aboutsummaryrefslogtreecommitdiff
path: root/networking/tls_sp_c32.c
diff options
context:
space:
mode:
authorDenys Vlasenko <vda.linux@googlemail.com>2021-04-27 13:09:44 +0200
committerDenys Vlasenko <vda.linux@googlemail.com>2021-04-27 13:09:44 +0200
commit646e85629497ee364e97101de4402d7141919144 (patch)
tree86913653cb8f3552f96061fa24360b6265bdd25a /networking/tls_sp_c32.c
parent48a18d15dfbfa41d137802e811a3abdfef012ac8 (diff)
downloadbusybox-w32-646e85629497ee364e97101de4402d7141919144.tar.gz
busybox-w32-646e85629497ee364e97101de4402d7141919144.tar.bz2
busybox-w32-646e85629497ee364e97101de4402d7141919144.zip
tls: shrink sp_256_mod_mul_norm_10
function old new delta sp_256_mod_mul_norm_10 1439 1405 -34 Signed-off-by: Denys Vlasenko <vda.linux@googlemail.com>
Diffstat (limited to 'networking/tls_sp_c32.c')
-rw-r--r--networking/tls_sp_c32.c145
1 files changed, 85 insertions, 60 deletions
diff --git a/networking/tls_sp_c32.c b/networking/tls_sp_c32.c
index f9c66b186..c5e887aad 100644
--- a/networking/tls_sp_c32.c
+++ b/networking/tls_sp_c32.c
@@ -460,51 +460,90 @@ static void sp_256_mont_inv_10(sp_digit* r, sp_digit* a)
460static void sp_256_mod_mul_norm_10(sp_digit* r, const sp_digit* a) 460static void sp_256_mod_mul_norm_10(sp_digit* r, const sp_digit* a)
461{ 461{
462 int64_t t[8]; 462 int64_t t[8];
463 int64_t a32[8];
464 int64_t o; 463 int64_t o;
465 464 uint32_t a32;
466 a32[0] = a[0];
467 a32[0] |= a[1] << 26;
468 a32[0] &= 0xffffffff;
469 a32[1] = (sp_digit)(a[1] >> 6);
470 a32[1] |= a[2] << 20;
471 a32[1] &= 0xffffffff;
472 a32[2] = (sp_digit)(a[2] >> 12);
473 a32[2] |= a[3] << 14;
474 a32[2] &= 0xffffffff;
475 a32[3] = (sp_digit)(a[3] >> 18);
476 a32[3] |= a[4] << 8;
477 a32[3] &= 0xffffffff;
478 a32[4] = (sp_digit)(a[4] >> 24);
479 a32[4] |= a[5] << 2;
480 a32[4] |= a[6] << 28;
481 a32[4] &= 0xffffffff;
482 a32[5] = (sp_digit)(a[6] >> 4);
483 a32[5] |= a[7] << 22;
484 a32[5] &= 0xffffffff;
485 a32[6] = (sp_digit)(a[7] >> 10);
486 a32[6] |= a[8] << 16;
487 a32[6] &= 0xffffffff;
488 a32[7] = (sp_digit)(a[8] >> 16);
489 a32[7] |= a[9] << 10;
490 a32[7] &= 0xffffffff;
491 465
492 /* 1 1 0 -1 -1 -1 -1 0 */ 466 /* 1 1 0 -1 -1 -1 -1 0 */
493 t[0] = 0 + a32[0] + a32[1] - a32[3] - a32[4] - a32[5] - a32[6];
494 /* 0 1 1 0 -1 -1 -1 -1 */ 467 /* 0 1 1 0 -1 -1 -1 -1 */
495 t[1] = 0 + a32[1] + a32[2] - a32[4] - a32[5] - a32[6] - a32[7];
496 /* 0 0 1 1 0 -1 -1 -1 */ 468 /* 0 0 1 1 0 -1 -1 -1 */
497 t[2] = 0 + a32[2] + a32[3] - a32[5] - a32[6] - a32[7];
498 /* -1 -1 0 2 2 1 0 -1 */ 469 /* -1 -1 0 2 2 1 0 -1 */
499 t[3] = 0 - a32[0] - a32[1] + 2 * a32[3] + 2 * a32[4] + a32[5] - a32[7];
500 /* 0 -1 -1 0 2 2 1 0 */ 470 /* 0 -1 -1 0 2 2 1 0 */
501 t[4] = 0 - a32[1] - a32[2] + 2 * a32[4] + 2 * a32[5] + a32[6];
