https://llvm.org/bugs/show_bug.cgi?id=28136
Bug ID: 28136
Summary: Regression in x86 codegen using _mm256 intrinsic
Product: clang
Version: 3.7
Hardware: PC
OS: All
Status: NEW
Severity: normal
Priority: P
Component: LLVM Codegen
Assignee: unassignedclangb...@nondot.org
Reporter: deadal...@gmail.com
CC: llvm-bugs@lists.llvm.org
Classification: Unclassified
Sample code:
#include <x86intrin.h>
__m256i foo(__m256i a, __m256i b)
{
a = _mm256_unpacklo_epi8(a, b);
return _mm256_permute4x64_epi64(a, _MM_SHUFFLE(3, 1, 2, 0));
}
Assembly output on 3.6 :
foo(long long __vector(4), long long __vector(4)): #
@foo(long long __vector(4), long long __vector(4))
vpunpcklbw ymm0, ymm0, ymm1 # ymm0 =
ymm0[0],ymm1[0],ymm0[1],ymm1[1],ymm0[2],ymm1[2],ymm0[3],ymm1[3],ymm0[4],ymm1[4],ymm0[5],ymm1[5],ymm0[6],ymm1[6],ymm0[7],ymm1[7],ymm0[16],ymm1[16],ymm0[17],ymm1[17],ymm0[18],ymm1[18],ymm0[19],ymm1[19],ymm0[20],ymm1[20],ymm0[21],ymm1[21],ymm0[22],ymm1[22],ymm0[23],ymm1[23]
vpermq ymm0, ymm0, -40 # ymm0 = ymm0[0,2,1,3]
ret
Assembly output on 3.7 onward :
.LCPI0_0:
.zero 1
.zero 1
.zero 1
.zero 1
.zero 1
.zero 1
.zero 1
.zero 1
.byte 0 # 0x0
.zero 1
.byte 1 # 0x1
.zero 1
.byte 2 # 0x2
.zero 1
.byte 3 # 0x3
.zero 1
.byte 4 # 0x4
.zero 1
.byte 5 # 0x5
.zero 1
.byte 6 # 0x6
.zero 1
.byte 7 # 0x7
.zero 1
.zero 1
.zero 1
.zero 1
.zero 1
.zero 1
.zero 1
.zero 1
.zero 1
.LCPI0_1:
.zero 1
.zero 1
.zero 1
.zero 1
.zero 1
.zero 1
.zero 1
.zero 1
.zero 1
.byte 0 # 0x0
.zero 1
.byte 1 # 0x1
.zero 1
.byte 2 # 0x2
.zero 1
.byte 3 # 0x3
.zero 1
.byte 4 # 0x4
.zero 1
.byte 5 # 0x5
.zero 1
.byte 6 # 0x6
.zero 1
.byte 7 # 0x7
.zero 1
.zero 1
.zero 1
.zero 1
.zero 1
.zero 1
.zero 1
.zero 1
.LCPI0_2:
.byte 255 # 0xff
.byte 0 # 0x0
.byte 255 # 0xff
.byte 0 # 0x0
.byte 255 # 0xff
.byte 0 # 0x0
.byte 255 # 0xff
.byte 0 # 0x0
.byte 255 # 0xff
.byte 0 # 0x0
.byte 255 # 0xff
.byte 0 # 0x0
.byte 255 # 0xff
.byte 0 # 0x0
.byte 255 # 0xff
.byte 0 # 0x0
.byte 255 # 0xff
.byte 0 # 0x0
.byte 255 # 0xff
.byte 0 # 0x0
.byte 255 # 0xff
.byte 0 # 0x0
.byte 255 # 0xff
.byte 0 # 0x0
.byte 255 # 0xff
.byte 0 # 0x0
.byte 255 # 0xff
.byte 0 # 0x0
.byte 255 # 0xff
.byte 0 # 0x0
.byte 255 # 0xff
.byte 0 # 0x0
foo(long long __vector(4), long long __vector(4)): #
@foo(long long __vector(4), long long __vector(4))
vpunpcklbw ymm2, ymm0, ymm0 # ymm2 =
ymm0[0,0,1,1,2,2,3,3,4,4,5,5,6,6,7,7,16,16,17,17,18,18,19,19,20,20,21,21,22,22,23,23]
vperm2i128 ymm0, ymm0, ymm0, 35 # ymm0 = ymm0[2,3,0,1]
vpshufb ymm0, ymm0, ymmword ptr [rip + .LCPI0_0] # ymm0 =
ymm0[u,u,u,u,u,u,u,u,0,u,1,u,2,u,3,u,20,u,21,u,22,u,23,u,u,u,u,u,u,u,u,u]
vpblendd ymm0, ymm2, ymm0, 60 # ymm0 =
ymm2[0,1],ymm0[2,3,4,5],ymm2[6,7]
vpunpcklbw ymm2, ymm0, ymm1 # ymm2 =
ymm0[0],ymm1[0],ymm0[1],ymm1[1],ymm0[2],ymm1[2],ymm0[3],ymm1[3],ymm0[4],ymm1[4],ymm0[5],ymm1[5],ymm0[6],ymm1[6],ymm0[7],ymm1[7],ymm0[16],ymm1[16],ymm0[17],ymm1[17],ymm0[18],ymm1[18],ymm0[19],ymm1[19],ymm0[20],ymm1[20],ymm0[21],ymm1[21],ymm0[22],ymm1[22],ymm0[23],ymm1[23]
vperm2i128 ymm1, ymm0, ymm1, 35 # ymm1 = ymm1[2,3,0,1]
vpshufb ymm1, ymm1, ymmword ptr [rip + .LCPI0_1] # ymm1 =
ymm1[u,u,u,u,u,u,u,u,u,0,u,1,u,2,u,3,u,20,u,21,u,22,u,23,u,u,u,u,u,u,u,u]
vpblendd ymm1, ymm2, ymm1, 60 # ymm1 =
ymm2[0,1],ymm1[2,3,4,5],ymm2[6,7]
vmovdqa ymm2, ymmword ptr [rip + .LCPI0_2] # ymm2 =
[255,0,255,0,255,0,255,0,255,0,255,0,255,0,255,0,255,0,255,0,255,0,255,0,255,0,255,0,255,0,255,0]
vpblendvb ymm0, ymm1, ymm0, ymm2
ret
--
You are receiving this mail because:
You are on the CC list for the bug.
_______________________________________________
llvm-bugs mailing list
llvm-bugs@lists.llvm.org
http://lists.llvm.org/cgi-bin/mailman/listinfo/llvm-bugs
