Closed aosokin closed 8 years ago
Hi,
tutorial t03_usingCUDA.m runs the computation in double precision although in comments it is said that computations are done in single precision.
One my configuration line [u4, conv4, i4, time4] = asetsBinaryMF3D(gpuArray(Cs), gpuArray(Ct), gpuArray(alpha), pars); takes 5 seconds.
If I call [u4, conv4, i4, time4] = asetsBinaryMF3D(gpuArray(single(Cs)), gpuArray(single(Ct)), gpuArray(single(alpha)), pars); it takes 1.5 seconds.
Thank you for pointing this out. Your suggestions have been changed and pushed.
Hi,
tutorial t03_usingCUDA.m runs the computation in double precision although in comments it is said that computations are done in single precision.
One my configuration line [u4, conv4, i4, time4] = asetsBinaryMF3D(gpuArray(Cs), gpuArray(Ct), gpuArray(alpha), pars); takes 5 seconds.
If I call [u4, conv4, i4, time4] = asetsBinaryMF3D(gpuArray(single(Cs)), gpuArray(single(Ct)), gpuArray(single(alpha)), pars); it takes 1.5 seconds.