1 | ; $Id: bignum-amd64-x86.asm 82968 2020-02-04 10:35:17Z vboxsync $
|
---|
2 | ;; @file
|
---|
3 | ; IPRT - Big Integer Numbers, AMD64 and X86 Assembly Workers
|
---|
4 | ;
|
---|
5 |
|
---|
6 | ;
|
---|
7 | ; Copyright (C) 2006-2020 Oracle Corporation
|
---|
8 | ;
|
---|
9 | ; This file is part of VirtualBox Open Source Edition (OSE), as
|
---|
10 | ; available from http://www.virtualbox.org. This file is free software;
|
---|
11 | ; you can redistribute it and/or modify it under the terms of the GNU
|
---|
12 | ; General Public License (GPL) as published by the Free Software
|
---|
13 | ; Foundation, in version 2 as it comes in the "COPYING" file of the
|
---|
14 | ; VirtualBox OSE distribution. VirtualBox OSE is distributed in the
|
---|
15 | ; hope that it will be useful, but WITHOUT ANY WARRANTY of any kind.
|
---|
16 | ;
|
---|
17 | ; The contents of this file may alternatively be used under the terms
|
---|
18 | ; of the Common Development and Distribution License Version 1.0
|
---|
19 | ; (CDDL) only, as it comes in the "COPYING.CDDL" file of the
|
---|
20 | ; VirtualBox OSE distribution, in which case the provisions of the
|
---|
21 | ; CDDL are applicable instead of those of the GPL.
|
---|
22 | ;
|
---|
23 | ; You may elect to license modified versions of this file under the
|
---|
24 | ; terms and conditions of either the GPL or the CDDL or both.
|
---|
25 | ;
|
---|
26 |
|
---|
27 |
|
---|
28 | ;*********************************************************************************************************************************
|
---|
29 | ;* Header Files *
|
---|
30 | ;*********************************************************************************************************************************
|
---|
31 | %define RT_ASM_WITH_SEH64
|
---|
32 | %include "iprt/asmdefs.mac"
|
---|
33 | %include "internal/bignum.mac"
|
---|
34 |
|
---|
35 |
|
---|
36 | ;*********************************************************************************************************************************
|
---|
37 | ;* Defined Constants And Macros *
|
---|
38 | ;*********************************************************************************************************************************
|
---|
39 | %ifdef RT_ARCH_AMD64
|
---|
40 | %macro sahf 0
|
---|
41 | %error "SAHF not supported on ancient AMD64"
|
---|
42 | %endmacro
|
---|
43 | %macro lahf 0
|
---|
44 | %error "LAHF not supported on ancient AMD64"
|
---|
45 | %endmacro
|
---|
46 | %endif
|
---|
47 |
|
---|
48 |
|
---|
49 | BEGINCODE
|
---|
50 |
|
---|
51 | ;;
|
---|
52 | ; Subtracts a number (pauSubtrahend) from a larger number (pauMinuend) and
|
---|
53 | ; stores the result in pauResult.
|
---|
54 | ;
|
---|
55 | ; All three numbers are zero padded such that a borrow can be carried one (or
|
---|
56 | ; two for 64-bit) elements beyond the end of the largest number.
|
---|
57 | ;
|
---|
58 | ; @returns nothing.
|
---|
59 | ; @param pauResult x86:[ebp + 8] gcc:rdi msc:rcx
|
---|
60 | ; @param pauMinuend x86:[ebp + 12] gcc:rsi msc:rdx
|
---|
61 | ; @param pauSubtrahend x86:[ebp + 16] gcc:rdx msc:r8
|
---|
62 | ; @param cUsed x86:[ebp + 20] gcc:rcx msc:r9
|
---|
63 | ;
|
---|
64 | BEGINPROC rtBigNumMagnitudeSubAssemblyWorker
|
---|
65 | push xBP
|
---|
66 | SEH64_PUSH_xBP
|
---|
67 | mov xBP, xSP
|
---|
68 | SEH64_SET_FRAME_xBP 0
|
---|
69 | SEH64_END_PROLOGUE
|
---|
70 |
|
---|
71 | %ifdef RT_ARCH_AMD64
|
---|
72 | %ifdef ASM_CALL64_GCC
|
---|
73 | %define pauResult rdi
|
---|
74 | %define pauMinuend rsi
|
---|
75 | %define pauSubtrahend rdx
|
---|
76 | %define cUsed ecx
|
---|
77 | %else
|
---|
78 | %define pauResult rcx
|
---|
79 | %define pauMinuend rdx
|
---|
80 | %define pauSubtrahend r8
|
---|
81 | %define cUsed r9d
|
---|
82 | %endif
|
---|
83 | xor r11d, r11d ; index register.
|
---|
84 |
|
---|
85 | %if RTBIGNUM_ELEMENT_SIZE == 4
|
---|
86 | add cUsed, 1 ; cUsed = RT_ALIGN(cUsed, 2) / 2
|
---|
87 | shr cUsed, 1
|
---|
88 | %endif
|
---|
89 | cmp cUsed, 8 ; Skip the big loop if small number.
