Open feiliz opened 8 years ago
if i used "uC.f8v = C + (coord.y * ldc + coord.x)/8;", will it fix it ?
Did you test it with the latest devel or master branches?
@feiliz Hi, did you run the clBlas on the android with mali GPU? Do you have any instructions about how to complie?
could you give some suggust about it?
02-26 12:23:26.774: D/clblas(18822): startbuild-------:pgran->wgDim = 2, pgran->wgSize[0] = 8, pgran->wgSize[1] = 8, pgran->wfSize = 64 02-26 12:23:26.774: D/clblas(18822): device name = Mali-T880 02-26 12:23:26.774: D/clblas(18822): error = -11 02-26 12:23:26.774: D/clblas(18822): memory pattern = Cached global memory based block gemm, computing kernel generator 02-26 12:23:26.774: D/clblas(18822): Subproblem dimensions: dims[0].itemY = 32, dims[0].itemX = 64, dims[0].y = 32, dims[0].x = 64, dims[0].bwidth = 4; ; dims[1].itemY = 4, dims[1].itemX = 8, dims[1].y = 4, dims[1].x = 8, dims[1].bwidth = 4; ; 02-26 12:23:26.774: D/clblas(18822): Parallelism granularity: pgran->wgDim = 2, pgran->wgSize[0] = 8, pgran->wgSize[1] = 8, pgran->wfSize = 64 02-26 12:23:26.774: D/clblas(18822): Kernel extra flags: 536878080 02-26 12:23:26.774: D/clblas(18822): Source: GPtr uC; uC.f = C + (coord.y * ldc + coord.x)/8; __global float8 *pC = uC.f8v;
02-26 12:23:26.774: D/clblas(18822): Build log: 02-26 12:23:26.774: D/clblas(18822):