MAGMA Forum

yes this patch is to be applied to strsm too. And the memset for d_x is also going to be in future RC or release. Thanks a lot for the patience and reminder on d_x.

Folks Sorry for the long wait. For the problem caused by dtrsm, here is a quick patch, and please let me know if this helps: du:disco ~/disco-home/sandbox/clean-svn/run/magmablas> diff dtrsm_tesla.cu ../../web/magma_1.0.0-rc3/magmablas/dtrsm_tesla.cu 1790d1789 < if (i*2>=M) break; 1825d1823 < if (i*...

great! these should be good tests but if you see any other, feel free to let us know. thanks.

here is a quick patch. please let me know if this helps: du:disco ~/disco-home/sandbox/clean-svn/web/magma_1.0.0-rc3/magmablas> diff dtrsm_tesla.cu ../../../run/magmablas/dtrsm_tesla.cu 1823a1824 > if (i*2>=M) break; This line is added at the end of the 'for (int i=BLOCK_SIZE; i<NB; i*=2)' loop

I planted a printf in dtrsm so that when magma_dtrsm is called, we know it for sure. This is what I'm seeing so far: du:disco ~/disco-home/sandbox/bugfix> ./nan <- 11:08AM [ [ 0.391137 , 0.37845 , 0.676826 , 0.828647 ]; [ 0.572013 , 1.29715 , 1.72296 , 3.94056 ]; [ 0.382288 , 0.33823 , 1.39019 , 2.5...

thanks, we'll get back to you asap.

Hello Katayama,

Could you please give some more details about your running environment? For example, version of CUDA, nvcc, driver, which GPU. Also the exact size of g that causes NaN or wrong result. This so that we could repeat the issue on our side and pinpoint the bug.

thanks!

Hi all

We are trying to locate the bug.

First off, could you change NB to different value (a multiple of 16) and see if you still get NaN?

Thanks.

Hi all

We are trying to locate the bug.

First off, could you change NB to different value (a multiple of 16) and see if you still get NaN?

Thanks.