-
Notifications
You must be signed in to change notification settings - Fork 1.5k
New issue
Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.
By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.
Already on GitHub? Sign in to your account
Please support AMD Bulldozer #118
Comments
Hi @BagOfMostlyWater , We didn't support AMD Bulldozer now. Because we didn't have the Bulldozer box, we didn't have the plan about optimization on Bulldozer. Try this.
This will build the library about AMD Shanghai. You can also try AMD ACML library which is optimized for Bulldozer. Thanks Xianyi |
On Sun, Jun 24, 2012 at 09:06:24PM -0700, Xianyi Zhang wrote:
Thank you, this enables the build to complete. Ryan |
Hi Ryan, I added Bulldozer detection into develop branch. You can download from this link. https://github.com/xianyi/OpenBLAS/tarball/develop Please test it on your AMD Opteron 62xx. Thanks Xianyi |
On Mon, Jun 25, 2012 at 02:33:30AM -0700, Xianyi Zhang wrote:
Hi Xianyi, Thanks for doing that. There were some core dumps during the make within Test of subprogram number 5 SCOPY Test of subprogram number 6 SSWAP It seems that all the tests with OPENBLAS_NUM_THREADS=2 are failing. gdb of sblat1: gdb of sblat2: gdb of sblat3: Just let me know if you'd like me to perform other steps and I'll send you Ryan |
Hi Ryan, Could you build the debug version?
Then, you can debug sblat1. Please try serial version (export OMP_NUM_THREADS=1 ) at first Thank you Xianyi |
On Mon, Jun 25, 2012 at 04:29:22PM -0700, Xianyi Zhang wrote:
Here are some results: $ OMP_NUM_THREADS=1 OPENBLAS_NUM_THREADS=1 gdb ./sblat1 Test of subprogram number 1 SDOT Test of subprogram number 2 SAXPY Test of subprogram number 3 SROTG Test of subprogram number 4 SROT Test of subprogram number 5 SCOPY Test of subprogram number 6 SSWAP Test of subprogram number 7 SNRM2 Test of subprogram number 8 SASUM Test of subprogram number 9 SSCAL Test of subprogram number 10 ISAMAX $ OMP_NUM_THREADS=2 OPENBLAS_NUM_THREADS=1 gdb ./sblat1 Test of subprogram number 1 SDOT Test of subprogram number 2 SAXPY Test of subprogram number 3 SROTG Test of subprogram number 4 SROT Test of subprogram number 5 SCOPY Test of subprogram number 6 SSWAP Test of subprogram number 7 SNRM2 Test of subprogram number 8 SASUM Test of subprogram number 9 SSCAL Test of subprogram number 10 ISAMAX $ OMP_NUM_THREADS=1 OPENBLAS_NUM_THREADS=2 gdb ./sblat1 Test of subprogram number 1 SDOT Test of subprogram number 2 SAXPY Test of subprogram number 3 SROTG Test of subprogram number 4 SROT Test of subprogram number 5 SCOPY Test of subprogram number 6 SSWAP Program received signal SIGSEGV, Segmentation fault. ubp_av=0x7fffffffe5c8, init=, fini=, rtld_fini=, stack_end=0x7fffffffe5b8) at libc-start.c:226 #7 0x0000000000401c29 in _start () $ OMP_NUM_THREADS=2 OPENBLAS_NUM_THREADS=2 gdb ./sblat1 Test of subprogram number 1 SDOT Test of subprogram number 2 SAXPY Test of subprogram number 3 SROTG Test of subprogram number 4 SROT Test of subprogram number 5 SCOPY Test of subprogram number 6 SSWAP Program received signal SIGSEGV, Segmentation fault. ubp_av=0x7fffffffe5c8, init=, fini=, rtld_fini=, stack_end=0x7fffffffe5b8) at libc-start.c:226 #7 0x0000000000401c29 in _start () In the last session, blas_quick_divide_table was sized 38520 so y was too Ryan |
Hi Ryan, I think this may relate to the issue about detecting the numbers of cores or threads. Because I didn't have access to Bulldozer, I cannot investigate this bug now. Thanks Xianyi |
On Tue, Jun 26, 2012 at 11:00:01PM -0700, Xianyi Zhang wrote:
Okay, thank you for taking the time to investigate this far. Ryan |
Hi Ryan, I asked a friend from AMD China to test OpenBLAS develop branch on AMD Opteron 42xx and 62xx. He built the library successfully on RHEL 6.2. He only meet an issue about force using gfortran compiler between f77, g77 and gfortran. I think this issue is the same to issue #119 on Intel Sandy Bridge. Thanks Xianyi |
Hi Ryan, Please test the newest develop branch. I think I fixed this bug. Thank you Xianyi |
On Fri, Jun 29, 2012 at 12:55:59AM -0700, Xianyi Zhang wrote:
I was able to successfully build with make DEBUG=1 NO_LAPACK=1 NO_CBLAS=1 and it completed all tests. I was also able to successfully build custom Ryan |
Merge in PL/openblas from dev/k.zaytseva/LM-538 to dev-riscv
Merge in PL/openblas from dev/k.zaytseva/LM-538 to dev-riscv
Merge in PL/openblas from dev/k.zaytseva/LM-538 to dev-riscv
OpenBLAS fails to build for me on Opteron 62xx. (Ubuntu 12.04, openblas 0.1alpha2.2)
Without DYNAMIC_ARCH=1
gemm.c: In function 'sgemm_':
gemm.c:392:37: error: 'GEMM_DEFAULT_OFFSET_A' undeclared (first use in this function)
With DYNAMIC_ARCH=1:
In file included from gemm3m.c:57:0:
gemm3m_level3.c: In function 'cgemm3m_cn':
gemm3m_level3.c:329:25: error: 'SGEMM_DEFAULT_UNROLL_M' undeclared (first use in this function)
Also:
getarch_2nd.c: In function 'main':
getarch_2nd.c:12:35: error: 'SGEMM_DEFAULT_UNROLL_M' undeclared (first use in this function)
getarch_2nd.c:12:35: note: each undeclared identifier is reported only once for each function it appears in
getarch_2nd.c:13:35: error: 'SGEMM_DEFAULT_UNROLL_N' undeclared (first use in this function)
getarch_2nd.c:14:35: error: 'DGEMM_DEFAULT_UNROLL_M' undeclared (first use in this function)
getarch_2nd.c:15:35: error: 'DGEMM_DEFAULT_UNROLL_N' undeclared (first use in this function)
getarch_2nd.c:19:35: error: 'CGEMM_DEFAULT_UNROLL_M' undeclared (first use in this function)
getarch_2nd.c:20:35: error: 'CGEMM_DEFAULT_UNROLL_N' undeclared (first use in this function)
getarch_2nd.c:21:35: error: 'ZGEMM_DEFAULT_UNROLL_M' undeclared (first use in this function)
getarch_2nd.c:22:35: error: 'ZGEMM_DEFAULT_UNROLL_N' undeclared (first use in this function)
getarch_2nd.c:29:50: error: 'SGEMM_DEFAULT_Q' undeclared (first use in this function)
getarch_2nd.c:30:50: error: 'DGEMM_DEFAULT_Q' undeclared (first use in this function)
getarch_2nd.c:31:50: error: 'CGEMM_DEFAULT_Q' undeclared (first use in this function)
getarch_2nd.c:32:50: error: 'ZGEMM_DEFAULT_Q' undeclared (first use in this function)
make[1]: *** [getarch_2nd] Error 1
The text was updated successfully, but these errors were encountered: