Re: [opensource-dev] Openjpeg/KDU the cold hard metrics
I found a simple fix to the OpenJPEG v2 library that corrects the previous crashes that occurred with the test harness. The fix is attached to https://jira.secondlife.com/browse/SNOW-361 I also updated Robin's spreadsheet to reflect the improved performance. OpenJPEG v2 is now marginally quicker than the OpenJPEG v1.3.0 that ships with Viewer 2. Previously the OpenJPEG v2 library was crashing when it skipped over packets for higher discard levels. The library was skipping over too few bytes. Earlier test results for OpenJPEG v2 were performing a full decode. The Reduce feature (that performs partial decodes when the image was incomplete) had to be disabled because of the crashes. Now that the crashes are fixed, the updated test results for v2 reflect the same partial decodes that OpenJPEG v1.3.0 was performing. The fix will likely be submitted to the OpenJPEG project tomorrow night, when I have more time. Sheet Spotter -Original Message- From: Robin Cornelius [mailto:robin.cornel...@gmail.com] Sent: September 23, 2010 5:37 AM To: Sheet Spotter Cc: opensource-dev Subject: Re: [opensource-dev] Openjpeg/KDU the cold hard metrics On Thu, Sep 23, 2010 at 3:40 AM, Sheet Spotter sheet.spot...@gmail.com wrote: There may be another option for obtaining a common set of J2C files for comparison. Viewer 2 installs 784 JPEG-2000 files (*.j2c) in the local_assets folder under the main install folder (e.g., C:\Program Files\SecondLifeViewer2). The j2k_metric test harness decoded all 784 files from the local_assets folder without errors. The OpenJPEG 1.3.0 library was 3.7 times slower than the KDU library that comes with Viewer 2.1.1.208043 (the current release version). Hi (again) Sheet and everyone else, I've started another sheet with a fresh set of results on another system of mine, please feel free to add your own tests rows https://spreadsheets.google.com/ccc?key=0AiSrUP47_VxIdEZ4NmlSY281UXFac0ZZTkV jWGJtV1Ehl=enauthkey=CIGY9M4O If you want to compare to KDU you need to compare to KDU on your system don't compare to by KDU results or anyone elses, thats not a fair comparision. OJP 2.0 is not playing nice with the local_asset texture set needs some work. Robin ___ Policies and (un)subscribe information available here: http://wiki.secondlife.com/wiki/OpenSource-Dev Please read the policies before posting to keep unmoderated posting privileges
Re: [opensource-dev] Openjpeg/KDU the cold hard metrics
I created a crude profiling tool to determine where to focus on improving the performance of OpenJPEG v2. The profiling tool records the number of function calls and combined execution time for every function in the main source files (openjpeg.c, j2k.c, dwt.c, tcd.c, t1.c, mqc.c, and raw.c). Results from running the test harness on the 748 files from the local_assets folder were recorded in a CSV (Comma-Separated Value) format. The results were placed here: http://pastebin.com/7K9PCFcB The three columns in the file are: 1. Function name 2. Number of times the function was called 3. Combined execution time for the function. The order the rows appear in the file is not significant. (You can sort the rows in any order you like after importing the file into a spreadsheet.) Enabling the profiling tool definitely distorts the results. Overall execution time for the test increased from 21 seconds to 129 seconds. The crude profiling tool was able to identify where most of the decoding time is spent. Only a small number of routines are responsible for most of the execution time. Hopefully this information can lead to improvements in the performance of the OpenJPEG v2 library. Sheet Spotter -Original Message- From: Robin Cornelius [mailto:robin.cornel...@gmail.com] Sent: September 23, 2010 5:37 AM To: Sheet Spotter Cc: opensource-dev Subject: Re: [opensource-dev] Openjpeg/KDU the cold hard metrics On Thu, Sep 23, 2010 at 3:40 AM, Sheet Spotter sheet.spot...@gmail.com wrote: There may be another option for obtaining a common set of J2C files for comparison. Viewer 2 installs 784 JPEG-2000 files (*.j2c) in the local_assets folder under the main install folder (e.g., C:\Program Files\SecondLifeViewer2). The j2k_metric test harness decoded all 784 files from the local_assets folder without errors. The OpenJPEG 1.3.0 library was 3.7 times slower than the KDU library that comes with Viewer 2.1.1.208043 (the current release version). Hi (again) Sheet and everyone else, I've started another sheet with a fresh set of results on another system of mine, please feel free to add your own tests rows https://spreadsheets.google.com/ccc?key=0AiSrUP47_VxIdEZ4NmlSY281UXFac0ZZTkV jWGJtV1Ehl=enauthkey=CIGY9M4O If you want to compare to KDU you need to compare to KDU on your system don't compare to by KDU results or anyone elses, thats not a fair comparision. OJP 2.0 is not playing nice with the local_asset texture set needs some work. Robin ___ Policies and (un)subscribe information available here: http://wiki.secondlife.com/wiki/OpenSource-Dev Please read the policies before posting to keep unmoderated posting privileges
