[pkg-gnupg-maint] Bug#848951: Bug#848951: gnupg: Utilize multiple cores on CPU for encryption and decryption (and compression)
Boyan Penkov
boyan.penkov at gmail.com
Wed Dec 21 13:33:38 UTC 2016
Hello Werner,
This sounds interesting, and I'd like to test it on my setups, however I
am quite new to gpg.
Is this straighforward? What changes should I make to gpg.conf to give
this a shot?
Cheers!
On 12/21/2016 03:11 AM, Werner Koch wrote:
> On Wed, 21 Dec 2016 06:57, witold.baryluk at gmail.com said:
>
>> Using cipher and compression algorithms that can utilize multiple cores
> It is not possible to parallelize encryption using the CFB mode as
> required by OpenPGP. In theory it would be possible to run the hashing
> (which is also run on the plaintext) on a different thread. However
> that would complicate matters a lot and I doubt that there will be a
> real benefit.
>
> What would really improve throughput is a different encryption mode to
> replace CFB and its SHA-1 based MDC. My suggestion to the WG is the use
> of OCB which would be a lot faster: On an X220 you get these values
>
> AES | nanosecs/byte mebibytes/sec cycles/byte
> CFB enc | 1.77 ns/B 537.9 MiB/s 4.08 c/B
> CFB dec | 0.373 ns/B 2557.6 MiB/s 0.858 c/B
> OCB enc | 0.436 ns/B 2189.4 MiB/s 1.00 c/B
> OCB dec | 0.452 ns/B 2107.9 MiB/s 1.04 c/B
>
> | nanosecs/byte mebibytes/sec cycles/byte
> SHA1 | 1.88 ns/B 507.7 MiB/s 4.32 c/B
>
> Thus the theoretical speedup would be
>
> CFB SHA1 OCB
> enc 1.77 + 1.88 = 3.65 0.44 8 times
> dec 0.37 + 1.88 = 2.25 0.45 5 times
>
>
>
> Shalom-Salam,
>
> Werner
>
--
Boyan Penkov
www.boyanpenkov.com
-------------- next part --------------
A non-text attachment was scrubbed...
Name: signature.asc
Type: application/pgp-signature
Size: 866 bytes
Desc: OpenPGP digital signature
URL: <http://lists.alioth.debian.org/pipermail/pkg-gnupg-maint/attachments/20161221/839cfb11/attachment.sig>
More information about the pkg-gnupg-maint
mailing list