Skip to content

Instantly share code, notes, and snippets.

View ahennequ's full-sized avatar

Arthur Hennequin ahennequ

View GitHub Profile
@ahennequ
ahennequ / tensorcore_mapping.cu
Created September 15, 2022 18:04
Use this program to find out about tensor core's accumulator warp register layout
#include <stdio.h>
// Check tensor core's warp register layout
// nvcc -arch=sm_75 tensorcore_mapping.cu -o mapping
// ./mapping
// Define some error checking macros.
#define cudaErrCheck(stat) { cudaErrCheck_((stat), __FILE__, __LINE__); }
void cudaErrCheck_(cudaError_t stat, const char *file, int line) {
if (stat != cudaSuccess) {