| 1 |
/* -*- Mode: Asm -*- */ |
| 2 |
/* Copyright (C) 1998-2025 Free Software Foundation, Inc. |
| 3 |
Contributed by Denis Chertykov <[email protected]> |
| 4 |
|
| 5 |
This file is free software; you can redistribute it and/or modify it |
| 6 |
under the terms of the GNU General Public License as published by the |
| 7 |
Free Software Foundation; either version 3, or (at your option) any |
| 8 |
later version. |
| 9 |
|
| 10 |
This file is distributed in the hope that it will be useful, but |
| 11 |
WITHOUT ANY WARRANTY; without even the implied warranty of |
| 12 |
MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU |
| 13 |
General Public License for more details. |
| 14 |
|
| 15 |
Under Section 7 of GPL version 3, you are granted additional |
| 16 |
permissions described in the GCC Runtime Library Exception, version |
| 17 |
3.1, as published by the Free Software Foundation. |
| 18 |
|
| 19 |
You should have received a copy of the GNU General Public License and |
| 20 |
a copy of the GCC Runtime Library Exception along with this program; |
| 21 |
see the files COPYING3 and COPYING.RUNTIME respectively. If not, see |
| 22 |
<http://www.gnu.org/licenses/>. */ |
| 23 |
|
| 24 |
#if defined (__AVR_TINY__) |
| 25 |
#define __zero_reg__ r17 |
| 26 |
#define __tmp_reg__ r16 |
| 27 |
#else |
| 28 |
#define __zero_reg__ r1 |
| 29 |
#define __tmp_reg__ r0 |
| 30 |
#endif |
| 31 |
#define __SREG__ 0x3f |
| 32 |
#if defined (__AVR_HAVE_SPH__) |
| 33 |
#define __SP_H__ 0x3e |
| 34 |
#endif |
| 35 |
#define __SP_L__ 0x3d |
| 36 |
#define __RAMPZ__ 0x3B |
| 37 |
#define __EIND__ 0x3C |
| 38 |
|
| 39 |
/* Most of the functions here are called directly from avr.md |
| 40 |
patterns, instead of using the standard libcall mechanisms. |
| 41 |
This can make better code because GCC knows exactly which |
| 42 |
of the call-used registers (not all of them) are clobbered. */ |
| 43 |
|
| 44 |
/* FIXME: At present, there is no SORT directive in the linker |
| 45 |
script so that we must not assume that different modules |
| 46 |
in the same input section like .libgcc.text.mul will be |
| 47 |
located close together. Therefore, we cannot use |
| 48 |
RCALL/RJMP to call a function like __udivmodhi4 from |
| 49 |
__divmodhi4 and have to use lengthy XCALL/XJMP even |
| 50 |
though they are in the same input section and all same |
| 51 |
input sections together are small enough to reach every |
| 52 |
location with a RCALL/RJMP instruction. */ |
| 53 |
|
| 54 |
#if defined (__AVR_HAVE_EIJMP_EICALL__) && !defined (__AVR_HAVE_ELPMX__) |
| 55 |
#error device not supported |
| 56 |
#endif |
| 57 |
|
| 58 |
.macro mov_l r_dest, r_src |
| 59 |
#if defined (__AVR_HAVE_MOVW__) |
| 60 |
movw \r_dest, \r_src |
| 61 |
#else |
| 62 |
mov \r_dest, \r_src |
| 63 |
#endif |
| 64 |
.endm |
| 65 |
|
| 66 |
.macro mov_h r_dest, r_src |
| 67 |
#if defined (__AVR_HAVE_MOVW__) |
| 68 |
; empty |
| 69 |
#else |
| 70 |
mov \r_dest, \r_src |
| 71 |
#endif |
| 72 |
.endm |
| 73 |
|
| 74 |
.macro wmov r_dest, r_src |
| 75 |
#if defined (__AVR_HAVE_MOVW__) |
| 76 |
movw \r_dest, \r_src |
| 77 |
#else |
| 78 |
mov \r_dest, \r_src |
| 79 |
mov \r_dest+1, \r_src+1 |
| 80 |
#endif |
| 81 |
.endm |
| 82 |
|
| 83 |
.macro mov4 r_dest, r_src |
| 84 |
wmov \r_dest, \r_src |
| 85 |
wmov \r_dest+2, \r_src+2 |
| 86 |
.endm |
| 87 |
|
| 88 |
#if defined (__AVR_HAVE_JMP_CALL__) |
| 89 |
#define XCALL call |
| 90 |
#define XJMP jmp |
| 91 |
#else |
| 92 |
#define XCALL rcall |
| 93 |
#define XJMP rjmp |
| 94 |
#endif |
| 95 |
|
| 96 |
#if defined (__AVR_HAVE_EIJMP_EICALL__) |
| 97 |
#define XICALL eicall |
| 98 |
#define XIJMP eijmp |
| 99 |
#else |
| 100 |
#define XICALL icall |
| 101 |
#define XIJMP ijmp |
| 102 |
#endif |
| 103 |
|
| 104 |
;;; [R]JMP to label \labl when \reg is positive (\reg.7 = 0). |
| 105 |
;;; Otherwise, fallthrough. |
| 106 |
.macro .branch_plus reg, labl |
| 107 |
#ifdef __AVR_ERRATA_SKIP_JMP_CALL__ |
| 108 |
;; Some cores have a problem skipping 2-word instructions |
| 109 |
tst \reg |
| 110 |
brmi .L..\@ |
| 111 |
#else |
| 112 |
sbrs \reg, 7 |
| 113 |
#endif /* skip erratum */ |
| 114 |
XJMP \labl |
| 115 |
.L..\@: |
| 116 |
.endm ; .branch_plus |
| 117 |
|
| 118 |
;; Prologue stuff |
| 119 |
|
| 120 |
.macro do_prologue_saves n_pushed n_frame=0 |
| 121 |
ldi r26, lo8(\n_frame) |
| 122 |
ldi r27, hi8(\n_frame) |
| 123 |
ldi r30, lo8(gs(.L_prologue_saves.\@)) |
| 124 |
ldi r31, hi8(gs(.L_prologue_saves.\@)) |
| 125 |
XJMP __prologue_saves__ + ((18 - (\n_pushed)) * 2) |
| 126 |
.L_prologue_saves.\@: |
| 127 |
.endm |
| 128 |
|
| 129 |
;; Epilogue stuff |
| 130 |
|
| 131 |
.macro do_epilogue_restores n_pushed n_frame=0 |
| 132 |
in r28, __SP_L__ |
| 133 |
#ifdef __AVR_HAVE_SPH__ |
| 134 |
in r29, __SP_H__ |
| 135 |
.if \n_frame > 63 |
| 136 |
subi r28, lo8(-\n_frame) |
| 137 |
sbci r29, hi8(-\n_frame) |
| 138 |
.elseif \n_frame > 0 |
| 139 |
adiw r28, \n_frame |
| 140 |
.endif |
| 141 |
#else |
| 142 |
clr r29 |
| 143 |
.if \n_frame > 0 |
| 144 |
subi r28, lo8(-\n_frame) |
| 145 |
.endif |
| 146 |
#endif /* HAVE SPH */ |
| 147 |
ldi r30, \n_pushed |
| 148 |
XJMP __epilogue_restores__ + ((18 - (\n_pushed)) * 2) |
| 149 |
.endm |
| 150 |
|
| 151 |
;; Support function entry and exit for convenience |
| 152 |
|
| 153 |
.macro wsubi r_arg1, i_arg2 |
| 154 |
#if defined (__AVR_TINY__) |
| 155 |
subi \r_arg1, lo8(\i_arg2) |
| 156 |
sbci \r_arg1+1, hi8(\i_arg2) |
| 157 |
#else |
| 158 |
sbiw \r_arg1, \i_arg2 |
| 159 |
#endif |
| 160 |
.endm |
| 161 |
|
| 162 |
.macro waddi r_arg1, i_arg2 |
| 163 |
#if defined (__AVR_TINY__) |
| 164 |
subi \r_arg1, lo8(-\i_arg2) |
| 165 |
sbci \r_arg1+1, hi8(-\i_arg2) |
| 166 |
#else |
| 167 |
adiw \r_arg1, \i_arg2 |
| 168 |
#endif |
| 169 |
.endm |
| 170 |
|
| 171 |
.macro DEFUN name |
| 172 |
.global \name |
| 173 |
.func \name |
| 174 |
\name: |
| 175 |
.endm |
| 176 |
|
| 177 |
.macro ENDF name |
| 178 |
.size \name, .-\name |
| 179 |
.endfunc |
| 180 |
.endm |
| 181 |
|
| 182 |
.macro FALIAS name |
| 183 |
.global \name |
| 184 |
.func \name |
| 185 |
\name: |
| 186 |
.size \name, .-\name |
| 187 |
.endfunc |
| 188 |
.endm |
| 189 |
|
| 190 |
;; Skip next instruction, typically a jump target |
| 191 |
#define skip cpse 16,16 |
| 192 |
|
| 193 |
;; Negate a 2-byte value held in consecutive registers |
| 194 |
.macro NEG2 reg |
| 195 |
com \reg+1 |
| 196 |
neg \reg |
| 197 |
sbci \reg+1, -1 |
| 198 |
.endm |
| 199 |
|
| 200 |
;; Negate a 4-byte value held in consecutive registers |
| 201 |
;; Sets the V flag for signed overflow tests if REG >= 16 |
| 202 |
.macro NEG4 reg |
| 203 |
com \reg+3 |
| 204 |
com \reg+2 |
| 205 |
com \reg+1 |
| 206 |
.if \reg >= 16 |
| 207 |
neg \reg |
| 208 |
sbci \reg+1, -1 |
| 209 |
sbci \reg+2, -1 |
| 210 |
sbci \reg+3, -1 |
| 211 |
.else |
| 212 |
com \reg |
| 213 |
adc \reg, __zero_reg__ |
| 214 |
adc \reg+1, __zero_reg__ |
| 215 |
adc \reg+2, __zero_reg__ |
| 216 |
adc \reg+3, __zero_reg__ |
| 217 |
.endif |
| 218 |
.endm |
| 219 |
|
| 220 |
#define exp_lo(N) hlo8 ((N) << 23) |
| 221 |
#define exp_hi(N) hhi8 ((N) << 23) |
| 222 |
|
| 223 |
|
| 224 |
.section .text.libgcc.mul, "ax", @progbits |
| 225 |
|
| 226 |
;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;; |
| 227 |
/* Note: mulqi3, mulhi3 are open-coded on the enhanced core. */ |
| 228 |
#if !defined (__AVR_HAVE_MUL__) |
| 229 |
/******************************************************* |
| 230 |
Multiplication 8 x 8 without MUL |
| 231 |
*******************************************************/ |
| 232 |
#if defined (L_mulqi3) |
| 233 |
|
| 234 |
#define r_arg2 r22 /* multiplicand */ |
| 235 |
#define r_arg1 r24 /* multiplier */ |
| 236 |
#define r_res __tmp_reg__ /* result */ |
| 237 |
|
| 238 |
DEFUN __mulqi3 |
| 239 |
clr r_res ; clear result |
| 240 |
__mulqi3_loop: |
| 241 |
sbrc r_arg1,0 |
| 242 |
add r_res,r_arg2 |
| 243 |
add r_arg2,r_arg2 ; shift multiplicand |
| 244 |
breq __mulqi3_exit ; while multiplicand != 0 |
| 245 |
lsr r_arg1 ; |
| 246 |
brne __mulqi3_loop ; exit if multiplier = 0 |
| 247 |
__mulqi3_exit: |
| 248 |
mov r_arg1,r_res ; result to return register |
| 249 |
ret |
| 250 |
ENDF __mulqi3 |
| 251 |
|
| 252 |
#undef r_arg2 |
| 253 |
#undef r_arg1 |
| 254 |
#undef r_res |
| 255 |
|
| 256 |
#endif /* defined (L_mulqi3) */ |
| 257 |
|
| 258 |
|
| 259 |
/******************************************************* |
| 260 |
Widening Multiplication 16 = 8 x 8 without MUL |
| 261 |
Multiplication 16 x 16 without MUL |
| 262 |
*******************************************************/ |
| 263 |
|
| 264 |
#define A0 22 |
| 265 |
#define A1 23 |
| 266 |
#define B0 24 |
| 267 |
#define BB0 20 |
| 268 |
#define B1 25 |
| 269 |
;; Output overlaps input, thus expand result in CC0/1 |
| 270 |
#define C0 24 |
| 271 |
#define C1 25 |
| 272 |
#define CC0 __tmp_reg__ |
| 273 |
#define CC1 21 |
| 274 |
|
| 275 |
#if defined (L_umulqihi3) |
| 276 |
;;; R25:R24 = (unsigned int) R22 * (unsigned int) R24 |
| 277 |
;;; (C1:C0) = (unsigned int) A0 * (unsigned int) B0 |
| 278 |
;;; Clobbers: __tmp_reg__, R21..R23 |
| 279 |
DEFUN __umulqihi3 |
| 280 |
clr A1 |
| 281 |
clr B1 |
| 282 |
XJMP __mulhi3 |
| 283 |
ENDF __umulqihi3 |
| 284 |
#endif /* L_umulqihi3 */ |
| 285 |
|
| 286 |
#if defined (L_mulqihi3) |
| 287 |
;;; R25:R24 = (signed int) R22 * (signed int) R24 |
| 288 |
;;; (C1:C0) = (signed int) A0 * (signed int) B0 |
| 289 |
;;; Clobbers: __tmp_reg__, R20..R23 |
| 290 |
DEFUN __mulqihi3 |
| 291 |
;; Sign-extend B0 |
| 292 |
clr B1 |
| 293 |
sbrc B0, 7 |
| 294 |
com B1 |
| 295 |
;; The multiplication runs twice as fast if A1 is zero, thus: |
| 296 |
;; Zero-extend A0 |
| 297 |
clr A1 |
| 298 |
#ifdef __AVR_HAVE_JMP_CALL__ |
| 299 |
;; Store B0 * sign of A |
| 300 |
clr BB0 |
| 301 |
sbrc A0, 7 |
| 302 |
mov BB0, B0 |
| 303 |
call __mulhi3 |
| 304 |
#else /* have no CALL */ |
| 305 |
;; Skip sign-extension of A if A >= 0 |
| 306 |
;; Same size as with the first alternative but avoids errata skip |
| 307 |
;; and is faster if A >= 0 |
| 308 |
sbrs A0, 7 |
| 309 |
rjmp __mulhi3 |
| 310 |
;; If A < 0 store B |
| 311 |
mov BB0, B0 |
| 312 |
rcall __mulhi3 |
| 313 |
#endif /* HAVE_JMP_CALL */ |
| 314 |
;; 1-extend A after the multiplication |
| 315 |
sub C1, BB0 |
| 316 |
ret |
| 317 |
ENDF __mulqihi3 |
| 318 |
#endif /* L_mulqihi3 */ |
| 319 |
|
| 320 |
#if defined (L_mulhi3) |
| 321 |
;;; R25:R24 = R23:R22 * R25:R24 |
| 322 |
;;; (C1:C0) = (A1:A0) * (B1:B0) |
| 323 |
;;; Clobbers: __tmp_reg__, R21..R23 |
| 324 |
DEFUN __mulhi3 |
| 325 |
|
| 326 |
;; Clear result |
| 327 |
clr CC0 |
| 328 |
clr CC1 |
| 329 |
rjmp 3f |
| 330 |
1: |
| 331 |
;; Bit n of A is 1 --> C += B << n |
| 332 |
add CC0, B0 |
| 333 |
adc CC1, B1 |
| 334 |
2: |
| 335 |
lsl B0 |
| 336 |
rol B1 |
| 337 |
3: |
| 338 |
;; If B == 0 we are ready |
| 339 |
wsubi B0, 0 |
| 340 |
breq 9f |
| 341 |
|
| 342 |
;; Carry = n-th bit of A |
| 343 |
lsr A1 |
| 344 |
ror A0 |
| 345 |
;; If bit n of A is set, then go add B * 2^n to C |
| 346 |
brcs 1b |
| 347 |
|
| 348 |
;; Carry = 0 --> The ROR above acts like CP A0, 0 |
| 349 |
;; Thus, it is sufficient to CPC the high part to test A against 0 |
| 350 |
cpc A1, __zero_reg__ |
| 351 |
;; Only proceed if A != 0 |
| 352 |
brne 2b |
| 353 |
9: |
| 354 |
;; Move Result into place |
| 355 |
mov C0, CC0 |
| 356 |
mov C1, CC1 |
| 357 |
ret |
| 358 |
ENDF __mulhi3 |
| 359 |
#endif /* L_mulhi3 */ |
| 360 |
|
| 361 |
#undef A0 |
| 362 |
#undef A1 |
| 363 |
#undef B0 |
| 364 |
#undef BB0 |
| 365 |
#undef B1 |
| 366 |
#undef C0 |
| 367 |
#undef C1 |
| 368 |
#undef CC0 |
| 369 |
#undef CC1 |
| 370 |
|
| 371 |
|
| 372 |
#define A0 22 |
| 373 |
#define A1 A0+1 |
| 374 |
#define A2 A0+2 |
| 375 |
#define A3 A0+3 |
| 376 |
|
| 377 |
#define B0 18 |
| 378 |
#define B1 B0+1 |
| 379 |
#define B2 B0+2 |
| 380 |
#define B3 B0+3 |
| 381 |
|
| 382 |
#define CC0 26 |
| 383 |
#define CC1 CC0+1 |
| 384 |
#define CC2 30 |
| 385 |
#define CC3 CC2+1 |
| 386 |
|
| 387 |
#define C0 22 |
| 388 |
#define C1 C0+1 |
| 389 |
#define C2 C0+2 |
| 390 |
#define C3 C0+3 |
| 391 |
|
| 392 |
/******************************************************* |
| 393 |
Widening Multiplication 32 = 16 x 16 without MUL |
| 394 |
*******************************************************/ |
| 395 |
|
| 396 |
#if defined (L_umulhisi3) |
| 397 |
DEFUN __umulhisi3 |
| 398 |
#ifdef __AVR_TINY__ |
| 399 |
;; Save callee saved regs. |
| 400 |
push B0 |
| 401 |
push B1 |
| 402 |
#endif /* AVR_TINY */ |
| 403 |
wmov B0, 24 |
| 404 |
;; Zero-extend B |
| 405 |
clr B2 |
| 406 |
clr B3 |
| 407 |
;; Zero-extend A |
| 408 |
wmov A2, B2 |
| 409 |
#ifdef __AVR_TINY__ |
| 410 |
;; Clear hi16 of the result so we can use __mulsi3_helper. |
| 411 |
wmov CC2, B2 |
| 412 |
XJMP __mulsi3_helper |
| 413 |
#else |
| 414 |
XJMP __mulsi3 |
| 415 |
#endif /* AVR_TINY? */ |
| 416 |
ENDF __umulhisi3 |
| 417 |
#endif /* L_umulhisi3 */ |
| 418 |
|
| 419 |
#if defined (L_mulhisi3) |
| 420 |
DEFUN __mulhisi3 |
| 421 |
#ifdef __AVR_TINY__ |
| 422 |