Re: [opensource-dev] Openjpeg/KDU the cold hard metrics
Does the execution time *include* the time of sub-functions called, or does it *exclude*? On Sun, Sep 26, 2010 at 3:02 AM, Sheet Spotter sheet.spot...@gmail.comwrote: I created a crude profiling tool to determine where to focus on improving the performance of OpenJPEG v2. ___ Policies and (un)subscribe information available here: http://wiki.secondlife.com/wiki/OpenSource-Dev Please read the policies before posting to keep unmoderated posting privileges
Re: [opensource-dev] Openjpeg/KDU the cold hard metrics
Execution time from the profile includes sub-functions. For example, the tcd_t1_decode function calls the t1_decode_cblks function multiple times. The execution time for tcd_t1_decode includes the time for every call to t1_decode_cblks. The t1_decode_cblk function was responsible for most of the execution time (121 out of 128 seconds). Almost all of this time in this function is split between three other functions t1_dec_sigpass (51 seconds), t1_dec_refpass (36 seconds), and t1_dec_cInpass (35 seconds). Each of these functions make millions of calls to only a few other functions. The profiler was written very simply. The profiler consists of a single CProfiler class in C++. The constructor records the start time using a high resolution timer. The destructor calculates the elapsed time and updates a collection of counters. Declaring a variable of the CProfiler class is all that's required to capture the execution time and call count. The constructor accepts a single argument, which is used as a unique identifier. Results are grouped by this identifier. The name of the function was used at the unique identifier. After running the simulation a static method is called to dump the statistics. The most expensive part of the coding was adding CProfiler variables to the start of each function. No attempt was made to subtract out the time spent in sub-routines. No attempt was made to subtract out the impact of the CProfiler class itself. The CProfiler class will not be submitted to the OpenJPEG source code. Comments in their Google group suggest they prefer to limit the source code to C and are not interested in using C++ objects. Sheet Spotter _ From: lee.sai...@gmail.com [mailto:lee.sai...@gmail.com] On Behalf Of Ponzu Sent: September 26, 2010 10:03 AM To: Sheet Spotter Cc: Robin Cornelius; opensource-dev Subject: Re: [opensource-dev] Openjpeg/KDU the cold hard metrics Does the execution time *include* the time of sub-functions called, or does it *exclude*? On Sun, Sep 26, 2010 at 3:02 AM, Sheet Spotter sheet.spot...@gmail.com wrote: I created a crude profiling tool to determine where to focus on improving the performance of OpenJPEG v2. ___ Policies and (un)subscribe information available here: http://wiki.secondlife.com/wiki/OpenSource-Dev Please read the policies before posting to keep unmoderated posting privileges
Re: [opensource-dev] Openjpeg/KDU the cold hard metrics
On Thu, Sep 23, 2010 at 3:40 AM, Sheet Spotter sheet.spot...@gmail.com wrote: There may be another option for obtaining a common set of J2C files for comparison. Viewer 2 installs 784 JPEG-2000 files (*.j2c) in the local_assets folder under the main install folder (e.g., C:\Program Files\SecondLifeViewer2). Doh!, good call. we could set the default path for the test to use those +1 on that idea. The j2k_metric test harness decoded all 784 files from the local_assets folder without errors. The OpenJPEG 1.3.0 library was 3.7 times slower than the KDU library that comes with Viewer 2.1.1.208043 (the current release version). That's significantly different to my latest results, but i need to validate on that texture set. I'm expecting we will see massive differences between relative performance from CPU to CPU. I'll put up a google doc that any one can add to and we can all (if you want) add our individual results against this set. I have been unsuccessful at implementing the additional functions to make the OpenJPEG v2 interface compatible with the earlier v1.3.0 interface. The test application aborted with buffer overrun errors. Oopsie! Did you see