March 2011 - squirrelfish-dev - lists.macosforge.org

tar1583@gmail.com is inviting you to freecharge.in
by tar1583＠gmail.com 11 Mar '11

11 Mar '11

1 0

true as 1 (int32)
by Zoltan Herczeg 04 Mar '11

04 Mar '11

Hi, I found something interesting in the v8 source code. They store the true and false as integers: 1 and 0. Actually it was a surprise to me, but "true + 1" is "2" in JavaScript, so this kind of object handling seems ok. Advantage of their approach: they don't need to care about boolean values for all arithmetic and conditional operations in JIT. This could reduce the source code (maintainability), and probably faster, since you don't need to check boolean values. Disadvantage: type resolving is a bit more difficult, but since that is a rare operation, I think the gain could be bigger. What is your opinion about it? Would it be possible to implement something like this for JSValue32_64? Would it worth it? Regards, Zoltan

3 5

Re: [squirrelfish] true as 1 (int32)
by Zoltan Herczeg 04 Mar '11

04 Mar '11

This is their initialization: { MaybeObject* maybe_obj = CreateOddball("true", Smi::FromInt(1), Oddball::kTrue); if (!maybe_obj->ToObject(&obj)) return false; } set_true_value(obj); { MaybeObject* maybe_obj = CreateOddball("false", Smi::FromInt(0), Oddball::kFalse); if (!maybe_obj->ToObject(&obj)) return false; } set_false_value(obj); Of course, you need to know in several places that it IS a boolean, but for arithmetic operations (most common in JIT), it seems an int is more than enough. (Smi thing is similar to the old JSValue32 approach) Regards, Zoltan > Hello, > > This is not the case, at least in the full-codegen, where v8 uses specific > heap values for the boolean True and False. > I am not sure what Crankshaft is able to optimize, but it does also use > these values. > > See src/heap.h line 56 > #define UNCONDITIONAL_STRONG_ROOT_LIST(V) > ... > V(Object, true_value, TrueValue) > \ > V(Object, false_value, FalseValue) > > For example ia32 then uses Factory::true_value() to access it. > (full-codegen-ia32.cc) > > 527 // Emit the inlined tests assumed by the stub. > 528 __ cmp(result_register(), Factory::undefined_value()); > 529 __ j(equal, if_false); > 530 __ cmp(result_register(), Factory::true_value()); > 531 __ j(equal, if_true); > 532 __ cmp(result_register(), Factory::false_value()); > 533 __ j(equal, if_false); > > > Alexandre > > > On Fri, Mar 4, 2011 at 1:37 PM, Zoltan Herczeg > <zherczeg(a)inf.u-szeged.hu>wrote: > >> Hi, >> >> I found something interesting in the v8 source code. They store the true >> and false as integers: 1 and 0. Actually it was a surprise to me, but >> "true + 1" is "2" in JavaScript, so this kind of object handling seems >> ok. >> >> Advantage of their approach: they don't need to care about boolean >> values >> for all arithmetic and conditional operations in JIT. This could reduce >> the source code (maintainability), and probably faster, since you don't >> need to check boolean values. >> >> Disadvantage: type resolving is a bit more difficult, but since that is >> a >> rare operation, I think the gain could be bigger. >> >> What is your opinion about it? Would it be possible to implement >> something >> like this for JSValue32_64? Would it worth it? >> >> Regards, >> Zoltan >> >> >> _______________________________________________ >> squirrelfish-dev mailing list >> squirrelfish-dev(a)lists.webkit.org >> http://lists.webkit.org/mailman/listinfo.cgi/squirrelfish-dev >> >

1 0