From jules at codesourcery.com Mon Jul 2 20:21:30 2007 From: jules at codesourcery.com (Jules Bergmann) Date: Mon, 02 Jul 2007 16:21:30 -0400 Subject: [patch] New configure options for use w/eclipse, misc fixes Message-ID: <46895E4A.9010207@codesourcery.com> This patch ... ... adds a new configure option "--disable-shared-acconfig" that moves defines currently on the compiler the command line (to enable shared acconfig.hpp across multiple variants in a binary package) into acconfig.hpp. This is useful when building a binary package to be used from eclipse, since there is less stuff to type into eclipse's phalanx of dialog boxes. ... adds a new configure option "--disable-fftw3-simd" that disables the builtin FFTW from using SIMD. Useful when cross-building for a Power platform (such as powerpc-eabi) that doesn't have altivec. ... fixes recursive builtin lapack make's to use $(MAKE). Fixes bug reported w/Solaris AMD64. ... removes missing define from expr/generator_block.hpp. Fixes vramp, I think. 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: misc.diff URL: From jules at codesourcery.com Thu Jul 5 16:39:57 2007 From: jules at codesourcery.com (Jules Bergmann) Date: Thu, 05 Jul 2007 12:39:57 -0400 Subject: [vsipl++] reductions-idx In-Reply-To: <4680BD48.6000004@codesourcery.com> References: <4680BD48.6000004@codesourcery.com> Message-ID: <468D1EDD.2060905@codesourcery.com> Assem Salama wrote: > Everyone, > This patch fixes a failure in reduction-idx test. Assem, This looks good, please check it in. -- Jules -- Jules Bergmann CodeSourcery jules at codesourcery.com (650) 331-3385 x705 From don at codesourcery.com Mon Jul 23 19:15:50 2007 From: don at codesourcery.com (Don McCoy) Date: Mon, 23 Jul 2007 13:15:50 -0600 Subject: [patch] External data access example Message-ID: <46A4FE66.8040100@codesourcery.com> This patch adds an example where the values of a VSIPL++ Matrix are altered using a direct pointer to the underlying data. This is useful for performing operations on data outside the VSIPL++ library. Regards, -- Don McCoy don (at) CodeSourcery (888) 776-0262 / (650) 331-3385, x712 -------------- next part -------------- An embedded and charset-unspecified text was scrubbed... Name: extdata.changes URL: -------------- next part -------------- An embedded and charset-unspecified text was scrubbed... Name: extdata.diff URL: From jules at codesourcery.com Tue Jul 24 12:17:25 2007 From: jules at codesourcery.com (Jules Bergmann) Date: Tue, 24 Jul 2007 08:17:25 -0400 Subject: [vsipl++] [patch] External data access example In-Reply-To: <46A4FE66.8040100@codesourcery.com> References: <46A4FE66.8040100@codesourcery.com> Message-ID: <46A5EDD5.7040209@codesourcery.com> Don McCoy wrote: > This patch adds an example where the values of a VSIPL++ Matrix are > altered using a direct pointer to the underlying data. This is useful > for performing operations on data outside the VSIPL++ library. Don, This patch looks good, please check it in. thanks, -- Jules -- Jules Bergmann CodeSourcery jules at codesourcery.com (650) 331-3385 x705 From Stefan.Joe-Yen at EssexCorp.com Wed Jul 25 17:06:36 2007 From: Stefan.Joe-Yen at EssexCorp.com (Joe-Yen, Stefan) Date: Wed, 25 Jul 2007 13:06:36 -0400 Subject: Configuration and compile problems with ICL Message-ID: <5CD1C9B961A59D4592F02D12CD93238B0813B1F9@STITCH.essexcorp.com> Hi, I am trying to build VSIPL++ with the Intel compiler and Cygwin on an Intel EMT64 System and running into a few problems. I have installed trial versions of the Intel Compiler (10.0.026), IPP library (5.2.057) and MKL library (9.1.025). We are using Visual C++ 2005 Express edition with the Platform SDK (Server 2003 R2) installed. I tried to run the configure script using the Cygwin shell. I started with the following configuration line from the quickstart guide: ./configure CXX=icl CC=icl CXXFLAGS="/Qcxx-features /Qvc8" --build=i686-cygwin --host=i686-mingw32 --enable-ipp=win --enable-fft=ipp --with-lapack=mkl_win --disable-mpi This initially failed when checking the compiler. I made sure the include library paths were available to the Cygwin shell and then the compiler check completed but configure failed on the IPP step as follows. ... Using IPP for Windows. checking for ipps.h... yes checking for ippsMul_32f... configure: error: not found. Trying without IPP, ./configure CXX=icl CC=icl CXXFLAGS="/Qcxx-features /Qvc8" --build=i686-cygwin --host=i686-mingw32 --with-lapack=mkl_win --disable-mpi then failed on the MKL step as follows: ... with_lapack: mkl_win Searching for LAPACK packages: mkl_win checking for LAPACK/MKL 8.x library for windown... not found configure: error: LAPACK enabled but no library found This looks like a version issue is there a workaround? Finally with both IPP and MKL removed, ./configure CXX=icl CC=icl CXXFLAGS="/Qcxx-features /Qvc8" --build=i686-cygwin --host=i686-mingw32 --disable-mpi I get this failure: ... checking for asm style... configure: error: cannot determine asm type. ============================================================= configure: error: built-in ATLAS configure FAILED. This seems to be addressed in a prior post so I'm not to concerned about it if it has been patched. Please let me know if you have any insights on these problems. I suspect it is a preferred installation configuration or a tool version issue but I can't find indications that the tools I am using are incorrect or detailed guidance on preferred tool configuration and setup in the documentation. Thanks. - Stefan Stefan Joe-Yen Cognitive Research Engineer Essex 321-837-7018 This electronic message and any files transmitted with it contain information which may be privileged and/or proprietary. The information is intended for use solely by the intended recipient(s). If you are not the intended recipient, be aware that any disclosure, copying, distribution or use of this information is prohibited. If you have received this electronic message in error, please advise the sender by reply email or by telephone (301-939-7000) and delete the message. -------------- next part -------------- An HTML attachment was scrubbed... URL: From jules at codesourcery.com Thu Jul 26 16:50:37 2007 From: jules at codesourcery.com (Jules Bergmann) Date: Thu, 26 Jul 2007 12:50:37 -0400 Subject: [patch] Fix FFTW3 split-complex Message-ID: <46A8D0DD.7080700@codesourcery.com> This fixes the FFTW3 split-complex breakage. 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: fftw.diff URL: From jules at codesourcery.com Fri Jul 27 12:07:02 2007 From: jules at codesourcery.com (Jules Bergmann) Date: Fri, 27 Jul 2007 08:07:02 -0400 Subject: [patch] Fix distributed maxval bug, split large tests Message-ID: <46A9DFE6.40203@codesourcery.com> This patch fixes a bug causing distributed maxval not to work with matrices and tensors. It also splits apart several large tests (coverage_unary and fft) into smaller tests. Patch applied. -- Jules Bergmann CodeSourcery jules at codesourcery.com (650) 331-3385 x705 -------------- next part -------------- An embedded and charset-unspecified text was scrubbed... Name: maxval.diff URL: From jules at codesourcery.com Mon Jul 30 12:54:45 2007 From: jules at codesourcery.com (Jules Bergmann) Date: Mon, 30 Jul 2007 08:54:45 -0400 Subject: [patch] FFTW split complex fixes; SIMD unaligned fixes Message-ID: <46ADDF95.2080602@codesourcery.com> This patch fixes and works around several bugs in split-complex FFT and unaligned SIMD expressions. Split-complex FFT support using FFTW was failing numerous tests. This was a combination of bugs and unimplemented functionality in VSIPL++, and bugs in FFTW. * The VSIPL++ split complex to complex transforms (both FFT and FFTM) did not handle inverse transforms at all. * Split real-to-complex and complex-to-real transforms for FFTM were unimplemented. * FFTW (3.1.2 at least) plans capture the gap between real and imaginary components for in-place complex FFTs, and for real-complex and complex-real FFTs. If the plan is executed with data with a different gap, all heck breaks loose. Work around is to copy data into buffers with same alignment as during planning. Unaligned SIMD (handling multiple unaligned vectors with the same alignment) was broken. Instead of incrementing past the unaligned section, it would decrement before. This caused it to skip the last aligned word of the expressions. Fix is to make increment_by_element explicit. Patch to be applied, pending test results. -- Jules -- Jules Bergmann CodeSourcery jules at codesourcery.com (650) 331-3385 x705 -------------- next part -------------- An embedded and charset-unspecified text was scrubbed... Name: fftw-simd.diff URL: From jules at codesourcery.com Tue Jul 31 13:11:27 2007 From: jules at codesourcery.com (Jules Bergmann) Date: Tue, 31 Jul 2007 09:11:27 -0400 Subject: [patch] Misc fixes Message-ID: <46AF34FF.8060902@codesourcery.com> This patch - disable FFTW split-complex altogether. Out-of-place complex split FFT appears broken on belgarath. - Fixes ifdef logic bug that prevented optimized convolution BEs from being used. - Adds a 2D conv benchmark 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: misc.diff URL: From jules at codesourcery.com Tue Jul 31 20:39:57 2007 From: jules at codesourcery.com (Jules Bergmann) Date: Tue, 31 Jul 2007 16:39:57 -0400 Subject: [patch] Misc bits for Mercury Message-ID: <46AF9E1D.8000902@codesourcery.com> This patch contains several fixes for Mercury/GHS. Patch applied. -- Jules Bergmann CodeSourcery jules at codesourcery.com (650) 331-3385 x705 -------------- next part -------------- An embedded and charset-unspecified text was scrubbed... Name: mc.diff URL: