[SIMD] Intel support for swizzle and shuffle and fix extract_lane and replace_lane #7125

hyjorc1 · 2022-12-04T02:59:10Z

`cef394f`

[SIMD] Intel support for swizzle and shuffle and fix extract_lane and replace_lane
https://bugs.webkit.org/show_bug.cgi?id=248728
rdar://102942396

Reviewed by Yusuke Suzuki.

This patch aims for three tasks:

    1. Add WASM SIMD operations `swizzle` and `shuffle`.
    https://github.com/WebAssembly/simd/blob/main/proposals/simd/SIMD.md#shuffling-using-immediate-indices

    2. Refactor MacroAssemblerX86_64.h and X86Assembler.h for `pinsr` and `pextr`,
    where X86Assembler should do simple instruction emission only. And MacroAssemblerX86
    should select the instruction.

    3. Fix WASM SIMD operation `replace_lane`. Previously, the operation is
    implemented with instruction `pinsr` in AVX format but only passing three parameters
    which is wrong.
        https://www.felixcloutier.com/x86/pinsrb:pinsrd:pinsrq
        https://www.felixcloutier.com/x86/pinsrw
        https://www.officedaytime.com/simd512e/simdimg/si.php?f=pinsrb
        https://www.officedaytime.com/simd512e/simdimg/si.php?f=pinsrw
        https://www.officedaytime.com/simd512e/simdimg/si.php?f=pinsrd
        https://www.officedaytime.com/simd512e/simdimg/si.php?f=pinsrq

* Source/JavaScriptCore/assembler/MacroAssemblerX86_64.h:
(JSC::MacroAssemblerX86_64::vectorReplaceLane):
(JSC::MacroAssemblerX86_64::vectorExtractLane):
(JSC::MacroAssemblerX86_64::vectorSwizzle):
* Source/JavaScriptCore/assembler/X86Assembler.h:
(JSC::X86Assembler::pinsrb):
(JSC::X86Assembler::pinsrw):
(JSC::X86Assembler::pinsrd):
(JSC::X86Assembler::pinsrq):
(JSC::X86Assembler::insertps):
(JSC::X86Assembler::unpcklpd):
(JSC::X86Assembler::vpextrb):
(JSC::X86Assembler::vpextrw):
(JSC::X86Assembler::vpextrd):
(JSC::X86Assembler::vpextrq):
(JSC::X86Assembler::X86InstructionFormatter::SingleInstructionBufferWriter::memoryModRM):
(JSC::X86Assembler::pinsr): Deleted.
(JSC::X86Assembler::pextr): Deleted.
(JSC::X86Assembler::vextractps): Deleted.
* Source/JavaScriptCore/b3/air/AirOpcode.opcodes:
* Source/JavaScriptCore/wasm/WasmAirIRGenerator.cpp:
(JSC::Wasm::AirIRGenerator::addReplaceLane):

Canonical link: https://commits.webkit.org/257400@main

dea943f

Misc	iOS, tvOS & watchOS	macOS	Linux	Windows
❌ 🧪 style	✅ 🛠 ios	✅ 🛠 mac	✅ 🛠 wpe	✅ 🛠 🧪 win
	✅ 🛠 ios-sim	✅ 🛠 mac-AS-debug	✅ 🛠 gtk	✅ 🛠 wincairo
✅ 🧪 webkitperl	~~🧪 ios-wk2~~	✅ 🧪 api-mac	~~🧪 gtk-wk2~~
	~~🧪 api-ios~~	~~🧪 mac-wk1~~	~~🧪 api-gtk~~
✅ 🛠 🧪 jsc	✅ 🛠 tv	~~🧪 mac-wk2~~	✅ 🛠 jsc-armv7
	✅ 🛠 tv-sim	~~🧪 mac-AS-debug-wk2~~	~~🧪 jsc-armv7-tests~~
✅ 🛠 🧪 merge	✅ 🛠 watch	✅ 🧪 mac-wk2-stress	✅ 🛠 jsc-mips
	✅ 🛠 watch-sim		✅ 🧪 jsc-mips-tests

webkit-early-warning-system · 2022-12-04T02:59:33Z

EWS run on previous version of this PR (hash 829c9e4)

