CUDA acceleration for trace and constraint commitments #1617
Conversation
| pollster = { version = "0.4", optional = true } | ||
|
|
||
| [target.'cfg(target_arch = "x86_64")'.dependencies] | ||
| miden-gpu = { path = "../../miden-gpu", optional = true } |
There was a problem hiding this comment.
This and line 33 need to be changed to a publicly released version.
| ProvingOptions::RECURSIVE_96_BITS.with_partitions(3, 8), | ||
| ProvingOptions::RECURSIVE_128_BITS.with_partitions(3, 8), | ||
| ProvingOptions::RECURSIVE_96_BITS.with_partitions(4, 8), | ||
| ProvingOptions::RECURSIVE_128_BITS.with_partitions(4, 8), |
There was a problem hiding this comment.
Please double check if these make sense to you.
There was a problem hiding this comment.
Yes - these make sense. I might've even removed the options with 3 partitions, but I guess no harm for keeping them.
There was a problem hiding this comment.
I removed the option with 3 partitions.
My intuition is that it's better to add it if somebody needs it (and can ensure that it works), rather than have an option that may or may not be used and therefore may or may not work in a future release.
| ProvingOptions::RECURSIVE_96_BITS.with_partitions(3, 8), | ||
| ProvingOptions::RECURSIVE_128_BITS.with_partitions(3, 8), | ||
| ProvingOptions::RECURSIVE_96_BITS.with_partitions(4, 8), | ||
| ProvingOptions::RECURSIVE_128_BITS.with_partitions(4, 8), |
There was a problem hiding this comment.
Yes - these make sense. I might've even removed the options with 3 partitions, but I guess no harm for keeping them.
air/src/options.rs
Outdated
| /// multiple devices. The number of partitions should be equal to the number of devices. | ||
| pub const fn with_partitions(mut self, num_partitions: usize) -> Self { | ||
| let hash_rate = match self.hash_fn { | ||
| HashFunction::Blake3_192 => 6, |
There was a problem hiding this comment.
I think this could be
There was a problem hiding this comment.
Changed to 8. The match is not really necessary but I kept it so that if HashFunctions enum ever changes, this code will be revisited.
Integrate CUDA hardware acceleration from miden-gpu. Supersedes #1474 and #1589
Blocked on the release of https://github.com/0xPolygonMiden/miden-gpu/pull/41. Make sure to update Cargo.toml