native_client/pnacl-llvm-testsuite.git
10 months agollvm-test-suite: only look for machine/endian.h for NaCl newlib. master
Jan Voung [Wed, 18 Dec 2013 23:00:52 +0000 (15:00 -0800)]
llvm-test-suite: only look for machine/endian.h for NaCl newlib.

The host-gcc build/glibc has <endian.h> but not <machine/endian.h>,
so the testsuite was failing to compile the host version.

BUG=none
R=stichnot@chromium.org

Review URL: https://codereview.chromium.org/118663002

10 months agoLLVM Testsuite: Fix endian.h include in security-sha
Derek Schuff [Fri, 13 Dec 2013 18:52:26 +0000 (10:52 -0800)]
LLVM Testsuite: Fix endian.h include in security-sha

My recent fix of PNaCl's newlib headers removed a duplicate copy of
newlib's endian.h. Update sha.c's include to match the right location.

R=stichnot@chromium.org
BUG=none

Review URL: https://codereview.chromium.org/114863002

10 months agoFix the PNaCl build, as PNaCl now lacks <fenv.h>.
Jim Stichnoth [Thu, 5 Dec 2013 19:22:55 +0000 (11:22 -0800)]
Fix the PNaCl build, as PNaCl now lacks <fenv.h>.

This fixes the recent FYI bot failures.

BUG= none
R=dschuff@chromium.org

Review URL: https://codereview.chromium.org/107003002

10 months agoRevert a local pnacl-specific patch to sqlite3.c.
Sam Clegg [Wed, 4 Dec 2013 18:42:46 +0000 (10:42 -0800)]
Revert a local pnacl-specific patch to sqlite3.c.

This patch is no longer needed since flock was added
to the toolchain headers:
https://codereview.chromium.org/57663004/

This patch was originally added in b11fdb813.

TEST=./pnacl/scripts/llvm-test.py --testsuite-configure --testsuite-run --testsuite-report  --arch x86-32 -v -c
R=mseaborn@chromium.org

Review URL: https://codereview.chromium.org/103443004

11 months agoAdd a retry mechanism to work around flakes.
Jim Stichnoth [Wed, 30 Oct 2013 16:37:39 +0000 (09:37 -0700)]
Add a retry mechanism to work around flakes.

For ARM tests, run.py is invoked with --retries=1 to give failed tests a second change.

This depends on the test being idempotent, since local/temporary files can't be cleaned up between retries.

Tests are rerun only when returning exit codes >=128, i.e. crashes.  Timeouts enforced by RunSafely.sh will need to be have separate retry handling.

This also causes NaCl and native tests to be run with stdout/stderr collation, so that tests won't fail due to interleaving differences.  The makefile should be invoked with "make COLLATE=/path/to/collate.py" to enable native test collation.

BUG= none
R=dschuff@chromium.org

Review URL: https://codereview.chromium.org/32193004

12 months agoMake two more tests pass, make progress on a third.
Jim Stichnoth [Thu, 17 Oct 2013 17:36:13 +0000 (10:36 -0700)]
Make two more tests pass, make progress on a third.

Disable the use of getrusage().
MultiSource/Applications/minisat/minisat: unexpected success

Manually declare u_int16_t and u_int32_t.
MultiSource/Applications/oggenc/oggenc: unexpected success

BUG= none
R=jvoung@chromium.org

Review URL: https://codereview.chromium.org/27586003

12 months agoFix a number of tests for PNaCl.
Jim Stichnoth [Wed, 16 Oct 2013 20:06:40 +0000 (13:06 -0700)]
Fix a number of tests for PNaCl.

Add a <memory.h> stub to llvm-test-suite/include/
MultiSource/Benchmarks/MallocBench/gs/gs: unexpected success
SingleSource/Benchmarks/Misc/oourafft: unexpected success
MultiSource/Benchmarks/MiBench/security-rijndael/security-rijndael: unexpected success
MultiSource/Applications/viterbi/viterbi: unexpected success
SingleSource/Benchmarks/CoyoteBench/huffbench: unexpected success
MultiSource/Benchmarks/Bullet/bullet: unexpected success

Suppress error on lack of rename() syscall
MultiSource/Applications/kimwitu++/kc: unexpected success

Disable floating point exception stuff
MultiSource/Benchmarks/tramp3d-v4/tramp3d-v4: unexpected success

Disable FP rounding mode stuff
SingleSource/Regression/C/uint64_to_float: unexpected success

Fake struct flock; failure to find home dir
MultiSource/Applications/sqlite3/sqlite3: unexpected success

Stub out utime()/umask(); disable ENV_HAVE_GETPASS/ENV_HAVE_TIMEGM
MultiSource/Benchmarks/7zip/7zip-benchmark: unexpected success

BUG= none
R=jvoung@chromium.org

Review URL: https://codereview.chromium.org/26238006

12 months agoFix/clarify a comment.
Jim Stichnoth [Mon, 14 Oct 2013 23:50:11 +0000 (16:50 -0700)]
Fix/clarify a comment.

12 months agoMerge remote-tracking branch 'origin/upstream/master' into test-upstream-merge
Jim Stichnoth [Mon, 14 Oct 2013 20:45:47 +0000 (13:45 -0700)]
Merge remote-tracking branch 'origin/upstream/master' into test-upstream-merge

Conflicts:
Makefile.programs
MultiSource/Makefile.multisrc
RunSafely.sh

13 months agoFix a buildbot test failure caused by paths longer than the expected 200 characters... upstream/master
cmatthews [Wed, 18 Sep 2013 04:24:57 +0000 (04:24 +0000)]
Fix a buildbot test failure caused by paths longer than the expected 200 characters.  Replace path buffer size with the standard PATH_MAX from limits.h. I think this is a very safe change, but please revert if it causes problems.

git-svn-id: http://llvm.org/svn/llvm-project/test-suite/trunk@190918 91177308-0d34-0410-b5e6-96231b3b80d8

13 months agoFix undefined behaviour by moving safety test before the memory loads. Patch
nicholas [Tue, 17 Sep 2013 05:26:07 +0000 (05:26 +0000)]
Fix undefined behaviour by moving safety test before the memory loads. Patch
by Jim Stichnoth!

git-svn-id: http://llvm.org/svn/llvm-project/test-suite/trunk@190848 91177308-0d34-0410-b5e6-96231b3b80d8

14 months agoAllow for test suite to be run under Qemu and other host emulators.
rkotler [Fri, 2 Aug 2013 23:58:19 +0000 (23:58 +0000)]
Allow for test suite to be run under Qemu and other host emulators.

