PowerPC specific sub-directories

PowerPC specific sub-directories

Split from
https://codereview.chromium.org/422063005/

This commit contains src/ppc and src/ic/ppc - the two PowerPC specific
sub-directories.

Another commit will contain the code to hook the PPC code in

BUG=

[andrew_low, 2014-09-15 17:50:26]

andrew_low@ca.ibm.com changed reviewers:
+ danno@chromium.org, rmcilroy@chromium.org, svenpanne@chromium.org

[Sven Panne, 2014-09-17 09:02:50]

LGTM (rubberstamped)

[andrew_low, 2014-09-17 13:13:41]

On 2014/09/17 09:02:50, Sven Panne wrote:
> LGTM (rubberstamped)

Thank you. IBM does have an ongoing commitment to maintain this code base in
this repository.

[danno, 2014-09-23 09:06:15]

NOT LGTM, see the copyright comment. I will only be able to land this once we've
gotten that sorted.

https://codereview.chromium.org/571173003/diff/10001/src/ic/ppc/access-compil...
File src/ic/ppc/access-compiler-ppc.cc (right):

https://codereview.chromium.org/571173003/diff/10001/src/ic/ppc/access-compil...
src/ic/ppc/access-compiler-ppc.cc:3: // Copyright IBM Corp. 2012, 2013. All
rights reserved.
Could you remove IBM copyright from all the new files in this directory? I've
been advised that we cannot accept additional copyright information other than
the standard V8 one. I can put you in touch with our legal department here if
you need further clarification, just let me know.

[andrew_low, 2014-09-23 13:15:03]

Ah.. the clash of two legal departments. This'll be fun.

Let me see what we can sort out on the IBM side and I'll let you know.

On 2014/09/23 09:06:15, danno wrote:
> NOT LGTM, see the copyright comment. I will only be able to land this once
we've
> gotten that sorted.
> 
>
https://codereview.chromium.org/571173003/diff/10001/src/ic/ppc/access-compil...
> File src/ic/ppc/access-compiler-ppc.cc (right):
> 
>
https://codereview.chromium.org/571173003/diff/10001/src/ic/ppc/access-compil...
> src/ic/ppc/access-compiler-ppc.cc:3: // Copyright IBM Corp. 2012, 2013. All
> rights reserved.
> Could you remove IBM copyright from all the new files in this directory? I've
> been advised that we cannot accept additional copyright information other than
> the standard V8 one. I can put you in touch with our legal department here if
> you need further clarification, just let me know.

[andrew_low, 2014-10-14 17:59:44]

All sorted out on the IBM side - the copyright has been modified as requested.

Please let me know if there are additional changes required.

[danno, 2014-10-20 08:28:43]

Getting very close. Just a few comments, mostly nits.

https://codereview.chromium.org/571173003/diff/20001/src/ppc/code-stubs-ppc.cc
File src/ppc/code-stubs-ppc.cc (right):

https://codereview.chromium.org/571173003/diff/20001/src/ppc/code-stubs-ppc.c...
src/ppc/code-stubs-ppc.cc:1: // Copyright 2012 the V8 project authors. All
rights reserved.
Here and elsewhere, please change the year to 2014

https://codereview.chromium.org/571173003/diff/20001/src/ppc/code-stubs-ppc.c...
src/ppc/code-stubs-ppc.cc:159: #if V8_TARGET_ARCH_PPC64
Here and elsewhere where possible, can you turn this into a 
if (V8_TARGET_ARCH_PPC) instead? This makes more code compile on both platforms
and reduces the number of #ifdefs.

https://codereview.chromium.org/571173003/diff/20001/src/ppc/code-stubs-ppc.c...
src/ppc/code-stubs-ppc.cc:301: #endif  // roohack
Do you still want the block above in the code?

https://codereview.chromium.org/571173003/diff/20001/src/ppc/code-stubs-ppc.c...
src/ppc/code-stubs-ppc.cc:984: // 
WriteInt32ToHeapNumberStub::GenerateFixedRegStubsAheadOfTime(isolate);
Really delete this line?

https://codereview.chromium.org/571173003/diff/20001/src/ppc/code-stubs-ppc.c...
src/ppc/code-stubs-ppc.cc:4385: #if ABI_USES_FUNCTION_DESCRIPTORS
Is it possible to use just an "if" and not "#if"?

https://codereview.chromium.org/571173003/diff/20001/src/ppc/codegen-ppc.cc
File src/ppc/codegen-ppc.cc (right):

https://codereview.chromium.org/571173003/diff/20001/src/ppc/codegen-ppc.cc#n...
src/ppc/codegen-ppc.cc:377: // heap_number: new heap number
Indentation?