;; Save callee saved regs. |
| 423 |
push B0 |
| 424 |
push B1 |
| 425 |
#endif /* AVR_TINY */ |
| 426 |
wmov B0, 24 |
| 427 |
;; Sign-extend B |
| 428 |
lsl r25 |
| 429 |
sbc B2, B2 |
| 430 |
mov B3, B2 |
| 431 |
;; Zero-extend A and __mulsi3 will run at least twice as fast |
| 432 |
;; compared to a sign-extended A. |
| 433 |
clr A2 |
| 434 |
clr A3 |
| 435 |
;; Clear hi16 of the result so we can use __mulsi3_helper. |
| 436 |
wmov CC2, A2 |
| 437 |
sbrs A1, 7 |
| 438 |
#ifdef __AVR_ERRATA_SKIP_JMP_CALL__ |
| 439 |
rjmp 1f |
| 440 |
#else |
| 441 |
XJMP __mulsi3_helper |
| 442 |
#endif /* ERRATA_SKIP */ |
| 443 |
;; If A < 0 then perform the B * 0xffff.... before the |
| 444 |
;; very multiplication by initializing the high part of the |
| 445 |
;; result CC with -B. |
| 446 |
sub CC2, B0 |
| 447 |
sbc CC3, B1 |
| 448 |
1: XJMP __mulsi3_helper |
| 449 |
ENDF __mulhisi3 |
| 450 |
#endif /* L_mulhisi3 */ |
| 451 |
|
| 452 |
|
| 453 |
/******************************************************* |
| 454 |
Multiplication 32 x 32 without MUL |
| 455 |
*******************************************************/ |
| 456 |
|
| 457 |
#if defined (L_mulsi3) |
| 458 |
DEFUN __mulsi3 |
| 459 |
#if defined (__AVR_TINY__) |
| 460 |
in r26, __SP_L__ ; safe to use X, as it is CC0/CC1 |
| 461 |
in r27, __SP_H__ |
| 462 |
subi r26, lo8(-3) ; Add 3 to point past return address |
| 463 |
sbci r27, hi8(-3) |
| 464 |
push B0 ; save callee saved regs |
| 465 |
push B1 |
| 466 |
ld B0, X+ ; load from caller stack |
| 467 |
ld B1, X+ |
| 468 |
ld B2, X+ |
| 469 |
ld B3, X |
| 470 |
#endif |
| 471 |
;; Clear result |
| 472 |
clr CC2 |
| 473 |
clr CC3 |
| 474 |
;; FALLTHRU |
| 475 |
ENDF __mulsi3 |
| 476 |
|
| 477 |
DEFUN __mulsi3_helper |
| 478 |
clr CC0 |
| 479 |
clr CC1 |
| 480 |
rjmp 3f |
| 481 |
|
| 482 |
1: ;; If bit n of A is set, then add B * 2^n to the result in CC |
| 483 |
;; CC += B |
| 484 |
add CC0,B0 $ adc CC1,B1 $ adc CC2,B2 $ adc CC3,B3 |
| 485 |
|
| 486 |
2: ;; B <<= 1 |
| 487 |
lsl B0 $ rol B1 $ rol B2 $ rol B3 |
| 488 |
|
| 489 |
3: ;; A >>= 1: Carry = n-th bit of A |
| 490 |
lsr A3 $ ror A2 $ ror A1 $ ror A0 |
| 491 |
|
| 492 |
brcs 1b |
| 493 |
;; Only continue if A != 0 |
| 494 |
sbci A1, 0 |
| 495 |
brne 2b |
| 496 |
wsubi A2, 0 |
| 497 |
brne 2b |
| 498 |
|
| 499 |
;; All bits of A are consumed: Copy result to return register C |
| 500 |
wmov C0, CC0 |
| 501 |
wmov C2, CC2 |
| 502 |
#if defined (__AVR_TINY__) |
| 503 |
pop B1 ; restore callee saved regs |
| 504 |
pop B0 |
| 505 |
#endif /* defined (__AVR_TINY__) */ |
| 506 |
|
| 507 |
ret |
| 508 |
ENDF __mulsi3_helper |
| 509 |
#endif /* L_mulsi3 */ |
| 510 |
|
| 511 |
#undef A0 |
| 512 |
#undef A1 |
| 513 |
#undef A2 |
| 514 |
#undef A3 |
| 515 |
#undef B0 |
| 516 |
#undef B1 |
| 517 |
#undef B2 |
| 518 |
#undef B3 |
| 519 |
#undef C0 |
| 520 |
#undef C1 |
| 521 |
#undef C2 |
| 522 |
#undef C3 |
| 523 |
#undef CC0 |
| 524 |
#undef CC1 |
| 525 |
#undef CC2 |
| 526 |
#undef CC3 |
| 527 |
|
| 528 |
#endif /* !defined (__AVR_HAVE_MUL__) */ |
| 529 |
;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;; |
| 530 |
|
| 531 |
;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;; |
| 532 |
#if defined (__AVR_HAVE_MUL__) |
| 533 |
#define A0 26 |
| 534 |
#define B0 18 |
| 535 |
#define C0 22 |
| 536 |
|
| 537 |
#define A1 A0+1 |
| 538 |
|
| 539 |
#define B1 B0+1 |
| 540 |
#define B2 B0+2 |
| 541 |
#define B3 B0+3 |
| 542 |
|
| 543 |
#define C1 C0+1 |
| 544 |
#define C2 C0+2 |
| 545 |
#define C3 C0+3 |
| 546 |
|
| 547 |
/******************************************************* |
| 548 |
Widening Multiplication 32 = 16 x 16 with MUL |
| 549 |
*******************************************************/ |
| 550 |
|
| 551 |
#if defined (L_mulhisi3) |
| 552 |
;;; R25:R22 = (signed long) R27:R26 * (signed long) R19:R18 |
| 553 |
;;; C3:C0 = (signed long) A1:A0 * (signed long) B1:B0 |
| 554 |
;;; Clobbers: __tmp_reg__ |
| 555 |
DEFUN __mulhisi3 |
| 556 |
XCALL __umulhisi3 |
| 557 |
;; Sign-extend B |
| 558 |
tst B1 |
| 559 |
brpl 1f |
| 560 |
sub C2, A0 |
| 561 |
sbc C3, A1 |
| 562 |
1: ;; Sign-extend A |
| 563 |
XJMP __usmulhisi3_tail |
| 564 |
ENDF __mulhisi3 |
| 565 |
#endif /* L_mulhisi3 */ |
| 566 |
|
| 567 |
#if defined (L_usmulhisi3) |
| 568 |
;;; R25:R22 = (signed long) R27:R26 * (unsigned long) R19:R18 |
| 569 |
;;; C3:C0 = (signed long) A1:A0 * (unsigned long) B1:B0 |
| 570 |
;;; Clobbers: __tmp_reg__ |
| 571 |
DEFUN __usmulhisi3 |
| 572 |
XCALL __umulhisi3 |
| 573 |
;; FALLTHRU |
| 574 |
ENDF __usmulhisi3 |
| 575 |
|
| 576 |
DEFUN __usmulhisi3_tail |
| 577 |
;; Sign-extend A |
| 578 |
sbrs A1, 7 |
| 579 |
ret |
| 580 |
sub C2, B0 |
| 581 |
sbc C3, B1 |
| 582 |
ret |
| 583 |
ENDF __usmulhisi3_tail |
| 584 |
#endif /* L_usmulhisi3 */ |
| 585 |
|
| 586 |
#if defined (L_umulhisi3) |
| 587 |
;;; R25:R22 = (unsigned long) R27:R26 * (unsigned long) R19:R18 |
| 588 |
;;; C3:C0 = (unsigned long) A1:A0 * (unsigned long) B1:B0 |
| 589 |
;;; Clobbers: __tmp_reg__ |
| 590 |
DEFUN __umulhisi3 |
| 591 |
mul A0, B0 |
| 592 |
movw C0, r0 |
| 593 |
mul A1, B1 |
| 594 |
movw C2, r0 |
| 595 |
mul A0, B1 |
| 596 |
add C1, r0 |
| 597 |
adc C2, r1 |
| 598 |
clr __zero_reg__ |
| 599 |
adc C3, __zero_reg__ |
| 600 |
mul A1, B0 |
| 601 |
add C1, r0 |
| 602 |
adc C2, r1 |
| 603 |
clr __zero_reg__ |
| 604 |
adc C3, __zero_reg__ |
| 605 |
ret |
| 606 |
ENDF __umulhisi3 |
| 607 |
#endif /* L_umulhisi3 */ |
| 608 |
|
| 609 |
/******************************************************* |
| 610 |
Widening Multiplication 32 = 16 x 32 with MUL |
| 611 |
*******************************************************/ |
| 612 |
|
| 613 |
#if defined (L_mulshisi3) |
| 614 |
;;; R25:R22 = (signed long) R27:R26 * R21:R18 |
| 615 |
;;; (C3:C0) = (signed long) A1:A0 * B3:B0 |
| 616 |
;;; Clobbers: __tmp_reg__ |
| 617 |
DEFUN __mulshisi3 |
| 618 |
#ifdef __AVR_ERRATA_SKIP_JMP_CALL__ |
| 619 |
;; Some cores have problem skipping 2-word instruction |
| 620 |
tst A1 |
| 621 |
brmi __mulohisi3 |
| 622 |
#else |
| 623 |
sbrs A1, 7 |
| 624 |
#endif /* __AVR_HAVE_JMP_CALL__ */ |
| 625 |
XJMP __muluhisi3 |
| 626 |
;; FALLTHRU |
| 627 |
ENDF __mulshisi3 |
| 628 |
|
| 629 |
;;; R25:R22 = (one-extended long) R27:R26 * R21:R18 |
| 630 |
;;; (C3:C0) = (one-extended long) A1:A0 * B3:B0 |
| 631 |
;;; Clobbers: __tmp_reg__ |
| 632 |
DEFUN __mulohisi3 |
| 633 |
XCALL __muluhisi3 |
| 634 |
;; One-extend R27:R26 (A1:A0) |
| 635 |
sub C2, B0 |
| 636 |
sbc C3, B1 |
| 637 |
ret |
| 638 |
ENDF __mulohisi3 |
| 639 |
#endif /* L_mulshisi3 */ |
| 640 |
|
| 641 |
#if defined (L_muluhisi3) |
| 642 |
;;; R25:R22 = (unsigned long) R27:R26 * R21:R18 |
| 643 |
;;; (C3:C0) = (unsigned long) A1:A0 * B3:B0 |
| 644 |
;;; Clobbers: __tmp_reg__ |
| 645 |
DEFUN __muluhisi3 |
| 646 |
XCALL __umulhisi3 |
| 647 |
mul A0, B3 |
| 648 |
add C3, r0 |
| 649 |
mul A1, B2 |
| 650 |
add C3, r0 |
| 651 |
mul A0, B2 |
| 652 |
add C2, r0 |
| 653 |
adc C3, r1 |
| 654 |
clr __zero_reg__ |
| 655 |
ret |
| 656 |
ENDF __muluhisi3 |
| 657 |
#endif /* L_muluhisi3 */ |
| 658 |
|
| 659 |
/******************************************************* |
| 660 |
Multiplication 32 x 32 with MUL |
| 661 |
*******************************************************/ |
| 662 |
|
| 663 |
#if defined (L_mulsi3) |
| 664 |
;;; R25:R22 = R25:R22 * R21:R18 |
| 665 |
;;; (C3:C0) = C3:C0 * B3:B0 |
| 666 |
;;; Clobbers: R26, R27, __tmp_reg__ |
| 667 |
DEFUN __mulsi3 |
| 668 |
movw A0, C0 |
| 669 |
push C2 |
| 670 |
push C3 |
| 671 |
XCALL __muluhisi3 |
| 672 |
pop A1 |
| 673 |
pop A0 |
| 674 |
;; A1:A0 now contains the high word of A |
| 675 |
mul A0, B0 |
| 676 |
add C2, r0 |
| 677 |
adc C3, r1 |
| 678 |
mul A0, B1 |
| 679 |
add C3, r0 |
| 680 |
mul A1, B0 |
| 681 |
add C3, r0 |
| 682 |
clr __zero_reg__ |
| 683 |
ret |
| 684 |
ENDF __mulsi3 |
| 685 |
#endif /* L_mulsi3 */ |
| 686 |
|
| 687 |
#undef A0 |
| 688 |
#undef A1 |
| 689 |
|
| 690 |
#undef B0 |
| 691 |
#undef B1 |
| 692 |
#undef B2 |
| 693 |
#undef B3 |
| 694 |
|
| 695 |
#undef C0 |
| 696 |
#undef C1 |
| 697 |
#undef C2 |
| 698 |
#undef C3 |
| 699 |
|
| 700 |
#endif /* __AVR_HAVE_MUL__ */ |
| 701 |
|
| 702 |
/******************************************************* |
| 703 |
Multiplication 24 x 24 with MUL |
| 704 |
*******************************************************/ |
| 705 |
|
| 706 |
#if defined (L_mulpsi3) |
| 707 |
|
| 708 |
;; A[0..2]: In: Multiplicand; Out: Product |
| 709 |
#define A0 22 |
| 710 |
#define A1 A0+1 |
| 711 |
#define A2 A0+2 |
| 712 |
|
| 713 |
;; B[0..2]: In: Multiplier |
| 714 |
#define B0 18 |
| 715 |
#define B1 B0+1 |
| 716 |
#define B2 B0+2 |
| 717 |
|
| 718 |
#if defined (__AVR_HAVE_MUL__) |
| 719 |
|
| 720 |
;; C[0..2]: Expand Result |
| 721 |
#define C0 22 |
| 722 |
#define C1 C0+1 |
| 723 |
#define C2 C0+2 |
| 724 |
|
| 725 |
;; R24:R22 *= R20:R18 |
| 726 |
;; Clobbers: r21, r25, r26, r27, __tmp_reg__ |
| 727 |
|
| 728 |
#define AA0 26 |
| 729 |
#define AA2 21 |
| 730 |
|
| 731 |
DEFUN __mulpsi3 |
| 732 |
wmov AA0, A0 |
| 733 |
mov AA2, A2 |
| 734 |
XCALL __umulhisi3 |
| 735 |
mul AA2, B0 $ add C2, r0 |
| 736 |
mul AA0, B2 $ add C2, r0 |
| 737 |
clr __zero_reg__ |
| 738 |
ret |
| 739 |
ENDF __mulpsi3 |
| 740 |
|
| 741 |
#undef AA2 |
| 742 |
#undef AA0 |
| 743 |
|
| 744 |
#undef C2 |
| 745 |
#undef C1 |
| 746 |
#undef C0 |
| 747 |
|
| 748 |
#else /* !HAVE_MUL */ |
| 749 |
;; C[0..2]: Expand Result |
| 750 |
#if defined (__AVR_TINY__) |
| 751 |
#define C0 16 |
| 752 |
#else |
| 753 |
#define C0 0 |
| 754 |
#endif /* defined (__AVR_TINY__) */ |
| 755 |
#define C1 C0+1 |
| 756 |
#define C2 21 |
| 757 |
|
| 758 |
;; R24:R22 *= R20:R18 |
| 759 |
;; Clobbers: __tmp_reg__, R18, R19, R20, R21 |
| 760 |
|
| 761 |
DEFUN __mulpsi3 |
| 762 |
#if defined (__AVR_TINY__) |
| 763 |
in r26,__SP_L__ |
| 764 |
in r27,__SP_H__ |
| 765 |
subi r26, lo8(-3) ; Add 3 to point past return address |
| 766 |
sbci r27, hi8(-3) |
| 767 |
push B0 ; save callee saved regs |
| 768 |
push B1 |
| 769 |
ld B0,X+ ; load from caller stack |
| 770 |
ld B1,X+ |
| 771 |
ld B2,X+ |
| 772 |
#endif /* defined (__AVR_TINY__) */ |
| 773 |
|
| 774 |
;; C[] = 0 |
| 775 |
clr __tmp_reg__ |
| 776 |
clr C2 |
| 777 |
|
| 778 |
0: ;; Shift N-th Bit of B[] into Carry. N = 24 - Loop |
| 779 |
LSR B2 $ ror B1 $ ror B0 |
| 780 |
|
| 781 |
;; If the N-th Bit of B[] was set... |
| 782 |
brcc 1f |
| 783 |
|
| 784 |
;; ...then add A[] * 2^N to the Result C[] |
| 785 |
ADD C0,A0 $ adc C1,A1 $ adc C2,A2 |
| 786 |
|
| 787 |
1: ;; Multiply A[] by 2 |
| 788 |
LSL A0 $ rol A1 $ rol A2 |
| 789 |
|
| 790 |
;; Loop until B[] is 0 |
| 791 |
subi B0,0 $ sbci B1,0 $ sbci B2,0 |
| 792 |
brne 0b |
| 793 |
|
| 794 |
;; Copy C[] to the return Register A[] |
| 795 |
wmov A0, C0 |
| 796 |
mov A2, C2 |
| 797 |
|
| 798 |
clr __zero_reg__ |
| 799 |
#if defined (__AVR_TINY__) |
| 800 |
pop B1 |
| 801 |
pop B0 |
| 802 |
#endif /* (__AVR_TINY__) */ |
| 803 |
ret |
| 804 |
ENDF __mulpsi3 |
| 805 |
|
| 806 |
#undef C2 |
| 807 |
#undef C1 |
| 808 |
#undef C0 |
| 809 |
|
| 810 |
#endif /* HAVE_MUL */ |
| 811 |
|
| 812 |
#undef B2 |
| 813 |
#undef B1 |
| 814 |
#undef B0 |
| 815 |
|
| 816 |
#undef A2 |
| 817 |
#undef A1 |
| 818 |
#undef A0 |
| 819 |
|
| 820 |
#endif /* L_mulpsi3 */ |
| 821 |
|
| 822 |
#if defined (L_mulsqipsi3) && defined (__AVR_HAVE_MUL__) |
| 823 |
|
| 824 |
;; A[0..2]: In: Multiplicand |
| 825 |
#define A0 22 |
| 826 |
#define A1 A0+1 |
| 827 |
#define A2 A0+2 |
| 828 |
|
| 829 |
;; BB: In: Multiplier |
| 830 |
#define BB 25 |
| 831 |
|
| 832 |
;; C[0..2]: Result |
| 833 |
#define C0 18 |
| 834 |
#define C1 C0+1 |
| 835 |
#define C2 C0+2 |
| 836 |
|
| 837 |
;; C[] = A[] * sign_extend (BB) |
| 838 |
DEFUN __mulsqipsi3 |
| 839 |
mul A0, BB |
| 840 |
movw C0, r0 |
| 841 |
mul A2, BB |
| 842 |
mov C2, r0 |
| 843 |
mul A1, BB |
| 844 |
add C1, r0 |
| 845 |
adc C2, r1 |
| 846 |
clr __zero_reg__ |
| 847 |
sbrs BB, 7 |
| 848 |
ret |
| 849 |
;; One-extend BB |
| 850 |
sub C1, A0 |
| 851 |
sbc C2, A1 |
| 852 |
ret |
| 853 |
ENDF __mulsqipsi3 |
| 854 |
|
| 855 |
#undef C2 |
| 856 |
#undef C1 |
| 857 |
#undef C0 |
| 858 |
|
| 859 |
#undef BB |
| 860 |
|
| 861 |
#undef A2 |
| 862 |
#undef A1 |
| 863 |
#undef A0 |
| 864 |
|
| 865 |
#endif /* L_mulsqipsi3 && HAVE_MUL */ |
| 866 |
|
| 867 |
/******************************************************* |
| 868 |
Multiplication 64 x 64 |
| 869 |
*******************************************************/ |
| 870 |
|
| 871 |
;; A[] = A[] * B[] |
| 872 |
|
| 873 |
;; A[0..7]: In: Multiplicand |
| 874 |
;; Out: Product |
| 875 |
#define A0 18 |
| 876 |
#define A1 A0+1 |
| 877 |
#define A2 A0+2 |
| 878 |
#define A3 A0+3 |
| 879 |
#define A4 A0+4 |
| 880 |
#define A5 A0+5 |
| 881 |
#define A6 A0+6 |
| 882 |
#define A7 A0+7 |
| 883 |
|
| 884 |
;; B[0..7]: In: Multiplier |
| 885 |
#define B0 10 |
| 886 |
#define B1 B0+1 |
| 887 |
#define B2 B0+2 |
| 888 |
#define B3 B0+3 |
| 889 |
#define B4 B0+4 |
| 890 |
#define B5 B0+5 |
| 891 |
#define B6 B0+6 |
| 892 |
#define B7 B0+7 |
| 893 |
|
