Improve the OOP ABI

Bruce Carneal bcarneal at gmail.com
Tue Jan 23 02:09:52 UTC 2024


On Monday, 22 January 2024 at 22:44:20 UTC, Walter Bright wrote:
> On 1/22/2024 1:11 PM, Bruce Carneal wrote:
>> Blake3 might be worth a look.  It's reportedly faster and 
>> stronger than md5.
>> 
>> https://github.com/BLAKE3-team/BLAKE3
>
> Thanks for the tip. I don't know how to evaluate a hash 
> function for uniqueness.

I've not looked into that aspect either.  I'd just assumed that 
all modern hashes were quite good and was just looking at the 
time to compute the hash.  I'm not at all concerned about 
accidental collisions apart from a broken implementation.  As you 
say, more likely that we get smacked by a meteor.

I've read hash comparisons claiming blake3 is anywhere from 3X to 
9X faster than MD5 on a single core wrt throughput.  No idea on 
how things would work out for this particular use case though, 
nor even if a 9X throughput upgrade would be noticeable.  I'd 
suspect not, just wanted to put forward an alternative.



More information about the Digitalmars-d mailing list