wow that looks fantastic! been using blake2sp from the blake2s_simd crate but this is way faster
-
-
-
On an AVX2 machine, if you don't enable --features=rayon, the difference from BLAKE2sp will be modest, since the two implementations have a lot in common. But on an AVX-512 machine, or with small inputs, or if you do enable Rayon, it's way faster :)
- Još 6 drugih odgovora
Novi razgovor -
-
-
Where's the break-even input size on your test c5 machine for when the coordination cost of using rayon is made up for by the parallelism?
-
If you look at the Performance section in our spec, there's this graph from the c5.metal instance. The break-even input size depends on the number of threads, so it's hard to give a single answer. On my Raspberry Pi 2, though, I've found that threading ASAP is beneficial.pic.twitter.com/JrRB77ABjt
- Još 10 drugih odgovora
Novi razgovor -
-
-
Nice do you have any ARM benchmarks ? (look for M6g in AWS, also Android would be interesting)
-
The BLAKE3 paper (https://github.com/BLAKE3-team/BLAKE3-specs/blob/master/blake3.pdf …) includes benchmarks on a 32-bit ARM1176. See Figure 5 in the Performance section.pic.twitter.com/3VY7Jg8leC
Kraj razgovora
Novi razgovor -
-
-
how is the performance boost in a single thread environment?
-
The red bar chart *is* a single threaded implementation :)
- Još 2 druga odgovora
Novi razgovor -
-
-
Hvala. Twitter će to iskoristiti za poboljšanje vaše vremenske crte. PoništiPoništi
-
-
-
When libsodium?
Hvala. Twitter će to iskoristiti za poboljšanje vaše vremenske crte. PoništiPoništi
-
Čini se da učitavanje traje već neko vrijeme.
Twitter je možda preopterećen ili ima kratkotrajnih poteškoća u radu. Pokušajte ponovno ili potražite dodatne informacije u odjeljku Status Twittera.