Patch by Doug Gilmore of Imagination technologies.

git-svn-id: http://llvm.org/svn/llvm-project/test-suite/trunk@187691 91177308-0d34-0410-b5e6-96231b3b80d8

15 months agosqlite3: Bring back -DSQLITE_THREADSAFE=0 to unbreak the ARM LNT bot.
d0k [Sat, 27 Jul 2013 20:48:37 +0000 (20:48 +0000)]
sqlite3: Bring back -DSQLITE_THREADSAFE=0 to unbreak the ARM LNT bot.

git-svn-id: http://llvm.org/svn/llvm-project/test-suite/trunk@187311 91177308-0d34-0410-b5e6-96231b3b80d8

15 months agonbench: just disable inclusion of malloc.h, LLVM doesn't support platforms that still...
d0k [Sat, 27 Jul 2013 18:49:59 +0000 (18:49 +0000)]
nbench: just disable inclusion of malloc.h, LLVM doesn't support platforms that still need it anyways.

Fixes the test on recent FreeBSDs.

git-svn-id: http://llvm.org/svn/llvm-project/test-suite/trunk@187308 91177308-0d34-0410-b5e6-96231b3b80d8

15 months agosqlite3: Remove preprocessor defines that should be disabled.
d0k [Sat, 27 Jul 2013 17:52:54 +0000 (17:52 +0000)]
sqlite3: Remove preprocessor defines that should be disabled.

The source code uses #ifdef so the =0 had no effect. Should unbreak tests on
FreeBSD which misses fdatasync.

git-svn-id: http://llvm.org/svn/llvm-project/test-suite/trunk@187307 91177308-0d34-0410-b5e6-96231b3b80d8

15 months ago[PowerPC] Support powerpc64le as a syntax-checking target.
wschmidt [Fri, 26 Jul 2013 01:36:25 +0000 (01:36 +0000)]
[PowerPC] Support powerpc64le as a syntax-checking target.

This patch modifies the config.guess file to do the appropriate thing
on a system identifying itself as powerpc64le Linux.  As there
currently is no such system, there is nothing to test.  This is done
for completeness to accompany the LLVM and Clang patches to support
powerpc64le as a syntax-checking target.

git-svn-id: http://llvm.org/svn/llvm-project/test-suite/trunk@187181 91177308-0d34-0410-b5e6-96231b3b80d8

16 months agoAdd the ability to have Mips specific single and multi source tests.
rkotler [Fri, 21 Jun 2013 06:48:43 +0000 (06:48 +0000)]
Add the ability to have Mips specific single and multi source tests.

git-svn-id: http://llvm.org/svn/llvm-project/test-suite/trunk@184530 91177308-0d34-0410-b5e6-96231b3b80d8

17 months agoFix use of EXECUTION_ENVIRONMENT_OVERRIDES.
ddunbar [Thu, 9 May 2013 01:15:27 +0000 (01:15 +0000)]
Fix use of EXECUTION_ENVIRONMENT_OVERRIDES.

git-svn-id: http://llvm.org/svn/llvm-project/test-suite/trunk@181499 91177308-0d34-0410-b5e6-96231b3b80d8

17 months agoForce ffp-contract=off on PowerPC/Linux
hfinkel [Mon, 6 May 2013 23:03:52 +0000 (23:03 +0000)]
Force ffp-contract=off on PowerPC/Linux

In order for the test-suite outputs to match the reference outputs we need to
specify -ffp-contract=off. Not only does extra FMA precision cause differences
between the outputs and the reference outputs, but also between our output and
that of gcc (because the two compilers sometimes make slightly different
contraction choices). Fortunately, providing -ffp-contract=off to both
compilers reveals agreement. Older versions of gcc did not support
-ffp-contract, and if using such a version, you'll need to use the reference
outputs (which you should be able to do now).

git-svn-id: http://llvm.org/svn/llvm-project/test-suite/trunk@181270 91177308-0d34-0410-b5e6-96231b3b80d8

17 months agoAlways enable hashed program output when using reference outputs
hfinkel [Mon, 6 May 2013 22:42:09 +0000 (22:42 +0000)]
Always enable hashed program output when using reference outputs

Because several of the reference outputs require hashed output, running
with reference outputs but without hashed program output yields anomalous
failures.

git-svn-id: http://llvm.org/svn/llvm-project/test-suite/trunk@181260 91177308-0d34-0410-b5e6-96231b3b80d8

17 months agoAdd other big-endian reference outputs
hfinkel [Mon, 6 May 2013 22:20:27 +0000 (22:20 +0000)]
Add other big-endian reference outputs

In the name of getting reasonable feedback from the buildbots (and other
test-suite runs that use the reference outputs), add big-endian reference
outputs for the remaining endian-sensitive tests. Some of these tests should
probably be fixed in other ways, and I'll continue to work on that in the
future.

git-svn-id: http://llvm.org/svn/llvm-project/test-suite/trunk@181259 91177308-0d34-0410-b5e6-96231b3b80d8

17 months agoFix office-ispell big-endian reference output
hfinkel [Mon, 6 May 2013 22:15:13 +0000 (22:15 +0000)]
Fix office-ispell big-endian reference output

As Daniel has mentioned, this test is completely broken (on both big and little
endian systems, it just errors out). On top of that, there seems to be some
endian sensitivity in the code that checks the file magic. As a result, on big
endian systems, the error message contains the full path of the input file
(which makes it impossible to create a reference output).

