CUDA: Instruction fetch issues -


i have written kernel has issues instruction fetching. more 75% of issue stall reasons due instruction fetch.

i have no clue of how improve this. nvidia not instruction fetch policies , nsight documentation doesn't bring light either: "instruction fetch stall reason if next assembly instruction has not yet been fetched."

is there way of how avoid issue (or reduce impact) ?enter image description here


Comments

Popular posts from this blog

ios - iPhone/iPad different view orientations in different views , and apple approval process -

monitor web browser programmatically in Android? -

c# - Using multiple datasets in RDLC -