https://jira.secondlife.com/browse/SNOW-361 ? i used that code to interface to OJP 2, you do need to build openjpeg with a compile flag to expose two functions that have been deprecated , USE_OPJ_DEPRECATED is what is needed in openjpeg. There was one small surprise. Only two threads were in use for both the OpenJPEG v1.3.0 and KDU tests. I expected KDU to use all eight available threads. I'm not overly concerned on threading for this reason, JPEG2000s often are GB/TB images, Nasa etc use them for planet surveys and all kinds of things, multi core support there is really valuable and will clearly speed up the total process. In SL we are decoding lots of small images so my argument is that KDU using all 4/8/what ever cores is little real world advantage over 4/8/whatever threads decoding separate images in parallel and for the end user the effects will be very similar (assuming a similar single thread decode performance) and infact depending on how KDU is written, it might even have a performance penalty if it spawns threads for each decode as starting threads has a cost, that cost is worth paying for single very large images, but could be too much for multiple images as the thread start costs adds up. I suspect the KDU api has an option to use/not use multi core so may be LL have some control over this and can optimised? Robin ___ Policies and (un)subscribe information available here: http://wiki.secondlife.com/wiki/OpenSource-Dev Please read the policies before posting to keep unmoderated posting privileges
Re: [opensource-dev] Openjpeg/KDU the cold hard metrics
On Thu, Sep 23, 2010 at 3:40 AM, Sheet Spotter sheet.spot...@gmail.com wrote: There may be another option for obtaining a common set of J2C files for comparison. Viewer 2 installs 784 JPEG-2000 files (*.j2c) in the local_assets folder under the main install folder (e.g., C:\Program Files\SecondLifeViewer2). The j2k_metric test harness decoded all 784 files from the local_assets folder without errors. The OpenJPEG 1.3.0 library was 3.7 times slower than the KDU library that comes with Viewer 2.1.1.208043 (the current release version). Hi (again) Sheet and everyone else, I've started another sheet with a fresh set of results on another system of mine, please feel free to add your own tests rows https://spreadsheets.google.com/ccc?key=0AiSrUP47_VxIdEZ4NmlSY281UXFac0ZZTkVjWGJtV1Ehl=enauthkey=CIGY9M4O If you want to compare to KDU you need to compare to KDU on your system don't compare to by KDU results or anyone elses, thats not a fair comparision. OJP 2.0 is not playing nice with the local_asset texture set needs some work. Robin ___ Policies and (un)subscribe information available here: http://wiki.secondlife.com/wiki/OpenSource-Dev Please read the policies before posting to keep unmoderated posting privileges
Re: [opensource-dev] Openjpeg/KDU the cold hard metrics
How many textures do you need? I bet I have a bunch that are perfectly OK for this purpose as well as for commercial use. And if you need more I can generate them using filters. I.e.; this is not a block. From: Robin Cornelius robin.cornel...@gmail.com To: Sheet Spotter sheet.spot...@gmail.com Cc: opensource-dev opensource-dev@lists.secondlife.com Sent: Wed, September 22, 2010 3:55:20 AM Subject: Re: [opensource-dev] Openjpeg/KDU the cold hard metrics On Wed, Sep 22, 2010 at 4:01 AM, Sheet Spotter sheet.spot...@gmail.com wrote: I would love to see Robin's test harness. I would also love to see the images that failed with v2. For the record I cannot give you (or anyone else) the images that failed, I was testing against a random selection of images pulled from the texture cache (and yes i know how to get a complete image out the cache) i wanted it to be a fair cross section of real SL textures not contrived test cases. I can probably people on an individual basis UUIDs of failed images but i'm afraid you will have to obtain the images yourselves I can't just go around distributing SL textures that are not mine. Maybe we can make a collection of test textures that we (the collective we) own, if an individual copyright holder has textures they can donate we can build a collection of test textures that can be used for metrics now and in the future. I started diving into the v2 branch of OpenJPEG a few days ago, after research JPEG 2000 for a week. Tackling the bugs from the failed images might be a good way to become more familiar with the source code. I can get full back traces to the offending lines which will be some help. I posted a very minor patch correcting some simple build problems with the DLL version of OpenJPEG V2. http://code.google.com/p/openjpeg/issues/detail?id=40 