Here I've slightly modified the test so that it only prints the base path name
on big endian systems (so that I can create a reference output), and I've added
that reference output. This change can be reverted once the test is actually
fixed to do something useful.

git-svn-id: http://llvm.org/svn/llvm-project/test-suite/trunk@181258 91177308-0d34-0410-b5e6-96231b3b80d8

18 months agoKeep Makefile.spec in sync with Makefile.programs.
ddunbar [Sat, 27 Apr 2013 00:07:47 +0000 (00:07 +0000)]
Keep Makefile.spec in sync with Makefile.programs.

git-svn-id: http://llvm.org/svn/llvm-project/test-suite/trunk@180650 91177308-0d34-0410-b5e6-96231b3b80d8

18 months ago[tests] Add big-endian specific reference outputs for some tests.
ddunbar [Fri, 26 Apr 2013 23:40:04 +0000 (23:40 +0000)]
[tests] Add big-endian specific reference outputs for some tests.
 - The outputs themselves are taken from Hal's PPC64 runs.

 - I only added outputs for tests where I verified that the code was not
   endian-portable, and where the tests did real work unrelated to that
   importability. There are other tests (like the bitfield ones, and the cast
   ones) which are not endian-portable but where the nature of the test is tied
   to the exact memory layout. Those tests I feel should be rewritten to be
   endian agnostic.

git-svn-id: http://llvm.org/svn/llvm-project/test-suite/trunk@180649 91177308-0d34-0410-b5e6-96231b3b80d8

18 months ago[tests] Fix automative-bitcount test for big-endian LP64 systems.
ddunbar [Fri, 26 Apr 2013 23:39:54 +0000 (23:39 +0000)]
[tests] Fix automative-bitcount test for big-endian LP64 systems.

git-svn-id: http://llvm.org/svn/llvm-project/test-suite/trunk@180648 91177308-0d34-0410-b5e6-96231b3b80d8

18 months agoAdd support for endian-specific reference outputs.
ddunbar [Fri, 26 Apr 2013 23:39:50 +0000 (23:39 +0000)]
Add support for endian-specific reference outputs.
 - Use naming convention such as ".reference_output.big-endian" or, for type
   specific outputs, something like ".reference_output.big-endian.small".

git-svn-id: http://llvm.org/svn/llvm-project/test-suite/trunk@180647 91177308-0d34-0410-b5e6-96231b3b80d8

18 months agoFix a typo when running with LARGE_PROBLEM_SIZE.
ddunbar [Thu, 25 Apr 2013 19:05:58 +0000 (19:05 +0000)]
Fix a typo when running with LARGE_PROBLEM_SIZE.

git-svn-id: http://llvm.org/svn/llvm-project/test-suite/trunk@180566 91177308-0d34-0410-b5e6-96231b3b80d8

18 months agoFix accesses to uninitialized stack space in alti.isamax.c.
wschmidt [Tue, 9 Apr 2013 18:27:31 +0000 (18:27 +0000)]
Fix accesses to uninitialized stack space in alti.isamax.c.

Hal Finkel recently discovered this test could fail when stack pages
are not pre-initialized to zero.  We eventually traced it down to a
loop in the source that accessed values outside the initialized
portion of an array.  With that fixed, we found another problem in the
epilogue of the same loop that caused some elements of the array to be
wrongly ignored.  The test now passes.

git-svn-id: http://llvm.org/svn/llvm-project/test-suite/trunk@179113 91177308-0d34-0410-b5e6-96231b3b80d8

18 months agollvm-test-suite: Add a rule to run pnacl-finalize before translating.
Jan Voung [Mon, 8 Apr 2013 23:09:53 +0000 (16:09 -0700)]
llvm-test-suite: Add a rule to run pnacl-finalize before translating.

BUG=http://code.google.com/p/nativeclient/issues/detail?id=3348

Review URL: https://codereview.chromium.org/13816002

18 months ago[tests] Modify TSVC static data layout to make benchmarks more stable.
ddunbar [Sun, 7 Apr 2013 03:21:06 +0000 (03:21 +0000)]
[tests] Modify TSVC static data layout to make benchmarks more stable.

As was previously written, the performance of these benchmarks can depend
significantly on the exact addresses assigned to the global data arrays at
runtime. Rewrite the tests to make the relative layout of each of the data
arrays fixed by storing them all in a single global structure. This makes the
behavior more predictable across platforms.

For example, previously on Mac OS X, two of the arrays would end up with
relative offsets at an exact multiple of 4K, which causes problems with some
Intel CPUs. This caused the performance of some benchmarks
(StatementReordering-flt in particular) to be dominated by a CPU specific hazard
instead of the actual benchmark code.

git-svn-id: http://llvm.org/svn/llvm-project/test-suite/trunk@178968 91177308-0d34-0410-b5e6-96231b3b80d8

18 months agofpcmp catching different number of digits
rengolin [Sat, 6 Apr 2013 18:26:47 +0000 (18:26 +0000)]
fpcmp catching different number of digits

teaching fpcmp to catch the case where the two number have different
number of digits (ex. 1.234 vs 1.2345), where before it would compare
the biggest number (1.2345) against the number after the smallest.

This change makes sqlite3 pass on ARM, since the FP output precision
is not specified.

git-svn-id: http://llvm.org/svn/llvm-project/test-suite/trunk@178957 91177308-0d34-0410-b5e6-96231b3b80d8

18 months agoDeclaration vs. Initialization in C89
rengolin [Fri, 5 Apr 2013 16:45:16 +0000 (16:45 +0000)]
Declaration vs. Initialization in C89

git-svn-id: http://llvm.org/svn/llvm-project/test-suite/trunk@178884 91177308-0d34-0410-b5e6-96231b3b80d8

18 months agoAdd return statement to linpack main routine.
uweigand [Fri, 5 Apr 2013 12:46:31 +0000 (12:46 +0000)]
Add return statement to linpack main routine.

The "main" routine in linpack-pc.c is missing a return statement.
This doesn't matter when it is compiled with a C99 compiler, but
when built with a C89 compiler, it may cause the benchmark to
return with a random exit code even after successful execution.

