[turbofan] Add basic support for calling to (a subset of) C functions.

[turbofan] Add basic support for calling to (a subset of) C functions.

This introduces some initial building blocks for calling out to
C/C++ functions directly from TurboFan generated code objects.

R=svenpanne@chromium.org

Committed: https://chromium.googlesource.com/v8/v8/+/a58ba8d80179bf5b6b7245590c82e47fda8c8a5e

[Benedikt Meurer, 2015-06-24 12:34:20]

bmeurer@chromium.org changed reviewers:
+ rodolph.perfetta@arm.com, v8-mips-ports@googlegroups.com

[Benedikt Meurer, 2015-06-24 12:34:20]

Hey Rodolph, MIPS people,

Please add the ports (mips, mips64, arm64).

Thanks,
Benedikt

[paul.l..., 2015-06-25 00:09:16]

Hi Benedikt - Balázs added mips, mips64 ports in
https://codereview.chromium.org/1211443003/

[Benedikt Meurer, 2015-06-25 04:34:29]

Thanks guys!

[Benedikt Meurer, 2015-06-25 05:58:05]

bmeurer@chromium.org changed reviewers:
+ jarin@chromium.org
- rodolph.perfetta@arm.com, v8-mips-ports@googlegroups.com

[Benedikt Meurer, 2015-06-25 05:58:05]

Hey Jaro,

Adding missing bits, it's not used yet. Please take a look.

Thanks,
Benedikt

[Sven Panne, 2015-06-25 07:15:59]

svenpanne@chromium.org changed reviewers:
+ svenpanne@chromium.org

[Sven Panne, 2015-06-25 07:16:00]

LGTM with nits + more test coverage

https://codereview.chromium.org/1205023002/diff/40001/src/compiler/arm/instru...
File src/compiler/arm/instruction-codes-arm.h (right):

https://codereview.chromium.org/1205023002/diff/40001/src/compiler/arm/instru...
src/compiler/arm/instruction-codes-arm.h:96: V(ArmPoke)                       \
I'm not totally happy with the "Poke" suffix, because this term in itself is
usually not semantically connected to stack manipulation. "WriteStackEntry"?
"StoreStackArgument"? Hmmm...

https://codereview.chromium.org/1205023002/diff/40001/test/cctest/compiler/te...
File test/cctest/compiler/test-run-machops.cc (right):

https://codereview.chromium.org/1205023002/diff/40001/test/cctest/compiler/te...
test/cctest/compiler/test-run-machops.cc:5195: 
What about float/double/int64_t/... arguments/return values? If those types are
not supported yet (not clear to me), there should at least be a comment about
this.

Furthermore, please add tests with tons of arguments (probably with mixed
types), otherwise test coverage is a bit :-P * * * for most calling conventions
(0 - 2 args are probably in registers only).

[Benedikt Meurer, 2015-06-25 07:31:10]

https://codereview.chromium.org/1205023002/diff/40001/src/compiler/arm/instru...
File src/compiler/arm/instruction-codes-arm.h (right):

https://codereview.chromium.org/1205023002/diff/40001/src/compiler/arm/instru...
src/compiler/arm/instruction-codes-arm.h:96: V(ArmPoke)                       \
We can fix that naming later. Poke is in line with the ARM64 naming.

https://codereview.chromium.org/1205023002/diff/40001/test/cctest/compiler/te...
File test/cctest/compiler/test-run-machops.cc (right):

https://codereview.chromium.org/1205023002/diff/40001/test/cctest/compiler/te...
test/cctest/compiler/test-run-machops.cc:5195: 
On 2015/06/25 07:15:59, Sven Panne wrote:
> What about float/double/int64_t/... arguments/return values? If those types
are
> not supported yet (not clear to me), there should at least be a comment about
> this.
> 
> Furthermore, please add tests with tons of arguments (probably with mixed
> types), otherwise test coverage is a bit :-P * * * for most calling
conventions
> (0 - 2 args are probably in registers only).

Acknowledged.

[Benedikt Meurer, 2015-06-25 08:28:52]

The CQ bit was checked by bmeurer@chromium.org

[Benedikt Meurer, 2015-06-25 08:28:52]

The patchset sent to the CQ was uploaded after l-g-t-m from
svenpanne@chromium.org
Link to the patchset: https://codereview.chromium.org/1205023002/#ps80001
(title: "Fix argument slots.")

[Benedikt Meurer, 2015-06-25 08:32:41]

Committed patchset #5 (id:80001) manually as
a58ba8d80179bf5b6b7245590c82e47fda8c8a5e (presubmit successful).

[Sven Panne, 2015-06-25 08:35:09]

https://codereview.chromium.org/1205023002/diff/80001/test/cctest/compiler/te...
File test/cctest/compiler/test-run-machops.cc (right):

https://codereview.chromium.org/1205023002/diff/80001/test/cctest/compiler/te...
test/cctest/compiler/test-run-machops.cc:5160: return a + b + c + d + e + f + g
+ h;
Using a non-commutative and non-associative operation would be more helpful
here: Permuted arguments can't be detected with "+".

https://codereview.chromium.org/1205023002/diff/80001/test/cctest/compiler/te...
test/cctest/compiler/test-run-machops.cc:5210: kMachInt32, function, param,
param, param, param,
Using the same actual parameter for all formal parameters might disguise bugs,
e.g. accidentally duplicating/leaving out/permuting args.