Yes i had to fix a few issues to get the dll to build, nothing major. Thank you for publishing your results, Robin. They are encouraging me to continue looking at OpenJPEG v2. Thanks OJP V2 looks like its worth the effort, so far using it with the viewer is failing badly. Encode now i've stopped it crashing is failing causing a bake loop of hell which consumes massive amounts of CPU time and spams the LL servers with bakes, so bad all around. As for the test harness, Yes let me tidy it slightly and I'll publish it, its really simple but i'm lazily working around one issue with error reporting currently. I also hopefully will have some new data tonight myself, looks like my results are artificially slow, BUT there remains a massive difference between KDU and openjpg, possibly worse than the x4 data i obtained, maybe in the order of x16-x20 slower but i need to rerun all the tests. This could also explain why SSE was showing odd results. Robin ___ Policies and (un)subscribe information available here: http://wiki.secondlife.com/wiki/OpenSource-Dev Please read the policies before posting to keep unmoderated posting privileges ___ Policies and (un)subscribe information available here: http://wiki.secondlife.com/wiki/OpenSource-Dev Please read the policies before posting to keep unmoderated posting privileges
Re: [opensource-dev] Openjpeg/KDU the cold hard metrics
On Wed, Sep 22, 2010 at 9:29 AM, Ann Otoole missannoto...@yahoo.com wrote: How many textures do you need? I bet I have a bunch that are perfectly OK for this purpose as well as for commercial use. And if you need more I can generate them using filters. I.e.; this is not a block. Thanks, I was running just under 300 for my tests, but we could do with a mix of RGB and RGBA textures + some that have lossless encoding and also some encoded on KDU and some on Openjpeg as there are differences. So any you want to donate to the cause would be most welcome, i can easily add a text file with a list of textures and copyright by who etc so its clear who actually owns them. Also i can encode if necessary but esp with the KDU ones i need to be 100% i've made them the same as the viewer does. We ideally want to get more that 300 and get a good test bed of say 1000 in total and then that will be very useful with a common set of textures for everyone to compare results with as many things in common as possible, this will then show differences on optimisations between different CPUs, operating systems etc and make it very easy for changes to openjpeg or any other decoder to be compared on mass. Regards Robin ___ Policies and (un)subscribe information available here: http://wiki.secondlife.com/wiki/OpenSource-Dev Please read the policies before posting to keep unmoderated posting privileges
Re: [opensource-dev] Openjpeg/KDU the cold hard metrics
On 22/09/2010 5:55 PM, Robin Cornelius wrote: On Wed, Sep 22, 2010 at 4:01 AM, Sheet Spottersheet.spot...@gmail.com wrote: I would love to see Robin's test harness. I would also love to see the images that failed with v2. For the record I cannot give you (or anyone else) the images that failed, I was testing against a random selection of images pulled from the texture cache (and yes i know how to get a complete image out the cache) It's not hard to do. I've got this little HTTP texture-proxy thing that queries the local computers here to share local viewer caches and save on bandwidth. And yes, the contents aren't the sort of thing you want to pass around willy-nilly. -- Tateru Nino http://dwellonit.taterunino.net/ ___ Policies and (un)subscribe information available here: http://wiki.secondlife.com/wiki/OpenSource-Dev Please read the policies before posting to keep unmoderated posting privileges
Re: [opensource-dev] Openjpeg/KDU the cold hard metrics
Doing this out of a local cache would make it clearer what the time for the library is. I suspect these numbers understate the kdu advantage. On Wed, Sep 22, 2010 at 7:30 AM, Tateru Nino tateru.n...@gmail.com wrote: On 22/09/2010 5:55 PM, Robin Cornelius wrote: On Wed, Sep 22, 2010 at 4:01 AM, Sheet Spottersheet.spot...@gmail.com wrote: I would love to see Robin's test harness. I would also love to see the images that failed with v2. For the record I cannot give you (or anyone else) the images that failed, I was testing against a random selection of images pulled from the texture cache (and yes i know how to get a complete image out the cache) It's not hard to do. I've got this little HTTP texture-proxy thing that queries the local computers here to share local viewer caches and save on bandwidth. And yes, the contents aren't the sort of thing you want to pass around willy-nilly. -- Tateru Nino http://dwellonit.taterunino.net/ ___ Policies and (un)subscribe information available here: http://wiki.secondlife.com/wiki/OpenSource-Dev Please read the policies before posting to keep unmoderated posting privileges ___ Policies and (un)subscribe information available here: http://wiki.secondlife.com/wiki/OpenSource-Dev Please read the policies before posting to keep unmoderated posting privileges
Re: [opensource-dev] Openjpeg/KDU the cold hard metrics
In theory, KDU 6.4 is supposed to be platform neutral - or at least that's one of the claims the mfr. is making about the new version. On Sep 22, 2010, at 11:35 AM, Arrehn Oberlander wrote: Is it possible for you to release your test harness for these comparisons, including the texture set? I noticed there isn't data for linux or mac platforms, which I am curious about. My eyeball tells me KDU is more optimized for windows than other platforms, but I'd like to test this empirically, as well as see how well my hardware compares to these results using the same test. ___ Policies and (un)subscribe information available here: http://wiki.secondlife.com/wiki/OpenSource-Dev Please read the policies before posting to keep unmoderated posting privileges ___ Policies and (un)subscribe information available here: http://wiki.secondlife.com/wiki/OpenSource-Dev Please read the policies before posting to keep unmoderated posting privileges
Re: [opensource-dev] Openjpeg/KDU the cold hard metrics
On 09/22/2010 02:12 PM, Lillian Yiyuan wrote: Doing this out of a local cache would make it clearer what the time for the library is. I suspect these numbers understate the kdu advantage. I'm pretty sure Robin's test harness is very minimal, if you are thinking that there's a bloated client attached to it, muddying the numbers. ___ Policies and (un)subscribe information available here: http://wiki.secondlife.com/wiki/OpenSource-Dev Please read the policies before posting to keep unmoderated posting privileges
Re: [opensource-dev] Openjpeg/KDU the cold hard metrics
There may be another option for obtaining a common set of J2C files for comparison. Viewer 2 installs 784 JPEG-2000 files (*.j2c) in the local_assets folder under the main install folder (e.g., C:\Program Files\SecondLifeViewer2). The j2k_metric test harness decoded all 784 files from the local_assets folder without errors. The OpenJPEG 1.3.0 library was 3.7 times slower than the KDU library that comes with Viewer 2.1.1.208043 (the current release version). I have been unsuccessful at implementing the additional functions to make the OpenJPEG v2 interface compatible with the earlier v1.3.0 interface. The test application aborted with buffer overrun errors. Oopsie! The tests were run under Windoze 7 on a Pentium i7-950 (3 GHz Hyperthreaded) using a RelWithDebInfo build run from the command line. (Running the same build from within VisualStudio 2005 was noticeably slower. There was one small surprise. Only two threads were in use for both the OpenJPEG v1.3.0 and KDU tests. I expected KDU to use all eight available threads. Sheet Spotter -Original Message- From: opensource-dev-boun...@lists.secondlife.com [mailto:opensource-dev-boun...@lists.secondlife.com] On Behalf Of Robin Cornelius Sent: September 22, 2010 2:00 PM To: Arrehn Oberlander Cc: opensource-dev Subject: Re: [opensource-dev] Openjpeg/KDU the cold hard metrics On Wed, Sep 22, 2010 at 7:35 PM, Arrehn Oberlander arr...@gmail.com wrote: Is it possible for you to release your test harness for these comparisons, including the texture set? I can't give out the texture set, hopefully as per a previous discussion on this topic, we can assemble a good texture set that we can all share and compare metrics. the code can be found at http://bitbucket.org/robincornelius/viewer-development-j2c-metrics The code although based on 2.X should still work on 1.X (my original tests were a 1.X build) but i wanted to clean this up a bit before release and to be honest we are concerned with openjpeg/kdu not the viewer for these tests, my test code itsself is painfully simple. Load a texture into an LLImage, run decode time it. repeat and rince. The code forms a seperate project j2k_metric which depends on llcommon/llimage/llimagej2coj and llvfs (cmake rules included to make it build out of the box. But to make it run you need to set the working directory to be the sharedlibs folder (or copy dlls/so's around) as i've not bothered to stage it yet (patches welcome) The exe accepts one command line argument and that is the path to a folder of textures, if no path is given it assumes a default path of C:\Textures to make things a tiny bit easier on windows. Hope this is enough to get people started Robin ___ Policies and (un)subscribe information available here: http://wiki.secondlife.com/wiki/OpenSource-Dev Please read the policies before posting to keep unmoderated posting privileges ___ Policies and (un)subscribe information available here: http://wiki.secondlife.com/wiki/OpenSource-Dev Please read the policies before posting to keep unmoderated posting privileges
