From jules at codesourcery.com Thu Nov 1 15:52:03 2007 From: jules at codesourcery.com (Jules Bergmann) Date: Thu, 01 Nov 2007 11:52:03 -0400 Subject: [patch] Fix Matlab IO Wall warnings Message-ID: <4729F623.6070906@codesourcery.com> Patch applied. -- Jules Bergmann CodeSourcery jules at codesourcery.com (650) 331-3385 x705 -------------- next part -------------- An embedded and charset-unspecified text was scrubbed... Name: wal.diff URL: From jules at codesourcery.com Thu Nov 1 15:58:10 2007 From: jules at codesourcery.com (Jules Bergmann) Date: Thu, 01 Nov 2007 11:58:10 -0400 Subject: [patch] Use FFTW from trunk Message-ID: <4729F792.8020603@codesourcery.com> Update to use 'fftw/trunk' (instead of 'fftw/vendor/3.1.2') for GPL snapshots. 'fftw/trunk' is 3.1.2, plus our GreenHills/MCOE changes from 3.0. Patch applied. -- Jules -- Jules Bergmann CodeSourcery jules at codesourcery.com (650) 331-3385 x705 From jules at codesourcery.com Thu Nov 1 15:58:45 2007 From: jules at codesourcery.com (Jules Bergmann) Date: Thu, 01 Nov 2007 11:58:45 -0400 Subject: [vsipl++] [patch] Use FFTW from trunk In-Reply-To: <4729F792.8020603@codesourcery.com> References: <4729F792.8020603@codesourcery.com> Message-ID: <4729F7B5.4070601@codesourcery.com> Oops. Patch attached. Jules Bergmann wrote: > Update to use 'fftw/trunk' (instead of 'fftw/vendor/3.1.2') for GPL > snapshots. 'fftw/trunk' is 3.1.2, plus our GreenHills/MCOE changes from > 3.0. > > Patch applied. > > -- Jules > -- Jules Bergmann CodeSourcery jules at codesourcery.com (650) 331-3385 x705 -------------- next part -------------- An embedded and charset-unspecified text was scrubbed... Name: cfg.diff URL: From jules at codesourcery.com Mon Nov 5 14:07:27 2007 From: jules at codesourcery.com (Jules Bergmann) Date: Mon, 05 Nov 2007 09:07:27 -0500 Subject: [patch] Separable filter Message-ID: <472F239F.7020804@codesourcery.com> This patch implements a subset of separable filter functionality, along with a unit test and a benchmark. Overall, this follows our image processing API draft: https://intranet.codesourcery.com/VSIPLImageProcessing This is in the vsip_csl::img namespace. I also added a new support type: support_min_zeropad. It computes the same support as support_min, however it produces an image the same size as the input, zero-filling outputs not computed (i.e. zero-padding the image border). This support type matches how the registration code uses its separable filter. It also matches with IPP's separable filter implementation. Such a support type would also let us extend Convolution to use SAL's f3x3 and f5x5 functions, which essentially compute a min-support result with zero-padding around the edges. On verona, performance is poor with the generic BE (260 MF/s at 1024x1024 image with 7x7 filter) and very good with the IPP BE (8413 MF/s). Ok to apply? -- Jules -- Jules Bergmann CodeSourcery jules at codesourcery.com (650) 331-3385 x705 -------------- next part -------------- An embedded and charset-unspecified text was scrubbed... Name: sfilt.diff URL: From mark at codesourcery.com Fri Nov 9 23:54:41 2007 From: mark at codesourcery.com (Mark Mitchell) Date: Fri, 09 Nov 2007 15:54:41 -0800 Subject: Live Webcast: Cell Broadband Engine by IBM and CodeSourcery Message-ID: <4734F341.2090401@codesourcery.com> Attached is an announcement about a joint presentation by IBM and CodeSourcery regarding the Cell Broadband Engine and some of the signal- and image-processing library development that CodeSourcery is doing for the Cell/B.E. This presentation will take place on November 15th, at 2PM EST/11AM PST. Max Aguilar, an IBM Master Inventor, will be explaining some of the key aspects of the Cell/B.E. hardware platform. I will talk about Sourcery VSIPL++ for the Cell/B.E., as well as two lower-level programming interfaces that will also be useful to developers of high-performance signal and image-processing programs. If you want to sign up for the e-Cast, visit: http://w.on24.com/r.htm?e=96733&s=1&k=66A24D760DE02DE3957621DCA9569C18&partnerref=code I hope you can join us! -- Mark Mitchell CodeSourcery mark at codesourcery.com (650) 331-3385 x713 -------------- next part -------------- An