Enable register spilling to shared memory#1132
Enable register spilling to shared memory#1132stephenswat wants to merge 1 commit intoacts-project:mainfrom
Conversation
|
|
I'm not 100% certain this works as intended like this, as this pragma is to be attached at the function scope. But we can try. |
|
This is interesting as we are not actively using the shared memory in our finding and fitting kernels. |
CUDA 13.0 enables the PTX assembler to spill registers to shared memory instead of local memory, which should both be much faster, and also reduce the local memory usage of our fitting and finding kernels which are currently bottlenecking our throughput.
13238d6 to
2e6b339
Compare
|
Physics performance summaryHere is a summary of the physics performance effects of this PR. Command used: Seeding performanceTotal number of seeds went from 33828 to 33828 (+0.0%) Track finding performanceTotal number of found tracks went from 5565 to 5569 (+0.1%) Track fitting performanceSeeding to track finding relative performanceNote This is an automated message produced on the explicit request of a human being. |
































CUDA 13.0 enables the PTX assembler to spill registers to shared memory instead of local memory, which should both be much faster, and also reduce the local memory usage of our fitting and finding kernels which are currently bottlenecking our throughput.