Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Jump to bottom

MultiShift, Masked shift and masked interleave #12

Open
wants to merge 106 commits into
base: master
Choose a base branch
from
Open

MultiShift, Masked shift and masked interleave #12

wants to merge 106 commits into from

Conversation

mazimkhan
Copy link
Member

@mazimkhan mazimkhan commented Nov 18, 2024 •
edited
Loading

Introduces the following operations for generic_ops-inl.h and arm_sve-inl.h where the use of sve intrinsics will provide a performance gain:

  • MultiShift - An arbitrary set of shifts performed on each byte in each 64-bit wide lane of a vector. A fuller description can be found in g3doc/quick_reference.md.
  • Masked shifts - shifting operations where a mask is applied for alternative behaviour where the mask is false.
  • Masked interleave operations which interleave either the even or odd elements of the two input vectors, unless the mask is false for a given lane. In this case, the lane is set to zero.

Tests have been introduced for all the listed operations.

malaterre and others added 30 commits May 30, 2024 13:57
@malaterre
Allow compilation of AVX2 on x86
216a960
@johnplatts
Added MinMagnitude and MaxMagnitude ops
620fec7
@lsrcz
add Get/Set for vectors and use them to implement Concat* operators
57a6c9b
@dependabot
Bump step-security/harden-runner from 2.10.1 to 2.10.2
bea0da5 
Bumps [step-security/harden-runner](https://github.com/step-security/harden-runner) from 2.10.1 to 2.10.2.
- [Release notes](https://github.com/step-security/harden-runner/releases)
- [Commits](step-security/harden-runner@91182cc...0080882)

---
updated-dependencies:
- dependency-name: step-security/harden-runner
  dependency-type: direct:production
  update-type: version-update:semver-patch
...

Signed-off-by: dependabot[bot] <[email protected]>
@copybara-github
Merge pull request google#2381 from google:dependabot/github_actions/…
5c7a693 
…step-security/harden-runner-2.10.2

PiperOrigin-RevId: 697981775
@Mousius
Revert to previous logic for GatherIndexN
3da3328
@jan-wassenberg @copybara-github
detect cache parameters
3cb3a91 
PiperOrigin-RevId: 698357330
@jan-wassenberg @copybara-github
replace non-test/trace fprintf with new hwy::Warn/HWY_WARN
8a0602d 
PiperOrigin-RevId: 698416280
@jan-wassenberg @copybara-github
topology fixes for M3
89b2678 
PiperOrigin-RevId: 698685120
@jan-wassenberg @copybara-github
Add topology support for Windows and Apple
d22ccd0 
Also reduce #if in mainline by factoring into helper functions, share code between Linux and FreeBSD
Use anon namespace instead of static

PiperOrigin-RevId: 699906189
@jan-wassenberg @copybara-github
add LSX/LASX targets. Refs google#2386
7e01a07 
PiperOrigin-RevId: 699936076
@copybara-github
Merge pull request google#2383 from Mousius:revert-gatherindexn
2b565e8 
PiperOrigin-RevId: 699943134
@ZequanWu @copybara-github
Remove VQSORT_SKIP workaround for compiler bug
02253c8 
The test passes with the most recent clang.

PiperOrigin-RevId: 700385250
@jan-wassenberg @copybara-github
Add BitsFromMask, promoting from detail::.
68b0fde 
Also split mask_test into mask_set_test,
remove unused overload in scalar,
modernize overloads (SFINAE instead of type tags).

PiperOrigin-RevId: 700701299
@jan-wassenberg @copybara-github
Add BitsFromMask, promoting from detail::.
bcf564e 
Also split mask_test into mask_set_test,
remove unused overload in scalar,
modernize overloads (SFINAE instead of type tags).

PiperOrigin-RevId: 700938851
@johnplatts
Added AVX10_2 and AVX10_2_512 targets
9b39ef2
@copybara-github
Merge pull request google#2395 from johnplatts:hwy_avx10_120124
b13a46f 
PiperOrigin-RevId: 701876987
@jan-wassenberg @copybara-github
(v2 of) Add BitsFromMask, promoting from detail::.
62c0a79 
Provide DFromM on all targets except SVE/RVV.
Also split mask_test into mask_set_test,
remove unused overload in scalar,
modernize overloads (SFINAE instead of type tags).
arm_sve required moving some sections earlier before their first usage.

PiperOrigin-RevId: 701919058
@johnplatts
Fix for RVV CMake detection if cross-compiling with Clang
07396f9
@johnplatts
Made changes to RVV Concat, Combine, ZeroExtendVector, and UpperHalf ops
914cb69
@copybara-github
Merge pull request google#2396 from johnplatts:hwy_rvv_cmake_fix_120224
9689e2c 
PiperOrigin-RevId: 702035129
@johnplatts
Enable tuples on RVV with Clang 17 or later
80839b5
@copybara-github
Merge pull request google#2397 from johnplatts:hwy_rvv_tuple_120224
fca5363 
PiperOrigin-RevId: 702229080
@dependabot
Bump actions/cache from 4.0.2 to 4.2.0
fccc82d 
Bumps [actions/cache](https://github.com/actions/cache) from 4.0.2 to 4.2.0.
- [Release notes](https://github.com/actions/cache/releases)
- [Changelog](https://github.com/actions/cache/blob/main/RELEASES.md)
- [Commits](actions/cache@0c45773...1bd1e32)

---
updated-dependencies:
- dependency-name: actions/cache
  dependency-type: direct:production
  update-type: version-update:semver-minor
...

Signed-off-by: dependabot[bot] <[email protected]>
@copybara-github
Merge pull request google#2401 from google:dependabot/github_actions/…
b9fa960 
…actions/cache-4.2.0

PiperOrigin-RevId: 703468488
@jan-wassenberg @copybara-github
add perf_counters
83b81ab 
PiperOrigin-RevId: 705134221
@jan-wassenberg @copybara-github
fix topology detection for some CPUs being offline (e.g. SMT off)
a076ade 
PiperOrigin-RevId: 705569288
@johnplatts
Fixes for ZSeries with GCC 9 or earlier or Clang 18 or earlier
eb4dc59
@copybara-github
Merge pull request google#2411 from johnplatts:jep_s390x_fix_121324
065050e 
PiperOrigin-RevId: 706272646
@jan-wassenberg @copybara-github
update thresholds for test failure
49674e1 
PiperOrigin-RevId: 706764322
wbb-ccl and others added 21 commits January 28, 2025 17:01
@wbb-ccl
Remove new ops that only have a generic implementation
83f183d
@copybara-github
Merge pull request google#2425 from cambridgeconsultants:cc_up_float_…
d547f91 
…operations

PiperOrigin-RevId: 720600683
@wbb-ccl
Fix bool_lanes typing
acd4f09
@copybara-github
Merge pull request google#2424 from cambridgeconsultants:cc_up_conver…
9cb44bc 
…t_promote_demote

PiperOrigin-RevId: 720897646
@mazimkhan @wbb-ccl
MulRound, MulLower and MulAddLower ops
cdd64d2
@wbb-ccl
Fix review comments
c264695 
Remove MulLower
Use MaskedMulOr instead
Replace MulAddLower with MaskedMulAddOr
@wbb-ccl
Remove redundant macro
ec5b0aa
@copybara-github
Merge pull request google#2429 from cambridgeconsultants:cc_up_mul_an…
b18b71d 
…d_arithmetic

PiperOrigin-RevId: 720957423
@johnplatts
Made changes to SVE_256 and SVE2_128 target detection
42feb2f
@mazimkhan @wbb-ccl
Masked compare and floating point classifications
df90905
@wbb-ccl
Fix review comments
41ff090 
Fix docs for masked floating-point classification
Rename MaskedComp* ops to Masked*
Remove redundant wrapper functions
@wbb-ccl
Remove MaskedIsInf & MaskedIsFinite
e8a840c
@mazimkhan @wbb-ccl
Load/Store, masked set and counting operations
ecb2f36
@copybara-github
Merge pull request google#2456 from johnplatts:hwy_sve_detect_enh_012925
2965981 
PiperOrigin-RevId: 721349271
@wbb-ccl
Fix review comments
6b90d90 
Rename SetOr* ops for consistency
Rename AllOnes/AllZeros to AllBits1/0
Remove MaskedLoadU, this is covered by MaskedLoad
Rename LowerHigher to InsertIntoUpper
Rework StoreTruncated, rename to TruncateStore
Rename macro arg
Avoid full-length load in LoadHigher
Optimise AllBits1
@wbb-ccl
Improve handling float_16 in TestInsertIntoUpper
4ef10e1
@wbb-ccl
Remove OrZero suffix
a74a04d
@wbb-ccl
Clarify AllBits0/1
6fe29f9
@wbb-ccl
Correct typo
ec0da10
@eustas @copybara-github
Add fake MulOdd to scalar
8e1476f 
That allows using `using` without extra `#if` guards.
`static_assert` disallows instantiation of this template.

PiperOrigin-RevId: 721414731
@copybara-github
Merge pull request google#2430 from cambridgeconsultants:cc_up_set_lo…
e96d4d3 
…ad_store_count_operations

PiperOrigin-RevId: 721470975
@wbb-ccl wbb-ccl force-pushed the cc_up_shift_and_interleave branch from 70d6581 to c4a44dd Compare January 31, 2025 11:01
copybara-github and others added 4 commits January 31, 2025 03:40
@copybara-github
Merge pull request google#2427 from cambridgeconsultants:cc_up_masked…
a74db0c 
…_compare

PiperOrigin-RevId: 721713013
@mazimkhan @wbb-ccl
MultiShift, Masked shift and masked interleave
f978890
@wbb-ccl
Fix review comments
5047921 
Remove OrZero suffixes
Remove masked interleaves. Will be in a future PR with x86 specialisations
Correct template naming
Correct multishift header comment
Optimise MultiShift
Add x86 TODOs for multishift
Improve MultiShift docs
@wbb-ccl
Correct incorrect op
1e907b0
@wbb-ccl wbb-ccl force-pushed the cc_up_shift_and_interleave branch from 37e7368 to 1e907b0 Compare January 31, 2025 12:21
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Reviewers
No reviews
Assignees
No one assigned
Labels
None yet
Projects
None yet
Milestone
No milestone
Development

Successfully merging this pull request may close these issues.