[Webkit-unassigned] [Bug 130638] [Win64] ASM LLINT is not enabled.
bugzilla-daemon at webkit.org
bugzilla-daemon at webkit.org
Fri Apr 11 15:59:00 PDT 2014
https://bugs.webkit.org/show_bug.cgi?id=130638
--- Comment #20 from Mark Lam <mark.lam at apple.com> 2014-04-11 15:59:19 PST ---
(From update of attachment 229169)
View in context: https://bugs.webkit.org/attachment.cgi?id=229169&action=review
> Source/JavaScriptCore/jit/Repatch.cpp:1048
> - stubJit.store32(MacroAssembler::TrustedImm32(reinterpret_cast<uint32_t>(structure->id())), MacroAssembler::Address(baseGPR, JSCell::structureIDOffset()));
> +#if USE(JSVALUE64)
> + uint32_t val = structure->id();
> +#else
> + uint32_t val = reinterpret_cast<uint32_t>(structure->id());
> +#endif
> + stubJit.store32(MacroAssembler::TrustedImm32(val), MacroAssembler::Address(baseGPR, JSCell::structureIDOffset()));
Did you explain why this is needed?
> Source/JavaScriptCore/llint/LowLevelInterpreter64.asm:209
> - cCall2(_llint_throw_stack_overflow_error, vm, protoCallFrame)
> + cCall2SlowPath(_llint_throw_stack_overflow_error, vm, protoCallFrame)
Not needed.
> Source/JavaScriptCore/llint/LowLevelInterpreter64.asm:327
> else
> - addp 16, sp
> + if X86_64_WIN
> + subp 32, sp
> + else
> + addp 16, sp
> + end
> call temp
> - subp 16, sp
> + if X86_64_WIN
> + addp 32, sp
> + else
> + subp 16, sp
> + end
> end
This will be easier to read and understand if you express it as:
elsif if X86_64_WIN
subp 32, sp
call temp
addp 32, sp
else
addp 16, sp
call temp
subp 16, sp
end
> Source/JavaScriptCore/llint/LowLevelInterpreter64.asm:383
> +# The signature of a slow path call is extern "C" SlowPathReturnType slowpathfunction(ExecState* exec, Instruction* pc).
> +# The size of the SlowPathReturnType is 16 bytes.
> +# On Win64, when the return type is larger than 8 bytes, we need to allocate space on the stack for the return value.
> +# On entry rcx (t2), should contain a pointer to this stack space. The other parameters are shifted to the right,
> +# rdx (t1) should contain the first argument, and r8 (t6) should contain the second argument.
> +# On return, rax contains a pointer to this stack value, and we then need to copy the 16 byte return value into rax (t0) and rdx (t1)
> +# since the return value is expected to be split between the two.
> +# See http://msdn.microsoft.com/en-us/library/7572ztz4.aspx
> +macro callSlowPathWin64(slowPath, arg1, arg2)
> + move arg1, t1
> + move arg2, t6
> + subp 80, sp
> + move sp, t2
> + addp 48, t2
> + call slowPath
> + addp 80, sp
> + move 8[t0], t1
> + move [t0], t0
> +end
You misunderstood me. I meant for you to move this inside cCall2. Is there any reason why that cannot be done?
> Source/JavaScriptCore/llint/LowLevelInterpreter64.asm:391
> +macro cCall2SlowPath(slowPath, arg1, arg2)
> + if X86_64_WIN
> + callSlowPathWin64(slowPath, arg1, arg2)
> + else
> + cCall2(slowPath, arg1, arg2)
> + end
> +end
This is not needed.
> Source/JavaScriptCore/llint/LowLevelInterpreter64.asm:395
> - cCall2(slowPath, cfr, PC)
> + cCall2SlowPath(slowPath, cfr, PC)
No need to change this. Leave as calling cCall2. Same with all similar cases below.
> Source/JavaScriptCore/llint/LowLevelInterpreter64.asm:415
> - cCall2(slowPath, cfr, PC)
> + cCall2SlowPath(slowPath, cfr, PC)
Ditto.
> Source/JavaScriptCore/llint/LowLevelInterpreter64.asm:422
> - cCall2(_llint_slow_path_handle_watchdog_timer, cfr, PC)
> + cCall2SlowPath(_llint_slow_path_handle_watchdog_timer, cfr, PC)
Ditto.
> Source/JavaScriptCore/llint/LowLevelInterpreter64.asm:434
> - cCall2(_llint_loop_osr, cfr, PC)
> + cCall2SlowPath(_llint_loop_osr, cfr, PC)
Ditto.
> Source/JavaScriptCore/llint/LowLevelInterpreter64.asm:555
> - cCall2(slowPath, cfr, PC) # This slowPath has the protocol: t0 = 0 => no error, t0 != 0 => error
> + cCall2SlowPath(slowPath, cfr, PC) # This slowPath has the protocol: t0 = 0 => no error, t0 != 0 => error
Ditto.
> Source/JavaScriptCore/llint/LowLevelInterpreter.asm:520
> - cCall2(osrSlowPath, cfr, PC)
> + cCall2SlowPath(osrSlowPath, cfr, PC)
Ditto.
--
Configure bugmail: https://bugs.webkit.org/userprefs.cgi?tab=email
------- You are receiving this mail because: -------
You are the assignee for the bug.
More information about the webkit-unassigned
mailing list