Fast-compiler: Add stack limit checks to back edges of while, do-while and for.

Fast-compiler: Add stack limit checks to back edges of while, do-while and for.
A few other tweaks.

[Lasse Reichstein, 2009-11-24 14:21:34]

Quick fix of missing stack check.
Seems to work, but I'll run more tests tomorrow too.

http://codereview.chromium.org/435020/diff/1/6
File src/heap.h (right):

http://codereview.chromium.org/435020/diff/1/6#newcode914
src/heap.h:914: (1 << MapWord::kMapPageIndexBits) * Page::kPageSize;
By popular demand: Make it obvious what map space is restricted by.

http://codereview.chromium.org/435020/diff/1/10
File src/x64/fast-codegen-x64.cc (right):

http://codereview.chromium.org/435020/diff/1/10#newcode392
src/x64/fast-codegen-x64.cc:392:
ASSERT(!Heap::InNewSpace(*Factory::the_hole_value()));
I can't see how we could have been hitting this at all.

[Mads Ager (chromium), 2009-11-24 14:46:32]

LGTM

http://codereview.chromium.org/435020/diff/1/2
File src/arm/fast-codegen-arm.cc (right):

http://codereview.chromium.org/435020/diff/1/2#newcode386
src/arm/fast-codegen-arm.cc:386:
ASSERT(!Heap::InNewSpace(*Factory::the_hole_value()));
Could you add a regression test for this?  Something with a const variable in a
context slot at the top level?

http://codereview.chromium.org/435020/diff/1/6
File src/heap.h (right):

http://codereview.chromium.org/435020/diff/1/6#newcode914
src/heap.h:914: (1 << MapWord::kMapPageIndexBits) * Page::kPageSize;
On 2009/11/24 14:21:36, Lasse Reichstein wrote:
> By popular demand: Make it obvious what map space is restricted by.

by the amount of bits available for map page index encoding.

Thanks.

[fschneider, 2009-11-25 01:02:59]

LGTM!

http://codereview.chromium.org/435020/diff/1/10
File src/x64/fast-codegen-x64.cc (right):

http://codereview.chromium.org/435020/diff/1/10#newcode392
src/x64/fast-codegen-x64.cc:392:
ASSERT(!Heap::InNewSpace(*Factory::the_hole_value()));
On 2009/11/24 14:21:36, Lasse Reichstein wrote:
> I can't see how we could have been hitting this at all.

Thanks for fixing this along the way. I'm not sure - maybe this assertion should
go away? For some reason I removed it only from IA32 + got it wrong on the other
platforms!? My fault...

[Lasse Reichstein, 2009-11-25 08:56:25]

http://codereview.chromium.org/435020/diff/1/2
File src/arm/fast-codegen-arm.cc (right):

http://codereview.chromium.org/435020/diff/1/2#newcode386
src/arm/fast-codegen-arm.cc:386:
ASSERT(!Heap::InNewSpace(*Factory::the_hole_value()));
The code is currently unreachable since we don't fully support const variables
in the fast compiler (we bail out on assignment to them).

http://codereview.chromium.org/435020/diff/1/6
File src/heap.h (right):

http://codereview.chromium.org/435020/diff/1/6#newcode914
src/heap.h:914: (1 << MapWord::kMapPageIndexBits) * Page::kPageSize;
Guess I should add a comment to make it more obvious :)

http://codereview.chromium.org/435020/diff/1/10
File src/x64/fast-codegen-x64.cc (right):

http://codereview.chromium.org/435020/diff/1/10#newcode392
src/x64/fast-codegen-x64.cc:392:
ASSERT(!Heap::InNewSpace(*Factory::the_hole_value()));
Since the code isn't hit at all now, it's an easy typo to miss.
At least we have found out why the code isn't hit (we bail out on assignment to
const variables), so we know that the code is untested and can be prepared when
we make it reachable.