For C-Ray GCC needed 20% less time, which is comparable tot he 28% increase of LLVM for Botan CAST-256.
Also GCC 4.8 delivered almost 19% more C/s for John the Ripper.
On the other hand, LLVM 3.3 really shines for Botan KASUMI (an increase of more than 60% is huge).
Wish: Tell me why.
What operations do the tests use? You have that for OpenMP, but not for the other tests.
Essentially the wish boils down to the question whether that improvement is relevant for my projects.
An example: Googling gave me this: KASUMI, the block cipher used in 3G telephony.
How are the chances, that LLVM is hand-optimized for that?
A second wish: Please mark the compilation time tests somehow. It?s really jarring to see a benchmark which looks like a huge difference, just to realize a few moments later that it is completely irrelevant for production code.
Announcement
Collapse
No announcement yet.
LLVM/Clang 3.3 Very Competitive To GCC 4.8
Collapse
X
-
Originally posted by Marc Driftmeyer View PostWhat's with the OpenMP tests thrown in? Wake me when both compiler suites are OpenMP ready.
Show me some OpenCL tests.
OpenCL is only usable with R600g (I think) on LLVM. GCC has no OpenCL support so you'd have LLVM giving you SOME results, and GCC sitting at 0.
Leave a comment:
-
Originally posted by Marc Driftmeyer View PostWhat's with the OpenMP tests thrown in? Wake me when both compiler suites are OpenMP ready.
Show me some OpenCL tests.
Leave a comment:
-
What's with the OpenMP tests thrown in? Wake me when both compiler suites are OpenMP ready.
Show me some OpenCL tests.
Leave a comment:
-
Great benchmark, but Intel icc could be added as reference, then it would be perfect.
Leave a comment:
-
Very interesting.
But as a future reference, Michael, there isn't much point in doing such tests on ffmpeg, since ffmpeg's critical code paths are mostly hand-tuned assembly. So it is no wonder that there is almost no difference there, and there never will be.
Leave a comment:
-
LLVM/Clang 3.3 Very Competitive To GCC 4.8
Phoronix: LLVM/Clang 3.3 Very Competitive To GCC 4.8
Benchmarks for sharing this weekend are looking at the performance of GCC 4.7, GCC 4.8, LLVM/Clang 3.2, and the latest LLVM/Clang 3.3 development code. How does the performance of the newly released GCC 4.8.0 compare to the yet-to-be-released LLVM/Clang 3.3? It's interesting.
Tags: None
Leave a comment: