Skip to content

Commit

Permalink
Faster simd when simd is the previous algo in the chain (1/16 probabi…
Browse files Browse the repository at this point in the history 
…lity):

Fugue +26%
Whirlpool +40%
Hamsi +40%
  • Loading branch information
@sp-hash
sp-hash committed May 9, 2018
1 parent 64f181e commit 4670dd1
Show file tree
Hide file tree
Showing 3 changed files with 983 additions and 7 deletions.

7 comments on commit 4670dd1

@Ravencoin-Miner
Copy link

@Ravencoin-Miner Ravencoin-Miner commented on 4670dd1 May 9, 2018 •
edited

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

(1/16)^2 probability not 1/16, roughly 0.4%

@sp-hash
Copy link
Owner Author

@sp-hash sp-hash commented on 4670dd1 May 9, 2018 •
edited

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

X16R: (1/16)^2 * 16 (to hit 2 pairs in a sequence) 1/16 (6.4%)
X16S; (1/16)*(1/15) (0.42%)

@pallas1
Copy link

@pallas1 pallas1 commented on 4670dd1 May 9, 2018

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Wrong it's (1/16)^2*15, simd as first and last should not be counted.

@buxstore
Copy link

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

result for 82fde399 does not validate on CPU! keccak 40DC5F13A26789BE

@sp-hash
Copy link
Owner Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

I have submitted a fix for the validation errors on some rigs

@buxstore
Copy link

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

N card？

@sp-hash
Copy link
Owner Author

@sp-hash sp-hash commented on 4670dd1 May 10, 2018 •
edited

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

yes
(1/16)^2 * 15 to find one pair in the sequence (5.8%)
((1/16)^2 * 14)^2 to find two pairs in the sequence (0.3%)
((1/16)^2 * 14)^3 to find three pairs in the sequence...
...
5.8%+0.3%

With 2 pairs in the sequence the gain is 2x bether

5.8%+(0.3%+0.3%)

So we end up with something like 1/16 :)

Please sign in to comment.