RPCS3/llvm-mirror
1
0
Fork 0
mirror of https://github.com/RPCS3/llvm-mirror.git synced 2024-11-23 19:23:23 +01:00
Code Issues Projects Releases Wiki Activity

[X86][SSE] Add more non-uniform exact sdiv vector tests covering all/none ashr paths

Browse Source 
llvm-svn: 339120
This commit is contained in:
Simon Pilgrim 2018-08-07 09:31:22 +00:00
parent 1017c28ba5
commit 90a599055d
1 changed files with 94 additions and 0 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

94
test/CodeGen/X86/sdiv-exact.ll
View File

@ -123,3 +123,97 @@ define <4 x i32> @test5(<4 x i32> %x) {
%div = sdiv exact <4 x i32> %x, <i32 24, i32 24, i32 25, i32 25>
ret <4 x i32> %div
}
define <4 x i32> @test6(<4 x i32> %x) {
; X86-LABEL: test6:
; X86: # %bb.0:
; X86-NEXT: pshufd {{.*#+}} xmm1 = xmm0[3,1,2,3]
; X86-NEXT: movd %xmm1, %eax
; X86-NEXT: sarl %eax
; X86-NEXT: imull $-991146299, %eax, %eax # imm = 0xC4EC4EC5
; X86-NEXT: movd %eax, %xmm1
; X86-NEXT: pshufd {{.*#+}} xmm2 = xmm0[2,3,0,1]
; X86-NEXT: movd %xmm2, %eax
; X86-NEXT: sarl %eax
; X86-NEXT: imull $-991146299, %eax, %eax # imm = 0xC4EC4EC5
; X86-NEXT: movd %eax, %xmm2
; X86-NEXT: punpckldq {{.*#+}} xmm2 = xmm2[0],xmm1[0],xmm2[1],xmm1[1]
; X86-NEXT: movd %xmm0, %eax
; X86-NEXT: sarl $3, %eax
; X86-NEXT: imull $-1431655765, %eax, %eax # imm = 0xAAAAAAAB
; X86-NEXT: movd %eax, %xmm1
; X86-NEXT: pshufd {{.*#+}} xmm0 = xmm0[1,1,2,3]
; X86-NEXT: movd %xmm0, %eax
; X86-NEXT: sarl $3, %eax
; X86-NEXT: imull $-1431655765, %eax, %eax # imm = 0xAAAAAAAB
; X86-NEXT: movd %eax, %xmm0
; X86-NEXT: punpckldq {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
; X86-NEXT: punpcklqdq {{.*#+}} xmm1 = xmm1[0],xmm2[0]
; X86-NEXT: movdqa %xmm1, %xmm0
; X86-NEXT: retl
;
; X64-LABEL: test6:
; X64: # %bb.0:
; X64-NEXT: vpextrd $1, %xmm0, %eax
; X64-NEXT: sarl $3, %eax
; X64-NEXT: imull $-1431655765, %eax, %eax # imm = 0xAAAAAAAB
; X64-NEXT: vmovd %xmm0, %ecx
; X64-NEXT: sarl $3, %ecx
; X64-NEXT: imull $-1431655765, %ecx, %ecx # imm = 0xAAAAAAAB
; X64-NEXT: vmovd %ecx, %xmm1
; X64-NEXT: vpinsrd $1, %eax, %xmm1, %xmm1
; X64-NEXT: vpextrd $2, %xmm0, %eax
; X64-NEXT: sarl %eax
; X64-NEXT: imull $-991146299, %eax, %eax # imm = 0xC4EC4EC5
; X64-NEXT: vpinsrd $2, %eax, %xmm1, %xmm1
; X64-NEXT: vpextrd $3, %xmm0, %eax
; X64-NEXT: sarl %eax
; X64-NEXT: imull $-991146299, %eax, %eax # imm = 0xC4EC4EC5
; X64-NEXT: vpinsrd $3, %eax, %xmm1, %xmm0
; X64-NEXT: retq
%div = sdiv exact <4 x i32> %x, <i32 24, i32 24, i32 26, i32 26>
ret <4 x i32> %div
}
define <4 x i32> @test7(<4 x i32> %x) {
; X86-LABEL: test7:
; X86: # %bb.0:
; X86-NEXT: pshufd {{.*#+}} xmm1 = xmm0[3,1,2,3]
; X86-NEXT: movd %xmm1, %eax
; X86-NEXT: imull $1749801491, %eax, %eax # imm = 0x684BDA13
; X86-NEXT: movd %eax, %xmm1
; X86-NEXT: pshufd {{.*#+}} xmm2 = xmm0[2,3,0,1]
; X86-NEXT: movd %xmm2, %eax
; X86-NEXT: imull $1749801491, %eax, %eax # imm = 0x684BDA13
; X86-NEXT: movd %eax, %xmm2
; X86-NEXT: punpckldq {{.*#+}} xmm2 = xmm2[0],xmm1[0],xmm2[1],xmm1[1]
; X86-NEXT: movd %xmm0, %eax
; X86-NEXT: imull $-1030792151, %eax, %eax # imm = 0xC28F5C29
; X86-NEXT: movd %eax, %xmm1
; X86-NEXT: pshufd {{.*#+}} xmm0 = xmm0[1,1,2,3]
; X86-NEXT: movd %xmm0, %eax
; X86-NEXT: imull $-1030792151, %eax, %eax # imm = 0xC28F5C29
; X86-NEXT: movd %eax, %xmm0
; X86-NEXT: punpckldq {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
; X86-NEXT: punpcklqdq {{.*#+}} xmm1 = xmm1[0],xmm2[0]
; X86-NEXT: movdqa %xmm1, %xmm0
; X86-NEXT: retl
;
; X64-LABEL: test7:
; X64: # %bb.0:
; X64-NEXT: vpextrd $1, %xmm0, %eax
; X64-NEXT: imull $-1030792151, %eax, %eax # imm = 0xC28F5C29
; X64-NEXT: vmovd %xmm0, %ecx
; X64-NEXT: imull $-1030792151, %ecx, %ecx # imm = 0xC28F5C29
; X64-NEXT: vmovd %ecx, %xmm1
; X64-NEXT: vpinsrd $1, %eax, %xmm1, %xmm1
; X64-NEXT: vpextrd $2, %xmm0, %eax
; X64-NEXT: imull $1749801491, %eax, %eax # imm = 0x684BDA13
; X64-NEXT: vpinsrd $2, %eax, %xmm1, %xmm1
; X64-NEXT: vpextrd $3, %xmm0, %eax
; X64-NEXT: imull $1749801491, %eax, %eax # imm = 0x684BDA13
; X64-NEXT: vpinsrd $3, %eax, %xmm1, %xmm0
; X64-NEXT: retq
%div = sdiv exact <4 x i32> %x, <i32 25, i32 25, i32 27, i32 27>
ret <4 x i32> %div
}