Fixed by adding a "return 0;" statment.

git-svn-id: http://llvm.org/svn/llvm-project/test-suite/trunk@178867 91177308-0d34-0410-b5e6-96231b3b80d8

18 months agoImproving lemon sort criteria makes test pass
rengolin [Thu, 4 Apr 2013 23:33:46 +0000 (23:33 +0000)]
Improving lemon sort criteria makes test pass

As with other tests, the output of lemon was scrambled due to
its sorting criteria. Changing the comparison functions made it
more stable and now it's similar on Linux x86_64, ARM and Darwin.

The reference output is an MD5 of the real output because the
latter is 40M and diff has some trouble parsing it on resource
limited machines, not to mention skimming through a multi-MB
diff on the logs.

git-svn-id: http://llvm.org/svn/llvm-project/test-suite/trunk@178810 91177308-0d34-0410-b5e6-96231b3b80d8

18 months agoRemoving sources of uncertainty from ClamAV
rengolin [Wed, 3 Apr 2013 15:07:42 +0000 (15:07 +0000)]
Removing sources of uncertainty from ClamAV

ClamAV is an anti-virus that scans directories by their
inode order, and thus being random in nature. This patch
does two things to address that: passes the files to scan
as a list (on the Makefile) and sorts the definition files
as they're read (readdb).

The result now should be identical on any machine/os/arch.

git-svn-id: http://llvm.org/svn/llvm-project/test-suite/trunk@178660 91177308-0d34-0410-b5e6-96231b3b80d8

18 months agoAdd test for atomic addition and subtraction. <rdar://problem/13549656>
void [Wed, 3 Apr 2013 00:04:26 +0000 (00:04 +0000)]
Add test for atomic addition and subtraction. <rdar://problem/13549656>

git-svn-id: http://llvm.org/svn/llvm-project/test-suite/trunk@178594 91177308-0d34-0410-b5e6-96231b3b80d8

18 months agoAdd a missing open-paren for $(TARGET_OS) reference.
willschm [Tue, 2 Apr 2013 18:53:25 +0000 (18:53 +0000)]
Add a missing open-paren for $(TARGET_OS) reference.

git-svn-id: http://llvm.org/svn/llvm-project/test-suite/trunk@178560 91177308-0d34-0410-b5e6-96231b3b80d8

18 months agoUse fgets() to remove a Darwin-specific warning from stdout of the test.
ddunbar [Tue, 2 Apr 2013 17:58:37 +0000 (17:58 +0000)]
Use fgets() to remove a Darwin-specific warning from stdout of the test.

git-svn-id: http://llvm.org/svn/llvm-project/test-suite/trunk@178555 91177308-0d34-0410-b5e6-96231b3b80d8

18 months agoIncreasing stability of sort in Burg
rengolin [Tue, 2 Apr 2013 16:15:23 +0000 (16:15 +0000)]
Increasing stability of sort in Burg

The state matrix was being generated slightly different
on x86_64 and ARM Linux than on the reference output because
the sort wasn't stable. This new version tries to stabilize
the comparison, but since the structures are big and nested,
there is room for further improvement, if needed.

git-svn-id: http://llvm.org/svn/llvm-project/test-suite/trunk@178547 91177308-0d34-0410-b5e6-96231b3b80d8

18 months agoTry to make anagram's sort more stable
rengolin [Tue, 2 Apr 2013 11:47:17 +0000 (11:47 +0000)]
Try to make anagram's sort more stable

The reference_output did not match the output of the
program, neither with current comparator, not with
the one changed in this patch. Though, the new version
should be more stable, and possibly produce the same
order in different library implementations.

If not, it's quite possible that the problem is not
in the comparator per se, but in how auGlobalFrequency
is built, which will warrant a deeper investigation.

git-svn-id: http://llvm.org/svn/llvm-project/test-suite/trunk@178535 91177308-0d34-0410-b5e6-96231b3b80d8

18 months agoAdd reference data output for the Altivec tests.
willschm [Mon, 1 Apr 2013 21:22:28 +0000 (21:22 +0000)]
Add reference data output for the Altivec tests.

git-svn-id: http://llvm.org/svn/llvm-project/test-suite/trunk@178501 91177308-0d34-0410-b5e6-96231b3b80d8

19 months ago483lxalancbmk needs more stack space.
grosbach [Fri, 29 Mar 2013 22:31:29 +0000 (22:31 +0000)]
483lxalancbmk needs more stack space.

For the reference workload, some Darwin targets don't have enough default
stack space.

git-svn-id: http://llvm.org/svn/llvm-project/test-suite/trunk@178379 91177308-0d34-0410-b5e6-96231b3b80d8

19 months agoTestcase for PR15555 (r177680).
void [Thu, 21 Mar 2013 23:30:50 +0000 (23:30 +0000)]
Testcase for PR15555 (r177680).

git-svn-id: http://llvm.org/svn/llvm-project/test-suite/trunk@177681 91177308-0d34-0410-b5e6-96231b3b80d8

19 months agoDisable the darwin -force_cpusubtype_ALL parameter for Linux targets.
willschm [Tue, 19 Mar 2013 15:24:08 +0000 (15:24 +0000)]
Disable the darwin -force_cpusubtype_ALL parameter for Linux targets.

git-svn-id: http://llvm.org/svn/llvm-project/test-suite/trunk@177405 91177308-0d34-0410-b5e6-96231b3b80d8

19 months agoDisable Shootout-C++ except on non-Darwin ARM
rengolin [Mon, 18 Mar 2013 21:27:35 +0000 (21:27 +0000)]
Disable Shootout-C++ except on non-Darwin ARM

Moving except test to EH dir to disable it on
non-Darwin ARM until EHABI support gets better

git-svn-id: http://llvm.org/svn/llvm-project/test-suite/trunk@177322 91177308-0d34-0410-b5e6-96231b3b80d8

19 months agoFix Darwin SjLj tests back
rengolin [Mon, 18 Mar 2013 20:39:40 +0000 (20:39 +0000)]
Fix Darwin SjLj tests back

git-svn-id: http://llvm.org/svn/llvm-project/test-suite/trunk@177316 91177308-0d34-0410-b5e6-96231b3b80d8

19 months agoDefine the RUNTIMELIMIT before we include other Makefiles.
mcrosier [Mon, 18 Mar 2013 19:18:49 +0000 (19:18 +0000)]
Define the RUNTIMELIMIT before we include other Makefiles.

