Get PTX for kernel about triton HOT 3 CLOSED

Hi! This is now part of the new Python API. You can do

binary = kernel[grid](...)
print(binary.asm('ptx'))

where binary is a handle to the fastest binary executed (it may depend on the input args through the auto-tuner. We are working on a better API that doesn't require executing the kernel (providing appropriate meta-parameters should be enough)

from triton.

Awesome, thank you!!

from triton.

FYI as of 06/29/2022, I think the syntax should be

binary = kernel[grid](...)
print(binary.asm['ptx'])

from triton.