Misc	iOS, tvOS & watchOS	macOS	Linux	Windows
✅ 🧪 style	✅ 🛠 ios	✅ 🛠 mac	✅ 🛠 wpe	✅ 🛠 🧪 win
	✅ 🛠 ios-sim	✅ 🛠 mac-AS-debug	✅ 🛠 gtk	✅ 🛠 wincairo
✅ 🧪 webkitperl	✅ 🧪 ios-wk2	✅ 🧪 api-mac	✅ 🧪 gtk-wk2
	✅ 🧪 api-ios	✅ 🧪 mac-wk1	✅ 🧪 api-gtk
✅ 🛠 🧪 jsc	✅ 🛠 tv	❌ 🧪 mac-wk2	✅ 🛠 jsc-armv7
	✅ 🛠 tv-sim	❌ 🧪 mac-AS-debug-wk2	✅ 🧪 jsc-armv7-tests
	✅ 🛠 watch	✅ 🧪 mac-wk2-stress	✅ 🛠 jsc-mips
	✅ 🛠 watch-sim		✅ 🧪 jsc-mips-tests

Source/JavaScriptCore/assembler/MacroAssemblerX86_64.h

Source/JavaScriptCore/assembler/X86Assembler.h

webkit-early-warning-system · 2022-12-05T04:05:36Z

EWS run on previous version of this PR (hash 0bb11d6)

Misc	iOS, tvOS & watchOS	macOS	Linux	Windows
❌ 🧪 style	✅ 🛠 ios	✅ 🛠 mac	✅ 🛠 wpe	✅ 🛠 🧪 win
	✅ 🛠 ios-sim	✅ 🛠 mac-AS-debug	✅ 🛠 gtk	✅ 🛠 wincairo
✅ 🧪 webkitperl	❌ 🧪 ios-wk2	✅ 🧪 api-mac	✅ 🧪 gtk-wk2
	✅ 🧪 api-ios	✅ 🧪 mac-wk1	✅ 🧪 api-gtk
✅ 🛠 🧪 jsc	✅ 🛠 tv	✅ 🧪 mac-wk2	✅ 🛠 jsc-armv7
	✅ 🛠 tv-sim	❌ 🧪 mac-AS-debug-wk2	✅ 🧪 jsc-armv7-tests
	✅ 🛠 watch	✅ 🧪 mac-wk2-stress	✅ 🛠 jsc-mips
	✅ 🛠 watch-sim		✅ 🧪 jsc-mips-tests

webkit-early-warning-system · 2022-12-06T06:19:07Z

Starting EWS tests for 2e2be09. Live statuses available at the PR page, #7125

webkit-early-warning-system · 2022-12-06T06:19:07Z

EWS run on previous version of this PR (hash 2e2be09)

Misc	iOS, tvOS & watchOS	macOS	Linux	Windows
❌ 🧪 style	✅ 🛠 ios	✅ 🛠 mac	✅ 🛠 wpe	~~🛠 🧪 win~~
	✅ 🛠 ios-sim	✅ 🛠 mac-AS-debug	✅ 🛠 gtk	⏳ 🛠 wincairo
✅ 🧪 webkitperl	~~🧪 ios-wk2~~	~~🧪 api-mac~~	~~🧪 gtk-wk2~~
	~~🧪 api-ios~~	~~🧪 mac-wk1~~	~~🧪 api-gtk~~
✅ 🛠 🧪 jsc	✅ 🛠 tv	~~🧪 mac-wk2~~	✅ 🛠 jsc-armv7
	✅ 🛠 tv-sim	~~🧪 mac-AS-debug-wk2~~	~~🧪 jsc-armv7-tests~~
	✅ 🛠 watch	~~🧪 mac-wk2-stress~~	✅ 🛠 jsc-mips
	✅ 🛠 watch-sim		✅ 🧪 jsc-mips-tests

webkit-early-warning-system · 2022-12-06T06:52:31Z

EWS run on previous version of this PR (hash b063c3c)

Misc	iOS, tvOS & watchOS	macOS	Linux	Windows
❌ 🧪 style	✅ 🛠 ios	⏳ 🛠 mac	✅ 🛠 wpe	~~🛠 🧪 win~~
	✅ 🛠 ios-sim	⏳ 🛠 mac-AS-debug	✅ 🛠 gtk	⏳ 🛠 wincairo
~~🧪 webkitperl~~	⏳ 🧪 ios-wk2	⏳ 🧪 api-mac	~~🧪 gtk-wk2~~
	⏳ 🧪 api-ios	⏳ 🧪 mac-wk1	~~🧪 api-gtk~~
⏳ 🛠 🧪 jsc	~~🛠 tv~~	⏳ 🧪 mac-wk2	✅ 🛠 jsc-armv7
	✅ 🛠 tv-sim	⏳ 🧪 mac-AS-debug-wk2	⏳ 🧪 jsc-armv7-tests
	~~🛠 watch~~	⏳ 🧪 mac-wk2-stress	✅ 🛠 jsc-mips
	✅ 🛠 watch-sim		⏳ 🧪 jsc-mips-tests

webkit-early-warning-system · 2022-12-06T06:57:27Z

EWS run on previous version of this PR (hash 2c999a9)