git-svn-id: http://llvm.org/svn/llvm-project/test-suite/trunk@177308 91177308-0d34-0410-b5e6-96231b3b80d8

19 months agoDisabling EH tests on ARM for now
rengolin [Mon, 18 Mar 2013 10:09:10 +0000 (10:09 +0000)]
Disabling EH tests on ARM for now

git-svn-id: http://llvm.org/svn/llvm-project/test-suite/trunk@177266 91177308-0d34-0410-b5e6-96231b3b80d8

19 months agosmallpt exceeds the 500s timeout for some -O0 configurations, so bump it a bit.
mcrosier [Fri, 15 Mar 2013 21:51:41 +0000 (21:51 +0000)]
smallpt exceeds the 500s timeout for some -O0 configurations, so bump it a bit.
rdar://13420880

git-svn-id: http://llvm.org/svn/llvm-project/test-suite/trunk@177191 91177308-0d34-0410-b5e6-96231b3b80d8

19 months agoMiBench: Add simple RNG to avoid library diffs
rengolin [Fri, 15 Mar 2013 12:38:44 +0000 (12:38 +0000)]
MiBench: Add simple RNG to avoid library diffs

Now the reference output is specific to the local implementation, avoiding
dependencies on the library that could change from release to release on
the same architectures.

Tested on x86_64 and ARMv7 (32-bits), so should work on any 32-bit and
64-bit architectures. It's also probably safe on 16-bit archs.

git-svn-id: http://llvm.org/svn/llvm-project/test-suite/trunk@177150 91177308-0d34-0410-b5e6-96231b3b80d8

19 months agoAdd simple RNG to avoid library differences
rengolin [Fri, 15 Mar 2013 11:52:08 +0000 (11:52 +0000)]
Add simple RNG to avoid library differences

Now the reference output is specific to the local implementation, avoiding
dependencies on the library that could change from release to release on
the same architectures.

Tested on x86_64 and ARMv7 (32-bits), so should work on any 32-bit and
64-bit architectures. It's also probably safe on 16-bit archs.

git-svn-id: http://llvm.org/svn/llvm-project/test-suite/trunk@177148 91177308-0d34-0410-b5e6-96231b3b80d8

19 months agoPerform the atomic operations on a global, rather then a local stack variable.
mcrosier [Tue, 12 Mar 2013 22:07:00 +0000 (22:07 +0000)]
Perform the atomic operations on a global, rather then a local stack variable.
This was causing failures on one of our internal -O0 testers.

The issue appears to be that spills/refills to/from the stack are clearing the
monitor and causing the atomic operations to fail, which results in the benchmark
going into an infinite loop.  This seems to only happen at -O0 because the
fast-regalloc is generating a reload in-between the excusive load and exclusive
store to a stack slot adjacent to the local variable x.
rdar://13363219

git-svn-id: http://llvm.org/svn/llvm-project/test-suite/trunk@176903 91177308-0d34-0410-b5e6-96231b3b80d8

19 months agoTypo.
mcrosier [Tue, 12 Mar 2013 19:14:28 +0000 (19:14 +0000)]
Typo.

git-svn-id: http://llvm.org/svn/llvm-project/test-suite/trunk@176885 91177308-0d34-0410-b5e6-96231b3b80d8

19 months agoBound the recursion to avoid stack overflow on unoptimized builds.
mcrosier [Wed, 6 Mar 2013 17:39:28 +0000 (17:39 +0000)]
Bound the recursion to avoid stack overflow on unoptimized builds.

git-svn-id: http://llvm.org/svn/llvm-project/test-suite/trunk@176561 91177308-0d34-0410-b5e6-96231b3b80d8

19 months agoForce signed char on McCat to pass on ARM
rengolin [Wed, 6 Mar 2013 10:05:32 +0000 (10:05 +0000)]
Force signed char on McCat to pass on ARM

git-svn-id: http://llvm.org/svn/llvm-project/test-suite/trunk@176549 91177308-0d34-0410-b5e6-96231b3b80d8

20 months agoWhen using -stdlib=libstdc++, it needs to be specified for the link as well.
bwilson [Thu, 28 Feb 2013 08:19:01 +0000 (08:19 +0000)]
When using -stdlib=libstdc++, it needs to be specified for the link as well.

git-svn-id: http://llvm.org/svn/llvm-project/test-suite/trunk@176239 91177308-0d34-0410-b5e6-96231b3b80d8

20 months agoChange 252.eon and 447.dealII to explicitly build with libstdc++.
bwilson [Thu, 28 Feb 2013 07:57:20 +0000 (07:57 +0000)]
Change 252.eon and 447.dealII to explicitly build with libstdc++.
They both have issues when using libc++.

git-svn-id: http://llvm.org/svn/llvm-project/test-suite/trunk@176238 91177308-0d34-0410-b5e6-96231b3b80d8

20 months agoFurther fixes for libc++. <rdar://problem/13302264>
bwilson [Wed, 27 Feb 2013 18:34:10 +0000 (18:34 +0000)]
Further fixes for libc++. <rdar://problem/13302264>

git-svn-id: http://llvm.org/svn/llvm-project/test-suite/trunk@176196 91177308-0d34-0410-b5e6-96231b3b80d8

20 months agoAttempt to fix a buildbot that is using libc++ by default. <rdar://13302264>
bwilson [Wed, 27 Feb 2013 18:28:28 +0000 (18:28 +0000)]
Attempt to fix a buildbot that is using libc++ by default. <rdar://13302264>

git-svn-id: http://llvm.org/svn/llvm-project/test-suite/trunk@176194 91177308-0d34-0410-b5e6-96231b3b80d8

20 months agoUse LLVMCC_EMITIR_FLAG rather than -emit-llvm in the 7zip Makefile, unbreaking
baldrick [Sun, 17 Feb 2013 15:21:11 +0000 (15:21 +0000)]
Use LLVMCC_EMITIR_FLAG rather than -emit-llvm in the 7zip Makefile, unbreaking
the dragonegg build.