|
---|
90 | jb .small_job
|
---|
91 |
|
---|
92 | mov r10d, cUsed
|
---|
93 | shr r10d, 3
|
---|
94 | clc
|
---|
95 | .big_loop:
|
---|
96 | mov rax, [pauMinuend + r11]
|
---|
97 | sbb rax, [pauSubtrahend + r11]
|
---|
98 | mov [pauResult + r11], rax
|
---|
99 | mov rax, [pauMinuend + r11 + 8]
|
---|
100 | sbb rax, [pauSubtrahend + r11 + 8]
|
---|
101 | mov [pauResult + r11 + 8], rax
|
---|
102 | mov rax, [pauMinuend + r11 + 16]
|
---|
103 | sbb rax, [pauSubtrahend + r11 + 16]
|
---|
104 | mov [pauResult + r11 + 16], rax
|
---|
105 | mov rax, [pauMinuend + r11 + 24]
|
---|
106 | sbb rax, [pauSubtrahend + r11 + 24]
|
---|
107 | mov [pauResult + r11 + 24], rax
|
---|
108 | mov rax, [pauMinuend + r11 + 32]
|
---|
109 | sbb rax, [pauSubtrahend + r11 + 32]
|
---|
110 | mov [pauResult + r11 + 32], rax
|
---|
111 | mov rax, [pauMinuend + r11 + 40]
|
---|
112 | sbb rax, [pauSubtrahend + r11 + 40]
|
---|
113 | mov [pauResult + r11 + 40], rax
|
---|
114 | mov rax, [pauMinuend + r11 + 48]
|
---|
115 | sbb rax, [pauSubtrahend + r11 + 48]
|
---|
116 | mov [pauResult + r11 + 48], rax
|
---|
117 | mov rax, [pauMinuend + r11 + 56]
|
---|
118 | sbb rax, [pauSubtrahend + r11 + 56]
|
---|
119 | mov [pauResult + r11 + 56], rax
|
---|
120 | lea r11, [r11 + 64]
|
---|
121 | dec r10d ; Does not change CF.
|
---|
122 | jnz .big_loop
|
---|
123 |
|
---|
124 | %if 0 ; Ancient AMD CPUs does have lahf/sahf, thus the mess in the %else.
|
---|
125 | lahf ; Save CF
|
---|
126 | and cUsed, 7 ; Up to seven odd rounds.
|
---|
127 | jz .done
|
---|
128 | sahf ; Restore CF.
|
---|
129 | jmp .small_loop ; Skip CF=1 (clc).
|
---|
130 | %else
|
---|
131 | jnc .no_carry
|
---|
132 | and cUsed, 7 ; Up to seven odd rounds.
|
---|
133 | jz .done
|
---|
134 | stc
|
---|
135 | jmp .small_loop ; Skip CF=1 (clc).
|
---|
136 | .no_carry:
|
---|
137 | and cUsed, 7 ; Up to seven odd rounds.
|
---|
138 | jz .done
|
---|
139 | %endif
|
---|
140 | .small_job:
|
---|
141 | clc
|
---|
142 | .small_loop:
|
---|
143 | mov rax, [pauMinuend + r11]
|
---|
144 | sbb rax, [pauSubtrahend + r11]
|
---|
145 | mov [pauResult + r11], rax
|
---|
146 | lea r11, [r11 + 8]
|
---|
147 | dec cUsed ; does not change CF.
|
---|
148 | jnz .small_loop
|
---|
149 | %ifdef RT_STRICT
|
---|
150 | jnc .done
|
---|
151 | int3
|
---|
152 | %endif
|
---|
153 | .done:
|
---|
154 |
|
---|
155 | %elifdef RT_ARCH_X86
|
---|
156 | push edi
|
---|
157 | push esi
|
---|
158 | push ebx
|
---|
159 |
|
---|
160 | mov edi, [ebp + 08h] ; pauResult
|
---|
161 | %define pauResult edi
|
---|
162 | mov ecx, [ebp + 0ch] ; pauMinuend
|
---|
163 | %define pauMinuend ecx
|
---|
164 | mov edx, [ebp + 10h] ; pauSubtrahend
|
---|
165 | %define pauSubtrahend edx
|
---|
166 | mov esi, [ebp + 14h] ; cUsed
|
---|
167 | %define cUsed esi
|
---|
168 |
|
---|
169 | xor ebx, ebx ; index register.
|
---|
170 |
|
---|
171 | cmp cUsed, 8 ; Skip the big loop if small number.
|
---|
172 | jb .small_job
|
---|
173 |
|
---|
174 | shr cUsed, 3
|
---|
175 | clc
|
---|
176 | .big_loop:
|
---|
177 | mov eax, [pauMinuend + ebx]
|
---|
178 | sbb eax, [pauSubtrahend + ebx]
|
---|
179 | mov [pauResult + ebx], eax
|
---|
180 | mov eax, [pauMinuend + ebx + 4]
|
---|
181 | sbb eax, [pauSubtrahend + ebx + 4]
|
---|
182 | mov [pauResult + ebx + 4], eax
|
---|
183 | mov eax, [pauMinuend + ebx + 8]
|
---|
184 | sbb eax, [pauSubtrahend + ebx + 8]
|
---|
185 | mov [pauResult + ebx + 8], eax
|
---|
186 | mov eax, [pauMinuend + ebx + 12]
|
---|
187 | sbb eax, [pauSubtrahend + ebx + 12]
|
---|
188 | mov [pauResult + ebx + 12], eax
|
---|
189 | mov eax, [pauMinuend + ebx + 16]
|
---|
190 | sbb eax, [pauSubtrahend + ebx + 16]
|
---|
191 | mov [pauResult + ebx + 16], eax
|
---|
192 | mov eax, [pauMinuend + ebx + 20]
|
---|
193 | sbb eax, [pauSubtrahend + ebx + 20]
|
---|
194 | mov [pauResult + ebx + 20], eax
|
---|
195 | mov eax, [pauMinuend + ebx + 24]
|
---|
196 | sbb eax, [pauSubtrahend + ebx + 24]
|
---|
197 | mov [pauResult + ebx + 24], eax
|
---|
198 | mov eax, [pauMinuend + ebx + 28]
|
---|
199 | sbb eax, [pauSubtrahend + ebx + 28]
|
---|
200 | mov [pauResult + ebx + 28], eax
|
---|
201 | lea ebx, [ebx + 32]
|
---|
202 | dec cUsed ; Does not change CF.