Misc	iOS, tvOS & watchOS	macOS	Linux	Windows
❌ 🧪 style	✅ 🛠 ios	✅ 🛠 mac	✅ 🛠 wpe	~~🛠 🧪 win~~
	✅ 🛠 ios-sim	✅ 🛠 mac-AS-debug	✅ 🛠 gtk	⏳ 🛠 wincairo
✅ 🧪 webkitperl	⏳ 🧪 ios-wk2	~~🧪 api-mac~~	~~🧪 gtk-wk2~~
	⏳ 🧪 api-ios	~~🧪 mac-wk1~~	⏳ 🧪 api-gtk
✅ 🛠 🧪 jsc	✅ 🛠 tv	~~🧪 mac-wk2~~	✅ 🛠 jsc-armv7
	✅ 🛠 tv-sim	~~🧪 mac-AS-debug-wk2~~	⏳ 🧪 jsc-armv7-tests
	✅ 🛠 watch	~~🧪 mac-wk2-stress~~	✅ 🛠 jsc-mips
	✅ 🛠 watch-sim		✅ 🧪 jsc-mips-tests

Constellation

r=me

Source/JavaScriptCore/assembler/MacroAssemblerX86_64.h

Source/JavaScriptCore/wasm/WasmAirIRGenerator.cpp

webkit-early-warning-system · 2022-12-06T07:29:52Z

EWS run on previous version of this PR (hash 8f5cea8)

Misc	iOS, tvOS & watchOS	macOS	Linux	Windows
❌ 🧪 style	✅ 🛠 ios	~~🛠 mac~~	✅ 🛠 wpe	~~🛠 🧪 win~~
	✅ 🛠 ios-sim	~~🛠 mac-AS-debug~~	~~🛠 gtk~~	~~🛠 wincairo~~
✅ 🧪 webkitperl	~~🧪 ios-wk2~~	~~🧪 api-mac~~	~~🧪 gtk-wk2~~
	~~🧪 api-ios~~	~~🧪 mac-wk1~~	~~🧪 api-gtk~~
~~🛠 🧪 jsc~~	✅ 🛠 tv	~~🧪 mac-wk2~~	~~🛠 jsc-armv7~~
	✅ 🛠 tv-sim	~~🧪 mac-AS-debug-wk2~~	~~🧪 jsc-armv7-tests~~
~~🛠 🧪 merge~~	✅ 🛠 watch	~~🧪 mac-wk2-stress~~	✅ 🛠 jsc-mips
	✅ 🛠 watch-sim		~~🧪 jsc-mips-tests~~

webkit-early-warning-system · 2022-12-06T07:40:58Z

Starting EWS tests for dea943f. Live statuses available at the PR page, #7125

webkit-early-warning-system · 2022-12-06T07:41:00Z

EWS run on current version of this PR (hash dea943f)

Misc	iOS, tvOS & watchOS	macOS	Linux	Windows
❌ 🧪 style	✅ 🛠 ios	✅ 🛠 mac	✅ 🛠 wpe	✅ 🛠 🧪 win
	✅ 🛠 ios-sim	✅ 🛠 mac-AS-debug	✅ 🛠 gtk	✅ 🛠 wincairo
✅ 🧪 webkitperl	~~🧪 ios-wk2~~	✅ 🧪 api-mac	~~🧪 gtk-wk2~~
	~~🧪 api-ios~~	~~🧪 mac-wk1~~	~~🧪 api-gtk~~
✅ 🛠 🧪 jsc	✅ 🛠 tv	~~🧪 mac-wk2~~	✅ 🛠 jsc-armv7
	✅ 🛠 tv-sim	~~🧪 mac-AS-debug-wk2~~	~~🧪 jsc-armv7-tests~~
✅ 🛠 🧪 merge	✅ 🛠 watch	✅ 🧪 mac-wk2-stress	✅ 🛠 jsc-mips
	✅ 🛠 watch-sim		✅ 🧪 jsc-mips-tests

