Summary:
LNT has this underused feature that lets you defined tests in python. I used
this feat...

llvm-svn: 253179

The test cases sqlite3 and tramp3d-v4 now work on power. Previously they caused a compiler cras...

Summary:
Additionally, allow the compiler to choose the default C++ ABI library
for MIPS targets...

Reviewers: dsanders

Subscribers: dsanders, llvm-commits

Differential Revision: http://reviews....

Reviewers: dsanders

Subscribers: dsanders, llvm-commits

Differential Revision: http://reviews....

Reviewed by Matthias Braun

llvm-svn: 252594

llvm-svn: 252263

- Writing/Reading files is not a good idea for a compiler/CPU benchmark;
It easily adds noise ...

llvm-svn: 251897

Do not stick with a fixed template but let cmake piece the .test file
together to avoid unnecess...

llvm-svn: 251895

llvm-svn: 251894

llvm-svn: 251893

llvm-svn: 251892

llvm-svn: 251891

llvm-svn: 251890

The script-generated CMake was copying the data files with a wildcard which makes it so that CMa...

This split allows us to have the benchmarking running machinery
(RunSafely.sh, timeit) for the R...

In effect both had /dev/null as input and didn't compute much.

llvm-svn: 251601

This way the build directory does not need to be a subdirectory of the
sourcedir.

llvm-svn: 251483

These were generated from a Python script written by James Molloy.
If you spot any problem, we'l...

The test-suite's minimum CMake version is 3.2, so this should aways be true, further USES_TERMIN...

This introduces a replacement for the Makefile based build system. This replacement build system...

Remove sqlite3 and tramp3d-v4 from the test suite for PowerPC until it is figured out why they f...

Summary:
LNT's benchmarks produce their output into files which are then compared in
various way...

consumer-lame and consumer-typeset are broken after r246877.
Fix consumer-lane by replacing strc...

llvm-svn: 246395

Now also with special casing to work around difference of strip command line syntax between linu...

It break LNT on Darwin because strip does not support --remove-section.

llvm-svn: 243481

llvm-svn: 243400

There is no need to go through the pain and maintain the
PROGRAMS_TO_SKIP variable in all the Ma...

Summary:
The frame_layout test requires support for dynamic stack realignment which
is not avail...

In r241675, print_element was added to polybench.h so make printing out the
arrays faster. Unfor...

Same as with reg_detect, the bechmark output wasn't bit exact. Adding
a FIXME to make sure we de...

This benchmark got some variance in the results due to the bit-exact
printing of the results and...

llvm-svn: 241677

Polybench had large execution time due to the successive call to
fprintf as much as 4000*4000 ti...

http://reviews.llvm.org/D9778
In order to prevent a similar run-time failure in the future, add ...

A substantial source of performance noisiness of the LNT test-suite
are programs running very sh...

Bullet currently needs ~1200s to execute under qemu.
Tramp3D-v4 currently needs ~520s to compile...

It never worked, and isn't supposed to.

It tests that clang properly emits the BLOCK_USE_STRET ...

The added test in this patch uses a template function, so that depending on the
template paramet...

I previously added a #include <unistd.h> to fix an implicit-declaration
problem of getopt(). Thi...

The iOS SDK clang seems to declare implicit function declarations with
-Werror so I decided to f...

implicit-function-declarations are errors by default on the iOS SDK.

llvm-svn: 231943

The previously used fallback sgtty.h is only provided by OS X but not by the
iOS SDK.

llvm-svn:...

This include does not seem to be necessary and is also not present on
iOS SDKs.

llvm-svn: 231833

Define a custom struct instead of relying on NSSize which is not
available in the iOS SDK.

llvm...

llvm-svn: 230715

llvm-svn: 227857

llvm-svn: 226619

The little-endian reference output uses a hyphen rather than an underscore.

llvm-svn: 226045

This should resolve the currently failing test on the ppc64le bot, which is showing as

...

The test fixed by this patch uses __builtin_altivec_lvsl and
__builtin_altivec_lvsr, which are n...

The test has a basic correctness check for every AArch64 Neon intrinsic,
except for a small numb...

llvm-svn: 220538

Some misc. fixes made after 1.0.1 require changes to the input file names for
403.gcc (and thus ...

-fsigned-char is necessary on PowerPC/Linux in order to correctly compile this
benchmark.

llvm-...

This is necessary both on x86_64 and PPC64 Linux using glibc 2.12 and 2.15. It
is related to the...

llvm-svn: 219968

The new Altivec tests I added to projects/test-suite yesterday
contained incorrect checking of t...

This patch adds tests for vec_lvsl and vec_lvsr for all supported
data types. This tests relate...

Do not try to automatically detect and use perf on Linux, thus giving
user more control of which...

It has some interesting data that can be useful, like CPU cycles,
number of branch misses.

llvm...

These programs have really high variance in their execution time, and
frequently produce false p...

llvm-svn: 215470

llvm-svn: 214978

llvm-svn: 214127

llvm-svn: 214126

llvm-svn: 214120

llvm-svn: 214118

A relatively recent patch (r209797) changed RunSafely.sh to use the timeit.sh
script, which uses...

<rdar://problem/17638809>

llvm-svn: 213208

The == operator doesn't exist in POSIX's test. We aren't using globs
anyway, so simplify some us...

This breaks concurrent benchmarking and the benefit is not significant.

llvm-svn: 211499

llvm-svn: 211128

llvm-svn: 210856

llvm-svn: 210788

The test1.c test in the SingleSource/UnitTests/Vector/Altivec
directory makes use of the vec_sld...

support 8 bytes atomic operations.

llvm-svn: 210434

The PowerPC VMX instruction set (also known as Altivec) contains a
number of instructions that h...

This adds a benchmarking-only mode to the test suite, which excludes those
programs unlikely to ...

llvm-svn: 210237

llvm-svn: 209797

llvm-svn: 209748

A wrapper over Linux perf tool which mimics original timeit tool to provide
more accurate time m...

MIPS target.

llvm-svn: 208847

r208341 changed the sqrt used in sphereflake, but only updated
'sphereflake.reference_output'. T...

The default value of XCORE_TARGET_NEEDS_MEMORY is 32 (MB).
It should be noted that the XCore emu...

llvm-svn: 208341

Reference generated from trunk clang for MIPS and verified with clang for
X86_64 (from clang_3.0...

According to C++ 17.6.4.2.1 these are undefined, and after r204082 they
will not compile against...

Correct expected sphereflake.reference_output.small

llvm-svn: 204147

Avoiding issues with different output in different systems.

llvm-svn: 204136

Standard libraries have slightly different results on different targets.
To make the output repr...

Adding a few more corner cases to EH testing, all passing on both
x86_64 and ARM. If this breaks...

Return dummy stopwatch timings for all targets.
Otherwise the timer may cause inconsistency betw...

MultiSource/Benchmarks/mafft

llvm-svn: 203454

llvm-svn: 203453