diff options
Diffstat (limited to 'README.md')
| -rw-r--r-- | README.md | 7 |
1 files changed, 2 insertions, 5 deletions
| @@ -44,7 +44,8 @@ And if `nvdebug` has been installed: | |||
| 44 | 44 | ||
| 45 | #### Known Issues | 45 | #### Known Issues |
| 46 | 46 | ||
| 47 | - `next_mask` will not override `stream_mask` on CUDA 12.0+ | 47 | - `next_mask` will not override `stream_mask` on CUDA 11.0+ |
| 48 | - _As of Feb 2024, a fix for this is coming soon..._ | ||
| 48 | - `global_mask` and `next_mask` cannot disable TPCs with IDs above 128 | 49 | - `global_mask` and `next_mask` cannot disable TPCs with IDs above 128 |
| 49 | - Only relevant on GPUs with over 128 TPCs, such as the RTX 6000 Ada | 50 | - Only relevant on GPUs with over 128 TPCs, such as the RTX 6000 Ada |
| 50 | - Untested on H100 (compute capability 9.0) | 51 | - Untested on H100 (compute capability 9.0) |
| @@ -74,7 +75,3 @@ How this works: | |||
| 74 | 3. If the test succeeded (returned zero) the loop aborts, otherwise it increments the offset to attempt and repeats. | 75 | 3. If the test succeeded (returned zero) the loop aborts, otherwise it increments the offset to attempt and repeats. |
| 75 | 76 | ||
| 76 | Once this loop aborts, take the found offset and add it into the switch statement for the appropriate CUDA version and CPU architecture. | 77 | Once this loop aborts, take the found offset and add it into the switch statement for the appropriate CUDA version and CPU architecture. |
| 77 | |||
| 78 | ## TODO | ||
| 79 | |||
| 80 | - Add a test to check that more-granularly-set compute masks override more-corsely-set ones. | ||