git-svn-id: http://llvm.org/svn/llvm-project/test-suite/trunk@175396 91177308-0d34-0410-b5e6-96231b3b80d8

20 months agoTargetConfig.mk should be included unconditionally - the defines are
lhames [Sun, 17 Feb 2013 05:29:32 +0000 (05:29 +0000)]
TargetConfig.mk should be included unconditionally - the defines are
required regardless of whether LLVM's configuration makefile is
available.

git-svn-id: http://llvm.org/svn/llvm-project/test-suite/trunk@175388 91177308-0d34-0410-b5e6-96231b3b80d8

20 months agoTurned back on Polybench tests on ARM since we now have Small Problem Size support.
mgottesman [Fri, 15 Feb 2013 03:15:41 +0000 (03:15 +0000)]
Turned back on Polybench tests on ARM since we now have Small Problem Size support.

rdar://11907111.

git-svn-id: http://llvm.org/svn/llvm-project/test-suite/trunk@175247 91177308-0d34-0410-b5e6-96231b3b80d8

20 months agoAdded Small Problem Size support to most tests in Benchmarks/Polybench/stencil/.
mgottesman [Fri, 15 Feb 2013 03:13:28 +0000 (03:13 +0000)]
Added Small Problem Size support to most tests in Benchmarks/Polybench/stencil/.

Specifically I added said support for:

1. adi.
2. fdtd-2d.
3. fdtd-apml.
4. jacobi-2d-imper.
5. seidel-2d.

rdar://11907111.

git-svn-id: http://llvm.org/svn/llvm-project/test-suite/trunk@175246 91177308-0d34-0410-b5e6-96231b3b80d8

20 months agoAdded SmallProblemSize support for SingleSource/Benchmarks/Polybench/medley/floyd...
mgottesman [Fri, 15 Feb 2013 02:59:06 +0000 (02:59 +0000)]
Added SmallProblemSize support for SingleSource/Benchmarks/Polybench/medley/floyd-warshall.

rdar://11907111.

git-svn-id: http://llvm.org/svn/llvm-project/test-suite/trunk@175245 91177308-0d34-0410-b5e6-96231b3b80d8

20 months agoAdded SmallProblemSize support for SingleSource/Benchmarks/Polybench/linear-algebra...
mgottesman [Fri, 15 Feb 2013 02:55:39 +0000 (02:55 +0000)]
Added SmallProblemSize support for SingleSource/Benchmarks/Polybench/linear-algebra/solvers/gramschmidt. rdar://11907111.

git-svn-id: http://llvm.org/svn/llvm-project/test-suite/trunk@175244 91177308-0d34-0410-b5e6-96231b3b80d8

20 months agoAdded SmallProblemSize support for SingleSource/Benchmarks/Polybench/datamining/...
mgottesman [Fri, 15 Feb 2013 02:45:29 +0000 (02:45 +0000)]
Added SmallProblemSize support for SingleSource/Benchmarks/Polybench/datamining/{correlation,covariance}. rdar://11907111.

git-svn-id: http://llvm.org/svn/llvm-project/test-suite/trunk@175243 91177308-0d34-0410-b5e6-96231b3b80d8

20 months agoAdded SmallProblemSize support for most Polybench Kernel Tests.
mgottesman [Fri, 15 Feb 2013 01:34:32 +0000 (01:34 +0000)]
Added SmallProblemSize support for most Polybench Kernel Tests.

Specifically I added it for: 2mm, 3mm, cholesky, doitgen, gemm, symm, syr2k,
syrk, and trmm.

I additionally converted gesummv to use hashed outputs as a bonus.

rdar://11907111.

git-svn-id: http://llvm.org/svn/llvm-project/test-suite/trunk@175240 91177308-0d34-0410-b5e6-96231b3b80d8

20 months agoUpdate Makefile.{singlesrc,multisrc} to use X_TARGET_FLAGS
willschm [Fri, 8 Feb 2013 14:50:04 +0000 (14:50 +0000)]
Update Makefile.{singlesrc,multisrc} to use X_TARGET_FLAGS

The LNT test harness handles TARGET_FLAGS specially, and prevents other
users of the variable from functioning as intended, i.e.   "TARGET_FLAGS
+= -maltivec" does not work as intended.
X_TARGET_FLAGS exists to work around this, but was not utilized in
Makefile.{singlesrc,multisrc}.   Update those makefiles to use
X_TARGET_FLAGS, and update Altivec/Makefile appropriately.

git-svn-id: http://llvm.org/svn/llvm-project/test-suite/trunk@174712 91177308-0d34-0410-b5e6-96231b3b80d8

21 months agoAdd missing include of Makefile.config to Altivec Makefile.
wschmidt [Tue, 22 Jan 2013 15:30:19 +0000 (15:30 +0000)]
Add missing include of Makefile.config to Altivec Makefile.

git-svn-id: http://llvm.org/svn/llvm-project/test-suite/trunk@173167 91177308-0d34-0410-b5e6-96231b3b80d8

21 months agoThis test fixes the special compile rules for 7zip so that .bc files can be properly...
mgottesman [Mon, 21 Jan 2013 21:13:08 +0000 (21:13 +0000)]
This test fixes the special compile rules for 7zip so that .bc files can be properly compiled if they are a target.

This fixes 'make TEST=profile'. Patch by Alastair Murray.

git-svn-id: http://llvm.org/svn/llvm-project/test-suite/trunk@173091 91177308-0d34-0410-b5e6-96231b3b80d8

21 months agoReorder some environment variable settings to be actually useful.
wschmidt [Mon, 21 Jan 2013 21:08:46 +0000 (21:08 +0000)]
Reorder some environment variable settings to be actually useful.

git-svn-id: http://llvm.org/svn/llvm-project/test-suite/trunk@173089 91177308-0d34-0410-b5e6-96231b3b80d8

21 months agoadd -lstdc++ to smallpt LDFLAGS
predmond [Mon, 21 Jan 2013 16:53:01 +0000 (16:53 +0000)]
add -lstdc++ to smallpt LDFLAGS

- necessary when TEST=profile

Patch from Alastair Murray

git-svn-id: http://llvm.org/svn/llvm-project/test-suite/trunk@173059 91177308-0d34-0410-b5e6-96231b3b80d8

21 months agoRe-enable #pragma openmp as suggested by Duncan Sands
predmond [Mon, 21 Jan 2013 16:49:00 +0000 (16:49 +0000)]
Re-enable #pragma openmp as suggested by Duncan Sands

- openmp is supported by dragonegg
- #pragma openmp is ignored by clang

git-svn-id: http://llvm.org/svn/llvm-project/test-suite/trunk@173057 91177308-0d34-0410-b5e6-96231b3b80d8

21 months agoExplicitly specify "signed char" when testing extensions.
tnorthover [Thu, 17 Jan 2013 13:24:38 +0000 (13:24 +0000)]
Explicitly specify "signed char" when testing extensions.

