Eventcounts are "lock-free condition variables".
The lock acquisition is removed from the hot path, significantly reducing contention and overhead.
Eventcounts were pioneered by Chris Thomasson and Dmitry Vyukov (from Go multithreading runtime, LLVM thread sanitizer, Relacy Race Detector and Eigen/Tensorflow multithreading runtime).
| $ ./bin/bls12_test.exe -m gmp_mont | |
| JIT 1 | |
| ctest:module=size | |
| ctest:module=naive | |
| i=0 curve=BLS12_381 | |
| G1 | |
| G2 | |
| GT | |
| G1::mulCT 656.335Kclk | |
| G1::mul 584.777Kclk |
| import benchy, random, streams | |
| import jsony, jason | |
| import eminim | |
| #import packedjson, packedjson/deserialiser | |
| import json | |
| import serialization | |
| import json_serialization except Json, toJson | |
| type Node = ref object | |
| active: bool |
https://github.com/MattPD/cpplinks / C++ Standard / C++20 / Coroutines
(draft; work in progress)
#coroutines (C++ Slack): https://cpplang.slack.com/archives/C5JS5JXT5
| title | author | excerpt |
|---|---|---|
Multithreading flavors: Choosing the right scheduler for the right job. |
Mamy Ratsimbazafy (mratsim) |
Demystifying multithreading for IO and multithreading for Compute. |
_Disclaimer: This post represent my current understanding as of December 2020 and might not age well.
With the advent of multicores CPU, developers are urge to blast through performance bottlenecks by exploiting more cores. Hence we are attracted to multithreading solutions like light is attracted by a black hole.
Eth2 call - ethereum/eth2.0-pm#196
Testing infra : changes plan for January next year
Proposal: accounting reform
| func dacPairing[HashLen]( | |
| sigsets: ptr UncheckedArray[SignatureSet[HashLen]], | |
| contexts: ptr UncheckedArray[ContextMultiAggregateVerify[DST]], | |
| numBatches: uint32, | |
| batchID: uint32, | |
| subsetStart: uint32, | |
| subsetStopEx: uint32 | |
| ): bool = | |
| ## Distribute pairing computation using a recursive divide-and-conquer (DAC) algorithm |