|
---|
203 | jnz .big_loop
|
---|
204 |
|
---|
205 | lahf ; Save CF
|
---|
206 | mov cUsed, [ebp + 14h] ; Up to three final rounds.
|
---|
207 | and cUsed, 7
|
---|
208 | jz .done
|
---|
209 | sahf ; Restore CF.
|
---|
210 | jmp .small_loop ; Skip CF=1 (clc).
|
---|
211 |
|
---|
212 | .small_job:
|
---|
213 | clc
|
---|
214 | .small_loop:
|
---|
215 | mov eax, [pauMinuend + ebx]
|
---|
216 | sbb eax, [pauSubtrahend + ebx]
|
---|
217 | mov [pauResult + ebx], eax
|
---|
218 | lea ebx, [ebx + 4]
|
---|
219 | dec cUsed ; Does not change CF
|
---|
220 | jnz .small_loop
|
---|
221 | %ifdef RT_STRICT
|
---|
222 | jnc .done
|
---|
223 | int3
|
---|
224 | %endif
|
---|
225 | .done:
|
---|
226 |
|
---|
227 | pop ebx
|
---|
228 | pop esi
|
---|
229 | pop edi
|
---|
230 | %else
|
---|
231 | %error "Unsupported arch"
|
---|
232 | %endif
|
---|
233 |
|
---|
234 | leave
|
---|
235 | ret
|
---|
236 | %undef pauResult
|
---|
237 | %undef pauMinuend
|
---|
238 | %undef pauSubtrahend
|
---|
239 | %undef cUsed
|
---|
240 | ENDPROC rtBigNumMagnitudeSubAssemblyWorker
|
---|
241 |
|
---|
242 |
|
---|
243 |
|
---|
244 | ;;
|
---|
245 | ; Subtracts a number (pauSubtrahend) from a larger number (pauMinuend) and
|
---|
246 | ; stores the result in pauResult.
|
---|
247 | ;
|
---|
248 | ; All three numbers are zero padded such that a borrow can be carried one (or
|
---|
249 | ; two for 64-bit) elements beyond the end of the largest number.
|
---|
250 | ;
|
---|
251 | ; @returns nothing.
|
---|
252 | ; @param pauResultMinuend x86:[ebp + 8] gcc:rdi msc:rcx
|
---|
253 | ; @param pauSubtrahend x86:[ebp + 12] gcc:rsi msc:rdx
|
---|
254 | ; @param cUsed x86:[ebp + 16] gcc:rdx msc:r8
|
---|
255 | ;
|
---|
256 | BEGINPROC rtBigNumMagnitudeSubThisAssemblyWorker
|
---|
257 | push xBP
|
---|
258 | SEH64_PUSH_xBP
|
---|
259 | mov xBP, xSP
|
---|
260 | SEH64_SET_FRAME_xBP 0
|
---|
261 | SEH64_END_PROLOGUE
|
---|
262 |
|
---|
263 | %ifdef RT_ARCH_AMD64
|
---|
264 | %ifdef ASM_CALL64_GCC
|
---|
265 | %define pauResultMinuend rdi
|
---|
266 | %define pauSubtrahend rsi
|
---|
267 | %define cUsed edx
|
---|
268 | %else
|
---|
269 | %define pauResultMinuend rcx
|
---|
270 | %define pauSubtrahend rdx
|
---|
271 | %define cUsed r8d
|
---|
272 | %endif
|
---|
273 | xor r11d, r11d ; index register.
|
---|
274 |
|
---|
275 | %if RTBIGNUM_ELEMENT_SIZE == 4
|
---|
276 | add cUsed, 1 ; cUsed = RT_ALIGN(cUsed, 2) / 2
|
---|
277 | shr cUsed, 1
|
---|
278 | %endif
|
---|
279 | cmp cUsed, 8 ; Skip the big loop if small number.
