gpu-microbench.git/copy_experiments, branch master

Don't assume that nvcc is available on PATH by default

2025-05-14T04:45:14+00:00

This change better aligns gpu-microbench with the assumptions of
the cuda_scheduling_examiner Makefile.

Ensure all benchmarks support --help and cleanup

2024-02-27T01:36:27+00:00

Add gl_copies benchmark and make target

2024-02-27T01:33:25+00:00

=> Uploads a large texture to the GPU on every frame
=> Requires the freeglut library (freeglut3-dev on Ubuntu)
=> Due to special requirements, not built by default
=> Written for Anderson and Bakita RTAS'24 paper

Fix timestamping bugs and clean up CPU side of mon_cross_ctx_copies

2024-02-21T19:05:34+00:00

- Add a more reliable heuristic for how long to run the CPU-side
  copy monitor.
- Check that copies sizes are appropriately divisible.
- Fix parsing issue for copies larger than 8192 GiB.
- Clean up, and document bugs in, the time synchronization code.
- Wait for user before logging copies.

Always synchronize GPU and CPU time in the current process

2024-02-20T01:53:09+00:00

Performing the synchronization in a child process was fragile,
high-overhead, and unnecessary for our use cases.

Support >4 GiB copy sizes in `mon_cross_ctx_copies`

2024-02-20T01:43:56+00:00

Split SAFE() macro into API-specific typed variants

2024-02-20T01:39:39+00:00

Avoids incorrectly interpreting CUDA Driver Library error codes as
though they were CUDA Runtime Library error codes (the numbering is
similar, but not identical).

Also cleans up how we initialize and terminate a context for
capability checking in `mon_cross_ctx_copies`.

Allow custom nvcc location when running make + ignore gdb history

2023-10-25T13:50:48+00:00

To specify a custom path to nvcc when building:
`make NVCC=/my/path/to/nvcc`

Add a "clean" target to the Makefiles

2023-07-05T20:49:03+00:00

Comments and check for successful cleanup in mon_cross_ctx_copies

2023-07-05T20:45:46+00:00