Dynamic shared memory allocation - Githubissues

HigherOrderCO / HVM

A massively parallel, optimal functional runtime in Rust

https://higherorderco.com

Apache License 2.0

10.52k stars 400 forks source link

Dynamic shared memory allocation #417

Open kings177 opened 3 months ago

kings177 commented 3 months ago

[x] Make a CUDA file to get max shared mem info.
[x] Capture output from CUDA file with memory info.
[x] Allow L_NODE_LEN and L_VARS_LEN to dynamically change.
[ ] Calculate performance loss per less memory allocated.
[ ] Return warning when set minimum memory for performance is not met.