|
---|
280 | jb .small_job
|
---|
281 |
|
---|
282 | mov r10d, cUsed
|
---|
283 | shr r10d, 3
|
---|
284 | clc
|
---|
285 | .big_loop:
|
---|
286 | mov rax, [pauSubtrahend + r11]
|
---|
287 | sbb [pauResultMinuend + r11], rax
|
---|
288 | mov rax, [pauSubtrahend + r11 + 8]
|
---|
289 | sbb [pauResultMinuend + r11 + 8], rax
|
---|
290 | mov rax, [pauSubtrahend + r11 + 16]
|
---|
291 | sbb [pauResultMinuend + r11 + 16], rax
|
---|
292 | mov rax, [pauSubtrahend + r11 + 24]
|
---|
293 | sbb [pauResultMinuend + r11 + 24], rax
|
---|
294 | mov rax, [pauSubtrahend + r11 + 32]
|
---|
295 | sbb [pauResultMinuend + r11 + 32], rax
|
---|
296 | mov rax, [pauSubtrahend + r11 + 40]
|
---|
297 | sbb [pauResultMinuend + r11 + 40], rax
|
---|
298 | mov rax, [pauSubtrahend + r11 + 48]
|
---|
299 | sbb [pauResultMinuend + r11 + 48], rax
|
---|
300 | mov rax, [pauSubtrahend + r11 + 56]
|
---|
301 | sbb [pauResultMinuend + r11 + 56], rax
|
---|
302 | lea r11, [r11 + 64]
|
---|
303 | dec r10d ; Does not change CF.
|
---|
304 | jnz .big_loop
|
---|
305 |
|
---|
306 | %if 0 ; Ancient AMD CPUs does have lahf/sahf, thus the mess in the %else.
|
---|
307 | lahf ; Save CF
|
---|
308 | and cUsed, 7 ; Up to seven odd rounds.
|
---|
309 | jz .done
|
---|
310 | sahf ; Restore CF.
|
---|
311 | jmp .small_loop ; Skip CF=1 (clc).
|
---|
312 | %else
|
---|
313 | jnc .no_carry
|
---|
314 | and cUsed, 7 ; Up to seven odd rounds.
|
---|
315 | jz .done
|
---|
316 | stc
|
---|
317 | jmp .small_loop ; Skip CF=1 (clc).
|
---|
318 | .no_carry:
|
---|
319 | and cUsed, 7 ; Up to seven odd rounds.
|
---|
320 | jz .done
|
---|
321 | %endif
|
---|
322 | .small_job:
|
---|
323 | clc
|
---|
324 | .small_loop:
|
---|
325 | mov rax, [pauSubtrahend + r11]
|
---|
326 | sbb [pauResultMinuend + r11], rax
|
---|
327 | lea r11, [r11 + 8]
|
---|
328 | dec cUsed ; does not change CF.
|
---|
329 | jnz .small_loop
|
---|
330 | %ifdef RT_STRICT
|
---|
331 | jnc .done
|
---|
332 | int3
|
---|
333 | %endif
|
---|
334 | .done:
|
---|
335 |
|
---|
336 | %elifdef RT_ARCH_X86
|
---|
337 | push edi
|
---|
338 | push ebx
|
---|
339 |
|
---|
340 | mov edi, [ebp + 08h] ; pauResultMinuend
|
---|
341 | %define pauResultMinuend edi
|
---|
342 | mov edx, [ebp + 0ch] ; pauSubtrahend
|
---|
343 | %define pauSubtrahend edx
|
---|
344 | mov ecx, [ebp + 10h] ; cUsed
|
---|
345 | %define cUsed ecx
|
---|
346 |
|
---|
347 | xor ebx, ebx ; index register.
|
---|
348 |
|
---|
349 | cmp cUsed, 8 ; Skip the big loop if small number.
|
---|
350 | jb .small_job
|
---|
351 |
|
---|
352 | shr cUsed, 3
|
---|
353 | clc
|
---|
354 | .big_loop:
|
---|
355 | mov eax, [pauSubtrahend + ebx]
|
---|
356 | sbb [pauResultMinuend + ebx], eax
|
---|
357 | mov eax, [pauSubtrahend + ebx + 4]
|
---|
358 | sbb [pauResultMinuend + ebx + 4], eax
|
---|
359 | mov eax, [pauSubtrahend + ebx + 8]
|
---|
360 | sbb [pauResultMinuend + ebx + 8], eax
|
---|
361 | mov eax, [pauSubtrahend + ebx + 12]
|
---|
362 | sbb [pauResultMinuend + ebx + 12], eax
|
---|
363 | mov eax, [pauSubtrahend + ebx + 16]
|
---|
364 | sbb [pauResultMinuend + ebx + 16], eax
|
---|
365 | mov eax, [pauSubtrahend + ebx + 20]
|
---|
366 | sbb [pauResultMinuend + ebx + 20], eax
|
---|
367 | mov eax, [pauSubtrahend + ebx + 24]
|
---|
368 | sbb [pauResultMinuend + ebx + 24], eax
|
---|
369 | mov eax, [pauSubtrahend + ebx + 28]
|
---|
370 | sbb [pauResultMinuend + ebx + 28], eax
|
---|
371 | lea ebx, [ebx + 32]
|
---|
372 | dec cUsed ; Does not change CF.
|
---|
373 | jnz .big_loop
|
---|
374 |
|
---|
375 | lahf ; Save CF
|
---|
376 | mov cUsed, [ebp + 10h] ; Up to seven odd rounds.
|
---|
377 | and cUsed, 7
|
---|
378 | jz .done
|
---|
379 | sahf ; Restore CF.
|
---|
380 | jmp .small_loop ; Skip CF=1 (clc).