The ARM ABIs specify that char is unsigned by default, in contrast to many
others. This test seemed more concerned about the extensions that get applied
to "signed char" than whether that's the default so just add "signed" to the
types.

git-svn-id: http://llvm.org/svn/llvm-project/test-suite/trunk@172716 91177308-0d34-0410-b5e6-96231b3b80d8

21 months agoRemoving Livermore Loops, since the parameters() function was adapted wrongly.
rengolin [Tue, 8 Jan 2013 15:44:53 +0000 (15:44 +0000)]
Removing Livermore Loops, since the parameters() function was adapted wrongly.
I'll re-add the benchmark later, when more thought is put into it.

git-svn-id: http://llvm.org/svn/llvm-project/test-suite/trunk@171859 91177308-0d34-0410-b5e6-96231b3b80d8

21 months agoadd smallpt benchmark to test-suite
predmond [Mon, 7 Jan 2013 15:25:10 +0000 (15:25 +0000)]
add smallpt benchmark to test-suite

original source from http://www.kevinbeason.com/smallpt/ with the following
modifications:
- commented out openmp pragma
- disabled writing to output image to file

git-svn-id: http://llvm.org/svn/llvm-project/test-suite/trunk@171745 91177308-0d34-0410-b5e6-96231b3b80d8

21 months agoSome of the kernels assumed that the pointers are aligned. This commit makes the
nadav [Fri, 4 Jan 2013 22:46:16 +0000 (22:46 +0000)]
Some of the kernels assumed that the pointers are aligned. This commit makes the
arrays aligned. This commit also increases the number of executions for some of the
faster loops.

This should fix build.lnt_clang-x86_64-darwin10-RA_i386-O3.

git-svn-id: http://llvm.org/svn/llvm-project/test-suite/trunk@171546 91177308-0d34-0410-b5e6-96231b3b80d8

22 months agoRevert 171018, 171019, 171116, 171138.
bwilson [Fri, 28 Dec 2012 16:19:14 +0000 (16:19 +0000)]
Revert 171018, 171019, 171116, 171138.

This new test has been failing consistently on multiple Apple buildbots,
and Nadav has not been around to fix it.

git-svn-id: http://llvm.org/svn/llvm-project/test-suite/trunk@171201 91177308-0d34-0410-b5e6-96231b3b80d8

22 months agoAdded -pthreads to LDFLAGS (thanks to Alastair Murray) and CFLAGS/CXXFLAGS (thanks...
mgottesman [Fri, 28 Dec 2012 10:47:42 +0000 (10:47 +0000)]
Added -pthreads to LDFLAGS (thanks to Alastair Murray) and CFLAGS/CXXFLAGS (thanks to my paranoia).

This fixes a compilation issue that arose since using -pthread on OS X
is not necessary while it is on Linux. The reason pthreads is not
necessary to be explicitly linked in on OS X is because it is already
in libSystem.dylib.

git-svn-id: http://llvm.org/svn/llvm-project/test-suite/trunk@171187 91177308-0d34-0410-b5e6-96231b3b80d8

22 months agoRemoved extraneous .o files that should not be in the test suite source bundle =/.
mgottesman [Fri, 28 Dec 2012 09:07:29 +0000 (09:07 +0000)]
Removed extraneous .o files that should not be in the test suite source bundle =/.

git-svn-id: http://llvm.org/svn/llvm-project/test-suite/trunk@171182 91177308-0d34-0410-b5e6-96231b3b80d8

22 months agoUpdate reference output to match what I'm seeing from this test.
bwilson [Thu, 27 Dec 2012 03:50:49 +0000 (03:50 +0000)]
Update reference output to match what I'm seeing from this test.

I don't claim to understand what the correct answer should be, but this
has been causing our buildbots to be broken for several days now and I just
want to get them to pass again.

git-svn-id: http://llvm.org/svn/llvm-project/test-suite/trunk@171138 91177308-0d34-0410-b5e6-96231b3b80d8

22 months agoChange the test order.
nadav [Wed, 26 Dec 2012 20:57:28 +0000 (20:57 +0000)]
Change the test order.

git-svn-id: http://llvm.org/svn/llvm-project/test-suite/trunk@171116 91177308-0d34-0410-b5e6-96231b3b80d8

22 months agoupdate the reference file
nadav [Mon, 24 Dec 2012 07:56:57 +0000 (07:56 +0000)]
update the reference file

git-svn-id: http://llvm.org/svn/llvm-project/test-suite/trunk@171019 91177308-0d34-0410-b5e6-96231b3b80d8

22 months agoAdd example21 to the test.
nadav [Mon, 24 Dec 2012 06:31:45 +0000 (06:31 +0000)]
Add example21 to the test.

git-svn-id: http://llvm.org/svn/llvm-project/test-suite/trunk@171018 91177308-0d34-0410-b5e6-96231b3b80d8

22 months agoFix warnings
nadav [Mon, 24 Dec 2012 06:05:58 +0000 (06:05 +0000)]
Fix warnings

git-svn-id: http://llvm.org/svn/llvm-project/test-suite/trunk@171017 91177308-0d34-0410-b5e6-96231b3b80d8

22 months agoRemove an unused variable to fix a warning.
nadav [Thu, 20 Dec 2012 00:28:59 +0000 (00:28 +0000)]
Remove an unused variable to fix a warning.

