Implement the arithmetic instructions for floats in WebAssembly
https://bugs.webkit.org/show_bug.cgi?id=149102
Patch by Sukolsak Sakshuwong <sukolsak@gmail.com> on 2015-09-14
Reviewed by Geoffrey Garen.
This patch implements the arithmetic instructions for floats (float32)
in WebAssembly by converting the float operands to doubles, performing
the equivalent double instructions, and converting the result back to
float. The asm.js spec says that "As proved in 'When is double rounding
innocuous?' (Figueroa 1995), both the 32- and 64-bit versions of
standard arithmetic operations produce equivalent results when given
32-bit inputs and coerced to 32-bit outputs."
(http://asmjs.org/spec/latest/#floatish)
This patch also pads WebAssembly call frames by maxFrameExtentForSlowPathCall,
so that there is no need to adjust the stack pointer every time we make
a slow path call.
* tests/stress/wasm-arithmetic-float32.js:
* tests/stress/wasm/arithmetic-float32.wasm:
* wasm/WASMFunctionCompiler.h:
(JSC::WASMFunctionCompiler::startFunction):
(JSC::WASMFunctionCompiler::buildUnaryF32):
(JSC::WASMFunctionCompiler::buildBinaryF32):
(JSC::WASMFunctionCompiler::callOperation):
(JSC::WASMFunctionCompiler::callAndUnboxResult):
(JSC::WASMFunctionCompiler::endFunction): Deleted.
(JSC::WASMFunctionCompiler::buildBinaryI32): Deleted.
* wasm/WASMFunctionParser.cpp:
(JSC::WASMFunctionParser::parseExpressionF32):
(JSC::WASMFunctionParser::parseUnaryExpressionF32):
(JSC::WASMFunctionParser::parseBinaryExpressionF32):
* wasm/WASMFunctionParser.h:
* wasm/WASMFunctionSyntaxChecker.h:
(JSC::WASMFunctionSyntaxChecker::buildUnaryF32):
(JSC::WASMFunctionSyntaxChecker::buildBinaryF32):
git-svn-id: http://svn.webkit.org/repository/webkit/trunk@189744 268f45cc-cd09-0410-ab3c-d52691b4dbfc
7 files changed