|
---|
381 |
|
---|
382 | .small_job:
|
---|
383 | clc
|
---|
384 | .small_loop:
|
---|
385 | mov eax, [pauSubtrahend + ebx]
|
---|
386 | sbb [pauResultMinuend + ebx], eax
|
---|
387 | lea ebx, [ebx + 4]
|
---|
388 | dec cUsed ; Does not change CF
|
---|
389 | jnz .small_loop
|
---|
390 | %ifdef RT_STRICT
|
---|
391 | jnc .done
|
---|
392 | int3
|
---|
393 | %endif
|
---|
394 | .done:
|
---|
395 |
|
---|
396 | pop ebx
|
---|
397 | pop edi
|
---|
398 | %else
|
---|
399 | %error "Unsupported arch"
|
---|
400 | %endif
|
---|
401 |
|
---|
402 | leave
|
---|
403 | ret
|
---|
404 | ENDPROC rtBigNumMagnitudeSubThisAssemblyWorker
|
---|
405 |
|
---|
406 |
|
---|
407 | ;;
|
---|
408 | ; Shifts an element array one bit to the left, returning the final carry value.
|
---|
409 | ;
|
---|
410 | ; On 64-bit hosts the array is always zero padded to a multiple of 8 bytes, so
|
---|
411 | ; we can use 64-bit operand sizes even if the element type is 32-bit.
|
---|
412 | ;
|
---|
413 | ; @returns The final carry value.
|
---|
414 | ; @param pauElements x86:[ebp + 8] gcc:rdi msc:rcx
|
---|
415 | ; @param cUsed x86:[ebp + 12] gcc:rsi msc:rdx
|
---|
416 | ; @param uCarry x86:[ebp + 16] gcc:rdx msc:r8
|
---|
417 | ;
|
---|
418 | BEGINPROC rtBigNumMagnitudeShiftLeftOneAssemblyWorker
|
---|
419 | push xBP
|
---|
420 | SEH64_PUSH_xBP
|
---|
421 | mov xBP, xSP
|
---|
422 | SEH64_SET_FRAME_xBP 0
|
---|
423 | SEH64_END_PROLOGUE
|
---|
424 |
|
---|
425 | %ifdef RT_ARCH_AMD64
|
---|
426 | %ifdef ASM_CALL64_GCC
|
---|
427 | %define pauElements rdi
|
---|
428 | %define cUsed esi
|
---|
429 | %define uCarry edx
|
---|
430 | %else
|
---|
431 | %define pauElements rcx
|
---|
432 | %define cUsed edx
|
---|
433 | %define uCarry r8d
|
---|
434 | %endif
|
---|
435 | %elifdef RT_ARCH_X86
|
---|
436 | %define pauElements ecx
|
---|
437 | mov pauElements, [ebp + 08h]
|
---|
438 | %define cUsed edx
|
---|
439 | mov cUsed, [ebp + 0ch]
|
---|
440 | %define uCarry eax
|
---|
441 | mov uCarry, [ebp + 10h]
|
---|
442 | %else
|
---|
443 | %error "Unsupported arch."
|
---|
444 | %endif
|
---|
445 | ; Lots to do?
|
---|
446 | cmp cUsed, 8
|
---|
447 | jae .big_loop_init
|
---|
448 |
|
---|
449 | ; Check for empty array.
|
---|
450 | test cUsed, cUsed
|
---|
451 | jz .no_elements
|
---|
452 | jmp .small_loop_init
|
---|
453 |
|
---|
454 | ; Big loop - 8 unrolled loop iterations.
|
---|
455 | .big_loop_init:
|
---|
456 | %ifdef RT_ARCH_AMD64
|
---|
457 | mov r11d, cUsed
|
---|
458 | %endif
|
---|
459 | shr cUsed, 3
|
---|
460 | test uCarry, uCarry ; clear the carry flag
|
---|
461 | jz .big_loop
|
---|
462 | stc
|
---|
463 | .big_loop:
|
---|
464 | %if RTBIGNUM_ELEMENT_SIZE == 8
|
---|
465 | rcl qword [pauElements], 1
|
---|
466 | rcl qword [pauElements + 8], 1
|
---|
467 | rcl qword [pauElements + 16], 1
|
---|
468 | rcl qword [pauElements + 24], 1
|
---|
469 | rcl qword [pauElements + 32], 1
|
---|
470 | rcl qword [pauElements + 40], 1
|
---|
471 | rcl qword [pauElements + 48], 1
|
---|
472 | rcl qword [pauElements + 56], 1
|
---|
473 | lea pauElements, [pauElements + 64]
|
---|
474 | %else
|
---|
475 | rcl dword [pauElements], 1
|
---|
476 | rcl dword [pauElements + 4], 1
|
---|
477 | rcl dword [pauElements + 8], 1
|
---|
478 | rcl dword [pauElements + 12], 1
|
---|
479 | rcl dword [pauElements + 16], 1
|
---|
480 | rcl dword [pauElements + 20], 1
|
---|
481 | rcl dword [pauElements + 24], 1
|
---|
482 | rcl dword [pauElements + 28], 1
|
---|
483 | lea pauElements, [pauElements + 32]
|
---|
484 | %endif
|
---|
485 | dec cUsed
|
---|
486 | jnz .big_loop
|
---|
487 |
|
---|
488 | ; More to do?
