Pass key and receiver in registers for keyed load IC on ARM...

Pass key and receiver in registers for keyed load IC on ARM

The calling convention for keyed load IC's on ARM now passes the key and receiver in registers r0 and r1.

The code path in the ARM full compiler for handling keyed property load now has the same structure as for ia32 where the keyed load IC is also called with key end receiver in registers.

This change have been tested with an exhaustive combinations of the flags

  --special-command="@ --nofull-compiler"
  --special-command="@ --always-full-compiler"
  --special-command="@ --noenable-vfp3"

to the test runner.

Committed: http://code.google.com/p/v8/source/detail?r=4608

[Erik Corry, 2010-05-07 08:53:25]

LGTM with comments.

http://codereview.chromium.org/2024002/diff/8001/9002
File src/arm/codegen-arm.cc (right):

http://codereview.chromium.org/2024002/diff/8001/9002#newcode5749
src/arm/codegen-arm.cc:5749: if (VirtualFrame::SpilledScope::is_spilled()) {
Can't we move this into Dup2()?

http://codereview.chromium.org/2024002/diff/8001/9006
File src/arm/ic-arm.cc (right):

http://codereview.chromium.org/2024002/diff/8001/9006#newcode1064
src/arm/ic-arm.cc:1064: GenerateUInt2Double(masm, value, loword, r4, 0);
value -> hiword

http://codereview.chromium.org/2024002/diff/8001/9006#newcode1069
src/arm/ic-arm.cc:1069: GenerateUInt2Double(masm, value, loword, r4, 1);
value -> hiword

http://codereview.chromium.org/2024002/diff/8001/9006#newcode1075
src/arm/ic-arm.cc:1075: // more. Also OK to clobber r1.
OK to clobber r1 if the allocation succeeds, but AllocateHeapNumber seems to
clobber r1 regardless!

http://codereview.chromium.org/2024002/diff/8001/9004
File src/arm/virtual-frame-arm.cc (right):

http://codereview.chromium.org/2024002/diff/8001/9004#newcode542
src/arm/virtual-frame-arm.cc:542: __ push(r1);
Oops!

http://codereview.chromium.org/2024002/diff/8001/9004#newcode548
src/arm/virtual-frame-arm.cc:548: __ push(r1);
Hmmm.

http://codereview.chromium.org/2024002/diff/8001/9004#newcode554
src/arm/virtual-frame-arm.cc:554: __ push(r0);
I don't get it.

[Søren Thygesen Gjesse, 2010-05-07 10:16:11]

http://codereview.chromium.org/2024002/diff/8001/9002
File src/arm/codegen-arm.cc (right):

http://codereview.chromium.org/2024002/diff/8001/9002#newcode5749
src/arm/codegen-arm.cc:5749: if (VirtualFrame::SpilledScope::is_spilled()) {
On 2010/05/07 08:53:25, Erik Corry wrote:
> Can't we move this into Dup2()?

Done. It needs a temporary register, which is why I did not do it in the first
place, but I can see the virtual frame can use ip.

http://codereview.chromium.org/2024002/diff/8001/9006
File src/arm/ic-arm.cc (right):

http://codereview.chromium.org/2024002/diff/8001/9006#newcode1064
src/arm/ic-arm.cc:1064: GenerateUInt2Double(masm, value, loword, r4, 0);
On 2010/05/07 08:53:25, Erik Corry wrote:
> value -> hiword

Done.

http://codereview.chromium.org/2024002/diff/8001/9006#newcode1069
src/arm/ic-arm.cc:1069: GenerateUInt2Double(masm, value, loword, r4, 1);
On 2010/05/07 08:53:25, Erik Corry wrote:
> value -> hiword

Done.

http://codereview.chromium.org/2024002/diff/8001/9006#newcode1075
src/arm/ic-arm.cc:1075: // more. Also OK to clobber r1.
On 2010/05/07 08:53:25, Erik Corry wrote:
> OK to clobber r1 if the allocation succeeds, but AllocateHeapNumber seems to
> clobber r1 regardless!

Good catch! Removed the comment on both r0 and r1 and used r2 and r3 instead.

Fixed all the other uses AllocateHeapNumber below to avoid clobbering r0 and r1
before any possible bailouts to the slow case.

http://codereview.chromium.org/2024002/diff/8001/9004
File src/arm/virtual-frame-arm.cc (right):

http://codereview.chromium.org/2024002/diff/8001/9004#newcode542
src/arm/virtual-frame-arm.cc:542: __ push(r1);
On 2010/05/07 08:53:25, Erik Corry wrote:
> Oops!

Done. This path is not hit by the tests.

http://codereview.chromium.org/2024002/diff/8001/9004#newcode548
src/arm/virtual-frame-arm.cc:548: __ push(r1);
On 2010/05/07 08:53:25, Erik Corry wrote:
> Hmmm.

What a mess - fixed. Also this path is not hit by the tests.

http://codereview.chromium.org/2024002/diff/8001/9004#newcode554
src/arm/virtual-frame-arm.cc:554: __ push(r0);
On 2010/05/07 08:53:25, Erik Corry wrote:
> I don't get it.

And fixed again. Also this path is not hit by the tests.