE-NEXT: .byte 2
|
|
; BE-NEXT: .byte 16
|
|
; BE-NEXT: .byte 3
|
|
; BE-NEXT: .byte 16
|
|
; BE-NEXT: .byte 0
|
|
; BE-NEXT: .byte 16
|
|
; BE-NEXT: .byte 0
|
|
; BE-NEXT: .byte 16
|
|
; BE-NEXT: .byte 0
|
|
; BE-NEXT: .byte 16
|
|
; BE-NEXT: .byte 0
|
|
; BE-NEXT: .byte 16
|
|
|
|
define <16 x i8> @test1(<16 x i8> %a) {
|
|
; LE-LABEL: test1:
|
|
; LE: # %bb.0:
|
|
; LE-NEXT: addis r3, r2, .LCPI0_0@toc@ha
|
|
; LE-NEXT: xxlxor vs36, vs36, vs36
|
|
; LE-NEXT: addi r3, r3, .LCPI0_0@toc@l
|
|
; LE-NEXT: lxvd2x vs0, 0, r3
|
|
; LE-NEXT: xxswapd vs35, vs0
|
|
; LE-NEXT: vperm v2, v4, v2, v3
|
|
; LE-NEXT: blr
|
|
;
|
|
; BE-LABEL: test1:
|
|
; BE: # %bb.0:
|
|
; BE-NEXT: addis r3, r2, .LCPI0_0@toc@ha
|
|
; BE-NEXT: xxlxor vs36, vs36, vs36
|
|
; BE-NEXT: addi r3, r3, .LCPI0_0@toc@l
|
|
; BE-NEXT: lxvw4x vs35, 0, r3
|
|
; BE-NEXT: vperm v2, v2, v4, v3
|
|
; BE-NEXT: blr
|
|
%shuffle = shufflevector <16 x i8> %a, <16 x i8> zeroinitializer, <16 x i32> <i32 0, i32 16, i32 1, i32 16, i32 2, i32 16, i32 3, i32 16, i32 poison, i32 16, i32 poison, i32 16, i32 poison, i32 16, i32 poison, i32 16>
|
|
ret <16 x i8> %shuffle
|
|
}
|
|
|
|
; LE-LABEL: .LCPI1_0:
|
|
; LE-NEXT: .byte 31
|
|
; LE-NEXT: .byte 15
|
|
; LE-NEXT: .byte 31
|
|
; LE-NEXT: .byte 14
|
|
; LE-NEXT: .byte 31
|
|
; LE-NEXT: .byte 13
|
|
; LE-NEXT: .byte 31
|
|
; LE-NEXT: .byte 12
|
|
; LE-NEXT: .byte 31
|
|
; LE-NEXT: .byte 11
|
|
; LE-NEXT: .byte 31
|
|
; LE-NEXT: .byte 10
|
|
; LE-NEXT: .byte 31
|
|
; LE-NEXT: .byte 9
|
|
; LE-NEXT: .byte 31
|
|
; LE-NEXT: .byte 8
|
|
|
|
define <16 x i8> @test2(<16 x i8> %a) {
|
|
; LE-LABEL: test2:
|
|
; LE: # %bb.0:
|
|
; LE-NEXT: addis r3, r2, .LCPI1_0@toc@ha
|
|
; LE-NEXT: xxlxor vs36, vs36, vs36
|
|
; LE-NEXT: addi r3, r3, .LCPI1_0@toc@l
|
|
; LE-NEXT: lxvd2x vs0, 0, r3
|
|
; LE-NEXT: xxswapd vs35, vs0
|
|
; LE-NEXT: vperm v2, v2, v4, v3
|
|
; LE-NEXT: blr
|
|
;
|
|
; BE-LABEL: test2:
|
|
; BE: # %bb.0:
|
|
; BE-NEXT: xxlxor vs35, vs35, vs35
|
|
; BE-NEXT: vmrghb v2, v3, v2
|
|
; BE-NEXT: blr
|
|
%shuffle = shufflevector <16 x i8> %a, <16 x i8> zeroinitializer, <16 x i32> <i32 16, i32 0, i32 16, i32 1, i32 16, i32 2, i32 16, i32 3, i32 16, i32 4, i32 poison, i32 5, i32 poison, i32 6, i32 poison, i32 7>
|
|
ret <16 x i8> %shuffle
|
|
}
|