Date: Thu, 27 Jun 2013 19:08:55 +0400 From: Solar Designer <solar@...nwall.com> To: john-dev@...ts.openwall.com Subject: Re: Parallella: bcrypt On Thu, Jun 27, 2013 at 04:54:31PM +0200, Katja Malvoni wrote: > In whole disassembly only ADD is used. That's something we'll need to try to deal with (a bit later). > Here it is: > 00000234 <_BF_encrypt>: > 234: d54c 4400 ldr r22,[sp,+0x2] [...] > 31a: 50ef 0402 mov r2,r12 > 31e: 9120 bgtu 240 <_BF_encrypt+0xc> > 320: 04e2 mov r0,r1 > 322: 194f 0402 rts We may try two things: 1. Interleave two instances of bcrypt. and/or 2. Rewrite this function in assembly. The compiler-generated code does look suboptimal. > Execution time when using all 16 cores is 294.676000 ms Why is it this much higher than for 1 core? Is most time spent on data transfers? Alexander
Powered by blists - more mailing lists
Confused about mailing lists and their use? Read about mailing lists on Wikipedia and check out these guidelines on proper formatting of your messages.