… replace_lane https://bugs.webkit.org/show_bug.cgi?id=248728 rdar://102942396 Reviewed by Yusuke Suzuki. This patch aims for three tasks: 1. Add WASM SIMD operations `swizzle` and `shuffle`. https://github.com/WebAssembly/simd/blob/main/proposals/simd/SIMD.md#shuffling-using-immediate-indices 2. Refactor MacroAssemblerX86_64.h and X86Assembler.h for `pinsr` and `pextr`, where X86Assembler should do simple instruction emission only. And MacroAssemblerX86 should select the instruction. 3. Fix WASM SIMD operation `replace_lane`. Previously, the operation is implemented with instruction `pinsr` in AVX format but only passing three parameters which is wrong. https://www.felixcloutier.com/x86/pinsrb:pinsrd:pinsrq https://www.felixcloutier.com/x86/pinsrw https://www.officedaytime.com/simd512e/simdimg/si.php?f=pinsrb https://www.officedaytime.com/simd512e/simdimg/si.php?f=pinsrw https://www.officedaytime.com/simd512e/simdimg/si.php?f=pinsrd https://www.officedaytime.com/simd512e/simdimg/si.php?f=pinsrq * Source/JavaScriptCore/assembler/MacroAssemblerX86_64.h: (JSC::MacroAssemblerX86_64::vectorReplaceLane): (JSC::MacroAssemblerX86_64::vectorExtractLane): (JSC::MacroAssemblerX86_64::vectorSwizzle): * Source/JavaScriptCore/assembler/X86Assembler.h: (JSC::X86Assembler::pinsrb): (JSC::X86Assembler::pinsrw): (JSC::X86Assembler::pinsrd): (JSC::X86Assembler::pinsrq): (JSC::X86Assembler::insertps): (JSC::X86Assembler::unpcklpd): (JSC::X86Assembler::vpextrb): (JSC::X86Assembler::vpextrw): (JSC::X86Assembler::vpextrd): (JSC::X86Assembler::vpextrq): (JSC::X86Assembler::X86InstructionFormatter::SingleInstructionBufferWriter::memoryModRM): (JSC::X86Assembler::pinsr): Deleted. (JSC::X86Assembler::pextr): Deleted. (JSC::X86Assembler::vextractps): Deleted. * Source/JavaScriptCore/b3/air/AirOpcode.opcodes: * Source/JavaScriptCore/wasm/WasmAirIRGenerator.cpp: (JSC::Wasm::AirIRGenerator::addReplaceLane): Canonical link: https://commits.webkit.org/257400@main

webkit-commit-queue · 2022-12-06T09:31:10Z

Committed 257400@main (cef394f): https://commits.webkit.org/257400@main

Reviewed commits have been landed. Closing PR #7125 and removing active labels.

hyjorc1 self-assigned this Dec 4, 2022

hyjorc1 added the JavaScriptCore For bugs in JavaScriptCore, the JS engine used by WebKit, other than kxmlcore issues. label Dec 4, 2022

hyjorc1 marked this pull request as ready for review December 4, 2022 03:00

hyjorc1 requested a review from a team as a code owner December 4, 2022 03:00

Constellation requested changes Dec 4, 2022

View reviewed changes

webkit-ews-buildbot added the merging-blocked Applied to prevent a change from being merged label Dec 4, 2022

hyjorc1 removed the merging-blocked Applied to prevent a change from being merged label Dec 5, 2022

hyjorc1 force-pushed the eng/wasm-simd-intel-swizzle branch from 829c9e4 to 0bb11d6 Compare December 5, 2022 04:05

webkit-ews-buildbot added the merging-blocked Applied to prevent a change from being merged label Dec 5, 2022

hyjorc1 removed the merging-blocked Applied to prevent a change from being merged label Dec 6, 2022

hyjorc1 force-pushed the eng/wasm-simd-intel-swizzle branch from 0bb11d6 to 2e2be09 Compare December 6, 2022 06:18

hyjorc1 changed the title ~~[SIMD] Intel support for swizzle and fix extract_plane and replace_lane~~ [SIMD] Intel support for swizzle and shuffle and fix extract_plane and replace_lane Dec 6, 2022

hyjorc1 force-pushed the eng/wasm-simd-intel-swizzle branch from 2e2be09 to b063c3c Compare December 6, 2022 06:51

hyjorc1 changed the title ~~[SIMD] Intel support for swizzle and shuffle and fix extract_plane and replace_lane~~ [SIMD] Intel support for swizzle and shuffle and fix extract_lane and replace_lane Dec 6, 2022

hyjorc1 force-pushed the eng/wasm-simd-intel-swizzle branch from b063c3c to 2c999a9 Compare December 6, 2022 06:54