502 /* 0 0 -1 -1 0 2 2 1 */ 471 /* 0 0 -1 -1 0 2 2 1 */
503 t[5] = 0 - a32[2] - a32[3] + 2 * a32[5] + 2 * a32[6] + a32[7];
504 /* -1 -1 0 0 0 1 3 2 */ 472 /* -1 -1 0 0 0 1 3 2 */
505 t[6] = 0 - a32[0] - a32[1] + a32[5] + 3 * a32[6] + 2 * a32[7];
506 /* 1 0 -1 -1 -1 -1 0 3 */ 473 /* 1 0 -1 -1 -1 -1 0 3 */
507 t[7] = 0 + a32[0] - a32[2] - a32[3] - a32[4] - a32[5] + 3 * a32[7]; 474 // t[] should be calculated from "a" (converted from 26-bit to 32-bit vector a32[8])
475 // according to the above matrix:
476 //t[0] = 0 + a32[0] + a32[1] - a32[3] - a32[4] - a32[5] - a32[6] ;
477 //t[1] = 0 + a32[1] + a32[2] - a32[4] - a32[5] - a32[6] - a32[7] ;
478 //t[2] = 0 + a32[2] + a32[3] - a32[5] - a32[6] - a32[7] ;
479 //t[3] = 0 - a32[0] - a32[1] + 2*a32[3] + 2*a32[4] + a32[5] - a32[7] ;
480 //t[4] = 0 - a32[1] - a32[2] + 2*a32[4] + 2*a32[5] + a32[6] ;
481 //t[5] = 0 - a32[2] - a32[3] + 2*a32[5] + 2*a32[6] + a32[7] ;
482 //t[6] = 0 - a32[0] - a32[1] + a32[5] + 3*a32[6] + 2*a32[7];
483 //t[7] = 0 + a32[0] - a32[2] - a32[3] - a32[4] - a32[5] + 3*a32[7];
484 // We can do it "piecemeal" after each a32[i] is known, no need to store entire a32[8] vector:
485
486#define A32 (int64_t)a32
487 a32 = a[0] | (a[1] << 26);
488 t[0] = 0 + A32;
489 t[3] = 0 - A32;
490 t[6] = 0 - A32;
491 t[7] = 0 + A32;
492
493 a32 = (a[1] >> 6) | (a[2] << 20);
494 t[0] += A32 ;
495 t[1] = 0 + A32;
496 t[3] -= A32 ;
497 t[4] = 0 - A32;
498 t[6] -= A32 ;
499
500 a32 = (a[2] >> 12) | (a[3] << 14);
501 t[1] += A32 ;
502 t[2] = 0 + A32;
503 t[4] -= A32 ;
504 t[5] = 0 - A32;
505 t[7] -= A32 ;
506
507 a32 = (a[3] >> 18) | (a[4] << 8);
508 t[0] -= A32 ;
509 t[2] += A32 ;
510 t[3] += 2*A32;
511 t[5] -= A32 ;
512 t[7] -= A32 ;
513
514 a32 = (a[4] >> 24) | (a[5] << 2) | (a[6] << 28);
515 t[0] -= A32 ;
516 t[1] -= A32 ;
517 t[3] += 2*A32;
518 t[4] += 2*A32;
519 t[7] -= A32 ;
520
521 a32 = (a[6] >> 4) | (a[7] << 22);
522 t[0] -= A32 ;
523 t[1] -= A32 ;
524 t[2] -= A32 ;
525 t[3] += A32 ;
526 t[4] += 2*A32;
527 t[5] += 2*A32;
528 t[6] += A32 ;
529 t[7] -= A32 ;
530
531 a32 = (a[7] >> 10) | (a[8] << 16);
532 t[0] -= A32 ;
533 t[1] -= A32 ;
534 t[2] -= A32 ;
535 t[4] += A32 ;
536 t[5] += 2*A32;
537 t[6] += 3*A32;
538
539 a32 = (a[8] >> 16) | (a[9] << 10);
540 t[1] -= A32 ;
541 t[2] -= A32 ;
542 t[3] -= A32 ;
543 t[5] += A32 ;
544 t[6] += 2*A32;
545 t[7] += 3*A32;
546#undef A32
508 547
509 t[1] += t[0] >> 32; t[0] &= 0xffffffff; 548 t[1] += t[0] >> 32; t[0] &= 0xffffffff;
510 t[2] += t[1] >> 32; t[1] &= 0xffffffff; 549 t[2] += t[1] >> 32; t[1] &= 0xffffffff;
@@ -526,30 +565,16 @@ static void sp_256_mod_mul_norm_10(sp_digit* r, const sp_digit* a)
526 t[6] += t[5] >> 32; t[5] &= 0xffffffff; 565 t[6] += t[5] >> 32; t[5] &= 0xffffffff;