|
---|
489 | pushf ; save carry flag (uCarry no longer used on x86).
|
---|
490 | %ifdef RT_ARCH_AMD64
|
---|
491 | mov cUsed, r11d
|
---|
492 | %else
|
---|
493 | mov cUsed, [ebp + 0ch]
|
---|
494 | %endif
|
---|
495 | and cUsed, 7
|
---|
496 | jz .restore_cf_and_return ; Jump if we're good and done.
|
---|
497 | popf ; Restore CF.
|
---|
498 | jmp .small_loop ; Deal with the odd rounds.
|
---|
499 | .restore_cf_and_return:
|
---|
500 | popf
|
---|
501 | jmp .carry_to_eax
|
---|
502 |
|
---|
503 | ; Small loop - One round at the time.
|
---|
504 | .small_loop_init:
|
---|
505 | test uCarry, uCarry ; clear the carry flag
|
---|
506 | jz .small_loop
|
---|
507 | stc
|
---|
508 | .small_loop:
|
---|
509 | %if RTBIGNUM_ELEMENT_SIZE == 8
|
---|
510 | rcl qword [pauElements], 1
|
---|
511 | lea pauElements, [pauElements + 8]
|
---|
512 | %else
|
---|
513 | rcl dword [pauElements], 1
|
---|
514 | lea pauElements, [pauElements + 4]
|
---|
515 | %endif
|
---|
516 | dec cUsed
|
---|
517 | jnz .small_loop
|
---|
518 |
|
---|
519 | ; Calculate return value.
|
---|
520 | .carry_to_eax:
|
---|
521 | mov eax, 0
|
---|
522 | jnc .return
|
---|
523 | inc eax
|
---|
524 | .return:
|
---|
525 | leave
|
---|
526 | ret
|
---|
527 |
|
---|
528 | .no_elements:
|
---|
529 | mov eax, uCarry
|
---|
530 | jmp .return
|
---|
531 | ENDPROC rtBigNumMagnitudeShiftLeftOneAssemblyWorker
|
---|
532 |
|
---|
533 |
|
---|
534 | ;;
|
---|
535 | ; Performs a 128-bit by 64-bit division on 64-bit and
|
---|
536 | ; a 64-bit by 32-bit divison on 32-bit.
|
---|
537 | ;
|
---|
538 | ; @returns nothing.
|
---|
539 | ; @param puQuotient x86:[ebp + 8] gcc:rdi msc:rcx Double element.
|
---|
540 | ; @param puRemainder x86:[ebp + 12] gcc:rsi msc:rdx Normal element.
|
---|
541 | ; @param uDividendHi x86:[ebp + 16] gcc:rdx msc:r8
|
---|
542 | ; @param uDividendLo x86:[ebp + 20] gcc:rcx msc:r9
|
---|
543 | ; @param uDivisior x86:[ebp + 24] gcc:r8 msc:[rbp + 30h]
|
---|
544 | ;
|
---|
545 | BEGINPROC rtBigNumElement2xDiv2xBy1x
|
---|
546 | push xBP
|
---|
547 | SEH64_PUSH_xBP
|
---|
548 | mov xBP, xSP
|
---|
549 | SEH64_SET_FRAME_xBP 0
|
---|
550 | SEH64_END_PROLOGUE
|
---|
551 |
|
---|
552 | %ifdef RT_ARCH_AMD64
|
---|
553 | %if RTBIGNUM_ELEMENT_SIZE == 4
|
---|
554 | %error "sorry not implemented yet."
|
---|
555 | sorry not implemented yet.
|
---|
556 | %endif
|
---|
557 |
|
---|
558 | %define uDividendHi rdx
|
---|
559 | %define uDividendLo rax
|
---|
560 | %ifdef ASM_CALL64_GCC
|
---|
561 | %define uDivisor r8
|
---|
562 | %define puQuotient rdi
|
---|
563 | %define puRemainder rsi
|
---|
564 | mov rax, rcx
|
---|
565 | %else
|
---|
566 | %define puQuotient rcx
|
---|
567 | %define puRemainder r11
|
---|
568 | %define uDivisor r10
|
---|
569 | mov r11, rdx
|
---|
570 | mov r10, [rbp + 30h]
|
---|
571 | mov rdx, r8
|
---|
572 | mov rax, r9
|
---|
573 | %endif
|
---|
574 |
|
---|
575 | %elifdef RT_ARCH_X86
|
---|
576 | push edi
|
---|
577 | push ebx
|
---|
578 |
|
---|
579 | %define uDividendHi edx
|
---|
580 | mov uDividendHi, [ebp + 10h]
|
---|
581 | %define uDividendLo eax
|
---|
582 | mov uDividendLo, [ebp + 14h]
|
---|
583 | %define uDivisor ecx
|
---|
584 | mov uDivisor, [ebp + 18h]
|
---|
585 | %define puQuotient edi
|
---|
586 | mov puQuotient, [ebp + 08h]
|
---|
587 | %define puRemainder ebx
|
---|
588 | mov puRemainder, [ebp + 0ch]
|
---|
589 | %else
|
---|
590 | %error "Unsupported arch."
|
---|
591 | %endif
|
---|
592 |
|
---|
593 | %ifdef RT_STRICT
|
---|
594 | ;
|
---|
595 | ; The dividend shall not be zero.