Constellation approved these changes Dec 6, 2022

View reviewed changes

Source/JavaScriptCore/assembler/MacroAssemblerX86_64.h Outdated Show resolved Hide resolved

Source/JavaScriptCore/assembler/MacroAssemblerX86_64.h Outdated Show resolved Hide resolved

Source/JavaScriptCore/wasm/WasmAirIRGenerator.cpp Outdated Show resolved Hide resolved

hyjorc1 force-pushed the eng/wasm-simd-intel-swizzle branch from 2c999a9 to 8f5cea8 Compare December 6, 2022 07:27

hyjorc1 added the merge-queue Applied to send a pull request to merge-queue label Dec 6, 2022

hyjorc1 removed the merge-queue Applied to send a pull request to merge-queue label Dec 6, 2022

hyjorc1 force-pushed the eng/wasm-simd-intel-swizzle branch from 8f5cea8 to dea943f Compare December 6, 2022 07:39

hyjorc1 added the merge-queue Applied to send a pull request to merge-queue label Dec 6, 2022

webkit-early-warning-system force-pushed the eng/wasm-simd-intel-swizzle branch from dea943f to cef394f Compare December 6, 2022 09:31

webkit-early-warning-system merged commit cef394f into WebKit:main Dec 6, 2022

webkit-commit-queue removed the merge-queue Applied to send a pull request to merge-queue label Dec 6, 2022

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

[SIMD] Intel support for swizzle and shuffle and fix extract_lane and replace_lane #7125

[SIMD] Intel support for swizzle and shuffle and fix extract_lane and replace_lane #7125

hyjorc1 commented Dec 4, 2022 •

edited by webkit-early-warning-system

webkit-early-warning-system commented Dec 4, 2022 •

edited

webkit-early-warning-system commented Dec 5, 2022 •

edited

webkit-early-warning-system commented Dec 6, 2022

webkit-early-warning-system commented Dec 6, 2022 •

edited

webkit-early-warning-system commented Dec 6, 2022 •

edited

webkit-early-warning-system commented Dec 6, 2022 •

edited

Constellation left a comment

webkit-early-warning-system commented Dec 6, 2022 •

edited

webkit-early-warning-system commented Dec 6, 2022

webkit-early-warning-system commented Dec 6, 2022 •

edited

webkit-commit-queue commented Dec 6, 2022

[SIMD] Intel support for swizzle and shuffle and fix extract_lane and replace_lane #7125

[SIMD] Intel support for swizzle and shuffle and fix extract_lane and replace_lane #7125

Conversation

hyjorc1 commented Dec 4, 2022 • edited by webkit-early-warning-system

cef394f

webkit-early-warning-system commented Dec 4, 2022 • edited

webkit-early-warning-system commented Dec 5, 2022 • edited

webkit-early-warning-system commented Dec 6, 2022

webkit-early-warning-system commented Dec 6, 2022 • edited

webkit-early-warning-system commented Dec 6, 2022 • edited

webkit-early-warning-system commented Dec 6, 2022 • edited

Constellation left a comment

Choose a reason for hiding this comment

webkit-early-warning-system commented Dec 6, 2022 • edited

webkit-early-warning-system commented Dec 6, 2022

webkit-early-warning-system commented Dec 6, 2022 • edited

webkit-commit-queue commented Dec 6, 2022

hyjorc1 commented Dec 4, 2022 •

edited by webkit-early-warning-system

`cef394f`

webkit-early-warning-system commented Dec 4, 2022 •

edited

webkit-early-warning-system commented Dec 5, 2022 •

edited

webkit-early-warning-system commented Dec 6, 2022 •

edited

webkit-early-warning-system commented Dec 6, 2022 •

edited

webkit-early-warning-system commented Dec 6, 2022 •

edited

webkit-early-warning-system commented Dec 6, 2022 •

edited

webkit-early-warning-system commented Dec 6, 2022 •

edited