Re: [opensource-dev] Openjpeg/KDU the cold hard metrics
There was one small surprise. Only two threads were in use for both the OpenJPEG v1.3.0 and KDU tests. I expected KDU to use all eight available threads. Remember that LL is still shipping a downright ancient KDU 4.* based llkdu.dll with the viewer. If they successfully update to 6.4.* or later, you'll probably see all cores used. ___ Policies and (un)subscribe information available here: http://wiki.secondlife.com/wiki/OpenSource-Dev Please read the policies before posting to keep unmoderated posting privileges
[opensource-dev] Openjpeg/KDU the cold hard metrics
Hi everyone I have been doing some direct comparison between KDU and Openjpeg and have some metrics to share with the rest of you (in case you have not seen them already flying around on #opensl/ or AWG chat) https://spreadsheets.google.com/ccc?key=0AiSrUP47_VxIdFY4Mi1VUkdnaXJsSVpldGRPZXoxc3chl=en#gid=0 The tests were done in a test harness using the LLImage class from the viewer. 296 Typical SL textures which were complete were used, Each texture was decoded to discard level 0 (fully) and the timing was only done for the call to decode(). This was repeated 10 times to average out some CPU load jitter and the results entered on to the spreadsheet. I did some variations on openjpeg 1.3 with builds on VC2005 and 2008 and some different SSE and optimisation settings, I also did a test with the Openjpeg V2 code from SVN, although I had to remove 5 textures from the test to prevent this from crashing. The SSE results are interesting, my CPU is an AMD Athlon II X4 64 bit quad core but with 2005 and 2008 SSE2 slows down the decode process,and 2008 also does a better optimisation job than 2005. I'm still not 100% sure why the build of openjpeg supplied by imprudence is better than my builds on 2005. I've also not done any tests on linux, but the basic pattern of KDU being way faster than openjpeg 1.X and 2.X being faster than 1.X i do not expect to change much, the only thing that might change is ups and downs for various optimisations and extra CPU instructions such as SSE behaving differently under gcc. Openjpeg 2 is really looking like its made a massive improvement in its speed, but currently its very unstable, as I said it crashes with 5 of my test textures (which appear to be perfectly normal RGBA textures of common dimensions). Its also crashing with encode for bakes and as far as i know it has issues/is not complete to handle truncated streams either, all of which make usage in SL right this moment not possible. I think openjpeg 2 is where opensource effort should be directed as its not _that_ far behind KDU for raw decode speed (compared to OJP 1.X) currently and yes I know the new KDU can use multiple cores, but so what, we can run multiple decode threads, its nots if we are decoding a GB image so that extra boost is probably not that significant for the type of SL usage. So this brings us to the state of Openjpeg 2.0, Now i know Dzontaz and Callum have done quite a bit with Openjpeg of various versions in the past, if either of them are watching do they have any insites to the current state of OJP 2.0 for SL usage (or does anyone else), as very little has happened in Openjpeg svn for a long while now I fear if we want this to work we may have to fix it and supply the patches upstream ourselves. I believe at one point Dzontas was using OJP 2 for download but OJP 1 for upload when testing it. Regards Robin ___ Policies and (un)subscribe information available here: http://wiki.secondlife.com/wiki/OpenSource-Dev Please read the policies before posting to keep unmoderated posting privileges
Re: [opensource-dev] Openjpeg/KDU the cold hard metrics
On 9/21/10, Robin Cornelius robin.cornel...@gmail.com wrote: I'm still not 100% sure why the build of openjpeg supplied by imprudence is better than my builds on 2005. Imprudence is using the openjpeg SVN trunk version from a few months ago, which is essentially a pre-release version of openjpeg 1.4 and is slightly more optimised than 1.3. You might want to try openjpeg from SVN yourself (the trunk version, not the 2.0 branch). Make sure to try it with SSE enabled too - it has some SSE-specific optimisations. ___ Policies and (un)subscribe information available here: http://wiki.secondlife.com/wiki/OpenSource-Dev Please read the policies before posting to keep unmoderated posting privileges