| 894 |
#ifndef __AVR_TINY__ |
| 895 |
#if defined (__AVR_HAVE_MUL__) |
| 896 |
;; Define C[] for convenience |
| 897 |
;; Notice that parts of C[] overlap A[] respective B[] |
| 898 |
#define C0 16 |
| 899 |
#define C1 C0+1 |
| 900 |
#define C2 20 |
| 901 |
#define C3 C2+1 |
| 902 |
#define C4 28 |
| 903 |
#define C5 C4+1 |
| 904 |
#define C6 C4+2 |
| 905 |
#define C7 C4+3 |
| 906 |
|
| 907 |
#if defined (L_muldi3) |
| 908 |
|
| 909 |
;; A[] *= B[] |
| 910 |
;; R25:R18 *= R17:R10 |
| 911 |
;; Ordinary ABI-Function |
| 912 |
|
| 913 |
DEFUN __muldi3 |
| 914 |
push r29 |
| 915 |
push r28 |
| 916 |
push r17 |
| 917 |
push r16 |
| 918 |
|
| 919 |
;; Counting in Words, we have to perform a 4 * 4 Multiplication |
| 920 |
|
| 921 |
;; 3 * 0 + 0 * 3 |
| 922 |
mul A7,B0 $ $ mov C7,r0 |
| 923 |
mul A0,B7 $ $ add C7,r0 |
| 924 |
mul A6,B1 $ $ add C7,r0 |
| 925 |
mul A6,B0 $ mov C6,r0 $ add C7,r1 |
| 926 |
mul B6,A1 $ $ add C7,r0 |
| 927 |
mul B6,A0 $ add C6,r0 $ adc C7,r1 |
| 928 |
|
| 929 |
;; 1 * 2 |
| 930 |
mul A2,B4 $ add C6,r0 $ adc C7,r1 |
| 931 |
mul A3,B4 $ $ add C7,r0 |
| 932 |
mul A2,B5 $ $ add C7,r0 |
| 933 |
|
| 934 |
push A5 |
| 935 |
push A4 |
| 936 |
push B1 |
| 937 |
push B0 |
| 938 |
push A3 |
| 939 |
push A2 |
| 940 |
|
| 941 |
;; 0 * 0 |
| 942 |
wmov 26, B0 |
| 943 |
XCALL __umulhisi3 |
| 944 |
wmov C0, 22 |
| 945 |
wmov C2, 24 |
| 946 |
|
| 947 |
;; 0 * 2 |
| 948 |
wmov 26, B4 |
| 949 |
XCALL __umulhisi3 $ wmov C4,22 $ add C6,24 $ adc C7,25 |
| 950 |
|
| 951 |
wmov 26, B2 |
| 952 |
;; 0 * 1 |
| 953 |
XCALL __muldi3_6 |
| 954 |
|
| 955 |
pop A0 |
| 956 |
pop A1 |
| 957 |
;; 1 * 1 |
| 958 |
wmov 26, B2 |
| 959 |
XCALL __umulhisi3 $ add C4,22 $ adc C5,23 $ adc C6,24 $ adc C7,25 |
| 960 |
|
| 961 |
pop r26 |
| 962 |
pop r27 |
| 963 |
;; 1 * 0 |
| 964 |
XCALL __muldi3_6 |
| 965 |
|
| 966 |
pop A0 |
| 967 |
pop A1 |
| 968 |
;; 2 * 0 |
| 969 |
XCALL __umulhisi3 $ add C4,22 $ adc C5,23 $ adc C6,24 $ adc C7,25 |
| 970 |
|
| 971 |
;; 2 * 1 |
| 972 |
wmov 26, B2 |
| 973 |
XCALL __umulhisi3 $ $ $ add C6,22 $ adc C7,23 |
| 974 |
|
| 975 |
;; A[] = C[] |
| 976 |
wmov A0, C0 |
| 977 |
;; A2 = C2 already |
| 978 |
wmov A4, C4 |
| 979 |
wmov A6, C6 |
| 980 |
|
| 981 |
pop r16 |
| 982 |
pop r17 |
| 983 |
pop r28 |
| 984 |
pop r29 |
| 985 |
ret |
| 986 |
ENDF __muldi3 |
| 987 |
#endif /* L_muldi3 */ |
| 988 |
|
| 989 |
#if defined (L_muldi3_6) |
| 990 |
;; A helper for some 64-bit multiplications with MUL available |
| 991 |
DEFUN __muldi3_6 |
| 992 |
__muldi3_6: |
| 993 |
XCALL __umulhisi3 |
| 994 |
add C2, 22 |
| 995 |
adc C3, 23 |
| 996 |
adc C4, 24 |
| 997 |
adc C5, 25 |
| 998 |
brcc 0f |
| 999 |
adiw C6, 1 |
| 1000 |
0: ret |
| 1001 |
ENDF __muldi3_6 |
| 1002 |
#endif /* L_muldi3_6 */ |
| 1003 |
|
| 1004 |
#undef C7 |
| 1005 |
#undef C6 |
| 1006 |
#undef C5 |
| 1007 |
#undef C4 |
| 1008 |
#undef C3 |
| 1009 |
#undef C2 |
| 1010 |
#undef C1 |
| 1011 |
#undef C0 |
| 1012 |
|
| 1013 |
#else /* !HAVE_MUL */ |
| 1014 |
|
| 1015 |
#if defined (L_muldi3) |
| 1016 |
|
| 1017 |
#define C0 26 |
| 1018 |
#define C1 C0+1 |
| 1019 |
#define C2 C0+2 |
| 1020 |
#define C3 C0+3 |
| 1021 |
#define C4 C0+4 |
| 1022 |
#define C5 C0+5 |
| 1023 |
#define C6 0 |
| 1024 |
#define C7 C6+1 |
| 1025 |
|
| 1026 |
#define Loop 9 |
| 1027 |
|
| 1028 |
;; A[] *= B[] |
| 1029 |
;; R25:R18 *= R17:R10 |
| 1030 |
;; Ordinary ABI-Function |
| 1031 |
|
| 1032 |
DEFUN __muldi3 |
| 1033 |
push r29 |
| 1034 |
push r28 |
| 1035 |
push Loop |
| 1036 |
|
| 1037 |
ldi C0, 64 |
| 1038 |
mov Loop, C0 |
| 1039 |
|
| 1040 |
;; C[] = 0 |
| 1041 |
clr __tmp_reg__ |
| 1042 |
wmov C0, 0 |
| 1043 |
wmov C2, 0 |
| 1044 |
wmov C4, 0 |
| 1045 |
|
| 1046 |
0: ;; Rotate B[] right by 1 and set Carry to the N-th Bit of B[] |
| 1047 |
;; where N = 64 - Loop. |
| 1048 |
;; Notice that B[] = B[] >>> 64 so after this Routine has finished, |
| 1049 |
;; B[] will have its initial Value again. |
| 1050 |
LSR B7 $ ror B6 $ ror B5 $ ror B4 |
| 1051 |
ror B3 $ ror B2 $ ror B1 $ ror B0 |
| 1052 |
|
| 1053 |
;; If the N-th Bit of B[] was set then... |
| 1054 |
brcc 1f |
| 1055 |
;; ...finish Rotation... |
| 1056 |
ori B7, 1 << 7 |
| 1057 |
|
| 1058 |
;; ...and add A[] * 2^N to the Result C[] |
| 1059 |
ADD C0,A0 $ adc C1,A1 $ adc C2,A2 $ adc C3,A3 |
| 1060 |
adc C4,A4 $ adc C5,A5 $ adc C6,A6 $ adc C7,A7 |
| 1061 |
|
| 1062 |
1: ;; Multiply A[] by 2 |
| 1063 |
LSL A0 $ rol A1 $ rol A2 $ rol A3 |
| 1064 |
rol A4 $ rol A5 $ rol A6 $ rol A7 |
| 1065 |
|
| 1066 |
dec Loop |
| 1067 |
brne 0b |
| 1068 |
|
| 1069 |
;; We expanded the Result in C[] |
| 1070 |
;; Copy Result to the Return Register A[] |
| 1071 |
wmov A0, C0 |
| 1072 |
wmov A2, C2 |
| 1073 |
wmov A4, C4 |
| 1074 |
wmov A6, C6 |
| 1075 |
|
| 1076 |
clr __zero_reg__ |
| 1077 |
pop Loop |
| 1078 |
pop r28 |
| 1079 |
pop r29 |
| 1080 |
ret |
| 1081 |
ENDF __muldi3 |
| 1082 |
|
| 1083 |
#undef Loop |
| 1084 |
|
| 1085 |
#undef C7 |
| 1086 |
#undef C6 |
| 1087 |
#undef C5 |
| 1088 |
#undef C4 |
| 1089 |
#undef C3 |
| 1090 |
#undef C2 |
| 1091 |
#undef C1 |
| 1092 |
#undef C0 |
| 1093 |
|
| 1094 |
#endif /* L_muldi3 */ |
| 1095 |
#endif /* HAVE_MUL */ |
| 1096 |
#endif /* if not __AVR_TINY__ */ |
| 1097 |
|
| 1098 |
#undef B7 |
| 1099 |
#undef B6 |
| 1100 |
#undef B5 |
| 1101 |
#undef B4 |
| 1102 |
#undef B3 |
| 1103 |
#undef B2 |
| 1104 |
#undef B1 |
| 1105 |
#undef B0 |
| 1106 |
|
| 1107 |
#undef A7 |
| 1108 |
#undef A6 |
| 1109 |
#undef A5 |
| 1110 |
#undef A4 |
| 1111 |
#undef A3 |
| 1112 |
#undef A2 |
| 1113 |
#undef A1 |
| 1114 |
#undef A0 |
| 1115 |
|
| 1116 |
/******************************************************* |
| 1117 |
Widening Multiplication 64 = 32 x 32 with MUL |
| 1118 |
*******************************************************/ |
| 1119 |
|
| 1120 |
#if defined (__AVR_HAVE_MUL__) |
| 1121 |
#define A0 r22 |
| 1122 |
#define A1 r23 |
| 1123 |
#define A2 r24 |
| 1124 |
#define A3 r25 |
| 1125 |
|
| 1126 |
#define B0 r18 |
| 1127 |
#define B1 r19 |
| 1128 |
#define B2 r20 |
| 1129 |
#define B3 r21 |
| 1130 |
|
| 1131 |
#define C0 18 |
| 1132 |
#define C1 C0+1 |
| 1133 |
#define C2 20 |
| 1134 |
#define C3 C2+1 |
| 1135 |
#define C4 28 |
| 1136 |
#define C5 C4+1 |
| 1137 |
#define C6 C4+2 |
| 1138 |
#define C7 C4+3 |
| 1139 |
|
| 1140 |
#if defined (L_umulsidi3) |
| 1141 |
|
| 1142 |
;; Unsigned widening 64 = 32 * 32 Multiplication with MUL |
| 1143 |
|
| 1144 |
;; R18[8] = R22[4] * R18[4] |
| 1145 |
;; |
| 1146 |
;; Ordinary ABI Function, but additionally sets |
| 1147 |
;; X = R20[2] = B2[2] |
| 1148 |
;; Z = R22[2] = A0[2] |
| 1149 |
DEFUN __umulsidi3 |
| 1150 |
clt |
| 1151 |
;; FALLTHRU |
| 1152 |
ENDF __umulsidi3 |
| 1153 |
;; T = sign (A) |
| 1154 |
DEFUN __umulsidi3_helper |
| 1155 |
push 29 $ push 28 ; Y |
| 1156 |
wmov 30, A2 |
| 1157 |
;; Counting in Words, we have to perform 4 Multiplications |
| 1158 |
;; 0 * 0 |
| 1159 |
wmov 26, A0 |
| 1160 |
XCALL __umulhisi3 |
| 1161 |
push 23 $ push 22 ; C0 |
| 1162 |
wmov 28, B0 |
| 1163 |
wmov 18, B2 |
| 1164 |
wmov C2, 24 |
| 1165 |
push 27 $ push 26 ; A0 |
| 1166 |
push 19 $ push 18 ; B2 |
| 1167 |
;; |
| 1168 |
;; 18 20 22 24 26 28 30 | B2, B3, A0, A1, C0, C1, Y |
| 1169 |
;; B2 C2 -- -- -- B0 A2 |
| 1170 |
;; 1 * 1 |
| 1171 |
wmov 26, 30 ; A2 |
| 1172 |
XCALL __umulhisi3 |
| 1173 |
;; Sign-extend A. T holds the sign of A |
| 1174 |
brtc 0f |
| 1175 |
;; Subtract B from the high part of the result |
| 1176 |
sub 22, 28 |
| 1177 |
sbc 23, 29 |
| 1178 |
sbc 24, 18 |
| 1179 |
sbc 25, 19 |
| 1180 |
0: wmov 18, 28 ;; B0 |
| 1181 |
wmov C4, 22 |
| 1182 |
wmov C6, 24 |
| 1183 |
;; |
| 1184 |
;; 18 20 22 24 26 28 30 | B2, B3, A0, A1, C0, C1, Y |
| 1185 |
;; B0 C2 -- -- A2 C4 C6 |
| 1186 |
;; |
| 1187 |
;; 1 * 0 |
| 1188 |
XCALL __muldi3_6 |
| 1189 |
;; 0 * 1 |
| 1190 |
pop 26 $ pop 27 ;; B2 |
| 1191 |
pop 18 $ pop 19 ;; A0 |
| 1192 |
XCALL __muldi3_6 |
| 1193 |
|
| 1194 |
;; Move result C into place and save A0 in Z |
| 1195 |
wmov 22, C4 |
| 1196 |
wmov 24, C6 |
| 1197 |
wmov 30, 18 ; A0 |
| 1198 |
pop C0 $ pop C1 |
| 1199 |
|
| 1200 |
;; Epilogue |
| 1201 |
pop 28 $ pop 29 ;; Y |
| 1202 |
ret |
| 1203 |
ENDF __umulsidi3_helper |
| 1204 |
#endif /* L_umulsidi3 */ |
| 1205 |
|
| 1206 |
|
| 1207 |
#if defined (L_mulsidi3) |
| 1208 |
|
| 1209 |
;; Signed widening 64 = 32 * 32 Multiplication |
| 1210 |
;; |
| 1211 |
;; R18[8] = R22[4] * R18[4] |
| 1212 |
;; Ordinary ABI Function |
| 1213 |
DEFUN __mulsidi3 |
| 1214 |
bst A3, 7 |
| 1215 |
sbrs B3, 7 ; Enhanced core has no skip bug |
| 1216 |
XJMP __umulsidi3_helper |
| 1217 |
|
| 1218 |
;; B needs sign-extension |
| 1219 |
push A3 |
| 1220 |
push A2 |
| 1221 |
XCALL __umulsidi3_helper |
| 1222 |
;; A0 survived in Z |
| 1223 |
sub r22, r30 |
| 1224 |
sbc r23, r31 |
| 1225 |
pop r26 |
| 1226 |
pop r27 |
| 1227 |
sbc r24, r26 |
| 1228 |
sbc r25, r27 |
| 1229 |
ret |
| 1230 |
ENDF __mulsidi3 |
| 1231 |
#endif /* L_mulsidi3 */ |
| 1232 |
|
| 1233 |
#undef A0 |
| 1234 |
#undef A1 |
| 1235 |
#undef A2 |
| 1236 |
#undef A3 |
| 1237 |
#undef B0 |
| 1238 |
#undef B1 |
| 1239 |
#undef B2 |
| 1240 |
#undef B3 |
| 1241 |
#undef C0 |
| 1242 |
#undef C1 |
| 1243 |
#undef C2 |
| 1244 |
#undef C3 |
| 1245 |
#undef C4 |
| 1246 |
#undef C5 |
| 1247 |
#undef C6 |
| 1248 |
#undef C7 |
| 1249 |
#endif /* HAVE_MUL */ |
| 1250 |
|
| 1251 |
/********************************************************** |
| 1252 |
Widening Multiplication 64 = 32 x 32 without MUL |
| 1253 |
**********************************************************/ |
| 1254 |
#ifndef __AVR_TINY__ /* if not __AVR_TINY__ */ |
| 1255 |
#if defined (L_mulsidi3) && !defined (__AVR_HAVE_MUL__) |
| 1256 |
#define A0 18 |
| 1257 |
#define A1 A0+1 |
| 1258 |
#define A2 A0+2 |
| 1259 |
#define A3 A0+3 |
| 1260 |
#define A4 A0+4 |
| 1261 |
#define A5 A0+5 |
| 1262 |
#define A6 A0+6 |
| 1263 |
#define A7 A0+7 |
| 1264 |
|
| 1265 |
#define B0 10 |
| 1266 |
#define B1 B0+1 |
| 1267 |
#define B2 B0+2 |
| 1268 |
#define B3 B0+3 |
| 1269 |
#define B4 B0+4 |
| 1270 |
#define B5 B0+5 |
| 1271 |
#define B6 B0+6 |
| 1272 |
#define B7 B0+7 |
| 1273 |
|
| 1274 |
#define AA0 22 |
| 1275 |
#define AA1 AA0+1 |
| 1276 |
#define AA2 AA0+2 |
| 1277 |
#define AA3 AA0+3 |
| 1278 |
|
| 1279 |
#define BB0 18 |
| 1280 |
#define BB1 BB0+1 |
| 1281 |
#define BB2 BB0+2 |
| 1282 |
#define BB3 BB0+3 |
| 1283 |
|
| 1284 |
#define Mask r30 |
| 1285 |
|
| 1286 |
;; Signed / Unsigned widening 64 = 32 * 32 Multiplication without MUL |
| 1287 |
;; |
| 1288 |
;; R18[8] = R22[4] * R18[4] |
| 1289 |
;; Ordinary ABI Function |
| 1290 |
DEFUN __mulsidi3 |
| 1291 |
set |
| 1292 |
skip |
| 1293 |
;; FALLTHRU |
| 1294 |
ENDF __mulsidi3 |
| 1295 |
|
| 1296 |
DEFUN __umulsidi3 |
| 1297 |
clt ; skipped |
| 1298 |
;; Save 10 Registers: R10..R17, R28, R29 |
| 1299 |
do_prologue_saves 10 |
| 1300 |
ldi Mask, 0xff |
| 1301 |
bld Mask, 7 |
| 1302 |
;; Move B into place... |
| 1303 |
wmov B0, BB0 |
| 1304 |
wmov B2, BB2 |
| 1305 |
;; ...and extend it |
| 1306 |
and BB3, Mask |
| 1307 |
lsl BB3 |
| 1308 |
sbc B4, B4 |
| 1309 |
mov B5, B4 |
| 1310 |
wmov B6, B4 |
| 1311 |
;; Move A into place... |
| 1312 |
wmov A0, AA0 |
| 1313 |
wmov A2, AA2 |
| 1314 |
;; ...and extend it |
| 1315 |
and AA3, Mask |
| 1316 |
lsl AA3 |
| 1317 |
sbc A4, A4 |
| 1318 |
mov A5, A4 |
| 1319 |
wmov A6, A4 |
| 1320 |
XCALL __muldi3 |
| 1321 |
do_epilogue_restores 10 |
| 1322 |
ENDF __umulsidi3 |
| 1323 |
|
| 1324 |
#undef A0 |
| 1325 |
#undef A1 |
| 1326 |
#undef A2 |
| 1327 |
#undef A3 |
| 1328 |
#undef A4 |
| 1329 |
#undef A5 |
| 1330 |
#undef A6 |
| 1331 |
#undef A7 |
| 1332 |
#undef B0 |
| 1333 |
#undef B1 |
| 1334 |
#undef B2 |
| 1335 |
#undef B3 |
| 1336 |
#undef B4 |
| 1337 |
#undef B5 |
| 1338 |
#undef B6 |
| 1339 |
#undef B7 |
| 1340 |
#undef AA0 |
| 1341 |
#undef AA1 |
| 1342 |
#undef AA2 |
| 1343 |
#undef AA3 |
| 1344 |
#undef BB0 |
| 1345 |
#undef BB1 |
| 1346 |
#undef BB2 |
| 1347 |
#undef BB3 |
| 1348 |
#undef Mask |
| 1349 |
#endif /* L_mulsidi3 && !HAVE_MUL */ |
| 1350 |
#endif /* if not __AVR_TINY__ */ |
| 1351 |
;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;; |
| 1352 |
|
| 1353 |
|
| 1354 |
.section .text.libgcc.div, "ax", @progbits |
| 1355 |
|
| 1356 |
/******************************************************* |
| 1357 |
Division 8 / 8 => (result + remainder) |
| 1358 |
*******************************************************/ |
| 1359 |
#define r_rem r25 /* remainder */ |
| 1360 |
#define r_arg1 r24 /* dividend, quotient */ |
| 1361 |
#define r_arg2 r22 /* divisor */ |
| 1362 |
#define r_cnt r23 /* loop count */ |
| 1363 |
|
| 1364 |
#if defined (L_udivmodqi4) |
| 1365 |
DEFUN __udivmodqi4 |
| 1366 |
clr r_rem ; clear remainder |
| 1367 |
ldi r_cnt,8 ; init loop counter |
| 1368 |
lsl r_arg1 ; shift dividend |
| 1369 |
__udivmodqi4_loop: |
| 1370 |
rol r_rem ; shift dividend into remainder |
| 1371 |
cp r_rem,r_arg2 ; compare remainder & divisor |
| 1372 |
brcs __udivmodqi4_ep ; remainder <= divisor |
| 1373 |
sub r_rem,r_arg2 ; restore remainder |
| 1374 |
__udivmodqi4_ep: |
| 1375 |