[Sven Panne, 2014-11-05 10:44:35]

On 2014/10/20 08:28:43, danno wrote:
> Getting very close. Just a few comments, mostly nits. [...]

What's the status of this CL? If the nits were addressed, we could land it...

[andrew_low, 2014-11-05 13:26:57]

On 2014/11/05 10:44:35, Sven Panne wrote:
> On 2014/10/20 08:28:43, danno wrote:
> > Getting very close. Just a few comments, mostly nits. [...]
> 
> What's the status of this CL? If the nits were addressed, we could land it...

Apologies. I got caught up in problems getting the code caught up to bleeding
edge which I assume is something you'd want to have in addition to addressing
the small nits.

If you only want the nits addressed - and are ok with the code being currently a
bit more than 2 weeks out of date, I can do that quickly.

[Sven Panne, 2014-11-06 07:23:15]

On 2014/11/05 13:26:57, andrew_low wrote:
> Apologies. I got caught up in problems getting the code caught up to bleeding
> edge which I assume is something you'd want to have in addition to addressing
> the small nits.
> 
> If you only want the nits addressed - and are ok with the code being currently
a
> bit more than 2 weeks out of date, I can do that quickly.

No need to hurry, I was just asking to see if there's something we can do on our
side to get this landed. From time to time I do a "mark/sweep" through my list
of CLs to be reviewed, that was the only reason, not some deadline etc. :-)

[andrew_low, 2014-11-07 18:03:17]

Code is caught up to fairly current (as of Nov 5th). 

I've gone through the nit comments and addressed them.

We need to discuss/agree on the conversion of #if to if as it appears to be more
of a style issue (and may introduce a lot of changes).

Also, one small "beg for forgiveness" comment. We will add it to the list of
things to implement properly, or get back on why it doesn't work as expected.
(it didn't, but that was a while back and may have been another bug)

https://codereview.chromium.org/571173003/diff/20001/src/ppc/code-stubs-ppc.cc
File src/ppc/code-stubs-ppc.cc (right):

https://codereview.chromium.org/571173003/diff/20001/src/ppc/code-stubs-ppc.c...
src/ppc/code-stubs-ppc.cc:1: // Copyright 2012 the V8 project authors. All
rights reserved.
On 2014/10/20 08:28:43, danno wrote:
> Here and elsewhere, please change the year to 2014

Done.

https://codereview.chromium.org/571173003/diff/20001/src/ppc/code-stubs-ppc.c...
src/ppc/code-stubs-ppc.cc:159: #if V8_TARGET_ARCH_PPC64
On 2014/10/20 08:28:43, danno wrote:
> Here and elsewhere where possible, can you turn this into a 
> if (V8_TARGET_ARCH_PPC) instead? This makes more code compile on both
platforms
> and reduces the number of #ifdefs.

I understand the request, but this seems to be a style that would be unique to
PowerPC (specifically the src/ppc and src/ic/ppc directories). I don't see any
usage of this pattern in the other platform directories.

This would probably result in 250-300 change areas (1000+ lines of code change)
- I'm fine with doing it if that is what it takes, but I really don't see much
benefit.

https://codereview.chromium.org/571173003/diff/20001/src/ppc/code-stubs-ppc.c...
src/ppc/code-stubs-ppc.cc:301: #endif  // roohack
On 2014/10/20 08:28:43, danno wrote:
> Do you still want the block above in the code?

No, removed.

https://codereview.chromium.org/571173003/diff/20001/src/ppc/code-stubs-ppc.c...
src/ppc/code-stubs-ppc.cc:984: // 
WriteInt32ToHeapNumberStub::GenerateFixedRegStubsAheadOfTime(isolate);
On 2014/10/20 08:28:43, danno wrote:
> Really delete this line?

When we were doing this work originally including this line caused a crash on
startup.

At this point - by running without this line, it caused this code
https://codereview.chromium.org/571173003/diff/20001/src/ppc/code-stubs-ppc.c...
to not be needed. 

I'd like to simply beg forgiveness for this missing part of the implementation
to help us get past this code contribution. We'll take another run at including
it in the near future.

https://codereview.chromium.org/571173003/diff/20001/src/ppc/code-stubs-ppc.c...
src/ppc/code-stubs-ppc.cc:4385: #if ABI_USES_FUNCTION_DESCRIPTORS
On 2014/10/20 08:28:43, danno wrote:
> Is it possible to use just an "if" and not "#if"?

Yes, but I've got the same comment as before. Yes, we'd run more code through
the compiler for a given bit size -- and it'd go an optimize the always false
case out, but this mostly seems like a code-style thing. 

