April 2010 - squirrelfish-dev - lists.macosforge.org

Does Squiirel fish JIT supports armv7 ?
by s k 01 May '10

01 May '10

Hi All, I am planning to port squirrelfish for ARMv7. Does jit and assembler os squirrelfish support arm v7 architecture? If not so, how can i implement? Any Ideas regarding this.....? I too want to contribute some code for the arm v7 support. After going through webkit buglist. JIT supports only armv4 version? Thanks & Regards, Suresh Kumar

3 3

SquirrelFis Internals
by s k 23 Apr '10

23 Apr '10

Hi Squirrelfish Developers, Hope you are all fine. This is skra, new entrant in the field of JavaScript engine. While googling about JavaScript engines, i heard of the name "SquirrelFish Extreme- an ultimate fastest JavaScript engine". Thanks to all for contributing their ideas. Just iam having some technical queries about squirrelfish. How can we benchmark squirrelfish as fastest. How can we compare with spider monkey or V8. How can i know the internals of squirrelfish. After perusing the source code, I understand the hierarchy of working procedure(as for my understanding) JavaScript Src - > parser -> yarr -> WREC > icu > compiler(using generator) produces byte code. Generated byte code is converted to machine code using JIT and assembler. If my understanding is wrong please correct me. What is the use of wtf and pcre module. How runtime module comes into picture. Thanks & Regards, skar......[?]

2 1

Improving timeout responsiveness
by Kent Hansen 23 Apr '10

23 Apr '10

Hi, Currently JSC doesn't take into account that a native function call can take an arbitrary time to complete. This means that TimeoutChecker::didTimeout() can be called less frequently than desired (see https://bugs.webkit.org/show_bug.cgi?id=38041 for how this also affects the Qt API). Does anyone have ideas other than brute-forcing the check interval to 1? Has using a high-resolution timer in the inner JSC loop been considered? I guess the two-level timeout check (cheap (loop) counter, followed by elapsed time check) was done for performance reasons. The alternatives I see are 1) Replace the two-level check by a single, reasonably precise check based on elapsed time (if this is possible without killing performance); or 2) Provide a way for native function calls to affect the tick counter. 2) could be done by extending the NativeFunction signature to take an unsigned& ticks as argument. Long-running functions would do the timing explicitly, and use a helper function to convert the elapsed time to ticks. Regards, Kent

1 0

Constant propagation
by Kent Hansen 20 Apr '10

20 Apr '10

Hi, I don't see constant propagation mentioned at http://trac.webkit.org/wiki/SquirrelFishPerfIdeas. Has it been considered? (On a related note, is the above page up-to-date?) With JSC as of r57470: var a = 3; a + 7; produces the bytecode ... [ 7] mov r-15, 3(@k1) [ 10] add r0, r-15, 7(@k2) ... Worth pursuing? Reasonably feasible to implement in the compiler? Regards, Kent

4 5

Reviewer for Tiger vs Leopard VMTags issue?
by Kent Hansen 07 Apr '10

07 Apr '10

Hi, I'm looking for a reviewer for https://bugs.webkit.org/show_bug.cgi?id=34888 Rather than spamming more individuals, I thought I'd spam this list first to see if there are any takers. Regards, Kent

1 0