I've got strange error with magmablas_dtrsm.
I solve triangular linear system with right sise Y(nx2).
magma_dgetrs_gpu get me
Invalid __global__ read of size 8
========= at 0x00000150 in diag_dtrtri_kernel_lower
========= by thread (12,0,0) in block (0,0,0)
========= Address 0x200200880 is out of bounds
If i solve it with cublasDtrsm all is ok.
My matrix size: (20x20), ........(5010x5010);
I've got GeForce GT 540M on notebook and GeForce GTX 560 on computer.
What i do wrong?