git-svn-id: http://llvm.org/svn/llvm-project/test-suite/trunk@170621 91177308-0d34-0410-b5e6-96231b3b80d8

22 months agoMade 7zip makefile friendly for case-sensitive file systems.
mgottesman [Mon, 17 Dec 2012 19:21:32 +0000 (19:21 +0000)]
Made 7zip makefile friendly for case-sensitive file systems.

git-svn-id: http://llvm.org/svn/llvm-project/test-suite/trunk@170356 91177308-0d34-0410-b5e6-96231b3b80d8

22 months agoEnsure that 7zip-benchmark links against a c++ standard library, ala d0k's wisdom...
mgottesman [Mon, 17 Dec 2012 05:30:43 +0000 (05:30 +0000)]
Ensure that 7zip-benchmark links against a c++ standard library, ala d0k's wisdom in r169995.

git-svn-id: http://llvm.org/svn/llvm-project/test-suite/trunk@170307 91177308-0d34-0410-b5e6-96231b3b80d8

22 months agoAdded test 7zip-benchmark.
mgottesman [Mon, 17 Dec 2012 05:23:25 +0000 (05:23 +0000)]
Added test 7zip-benchmark.

git-svn-id: http://llvm.org/svn/llvm-project/test-suite/trunk@170306 91177308-0d34-0410-b5e6-96231b3b80d8

22 months agoWe do not support Alpha any more, and whatever bugs GCC-alpha once had are not intere...
nadav [Fri, 14 Dec 2012 22:13:53 +0000 (22:13 +0000)]
We do not support Alpha any more, and whatever bugs GCC-alpha once had are not interesting.

git-svn-id: http://llvm.org/svn/llvm-project/test-suite/trunk@170249 91177308-0d34-0410-b5e6-96231b3b80d8

22 months agoMake sure ALAC links a c++ standard library.
d0k [Wed, 12 Dec 2012 17:48:08 +0000 (17:48 +0000)]
Make sure ALAC links a c++ standard library.

This is a bit hacky, but the same way as all other tests do it.

git-svn-id: http://llvm.org/svn/llvm-project/test-suite/trunk@169995 91177308-0d34-0410-b5e6-96231b3b80d8

22 months agoChange main.cpp to assume that an input wav file => output caff and vis-a-versa....
mgottesman [Tue, 11 Dec 2012 22:42:01 +0000 (22:42 +0000)]
Change main.cpp to assume that an input wav file => output caff and vis-a-versa. Also make reference output correct.

git-svn-id: http://llvm.org/svn/llvm-project/test-suite/trunk@169926 91177308-0d34-0410-b5e6-96231b3b80d8

22 months agoRemoved use of implementation defined behavior unsigned x = 'multiletter char' for...
mgottesman [Tue, 11 Dec 2012 22:41:58 +0000 (22:41 +0000)]
Removed use of implementation defined behavior unsigned x = 'multiletter char' for proper integers.

git-svn-id: http://llvm.org/svn/llvm-project/test-suite/trunk@169925 91177308-0d34-0410-b5e6-96231b3b80d8

22 months agoFixed incorrect case of folder MultiSource so ALAC can load MultiSource/Makefile...
mgottesman [Tue, 11 Dec 2012 21:45:40 +0000 (21:45 +0000)]
Fixed incorrect case of folder MultiSource so ALAC can load MultiSource/Makefile.multisrc on case sensitive systems.

git-svn-id: http://llvm.org/svn/llvm-project/test-suite/trunk@169918 91177308-0d34-0410-b5e6-96231b3b80d8

22 months agostepanov_v1p2: Fix a possible non-determism if subsequent clock() calls return
ddunbar [Tue, 11 Dec 2012 21:16:50 +0000 (21:16 +0000)]
stepanov_v1p2: Fix a possible non-determism if subsequent clock() calls return
an identical value.
 - In such a case this test could divide by 0 and end up printing an inf.

git-svn-id: http://llvm.org/svn/llvm-project/test-suite/trunk@169914 91177308-0d34-0410-b5e6-96231b3b80d8

22 months agoFixed path to ALAC in LICENSE.TXT.
mgottesman [Tue, 11 Dec 2012 20:48:42 +0000 (20:48 +0000)]
Fixed path to ALAC in LICENSE.TXT.

git-svn-id: http://llvm.org/svn/llvm-project/test-suite/trunk@169913 91177308-0d34-0410-b5e6-96231b3b80d8

22 months agoSeparated ALAC into two separate tests, one for decode and the other for encode ala...
mgottesman [Tue, 11 Dec 2012 20:46:43 +0000 (20:46 +0000)]
Separated ALAC into two separate tests, one for decode and the other for encode ala Daniel Dunbar's suggestion.

git-svn-id: http://llvm.org/svn/llvm-project/test-suite/trunk@169912 91177308-0d34-0410-b5e6-96231b3b80d8

22 months agoForgot to add ALAC to LICENSE.txt. There is now a note therein.
mgottesman [Tue, 11 Dec 2012 06:07:42 +0000 (06:07 +0000)]
Forgot to add ALAC to LICENSE.txt. There is now a note therein.

This is necessary since ALAC has been released under the Apple Open Source License, not the NCSA/University of Illinois open-source license.

git-svn-id: http://llvm.org/svn/llvm-project/test-suite/trunk@169820 91177308-0d34-0410-b5e6-96231b3b80d8

22 months agoAdded test ALAC (Apple Lossless Audio Codec).
mgottesman [Tue, 11 Dec 2012 01:58:37 +0000 (01:58 +0000)]
Added test ALAC (Apple Lossless Audio Codec).

git-svn-id: http://llvm.org/svn/llvm-project/test-suite/trunk@169808 91177308-0d34-0410-b5e6-96231b3b80d8

22 months agoThis test uses the C++ standard library, so it needs to link against it. This is
nicholas [Sat, 8 Dec 2012 23:49:06 +0000 (23:49 +0000)]
This test uses the C++ standard library, so it needs to link against it. This is
how SingleSource/UnitTests/Makefile does it.

git-svn-id: http://llvm.org/svn/llvm-project/test-suite/trunk@169685 91177308-0d34-0410-b5e6-96231b3b80d8