Yifan Sun
Yifan Sun
`-unified-gpus` allows the memory to be allocated evenly on these GPUs and threads be distributed evenly on these GPUs. `-use-unified-memory` allows page migration. There may be no migration if the...
Could you please create a `main.go` file so that we can reproduce this bug?
We are using the .sqlite3 file instead of CSV files now. The goal is to create only one file per simulation run. So we will update README. I tried to...
A fully connected layer is basically a GEMM. So, you can use the GEMM benchmark instead. That is why we do not have a fully connected layer benchmark.
The compiler that has been stable to use is the one in ROCm 3.8. We have been using the ROCm docker image to compile HSACO. Can you try to see...
Yes. We have always been using Linux to compile the HSACOs.
We have not been creating HSACOs for a long time. What error did you get?
@ch1y0q You are right in the additional context point 1. The RDMA-based access is at cache-line granularity. So, they do not involve page migration. Since you turned on `-unified-memory`, which...
@ch1y0q v3.0.0 should be a reasonable version that has a relatively stable unified-memory implementation. Option 2 is to go back to the original version of MGPUSim on GitLab https://gitlab.com/akita/mgpusim. Option...
Unified memory is known to have a hanging problem. We are in the process of reimplementing the unified memory system. Closed for now.