This is part of why I want the win64 jit so bad. I got the jit and dfg working with https://bugs.webkit.org/show_bug.cgi?id=130638 and I just got the ftl to compile and link. libllvmForJSC compiles and mostly links.
Created attachment 233288 [details] Patch
Comment on attachment 233288 [details] Patch Attachment 233288 [details] did not pass mac-wk2-ews (mac-wk2): Output: http://webkit-queues.appspot.com/results/5327732833517568 New failing tests: media/W3C/video/readyState/readyState_during_canplay.html
Created attachment 233298 [details] Archive of layout-test-results from webkit-ews-09 for mac-mountainlion-wk2 The attached test failures were seen while running run-webkit-tests on the mac-wk2-ews. Bot: webkit-ews-09 Port: mac-mountainlion-wk2 Platform: Mac OS X 10.8.5
The only possible behavior change on mac would be if std::nan("") is not the same as 0.0 / 0.0.
Comment on attachment 233288 [details] Patch View in context: https://bugs.webkit.org/attachment.cgi?id=233288&action=review > Source/JavaScriptCore/disassembler/LLVMDisassembler.cpp:57 > + _snprintf( Can we have one common #define or something in WTF so that we don't have to #if PLATFORM in so many places just for this function? > Source/JavaScriptCore/ftl/FTLLowerDFGToLLVM.cpp:1439 > + right, m_out.constDouble(std::nan(""))))); We should use PNaN here. It's a JSC thing.
Created attachment 233336 [details] Patch
(In reply to comment #5) > (From update of attachment 233288 [details]) > View in context: https://bugs.webkit.org/attachment.cgi?id=233288&action=review > > > Source/JavaScriptCore/disassembler/LLVMDisassembler.cpp:57 > > + _snprintf( > > Can we have one common #define or something in WTF so that we don't have to #if PLATFORM in so many places just for this function? I'll just leave that out of this patch until I can get a chance to test it. There's also an issue with _snprintf not always writing a null character at the end if the string is the exact length of the buffer. > > > Source/JavaScriptCore/ftl/FTLLowerDFGToLLVM.cpp:1439 > > + right, m_out.constDouble(std::nan(""))))); > > We should use PNaN here. It's a JSC thing. Done.
Comment on attachment 233336 [details] Patch View in context: https://bugs.webkit.org/attachment.cgi?id=233336&action=review > Source/JavaScriptCore/ftl/FTLLowerDFGToLLVM.cpp:1439 > - right, m_out.constDouble(0.0 / 0.0)))); > + right, m_out.constDouble(PNaN)))); I don't think it's right to change the indentation here.
(In reply to comment #8) > I don't think it's right to change the indentation here. Style bot does, but I didn't change it in the commit. http://trac.webkit.org/changeset/170130