rol r_arg1 ; shift dividend (with CARRY) |
| 1376 |
dec r_cnt ; decrement loop counter |
| 1377 |
brne __udivmodqi4_loop |
| 1378 |
com r_arg1 ; complement result |
| 1379 |
; because C flag was complemented in loop |
| 1380 |
ret |
| 1381 |
ENDF __udivmodqi4 |
| 1382 |
#endif /* defined (L_udivmodqi4) */ |
| 1383 |
|
| 1384 |
#if defined (L_divmodqi4) |
| 1385 |
DEFUN __divmodqi4 |
| 1386 |
bst r_arg1,7 ; store sign of dividend |
| 1387 |
mov __tmp_reg__,r_arg1 |
| 1388 |
eor __tmp_reg__,r_arg2; r0.7 is sign of result |
| 1389 |
sbrc r_arg1,7 |
| 1390 |
neg r_arg1 ; dividend negative : negate |
| 1391 |
sbrc r_arg2,7 |
| 1392 |
neg r_arg2 ; divisor negative : negate |
| 1393 |
XCALL __udivmodqi4 ; do the unsigned div/mod |
| 1394 |
brtc __divmodqi4_1 |
| 1395 |
neg r_rem ; correct remainder sign |
| 1396 |
__divmodqi4_1: |
| 1397 |
sbrc __tmp_reg__,7 |
| 1398 |
neg r_arg1 ; correct result sign |
| 1399 |
__divmodqi4_exit: |
| 1400 |
ret |
| 1401 |
ENDF __divmodqi4 |
| 1402 |
#endif /* defined (L_divmodqi4) */ |
| 1403 |
|
| 1404 |
#undef r_rem |
| 1405 |
#undef r_arg1 |
| 1406 |
#undef r_arg2 |
| 1407 |
#undef r_cnt |
| 1408 |
|
| 1409 |
|
| 1410 |
/******************************************************* |
| 1411 |
Division 16 / 16 => (result + remainder) |
| 1412 |
*******************************************************/ |
| 1413 |
#define r_remL r26 /* remainder Low */ |
| 1414 |
#define r_remH r27 /* remainder High */ |
| 1415 |
|
| 1416 |
/* return: remainder */ |
| 1417 |
#define r_arg1L r24 /* dividend Low */ |
| 1418 |
#define r_arg1H r25 /* dividend High */ |
| 1419 |
|
| 1420 |
/* return: quotient */ |
| 1421 |
#define r_arg2L r22 /* divisor Low */ |
| 1422 |
#define r_arg2H r23 /* divisor High */ |
| 1423 |
|
| 1424 |
#define r_cnt r21 /* loop count */ |
| 1425 |
|
| 1426 |
#if defined (L_udivmodhi4) |
| 1427 |
DEFUN __udivmodhi4 |
| 1428 |
sub r_remL,r_remL |
| 1429 |
sub r_remH,r_remH ; clear remainder and carry |
| 1430 |
ldi r_cnt,17 ; init loop counter |
| 1431 |
rjmp __udivmodhi4_ep ; jump to entry point |
| 1432 |
__udivmodhi4_loop: |
| 1433 |
rol r_remL ; shift dividend into remainder |
| 1434 |
rol r_remH |
| 1435 |
cp r_remL,r_arg2L ; compare remainder & divisor |
| 1436 |
cpc r_remH,r_arg2H |
| 1437 |
brcs __udivmodhi4_ep ; remainder < divisor |
| 1438 |
sub r_remL,r_arg2L ; restore remainder |
| 1439 |
sbc r_remH,r_arg2H |
| 1440 |
__udivmodhi4_ep: |
| 1441 |
rol r_arg1L ; shift dividend (with CARRY) |
| 1442 |
rol r_arg1H |
| 1443 |
dec r_cnt ; decrement loop counter |
| 1444 |
brne __udivmodhi4_loop |
| 1445 |
com r_arg1L |
| 1446 |
com r_arg1H |
| 1447 |
; div/mod results to return registers, as for the div() function |
| 1448 |
mov_l r_arg2L, r_arg1L ; quotient |
| 1449 |
mov_h r_arg2H, r_arg1H |
| 1450 |
mov_l r_arg1L, r_remL ; remainder |
| 1451 |
mov_h r_arg1H, r_remH |
| 1452 |
ret |
| 1453 |
ENDF __udivmodhi4 |
| 1454 |
#endif /* defined (L_udivmodhi4) */ |
| 1455 |
|
| 1456 |
#if defined (L_divmodhi4) |
| 1457 |
DEFUN __divmodhi4 |
| 1458 |
.global _div |
| 1459 |
_div: |
| 1460 |
bst r_arg1H,7 ; store sign of dividend |
| 1461 |
mov __tmp_reg__,r_arg2H |
| 1462 |
brtc 0f |
| 1463 |
com __tmp_reg__ ; r0.7 is sign of result |
| 1464 |
rcall __divmodhi4_neg1 ; dividend negative: negate |
| 1465 |
0: |
| 1466 |
sbrc r_arg2H,7 |
| 1467 |
rcall __divmodhi4_neg2 ; divisor negative: negate |
| 1468 |
XCALL __udivmodhi4 ; do the unsigned div/mod |
| 1469 |
sbrc __tmp_reg__,7 |
| 1470 |
rcall __divmodhi4_neg2 ; correct remainder sign |
| 1471 |
brtc __divmodhi4_exit |
| 1472 |
__divmodhi4_neg1: |
| 1473 |
;; correct dividend/remainder sign |
| 1474 |
com r_arg1H |
| 1475 |
neg r_arg1L |
| 1476 |
sbci r_arg1H,0xff |
| 1477 |
ret |
| 1478 |
__divmodhi4_neg2: |
| 1479 |
;; correct divisor/result sign |
| 1480 |
com r_arg2H |
| 1481 |
neg r_arg2L |
| 1482 |
sbci r_arg2H,0xff |
| 1483 |
__divmodhi4_exit: |
| 1484 |
ret |
| 1485 |
ENDF __divmodhi4 |
| 1486 |
#endif /* defined (L_divmodhi4) */ |
| 1487 |
|
| 1488 |
#undef r_remH |
| 1489 |
#undef r_remL |
| 1490 |
|
| 1491 |
#undef r_arg1H |
| 1492 |
#undef r_arg1L |
| 1493 |
|
| 1494 |
#undef r_arg2H |
| 1495 |
#undef r_arg2L |
| 1496 |
|
| 1497 |
#undef r_cnt |
| 1498 |
|
| 1499 |
/******************************************************* |
| 1500 |
Division 24 / 24 => (result + remainder) |
| 1501 |
*******************************************************/ |
| 1502 |
|
| 1503 |
;; A[0..2]: In: Dividend; Out: Quotient |
| 1504 |
#define A0 22 |
| 1505 |
#define A1 A0+1 |
| 1506 |
#define A2 A0+2 |
| 1507 |
|
| 1508 |
;; B[0..2]: In: Divisor; Out: Remainder |
| 1509 |
#define B0 18 |
| 1510 |
#define B1 B0+1 |
| 1511 |
#define B2 B0+2 |
| 1512 |
|
| 1513 |
;; C[0..2]: Expand remainder |
| 1514 |
#define C0 __zero_reg__ |
| 1515 |
#define C1 26 |
| 1516 |
#define C2 25 |
| 1517 |
|
| 1518 |
;; Loop counter |
| 1519 |
#define r_cnt 21 |
| 1520 |
|
| 1521 |
#if defined (L_udivmodpsi4) |
| 1522 |
;; R24:R22 = R24:R24 udiv R20:R18 |
| 1523 |
;; R20:R18 = R24:R22 umod R20:R18 |
| 1524 |
;; Clobbers: R21, R25, R26 |
| 1525 |
|
| 1526 |
DEFUN __udivmodpsi4 |
| 1527 |
; init loop counter |
| 1528 |
ldi r_cnt, 24+1 |
| 1529 |
; Clear remainder and carry. C0 is already 0 |
| 1530 |
clr C1 |
| 1531 |
sub C2, C2 |
| 1532 |
; jump to entry point |
| 1533 |
rjmp __udivmodpsi4_start |
| 1534 |
__udivmodpsi4_loop: |
| 1535 |
; shift dividend into remainder |
| 1536 |
rol C0 |
| 1537 |
rol C1 |
| 1538 |
rol C2 |
| 1539 |
; compare remainder & divisor |
| 1540 |
cp C0, B0 |
| 1541 |
cpc C1, B1 |
| 1542 |
cpc C2, B2 |
| 1543 |
brcs __udivmodpsi4_start ; remainder <= divisor |
| 1544 |
sub C0, B0 ; restore remainder |
| 1545 |
sbc C1, B1 |
| 1546 |
sbc C2, B2 |
| 1547 |
__udivmodpsi4_start: |
| 1548 |
; shift dividend (with CARRY) |
| 1549 |
rol A0 |
| 1550 |
rol A1 |
| 1551 |
rol A2 |
| 1552 |
; decrement loop counter |
| 1553 |
dec r_cnt |
| 1554 |
brne __udivmodpsi4_loop |
| 1555 |
com A0 |
| 1556 |
com A1 |
| 1557 |
com A2 |
| 1558 |
; div/mod results to return registers |
| 1559 |
; remainder |
| 1560 |
mov B0, C0 |
| 1561 |
mov B1, C1 |
| 1562 |
mov B2, C2 |
| 1563 |
clr __zero_reg__ ; C0 |
| 1564 |
ret |
| 1565 |
ENDF __udivmodpsi4 |
| 1566 |
#endif /* defined (L_udivmodpsi4) */ |
| 1567 |
|
| 1568 |
#if defined (L_divmodpsi4) |
| 1569 |
;; R24:R22 = R24:R22 div R20:R18 |
| 1570 |
;; R20:R18 = R24:R22 mod R20:R18 |
| 1571 |
;; Clobbers: T, __tmp_reg__, R21, R25, R26 |
| 1572 |
|
| 1573 |
DEFUN __divmodpsi4 |
| 1574 |
; R0.7 will contain the sign of the result: |
| 1575 |
; R0.7 = A.sign ^ B.sign |
| 1576 |
mov __tmp_reg__, B2 |
| 1577 |
; T-flag = sign of dividend |
| 1578 |
bst A2, 7 |
| 1579 |
brtc 0f |
| 1580 |
com __tmp_reg__ |
| 1581 |
; Adjust dividend's sign |
| 1582 |
rcall __divmodpsi4_negA |
| 1583 |
0: |
| 1584 |
; Adjust divisor's sign |
| 1585 |
sbrc B2, 7 |
| 1586 |
rcall __divmodpsi4_negB |
| 1587 |
|
| 1588 |
; Do the unsigned div/mod |
| 1589 |
XCALL __udivmodpsi4 |
| 1590 |
|
| 1591 |
; Adjust quotient's sign |
| 1592 |
sbrc __tmp_reg__, 7 |
| 1593 |
rcall __divmodpsi4_negA |
| 1594 |
|
| 1595 |
; Adjust remainder's sign |
| 1596 |
brtc __divmodpsi4_end |
| 1597 |
|
| 1598 |
__divmodpsi4_negB: |
| 1599 |
; Correct divisor/remainder sign |
| 1600 |
com B2 |
| 1601 |
com B1 |
| 1602 |
neg B0 |
| 1603 |
sbci B1, -1 |
| 1604 |
sbci B2, -1 |
| 1605 |
ret |
| 1606 |
|
| 1607 |
; Correct dividend/quotient sign |
| 1608 |
__divmodpsi4_negA: |
| 1609 |
com A2 |
| 1610 |
com A1 |
| 1611 |
neg A0 |
| 1612 |
sbci A1, -1 |
| 1613 |
sbci A2, -1 |
| 1614 |
__divmodpsi4_end: |
| 1615 |
ret |
| 1616 |
|
| 1617 |
ENDF __divmodpsi4 |
| 1618 |
#endif /* defined (L_divmodpsi4) */ |
| 1619 |
|
| 1620 |
#undef A0 |
| 1621 |
#undef A1 |
| 1622 |
#undef A2 |
| 1623 |
|
| 1624 |
#undef B0 |
| 1625 |
#undef B1 |
| 1626 |
#undef B2 |
| 1627 |
|
| 1628 |
#undef C0 |
| 1629 |
#undef C1 |
| 1630 |
#undef C2 |
| 1631 |
|
| 1632 |
#undef r_cnt |
| 1633 |
|
| 1634 |
/******************************************************* |
| 1635 |
Division 32 / 32 => (result + remainder) |
| 1636 |
*******************************************************/ |
| 1637 |
#define r_remHH r31 /* remainder High */ |
| 1638 |
#define r_remHL r30 |
| 1639 |
#define r_remH r27 |
| 1640 |
#define r_remL r26 /* remainder Low */ |
| 1641 |
|
| 1642 |
/* return: remainder */ |
| 1643 |
#define r_arg1HH r25 /* dividend High */ |
| 1644 |
#define r_arg1HL r24 |
| 1645 |
#define r_arg1H r23 |
| 1646 |
#define r_arg1L r22 /* dividend Low */ |
| 1647 |
|
| 1648 |
/* return: quotient */ |
| 1649 |
#define r_arg2HH r21 /* divisor High */ |
| 1650 |
#define r_arg2HL r20 |
| 1651 |
#define r_arg2H r19 |
| 1652 |
#define r_arg2L r18 /* divisor Low */ |
| 1653 |
|
| 1654 |
#define r_cnt __zero_reg__ /* loop count (0 after the loop!) */ |
| 1655 |
|
| 1656 |
#if defined (L_udivmodsi4) |
| 1657 |
DEFUN __udivmodsi4 |
| 1658 |
ldi r_remL, 33 ; init loop counter |
| 1659 |
mov r_cnt, r_remL |
| 1660 |
sub r_remL,r_remL |
| 1661 |
sub r_remH,r_remH ; clear remainder and carry |
| 1662 |
mov_l r_remHL, r_remL |
| 1663 |
mov_h r_remHH, r_remH |
| 1664 |
rjmp __udivmodsi4_ep ; jump to entry point |
| 1665 |
__udivmodsi4_loop: |
| 1666 |
rol r_remL ; shift dividend into remainder |
| 1667 |
rol r_remH |
| 1668 |
rol r_remHL |
| 1669 |
rol r_remHH |
| 1670 |
cp r_remL,r_arg2L ; compare remainder & divisor |
| 1671 |
cpc r_remH,r_arg2H |
| 1672 |
cpc r_remHL,r_arg2HL |
| 1673 |
cpc r_remHH,r_arg2HH |
| 1674 |
brcs __udivmodsi4_ep ; remainder <= divisor |
| 1675 |
sub r_remL,r_arg2L ; restore remainder |
| 1676 |
sbc r_remH,r_arg2H |
| 1677 |
sbc r_remHL,r_arg2HL |
| 1678 |
sbc r_remHH,r_arg2HH |
| 1679 |
__udivmodsi4_ep: |
| 1680 |
rol r_arg1L ; shift dividend (with CARRY) |
| 1681 |
rol r_arg1H |
| 1682 |
rol r_arg1HL |
| 1683 |
rol r_arg1HH |
| 1684 |
dec r_cnt ; decrement loop counter |
| 1685 |
brne __udivmodsi4_loop |
| 1686 |
; __zero_reg__ now restored (r_cnt == 0) |
| 1687 |
com r_arg1L |
| 1688 |
com r_arg1H |
| 1689 |
com r_arg1HL |
| 1690 |
com r_arg1HH |
| 1691 |
; div/mod results to return registers, as for the ldiv() function |
| 1692 |
mov_l r_arg2L, r_arg1L ; quotient |
| 1693 |
mov_h r_arg2H, r_arg1H |
| 1694 |
mov_l r_arg2HL, r_arg1HL |
| 1695 |
mov_h r_arg2HH, r_arg1HH |
| 1696 |
mov_l r_arg1L, r_remL ; remainder |
| 1697 |
mov_h r_arg1H, r_remH |
| 1698 |
mov_l r_arg1HL, r_remHL |
| 1699 |
mov_h r_arg1HH, r_remHH |
| 1700 |
ret |
| 1701 |
ENDF __udivmodsi4 |
| 1702 |
#endif /* defined (L_udivmodsi4) */ |
| 1703 |
|
| 1704 |
#if defined (L_divmodsi4) |
| 1705 |
DEFUN __divmodsi4 |
| 1706 |
mov __tmp_reg__,r_arg2HH |
| 1707 |
bst r_arg1HH,7 ; store sign of dividend |
| 1708 |
brtc 0f |
| 1709 |
com __tmp_reg__ ; r0.7 is sign of result |
| 1710 |
XCALL __negsi2 ; dividend negative: negate |
| 1711 |
0: |
| 1712 |
sbrc r_arg2HH,7 |
| 1713 |
rcall __divmodsi4_neg2 ; divisor negative: negate |
| 1714 |
XCALL __udivmodsi4 ; do the unsigned div/mod |
| 1715 |
sbrc __tmp_reg__, 7 ; correct quotient sign |
| 1716 |
rcall __divmodsi4_neg2 |
| 1717 |
brtc __divmodsi4_exit ; correct remainder sign |
| 1718 |
XJMP __negsi2 |
| 1719 |
__divmodsi4_neg2: |
| 1720 |
;; correct divisor/quotient sign |
| 1721 |
com r_arg2HH |
| 1722 |
com r_arg2HL |
| 1723 |
com r_arg2H |
| 1724 |
neg r_arg2L |
| 1725 |
sbci r_arg2H,0xff |
| 1726 |
sbci r_arg2HL,0xff |
| 1727 |
sbci r_arg2HH,0xff |
| 1728 |
__divmodsi4_exit: |
| 1729 |
ret |
| 1730 |
ENDF __divmodsi4 |
| 1731 |
#endif /* defined (L_divmodsi4) */ |
| 1732 |
|
| 1733 |
#if defined (L_negsi2) |
| 1734 |
;; (set (reg:SI 22) |
| 1735 |
;; (neg:SI (reg:SI 22))) |
| 1736 |
;; Sets the V flag for signed overflow tests |
| 1737 |
DEFUN __negsi2 |
| 1738 |
NEG4 22 |
| 1739 |
ret |
| 1740 |
ENDF __negsi2 |
| 1741 |
#endif /* L_negsi2 */ |
| 1742 |
|
| 1743 |
#undef r_remHH |
| 1744 |
#undef r_remHL |
| 1745 |
#undef r_remH |
| 1746 |
#undef r_remL |
| 1747 |
#undef r_arg1HH |
| 1748 |
#undef r_arg1HL |
| 1749 |
#undef r_arg1H |
| 1750 |
#undef r_arg1L |
| 1751 |
#undef r_arg2HH |
| 1752 |
#undef r_arg2HL |
| 1753 |
#undef r_arg2H |
| 1754 |
#undef r_arg2L |
| 1755 |
#undef r_cnt |
| 1756 |
|
| 1757 |
/* *di routines use registers below R19 and won't work with tiny arch |
| 1758 |
right now. */ |
| 1759 |
|
| 1760 |
#if !defined (__AVR_TINY__) |
| 1761 |
/******************************************************* |
| 1762 |
Division 64 / 64 |
| 1763 |
Modulo 64 % 64 |
| 1764 |
*******************************************************/ |
| 1765 |
|
| 1766 |
;; Use Speed-optimized Version on "big" Devices, i.e. Devices with |
| 1767 |
;; at least 16k of Program Memory. For smaller Devices, depend |
| 1768 |
;; on MOVW and SP Size. There is a Connexion between SP Size and |
| 1769 |
;; Flash Size so that SP Size can be used to test for Flash Size. |
| 1770 |
|
| 1771 |
#if defined (__AVR_HAVE_JMP_CALL__) |
| 1772 |
# define SPEED_DIV 8 |
| 1773 |
#elif defined (__AVR_HAVE_MOVW__) && defined (__AVR_HAVE_SPH__) |
