[webkit-changes] [WebKit/WebKit] 0def43: [SIMD] Intel support for load and store operations...

[Yijia Huang]

  Branch: refs/heads/main
  Home:   https://github.com/WebKit/WebKit
  Commit: 0def43da76cfb0f0fce99e5ff7a3e2b57f8e7dce
      https://github.com/WebKit/WebKit/commit/0def43da76cfb0f0fce99e5ff7a3e2b57f8e7dce
  Author: Yijia Huang <hyjorc1 at gmail.com>
  Date:   2022-12-13 (Tue, 13 Dec 2022)

  Changed paths:
    M Source/JavaScriptCore/assembler/MacroAssemblerX86_64.h
    M Source/JavaScriptCore/assembler/X86Assembler.h
    M Source/JavaScriptCore/b3/air/AirOpcode.opcodes
    M Source/JavaScriptCore/wasm/WasmAirIRGenerator64.cpp

  Log Message:
  -----------
  [SIMD] Intel support for load and store operations and fix operation shuffle
https://bugs.webkit.org/show_bug.cgi?id=249073
rdar://103214919

Reviewed by Yusuke Suzuki.

This patch achieves four tasks:
1. Add load and store operations for WASM SIMD on Intel.
   https://github.com/WebAssembly/simd/blob/main/proposals/simd/SIMD.md#load-and-store
2. And fix WASM SIMD operation i8x16.shuffle.
3. Fix instruction encoding in X86Assember.h according to the new conventions.
4. Define and use AVX functions for operations replace_lane, extract_lane, and splat.

* Source/JavaScriptCore/assembler/MacroAssemblerX86_64.h:
(JSC::MacroAssemblerX86_64::vectorReplaceLaneAVX):
(JSC::MacroAssemblerX86_64::vectorReplaceLane):
(JSC::MacroAssemblerX86_64::vectorExtractLane):
(JSC::MacroAssemblerX86_64::vectorExtractLaneAVX):
(JSC::MacroAssemblerX86_64::vectorAbs):
(JSC::MacroAssemblerX86_64::vectorTruncSatUnsignedFloat64):
(JSC::MacroAssemblerX86_64::vectorExtendHigh):
(JSC::MacroAssemblerX86_64::vectorSplatAVX):
(JSC::MacroAssemblerX86_64::vectorSplat):
(JSC::MacroAssemblerX86_64::vectorLoad8Splat):
(JSC::MacroAssemblerX86_64::vectorLoad16Splat):
(JSC::MacroAssemblerX86_64::vectorLoad32Splat):
(JSC::MacroAssemblerX86_64::vectorLoad64Splat):
(JSC::MacroAssemblerX86_64::vectorLoad8Lane):
(JSC::MacroAssemblerX86_64::vectorLoad16Lane):
(JSC::MacroAssemblerX86_64::vectorLoad32Lane):
(JSC::MacroAssemblerX86_64::vectorLoad64Lane):
(JSC::MacroAssemblerX86_64::vectorStore8Lane):
(JSC::MacroAssemblerX86_64::vectorStore16Lane):
(JSC::MacroAssemblerX86_64::vectorStore32Lane):
(JSC::MacroAssemblerX86_64::vectorStore64Lane):
* Source/JavaScriptCore/assembler/X86Assembler.h:
(JSC::X86Assembler::pinsrb_i8rr):
(JSC::X86Assembler::pinsrw_i8rr):
(JSC::X86Assembler::pinsrd_i8rr):
(JSC::X86Assembler::pinsrq_i8rr):
(JSC::X86Assembler::insertps_i8rr):
(JSC::X86Assembler::pextrb_i8rr):
(JSC::X86Assembler::pextrw_i8rr):
(JSC::X86Assembler::pextrd_i8rr):
(JSC::X86Assembler::pextrq_i8rr):
(JSC::X86Assembler::pshufd_i8rr):
(JSC::X86Assembler::pshuflw_i8rr):
(JSC::X86Assembler::shufps_i8rr):
(JSC::X86Assembler::shufpd_i8rr):
(JSC::X86Assembler::pblendw_rr):
(JSC::X86Assembler::packusdw_rr):
(JSC::X86Assembler::pmovsxbw):
(JSC::X86Assembler::pmovzxbw):
(JSC::X86Assembler::pmovsxwd):
(JSC::X86Assembler::pmovzxwd):
(JSC::X86Assembler::pmovsxdq):
(JSC::X86Assembler::pmovzxdq):
(JSC::X86Assembler::vpinsrb_i8mrr):
(JSC::X86Assembler::vpinsrb_i8rrr):
(JSC::X86Assembler::vpinsrw_i8mrr):
(JSC::X86Assembler::vpinsrw_i8rrr):
(JSC::X86Assembler::vpinsrd_i8mrr):
(JSC::X86Assembler::vpinsrd_i8rrr):
(JSC::X86Assembler::vpinsrq_i8mrr):
(JSC::X86Assembler::vpinsrq_i8rrr):
(JSC::X86Assembler::vinsertps_i8rrr):
(JSC::X86Assembler::vmovddup_rr):
(JSC::X86Assembler::vmovddup_mr):
(JSC::X86Assembler::vmovapd_rr):
(JSC::X86Assembler::vbroadcastss_mr):
(JSC::X86Assembler::vmovq_rr):
(JSC::X86Assembler::vpunpcklqdq_rrr):
(JSC::X86Assembler::vunpcklpd_rrr):
(JSC::X86Assembler::vpextrb_i8rr):
(JSC::X86Assembler::vpextrb_i8rm):
(JSC::X86Assembler::vpextrw_i8rr):
(JSC::X86Assembler::vpextrw_i8rm):
(JSC::X86Assembler::vpextrd_i8rr):
(JSC::X86Assembler::vpextrd_i8rm):
(JSC::X86Assembler::vpextrq_i8rr):
(JSC::X86Assembler::vpextrq_i8rm):
(JSC::X86Assembler::vshufps_i8rrr):
(JSC::X86Assembler::vshufpd_i8rrr):
(JSC::X86Assembler::vpshuflw_i8rr):
(JSC::X86Assembler::vpshufd_i8rr):
(JSC::X86Assembler::vpackusdw_rrr):
(JSC::X86Assembler::vpmovsxbw_rr):
(JSC::X86Assembler::vpmovzxbw_rr):
(JSC::X86Assembler::vpmovsxwd_rr):
(JSC::X86Assembler::vpmovzxwd_rr):
(JSC::X86Assembler::vpmovsxdq_rr):
(JSC::X86Assembler::vpmovzxdq_rr):
(JSC::X86Assembler::vmovaps_rr):
(JSC::X86Assembler::vmovshdup_rr):
(JSC::X86Assembler::vmovhlps_rrr):
(JSC::X86Assembler::vmovsd_rrr):
(JSC::X86Assembler::vpblendw_i8rrr):
(JSC::X86Assembler::X86InstructionFormatter::SingleInstructionBufferWriter::memoryModRM):
(JSC::X86Assembler::pinsrb_rr): Deleted.
(JSC::X86Assembler::pinsrw_rr): Deleted.
(JSC::X86Assembler::pinsrd_rr): Deleted.
(JSC::X86Assembler::pinsrq_rr): Deleted.
(JSC::X86Assembler::insertps_rr): Deleted.
(JSC::X86Assembler::pextrb_rr): Deleted.
(JSC::X86Assembler::pextrw_rr): Deleted.
(JSC::X86Assembler::pextrd_rr): Deleted.
(JSC::X86Assembler::pextrq_rr): Deleted.
(JSC::X86Assembler::pshufd_rr): Deleted.
(JSC::X86Assembler::pshuflw_rr): Deleted.
(JSC::X86Assembler::shufps_rr): Deleted.
(JSC::X86Assembler::shufpd_rr): Deleted.
(JSC::X86Assembler::vpextrb_rr): Deleted.
(JSC::X86Assembler::vpextrw_rr): Deleted.
(JSC::X86Assembler::vpextrd_rr): Deleted.
(JSC::X86Assembler::vpextrq_rr): Deleted.
(JSC::X86Assembler::vshufps_rrr): Deleted.
* Source/JavaScriptCore/b3/air/AirOpcode.opcodes:
* Source/JavaScriptCore/wasm/WasmAirIRGenerator.cpp:
(JSC::Wasm::AirIRGenerator::addSIMDSwizzleHelperX86):
(JSC::Wasm::AirIRGenerator::addSIMDV_VV):
(JSC::Wasm::AirIRGenerator::addSIMDShuffle):
(JSC::Wasm::AirIRGenerator::addSIMDLoadSplat):
(JSC::Wasm::AirIRGenerator::addSIMDSwizzle): Deleted.

Canonical link: https://commits.webkit.org/257818@main