527 t[7] += t[6] >> 32; t[6] &= 0xffffffff; 566 t[7] += t[6] >> 32; t[6] &= 0xffffffff;
528 567
529 r[0] = (sp_digit)(t[0]) & 0x3ffffff; 568 r[0] = 0x3ffffff & ((sp_digit)(t[0]));
530 r[1] = (sp_digit)(t[0] >> 26); 569 r[1] = 0x3ffffff & ((sp_digit)(t[0] >> 26) | ((sp_digit)t[1] << 6));
531 r[1] |= t[1] << 6; 570 r[2] = 0x3ffffff & ((sp_digit)(t[1] >> 20) | ((sp_digit)t[2] << 12));
532 r[1] &= 0x3ffffff; 571 r[3] = 0x3ffffff & ((sp_digit)(t[2] >> 14) | ((sp_digit)t[3] << 18));
533 r[2] = (sp_digit)(t[1] >> 20); 572 r[4] = 0x3ffffff & ((sp_digit)(t[3] >> 8) | ((sp_digit)t[4] << 24));
534 r[2] |= t[2] << 12; 573 r[5] = 0x3ffffff & ((sp_digit)t[4] >> 2); /* small shift, ok to cast t[4] to narrower type */
535 r[2] &= 0x3ffffff; 574 r[6] = 0x3ffffff & ((sp_digit)(t[4] >> 28) | ((sp_digit)t[5] << 4));
536 r[3] = (sp_digit)(t[2] >> 14); 575 r[7] = 0x3ffffff & ((sp_digit)(t[5] >> 22) | ((sp_digit)t[6] << 10));
537 r[3] |= t[3] << 18; 576 r[8] = 0x3ffffff & ((sp_digit)(t[6] >> 16) | ((sp_digit)t[7] << 16));
538 r[3] &= 0x3ffffff; 577 r[9] = ((sp_digit)(t[7] >> 10));
539 r[4] = (sp_digit)(t[3] >> 8);
540 r[4] |= t[4] << 24;
541 r[4] &= 0x3ffffff;
542 r[5] = (sp_digit)(t[4] >> 2) & 0x3ffffff;
543 r[6] = (sp_digit)(t[4] >> 28);
544 r[6] |= t[5] << 4;
545 r[6] &= 0x3ffffff;
546 r[7] = (sp_digit)(t[5] >> 22);
547 r[7] |= t[6] << 10;
548 r[7] &= 0x3ffffff;
549 r[8] = (sp_digit)(t[6] >> 16);
550 r[8] |= t[7] << 16;
551 r[8] &= 0x3ffffff;
552 r[9] = (sp_digit)(t[7] >> 10);
553} 578}
554 579
555/* Map the Montgomery form projective co-ordinate point to an affine point. 580/* Map the Montgomery form projective co-ordinate point to an affine point.
@@ -795,7 +820,7 @@ static void sp_256_ecc_mulmod_base_10(sp_point* r, sp_digit* k /*, int map*/)
795 0x6b,0x17,0xd1,0xf2,0xe1,0x2c,0x42,0x47,0xf8,0xbc,0xe6,0xe5,0x63,0xa4,0x40,0xf2,0x77,0x03,0x7d,0x81,0x2d,0xeb,0x33,0xa0,0xf4,0xa1,0x39,0x45,0xd8,0x98,0xc2,0x96, 820 0x6b,0x17,0xd1,0xf2,0xe1,0x2c,0x42,0x47,0xf8,0xbc,0xe6,0xe5,0x63,0xa4,0x40,0xf2,0x77,0x03,0x7d,0x81,0x2d,0xeb,0x33,0xa0,0xf4,0xa1,0x39,0x45,0xd8,0x98,0xc2,0x96,
796 /* y */ 821 /* y */
797 0x4f,0xe3,0x42,0xe2,0xfe,0x1a,0x7f,0x9b,0x8e,0xe7,0xeb,0x4a,0x7c,0x0f,0x9e,0x16,0x2b,0xce,0x33,0x57,0x6b,0x31,0x5e,0xce,0xcb,0xb6,0x40,0x68,0x37,0xbf,0x51,0xf5, 822 0x4f,0xe3,0x42,0xe2,0xfe,0x1a,0x7f,0x9b,0x8e,0xe7,0xeb,0x4a,0x7c,0x0f,0x9e,0x16,0x2b,0xce,0x33,0x57,0x6b,0x31,0x5e,0xce,0xcb,0xb6,0x40,0x68,0x37,0xbf,0x51,0xf5,
798 /* z will be set to 0, infinity flag to "false" */ 823 /* z will be set to 1, infinity flag to "false" */
799 }; 824 };
800 sp_point p256_base; 825 sp_point p256_base;
801 826
generated by cgit v1.2.3-55-g6feb (git 2.39.1) at 2025-03-18 05:39:37 +0000