| 1774 |
# define SPEED_DIV 16 |
| 1775 |
#else |
| 1776 |
# define SPEED_DIV 0 |
| 1777 |
#endif |
| 1778 |
|
| 1779 |
;; A[0..7]: In: Dividend; |
| 1780 |
;; Out: Quotient (T = 0) |
| 1781 |
;; Out: Remainder (T = 1) |
| 1782 |
#define A0 18 |
| 1783 |
#define A1 A0+1 |
| 1784 |
#define A2 A0+2 |
| 1785 |
#define A3 A0+3 |
| 1786 |
#define A4 A0+4 |
| 1787 |
#define A5 A0+5 |
| 1788 |
#define A6 A0+6 |
| 1789 |
#define A7 A0+7 |
| 1790 |
|
| 1791 |
;; B[0..7]: In: Divisor; Out: Clobber |
| 1792 |
#define B0 10 |
| 1793 |
#define B1 B0+1 |
| 1794 |
#define B2 B0+2 |
| 1795 |
#define B3 B0+3 |
| 1796 |
#define B4 B0+4 |
| 1797 |
#define B5 B0+5 |
| 1798 |
#define B6 B0+6 |
| 1799 |
#define B7 B0+7 |
| 1800 |
|
| 1801 |
;; C[0..7]: Expand remainder; Out: Remainder (unused) |
| 1802 |
#define C0 8 |
| 1803 |
#define C1 C0+1 |
| 1804 |
#define C2 30 |
| 1805 |
#define C3 C2+1 |
| 1806 |
#define C4 28 |
| 1807 |
#define C5 C4+1 |
| 1808 |
#define C6 26 |
| 1809 |
#define C7 C6+1 |
| 1810 |
|
| 1811 |
;; Holds Signs during Division Routine |
| 1812 |
#define SS __tmp_reg__ |
| 1813 |
|
| 1814 |
;; Bit-Counter in Division Routine |
| 1815 |
#define R_cnt __zero_reg__ |
| 1816 |
|
| 1817 |
;; Scratch Register for Negation |
| 1818 |
#define NN r31 |
| 1819 |
|
| 1820 |
#if defined (L_udivdi3) |
| 1821 |
|
| 1822 |
;; R25:R18 = R24:R18 umod R17:R10 |
| 1823 |
;; Ordinary ABI-Function |
| 1824 |
|
| 1825 |
DEFUN __umoddi3 |
| 1826 |
set |
| 1827 |
rjmp __udivdi3_umoddi3 |
| 1828 |
ENDF __umoddi3 |
| 1829 |
|
| 1830 |
;; R25:R18 = R24:R18 udiv R17:R10 |
| 1831 |
;; Ordinary ABI-Function |
| 1832 |
|
| 1833 |
DEFUN __udivdi3 |
| 1834 |
clt |
| 1835 |
ENDF __udivdi3 |
| 1836 |
|
| 1837 |
DEFUN __udivdi3_umoddi3 |
| 1838 |
push C0 |
| 1839 |
push C1 |
| 1840 |
push C4 |
| 1841 |
push C5 |
| 1842 |
XCALL __udivmod64 |
| 1843 |
pop C5 |
| 1844 |
pop C4 |
| 1845 |
pop C1 |
| 1846 |
pop C0 |
| 1847 |
ret |
| 1848 |
ENDF __udivdi3_umoddi3 |
| 1849 |
#endif /* L_udivdi3 */ |
| 1850 |
|
| 1851 |
#if defined (L_udivmod64) |
| 1852 |
|
| 1853 |
;; Worker Routine for 64-Bit unsigned Quotient and Remainder Computation |
| 1854 |
;; No Registers saved/restored; the Callers will take Care. |
| 1855 |
;; Preserves B[] and T-flag |
| 1856 |
;; T = 0: Compute Quotient in A[] |
| 1857 |
;; T = 1: Compute Remainder in A[] and shift SS one Bit left |
| 1858 |
|
| 1859 |
DEFUN __udivmod64 |
| 1860 |
|
| 1861 |
;; Clear Remainder (C6, C7 will follow) |
| 1862 |
clr C0 |
| 1863 |
clr C1 |
| 1864 |
wmov C2, C0 |
| 1865 |
wmov C4, C0 |
| 1866 |
ldi C7, 64 |
| 1867 |
|
| 1868 |
#if SPEED_DIV == 0 || SPEED_DIV == 16 |
| 1869 |
;; Initialize Loop-Counter |
| 1870 |
mov R_cnt, C7 |
| 1871 |
wmov C6, C0 |
| 1872 |
#endif /* SPEED_DIV */ |
| 1873 |
|
| 1874 |
#if SPEED_DIV == 8 |
| 1875 |
|
| 1876 |
push A7 |
| 1877 |
clr C6 |
| 1878 |
|
| 1879 |
1: ;; Compare shifted Devidend against Divisor |
| 1880 |
;; If -- even after Shifting -- it is smaller... |
| 1881 |
CP A7,B0 $ cpc C0,B1 $ cpc C1,B2 $ cpc C2,B3 |
| 1882 |
cpc C3,B4 $ cpc C4,B5 $ cpc C5,B6 $ cpc C6,B7 |
| 1883 |
brcc 2f |
| 1884 |
|
| 1885 |
;; ...then we can subtract it. Thus, it is legal to shift left |
| 1886 |
$ mov C6,C5 $ mov C5,C4 $ mov C4,C3 |
| 1887 |
mov C3,C2 $ mov C2,C1 $ mov C1,C0 $ mov C0,A7 |
| 1888 |
mov A7,A6 $ mov A6,A5 $ mov A5,A4 $ mov A4,A3 |
| 1889 |
mov A3,A2 $ mov A2,A1 $ mov A1,A0 $ clr A0 |
| 1890 |
|
| 1891 |
;; 8 Bits are done |
| 1892 |
subi C7, 8 |
| 1893 |
brne 1b |
| 1894 |
|
| 1895 |
;; Shifted 64 Bits: A7 has traveled to C7 |
| 1896 |
pop C7 |
| 1897 |
;; Divisor is greater than Dividend. We have: |
| 1898 |
;; A[] % B[] = A[] |
| 1899 |
;; A[] / B[] = 0 |
| 1900 |
;; Thus, we can return immediately |
| 1901 |
rjmp 5f |
| 1902 |
|
| 1903 |
2: ;; Initialze Bit-Counter with Number of Bits still to be performed |
| 1904 |
mov R_cnt, C7 |
| 1905 |
|
| 1906 |
;; Push of A7 is not needed because C7 is still 0 |
| 1907 |
pop C7 |
| 1908 |
clr C7 |
| 1909 |
|
| 1910 |
#elif SPEED_DIV == 16 |
| 1911 |
|
| 1912 |
;; Compare shifted Dividend against Divisor |
| 1913 |
cp A7, B3 |
| 1914 |
cpc C0, B4 |
| 1915 |
cpc C1, B5 |
| 1916 |
cpc C2, B6 |
| 1917 |
cpc C3, B7 |
| 1918 |
brcc 2f |
| 1919 |
|
| 1920 |
;; Divisor is greater than shifted Dividen: We can shift the Dividend |
| 1921 |
;; and it is still smaller than the Divisor --> Shift one 32-Bit Chunk |
| 1922 |
wmov C2,A6 $ wmov C0,A4 |
| 1923 |
wmov A6,A2 $ wmov A4,A0 |
| 1924 |
wmov A2,C6 $ wmov A0,C4 |
| 1925 |
|
| 1926 |
;; Set Bit Counter to 32 |
| 1927 |
lsr R_cnt |
| 1928 |
2: |
| 1929 |
#elif SPEED_DIV |
| 1930 |
#error SPEED_DIV = ? |
| 1931 |
#endif /* SPEED_DIV */ |
| 1932 |
|
| 1933 |
;; The very Division + Remainder Routine |
| 1934 |
|
| 1935 |
3: ;; Left-shift Dividend... |
| 1936 |
lsl A0 $ rol A1 $ rol A2 $ rol A3 |
| 1937 |
rol A4 $ rol A5 $ rol A6 $ rol A7 |
| 1938 |
|
| 1939 |
;; ...into Remainder |
| 1940 |
rol C0 $ rol C1 $ rol C2 $ rol C3 |
| 1941 |
rol C4 $ rol C5 $ rol C6 $ rol C7 |
| 1942 |
|
| 1943 |
;; Compare Remainder and Divisor |
| 1944 |
CP C0,B0 $ cpc C1,B1 $ cpc C2,B2 $ cpc C3,B3 |
| 1945 |
cpc C4,B4 $ cpc C5,B5 $ cpc C6,B6 $ cpc C7,B7 |
| 1946 |
|
| 1947 |
brcs 4f |
| 1948 |
|
| 1949 |
;; Divisor fits into Remainder: Subtract it from Remainder... |
| 1950 |
SUB C0,B0 $ sbc C1,B1 $ sbc C2,B2 $ sbc C3,B3 |
| 1951 |
sbc C4,B4 $ sbc C5,B5 $ sbc C6,B6 $ sbc C7,B7 |
| 1952 |
|
| 1953 |
;; ...and set according Bit in the upcoming Quotient |
| 1954 |
;; The Bit will travel to its final Position |
| 1955 |
ori A0, 1 |
| 1956 |
|
| 1957 |
4: ;; This Bit is done |
| 1958 |
dec R_cnt |
| 1959 |
brne 3b |
| 1960 |
;; __zero_reg__ is 0 again |
| 1961 |
|
| 1962 |
;; T = 0: We are fine with the Quotient in A[] |
| 1963 |
;; T = 1: Copy Remainder to A[] |
| 1964 |
5: brtc 6f |
| 1965 |
wmov A0, C0 |
| 1966 |
wmov A2, C2 |
| 1967 |
wmov A4, C4 |
| 1968 |
wmov A6, C6 |
| 1969 |
;; Move the Sign of the Result to SS.7 |
| 1970 |
lsl SS |
| 1971 |
|
| 1972 |
6: ret |
| 1973 |
|
| 1974 |
ENDF __udivmod64 |
| 1975 |
#endif /* L_udivmod64 */ |
| 1976 |
|
| 1977 |
|
| 1978 |
#if defined (L_divdi3) |
| 1979 |
|
| 1980 |
;; R25:R18 = R24:R18 mod R17:R10 |
| 1981 |
;; Ordinary ABI-Function |
| 1982 |
|
| 1983 |
DEFUN __moddi3 |
| 1984 |
set |
| 1985 |
rjmp __divdi3_moddi3 |
| 1986 |
ENDF __moddi3 |
| 1987 |
|
| 1988 |
;; R25:R18 = R24:R18 div R17:R10 |
| 1989 |
;; Ordinary ABI-Function |
| 1990 |
|
| 1991 |
DEFUN __divdi3 |
| 1992 |
clt |
| 1993 |
ENDF __divdi3 |
| 1994 |
|
| 1995 |
DEFUN __divdi3_moddi3 |
| 1996 |
#if SPEED_DIV |
| 1997 |
mov r31, A7 |
| 1998 |
or r31, B7 |
| 1999 |
brmi 0f |
| 2000 |
;; Both Signs are 0: the following Complexitiy is not needed |
| 2001 |
XJMP __udivdi3_umoddi3 |
| 2002 |
#endif /* SPEED_DIV */ |
| 2003 |
|
| 2004 |
0: ;; The Prologue |
| 2005 |
;; Save 12 Registers: Y, 17...8 |
| 2006 |
;; No Frame needed |
| 2007 |
do_prologue_saves 12 |
| 2008 |
|
| 2009 |
;; SS.7 will contain the Sign of the Quotient (A.sign * B.sign) |
| 2010 |
;; SS.6 will contain the Sign of the Remainder (A.sign) |
| 2011 |
mov SS, A7 |
| 2012 |
asr SS |
| 2013 |
;; Adjust Dividend's Sign as needed |
| 2014 |
#if SPEED_DIV |
| 2015 |
;; Compiling for Speed we know that at least one Sign must be < 0 |
| 2016 |
;; Thus, if A[] >= 0 then we know B[] < 0 |
| 2017 |
brpl 22f |
| 2018 |
#else |
| 2019 |
brpl 21f |
| 2020 |
#endif /* SPEED_DIV */ |
| 2021 |
|
| 2022 |
XCALL __negdi2 |
| 2023 |
|
| 2024 |
;; Adjust Divisor's Sign and SS.7 as needed |
| 2025 |
21: tst B7 |
| 2026 |
brpl 3f |
| 2027 |
22: ldi NN, 1 << 7 |
| 2028 |
eor SS, NN |
| 2029 |
|
| 2030 |
ldi NN, -1 |
| 2031 |
com B4 $ com B5 $ com B6 $ com B7 |
| 2032 |
$ com B1 $ com B2 $ com B3 |
| 2033 |
NEG B0 |
| 2034 |
$ sbc B1,NN $ sbc B2,NN $ sbc B3,NN |
| 2035 |
sbc B4,NN $ sbc B5,NN $ sbc B6,NN $ sbc B7,NN |
| 2036 |
|
| 2037 |
3: ;; Do the unsigned 64-Bit Division/Modulo (depending on T-flag) |
| 2038 |
XCALL __udivmod64 |
| 2039 |
|
| 2040 |
;; Adjust Result's Sign |
| 2041 |
#ifdef __AVR_ERRATA_SKIP_JMP_CALL__ |
| 2042 |
tst SS |
| 2043 |
brpl 4f |
| 2044 |
#else |
| 2045 |
sbrc SS, 7 |
| 2046 |
#endif /* __AVR_HAVE_JMP_CALL__ */ |
| 2047 |
XCALL __negdi2 |
| 2048 |
|
| 2049 |
4: ;; Epilogue: Restore 12 Registers and return |
| 2050 |
do_epilogue_restores 12 |
| 2051 |
|
| 2052 |
ENDF __divdi3_moddi3 |
| 2053 |
|
| 2054 |
#endif /* L_divdi3 */ |
| 2055 |
|
| 2056 |
#undef R_cnt |
| 2057 |
#undef SS |
| 2058 |
#undef NN |
| 2059 |
|
| 2060 |
.section .text.libgcc, "ax", @progbits |
| 2061 |
|
| 2062 |
#define TT __tmp_reg__ |
| 2063 |
|
| 2064 |
#if defined (L_adddi3) |
| 2065 |
;; (set (reg:DI 18) |
| 2066 |
;; (plus:DI (reg:DI 18) |
| 2067 |
;; (reg:DI 10))) |
| 2068 |
;; Sets the V flag for signed overflow tests |
| 2069 |
;; Sets the C flag for unsigned overflow tests |
| 2070 |
DEFUN __adddi3 |
| 2071 |
ADD A0,B0 $ adc A1,B1 $ adc A2,B2 $ adc A3,B3 |
| 2072 |
adc A4,B4 $ adc A5,B5 $ adc A6,B6 $ adc A7,B7 |
| 2073 |
ret |
| 2074 |
ENDF __adddi3 |
| 2075 |
#endif /* L_adddi3 */ |
| 2076 |
|
| 2077 |
#if defined (L_adddi3_s8) |
| 2078 |
;; (set (reg:DI 18) |
| 2079 |
;; (plus:DI (reg:DI 18) |
| 2080 |
;; (sign_extend:SI (reg:QI 26)))) |
| 2081 |
;; Sets the V flag for signed overflow tests |
| 2082 |
;; Sets the C flag for unsigned overflow tests provided 0 <= R26 < 128 |
| 2083 |
DEFUN __adddi3_s8 |
| 2084 |
clr TT |
| 2085 |
sbrc r26, 7 |
| 2086 |
com TT |
| 2087 |
ADD A0,r26 $ adc A1,TT $ adc A2,TT $ adc A3,TT |
| 2088 |
adc A4,TT $ adc A5,TT $ adc A6,TT $ adc A7,TT |
| 2089 |
ret |
| 2090 |
ENDF __adddi3_s8 |
| 2091 |
#endif /* L_adddi3_s8 */ |
| 2092 |
|
| 2093 |
#if defined (L_subdi3) |
| 2094 |
;; (set (reg:DI 18) |
| 2095 |
;; (minus:DI (reg:DI 18) |
| 2096 |
;; (reg:DI 10))) |
| 2097 |
;; Sets the V flag for signed overflow tests |
| 2098 |
;; Sets the C flag for unsigned overflow tests |
| 2099 |
DEFUN __subdi3 |
| 2100 |
SUB A0,B0 $ sbc A1,B1 $ sbc A2,B2 $ sbc A3,B3 |
| 2101 |
sbc A4,B4 $ sbc A5,B5 $ sbc A6,B6 $ sbc A7,B7 |
| 2102 |
ret |
| 2103 |
ENDF __subdi3 |
| 2104 |
#endif /* L_subdi3 */ |
| 2105 |
|
| 2106 |
#if defined (L_cmpdi2) |
| 2107 |
;; (set (cc0) |
| 2108 |
;; (compare (reg:DI 18) |
| 2109 |
;; (reg:DI 10))) |
| 2110 |
DEFUN __cmpdi2 |
| 2111 |
CP A0,B0 $ cpc A1,B1 $ cpc A2,B2 $ cpc A3,B3 |
| 2112 |
cpc A4,B4 $ cpc A5,B5 $ cpc A6,B6 $ cpc A7,B7 |
| 2113 |
ret |
| 2114 |
ENDF __cmpdi2 |
| 2115 |
#endif /* L_cmpdi2 */ |
| 2116 |
|
| 2117 |
#if defined (L_cmpdi2_s8) |
| 2118 |
;; (set (cc0) |
| 2119 |
;; (compare (reg:DI 18) |
| 2120 |
;; (sign_extend:SI (reg:QI 26)))) |
| 2121 |
DEFUN __cmpdi2_s8 |
| 2122 |
clr TT |
| 2123 |
sbrc r26, 7 |
| 2124 |
com TT |
| 2125 |
CP A0,r26 $ cpc A1,TT $ cpc A2,TT $ cpc A3,TT |
| 2126 |
cpc A4,TT $ cpc A5,TT $ cpc A6,TT $ cpc A7,TT |
| 2127 |
ret |
| 2128 |
ENDF __cmpdi2_s8 |
| 2129 |
#endif /* L_cmpdi2_s8 */ |
| 2130 |
|
| 2131 |
#if defined (L_negdi2) |
| 2132 |
;; (set (reg:DI 18) |
| 2133 |
;; (neg:DI (reg:DI 18))) |
| 2134 |
;; Sets the V flag for signed overflow tests |
| 2135 |
DEFUN __negdi2 |
| 2136 |
|
| 2137 |
com A4 $ com A5 $ com A6 $ com A7 |
| 2138 |
$ com A1 $ com A2 $ com A3 |
| 2139 |
NEG A0 |
| 2140 |
$ sbci A1,-1 $ sbci A2,-1 $ sbci A3,-1 |
| 2141 |
sbci A4,-1 $ sbci A5,-1 $ sbci A6,-1 $ sbci A7,-1 |
| 2142 |
ret |
| 2143 |
|
| 2144 |
ENDF __negdi2 |
| 2145 |
#endif /* L_negdi2 */ |
| 2146 |
|
| 2147 |
#undef TT |
| 2148 |
|
| 2149 |
#undef C7 |
| 2150 |
#undef C6 |
| 2151 |
#undef C5 |
| 2152 |
#undef C4 |
| 2153 |
#undef C3 |
| 2154 |
#undef C2 |
| 2155 |
#undef C1 |
| 2156 |
#undef C0 |
| 2157 |
|
| 2158 |
#undef B7 |
| 2159 |
#undef B6 |
| 2160 |
#undef B5 |
| 2161 |
#undef B4 |
| 2162 |
#undef B3 |
| 2163 |
#undef B2 |
| 2164 |
#undef B1 |
| 2165 |
#undef B0 |
| 2166 |
|
| 2167 |
#undef A7 |
| 2168 |
#undef A6 |
| 2169 |
#undef A5 |
| 2170 |
#undef A4 |
| 2171 |
#undef A3 |
| 2172 |
#undef A2 |
| 2173 |
#undef A1 |
| 2174 |
#undef A0 |
| 2175 |
|
| 2176 |
#endif /* !defined (__AVR_TINY__) */ |
| 2177 |
|
| 2178 |
|
| 2179 |
.section .text.libgcc.prologue, "ax", @progbits |
| 2180 |
|
| 2181 |
/********************************** |
| 2182 |
* This is a prologue subroutine |
| 2183 |
**********************************/ |
| 2184 |
#if !defined (__AVR_TINY__) |
| 2185 |
#if defined (L_prologue) |
| 2186 |
|
| 2187 |
;; This function does not clobber T-flag; 64-bit division relies on it |
| 2188 |
DEFUN __prologue_saves__ |
| 2189 |
push r2 |
| 2190 |
push r3 |
| 2191 |
push r4 |
| 2192 |
push r5 |
| 2193 |
push r6 |
| 2194 |
push r7 |
| 2195 |
push r8 |
| 2196 |
push r9 |
| 2197 |
push r10 |
| 2198 |
push r11 |
| 2199 |
push r12 |
| 2200 |
push r13 |
| 2201 |
push r14 |
| 2202 |
push r15 |
| 2203 |
push r16 |
| 2204 |
push r17 |
| 2205 |
push r28 |
| 2206 |
push r29 |
| 2207 |
#if !defined (__AVR_HAVE_SPH__) |
| 2208 |
in r28,__SP_L__ |
| 2209 |
sub r28,r26 |
| 2210 |
out __SP_L__,r28 |
| 2211 |
clr r29 |
| 2212 |
#elif defined (__AVR_XMEGA__) |
| 2213 |
in r28,__SP_L__ |