|
---|
596 | ;
|
---|
597 | test uDivisor, uDivisor
|
---|
598 | jnz .divisor_not_zero
|
---|
599 | int3
|
---|
600 | .divisor_not_zero:
|
---|
601 | %endif
|
---|
602 |
|
---|
603 | ;
|
---|
604 | ; Avoid division overflow. This will calculate the high part of the quotient.
|
---|
605 | ;
|
---|
606 | mov RTBIGNUM_ELEMENT_PRE [puQuotient + RTBIGNUM_ELEMENT_SIZE], 0
|
---|
607 | cmp uDividendHi, uDivisor
|
---|
608 | jb .do_divide
|
---|
609 | push xAX
|
---|
610 | mov xAX, xDX
|
---|
611 | xor edx, edx
|
---|
612 | div uDivisor
|
---|
613 | mov RTBIGNUM_ELEMENT_PRE [puQuotient + RTBIGNUM_ELEMENT_SIZE], xAX
|
---|
614 | pop xAX
|
---|
615 |
|
---|
616 | ;
|
---|
617 | ; Perform the division and store the result.
|
---|
618 | ;
|
---|
619 | .do_divide:
|
---|
620 | div uDivisor
|
---|
621 | mov RTBIGNUM_ELEMENT_PRE [puQuotient], xAX
|
---|
622 | mov RTBIGNUM_ELEMENT_PRE [puRemainder], xDX
|
---|
623 |
|
---|
624 |
|
---|
625 | %ifdef RT_ARCH_X86
|
---|
626 | pop ebx
|
---|
627 | pop edi
|
---|
628 | %endif
|
---|
629 | leave
|
---|
630 | ret
|
---|
631 | ENDPROC rtBigNumElement2xDiv2xBy1x
|
---|
632 |
|
---|
633 |
|
---|
634 | ;;
|
---|
635 | ; Performs the core of long multiplication.
|
---|
636 | ;
|
---|
637 | ; @returns nothing.
|
---|
638 | ; @param pauResult x86:[ebp + 8] gcc:rdi msc:rcx Initialized to zero.
|
---|
639 | ; @param pauMultiplier x86:[ebp + 12] gcc:rsi msc:rdx
|
---|
640 | ; @param cMultiplier x86:[ebp + 16] gcc:rdx msc:r8
|
---|
641 | ; @param pauMultiplicand x86:[ebp + 20] gcc:rcx msc:r9
|
---|
642 | ; @param cMultiplicand x86:[ebp + 24] gcc:r8 msc:[rbp + 30h]
|
---|
643 | ;
|
---|
644 | BEGINPROC rtBigNumMagnitudeMultiplyAssemblyWorker
|
---|
645 | push xBP
|
---|
646 | SEH64_PUSH_xBP
|
---|
647 | mov xBP, xSP
|
---|
648 | SEH64_SET_FRAME_xBP 0
|
---|
649 | SEH64_END_PROLOGUE
|
---|
650 |
|
---|
651 | %ifdef RT_ARCH_AMD64
|
---|
652 | %if RTBIGNUM_ELEMENT_SIZE == 4
|
---|
653 | %error "sorry not implemented yet."
|
---|
654 | sorry not implemented yet.
|
---|
655 | %endif
|
---|
656 |
|
---|
657 | %ifdef ASM_CALL64_GCC
|
---|
658 | %define pauResult rdi
|
---|
659 | %define pauMultiplier rsi
|
---|
660 | %define cMultiplier r9
|
---|
661 | %define pauMultiplicand rcx
|
---|
662 | %define cMultiplicand r8
|
---|
663 | mov r9d, edx ; cMultiplier
|
---|
664 | mov r8d, r8d ; cMultiplicand - paranoia
|
---|
665 | %define uMultiplier r10
|
---|
666 | %define iMultiplicand r11
|
---|
667 | %else
|
---|
668 | %define pauResult rcx
|
---|
669 | %define pauMultiplier r11
|
---|
670 | %define cMultiplier r8
|
---|
671 | %define pauMultiplicand r9
|
---|
672 | %define cMultiplicand r10
|
---|
673 | mov pauMultiplier, rdx
|
---|
674 | mov r10d, dword [rbp + 30h] ; cMultiplicand
|
---|
675 | mov r8d, r8d ; cMultiplier - paranoia
|
---|
676 | %define uMultiplier r12
|
---|
677 | push r12
|
---|
678 | %define iMultiplicand r13
|
---|
679 | push r13
|
---|
680 | %endif
|
---|
681 |
|
---|
682 | %elifdef RT_ARCH_X86
|
---|
683 | push edi
|
---|
684 | push esi
|
---|
685 | push ebx
|
---|
686 | sub esp, 10h
|
---|
687 | %define pauResult edi
|
---|
688 | mov pauResult, [ebp + 08h]
|
---|
689 | %define pauMultiplier dword [ebp + 0ch]
|
---|
690 | %define cMultiplier dword [ebp + 10h]
|
---|
691 | %define pauMultiplicand ecx
|
---|
692 | mov pauMultiplicand, [ebp + 14h]
|
---|
693 | %define cMultiplicand dword [ebp + 18h]
|
---|
694 | %define uMultiplier dword [ebp - 10h]
|
---|
695 | %define iMultiplicand ebx
|
---|
696 |
|
---|
697 | %else
|
---|
698 | %error "Unsupported arch."
