3 ms·
If the authors of Blake3 are to be believed, MD5 is slower than Blake3, Blake2, and SHA1. And on par with SHA512. (On a given architecture) https://raw.github
by ReactiveJelly 6y ago
If the authors of Blake3 are to be believed, MD5 is slower than Blake3, Blake2, and SHA1. And on par with SHA512.
(On a given architecture)
https://raw.githubusercontent.com/BLAKE3-team/BLAKE3/master/media/speed.svg https://raw.githubusercontent.com/BLAKE3-team/BLAKE3/master/...
I could understand if hashing is a large portion of their CPU time. But if it's not, why not compute MD5 and also something good and store both hashes?
- oconnor663 6y ago> (On a given architecture) Yes this is a good caveat. When I've tried the OpenSSL implementations on older ARM chips (a Raspberry Pi Zero), I've seen MD5 come out on top. I'm not sure exactly why it switches.
- perihelions 6y agoI think they're just inverted on x86 because it provides hardware acceleration for SHA-1 and SHA-2. https://en.wikipedia.org/wiki/Intel_SHA_extensions https://en.wikipedia.org/wiki/Intel_SHA_extensions https://github.com/openssl/openssl/blob/master/crypto/sha/asm/sha1-x86_64.pl#L412 https://github.com/openssl/openssl/blob/master/crypto/sha/as...
- oconnor663 6y agoI find SHA-1 to be faster than MD5 on Skylake, which does not have the SHA extensions.