Vectorise _mm_minpos_epu16 #551
Merged
Conversation
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Use vector instructions to find the index of the minimum element for
_mm_minpos_epu16. This avoids going through several comparisons and is
generally much faster when the minimum value is not just the first
element.
Example codegen for _mm_minpos_epu16(a) with GCC 11.2.0 (-O3)
Prior to commit:
uminv h1, v0.8h
sub sp, sp, 0x10
umov w1, v0.h[0]
str h1, [sp, 14]
ldrh w0, [sp, 14]
cmp w1, w0
b.eq 0x400aec
movi v1.4s, 0x0
ext v0.16b, v0.16b, v1.16b, 2
umov w1, v0.h[0]
cmp w1, w0
b.eq 0x400b10
ext v0.16b, v0.16b, v1.16b, 2
umov w1, v0.h[0]
cmp w1, w0
b.eq 0x400b18
ext v0.16b, v0.16b, v1.16b, 2
umov w1, v0.h[0]
cmp w1, w0
b.eq 0x400b20
ext v0.16b, v0.16b, v1.16b, 2
umov w1, v0.h[0]
cmp w1, w0
b.eq 0x400b28
ext v0.16b, v0.16b, v1.16b, 2
umov w1, v0.h[0]
cmp w1, w0
b.eq 0x400b30
ext v0.16b, v0.16b, v1.16b, 2
umov w1, v0.h[0]
cmp w1, w0
b.eq 0x400b38
ext v0.16b, v0.16b, v1.16b, 2
fmov s1, wzr
umov w1, v0.h[0]
cmp w1, w0
b.eq 0x400b40
movi v0.4s, 0x0
add x0, sp, 0xe
ld1 {v0.h}[0], [x0]
add sp, sp, 0x10
mov v0.h[1], v1.h[0]
ret
mov w0, 0x0
and w0, w0, 0xffff
movi v0.4s, 0x0
fmov s1, w0
add x0, sp, 0xe
ld1 {v0.h}[0], [x0]
add sp, sp, 0x10
mov v0.h[1], v1.h[0]
ret
mov w0, 0x1
b 0x400af0
mov w0, 0x2
b 0x400af0
mov w0, 0x3
b 0x400af0
mov w0, 0x4
b 0x400af0
mov w0, 0x5
b 0x400af0
mov w0, 0x6
b 0x400af0
mov w0, 0x7
After commit:
mov v2.16b, v0.16b
uminv h0, v0.8h
adrp x0, 0x400000
movi v3.4s, 0x0
ldr q4, [x0, 2864]
dup v1.8h, v0.h[0]
mov v3.h[0], v0.h[0]
cmeq v1.8h, v1.8h, v2.8h
mov v0.16b, v3.16b
orn v1.16b, v4.16b, v1.16b
uminv h1, v1.8h
mov v0.h[1], v1.h[0]
Co-authored-by: George Steed <george.steed@arm.com>
Co-authored-by: Aymen Qader <aymen.qader@arm.com>
howjmay
approved these changes
Nov 1, 2022
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
Use vector instructions to find the index of the minimum element for _mm_minpos_epu16. This avoids going through several comparisons and is generally much faster when the minimum value is not just the first element.
Codegen difference w/ GCC is included in the commit message.