.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
+
+.global rotate64_1bit_left
+rotate64_1bit_left:
+ bst r25, 7
+ rol r18
+ rol r19
+ rol r20
+ rol r21
+ rol r22
+ rol r23
+ rol r24
+ rol r25
+ bld r18, 0
+ ret
+
+.global rotate64_1bit_right
+rotate64_1bit_right:
+ bst r18, 0
+ ror r25
+ ror r24
+ ror r23
+ ror r22
+ ror r21
+ ror r20
+ ror r19
+ ror r18
+ bld r25, 7
+ ret
+
+.global rotate64_nbit_autodir
+rotate64_nbit_autodir:
+ lsr r16
+ brcc rotate64_nbit_left
+.global rotate64_nbit_right
+rotate64_nbit_right:
+ ldi r30, pm_lo8(rotate64_1bit_right)
+ ldi r31, pm_hi8(rotate64_1bit_right)
+ rjmp icall_r16_times
+.global rotate64_nbit_left
+rotate64_nbit_left:
+ ldi r30, pm_lo8(rotate64_1bit_left)
+ ldi r31, pm_hi8(rotate64_1bit_left)
+icall_r16_times:
+1: dec r16
+ brmi 2f
+ icall
+ rjmp 1b
+2:
+ ret
+
+rotate64_1byte_left:
+ mov r0, r25
+ mov r25, r24
+ mov r24, r23
+ mov r23, r22
+ mov r22, r21
+ mov r21, r20
+ mov r20, r19
+ mov r19, r18
+ mov r18, r0
+ ret
+
+rotate64_2byte_left:
+ movw r0, r24
+ movw r24, r22
+ movw r22, r20
+ movw r20, r18
+ movw r18, r0
+ ret
+
+rotate64_3byte_left:
+ mov r0, r25
+ mov r25, r22
+ mov r22, r19
+ mov r19, r24
+ mov r24, r21
+ mov r21, r18
+ mov r18, r23
+ mov r23, r20
+ mov r20, r0
+ ret
+
+rotate64_4byte_left:
+ movw r0, r24
+ movw r24, r20
+ movw r20, r0
+ movw r0, r22
+ movw r22, r18
+ movw r18, r0
+ ret
+
+rotate64_5byte_left:
+ mov r0, r25
+ mov r25, r20
+ mov r20, r23
+ mov r23, r18
+ mov r18, r21
+ mov r21, r24
+ mov r24, r19
+ mov r19, r22
+ mov r22, r0
+ ret
+
+rotate64_6byte_left:
+ movw r0, r18
+ movw r18, r20
+ movw r20, r22
+ movw r22, r24
+ movw r24, r0
+ ret
+
+rotate64_7byte_left:
+ mov r0, r18
+ mov r18, r19
+ mov r19, r20
+ mov r20, r21
+ mov r21, r22
+ mov r22, r23
+ mov r23, r24
+ mov r24, r25
+ mov r25, r0
+ ret
+
+
+byte_rot_jmp_table:
+ ret
+ rjmp rotate64_1byte_left
+ rjmp rotate64_2byte_left
+ rjmp rotate64_3byte_left
+ rjmp rotate64_4byte_left
+ rjmp rotate64_5byte_left
+ rjmp rotate64_6byte_left
+ rjmp rotate64_7byte_left
+
+.global rotate64left_code
+rotate64left_code:
+ ldi r30, pm_lo8(byte_rot_jmp_table)
+ ldi r31, pm_hi8(byte_rot_jmp_table)
+ mov r0, r16
+ andi r16, 0x70
+ swap r16
+ add r30, r16
+ adc r31, r1
+ mov r16, r0
+ andi r16, 0x0f
+ icall
+ clr r1
+ rjmp rotate64_nbit_autodir
+
+
/*
void keccak_theta (uint64_t *a, uint64_t *b){
// uint64_t b[5][5];
}
*/
-theta_2a:
-/*
+/*********************************************
+ * theta_2a
+ *********************************************
input:
r24:r25 = a ; uint64_t a[5][5]
X = b ; uint64_t *b
X += 8
Z = r24:r25 + 7 + 4 * 40
*/
+theta_2a:
ldi r20, 8
10:
movw ZL, r24
brne 10b
ret
-.global keccak_theta
-keccak_theta:
+/*********************************************
+ * 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
+
