Subzero: Improve handling of alloca instructions of constant size.

Subzero: Improve handling of alloca instructions of constant size.

PNaCl simplifies varargs calls by creating a known-size argument array with an alloca instruction, and passing the address of that argument array.  These alloca instructions don't necessarily require use of a frame pointer, freeing up the frame pointer register for normal register allocation.

These varargs calls sometimes show up in cold paths of hot functions, so increasing the number of registers available to the register allocator can produce tangible gains.

This patch does a simple recognition of these alloca patterns, and on x86 doesn't force a frame pointer if all alloca instructions are suitable.

Future work is to avoid saving the alloca result as a local variable, and instead rematerialize the address as needed with respect to the stack or frame pointer.

BUG= none
R=jpp@chromium.org

Committed: https://gerrit.chromium.org/gerrit/gitweb?p=native_client/pnacl-subzero.git;a=commit;h=55f931f65f307116c2def40c3b11ec93998a907e

[Jim Stichnoth, 2015-09-23 15:20:22]

stichnot@chromium.org changed reviewers:
+ jpp@chromium.org, sehr@google.com

[Jim Stichnoth, 2015-09-23 15:21:12]

I'll be adding more specific tests later today.

[John, 2015-09-23 15:28:59]

lgtm

https://codereview.chromium.org/1361803002/diff/40001/src/IceTargetLoweringAR...
File src/IceTargetLoweringARM32.h (right):

https://codereview.chromium.org/1361803002/diff/40001/src/IceTargetLoweringAR...
src/IceTargetLoweringARM32.h:113: size_t BasicFrameOffset, size_t StackAdjBytes,
What's the point of the ARM changes? you're passing an extra parameter, but it
seems you're not using it?... maybe revert and backport this later?

[Jim Stichnoth, 2015-09-23 20:17:02]

https://codereview.chromium.org/1361803002/diff/40001/src/IceTargetLoweringAR...
File src/IceTargetLoweringARM32.h (right):

https://codereview.chromium.org/1361803002/diff/40001/src/IceTargetLoweringAR...
src/IceTargetLoweringARM32.h:113: size_t BasicFrameOffset, size_t StackAdjBytes,
On 2015/09/23 15:28:59, John wrote:
> What's the point of the ARM changes? you're passing an extra parameter, but it
> seems you're not using it?... maybe revert and backport this later?

Good point - for some reason I thought finishArgumentLowering() was virtual and
I needed to change the signature everywhere.  Reverted.

[Jim Stichnoth, 2015-09-23 22:33:57]

I added a couple basic tests of the new lowering.

It would be nice to test that various esp and ebp offsets and adjustments are
correct, both before and after the alloca instructions execute, but it seems
really hard, if not impossible, to do this in lit.  I guess spec2k (especially
gcc) is the better test...

[Jim Stichnoth, 2015-09-23 23:33:12]

Committed patchset #5 (id:80001) manually as
55f931f65f307116c2def40c3b11ec93998a907e (presubmit successful).