feat(msm-gpu-exploration): integrate zprize msm #128
Conversation
…chmarking at once
…x 10 benchmark data can be gen in 5 min
moven0831
force-pushed
the
feat/integrate-zprize-msm
branch
from
ace73e7
to
66b0ae2
Compare
Cargo.toml
Outdated
| @@ -7,3 +7,10 @@ exclude = ["mopro-cli-example"] | |||
| # NOTE: Forked wasmer to work around memory limits | |||
| # See https://github.com/wasmerio/wasmer/commit/09c7070 | |||
| wasmer = { git = "https://github.com/oskarth/wasmer.git", rev = "09c7070" } | |||
|
|
|||
| # NOTE: For gpu exploration on zprize works | |||
There was a problem hiding this comment.
Can we put all of these under the gpu explorations feature flag perhaps? As optional
There was a problem hiding this comment.
I've added optional = true for the patch in commits fc0c972, which will only be included when --features=gpu-benchmarks is activated.
| @@ -921,6 +921,15 @@ public func add(a: UInt32, b: UInt32) -> UInt32 { | |||
| ) | |||
There was a problem hiding this comment.
FYI we will probably get rid of mopro-ios and use mopro-cli - perhaps we can initialize a new project with mopro-cli and put in examples/templates?
There was a problem hiding this comment.
sounds great! we will start this template ASAP.
There was a problem hiding this comment.
So ideally the whole swift test app will use mopro-cli and in examples/templates?
@oskarth
There was a problem hiding this comment.
yes! we have the basic template for simple apps, but benchmarks should probably be separate, can init with mopro-cli init
There was a problem hiding this comment.
Hi @oskarth, maybe we will create one after this PR since we have to import the msm functions from this PR
|
Hi @oskarth
I think add a |
|
iOS CI failing, not sure why |
moven0831
force-pushed
the
feat/integrate-zprize-msm
branch
from
6639bd1
to
c1243e0
Compare
… now test_ios shoud success
moven0831
force-pushed
the
feat/integrate-zprize-msm
branch
from
c1243e0
to
7103cc5
Compare
| // let pointUrl = URL(string: "https://mopro-msm.s3.eu-north-1.amazonaws.com/vectors/16x10/points") | ||
| // let scalarUrl = URL(string: "https://mopro-msm.s3.eu-north-1.amazonaws.com/vectors/16x10/scalars") | ||
| // func runDownloadAction() { | ||
| // let pointStart = CFAbsoluteTimeGetCurrent() | ||
| // FileDownloader.loadFileAsync(url: self.pointUrl!) { (path, error) in | ||
| // print("Points File downloaded to : \(path!)") | ||
| // let pointEnd = CFAbsoluteTimeGetCurrent() | ||
| // print("Download points took:", pointEnd - pointStart, "s") | ||
| // } | ||
|
|
||
| // let scalarStart = CFAbsoluteTimeGetCurrent() | ||
| // FileDownloader.loadFileAsync(url: self.scalarUrl!) { (path, error) in | ||
| // print("Scalars File downloaded to : \(path!)") | ||
| // let scalarEnd = CFAbsoluteTimeGetCurrent() | ||
| // print("Download scalars took:", scalarEnd - scalarStart, "s") | ||
| // } | ||
|
|
||
| // } |
There was a problem hiding this comment.
If we can confirm that the benchmarking instances can always be generated on both iOS simulator and mobile device, we should consider removing these comments in the later PR.
Overall
@moven0831 and I integrated the work of TrapdoorTech in zprize into
mopro-core,mopro-ffiand run the benchmark on myiphone 14 pro.And here is the result:
Result running on FoodChain's Iphone 14 pro
logging:
result screenshot:
arkworkscomponents they used were in version0.3.arkworksMSM algorithm (version:0.4) has been modified and probably taken some of the awarded algorithm in zprize as references, so that is why runs faster than the old work as shown in the result.Report
related issue
#22