Fix prehash crashing on newer NVIDIA drivers #54
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
Using any NVIDIA driver that is part of a 5XX release branch (such as any drivers released after and including 510.39.01 for Linux or 511.23 for Windows) will cause the miner to crash with cudaErrorIllegalAddress (700) at the prehash stage. Downgrading to an older driver such as 470.141.03 has been a stop-gap solution to getting the miner running again without this patch.
The issue is coming from this line:
Autolykos2_NV_Miner/secp256k1/src/prehash.cu
Line 251 in 4de5cea
Simply signaling to the compiler to disable unrolling for this one loop seems to be enough to make the driver happy with no measurable performance difference as far as I have measured.
Fixes #51 #53