gpgpu-sim / gpgpu-sim_distribution

GPGPU-Sim provides a detailed simulation model of contemporary NVIDIA GPUs running CUDA and/or OpenCL workloads. It includes support for features such as TensorCores and CUDA Dynamic Parallelism as well as a performance visualization tool, AerialVisoin, and an integrated energy model, GPUWattch.
Other
1.03k stars 486 forks source link

Does GPGPU-Sim support Gunrock execution? #240

Open AKKamath opened 2 years ago

AKKamath commented 2 years ago

Can the execution-driven version of GPGPU-Sim support large, multi-file binaries like Gunrock? When I try to run it using GPGPU-Sim, I'm getting an error that PTX is not found, despite using appropriate "--cudart=shared" flags.

I edited the GPGPU-Sim to use "-all" when running cuobjdump, which resolved this problem but created new errors when parsing. I was wondering whether this was a limitation of GPGPU-Sim.

ishitachaturvedi commented 2 years ago

Hi, were you able to resolve the issue?

sxzhang1993 commented 2 years ago

Hello how did you use the -all option?

AKKamath commented 2 years ago

In the /libcuda/cuda_runtime_api.cc file, you will see multiple places (Line 3005, 3022, 3099, 3179) snprintf is called, where cuobjdump is part of the string. Here, I added -all.

E.g., line 3005: "$CUDA_INSTALL_PATH/bin/cuobjdump -lptx %s | cut -d \":\" -f 2 | " Becomes: "$CUDA_INSTALL_PATH/bin/cuobjdump -all -lptx %s | cut -d \":\" -f 2 | "