| 2214 |
in r29,__SP_H__ |
| 2215 |
sub r28,r26 |
| 2216 |
sbc r29,r27 |
| 2217 |
out __SP_L__,r28 |
| 2218 |
out __SP_H__,r29 |
| 2219 |
#else |
| 2220 |
in r28,__SP_L__ |
| 2221 |
in r29,__SP_H__ |
| 2222 |
sub r28,r26 |
| 2223 |
sbc r29,r27 |
| 2224 |
in __tmp_reg__,__SREG__ |
| 2225 |
cli |
| 2226 |
out __SP_H__,r29 |
| 2227 |
out __SREG__,__tmp_reg__ |
| 2228 |
out __SP_L__,r28 |
| 2229 |
#endif /* #SP = 8/16 */ |
| 2230 |
|
| 2231 |
XIJMP |
| 2232 |
|
| 2233 |
ENDF __prologue_saves__ |
| 2234 |
#endif /* defined (L_prologue) */ |
| 2235 |
|
| 2236 |
/* |
| 2237 |
* This is an epilogue subroutine |
| 2238 |
*/ |
| 2239 |
#if defined (L_epilogue) |
| 2240 |
|
| 2241 |
DEFUN __epilogue_restores__ |
| 2242 |
ldd r2,Y+18 |
| 2243 |
ldd r3,Y+17 |
| 2244 |
ldd r4,Y+16 |
| 2245 |
ldd r5,Y+15 |
| 2246 |
ldd r6,Y+14 |
| 2247 |
ldd r7,Y+13 |
| 2248 |
ldd r8,Y+12 |
| 2249 |
ldd r9,Y+11 |
| 2250 |
ldd r10,Y+10 |
| 2251 |
ldd r11,Y+9 |
| 2252 |
ldd r12,Y+8 |
| 2253 |
ldd r13,Y+7 |
| 2254 |
ldd r14,Y+6 |
| 2255 |
ldd r15,Y+5 |
| 2256 |
ldd r16,Y+4 |
| 2257 |
ldd r17,Y+3 |
| 2258 |
ldd r26,Y+2 |
| 2259 |
#if !defined (__AVR_HAVE_SPH__) |
| 2260 |
ldd r29,Y+1 |
| 2261 |
add r28,r30 |
| 2262 |
out __SP_L__,r28 |
| 2263 |
mov r28, r26 |
| 2264 |
#elif defined (__AVR_XMEGA__) |
| 2265 |
ldd r27,Y+1 |
| 2266 |
add r28,r30 |
| 2267 |
adc r29,__zero_reg__ |
| 2268 |
out __SP_L__,r28 |
| 2269 |
out __SP_H__,r29 |
| 2270 |
wmov 28, 26 |
| 2271 |
#else |
| 2272 |
ldd r27,Y+1 |
| 2273 |
add r28,r30 |
| 2274 |
adc r29,__zero_reg__ |
| 2275 |
in __tmp_reg__,__SREG__ |
| 2276 |
cli |
| 2277 |
out __SP_H__,r29 |
| 2278 |
out __SREG__,__tmp_reg__ |
| 2279 |
out __SP_L__,r28 |
| 2280 |
mov_l r28, r26 |
| 2281 |
mov_h r29, r27 |
| 2282 |
#endif /* #SP = 8/16 */ |
| 2283 |
ret |
| 2284 |
ENDF __epilogue_restores__ |
| 2285 |
#endif /* defined (L_epilogue) */ |
| 2286 |
#endif /* !defined (__AVR_TINY__) */ |
| 2287 |
|
| 2288 |
#ifdef L_exit |
| 2289 |
.section .fini9,"ax",@progbits |
| 2290 |
DEFUN _exit |
| 2291 |
.weak exit |
| 2292 |
exit: |
| 2293 |
ENDF _exit |
| 2294 |
|
| 2295 |
/* Code from .fini8 ... .fini1 sections inserted by ld script. */ |
| 2296 |
|
| 2297 |
.section .fini0,"ax",@progbits |
| 2298 |
cli |
| 2299 |
__stop_program: |
| 2300 |
rjmp __stop_program |
| 2301 |
#endif /* defined (L_exit) */ |
| 2302 |
|
| 2303 |
#ifdef L_cleanup |
| 2304 |
.weak _cleanup |
| 2305 |
.func _cleanup |
| 2306 |
_cleanup: |
| 2307 |
ret |
| 2308 |
.endfunc |
| 2309 |
#endif /* defined (L_cleanup) */ |
| 2310 |
|
| 2311 |
|
| 2312 |
.section .text.libgcc, "ax", @progbits |
| 2313 |
|
| 2314 |
#ifdef L_tablejump2 |
| 2315 |
DEFUN __tablejump2__ |
| 2316 |
lsl r30 |
| 2317 |
rol r31 |
| 2318 |
#if defined (__AVR_HAVE_EIJMP_EICALL__) |
| 2319 |
;; Word address of gs() jumptable entry in R24:Z |
| 2320 |
rol r24 |
| 2321 |
out __RAMPZ__, r24 |
| 2322 |
#elif defined (__AVR_HAVE_ELPM__) |
| 2323 |
;; Word address of jumptable entry in Z |
| 2324 |
clr __tmp_reg__ |
| 2325 |
rol __tmp_reg__ |
| 2326 |
out __RAMPZ__, __tmp_reg__ |
| 2327 |
#endif |
| 2328 |
|
| 2329 |
;; Read word address from jumptable and jump |
| 2330 |
|
| 2331 |
#if defined (__AVR_HAVE_ELPMX__) |
| 2332 |
elpm __tmp_reg__, Z+ |
| 2333 |
elpm r31, Z |
| 2334 |
mov r30, __tmp_reg__ |
| 2335 |
#ifdef __AVR_HAVE_RAMPD__ |
| 2336 |
;; Reset RAMPZ to 0 so that EBI devices don't read garbage from RAM |
| 2337 |
out __RAMPZ__, __zero_reg__ |
| 2338 |
#endif /* RAMPD */ |
| 2339 |
XIJMP |
| 2340 |
#elif defined (__AVR_HAVE_ELPM__) |
| 2341 |
elpm |
| 2342 |
push r0 |
| 2343 |
adiw r30, 1 |
| 2344 |
elpm |
| 2345 |
push r0 |
| 2346 |
ret |
| 2347 |
#elif defined (__AVR_HAVE_LPMX__) |
| 2348 |
lpm __tmp_reg__, Z+ |
| 2349 |
lpm r31, Z |
| 2350 |
mov r30, __tmp_reg__ |
| 2351 |
ijmp |
| 2352 |
#elif defined (__AVR_TINY__) |
| 2353 |
wsubi 30, -(__AVR_TINY_PM_BASE_ADDRESS__) ; Add PM offset to Z |
| 2354 |
ld __tmp_reg__, Z+ |
| 2355 |
ld r31, Z ; Use ld instead of lpm to load Z |
| 2356 |
mov r30, __tmp_reg__ |
| 2357 |
ijmp |
| 2358 |
#else |
| 2359 |
lpm |
| 2360 |
push r0 |
| 2361 |
adiw r30, 1 |
| 2362 |
lpm |
| 2363 |
push r0 |
| 2364 |
ret |
| 2365 |
#endif |
| 2366 |
ENDF __tablejump2__ |
| 2367 |
#endif /* L_tablejump2 */ |
| 2368 |
|
| 2369 |
#if defined(__AVR_TINY__) |
| 2370 |
#ifdef L_copy_data |
| 2371 |
.section .init4,"ax",@progbits |
| 2372 |
.global __do_copy_data |
| 2373 |
__do_copy_data: |
| 2374 |
ldi r18, hi8(__data_end) |
| 2375 |
ldi r26, lo8(__data_start) |
| 2376 |
ldi r27, hi8(__data_start) |
| 2377 |
ldi r30, lo8(__data_load_start + __AVR_TINY_PM_BASE_ADDRESS__) |
| 2378 |
ldi r31, hi8(__data_load_start + __AVR_TINY_PM_BASE_ADDRESS__) |
| 2379 |
rjmp .L__do_copy_data_start |
| 2380 |
.L__do_copy_data_loop: |
| 2381 |
ld r19, z+ |
| 2382 |
st X+, r19 |
| 2383 |
.L__do_copy_data_start: |
| 2384 |
cpi r26, lo8(__data_end) |
| 2385 |
cpc r27, r18 |
| 2386 |
brne .L__do_copy_data_loop |
| 2387 |
#endif |
| 2388 |
#else |
| 2389 |
#ifdef L_copy_data |
| 2390 |
.section .init4,"ax",@progbits |
| 2391 |
DEFUN __do_copy_data |
| 2392 |
#if defined(__AVR_HAVE_ELPMX__) |
| 2393 |
ldi r17, hi8(__data_end) |
| 2394 |
ldi r26, lo8(__data_start) |
| 2395 |
ldi r27, hi8(__data_start) |
| 2396 |
ldi r30, lo8(__data_load_start) |
| 2397 |
ldi r31, hi8(__data_load_start) |
| 2398 |
ldi r16, hh8(__data_load_start) |
| 2399 |
out __RAMPZ__, r16 |
| 2400 |
rjmp .L__do_copy_data_start |
| 2401 |
.L__do_copy_data_loop: |
| 2402 |
elpm r0, Z+ |
| 2403 |
st X+, r0 |
| 2404 |
.L__do_copy_data_start: |
| 2405 |
cpi r26, lo8(__data_end) |
| 2406 |
cpc r27, r17 |
| 2407 |
brne .L__do_copy_data_loop |
| 2408 |
#elif !defined(__AVR_HAVE_ELPMX__) && defined(__AVR_HAVE_ELPM__) |
| 2409 |
ldi r17, hi8(__data_end) |
| 2410 |
ldi r26, lo8(__data_start) |
| 2411 |
ldi r27, hi8(__data_start) |
| 2412 |
ldi r30, lo8(__data_load_start) |
| 2413 |
ldi r31, hi8(__data_load_start) |
| 2414 |
ldi r16, hh8(__data_load_start - 0x10000) |
| 2415 |
.L__do_copy_data_carry: |
| 2416 |
inc r16 |
| 2417 |
out __RAMPZ__, r16 |
| 2418 |
rjmp .L__do_copy_data_start |
| 2419 |
.L__do_copy_data_loop: |
| 2420 |
elpm |
| 2421 |
st X+, r0 |
| 2422 |
adiw r30, 1 |
| 2423 |
brcs .L__do_copy_data_carry |
| 2424 |
.L__do_copy_data_start: |
| 2425 |
cpi r26, lo8(__data_end) |
| 2426 |
cpc r27, r17 |
| 2427 |
brne .L__do_copy_data_loop |
| 2428 |
#elif !defined(__AVR_HAVE_ELPMX__) && !defined(__AVR_HAVE_ELPM__) |
| 2429 |
ldi r17, hi8(__data_end) |
| 2430 |
ldi r26, lo8(__data_start) |
| 2431 |
ldi r27, hi8(__data_start) |
| 2432 |
ldi r30, lo8(__data_load_start) |
| 2433 |
ldi r31, hi8(__data_load_start) |
| 2434 |
rjmp .L__do_copy_data_start |
| 2435 |
.L__do_copy_data_loop: |
| 2436 |
#if defined (__AVR_HAVE_LPMX__) |
| 2437 |
lpm r0, Z+ |
| 2438 |
#else |
| 2439 |
lpm |
| 2440 |
adiw r30, 1 |
| 2441 |
#endif |
| 2442 |
st X+, r0 |
| 2443 |
.L__do_copy_data_start: |
| 2444 |
cpi r26, lo8(__data_end) |
| 2445 |
cpc r27, r17 |
| 2446 |
brne .L__do_copy_data_loop |
| 2447 |
#endif /* !defined(__AVR_HAVE_ELPMX__) && !defined(__AVR_HAVE_ELPM__) */ |
| 2448 |
#if defined (__AVR_HAVE_ELPM__) && defined (__AVR_HAVE_RAMPD__) |
| 2449 |
;; Reset RAMPZ to 0 so that EBI devices don't read garbage from RAM |
| 2450 |
out __RAMPZ__, __zero_reg__ |
| 2451 |
#endif /* ELPM && RAMPD */ |
| 2452 |
ENDF __do_copy_data |
| 2453 |
#endif /* L_copy_data */ |
| 2454 |
#endif /* !defined (__AVR_TINY__) */ |
| 2455 |
|
| 2456 |
/* __do_clear_bss is only necessary if there is anything in .bss section. */ |
| 2457 |
|
| 2458 |
#ifdef L_clear_bss |
| 2459 |
.section .init4,"ax",@progbits |
| 2460 |
DEFUN __do_clear_bss |
| 2461 |
ldi r18, hi8(__bss_end) |
| 2462 |
ldi r26, lo8(__bss_start) |
| 2463 |
ldi r27, hi8(__bss_start) |
| 2464 |
rjmp .do_clear_bss_start |
| 2465 |
.do_clear_bss_loop: |
| 2466 |
st X+, __zero_reg__ |
| 2467 |
.do_clear_bss_start: |
| 2468 |
cpi r26, lo8(__bss_end) |
| 2469 |
cpc r27, r18 |
| 2470 |
brne .do_clear_bss_loop |
| 2471 |
ENDF __do_clear_bss |
| 2472 |
#endif /* L_clear_bss */ |
| 2473 |
|
| 2474 |
/* __do_global_ctors and __do_global_dtors are only necessary |
| 2475 |
if there are any constructors/destructors. */ |
| 2476 |
|
| 2477 |
#if defined(__AVR_TINY__) |
| 2478 |
#define cdtors_tst_reg r18 |
| 2479 |
#else |
| 2480 |
#define cdtors_tst_reg r17 |
| 2481 |
#endif |
| 2482 |
|
| 2483 |
#ifdef L_ctors |
| 2484 |
.section .init6,"ax",@progbits |
| 2485 |
DEFUN __do_global_ctors |
| 2486 |
ldi cdtors_tst_reg, pm_hi8(__ctors_start) |
| 2487 |
ldi r28, pm_lo8(__ctors_end) |
| 2488 |
ldi r29, pm_hi8(__ctors_end) |
| 2489 |
#ifdef __AVR_HAVE_EIJMP_EICALL__ |
| 2490 |
ldi r16, pm_hh8(__ctors_end) |
| 2491 |
#endif /* HAVE_EIJMP */ |
| 2492 |
rjmp .L__do_global_ctors_start |
| 2493 |
.L__do_global_ctors_loop: |
| 2494 |
wsubi 28, 1 |
| 2495 |
#ifdef __AVR_HAVE_EIJMP_EICALL__ |
| 2496 |
sbc r16, __zero_reg__ |
| 2497 |
mov r24, r16 |
| 2498 |
#endif /* HAVE_EIJMP */ |
| 2499 |
mov_h r31, r29 |
| 2500 |
mov_l r30, r28 |
| 2501 |
XCALL __tablejump2__ |
| 2502 |
.L__do_global_ctors_start: |
| 2503 |
cpi r28, pm_lo8(__ctors_start) |
| 2504 |
cpc r29, cdtors_tst_reg |
| 2505 |
#ifdef __AVR_HAVE_EIJMP_EICALL__ |
| 2506 |
ldi r24, pm_hh8(__ctors_start) |
| 2507 |
cpc r16, r24 |
| 2508 |
#endif /* HAVE_EIJMP */ |
| 2509 |
brne .L__do_global_ctors_loop |
| 2510 |
ENDF __do_global_ctors |
| 2511 |
#endif /* L_ctors */ |
| 2512 |
|
| 2513 |
#ifdef L_dtors |
| 2514 |
.section .fini6,"ax",@progbits |
| 2515 |
DEFUN __do_global_dtors |
| 2516 |
ldi cdtors_tst_reg, pm_hi8(__dtors_end) |
| 2517 |
ldi r28, pm_lo8(__dtors_start) |
| 2518 |
ldi r29, pm_hi8(__dtors_start) |
| 2519 |
#ifdef __AVR_HAVE_EIJMP_EICALL__ |
| 2520 |
ldi r16, pm_hh8(__dtors_start) |
| 2521 |
#endif /* HAVE_EIJMP */ |
| 2522 |
rjmp .L__do_global_dtors_start |
| 2523 |
.L__do_global_dtors_loop: |
| 2524 |
#ifdef __AVR_HAVE_EIJMP_EICALL__ |
| 2525 |
mov r24, r16 |
| 2526 |
#endif /* HAVE_EIJMP */ |
| 2527 |
mov_h r31, r29 |
| 2528 |
mov_l r30, r28 |
| 2529 |
XCALL __tablejump2__ |
| 2530 |
waddi 28, 1 |
| 2531 |
#ifdef __AVR_HAVE_EIJMP_EICALL__ |
| 2532 |
adc r16, __zero_reg__ |
| 2533 |
#endif /* HAVE_EIJMP */ |
| 2534 |
.L__do_global_dtors_start: |
| 2535 |
cpi r28, pm_lo8(__dtors_end) |
| 2536 |
cpc r29, cdtors_tst_reg |
| 2537 |
#ifdef __AVR_HAVE_EIJMP_EICALL__ |
| 2538 |
ldi r24, pm_hh8(__dtors_end) |
| 2539 |
cpc r16, r24 |
| 2540 |
#endif /* HAVE_EIJMP */ |
| 2541 |
brne .L__do_global_dtors_loop |
| 2542 |
ENDF __do_global_dtors |
| 2543 |
#endif /* L_dtors */ |
| 2544 |
|
| 2545 |
#undef cdtors_tst_reg |
| 2546 |
|
| 2547 |
.section .text.libgcc, "ax", @progbits |
| 2548 |
|
| 2549 |
#if !defined (__AVR_TINY__) |
| 2550 |
;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;; |
| 2551 |
;; Loading n bytes from Flash; n = 3,4 |
| 2552 |
;; R22... = Flash[Z] |
| 2553 |
;; Clobbers: __tmp_reg__ |
| 2554 |
|
| 2555 |
#if (defined (L_load_3) \ |
| 2556 |
|| defined (L_load_4)) \ |
| 2557 |
&& !defined (__AVR_HAVE_LPMX__) |
| 2558 |
|
| 2559 |
;; Destination |
| 2560 |
#define D0 22 |
| 2561 |
#define D1 D0+1 |
| 2562 |
#define D2 D0+2 |
| 2563 |
#define D3 D0+3 |
| 2564 |
|
| 2565 |
.macro .load dest, n |
| 2566 |
lpm |
| 2567 |
mov \dest, r0 |
| 2568 |
.if \dest != D0+\n-1 |
| 2569 |
adiw r30, 1 |
| 2570 |
.else |
| 2571 |
sbiw r30, \n-1 |
| 2572 |
.endif |
| 2573 |
.endm |
| 2574 |
|
| 2575 |
#if defined (L_load_3) |
| 2576 |
DEFUN __load_3 |
| 2577 |
push D3 |
| 2578 |
XCALL __load_4 |
| 2579 |
pop D3 |
| 2580 |
ret |
| 2581 |
ENDF __load_3 |
| 2582 |
#endif /* L_load_3 */ |
| 2583 |
|
| 2584 |
#if defined (L_load_4) |
| 2585 |
DEFUN __load_4 |
| 2586 |
.load D0, 4 |
| 2587 |
.load D1, 4 |
| 2588 |
.load D2, 4 |
| 2589 |
.load D3, 4 |
| 2590 |
ret |
| 2591 |
ENDF __load_4 |
| 2592 |
#endif /* L_load_4 */ |
| 2593 |
|
| 2594 |
#endif /* L_load_3 || L_load_3 */ |
| 2595 |
#endif /* !defined (__AVR_TINY__) */ |
| 2596 |
|
| 2597 |
#if !defined (__AVR_TINY__) |
| 2598 |
;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;; |
| 2599 |
;; Loading n bytes from Flash or RAM; n = 1,2,3,4 |
| 2600 |
;; R22... = Flash[R21:Z] or RAM[Z] depending on R21.7 |
| 2601 |
;; Clobbers: __tmp_reg__, R21, R30, R31 |
| 2602 |
|
| 2603 |
#if (defined (L_xload_1) \ |
| 2604 |
|| defined (L_xload_2) \ |
| 2605 |
|| defined (L_xload_3) \ |
| 2606 |
|| defined (L_xload_4)) |
| 2607 |
|
| 2608 |
;; Destination |
| 2609 |
#define D0 22 |
| 2610 |
#define D1 D0+1 |
| 2611 |
#define D2 D0+2 |
| 2612 |
#define D3 D0+3 |
| 2613 |
|
| 2614 |
;; Register containing bits 16+ of the address |
| 2615 |
|
| 2616 |
#define HHI8 21 |
| 2617 |
|
| 2618 |
#if defined (L_xload_1) |
| 2619 |
DEFUN __xload_1 |
| 2620 |
#if defined (__AVR_HAVE_LPMX__) && !defined (__AVR_HAVE_ELPM__) |
| 2621 |
sbrc HHI8, 7 |
| 2622 |
ld D0, Z |
| 2623 |
sbrs HHI8, 7 |
| 2624 |
lpm D0, Z |
| 2625 |
ret |
| 2626 |
#else |
| 2627 |
.branch_plus HHI8, __fload_1 |
| 2628 |
ld D0, Z |
| 2629 |
ret |
| 2630 |
#endif /* LPMx && ! ELPM */ |
