github/ZeroTierOne
1
0
Fork 0
mirror of https://github.com/ZeroTier/ZeroTierOne synced 2025-08-21 13:54:15 -07:00
Code Issues Projects Releases Packages Wiki Activity Actions 6

Use X64 ASM ed25519 signatures on Linux/x64, which are about 10X faster. Will matter a lot for network controllers, not so much for other things.

Browse source
This commit is contained in:
Adam Ierymenko 2018-03-13 06:51:17 -07:00
commit beb170e4fb
68 changed files with 37123 additions and 68 deletions
Download patch file Download diff file Expand all files Collapse all files

102
ext/ed25519-amd64-asm/ge25519_double_scalarmult.c Normal file
View file

@ -0,0 +1,102 @@
#include "fe25519.h"
#include "sc25519.h"
#include "ge25519.h"
#define S1_SWINDOWSIZE 5
#define PRE1_SIZE (1<<(S1_SWINDOWSIZE-2))
#define S2_SWINDOWSIZE 7
#define PRE2_SIZE (1<<(S2_SWINDOWSIZE-2))
ge25519_niels pre2[PRE2_SIZE] = {
#include "ge25519_base_slide_multiples.data"
};
static const fe25519 ec2d = {{0xEBD69B9426B2F146, 0x00E0149A8283B156, 0x198E80F2EEF3D130, 0xA406D9DC56DFFCE7}};
static void setneutral(ge25519 *r)
{
fe25519_setint(&r->x,0);
fe25519_setint(&r->y,1);
fe25519_setint(&r->z,1);
fe25519_setint(&r->t,0);
}
/* computes [s1]p1 + [s2]p2 */
void ge25519_double_scalarmult_vartime(ge25519_p3 *r, const ge25519_p3 *p1, const sc25519 *s1, const sc25519 *s2)
{
signed char slide1[256], slide2[256];
ge25519_pniels pre1[PRE1_SIZE], neg;
ge25519_p3 d1;
ge25519_p1p1 t;
ge25519_niels nneg;
fe25519 d;
int i;
sc25519_slide(slide1, s1, S1_SWINDOWSIZE);
sc25519_slide(slide2, s2, S2_SWINDOWSIZE);
/* precomputation */
pre1[0] = *(ge25519_pniels *)p1;
ge25519_dbl_p1p1(&t,(ge25519_p2 *)pre1); ge25519_p1p1_to_p3(&d1, &t);
/* Convert pre[0] to projective Niels representation */
d = pre1[0].ysubx;
fe25519_sub(&pre1[0].ysubx, &pre1[0].xaddy, &pre1[0].ysubx);
fe25519_add(&pre1[0].xaddy, &pre1[0].xaddy, &d);
fe25519_mul(&pre1[0].t2d, &pre1[0].t2d, &ec2d);
for(i=0;i<PRE1_SIZE-1;i++)
{
ge25519_pnielsadd_p1p1(&t, &d1, &pre1[i]); ge25519_p1p1_to_p3((ge25519_p3 *)&pre1[i+1], &t);
/* Convert pre1[i+1] to projective Niels representation */
d = pre1[i+1].ysubx;
fe25519_sub(&pre1[i+1].ysubx, &pre1[i+1].xaddy, &pre1[i+1].ysubx);
fe25519_add(&pre1[i+1].xaddy, &pre1[i+1].xaddy, &d);
fe25519_mul(&pre1[i+1].t2d, &pre1[i+1].t2d, &ec2d);
}
setneutral(r);
for (i = 255;i >= 0;--i) {
if (slide1[i] || slide2[i]) goto firstbit;
}
for(;i>=0;i--)
{
firstbit:
ge25519_dbl_p1p1(&t, (ge25519_p2 *)r);
if(slide1[i]>0)
{
ge25519_p1p1_to_p3(r, &t);
ge25519_pnielsadd_p1p1(&t, r, &pre1[slide1[i]/2]);
}
else if(slide1[i]<0)
{
ge25519_p1p1_to_p3(r, &t);
neg = pre1[-slide1[i]/2];
d = neg.ysubx;
neg.ysubx = neg.xaddy;
neg.xaddy = d;
fe25519_neg(&neg.t2d, &neg.t2d);
ge25519_pnielsadd_p1p1(&t, r, &neg);
}
if(slide2[i]>0)
{
ge25519_p1p1_to_p3(r, &t);
ge25519_nielsadd_p1p1(&t, r, &pre2[slide2[i]/2]);
}
else if(slide2[i]<0)
{
ge25519_p1p1_to_p3(r, &t);
nneg = pre2[-slide2[i]/2];
d = nneg.ysubx;
nneg.ysubx = nneg.xaddy;
nneg.xaddy = d;
fe25519_neg(&nneg.t2d, &nneg.t2d);
ge25519_nielsadd_p1p1(&t, r, &nneg);
}
ge25519_p1p1_to_p2((ge25519_p2 *)r, &t);
}
}