X-Git-Url: https://git.cryptolib.org/?a=blobdiff_plain;f=keccak%2Fkeccak-asm.S;h=95d8aa6289ff484ec781862a4f1d2208e0245f16;hb=7a138263197006d74661fbe446f33d6014777245;hp=6887a755d573fded11f9ec6e73d72a7907c706f9;hpb=6a1c5e57c5aaa60e7f859893866acc7a0c5c6f22;p=avr-crypto-lib.git diff --git a/keccak/keccak-asm.S b/keccak/keccak-asm.S index 6887a75..95d8aa6 100644 --- a/keccak/keccak-asm.S +++ b/keccak/keccak-asm.S @@ -28,6 +28,18 @@ .nolist #include "avr-asm-macros.S" .list + +.equ __zero_reg__, 1 + +.global rho_pi_idx_table +rho_pi_idx_table: + .irp i, 0, 1, 2, 3, 4 + .irp j, 0, 1, 2, 3, 4 + .byte (((2 * \j + 3 * \i) % 5) * 5 + \i) * 8 + .endr + .endr + +.align 2 /* void keccak_theta (uint64_t *a, uint64_t *b){ // uint64_t b[5][5]; @@ -37,18 +49,107 @@ } */ +/********************************************* + * theta_2a + ********************************************* + input: + r24:r25 = a ; uint64_t a[5][5] + X = b ; uint64_t *b + output: + a[0..4][0] ^= b + r20 = 0 + r21 = XX + r22 = XX + r24:r25 += 8 + X += 8 + Z = r24:r25 + 7 + 4 * 40 +*/ +theta_2a: + ldi r20, 8 +10: + movw ZL, r24 + ld r21, X+ + .irp r, 0, 1, 2, 3, 4 + ld r22, Z + eor r22, r21 + st Z, r22 + .if \r != 4 + adiw ZL, 40 + .endif + .endr + adiw r24, 1 + dec r20 + brne 10b + ret + +/********************************************* + * theta_2b + ********************************************* + input: + r24:r25 = a+1 ; uint64_t a[5][5] + X = b ; uint64_t *b + output: + a[0..4][0] ^= rol(b,1) + r19 = XX + r20 = 0 + r21 = XX + r22 = XX + r24:r25 += 8 + X += 8 + Z = r24:r25 + 7 + 4 * 40 +*/ +theta_2b: + ldi r20, 7 + ld r19, X+ + lsl r19 + rol __zero_reg__ +10: + movw ZL, r24 + ld r21, X+ + ror __zero_reg__ + rol r21 + rol __zero_reg__ + .irp r, 0, 1, 2, 3, 4 + ld r22, Z + eor r22, r21 + st Z, r22 + .if \r != 4 + adiw ZL, 40 + .endif + .endr + adiw r24, 1 + dec r20 + brne 10b + add r19, __zero_reg__ + sbiw r24, 8 + movw ZL, r24 + .irp r, 0, 1, 2, 3, 4 + ld r22, Z + eor r22, r19 + st Z, r22 + .if \r != 4 + adiw ZL, 40 + .endif + .endr + adiw r24, 9 + clr __zero_reg__ + ret + + .global keccak_theta keccak_theta: - movw r30, r24 - movw r26, r22 + push_range 2, 8 + push r16 + push_range 28, 29 -; .irp offset, 0, 1, 2, 3, 4 + movw r30, r24 ; Z = a + movw r26, r22 ; X = b + push_range 22, 25 ldi r19, 5 10: ldi r20, 8 20: - ld r22, Z adiw ZL, 40 ld r21, Z @@ -62,10 +163,8 @@ keccak_theta: adiw ZL, 40 ld r21, Z eor r22, r21 - adiw r24, 1 movw r30, r24 - st X+, r22 dec r20 brne 20b @@ -73,6 +172,131 @@ keccak_theta: adiw XL, 8 * 4 dec r19 brne 10b -; .endr +/* + for(i = 0; i < 5; ++i){ + for(j = 0; j < 5; ++j){ + a[j][i] ^= b[(4 + i) % 5][0]; + } + } + +*/ +/* a[0..4][0]{0..7} ^= b[4][0]{0..7} */ + sbiw XL, 5 * 8 + sbiw r24, 40 + rcall theta_2a +/* a[0..4][1]{0..7} ^= b[0][0]{0..7} */ + subi XL, lo8(4 * 5 * 8 + 8) + sbci XH, hi8(4 * 5 * 8 + 8) + rcall theta_2a +/* a[0..4][2]{0..7} ^= b[1][0]{0..7} */ + adiw XL, 4 * 8 + rcall theta_2a +/* a[0..4][3]{0..7} ^= b[2][0]{0..7} */ + adiw XL, 4 * 8 + rcall theta_2a +/* a[0..4][4]{0..7} ^= b[3][0]{0..7} */ + adiw XL, 4 * 8 + rcall theta_2a +/* + for(i = 0; i < 5; ++i){ + for(j = 0; j < 5; ++j){ + a[j][i] ^= rotate64_1bit_left(b[(i + 1) % 5][0]); + } + } +*/ +/* a[0..4][0]{0..7} ^= rol(b[1][0]{0..7}) */ + subi r24, lo8(5 * 8 - 1) + sbci r25, hi8(5 * 8 - 1) + subi XL, lo8(2 * 5 * 8 + 8) + sbci XH, hi8(2 * 5 * 8 + 8) + rcall theta_2b +/* a[0..4][1]{0..7} ^= rol(b[2][0]{0..7}) */ + adiw XL, 4 * 8 + rcall theta_2b +/* a[0..4][21]{0..7} ^= rol(b[3][0]{0..7}) */ + adiw XL, 4 * 8 + rcall theta_2b +/* a[0..4][3]{0..7} ^= rol(b[4][0]{0..7}) */ + adiw XL, 4 * 8 + rcall theta_2b +/* a[0..4][4]{0..7} ^= rol(b[0][0]{0..7}) */ + subi XL, lo8(4 * 5 * 8 + 8) + sbci XH, hi8(4 * 5 * 8 + 8) + rcall theta_2b + +; ret +/* + rho & pi + for(i = 0; i < 5; ++i){ + for(j = 0; j < 5; ++j){ + b[(2 * i + 3 * j) % 5][j] = + rotate64left_code(a[j][i], pgm_read_byte(&(keccak_rotate_codes[i][j]))); + } + } + + -- or -- + + const uint8_t* rot_code = (const uint8_t*)keccak_rotate_codes; + const uint8_t* idx_idx = (const uint8_t*)rho_pi_idx_table; + uint64_t *a_tmp = (uint64_t*)a; + for(i = 0; i < 25; ++i){ + *((uint64_t*)(((uint8_t*)b) + pgm_read_byte(idx_idx++))) = + rotate64left_code(*a_tmp++, pgm_read_byte(rot_code++)); + + } + +*/ + +.equ B_REG_L, 6 +.equ B_REG_H, 7 + + ldi r18, lo8(keccak_rotate_codes) + ldi r19, hi8(keccak_rotate_codes) + movw r2, r18 + ldi r18, lo8(rho_pi_idx_table) + ldi r19, hi8(rho_pi_idx_table) + movw r4, r18 + ldi r16, 25 + mov r8, r16 + + pop YH + pop YL + pop B_REG_H + pop B_REG_L + +10: + ld r18, Y+ + ld r19, Y+ + ld r20, Y+ + ld r21, Y+ + ld r22, Y+ + ld r23, Y+ + ld r24, Y+ + ld r25, Y+ + movw ZL, r2 + lpm r16, Z+ + movw r2, ZL + call rotate64left_code + movw ZL, r4 + lpm r16, Z+ + movw r4, ZL + movw XL, B_REG_L + add XL, r16 + adc XH, __zero_reg__ + st X+, r18 + st X+, r19 + st X+, r20 + st X+, r21 + st X+, r22 + st X+, r23 + st X+, r24 + st X+, r25 + + dec r8 + brne 10b + + pop_range 28, 29 + pop r16 + pop_range 2, 8 ret