]> git.ipfire.org Git - thirdparty/nettle.git/commit
projects / thirdparty / nettle.git / commit
? search:
summary | shortlog | log | commit | commitdiff | tree
(parent: 95b29fe) | patch
"PowerPC64" Add optimized AES [Enc|Dec]
authorMaamoun TK <maamoun.tk@googlemail.com>
Tue, 14 Jul 2020 11:43:06 +0000 (14:43 +0300)
committerNiels Möller <nisse@lysator.liu.se>
Wed, 26 Aug 2020 16:13:51 +0000 (18:13 +0200)
commit9df0079edc6c2552ee2c162d9b569f7cccb5e215
treef0c8e2d7caa51cb0d81a30ed5e31168e065b0dcdtree
parent95b29fe510cc35cf617e730ed7fa951e982e1532commit | diff
"PowerPC64" Add optimized AES [Enc|Dec]

I measured the latency and throughput of vcipher/vncipher/vxor instructions
for POWER8
vcipher/vncipher
throughput 6 instructions per cycle
latency 0.91 clock cycles
vxor
throughput 6 instructions per cycle
latency 0.32 clock cycles
So the ideal option for POWER8 is processing 8 blocks, it has +12%
performance over processing 4 blocks.
powerpc64/P8/aes-decrypt-internal.asm [new file with mode: 0644] blob
powerpc64/P8/aes-encrypt-internal.asm [new file with mode: 0644] blob
Mirror of https://git.lysator.liu.se/nettle/nettle.git