[<prev] [next>] [thread-next>] [day] [month] [year] [list]
Date: Fri, 29 May 2015 22:52:41 +0200
From: magnum <john.magnum@...hmail.com>
To: john-dev@...ts.openwall.com
Subject: Interleaving of intrinsics
Solar,
Here's a GitHub issue where we discuss interleaving and present some
benchmarks:
https://github.com/magnumripper/JohnTheRipper/issues/1217
I think you will have some educated thoughts about this; Here's part of
our current SHA-1:
#define SHA1_PARA_DO(x) for((x)=0;(x)<SIMD_PARA_SHA1;(x)++)
#define SHA1_ROUND2(a,b,c,d,e,F,t) \
SHA1_PARA_DO(i) tmp3[i] = tmpR[i*16+(t&0xF)]; \
SHA1_EXPAND2(t+16) \
F(b,c,d) \
SHA1_PARA_DO(i) e[i] = vadd_epi32( e[i], tmp[i] ); \
SHA1_PARA_DO(i) tmp[i] = vroti_epi32(a[i], 5); \
SHA1_PARA_DO(i) e[i] = vadd_epi32( e[i], tmp[i] ); \
SHA1_PARA_DO(i) e[i] = vadd_epi32( e[i], cst ); \
SHA1_PARA_DO(i) e[i] = vadd_epi32( e[i], tmp3[i] ); \
SHA1_PARA_DO(i) b[i] = vroti_epi32(b[i], 30);
And here's a similar part of SHA256:
#define SHA256_STEP0(a,b,c,d,e,f,g,h,x,K) \
{ \
SHA256_PARA_DO(i) \
{ \
w = _w[i].w; \
tmp1[i] = vadd_epi32(h[i], S1(e[i])); \
tmp1[i] = vadd_epi32(tmp1[i], Ch(e[i],f[i],g[i])); \
tmp1[i] = vadd_epi32(tmp1[i], vset1_epi32(K)); \
tmp1[i] = vadd_epi32(tmp1[i], w[x]); \
tmp2[i] = vadd_epi32(S0(a[i]),Maj(a[i],b[i],c[i])); \
d[i] = vadd_epi32(tmp1[i], d[i]); \
h[i] = vadd_epi32(tmp1[i], tmp2[i]); \
} \
}
This file is -O3 (from a pragma) so I guess both cases will be unrolled
but there is obviously a big difference after just unrolling. Assuming a
perfect optimizer it wouldn't matter but assuming a non-perfect one, is
the former better? I'm guessing SHA-1 was written that way for a reason?
magnum
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.
