[webkit-changes] [WebKit/WebKit] 6e034b: [JSC] Fix ArithMin/ArithMax handling for double, a...

Yusuke Suzuki noreply at github.com
Wed Oct 12 19:45:22 PDT 2022


  Branch: refs/heads/main
  Home:   https://github.com/WebKit/WebKit
  Commit: 6e034bbd26837f492aa03d5bd1e5ce197b6bc61b
      https://github.com/WebKit/WebKit/commit/6e034bbd26837f492aa03d5bd1e5ce197b6bc61b
  Author: Yusuke Suzuki <ysuzuki at apple.com>
  Date:   2022-10-12 (Wed, 12 Oct 2022)

  Changed paths:
    M JSTests/stress/arith-min-max-multiple.js
    M Source/JavaScriptCore/dfg/DFGSpeculativeJIT.cpp
    M Source/JavaScriptCore/ftl/FTLLowerDFGToB3.cpp

  Log Message:
  -----------
  [JSC] Fix ArithMin/ArithMax handling for double, and Int32 speculation
https://bugs.webkit.org/show_bug.cgi?id=246422
rdar://101048572

Reviewed by Justin Michaud.

This patch fixes two issues: one is just a bug fix and one is a bit theoretical.

1. FTL ArithMin/ArithMax double case is skipping the first parameter accidentally. This patch fixes it.
2. DFG ArithMin/ArithMax Int32 case can OSR exit after clobbering first parameter's register, which can theoretically
   puts different output if type speculation failed in the second or later parameters.

* JSTests/stress/arith-min-max-multiple.js:
(test7):
(test8):
* Source/JavaScriptCore/dfg/DFGSpeculativeJIT.cpp:
* Source/JavaScriptCore/ftl/FTLLowerDFGToB3.cpp:
(JSC::FTL::DFG::LowerDFGToB3::compileArithMinOrMax):

Canonical link: https://commits.webkit.org/255465@main




More information about the webkit-changes mailing list