Replies: 1 comment 1 reply
-
I think lots of Turing+ features are not supported in gppgu-sim PTX mode. To support that, change the front end to parse for the s8 identifier in the PTX instruction. |
Beta Was this translation helpful? Give feedback.
1 reply
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
-
I use the gpgpusim in accel-sim, when i want to run cutlass-bench/examples/05_wmma_gemm with turing architecture, I encountered the issue "Aborted (core dumped). The detailed error messages are: “./ptx_ir.h:1053: int ptx_instruction::get_type2() const: Assertion `m_scalar_type.size() == 2' failed.” and "error: set_wmma_vector_operands".
Before this, I had already resolved some issues, which were mentioned in gpgpu-sim/gpgpu-sim_distribution#296
Is this issue occurring because GPGPU-Sim does not yet support using tensor cores for s8 matrix multiplication and can only perform float16 matrix multiplication?
Thank you very much. I hope you can provide assistance.
Beta Was this translation helpful? Give feedback.
All reactions