maxas
maxas copied to clipboard
Assembler for NVIDIA Maxwell architecture
Hi! Thank you for this repo! It is very helpful to me!!! I have a question, in the wiki part, the last comparison figure has a Max64-8 or Max64-16. I...
Wiki / Getting Started page indicates that the microbenchmark program be built with Visual Studio, but there are no project or solution files in the distribution.
why no updates?
I have added CSET and CSETP as I require them for some algorithms.
https://github.com/NervanaSystems/maxas/wiki/SGEMM 
using microbench, with .cu modified to be simply: ``` extern "C" __global__ void microbench(int *out, int *clocks, int *in) { out[0] = 7.0f; out[2] = 5.0f; clocks[1] = 9.0f; }...
Address https://github.com/NervanaSystems/maxas/issues/10
Address https://groups.google.com/forum/#!topic/maxas-discuss/4rovrjSRzKA partially (It would be nice to modify the parser too, and remove the redundant output, in the next line of the currently parsed '-:-:-:-:-' syntax, but ... baby-steps)
Per the message in microbench.cpp, minor should be >=0: ``` printf("No compute 5.0 device found, exiting.\n"); ``` But hte check is for `>=2`: ``` if (major >= 5 && minor...