| 2631 |
ENDF __xload_1 |
| 2632 |
#endif /* L_xload_1 */ |
| 2633 |
|
| 2634 |
#if defined (L_xload_2) |
| 2635 |
DEFUN __xload_2 |
| 2636 |
.branch_plus HHI8, __fload_2 |
| 2637 |
ld D0, Z+ |
| 2638 |
ld D1, Z+ |
| 2639 |
ret |
| 2640 |
ENDF __xload_2 |
| 2641 |
#endif /* L_xload_2 */ |
| 2642 |
|
| 2643 |
#if defined (L_xload_3) |
| 2644 |
DEFUN __xload_3 |
| 2645 |
.branch_plus HHI8, __fload_3 |
| 2646 |
ld D0, Z+ |
| 2647 |
ld D1, Z+ |
| 2648 |
ld D2, Z+ |
| 2649 |
ret |
| 2650 |
ENDF __xload_3 |
| 2651 |
#endif /* L_xload_3 */ |
| 2652 |
|
| 2653 |
#if defined (L_xload_4) |
| 2654 |
DEFUN __xload_4 |
| 2655 |
.branch_plus HHI8, __fload_4 |
| 2656 |
ld D0, Z+ |
| 2657 |
ld D1, Z+ |
| 2658 |
ld D2, Z+ |
| 2659 |
ld D3, Z+ |
| 2660 |
ret |
| 2661 |
ENDF __xload_4 |
| 2662 |
#endif /* L_xload_4 */ |
| 2663 |
|
| 2664 |
#endif /* L_xload_{1|2|3|4} */ |
| 2665 |
#endif /* if !defined (__AVR_TINY__) */ |
| 2666 |
|
| 2667 |
#if !defined (__AVR_TINY__) |
| 2668 |
;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;; |
| 2669 |
;; Loading n bytes from Flash; n = 1,2,3,4 |
| 2670 |
;; R22... = Flash[R21:Z] |
| 2671 |
;; Clobbers: __tmp_reg__, R21, R30, R31 |
| 2672 |
|
| 2673 |
#if (defined (L_fload_1) \ |
| 2674 |
|| defined (L_fload_2) \ |
| 2675 |
|| defined (L_fload_3) \ |
| 2676 |
|| defined (L_fload_4)) |
| 2677 |
|
| 2678 |
;; Destination |
| 2679 |
#define D0 22 |
| 2680 |
#define D1 D0+1 |
| 2681 |
#define D2 D0+2 |
| 2682 |
#define D3 D0+3 |
| 2683 |
|
| 2684 |
;; Register containing bits 16+ of the address |
| 2685 |
|
| 2686 |
#define HHI8 21 |
| 2687 |
|
| 2688 |
.macro .fload dest, n |
| 2689 |
#if defined (__AVR_HAVE_ELPM__) |
| 2690 |
.if \dest == D0 |
| 2691 |
out __RAMPZ__, HHI8 |
| 2692 |
.endif |
| 2693 |
#endif /* __AVR_HAVE_ELPM__ */ |
| 2694 |
#if defined (__AVR_HAVE_ELPMX__) |
| 2695 |
elpm \dest, Z+ |
| 2696 |
#elif defined (__AVR_HAVE_ELPM__) |
| 2697 |
elpm |
| 2698 |
mov \dest, r0 |
| 2699 |
.if \dest != D0+\n-1 |
| 2700 |
adiw r30, 1 |
| 2701 |
adc HHI8, __zero_reg__ |
| 2702 |
out __RAMPZ__, HHI8 |
| 2703 |
.endif |
| 2704 |
#elif defined (__AVR_HAVE_LPMX__) |
| 2705 |
lpm \dest, Z+ |
| 2706 |
#else |
| 2707 |
lpm |
| 2708 |
mov \dest, r0 |
| 2709 |
.if \dest != D0+\n-1 |
| 2710 |
adiw r30, 1 |
| 2711 |
.endif |
| 2712 |
#endif |
| 2713 |
#if defined (__AVR_HAVE_ELPM__) && defined (__AVR_HAVE_RAMPD__) |
| 2714 |
.if \dest == D0+\n-1 |
| 2715 |
;; Reset RAMPZ to 0 so that EBI devices don't read garbage from RAM |
| 2716 |
out __RAMPZ__, __zero_reg__ |
| 2717 |
.endif |
| 2718 |
#endif |
| 2719 |
.endm ; .fload |
| 2720 |
|
| 2721 |
#if defined (L_fload_1) |
| 2722 |
DEFUN __fload_1 |
| 2723 |
#if defined (__AVR_HAVE_LPMX__) && !defined (__AVR_HAVE_ELPM__) |
| 2724 |
lpm D0, Z |
| 2725 |
ret |
| 2726 |
#else |
| 2727 |
.fload D0, 1 |
| 2728 |
ret |
| 2729 |
#endif /* LPMx && ! ELPM */ |
| 2730 |
ENDF __fload_1 |
| 2731 |
#endif /* L_fload_1 */ |
| 2732 |
|
| 2733 |
#if defined (L_fload_2) |
| 2734 |
DEFUN __fload_2 |
| 2735 |
.fload D0, 2 |
| 2736 |
.fload D1, 2 |
| 2737 |
ret |
| 2738 |
ENDF __fload_2 |
| 2739 |
#endif /* L_fload_2 */ |
| 2740 |
|
| 2741 |
#if defined (L_fload_3) |
| 2742 |
DEFUN __fload_3 |
| 2743 |
.fload D0, 3 |
| 2744 |
.fload D1, 3 |
| 2745 |
.fload D2, 3 |
| 2746 |
ret |
| 2747 |
ENDF __fload_3 |
| 2748 |
#endif /* L_fload_3 */ |
| 2749 |
|
| 2750 |
#if defined (L_fload_4) |
| 2751 |
DEFUN __fload_4 |
| 2752 |
.fload D0, 4 |
| 2753 |
.fload D1, 4 |
| 2754 |
.fload D2, 4 |
| 2755 |
.fload D3, 4 |
| 2756 |
ret |
| 2757 |
ENDF __fload_4 |
| 2758 |
#endif /* L_fload_4 */ |
| 2759 |
|
| 2760 |
#endif /* L_fload_{1|2|3|4} */ |
| 2761 |
#endif /* if !defined (__AVR_TINY__) */ |
| 2762 |
|
| 2763 |
|
| 2764 |
#if !defined (__AVR_TINY__) |
| 2765 |
#if defined (L_strlen_memx) |
| 2766 |
DEFUN __strlen_memx |
| 2767 |
#ifdef __AVR_ERRATA_SKIP_JMP_CALL__ |
| 2768 |
tst r24 |
| 2769 |
brmi 1f |
| 2770 |
#else |
| 2771 |
sbrs r24, 7 |
| 2772 |
#endif |
| 2773 |
XJMP strlen_PF ; AVR-LibC |
| 2774 |
1: wmov 24, 22 |
| 2775 |
XJMP strlen ; AVR-LibC |
| 2776 |
ENDF __strlen_memx |
| 2777 |
#endif /* L_strlen_memx */ |
| 2778 |
#endif /* if !defined (__AVR_TINY__) */ |
| 2779 |
|
| 2780 |
|
| 2781 |
#if !defined (__AVR_TINY__) |
| 2782 |
;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;; |
| 2783 |
;; memcopy from Address Space __memx to RAM |
| 2784 |
;; R23:Z = Source Address |
| 2785 |
;; X = Destination Address |
| 2786 |
;; Clobbers: __tmp_reg__, R23, R24, R25, X, Z |
| 2787 |
|
| 2788 |
#if defined (L_movmemx) |
| 2789 |
|
| 2790 |
#define HHI8 23 |
| 2791 |
#define LOOP 24 |
| 2792 |
|
| 2793 |
DEFUN __movmemx_qi |
| 2794 |
;; #Bytes to copy fits in 8 Bits (1..255) |
| 2795 |
;; Zero-extend Loop Counter |
| 2796 |
clr LOOP+1 |
| 2797 |
;; FALLTHRU |
| 2798 |
ENDF __movmemx_qi |
| 2799 |
|
| 2800 |
DEFUN __movmemx_hi |
| 2801 |
|
| 2802 |
.branch_plus HHI8, __movmemf_hi |
| 2803 |
|
| 2804 |
;; Read 1 Byte from RAM... |
| 2805 |
1: ld r0, Z+ |
| 2806 |
;; and store that Byte to RAM Destination |
| 2807 |
st X+, r0 |
| 2808 |
sbiw LOOP, 1 |
| 2809 |
brne 1b |
| 2810 |
ret |
| 2811 |
ENDF __movmemx_hi |
| 2812 |
|
| 2813 |
#undef HHI8 |
| 2814 |
#undef LOOP |
| 2815 |
|
| 2816 |
#endif /* L_movmemx */ |
| 2817 |
|
| 2818 |
;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;; |
| 2819 |
;; memcopy from Address Space __flashx to RAM |
| 2820 |
;; R23:Z = Source Address |
| 2821 |
;; X = Destination Address |
| 2822 |
;; Clobbers: __tmp_reg__, R23, R24, R25, X, Z |
| 2823 |
|
| 2824 |
#if defined (L_movmemf) |
| 2825 |
|
| 2826 |
#define HHI8 23 |
| 2827 |
#define LOOP 24 |
| 2828 |
|
| 2829 |
DEFUN __movmemf_qi |
| 2830 |
;; #Bytes to copy fits in 8 Bits (1..255) |
| 2831 |
;; Zero-extend Loop Counter |
| 2832 |
clr LOOP+1 |
| 2833 |
;; FALLTHRU |
| 2834 |
ENDF __movmemf_qi |
| 2835 |
|
| 2836 |
DEFUN __movmemf_hi |
| 2837 |
|
| 2838 |
;; Read from Flash |
| 2839 |
|
| 2840 |
#if defined (__AVR_HAVE_ELPM__) |
| 2841 |
out __RAMPZ__, HHI8 |
| 2842 |
#endif |
| 2843 |
|
| 2844 |
0: ;; Load 1 Byte from Flash... |
| 2845 |
|
| 2846 |
#if defined (__AVR_HAVE_ELPMX__) |
| 2847 |
elpm r0, Z+ |
| 2848 |
#elif defined (__AVR_HAVE_ELPM__) |
| 2849 |
elpm |
| 2850 |
adiw r30, 1 |
| 2851 |
adc HHI8, __zero_reg__ |
| 2852 |
out __RAMPZ__, HHI8 |
| 2853 |
#elif defined (__AVR_HAVE_LPMX__) |
| 2854 |
lpm r0, Z+ |
| 2855 |
#else |
| 2856 |
lpm |
| 2857 |
adiw r30, 1 |
| 2858 |
#endif |
| 2859 |
|
| 2860 |
;; ...and store that Byte to RAM Destination |
| 2861 |
st X+, r0 |
| 2862 |
sbiw LOOP, 1 |
| 2863 |
brne 0b |
| 2864 |
#if defined (__AVR_HAVE_ELPM__) && defined (__AVR_HAVE_RAMPD__) |
| 2865 |
;; Reset RAMPZ to 0 so that EBI devices don't read garbage from RAM |
| 2866 |
out __RAMPZ__, __zero_reg__ |
| 2867 |
#endif /* ELPM && RAMPD */ |
| 2868 |
ret |
| 2869 |
ENDF __movmemf_hi |
| 2870 |
|
| 2871 |
#undef HHI8 |
| 2872 |
#undef LOOP |
| 2873 |
|
| 2874 |
#endif /* L_movmemf */ |
| 2875 |
#endif /* !defined (__AVR_TINY__) */ |
| 2876 |
|
| 2877 |
|
| 2878 |
.section .text.libgcc.builtins, "ax", @progbits |
| 2879 |
|
| 2880 |
/********************************** |
| 2881 |
* Find first set Bit (ffs) |
| 2882 |
**********************************/ |
| 2883 |
|
| 2884 |
#if defined (L_ffssi2) |
| 2885 |
;; find first set bit |
| 2886 |
;; r25:r24 = ffs32 (r25:r22) |
| 2887 |
;; clobbers: r22, r26 |
| 2888 |
DEFUN __ffssi2 |
| 2889 |
clr r26 |
| 2890 |
tst r22 |
| 2891 |
brne 1f |
| 2892 |
subi r26, -8 |
| 2893 |
or r22, r23 |
| 2894 |
brne 1f |
| 2895 |
subi r26, -8 |
| 2896 |
or r22, r24 |
| 2897 |
brne 1f |
| 2898 |
subi r26, -8 |
| 2899 |
or r22, r25 |
| 2900 |
brne 1f |
| 2901 |
ret |
| 2902 |
1: mov r24, r22 |
| 2903 |
XJMP __loop_ffsqi2 |
| 2904 |
ENDF __ffssi2 |
| 2905 |
#endif /* defined (L_ffssi2) */ |
| 2906 |
|
| 2907 |
#if defined (L_ffshi2) |
| 2908 |
;; find first set bit |
| 2909 |
;; r25:r24 = ffs16 (r25:r24) |
| 2910 |
;; clobbers: r26 |
| 2911 |
DEFUN __ffshi2 |
| 2912 |
clr r26 |
| 2913 |
#ifdef __AVR_ERRATA_SKIP_JMP_CALL__ |
| 2914 |
;; Some cores have problem skipping 2-word instruction |
| 2915 |
tst r24 |
| 2916 |
breq 2f |
| 2917 |
#else |
| 2918 |
cpse r24, __zero_reg__ |
| 2919 |
#endif /* __AVR_HAVE_JMP_CALL__ */ |
| 2920 |
1: XJMP __loop_ffsqi2 |
| 2921 |
2: ldi r26, 8 |
| 2922 |
or r24, r25 |
| 2923 |
brne 1b |
| 2924 |
ret |
| 2925 |
ENDF __ffshi2 |
| 2926 |
#endif /* defined (L_ffshi2) */ |
| 2927 |
|
| 2928 |
#if defined (L_loop_ffsqi2) |
| 2929 |
;; Helper for ffshi2, ffssi2 |
| 2930 |
;; r25:r24 = r26 + zero_extend16 (ffs8(r24)) |
| 2931 |
;; r24 must be != 0 |
| 2932 |
;; clobbers: r26 |
| 2933 |
DEFUN __loop_ffsqi2 |
| 2934 |
inc r26 |
| 2935 |
lsr r24 |
| 2936 |
brcc __loop_ffsqi2 |
| 2937 |
mov r24, r26 |
| 2938 |
clr r25 |
| 2939 |
ret |
| 2940 |
ENDF __loop_ffsqi2 |
| 2941 |
#endif /* defined (L_loop_ffsqi2) */ |
| 2942 |
|
| 2943 |
|
| 2944 |
/********************************** |
| 2945 |
* Count trailing Zeros (ctz) |
| 2946 |
**********************************/ |
| 2947 |
|
| 2948 |
#if defined (L_ctzsi2) |
| 2949 |
;; count trailing zeros |
| 2950 |
;; r25:r24 = ctz32 (r25:r22) |
| 2951 |
;; clobbers: r26, r22 |
| 2952 |
;; ctz(0) = 255 |
| 2953 |
;; Note that ctz(0) in undefined for GCC |
| 2954 |
DEFUN __ctzsi2 |
| 2955 |
XCALL __ffssi2 |
| 2956 |
dec r24 |
| 2957 |
ret |
| 2958 |
ENDF __ctzsi2 |
| 2959 |
#endif /* defined (L_ctzsi2) */ |
| 2960 |
|
| 2961 |
#if defined (L_ctzhi2) |
| 2962 |
;; count trailing zeros |
| 2963 |
;; r25:r24 = ctz16 (r25:r24) |
| 2964 |
;; clobbers: r26 |
| 2965 |
;; ctz(0) = 255 |
| 2966 |
;; Note that ctz(0) in undefined for GCC |
| 2967 |
DEFUN __ctzhi2 |
| 2968 |
XCALL __ffshi2 |
| 2969 |
dec r24 |
| 2970 |
ret |
| 2971 |
ENDF __ctzhi2 |
| 2972 |
#endif /* defined (L_ctzhi2) */ |
| 2973 |
|
| 2974 |
|
| 2975 |
/********************************** |
| 2976 |
* Count leading Zeros (clz) |
| 2977 |
**********************************/ |
| 2978 |
|
| 2979 |
#if defined (L_clzdi2) |
| 2980 |
;; count leading zeros |
| 2981 |
;; r25:r24 = clz64 (r25:r18) |
| 2982 |
;; clobbers: r22, r23, r26 |
| 2983 |
DEFUN __clzdi2 |
| 2984 |
XCALL __clzsi2 |
| 2985 |
sbrs r24, 5 |
| 2986 |
ret |
| 2987 |
mov_l r22, r18 |
| 2988 |
mov_h r23, r19 |
| 2989 |
mov_l r24, r20 |
| 2990 |
mov_h r25, r21 |
| 2991 |
XCALL __clzsi2 |
| 2992 |
subi r24, -32 |
| 2993 |
ret |
| 2994 |
ENDF __clzdi2 |
| 2995 |
#endif /* defined (L_clzdi2) */ |
| 2996 |
|
| 2997 |
#if defined (L_clzsi2) |
| 2998 |
;; count leading zeros |
| 2999 |
;; r25:r24 = clz32 (r25:r22) |
| 3000 |
;; clobbers: r26 |
| 3001 |
DEFUN __clzsi2 |
| 3002 |
XCALL __clzhi2 |
| 3003 |
sbrs r24, 4 |
| 3004 |
ret |
| 3005 |
mov_l r24, r22 |
| 3006 |
mov_h r25, r23 |
| 3007 |
XCALL __clzhi2 |
| 3008 |
subi r24, -16 |
| 3009 |
ret |
| 3010 |
ENDF __clzsi2 |
| 3011 |
#endif /* defined (L_clzsi2) */ |
| 3012 |
|
| 3013 |
#if defined (L_clzhi2) |
| 3014 |
;; count leading zeros |
| 3015 |
;; r25:r24 = clz16 (r25:r24) |
| 3016 |
;; clobbers: r26 |
| 3017 |
DEFUN __clzhi2 |
| 3018 |
clr r26 |
| 3019 |
tst r25 |
| 3020 |
brne 1f |
| 3021 |
or r25, r24 |
| 3022 |
breq 0f |
| 3023 |
subi r26, -8 |
| 3024 |
1: cpi r25, 16 |
| 3025 |
brsh 3f |
| 3026 |
subi r26, -3 |
| 3027 |
swap r25 |
| 3028 |
2: inc r26 |
| 3029 |
3: lsl r25 |
| 3030 |
brcc 2b |
| 3031 |
mov r24, r26 |
| 3032 |
clr r25 |
| 3033 |
ret |
| 3034 |
0: ldi r24, 16 |
| 3035 |
ret |
| 3036 |
ENDF __clzhi2 |
| 3037 |
#endif /* defined (L_clzhi2) */ |
| 3038 |
|
| 3039 |
|
| 3040 |
/********************************** |
| 3041 |
* Parity |
| 3042 |
**********************************/ |
| 3043 |
|
| 3044 |
#if defined (L_paritydi2) |
| 3045 |
;; r25:r24 = parity64 (r25:r18) |
| 3046 |
;; clobbers: __tmp_reg__ |
| 3047 |
DEFUN __paritydi2 |
| 3048 |
eor r24, r18 |
| 3049 |
eor r24, r19 |
| 3050 |
eor r24, r20 |
| 3051 |
eor r24, r21 |
| 3052 |
XJMP __paritysi2 |
| 3053 |
ENDF __paritydi2 |
| 3054 |
#endif /* defined (L_paritydi2) */ |
| 3055 |
|
| 3056 |
#if defined (L_paritysi2) |
| 3057 |
;; r25:r24 = parity32 (r25:r22) |
| 3058 |
;; clobbers: __tmp_reg__ |
| 3059 |
DEFUN __paritysi2 |
| 3060 |
eor r24, r22 |
| 3061 |
eor r24, r23 |
| 3062 |
XJMP __parityhi2 |
| 3063 |
ENDF __paritysi2 |
| 3064 |
#endif /* defined (L_paritysi2) */ |
| 3065 |
|
| 3066 |
#if defined (L_parityhi2) |
| 3067 |
;; r25:r24 = parity16 (r25:r24) |
| 3068 |
;; clobbers: __tmp_reg__ |
| 3069 |
DEFUN __parityhi2 |
| 3070 |
eor r24, r25 |
| 3071 |
;; FALLTHRU |
| 3072 |
ENDF __parityhi2 |
| 3073 |
|
| 3074 |
;; r25:r24 = parity8 (r24) |
| 3075 |
;; clobbers: __tmp_reg__ |
| 3076 |
DEFUN __parityqi2 |
| 3077 |
;; parity is in r24[0..7] |
| 3078 |
mov __tmp_reg__, r24 |
| 3079 |
swap __tmp_reg__ |
| 3080 |
eor r24, __tmp_reg__ |
| 3081 |
;; parity is in r24[0..3] |
| 3082 |
subi r24, -4 |
| 3083 |
andi r24, -5 |
| 3084 |
subi r24, -6 |
| 3085 |
;; parity is in r24[0,3] |
| 3086 |
sbrc r24, 3 |
| 3087 |
inc r24 |
| 3088 |
;; parity is in r24[0] |
| 3089 |
andi r24, 1 |
| 3090 |
clr r25 |