embedded message was scrubbed... From: "Embedded Ecast" Subject: Live Webcast: Cell Broadband Engine by IBM and CodeSourcery Date: Fri, 9 Nov 2007 15:32:09 -0500 Size: 16273 URL: From howes at ll.mit.edu Tue Nov 27 19:12:27 2007 From: howes at ll.mit.edu (Brad Howes) Date: Tue, 27 Nov 2007 14:12:27 -0500 Subject: [1.3] Mem Leak in vsip::llsqsol? Message-ID: Hi, I'm got what appears to be a memory problem with vsip::llsqsol ( Matrix, Matrix ). If I comment out my call, then the memory remains constant. Any known problems with 1.3 and this routine? Any hints on how to work around it? I've got another solution using vsip::svd, but it has its own problems right now. Thanks, Brad -- Brad Howes Group 42 MIT Lincoln Laboratory ? 244 Wood St. ? Lexington, MA 02173 Phone: 781.981.5292 ? Fax: 781.981.3495 ? Secretary: 781.981.7420 -------------- next part -------------- An HTML attachment was scrubbed... URL: From jules at codesourcery.com Tue Nov 27 20:14:39 2007 From: jules at codesourcery.com (Jules Bergmann) Date: Tue, 27 Nov 2007 15:14:39 -0500 Subject: [vsipl++] [1.3] Mem Leak in vsip::llsqsol? In-Reply-To: References: Message-ID: <474C7AAF.2070603@codesourcery.com> Brad Howes wrote: > Hi, > > I'm got what appears to be a memory problem with vsip::llsqsol( Matrix, > Matrix ). If I comment out my call, then the memory remains constant. > Any known problems with 1.3 and this routine? Any hints on how to work > around it? I've got another solution using vsip::svd, but it has its own > problems right now. > > Thanks, Brad, Unfortunately we weren't aware of this memory leak (and hence don't have an immediate work around to suggest). However, I'm sure we can either fix it or find a work around. A couple of questions: What version of VSIPL++ are you using? What backend library are you using for linear algebra? I.e. ATLAS, MKL, SAL, C-VSIPL? Also, what problems are you encountering with SVD? thanks, -- Jules -- Jules Bergmann CodeSourcery jules at codesourcery.com (650) 331-3385 x705 From don at codesourcery.com Fri Nov 30 00:08:52 2007 From: don at codesourcery.com (Don McCoy) Date: Thu, 29 Nov 2007 17:08:52 -0700 Subject: [vsipl++] [1.3] Mem Leak in vsip::llsqsol? In-Reply-To: References: Message-ID: <474F5494.1000805@codesourcery.com> Brad, I have a fix for your problem! Since you are building from source, it should be easy to apply the fix using the command patch -p0 < temp_buffer.diff issued from the top-level directory (the one containing 'src'), or by editing the file in question manually. Please let me know if this works for you and thanks again for letting us know about the problem! Team, I tested this patch against Sourcery VSIPL++ v1.3 under x86 Linux, but it should be applicable to all build variations and all branches. Ok to commit? -- Don McCoy don (at) CodeSourcery (888) 776-0262 / (650) 331-3385, x712 -------------- next part -------------- An embedded and charset-unspecified text was scrubbed... Name: temp_buffer.changes URL: -------------- next part -------------- An embedded and charset-unspecified text was scrubbed... Name: temp_buffer.diff URL: From jules at codesourcery.com Fri Nov 30 13:19:07 2007 From: jules at codesourcery.com (Jules Bergmann) Date: Fri, 30 Nov 2007 08:19:07 -0500 Subject: [vsipl++] [1.3] Mem Leak in vsip::llsqsol? In-Reply-To: <474F5494.1000805@codesourcery.com> References: <474F5494.1000805@codesourcery.com> Message-ID: <47500DCB.9090504@codesourcery.com> Don McCoy wrote: > Brad, > > I have a fix for your problem! Since you are building from source, it > should be easy to apply the fix using the command > > patch -p0 < temp_buffer.diff > > issued from the top-level directory (the one containing 'src'), or by > editing the file in question manually. > > Please let me know if this works for you and thanks again for letting us > know about the problem! > > > Team, > > I tested this patch against Sourcery VSIPL++ v1.3 under x86 Linux, but > it should be applicable to all build variations and all branches. Ok to > commit? > > Don, Thanks for tracking this down! The patch looks good. Please commit. thanks, -- Jules -- Jules Bergmann CodeSourcery jules at codesourcery.com (650) 331-3385 x705