Parallel Thread Execution (PTX)

Ran into this term while working on the open-source GPU accelerated particle filter built on top of range_libc https://github.com/kctess5/range_libc/

Resources

• https://docs.nvidia.com/cuda/parallel-thread-execution/

• Source: https://www.youtube.com/watch?v=7lJZHbg0EQ4

So NVIDIA’s assembly is called SASS, but its not very public knowledge? No, PTX is the assembly. SASS is an even lower level of PTX.

You can generate PTX from a .cu file by using:

nvcc -ptx my_kernel.cu -o my_kernel.ptx