Subzero, MIPS32: Enable MIPS support in lit tests

This patch enables running a couple more of lit tests for MIPS32

R=stichnot@chromium.org

Committed: https://gerrit.chromium.org/gerrit/gitweb?p=native_client/pnacl-subzero.git;a=commit;h=7ad028ea364d4795fd3fa441e36341db6849941e

[Stefan Maksimovic, 2016-10-28 15:12:01]

makdstefan@gmail.com changed reviewers:
+ jaydeep.patil@imgtec.com, sagar.thakur@imgtec.com, stichnot@chromium.org

[Jim Stichnoth, 2016-10-28 17:42:07]

https://codereview.chromium.org/2448193008/diff/1/tests_lit/llvm2ice_tests/64...
File tests_lit/llvm2ice_tests/64bit.pnacl.ll (right):

https://codereview.chromium.org/2448193008/diff/1/tests_lit/llvm2ice_tests/64...
tests_lit/llvm2ice_tests/64bit.pnacl.ll:123: ; MIPS32-O2: 	sw	a3,20(sp)
You may want to reconsider whether to use hard-coded stack offsets in the lit
tests.

This is especially relevant for O2 mode, where future optimizations may change
register assignments and stack frame layout, and then there's a chance that all
such tests need to be reexamined and fixed.

It's not so relevant for Om1 mode, where we try to do minimal optimizations and
transformations so the output is much more likely to be stable.

I see that the x86 tests are also at risk of being unstable, but the tests
haven't been changed in the last 20 months, so maybe I'm being overly cautious
here.

So it's your decision whether to generalize the test or leave it as is.

https://codereview.chromium.org/2448193008/diff/1/tests_lit/llvm2ice_tests/co...
File tests_lit/llvm2ice_tests/convert.ll (right):

https://codereview.chromium.org/2448193008/diff/1/tests_lit/llvm2ice_tests/co...
tests_lit/llvm2ice_tests/convert.ll:460: 
Remove trailing "whitespace" (last blank line) to make git happy.

https://codereview.chromium.org/2448193008/diff/1/tests_lit/llvm2ice_tests/fp...
File tests_lit/llvm2ice_tests/fp.call_ret.ll (right):

https://codereview.chromium.org/2448193008/diff/1/tests_lit/llvm2ice_tests/fp...
tests_lit/llvm2ice_tests/fp.call_ret.ll:135: ; MIPS32: 	lwc1	$f0,0(v0)	164:
R_MIPS_LO16	.L$float$3f9d70a4
Can you reduce some of the whitespace (here and below) to pull the lines into
80-col?

Note that FileCheck treats whitespace of different lengths the same.

https://codereview.chromium.org/2448193008/diff/1/tests_lit/llvm2ice_tests/fu...
File tests_lit/llvm2ice_tests/fused-alloca.ll (right):

https://codereview.chromium.org/2448193008/diff/1/tests_lit/llvm2ice_tests/fu...
tests_lit/llvm2ice_tests/fused-alloca.ll:35: ; MIPS32: 	addiu	sp,sp,-48
More examples of hard-coded stack frame size and offsets.

However, it's probably reasonable for alloca-specific tests.

https://codereview.chromium.org/2448193008/diff/1/tests_lit/llvm2ice_tests/ve...
File tests_lit/llvm2ice_tests/vector-arg.ll (right):

https://codereview.chromium.org/2448193008/diff/1/tests_lit/llvm2ice_tests/ve...
tests_lit/llvm2ice_tests/vector-arg.ll:32: ; MIPS32: 	lw	v0,16(sp)
Here are a lot more instances of "dangerously" hard-coding stack offsets into
the tests.  Again, your decision whether to modify or keep as-is.

[Stefan Maksimovic, 2016-10-31 10:33:08]

https://codereview.chromium.org/2448193008/diff/1/tests_lit/llvm2ice_tests/64...
File tests_lit/llvm2ice_tests/64bit.pnacl.ll (right):

https://codereview.chromium.org/2448193008/diff/1/tests_lit/llvm2ice_tests/64...
tests_lit/llvm2ice_tests/64bit.pnacl.ll:123: ; MIPS32-O2: 	sw	a3,20(sp)
On 2016/10/28 17:42:07, Jim Stichnoth wrote:
> You may want to reconsider whether to use hard-coded stack offsets in the lit
> tests.
> 
> This is especially relevant for O2 mode, where future optimizations may change
> register assignments and stack frame layout, and then there's a chance that
all
> such tests need to be reexamined and fixed.
> 
> It's not so relevant for Om1 mode, where we try to do minimal optimizations
and
> transformations so the output is much more likely to be stable.
> 
> I see that the x86 tests are also at risk of being unstable, but the tests
> haven't been changed in the last 20 months, so maybe I'm being overly cautious
> here.
> 
> So it's your decision whether to generalize the test or leave it as is.

Done.

https://codereview.chromium.org/2448193008/diff/1/tests_lit/llvm2ice_tests/co...
File tests_lit/llvm2ice_tests/convert.ll (right):

https://codereview.chromium.org/2448193008/diff/1/tests_lit/llvm2ice_tests/co...
tests_lit/llvm2ice_tests/convert.ll:460: 
On 2016/10/28 17:42:07, Jim Stichnoth wrote:
> Remove trailing "whitespace" (last blank line) to make git happy.

Done.

https://codereview.chromium.org/2448193008/diff/1/tests_lit/llvm2ice_tests/fp...
File tests_lit/llvm2ice_tests/fp.call_ret.ll (right):

https://codereview.chromium.org/2448193008/diff/1/tests_lit/llvm2ice_tests/fp...
tests_lit/llvm2ice_tests/fp.call_ret.ll:135: ; MIPS32: 	lwc1	$f0,0(v0)	164:
R_MIPS_LO16	.L$float$3f9d70a4
On 2016/10/28 17:42:07, Jim Stichnoth wrote:
> Can you reduce some of the whitespace (here and below) to pull the lines into
> 80-col?
> 
> Note that FileCheck treats whitespace of different lengths the same.

Done.

https://codereview.chromium.org/2448193008/diff/1/tests_lit/llvm2ice_tests/fu...
File tests_lit/llvm2ice_tests/fused-alloca.ll (right):

https://codereview.chromium.org/2448193008/diff/1/tests_lit/llvm2ice_tests/fu...
tests_lit/llvm2ice_tests/fused-alloca.ll:35: ; MIPS32: 	addiu	sp,sp,-48
On 2016/10/28 17:42:07, Jim Stichnoth wrote:
> More examples of hard-coded stack frame size and offsets.
> 
> However, it's probably reasonable for alloca-specific tests.

I chose to edit them anyway, if that's ok?

https://codereview.chromium.org/2448193008/diff/1/tests_lit/llvm2ice_tests/ve...
File tests_lit/llvm2ice_tests/vector-arg.ll (right):

https://codereview.chromium.org/2448193008/diff/1/tests_lit/llvm2ice_tests/ve...
tests_lit/llvm2ice_tests/vector-arg.ll:32: ; MIPS32: 	lw	v0,16(sp)
On 2016/10/28 17:42:07, Jim Stichnoth wrote:
> Here are a lot more instances of "dangerously" hard-coding stack offsets into
> the tests.  Again, your decision whether to modify or keep as-is.

Done.

[Jim Stichnoth, 2016-10-31 13:40:24]

lgtm

[Jim Stichnoth, 2016-10-31 13:41:49]

Description was changed from

==========
This patch enables running a couple more of lit tests for MIPS32
==========

to

==========
This patch enables running a couple more of lit tests for MIPS32

R=stichnot@chromium.org

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

[Jim Stichnoth, 2016-10-31 13:41:50]

Committed patchset #2 (id:20001) manually as
7ad028ea364d4795fd3fa441e36341db6849941e (presubmit successful).