[Feature]: Run kernel with cooperative thread blocks and performance measurement ability

ROCm / clr

MIT License

104 stars 50 forks source link

[Feature]: Run kernel with cooperative thread blocks and performance measurement ability #60

Closed averinevg closed 8 months ago

averinevg commented 9 months ago

Suggestion Description

Currently hipExtModuleLaunchKernel() is the only function which has the performance measurement capability. I think it would be a good idea to add a function with the same list of parameters for launching kernels with cooperative thread groups.

CC @junliume

Operating System

No response

GPU

No response

ROCm Component

No response

averinevg commented 8 months ago

Closed in favor of SWDEV-448157