Gina Sitaraman

Unable to use an internal RCCL build

4

Setting the following. env. variables does not result in the provided RCCL lib and include file to be used in the COSMA build: ``` export RCCL_ROOT_DIR="/path/to/install/rccl" export RCCL_INCLUDE_DIR="/path/to/install/rccl/include" export RCCL_LIB_DIR="/path/to/install/rccl/lib"...

Need for pinned host buffers in PW GPU backend

1

I was looking at a profile of 64-H2O-RPA which I ran with the PW GPU backend and observed that offloadMemcpyAsync calls were slower than expected. Tracing the caller path, I...

[Documentation]: Fix typo

2

### Description of errors Please fix typo "differentce" to "difference" in https://rocm.docs.amd.com/projects/HIP/en/latest/user_guide/faq.html#can-hip-api-support-c-style-application-what-is-the-differentce-between-c-and-c ### Attach any links, screenshots, or additional evidence you think will be helpful. _No response_

Add hipstdpar support to BabelStream

3

This PR adds support for offload to AMD GPUs using the `par_unseq` execution policy in C++ standard parallelism algorithms. To trigger the GPU offload of all parallel algorithms, the `--hipstdpar`...

Running `omniperf profile` after `omniperf profile --roof-only` erases all roofline pdfs

1

**Describe the bug** If we profiled the application with `omniperf profile` and then obtained roofline chart PDFs using `omniperf profile --kernel-names --roof-only`, then we have both the profiling data and...

Can there be an easier way to list launch params for a given kernel dispatch?

1

**Is your feature request related to a problem? Please describe.** Today, we have to do something like this to get launch parameters for a given dispatch. ``` omniperf analyze –p...

Unable to compare 2 kernels from same workload

1

It would be nice to easily compare 2 kernels from the same workload where counters were collected for all kernels. I would like to use a command such as: ```...

Improve documentation for usage with multi-process runs

6

Could some guidance be added in the documentation for using omniperf with MPI jobs? Should we collect profiles with omniperf for one rank only using a wrapper script that does...

[Documentation]: Could we have consistent naming for `roctx`?

1

### Description of errors In this page: [https://rocm.docs.amd.com/projects/roctracer/en/latest/reference/roctx-spec.html#roctx-spec](https://rocm.docs.amd.com/projects/roctracer/en/latest/reference/roctx-spec.html#roctx-spec) we see `roctx` being referred to as `ROC-TX` and `rocTX` in the very first line. There is no consistent naming for this...

[Documentation]: Add info about which meta package omnitrace and omniperf are part of

1

### Description of errors Today there is no reference to omnitrace and omniperf packages in [our ROCm package manager integration docs](https://rocm.docs.amd.com/projects/install-on-linux/en/latest/reference/package-manager-integration.html#packages-in-rocm-programming-models). We would like these tools to be packaged in...