+; a[i][j] = b[i][j] ^ ((~(b[i][(j + 1) % 5])) & (b[i][(j + 2) % 5]));
+
+/*********************************************
+ * chi_step
+ *********************************************
+ input:
+ Y = a; uint8t *a;
+ X = b; uint8t *b;
+ Z = c; uint8t *c;
+ output:
+ a[0..7] ^= ~b[0..7] & c[0..7]
+ X += 8
+ Y += 8
+ Z += 8
+ r16 = 0
+ trash r21, r22, r23
+*/
+chi_step:
+ ldi r16, 8
+10:
+ ld r21, Y
+ ld r22, X+
+ ld r23, Z+
+ com r22
+ and r22, r23
+ eor r21, r22
+ st Y+, r21
+ dec r16
+ brne 10b
+ ret
+
+.global keccak_f1600
+keccak_f1600:
+ push_range 2, 9
+ push r16
+ push_range 28, 29
+
+ stack_alloc_large 200, r26, r27
+ adiw XL, 1
+
+ clr r9
+5:
movw r30, r24 ; Z = a
- movw r26, r22 ; X = b
+
ldi r19, 5
10:
ldi r20, 8
a[j][i] ^= b[(4 + i) % 5][0];
}
}
- 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} ^= b[4][0]{0..7} */
sbiw XL, 5 * 8
-
sbiw r24, 40
-/* a[0..4][0]{0..7} ^= b[4][0]{0..7} */
rcall theta_2a
/* a[0..4][1]{0..7} ^= b[0][0]{0..7} */
subi XL, lo8(4 * 5 * 8 + 8)
/* 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
+; 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])));
+ }
+ }
- ldi r20, 8
+ -- 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
+
+ sbiw r24, 5 * 8 + 1
+ movw YL, r24
+ sbiw XL, 8
+ movw B_REG_L, XL
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
+ 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
+ rcall 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
+/*
+ -- chi --
+ for(i = 0; i < 5; ++i){
+ a[i][0] ^= ((~(b[i][1])) & (b[i][2]));
+ a[i][1] ^= ((~(b[i][2])) & (b[i][3]));
+ a[i][2] ^= ((~(b[i][3])) & (b[i][4]));
+ a[i][3] ^= ((~(b[i][4])) & (b[i][0]));
+ a[i][4] ^= ((~(b[i][0])) & (b[i][1]));
+
+ }
+*/
+ ; memcpy(a, b, 200)
+ ; X points at b + 32 + 8 = b + 40 = b[1][0] has to point to b[0][0]
+ ldi r16, 200
+ sbiw XL, 5 * 8
+ movw ZL, XL
+ subi YL, lo8(5 * 5 * 8)
+ sbci YH, hi8(5 * 5 * 8)
+ movw r2, YL
+10:
+ ld r22, X+
+ st Y+, r22
+ dec r16
+ brne 10b
+
+ ; Z points at b
+ movw XL, ZL
+ movw r4, ZL
+ adiw XL, 8
+ adiw ZL, 16
+ movw YL, r2
+ ldi r18, 5
+10:
+ rcall chi_step
+ rcall chi_step
+ rcall chi_step
+ sbiw ZL, 5 * 8
+ rcall chi_step
+ sbiw XL, 5 * 8
+ rcall chi_step
+ adiw XL, 5 * 8
+ adiw ZL, 5 * 8
+ dec r18
+ brne 10b
+
+ /* -- iota -- */
+ ldi r30, lo8(keccak_rc_comp)
+ ldi r31, hi8(keccak_rc_comp)
+ add r30, r9
+ adc r31, __zero_reg__
+ lpm r20, Z+
+ movw YL, r2
+ ldi r21, 0x80
+ bst r20, 6
+ brtc 10f
+ ldd r22, Y+7
+ eor r22, r21
+ std Y+7, r22
+10:
+ bst r20, 5
+ brtc 10f
+ ldd r22, Y+3
+ eor r22, r21
+ std Y+3, r22
+10:
+ bst r20, 4
+ brtc 10f
+ ldd r22, Y+1
+ eor r22, r21
+ std Y+1, r22
+10:
+ andi r20, 0x8f
+ ld r22, Y
+ eor r22, r20
+ st Y, r22
+
+ inc r9
+ mov r16, r9
+ cpi r16, 24
+ breq 20f
+ movw r24, YL
+ movw r26, r4
+ rjmp 5b
+20:
+
+ stack_free_large3 200
+ pop_range 28, 29
+ pop r16
+ pop_range 2, 9
ret