Make py import not assume dir is "pnacl-subzero". Avoid autovect in crosstest.

Make py import not assume dir is "pnacl-subzero". Avoid autovect in crosstest.

Derek's CL to check out subzero calls the source directory
"subzero", and the file header comments call the directory
"subzero". Just make the python sys.path munging for
importing pydir more generic.

Also change crosstest to not run the raw LLVM "opt" with
optimizations (only use it for ABI stabilization passes).
Instead run pnacl-clang with -O2. Otherwise, newer NACL_SDK
versions include a newer LLVM "opt" binary which
autovectorizes and may generate vector IR that is not
handled by Subzero yet.

E.g.,
LLVM ERROR: Invalid PNaCl instruction:   %1 = insertelement <4 x i32> undef, i32 %0, i32 0
w/ pepper_canary to version 37, revision 274873


BUG=none
TEST=make -f Makefile.standalone check
R=stichnot@chromium.org, wala@chromium.org

Committed: https://gerrit.chromium.org/gerrit/gitweb?p=native_client/pnacl-subzero.git;a=commit;h=1248a6d

[Jim Stichnoth, 2014-06-05 21:55:08]

Adding Matt who has also encountered the autovect issue.

[wala, 2014-06-05 22:14:30]

lgtm

[Jim Stichnoth, 2014-06-05 23:06:10]

Also lgtm.

I'm a tiny bit wary of taking away -O2 from build-pnacl-ir.py (though I guess
there is a conservation of -O2 between clang and opt).  One goal of the script
is to create tests that are representative of real-world code which would
presumably be compiled with -O2.  As such, we probably should have had both
clang and opt compiling with -O2.  Maybe we can make that happen once Subzero
has vector support.

[jvoung (off chromium), 2014-06-05 23:34:10]

On 2014/06/05 23:06:10, stichnot wrote:
> Also lgtm.
> 
> I'm a tiny bit wary of taking away -O2 from build-pnacl-ir.py (though I guess
> there is a conservation of -O2 between clang and opt).  One goal of the script
> is to create tests that are representative of real-world code which would
> presumably be compiled with -O2.  As such, we probably should have had both
> clang and opt compiling with -O2.  Maybe we can make that happen once Subzero
> has vector support.

I think running clang with -O2 would actually be more standard w/ how people use
PNaCl's tools. Usually people just have CC=pnacl-clang, CXX=pnacl-clang++,
LD=pnacl-clang++ in their build scripts, and so don't invoke opt by themselves.

For PNaCl we internally set up pnacl-clang++/pnacl-ld to run "opt" at link time,
and use -std-link-opts instead of -O2. I guess it is a bit different from the
normally pnacl-usage, in that there would have been a -std-link-opts run in
between the ABI passes. Otherwise, I think clang -O2 and opt -O2 are roughly the
same -- both would have included -slp-vectorizer, -loop-vectorizer and so clang
would have had the same problem. It's just that "pnacl-clang.py" filters out
those passes with -fno-vectorize, etc. when invoking clang.

[jvoung (off chromium), 2014-06-06 15:48:01]

Committed patchset #2 manually as r1248a6d (presubmit successful).