Closed MarbolanGos closed 12 years ago
Hi,
Thank you for this patch. I will add this patch into develop branch. You can also fork this project. Then, you can click "Pull Request" button on github.com
Thank you again!~
Xianyi
Hi,
After applying this patch, this patch didn't work. Output: lapack-3.4.0.tgz check sum is wrong (Please use orignal). /bin/sh: cannot create lapack-3.4.0/make.inc: Directory nonexistent ....
Then, I changed md5 checking for lapack-3.4.0. However, it failed on the patching. Outputs: tar zxf lapack-3.4.0.tgz patching file INSTALL/Makefile Hunk #1 FAILED at 27. 1 out of 1 hunk FAILED -- saving rejects to file INSTALL/Makefile.rej patching file Makefile Hunk #1 FAILED at 20. ...
Thanks
Xianyi
You are right I changed other things. Sorry...
Thanks for updating lapack.
Hi,
You can send the patch to me (traits.zhang at gmail dot com). Alternatively, you can use "Pull Request" on github.com
Thanks
Xianyi
Hi, I just upgraded LAPACK to 3.4.0 version in develop branch. You can download it from https://github.com/xianyi/OpenBLAS/zipball/develop
Could you help me test LAPACK 3.4.0 ?
Thanks
Xianyi
Hi, Sorry for not responding I was busy those days. Thank you for taking care of it. I tried your archive on 3 different computers with intel based CPUs all compiled well and I had no errors on lapack tests. All on Linux stations (64 bits).
Hi, I compiled the LAPACK-3.4.0 with MinGW-w64 without error. It runs with my small test code. However, I haven't figure out how to install python with MinGW-w64 and cannot run the build-in lapack_testing.py yet.
Best, Upsha
Hi Upsha,
Thank you for the test.
Xianyi
Does one get a faster LAPACK by using this patch, rather than by compiling LAPACK directly and linking it to openblas?
Hi Viral,
I think Mr. Goto optimized some LAPACK functions such as LU factorization. Thus, it will be faster than original LAPACK.
Thanks
Xianyi
Ok, in that case I will try it out once you merge it into master and make the next release.
-viral
On Jan 1, 2012, at 7:32 PM, Xianyi Zhang wrote:
Hi Viral,
I think Mr. Goto optimized some LAPACK functions such as LU factorization. Thus, it will be faster than original LAPACK.
Thanks
Xianyi
Reply to this email directly or view it on GitHub: https://github.com/xianyi/OpenBLAS/issues/72#issuecomment-3324559
As lapack was updated I was thinking is it possible easily to use it. I made those changes to the Makefile:
--- Makefile.old 2011-11-23 13:36:44.000000000 +0100 +++ Makefile 2011-11-23 13:38:38.000000000 +0100 @@ -195,47 +195,47 @@ netlib :
else -netlib : lapack-3.1.1 patch.for_lapack-3.1.1 lapack-3.1.1/make.inc +netlib : lapack-3.4.0 patch.for_lapack-3.4.0 lapack-3.4.0/make.inc ifndef NOFORTRAN
-prof_lapack : lapack-3.1.1 lapack-3.1.1/make.inc
-lapack-3.1.1/make.inc : +lapack-3.4.0/make.inc : ifndef NOFORTRAN
-lapack-3.1.1 : lapack-3.1.1.tgz +lapack-3.4.0 : lapack-3.4.0.tgz ifndef NOFORTRAN
$(MD5SUM) lapack-3.1.1.tgz | $(AWK) '{print $$1}'
= 00b21551a899bcfbaa7b8443e1faeef9; then \$(MD5SUM) lapack-3.4.0.tgz | $(AWK) '{print $$1}'
= 00b21551a899bcfbaa7b8443e1faeef9; then \ echo $(TAR) zxf $< ;\-lapack-3.1.1.tgz : +lapack-3.4.0.tgz : ifndef NOFORTRAN
-wget http://www.netlib.org/lapack/lapack-3.4.0.tgz endif
large.tgz : @@ -248,21 +248,21 @@ -wget http://www.netlib.org/lapack/timing/timing.tgz endif
-lapack-timing : lapack-3.1.1 large.tgz timing.tgz +lapack-timing : lapack-3.4.0 large.tgz timing.tgz ifndef NOFORTRAN
make -C lapack-3.4.0/TIMING endif
lapack-test :
$(GREP) failed lapack-3.4.0/TESTING/*.out
dummy :
@@ -280,9 +280,9 @@
endif
Copied patch.for_lapack-3.1.1 into patch.for_lapack-3.4.0 and all compiled correctly in my: OpenBLAS build complete.
OS ... Linux
Architecture ... x86_64
BINARY ... 64bit
C compiler ... GCC (command line : gcc) Fortran compiler ... GFORTRAN (command line : gfortran) Library Name ... libopenblas_penrynp-r0.1alpha2.4.a (Multi threaded; Max num-threads is 8)