Ignore:
Timestamp:
Mar 24, 2018, 2:43:22 PM (7 years ago)
Author:
Yusuke Suzuki
Message:

[DFG] Introduces fused compare and jump
​https://bugs.webkit.org/show_bug.cgi?id=177100

Reviewed by Mark Lam.

JSTests:

  • stress/fused-jeq-slow.js: Added.

(shouldBe):
(testJEQ):
(testJNEQB):
(testJEQB):
(testJNEQF):
(testJEQF):

  • stress/fused-jeq.js: Added.

(shouldBe):
(testJEQ):
(testJNEQB):
(testJEQB):
(testJNEQF):
(testJEQF):

  • stress/fused-jstricteq-slow.js: Added.

(shouldBe):
(testJSTRICTEQ):
(testJNSTRICTEQB):
(testJSTRICTEQB):
(testJNSTRICTEQF):
(testJSTRICTEQF):

  • stress/fused-jstricteq.js: Added.

(shouldBe):
(testJSTRICTEQ):
(testJNSTRICTEQB):
(testJSTRICTEQB):
(testJNSTRICTEQF):
(testJSTRICTEQF):

Source/JavaScriptCore:

This patch introduces op_jeq, op_jneq, op_jstricteq, and op_jnstricteq.
It offers 3 benefit.

  1. They are introduced due to the similar purpose to op_jless etc. It aligns

op_eq families to op_jless families.

  1. It reduces the size of bytecode to represent the typical code sequence.
  1. It offers the way to fuse check and jump in DFG code generation. Since

we have MovHint between Branch and CompareEq/CompareStrictEq previously,
we cannot do this optimization. It reduces the machine code size in DFG too.

It slightly improves Octane/boyer.

boyer 6.18038+-0.05002 6.06990+-0.04176 definitely 1.0182x faster

  • bytecode/BytecodeDumper.cpp:

(JSC::BytecodeDumper<Block>::dumpBytecode):

  • bytecode/BytecodeList.json:
  • bytecode/BytecodeUseDef.h:

(JSC::computeUsesForBytecodeOffset):
(JSC::computeDefsForBytecodeOffset):

  • bytecode/Opcode.h:

(JSC::isBranch):

  • bytecode/PreciseJumpTargetsInlines.h:

(JSC::extractStoredJumpTargetsForBytecodeOffset):

  • bytecompiler/BytecodeGenerator.cpp:

(JSC::BytecodeGenerator::emitJumpIfTrue):
(JSC::BytecodeGenerator::emitJumpIfFalse):

  • dfg/DFGByteCodeParser.cpp:

(JSC::DFG::ByteCodeParser::parseBlock):

  • dfg/DFGCapabilities.cpp:

(JSC::DFG::capabilityLevel):

  • dfg/DFGOperations.cpp:
  • dfg/DFGOperations.h:
  • dfg/DFGSpeculativeJIT.cpp:

(JSC::DFG::SpeculativeJIT::compileStrictEq):

  • jit/JIT.cpp:

(JSC::JIT::privateCompileMainPass):
(JSC::JIT::privateCompileSlowCases):

  • jit/JIT.h:
  • jit/JITOpcodes.cpp:

(JSC::JIT::emit_op_jeq):
(JSC::JIT::emit_op_neq):
(JSC::JIT::emit_op_jneq):
(JSC::JIT::compileOpStrictEq):
(JSC::JIT::emit_op_stricteq):
(JSC::JIT::emit_op_nstricteq):
(JSC::JIT::compileOpStrictEqJump):
(JSC::JIT::emit_op_jstricteq):
(JSC::JIT::emit_op_jnstricteq):
(JSC::JIT::emitSlow_op_jstricteq):
(JSC::JIT::emitSlow_op_jnstricteq):
(JSC::JIT::emitSlow_op_jeq):
(JSC::JIT::emitSlow_op_jneq):

  • jit/JITOpcodes32_64.cpp:

(JSC::JIT::emitSlow_op_eq):
(JSC::JIT::emit_op_jeq):
(JSC::JIT::compileOpEqJumpSlow):
(JSC::JIT::emitSlow_op_jeq):
(JSC::JIT::emit_op_jneq):
(JSC::JIT::emitSlow_op_jneq):
(JSC::JIT::compileOpStrictEq):
(JSC::JIT::emit_op_stricteq):
(JSC::JIT::emit_op_nstricteq):
(JSC::JIT::compileOpStrictEqJump):
(JSC::JIT::emit_op_jstricteq):
(JSC::JIT::emit_op_jnstricteq):
(JSC::JIT::emitSlow_op_jstricteq):
(JSC::JIT::emitSlow_op_jnstricteq):

  • jit/JITOperations.cpp:
  • jit/JITOperations.h:
  • llint/LLIntSlowPaths.cpp:

(JSC::LLInt::LLINT_SLOW_PATH_DECL):

  • llint/LLIntSlowPaths.h:
  • llint/LowLevelInterpreter.asm:
  • llint/LowLevelInterpreter32_64.asm:
  • llint/LowLevelInterpreter64.asm:
File:
1 edited

Legend:

Unmodified
Added
Removed

  • trunk/Source/JavaScriptCore/bytecode/BytecodeUseDef.h

    r229852 r229957  
    8686    case op_jngreatereq:
    8787    case op_jless:
     88    case op_jeq:
     89    case op_jneq:
     90    case op_jstricteq:
     91    case op_jnstricteq:
    8892    case op_jbelow:
    8993    case op_jbeloweq:
    … …  
    348352    case op_jngreater:
    349353    case op_jngreatereq:
     354    case op_jeq:
     355    case op_jneq:
     356    case op_jstricteq:
     357    case op_jnstricteq:
    350358    case op_jbelow:
    351359    case op_jbeloweq:
Note: See TracChangeset for help on using the changeset viewer.