| 3091 |
ret |
| 3092 |
ENDF __parityqi2 |
| 3093 |
#endif /* defined (L_parityhi2) */ |
| 3094 |
|
| 3095 |
|
| 3096 |
/********************************** |
| 3097 |
* Population Count |
| 3098 |
**********************************/ |
| 3099 |
|
| 3100 |
#if defined (L_popcounthi2) |
| 3101 |
;; population count |
| 3102 |
;; r25:r24 = popcount16 (r25:r24) |
| 3103 |
;; clobbers: __tmp_reg__ |
| 3104 |
DEFUN __popcounthi2 |
| 3105 |
XCALL __popcountqi2 |
| 3106 |
push r24 |
| 3107 |
mov r24, r25 |
| 3108 |
XCALL __popcountqi2 |
| 3109 |
clr r25 |
| 3110 |
;; FALLTHRU |
| 3111 |
ENDF __popcounthi2 |
| 3112 |
|
| 3113 |
DEFUN __popcounthi2_tail |
| 3114 |
pop __tmp_reg__ |
| 3115 |
add r24, __tmp_reg__ |
| 3116 |
ret |
| 3117 |
ENDF __popcounthi2_tail |
| 3118 |
#endif /* defined (L_popcounthi2) */ |
| 3119 |
|
| 3120 |
#if defined (L_popcountsi2) |
| 3121 |
;; population count |
| 3122 |
;; r25:r24 = popcount32 (r25:r22) |
| 3123 |
;; clobbers: __tmp_reg__ |
| 3124 |
DEFUN __popcountsi2 |
| 3125 |
XCALL __popcounthi2 |
| 3126 |
push r24 |
| 3127 |
mov_l r24, r22 |
| 3128 |
mov_h r25, r23 |
| 3129 |
XCALL __popcounthi2 |
| 3130 |
XJMP __popcounthi2_tail |
| 3131 |
ENDF __popcountsi2 |
| 3132 |
#endif /* defined (L_popcountsi2) */ |
| 3133 |
|
| 3134 |
#if defined (L_popcountdi2) |
| 3135 |
;; population count |
| 3136 |
;; r25:r24 = popcount64 (r25:r18) |
| 3137 |
;; clobbers: r22, r23, __tmp_reg__ |
| 3138 |
DEFUN __popcountdi2 |
| 3139 |
XCALL __popcountsi2 |
| 3140 |
push r24 |
| 3141 |
mov_l r22, r18 |
| 3142 |
mov_h r23, r19 |
| 3143 |
mov_l r24, r20 |
| 3144 |
mov_h r25, r21 |
| 3145 |
XCALL __popcountsi2 |
| 3146 |
XJMP __popcounthi2_tail |
| 3147 |
ENDF __popcountdi2 |
| 3148 |
#endif /* defined (L_popcountdi2) */ |
| 3149 |
|
| 3150 |
#if defined (L_popcountqi2) |
| 3151 |
;; population count |
| 3152 |
;; r24 = popcount8 (r24) |
| 3153 |
;; clobbers: __tmp_reg__ |
| 3154 |
DEFUN __popcountqi2 |
| 3155 |
mov __tmp_reg__, r24 ; oeoeoeoe |
| 3156 |
andi r24, 0xAA ; o0o0o0o0 |
| 3157 |
lsr r24 ; 0o0o0o0o |
| 3158 |
;; Four values 0, 1 or 2: # bits set o+e |
| 3159 |
sub __tmp_reg__, r24 ; 44332211 |
| 3160 |
mov r24, __tmp_reg__ ; 44332211 |
| 3161 |
andi r24, 0x33 ; 00330011 |
| 3162 |
eor __tmp_reg__, r24 ; 44002200 |
| 3163 |
lsr __tmp_reg__ ; 04400220 |
| 3164 |
lsr __tmp_reg__ ; 00440022 |
| 3165 |
add r24, __tmp_reg__ ; 04210421 |
| 3166 |
mov __tmp_reg__, r24 ; h421l421 |
| 3167 |
swap __tmp_reg__ ; l421h421 |
| 3168 |
add r24, __tmp_reg__ ; 84218421 |
| 3169 |
andi r24, 0xf ; 00008421 /17 |
| 3170 |
ret |
| 3171 |
ENDF __popcountqi2 |
| 3172 |
#endif /* defined (L_popcountqi2) */ |
| 3173 |
|
| 3174 |
|
| 3175 |
/********************************** |
| 3176 |
* Swap bytes |
| 3177 |
**********************************/ |
| 3178 |
|
| 3179 |
;; swap two registers with different register number |
| 3180 |
.macro bswap a, b |
| 3181 |
eor \a, \b |
| 3182 |
eor \b, \a |
| 3183 |
eor \a, \b |
| 3184 |
.endm |
| 3185 |
|
| 3186 |
#if defined (L_bswapsi2) |
| 3187 |
;; swap bytes |
| 3188 |
;; r25:r22 = bswap32 (r25:r22) |
| 3189 |
DEFUN __bswapsi2 |
| 3190 |
bswap r22, r25 |
| 3191 |
bswap r23, r24 |
| 3192 |
ret |
| 3193 |
ENDF __bswapsi2 |
| 3194 |
#endif /* defined (L_bswapsi2) */ |
| 3195 |
|
| 3196 |
#if defined (L_bswapdi2) |
| 3197 |
;; swap bytes |
| 3198 |
;; r25:r18 = bswap64 (r25:r18) |
| 3199 |
DEFUN __bswapdi2 |
| 3200 |
bswap r18, r25 |
| 3201 |
bswap r19, r24 |
| 3202 |
bswap r20, r23 |
| 3203 |
bswap r21, r22 |
| 3204 |
ret |
| 3205 |
ENDF __bswapdi2 |
| 3206 |
#endif /* defined (L_bswapdi2) */ |
| 3207 |
|
| 3208 |
|
| 3209 |
/********************************** |
| 3210 |
* 64-bit shifts |
| 3211 |
**********************************/ |
| 3212 |
|
| 3213 |
#if defined (L_ashrdi3) |
| 3214 |
|
| 3215 |
#define SS __zero_reg__ |
| 3216 |
|
| 3217 |
;; Arithmetic shift right |
| 3218 |
;; r25:r18 = ashr64 (r25:r18, r17:r16) |
| 3219 |
DEFUN __ashrdi3 |
| 3220 |
sbrc r25, 7 |
| 3221 |
com SS |
| 3222 |
;; FALLTHRU |
| 3223 |
ENDF __ashrdi3 |
| 3224 |
|
| 3225 |
;; Logic shift right |
| 3226 |
;; r25:r18 = lshr64 (r25:r18, r17:r16) |
| 3227 |
DEFUN __lshrdi3 |
| 3228 |
;; Signs are in SS (zero_reg) |
| 3229 |
mov __tmp_reg__, r16 |
| 3230 |
0: cpi r16, 8 |
| 3231 |
brlo 2f |
| 3232 |
subi r16, 8 |
| 3233 |
mov r18, r19 |
| 3234 |
mov r19, r20 |
| 3235 |
mov r20, r21 |
| 3236 |
mov r21, r22 |
| 3237 |
mov r22, r23 |
| 3238 |
mov r23, r24 |
| 3239 |
mov r24, r25 |
| 3240 |
mov r25, SS |
| 3241 |
rjmp 0b |
| 3242 |
1: asr SS |
| 3243 |
ror r25 |
| 3244 |
ror r24 |
| 3245 |
ror r23 |
| 3246 |
ror r22 |
| 3247 |
ror r21 |
| 3248 |
ror r20 |
| 3249 |
ror r19 |
| 3250 |
ror r18 |
| 3251 |
2: dec r16 |
| 3252 |
brpl 1b |
| 3253 |
clr __zero_reg__ |
| 3254 |
mov r16, __tmp_reg__ |
| 3255 |
ret |
| 3256 |
ENDF __lshrdi3 |
| 3257 |
|
| 3258 |
#undef SS |
| 3259 |
|
| 3260 |
#endif /* defined (L_ashrdi3) */ |
| 3261 |
|
| 3262 |
#if defined (L_ashldi3) |
| 3263 |
;; Shift left |
| 3264 |
;; r25:r18 = ashl64 (r25:r18, r17:r16) |
| 3265 |
;; This function does not clobber T. |
| 3266 |
DEFUN __ashldi3 |
| 3267 |
mov __tmp_reg__, r16 |
| 3268 |
0: cpi r16, 8 |
| 3269 |
brlo 2f |
| 3270 |
mov r25, r24 |
| 3271 |
mov r24, r23 |
| 3272 |
mov r23, r22 |
| 3273 |
mov r22, r21 |
| 3274 |
mov r21, r20 |
| 3275 |
mov r20, r19 |
| 3276 |
mov r19, r18 |
| 3277 |
clr r18 |
| 3278 |
subi r16, 8 |
| 3279 |
rjmp 0b |
| 3280 |
1: lsl r18 |
| 3281 |
rol r19 |
| 3282 |
rol r20 |
| 3283 |
rol r21 |
| 3284 |
rol r22 |
| 3285 |
rol r23 |
| 3286 |
rol r24 |
| 3287 |
rol r25 |
| 3288 |
2: dec r16 |
| 3289 |
brpl 1b |
| 3290 |
mov r16, __tmp_reg__ |
| 3291 |
ret |
| 3292 |
ENDF __ashldi3 |
| 3293 |
#endif /* defined (L_ashldi3) */ |
| 3294 |
|
| 3295 |
#if defined (L_rotldi3) |
| 3296 |
;; Rotate left |
| 3297 |
;; r25:r18 = rotl64 (r25:r18, r17:r16) |
| 3298 |
DEFUN __rotldi3 |
| 3299 |
push r16 |
| 3300 |
0: cpi r16, 8 |
| 3301 |
brlo 2f |
| 3302 |
subi r16, 8 |
| 3303 |
mov __tmp_reg__, r25 |
| 3304 |
mov r25, r24 |
| 3305 |
mov r24, r23 |
| 3306 |
mov r23, r22 |
| 3307 |
mov r22, r21 |
| 3308 |
mov r21, r20 |
| 3309 |
mov r20, r19 |
| 3310 |
mov r19, r18 |
| 3311 |
mov r18, __tmp_reg__ |
| 3312 |
rjmp 0b |
| 3313 |
1: lsl r18 |
| 3314 |
rol r19 |
| 3315 |
rol r20 |
| 3316 |
rol r21 |
| 3317 |
rol r22 |
| 3318 |
rol r23 |
| 3319 |
rol r24 |
| 3320 |
rol r25 |
| 3321 |
adc r18, __zero_reg__ |
| 3322 |
2: dec r16 |
| 3323 |
brpl 1b |
| 3324 |
pop r16 |
| 3325 |
ret |
| 3326 |
ENDF __rotldi3 |
| 3327 |
#endif /* defined (L_rotldi3) */ |
| 3328 |
|
| 3329 |
|
| 3330 |
.section .text.libgcc.fmul, "ax", @progbits |
| 3331 |
|
| 3332 |
/***********************************************************/ |
| 3333 |
;;; Softmul versions of FMUL, FMULS and FMULSU to implement |
| 3334 |
;;; __builtin_avr_fmul* if !AVR_HAVE_MUL |
| 3335 |
/***********************************************************/ |
| 3336 |
|
| 3337 |
#define A1 24 |
| 3338 |
#define B1 25 |
| 3339 |
#define C0 22 |
| 3340 |
#define C1 23 |
| 3341 |
#define A0 __tmp_reg__ |
| 3342 |
|
| 3343 |
#ifdef L_fmuls |
| 3344 |
;;; r23:r22 = fmuls (r24, r25) like in FMULS instruction |
| 3345 |
;;; Clobbers: r24, r25, __tmp_reg__ |
| 3346 |
DEFUN __fmuls |
| 3347 |
;; A0.7 = negate result? |
| 3348 |
mov A0, A1 |
| 3349 |
eor A0, B1 |
| 3350 |
;; B1 = |B1| |
| 3351 |
sbrc B1, 7 |
| 3352 |
neg B1 |
| 3353 |
XJMP __fmulsu_exit |
| 3354 |
ENDF __fmuls |
| 3355 |
#endif /* L_fmuls */ |
| 3356 |
|
| 3357 |
#ifdef L_fmulsu |
| 3358 |
;;; r23:r22 = fmulsu (r24, r25) like in FMULSU instruction |
| 3359 |
;;; Clobbers: r24, r25, __tmp_reg__ |
| 3360 |
DEFUN __fmulsu |
| 3361 |
;; A0.7 = negate result? |
| 3362 |
mov A0, A1 |
| 3363 |
;; FALLTHRU |
| 3364 |
ENDF __fmulsu |
| 3365 |
|
| 3366 |
;; Helper for __fmuls and __fmulsu |
| 3367 |
DEFUN __fmulsu_exit |
| 3368 |
;; A1 = |A1| |
| 3369 |
sbrc A1, 7 |
| 3370 |
neg A1 |
| 3371 |
#ifdef __AVR_ERRATA_SKIP_JMP_CALL__ |
| 3372 |
;; Some cores have problem skipping 2-word instruction |
| 3373 |
tst A0 |
| 3374 |
brmi 1f |
| 3375 |
#else |
| 3376 |
sbrs A0, 7 |
| 3377 |
#endif /* __AVR_HAVE_JMP_CALL__ */ |
| 3378 |
XJMP __fmul |
| 3379 |
1: XCALL __fmul |
| 3380 |
;; C = -C iff A0.7 = 1 |
| 3381 |
NEG2 C0 |
| 3382 |
ret |
| 3383 |
ENDF __fmulsu_exit |
| 3384 |
#endif /* L_fmulsu */ |
| 3385 |
|
| 3386 |
|
| 3387 |
#ifdef L_fmul |
| 3388 |
;;; r22:r23 = fmul (r24, r25) like in FMUL instruction |
| 3389 |
;;; Clobbers: r24, r25, __tmp_reg__ |
| 3390 |
DEFUN __fmul |
| 3391 |
; clear result |
| 3392 |
clr C0 |
| 3393 |
clr C1 |
| 3394 |
clr A0 |
| 3395 |
1: tst B1 |
| 3396 |
;; 1.0 = 0x80, so test for bit 7 of B to see if A must to be added to C. |
| 3397 |
2: brpl 3f |
| 3398 |
;; C += A |
| 3399 |
add C0, A0 |
| 3400 |
adc C1, A1 |
| 3401 |
3: ;; A >>= 1 |
| 3402 |
lsr A1 |
| 3403 |
ror A0 |
| 3404 |
;; B <<= 1 |
| 3405 |
lsl B1 |
| 3406 |
brne 2b |
| 3407 |
ret |
| 3408 |
ENDF __fmul |
| 3409 |
#endif /* L_fmul */ |
| 3410 |
|
| 3411 |
#undef A0 |
| 3412 |
#undef A1 |
| 3413 |
#undef B1 |
| 3414 |
#undef C0 |
| 3415 |
#undef C1 |
| 3416 |
|
| 3417 |
|
| 3418 |
|
| 3419 |
/********************************** |
| 3420 |
* Floating-Point |
| 3421 |
**********************************/ |
| 3422 |
|
| 3423 |
#if defined (L_powif) |
| 3424 |
#ifndef __AVR_TINY__ |
| 3425 |
|
| 3426 |
;; float output and arg #1 |
| 3427 |
#define A0 22 |
| 3428 |
#define A1 A0 + 1 |
| 3429 |
#define A2 A0 + 2 |
| 3430 |
#define A3 A0 + 3 |
| 3431 |
|
| 3432 |
;; float arg #2 |
| 3433 |
#define B0 18 |
| 3434 |
#define B1 B0 + 1 |
| 3435 |
#define B2 B0 + 2 |
| 3436 |
#define B3 B0 + 3 |
| 3437 |
|
| 3438 |
;; float X: input and iterated squares |
| 3439 |
#define X0 10 |
| 3440 |
#define X1 X0 + 1 |
| 3441 |
#define X2 X0 + 2 |
| 3442 |
#define X3 X0 + 3 |
| 3443 |
|
| 3444 |
;; float Y: expand result |
| 3445 |
#define Y0 14 |
| 3446 |
#define Y1 Y0 + 1 |
| 3447 |
#define Y2 Y0 + 2 |
| 3448 |
#define Y3 Y0 + 3 |
| 3449 |
|
| 3450 |
;; .7 = Sign of I. |
| 3451 |
;; .0 == 0 => Y = 1.0f implicitly. |
| 3452 |
#define Flags R9 |
| 3453 |
#define Y_set 0 |
| 3454 |
|
| 3455 |
;;; Integer exponent input. |
| 3456 |
#define I0 28 |
| 3457 |
#define I1 I0+1 |
| 3458 |
|
| 3459 |
#define ONE 0x3f800000 |
| 3460 |
|
| 3461 |
DEFUN __powisf2 |
| 3462 |
;; Save 11 Registers: R9...R17, R28, R29 |
| 3463 |
do_prologue_saves 11 |
| 3464 |
|
| 3465 |
;; Fill local vars with input parameters. |
| 3466 |
wmov I0, 20 |
| 3467 |
mov4 X0, A0 |
| 3468 |
;; Save sign of exponent for later. |
| 3469 |
mov Flags, I1 |
| 3470 |
;; I := abs (I) |
| 3471 |
tst I1 |
| 3472 |
brpl 1f |
| 3473 |
NEG2 I0 |
| 3474 |
1: |
| 3475 |
;; Y := (I % 2) ? X : 1.0f |
| 3476 |
;; (When we come from below, this is like SET, i.e. Flags.Y_set := 1). |
| 3477 |
bst I0, 0 |
| 3478 |
;; Flags.Y_set = false means that we have to assume Y = 1.0f below. |
| 3479 |
bld Flags, Y_set |
| 3480 |
2: ;; We have A == X when we come from above. |
| 3481 |
mov4 Y0, A0 |
| 3482 |
|
| 3483 |
.Loop: |
| 3484 |
;; while (I >>= 1) |
| 3485 |
lsr I1 |
| 3486 |
ror I0 |
| 3487 |
sbiw I0, 0 |
| 3488 |
breq .Loop_done |
| 3489 |
|
| 3490 |
;; X := X * X |
| 3491 |
mov4 A0, X0 |
| 3492 |
#ifdef __WITH_AVRLIBC__ |
| 3493 |
XCALL squaref |
| 3494 |
#else |
| 3495 |
mov4 B0, X0 |
| 3496 |
XCALL __mulsf3 |
| 3497 |
#endif /* Have AVR-LibC? */ |
| 3498 |
mov4 X0, A0 |
| 3499 |
|
| 3500 |
;; if (I % 2 == 1) Y := Y * X |
| 3501 |
bst I0, 0 |
| 3502 |
brtc .Loop |
| 3503 |
bst Flags, Y_set |
| 3504 |
;; When Y is not set => Y := Y * X = 1.0f * X (= A) |
| 3505 |
;; Plus, we have to set Y_set = 1 (= I0.0) |
| 3506 |
brtc 1b |
| 3507 |
;; Y is already set: Y := X * Y (= A * Y) |
| 3508 |
mov4 B0, Y0 |
| 3509 |
XCALL __mulsf3 |
| 3510 |
rjmp 2b |
| 3511 |
|
| 3512 |
;; End while |
| 3513 |
.Loop_done: |
| 3514 |
|
| 3515 |
;; A := 1.0f |
| 3516 |
ldi A3, hhi8(ONE) |
| 3517 |
ldi A2, hlo8(ONE) |
| 3518 |
ldi A1, hi8(ONE) |
| 3519 |
ldi A0, lo8(ONE) |
| 3520 |
|
| 3521 |
;; When Y is still not set, the result is 1.0f (= A). |
| 3522 |
bst Flags, Y_set |
| 3523 |
brtc .Lret |
| 3524 |
|
| 3525 |
;; if (I was < 0) Y = 1.0f / Y |
| 3526 |
tst Flags |
| 3527 |
brmi 1f |
| 3528 |
;; A := Y |
| 3529 |
mov4 A0, Y0 |
| 3530 |
rjmp .Lret |
| 3531 |
1: ;; A := 1 / Y = A / Y |
| 3532 |
mov4 B0, Y0 |
| 3533 |
XCALL __divsf3 |
| 3534 |
|
| 3535 |
.Lret: |
| 3536 |
do_epilogue_restores 11 |
| 3537 |
ENDF __powisf2 |
| 3538 |
|
| 3539 |
#undef A0 |
| 3540 |
#undef A1 |
| 3541 |
#undef A2 |
| 3542 |
#undef A3 |
| 3543 |
|
| 3544 |
#undef B0 |
| 3545 |
#undef B1 |
| 3546 |
#undef B2 |
| 3547 |
#undef B3 |
| 3548 |
|
| 3549 |
#undef X0 |
| 3550 |
#undef X1 |
| 3551 |
#undef X2 |
| 3552 |
#undef X3 |
| 3553 |
|
| 3554 |
#undef Y0 |
| 3555 |
#undef Y1 |
| 3556 |
#undef Y2 |
| 3557 |
#undef Y3 |
| 3558 |
|
| 3559 |
#undef I0 |
| 3560 |
#undef I1 |
| 3561 |
#undef ONE |
| 3562 |
|
| 3563 |
#endif /* __AVR_TINY__ */ |
| 3564 |
#endif /* L_powif */ |
| 3565 |
|
| 3566 |
#include "lib1funcs-fixed.S" |