Search found 33 matches

by evanlezar
Fri Sep 14, 2012 3:38 am
Forum: User discussion
Topic: Problem with testing_zgesv
Replies: 12
Views: 8524

Re: Problem with testing_zgesv

Thanks Stan, I also recall seeing that message in the release notes a while back, but since it is not mentioned in the CUDA 4.1 or 4.2 release notes (I have not checked 4.0), I assumed this was no longer an issue. Could it be that since the new CUBLAS interface passes the scaling values (alpha in th...
by evanlezar
Thu Sep 13, 2012 9:32 am
Forum: User discussion
Topic: 64-bit integer in MAGMA
Replies: 14
Views: 9010

Re: 64-bit integer in MAGMA

More from my investigation: When using the MAGMA_NGR_NB environment variable to ensure that the memory for dAT (which I have moved to a separate allocation) does not exceed 4GB, then the cuda-memcheck errors disappear. Could it be that somewhere 32bit pointers are being used internally? I am using t...
by evanlezar
Thu Sep 13, 2012 8:31 am
Forum: User discussion
Topic: Problem with testing_zgesv
Replies: 12
Views: 8524

Re: Problem with testing_zgesv

Just thought I would ping this thread and ask if there is anything special that we should be doing with the Intel compiler? I have been picking up similar messages, but specifically for VERY LARGE problems (see the 64bit Integer thread), and am also using the Intel compiler. Unfortunately simply swi...
by evanlezar
Wed Sep 12, 2012 11:57 am
Forum: User discussion
Topic: 64-bit integer in MAGMA
Replies: 14
Views: 9010

Re: 64-bit integer in MAGMA

Some more from my side: I have run the example I have here through cuda-memcheck and get a number of messages such as the following: ========= Invalid __global__ write of size 4 ========= at 0x00000d10 in ../../magmablas/stranspose-v2.cu:50:stranspose3_32 ========= by thread (0,5,0) in block (13,1,0...
by evanlezar
Tue Sep 11, 2012 5:07 am
Forum: User discussion
Topic: 64-bit integer in MAGMA
Replies: 14
Views: 9010

Re: 64-bit integer in MAGMA

Thanks Mark, After disabling the sgemm and strsm calls, I end up with a launch failure in magmablas_spermute_long3. I know that this is supposed to swap the columns of the transposed matrix, but the exact working is a little tricky to piece together. Is there a paper or something that describes the ...
by evanlezar
Fri Sep 07, 2012 10:03 am
Forum: User discussion
Topic: 64-bit integer in MAGMA
Replies: 14
Views: 9010

Re: 64-bit integer in MAGMA

Continuing the discussion here, although there are now some other issues: I am currently running a large random single precision real matrix problem -- M=71710 N=71710 LDA=71712 -- and the computation continues until at one point the program SEGFAULT in a call to cudaStreamCreate(): #0 0x00007fffea9...
by evanlezar
Fri Aug 03, 2012 11:14 am
Forum: User discussion
Topic: 64-bit integer in MAGMA
Replies: 14
Views: 9010

Re: 64-bit integer in MAGMA

Thanks Mark, As you say, problems should start occurring at around 50,000x50,000 matrix elements. With my modified code -- which I have sent to you by email -- I no longer get these crashes for a single precision run that exceeds this. Unfortunately, although the primary indexing problems seem to be...
by evanlezar
Mon Jul 16, 2012 11:16 am
Forum: User discussion
Topic: 64-bit integer in MAGMA
Replies: 14
Views: 9010

Re: 64-bit integer in MAGMA

After changing the typedef, I also get a number of warnings when calling CUBLAS functions. Obviously one is then passing 64bit integers to the int parameters in CUBLAS (such as ZGEMM). Although it is theoretically possible for me to link with the 8-byte MKL interface, we also support a number of oth...
by evanlezar
Mon Jul 16, 2012 2:35 am
Forum: User discussion
Topic: 64-bit integer in MAGMA
Replies: 14
Views: 9010

Re: 64-bit integer in MAGMA

Hi Stan, for 64bit integers, do you mean that I change the typedef of magma_int_t to a 64bit equivalent? I have not tried this yet.
by evanlezar
Fri Jul 13, 2012 10:18 am
Forum: User discussion
Topic: 64-bit integer in MAGMA
Replies: 14
Views: 9010

Re: 64-bit integer in MAGMA

Just a related note. There is a bigger issue than 64bit integers. I am trying to solve a pretty large problem (87612x87612) using the zgetrf on multiple GPUs. At the moment, MAGMA crashes with a segmentation violation when I = 24576 in zgetrf3_ooc. This is about the point where I*lda (lda=87616) ove...