Issue 608031: Change KeyedStoreIC interface to take value, key, and receiver in registers.

Issue 608031: Change KeyedStoreIC interface to take value, key, and receiver in registers. (Closed)

Created:
10 years, 10 months ago by William Hesse

Modified:
9 years, 7 months ago

Reviewers:
fschneider

CC:
v8-dev

Base URL:
http://v8.googlecode.com/svn/branches/bleeding_edge/

Visibility:
Public.

Description

Change KeyedStoreIC interface to take value, key, and receiver in registers. Committed: http://code.google.com/p/v8/source/detail?r=3955

Patch Set 5 : '' #

Created: 10 years, 10 months ago

Download [raw] [tar.bz2]

	Unified diffs	Side-by-side diffs	Delta from patch set	Stats (+275 lines, -227 lines)			Patch
M	src/ia32/codegen-ia32.cc	View	1 2 3 4	6 chunks	+64 lines, -14 lines	0 comments	Download
M	src/ia32/full-codegen-ia32.cc	View	4	4 chunks	+18 lines, -12 lines	0 comments	Download
M	src/ia32/ic-ia32.cc	View	1 2 3 4	10 chunks	+116 lines, -139 lines	0 comments	Download
M	src/ia32/stub-cache-ia32.cc	View	1 2 3 4	1 chunk	+2 lines, -7 lines	0 comments	Download
M	src/ia32/virtual-frame-ia32.h	View	1 2 3 4	1 chunk	+8 lines, -0 lines	0 comments	Download
M	src/ia32/virtual-frame-ia32.cc	View	1 2 3 4	3 chunks	+67 lines, -55 lines	0 comments	Download

Messages

Total messages: 3 (0 generated)

Expand Messages | Collapse Messages

fschneider

LGTM. http://codereview.chromium.org/608031/diff/6009/6010 File src/ia32/codegen-ia32.cc (right): http://codereview.chromium.org/608031/diff/6009/6010#newcode6751 src/ia32/codegen-ia32.cc:6751: __ mov(scratch_, receiver_); // Safe even if key_ ...

10 years, 10 months ago (2010-02-25 11:53:52 UTC) #2

William Hesse

10 years, 9 months ago (2010-03-01 16:16:14 UTC) #3

We forgot to fix the DebugBreakIC... to know that the registers contained GCable
values.  Fixed in http://codereview.chromium.org/660257

http://codereview.chromium.org/608031/diff/6009/6014
File src/ia32/ic-ia32.cc (right):

http://codereview.chromium.org/608031/diff/6009/6014#newcode817
src/ia32/ic-ia32.cc:817: __ fstp_s(Operand(edi, ebx, times_4, 0));
Not valid. ret(4) pops 4 bytes _after_ popping the return address.

On 2010/02/25 11:53:52, fschneider wrote:
> Tiny optimization: To remove the pop-instruction instead of 
> 
> __ push(ecx)
> __ fild_s(...)
> __ pop(ecx)
> __ fstp_s(...)
> __ ret(0)
> 
> you could just do
> 
> __ push(ecx)
> __ fild_s(...)
> __ fstp_s(...)
> __ ret(4)
> 
> And hoist __ ret(0) into the switch-statement for the other cases.

Expand Messages | Collapse Messages