search for: oggen

...tobias at grosser.es> wrote: >>On 07/13/2013 10:13 AM, Star Tan wrote: >>> Hi Tobias, >> >>Hi Star, >> >>thanks for the update. I copied the polly mailing list, as I believe >>this discussion is very valuable. >> >>> I tried to dig into oggenc.ll, one of the two cases we had for >>> scop-detect compile-time overhead, but I may need your help. >>> >>> First, we cannot simply reduce code size by changing part of function >>> defines to declares. Experimental results shows that the relative >>>...

...s at grosser.es> wrote: >>> On 07/13/2013 10:13 AM, Star Tan wrote: >>>> Hi Tobias, >>> >>> Hi Star, >[...] >>> Before we write a patch, we should do some profiling to understand where >>> the overhead comes from. I propose you generate oggenc*16 or even >>> oggen*32 to ensure we get to about 90% Polly-Detect overhead. >>> >>> I would then run Polly under linux 'perf'. Using 'perf record polly-opt >>> ...' and then 'perf report'. If we are lucky, this points us exactly to >&...

...-time overhead comes from the string buffer operation for "INVALID" MACRO in the polly-detect pass. Attached is a hack patch file that simply remove the string buffer operation. This patch file can significantly reduce compile-time overhead when compiling big source code. For example, for oggen*8.ll, the compile time is reduced from 40.5261 ( 51.2%) to 5.8813s (15.9%) with this patch file. However, this patch file does not solve this problem. It only shows the reason why polly-detect pass leads to significant compile-time overhead. Best wishes, Star Tan. At 2013-07-14 17:26:19,"...

...overhead comes from the string buffer operation for "INVALID" MACRO in the polly-detect pass. > Attached is a hack patch file that simply remove the string buffer operation. This patch file can significantly reduce compile-time overhead when compiling big source code. For example, for oggen*8.ll, the compile time is reduced from 40.5261 ( 51.2%) to 5.8813s (15.9%) with this patch file. On top of your patch, I have removed from ScopDetection.cpp all printing of LLVM values, like this: - INVALID(AffFunc, "Non affine access function: " << *AccessFunction); + INVA...

...gt;>>> On 07/13/2013 10:13 AM, Star Tan wrote: >>>>> Hi Tobias, >>>> >>>> Hi Star, >>[...] >>>> Before we write a patch, we should do some profiling to understand where >>>> the overhead comes from. I propose you generate oggenc*16 or even >>>> oggen*32 to ensure we get to about 90% Polly-Detect overhead. >>>> >>>> I would then run Polly under linux 'perf'. Using 'perf record polly-opt >>>> ...' and then 'perf report'. If we are lucky, this points us...

...overhead comes from the string buffer operation for "INVALID" MACRO in the polly-detect pass. > Attached is a hack patch file that simply remove the string buffer operation. This patch file can significantly reduce compile-time overhead when compiling big source code. For example, for oggen*8.ll, the compile time is reduced from 40.5261 ( 51.2%) to 5.8813s (15.9%) with this patch file. Very nice analysis. I just tried it myself and can verify that for oggenc 16x, your patch reduces the scop-detection time from 90 seconds (80 %) to 0.5 seconds (2.5 %). I think there are two problems...

Hi Sebastian, Yes, you have pointed an important reason. If we comment this source code you have listed, then the compile-time overhead for oggenc*8.ll can be reduced from 40.5261 ( 51.2%) to 20.3100 ( 35.7%). I just sent another mail to explain why polly-detect pass leads to significant compile-time overhead. Besides the reason you have pointed, another reason is resulted from those string buffer operations in "INVALID" MACRO. If...

...cience. I 've a project which is > to implement ogg vorbis codec in a DSP for embedded applications. But > before to do that i've to install and evaluate the codec performances > on a PC. So I downloaded the source code from Ogg vobis website and > compiled every file (in folder Oggen) using the Readme's > recommandations. But I don't know where is the executable file (.out). > > Can you help me please? > > Best Regards > > Gwladys > > _______________________________________________ > ogg-dev mailing list > ogg-dev@xiph.org > http://li...

...r, which should only show critical information in graph, I hope such modification is acceptable. Results show that it has almost the same performance improvements as my previous hack patch file, i.e., reducing the compile-time of Polly-detect pass from 90s (>80%) to 0.5s (2.5%) when compiling oggenc 16X. Best wishes, Star Tan Postscript to Tobias: I have also implemented your initial proposal, which uses some class hierarchy to show different Failure information. Unfortunately, I found it would significantly complicate the source code. It not only introduces a lot of dynamic object &quo...