From d64a11755ef92091a5430b002ebf507be9ffe182 Mon Sep 17 00:00:00 2001 From: Carl Pearson Date: Thu, 30 Apr 2020 16:02:37 -0500 Subject: [PATCH] Update CONTRIBUTING.md --- CONTRIBUTING.md | 11 ++++++++++- 1 file changed, 10 insertions(+), 1 deletion(-) diff --git a/CONTRIBUTING.md b/CONTRIBUTING.md index 3df7242..6c9ffb7 100644 --- a/CONTRIBUTING.md +++ b/CONTRIBUTING.md @@ -43,10 +43,19 @@ Carl Pearson's docker hub account is used to push images up to [cwpearson/nvidia * [ ] multi-device * [ ] interacting with `.qdrep` files. * [ ] interacting with `.nsight-cuprof-report` files. +* [ ] Best Practices + * [ ] Fixing GPU frequency + * [ ] initial CUDA runtime cost + * [ ] Warmup Kernels + * [ ] `cuda-memcheck` race condition and sync check? +* Is stream 0 the default stream? +* Nsight System with MPI +* Nsight System with multi-GPU + ## Inspirations * https://developer.download.nvidia.com/video/gputechconf/gtc/2020/presentations/s22141-what-the-profiler-is-telling-you-how-to-get-the-most-performance-out-of-your-hardware.pdf * https://developer.nvidia.com/gtc/2020/video/s22141 * https://developer.download.nvidia.com/video/gputechconf/gtc/2020/presentations/s21351-scaling-the-transformer-model-implementation-in-pytorch-across-multiple-nodes.pdf - * https://developer.nvidia.com/gtc/2020/video/s21351 \ No newline at end of file + * https://developer.nvidia.com/gtc/2020/video/s21351