search

a16z / jolt

The simplest and most extensible zkVM. Fast and fully open source from a16z crypto and friends. ⚡
https://jolt.a16zcrypto.com
MIT License
619 stars 123 forks source link

Optimize / Parallelize InstructionLookups::polynomialize #293

Open sragss opened 4 months ago

sragss commented 4 months ago

For a 64 core machine at a cycle count of ~16M Jolt spends ~4% of its time on segment called InstructionLookups::polynomialize here. This segment allocates and computes the offline memory checking (a,v,t) polynomials and other inputs for the Jolt instruction execution.

Currently much of the time is spent serially. We should parallelize this to get a speedup as the core count increases.

Similar to #292.

It may be helpful to review the tracing strategy for performance testing.

YashBit commented 2 weeks ago

@sragss I am interested. Could you assign it to me?