At least this one is a much smaller set of changes we're talking about. Can you
state a more specific opinion on how much this matters?

https://codereview.chromium.org/571173003/diff/20001/src/ppc/codegen-ppc.cc
File src/ppc/codegen-ppc.cc (right):

https://codereview.chromium.org/571173003/diff/20001/src/ppc/codegen-ppc.cc#n...
src/ppc/codegen-ppc.cc:377: // heap_number: new heap number
On 2014/10/20 08:28:43, danno wrote:
> Indentation?

Done.

[andrew_low, 2014-11-07 19:25:28]

Turns out that the code I was going to beg forgiveness for can simply be
deleted.

https://codereview.chromium.org/571173003/diff/20001/src/ppc/code-stubs-ppc.cc
File src/ppc/code-stubs-ppc.cc (right):

https://codereview.chromium.org/571173003/diff/20001/src/ppc/code-stubs-ppc.c...
src/ppc/code-stubs-ppc.cc:984: // 
WriteInt32ToHeapNumberStub::GenerateFixedRegStubsAheadOfTime(isolate);
On 2014/11/07 18:03:17, andrew_low wrote:
> On 2014/10/20 08:28:43, danno wrote:
> > Really delete this line?
> 
> When we were doing this work originally including this line caused a crash on
> startup.
> 
> At this point - by running without this line, it caused this code
>
https://codereview.chromium.org/571173003/diff/20001/src/ppc/code-stubs-ppc.c...
> to not be needed. 
> 
> I'd like to simply beg forgiveness for this missing part of the implementation
> to help us get past this code contribution. We'll take another run at
including
> it in the near future.

Hah. One of the guys here looked at it, and it appears that this is 'dead' code
even on ARM. So it should be removed from PPC and ARM (along with the associated
code).

[Sven Panne, 2014-11-10 08:47:31]

On 2014/11/07 19:25:28, andrew_low wrote:
> [...] Hah. One of the guys here looked at it, and it appears that this is
'dead' code
> even on ARM. So it should be removed from PPC and ARM (along with the
associated
> code).

Good catch. It seems to be dead for 1.5 years now, starting at the removal VFP2
support. I'll try to nuke that from our code base.

[Sven Panne, 2014-11-10 08:50:28]

LGTM. I'll land it for you. The concerns about #ifs are mainly that you
construct a rather non-trivial hypercube of configurations without even trying
to compile most of them. Similar things have hit us in the past in several other
places, so consider yourself warned... :-D

[Sven Panne, 2014-11-10 09:02:32]

The code review tool sucks: I can't apply the CL because it's considered "too
big", so we'll have to revert to a 70's style workflow: Could you please email
me a tar file containing the ppc subdirectory? :-/

[andrew_low, 2014-11-10 15:14:50]

On 2014/11/10 09:02:32, Sven Panne wrote:
> The code review tool sucks: I can't apply the CL because it's considered "too
> big", so we'll have to revert to a 70's style workflow: Could you please email
> me a tar file containing the ppc subdirectory? :-/

email'd a tgz file (tar + gzip) to your @chromium.org address.

Should contain
src/ppc/assembler-ppc.cc
src/ppc/assembler-ppc.h
src/ppc/assembler-ppc-inl.h
src/ppc/builtins-ppc.cc
src/ppc/codegen-ppc.cc
src/ppc/codegen-ppc.h
src/ppc/code-stubs-ppc.cc
src/ppc/code-stubs-ppc.h
src/ppc/constants-ppc.cc
src/ppc/constants-ppc.h
src/ppc/cpu-ppc.cc
src/ppc/cscope.out
src/ppc/debug-ppc.cc
src/ppc/deoptimizer-ppc.cc
src/ppc/disasm-ppc.cc
src/ppc/frames-ppc.cc
src/ppc/frames-ppc.h
src/ppc/full-codegen-ppc.cc
src/ppc/interface-descriptors-ppc.cc
src/ppc/lithium-codegen-ppc.cc
src/ppc/lithium-codegen-ppc.h
src/ppc/lithium-gap-resolver-ppc.cc
src/ppc/lithium-gap-resolver-ppc.h
src/ppc/lithium-ppc.cc
src/ppc/lithium-ppc.h
src/ppc/macro-assembler-ppc.cc
src/ppc/macro-assembler-ppc.h
src/ppc/regexp-macro-assembler-ppc.cc
src/ppc/regexp-macro-assembler-ppc.h
src/ppc/simulator-ppc.cc
src/ppc/simulator-ppc.h
src/ic/ppc/access-compiler-ppc.cc
src/ic/ppc/handler-compiler-ppc.cc
src/ic/ppc/ic-compiler-ppc.cc
src/ic/ppc/ic-ppc.cc
src/ic/ppc/stub-cache-ppc.cc

that matches this patch

[Sven Panne, 2014-11-11 08:30:44]

Landed as https://codereview.chromium.org/714093002/, closing...