[llvm-bugs] [Bug 47194] New: Broadcasting the high word in AVX2
via llvm-bugs
llvm-bugs at lists.llvm.org
Sun Aug 16 17:50:59 PDT 2020
https://bugs.llvm.org/show_bug.cgi?id=47194
Bug ID: 47194
Summary: Broadcasting the high word in AVX2
Product: libraries
Version: trunk
Hardware: PC
OS: Linux
Status: NEW
Severity: enhancement
Priority: P
Component: Backend: X86
Assignee: unassignedbugs at nondot.org
Reporter: david.bolvansky at gmail.com
CC: craig.topper at gmail.com, llvm-bugs at lists.llvm.org,
llvm-dev at redking.me.uk, spatel+llvm at rotateright.com
From
https://stackoverflow.com/questions/57505111/broadcasting-the-high-word-in-avx2:
#include <immintrin.h>
__m256i bcast_mse(__m256i v) {
__m256i temp = _mm256_permute2x128_si256(v, v, 0x11);
__m256i ctrl = _mm256_set_epi8(
15, 14, 15, 14, 15, 14, 15, 14,
15, 14, 15, 14, 15, 14, 15, 14,
15, 14, 15, 14, 15, 14, 15, 14,
15, 14, 15, 14, 15, 14, 15, 14
);
return _mm256_shuffle_epi8(temp, ctrl);
}
Clang -O3 -mavx2:
bcast_mse(long long __vector(4)): # @bcast_mse(long long
__vector(4))
vpshufhw ymm0, ymm0, 255 # ymm0 =
ymm0[0,1,2,3,7,7,7,7,8,9,10,11,15,15,15,15]
vpshufd ymm0, ymm0, 170 # ymm0 = ymm0[2,2,2,2,6,6,6,6]
vpermq ymm0, ymm0, 170 # ymm0 = ymm0[2,2,2,2]
ret
GCC/ICC -O3 -mavx2:
bcast_mse(__m256i):
vperm2i128 ymm1, ymm0, ymm0, 17 #4.20
vpshufb ymm0, ymm1, YMMWORD PTR .L_2il0floatpacket.0[rip] #11.12
ret #11.12
.L_2il0floatpacket.0:
.long
0x0f0e0f0e,0x0f0e0f0e,0x0f0e0f0e,0x0f0e0f0e,0x0f0e0f0e,0x0f0e0f0e,0x0f0e0f0e,0x0f0e0f0e
Godbolt:
https://godbolt.org/z/cWb9Po
--
You are receiving this mail because:
You are on the CC list for the bug.
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.llvm.org/pipermail/llvm-bugs/attachments/20200817/9bf11192/attachment.html>
More information about the llvm-bugs
mailing list