llvm-svn: 269247

PREPARE: allows to specify steps that should happen before the benchmark
runs. While you should ...

llvm-svn: 269159

llvm-svn: 269158

svn crashed during the commit and it was somehow partial only apparently.

llvm-svn: 269078

These are unittests checking mostly for simd operations.
They are compiled with and without opti...

llvm-svn: 268584

Do not set CMAKE_C_FLAGS etc. in the cache files because these often get
overridden by the user....

The per-benchmark files should only append to CPPFLAGS/LDFLAGS and never
override them. Otherwis...

The Examples are not used in the normal test-suite run so nobody noticed
they were broken. Also ...

llvm-svn: 267627

- Rename the caches to iphoneos to better match possible watchos/tvos
additions.
- Do not chan...

cmake_minimum_required(VERSION 3.2) sets CMP0002 policy to NEW
which will produce an error when ...

This prevents name clashes for test targets CMake generates on arm.

llvm-svn: 267439

Differential Revision: http://reviews.llvm.org/D19423

llvm-svn: 267243

llvm-svn: 267213

These are similar in spirit like iOS.cmake from the llvm repository. The
main difference is that...

llvm-svn: 266785

llvm-svn: 266782

Also remove the unused test_suite_root setting.

llvm-svn: 266781

You can now specify a list of test modules in lit.site.cfg.
This allows to add external test mod...

The run module runs the benchmark. The trick is that you can leave it
out to not run the benchm...

llvm-svn: 266778

llvm-svn: 266777

We should preferably take llvm-size and llvm-profdata at the same path
as the clang binary.

llv...

CMAKE_C_FLAGS etc. already are cached and need no alternate description.

llvm-svn: 266769

llvm-svn: 266762

llvm-svn: 266678

LNT will now collected details about a run to help triage and diagnose
performance problems. LN...

After r256038, we pass --target to configure from LNT. LNT detects the
OS based on the target t...

llvm-svn: 265071

llvm-svn: 264907

llvm-svn: 264900

llvm-svn: 264622

This reverts most of the cmake hackery from r264499, instead
communicate to lit which tests are ...

llvm-svn: 264613

This should make all steps transparent and reproducible without using
the lit tool.

This also c...

llvm-svn: 264500

This fixes the problem of compiletimes not getting collected corrected.

llvm-svn: 264499

llvm-svn: 264498

This way we can get rid of the llvm_prepend() function.

llvm-svn: 264497

The cmake build has never set ARCH to x86_64 or x86_64h, so no need to
check for it (things work...

This was probably automatically generated by accident when we generated
the CMakeLists.txt files...

llvm-svn: 264494

llvm-svn: 264493

llvm-svn: 264492

llvm-svn: 264242

llvm-svn: 264240

llvm-svn: 264220

Specify the working directory explicitely (if the benchmark depends on
it) instead of relying on...

llvm-svn: 264101

llvm-svn: 264100

When profiling, we were only mutating the *original* runscript. But the run_under argument is es...

Lit complains if you give it metrics as floats.

llvm-svn: 264045

This fixes parsing for any subsequent options.

llvm-svn: 264044

Host `strip` tool cannot recognize binary format of files built for
another target architecture....

llvm-svn: 263901

This is what the rest of the testsuite does which makes it more
convenient to grep for errors di...

llvm-svn: 263615

llvm-svn: 263614

llvm-svn: 263613

llvm-svn: 263612

Previously we only collected compile_times, this adds a new metric
link_time. This currently has...

This also allows us to split run_under and remove device handling into
separate modules.

llvm-s...

Appending the last line with "exit XXX" is usually done by RunSafely. I
added the same capabilit...

The testplan object describes the sequence of steps that will get
executed to run the benchmark ...

llvm-svn: 263531

Use an llvm style 2-space indentation.

llvm-svn: 263492

llvm-svn: 263484

llvm-svn: 263465

llvm-svn: 263456

It's meant to be '.note*', not .note! We weren't stripping all the note sections, causing hashes...

These options were added to timeit.c by Matthias and are required for running SPEC with --use-pe...

llvm-svn: 263315

llvm-svn: 263306

llvm-svn: 263284

r261397 changed how LDFLAGS were added to the target under cmake from
using target_link_librarie...

llvm-svn: 263189

llvm-svn: 263188

There are some common configurations of the test-suite which really
should be encoded right into...

Instead of trying to list every benchmark in the subdirectories in
PROGRAMS_TO_SKIP (and failing...

No need to an internal variable BENCHMARKING_ONLY that is just set
depending on TEST_SUITE_BENCH...

The -std=gnu89 flag is necessary to build this benchmark with SPEC 2000 v1.3
(we need traditiona...

llvm-svn: 263147

This went undetected before, because:
- RunSafely sets exitcodes to zero anyway and only puts th...

llvm-svn: 263145

You can now use "llvm-lit -Dprevious=xxx.json" to make lit load the
results of a previous run an...

This simply records the filesize of the executable.

llvm-svn: 262306

We need it for several modules so it makes sense to determine the name
of the main executable on...

llvm-svn: 262304

I had a similar hashing support lying around in my personal repository
for weeks now, thinking t...

Stripping out certain sections of the binary under test is important to get a stable hash, and i...

This adds the "hash" metric as computed by the HashProgramOutput.sh script (used by the Makefile...

This allows a .test script to specify a command to get a metric for the test. For example:

METR...

llvm-svn: 261854

This does not include the fortran benchmarks.

llvm-svn: 261816

Mark some options that are untypical for the testsuite as advanced, and
clear the advanced statu...

Use `check_symbol_exists`() to check macros defined by compiler and deduce
enabled MIPS features...

These changes were needed to make the spec 2000 and spec 2006 benchmark
compile correctly on the...

llvm-svn: 261400