Re: [opensource-dev] Openjpeg/KDU the cold hard metrics
On Tue, Sep 21, 2010 at 7:44 PM, Aidan Thornton makos...@gmail.com wrote: On 9/21/10, Robin Cornelius robin.cornel...@gmail.com wrote: I'm still not 100% sure why the build of openjpeg supplied by imprudence is better than my builds on 2005. Imprudence is using the openjpeg SVN trunk version from a few months ago, which is essentially a pre-release version of openjpeg 1.4 and is slightly more optimised than 1.3. I built off revision 635 from trunk and tried with no SSE/SSE1 and SSE2 as per my results, SSE1 on 2008 gives the best effect, SSE2 is a disaster all around on my setup. Robin ___ Policies and (un)subscribe information available here: http://wiki.secondlife.com/wiki/OpenSource-Dev Please read the policies before posting to keep unmoderated posting privileges
Re: [opensource-dev] Openjpeg/KDU the cold hard metrics
Could you publish the test harness, I would like to try this on the mac with a mactools build of v2 On Tue, Sep 21, 2010 at 2:49 PM, Robin Cornelius robin.cornel...@gmail.com wrote: On Tue, Sep 21, 2010 at 7:44 PM, Aidan Thornton makos...@gmail.com wrote: On 9/21/10, Robin Cornelius robin.cornel...@gmail.com wrote: I'm still not 100% sure why the build of openjpeg supplied by imprudence is better than my builds on 2005. Imprudence is using the openjpeg SVN trunk version from a few months ago, which is essentially a pre-release version of openjpeg 1.4 and is slightly more optimised than 1.3. I built off revision 635 from trunk and tried with no SSE/SSE1 and SSE2 as per my results, SSE1 on 2008 gives the best effect, SSE2 is a disaster all around on my setup. Robin ___ Policies and (un)subscribe information available here: http://wiki.secondlife.com/wiki/OpenSource-Dev Please read the policies before posting to keep unmoderated posting privileges ___ Policies and (un)subscribe information available here: http://wiki.secondlife.com/wiki/OpenSource-Dev Please read the policies before posting to keep unmoderated posting privileges
Re: [opensource-dev] Openjpeg/KDU the cold hard metrics
I would love to see Robin's test harness. I would also love to see the images that failed with v2. I started diving into the v2 branch of OpenJPEG a few days ago, after research JPEG 2000 for a week. Tackling the bugs from the failed images might be a good way to become more familiar with the source code. I posted a very minor patch correcting some simple build problems with the DLL version of OpenJPEG V2. http://code.google.com/p/openjpeg/issues/detail?id=40 Does anyone have a sense of how actively the OpenJPEG source code is maintained? Thank you for publishing your results, Robin. They are encouraging me to continue looking at OpenJPEG v2. Sheet Spotter -Original Message- From: opensource-dev-boun...@lists.secondlife.com [mailto:opensource-dev-boun...@lists.secondlife.com] On Behalf Of Lillian Yiyuan Sent: September 21, 2010 4:24 PM Cc: opensource-dev Subject: Re: [opensource-dev] Openjpeg/KDU the cold hard metrics Could you publish the test harness, I would like to try this on the mac with a mactools build of v2 On Tue, Sep 21, 2010 at 2:49 PM, Robin Cornelius robin.cornel...@gmail.com wrote: On Tue, Sep 21, 2010 at 7:44 PM, Aidan Thornton makos...@gmail.com wrote: On 9/21/10, Robin Cornelius robin.cornel...@gmail.com wrote: I'm still not 100% sure why the build of openjpeg supplied by imprudence is better than my builds on 2005. Imprudence is using the openjpeg SVN trunk version from a few months ago, which is essentially a pre-release version of openjpeg 1.4 and is slightly more optimised than 1.3. I built off revision 635 from trunk and tried with no SSE/SSE1 and SSE2 as per my results, SSE1 on 2008 gives the best effect, SSE2 is a disaster all around on my setup. Robin ___ Policies and (un)subscribe information available here: http://wiki.secondlife.com/wiki/OpenSource-Dev Please read the policies before posting to keep unmoderated posting privileges ___ Policies and (un)subscribe information available here: http://wiki.secondlife.com/wiki/OpenSource-Dev Please read the policies before posting to keep unmoderated posting privileges ___ Policies and (un)subscribe information available here: http://wiki.secondlife.com/wiki/OpenSource-Dev Please read the policies before posting to keep unmoderated posting privileges