|
---|
699 | %endif
|
---|
700 |
|
---|
701 | ;
|
---|
702 | ; Check that the multiplicand isn't empty (avoids an extra jump in the inner loop).
|
---|
703 | ;
|
---|
704 | cmp cMultiplicand, 0
|
---|
705 | je .done
|
---|
706 |
|
---|
707 | ;
|
---|
708 | ; Loop thru each element in the multiplier.
|
---|
709 | ;
|
---|
710 | ; while (cMultiplier-- > 0)
|
---|
711 | .multiplier_loop:
|
---|
712 | cmp cMultiplier, 0
|
---|
713 | jz .done
|
---|
714 | dec cMultiplier
|
---|
715 |
|
---|
716 | ; uMultiplier = *pauMultiplier
|
---|
717 | %ifdef RT_ARCH_X86
|
---|
718 | mov edx, pauMultiplier
|
---|
719 | mov eax, [edx]
|
---|
720 | mov uMultiplier, eax
|
---|
721 | %else
|
---|
722 | mov uMultiplier, [pauMultiplier]
|
---|
723 | %endif
|
---|
724 | ; for (iMultiplicand = 0; iMultiplicand < cMultiplicand; iMultiplicand++)
|
---|
725 | xor iMultiplicand, iMultiplicand
|
---|
726 | .multiplicand_loop:
|
---|
727 | mov xAX, [pauMultiplicand + iMultiplicand * RTBIGNUM_ELEMENT_SIZE]
|
---|
728 | mul uMultiplier
|
---|
729 | add [pauResult + iMultiplicand * RTBIGNUM_ELEMENT_SIZE], xAX
|
---|
730 | adc [pauResult + iMultiplicand * RTBIGNUM_ELEMENT_SIZE + RTBIGNUM_ELEMENT_SIZE], xDX
|
---|
731 | jnc .next_multiplicand
|
---|
732 | lea xDX, [iMultiplicand + 2]
|
---|
733 | .next_adc:
|
---|
734 | adc RTBIGNUM_ELEMENT_PRE [pauResult + xDX * RTBIGNUM_ELEMENT_SIZE], 0
|
---|
735 | inc xDX
|
---|
736 | jc .next_adc
|
---|
737 |
|
---|
738 | .next_multiplicand:
|
---|
739 | inc iMultiplicand ; iMultiplicand++
|
---|
740 | cmp iMultiplicand, cMultiplicand ; iMultiplicand < cMultiplicand
|
---|
741 | jb .multiplicand_loop
|
---|
742 |
|
---|
743 | ; Advance and loop on multiplier.
|
---|
744 | add pauMultiplier, RTBIGNUM_ELEMENT_SIZE
|
---|
745 | add pauResult, RTBIGNUM_ELEMENT_SIZE
|
---|
746 | jmp .multiplier_loop
|
---|
747 |
|
---|
748 | .done:
|
---|
749 |
|
---|
750 | %ifdef RT_ARCH_AMD64
|
---|
751 | %ifdef ASM_CALL64_GCC
|
---|
752 | %else
|
---|
753 | pop r13
|
---|
754 | pop r12
|
---|
755 | %endif
|
---|
756 | %elifdef RT_ARCH_X86
|
---|
757 | add esp, 10h
|
---|
758 | pop ebx
|
---|
759 | pop esi
|
---|
760 | pop edi
|
---|
761 | %endif
|
---|
762 | leave
|
---|
763 | ret
|
---|
764 | ENDPROC rtBigNumMagnitudeMultiplyAssemblyWorker
|
---|
765 |
|
---|
766 | ;;
|
---|
767 | ; Assembly implementation of the D4 step of Knuth's division algorithm.
|
---|
768 | ;
|
---|
769 | ; This subtracts Divisor * Qhat from the dividend at the current J index.
|
---|
770 | ;
|
---|
771 | ; @returns true if negative result (unlikely), false if positive.
|
---|
772 | ; @param pauDividendJ x86:[ebp + 8] gcc:rdi msc:rcx Initialized to zero.
|
---|
773 | ; @param pauDivisor x86:[ebp + 12] gcc:rsi msc:rdx
|
---|
774 | ; @param cDivisor x86:[ebp + 16] gcc:edx msc:r8d
|
---|
775 | ; @param uQhat x86:[ebp + 16] gcc:rcx msc:r9
|
---|
776 | ;
|
---|
777 | BEGINPROC rtBigNumKnuthD4_MulSub
|
---|
778 | push xBP
|
---|
779 | SEH64_PUSH_xBP
|
---|
780 | mov xBP, xSP
|
---|
781 | SEH64_SET_FRAME_xBP 0
|
---|
782 | SEH64_END_PROLOGUE
|
---|
783 |
|
---|
784 | %ifdef RT_ARCH_AMD64
|
---|
785 | %if RTBIGNUM_ELEMENT_SIZE == 4
|
---|
786 | %error "sorry not implemented yet."
|
---|
787 | sorry not implemented yet.
|
---|
788 | %endif
|
---|
789 |
|
---|
790 | %ifdef ASM_CALL64_GCC
|
---|
791 | %define pauDividendJ rdi
|
---|
792 | %define pauDivisor rsi
|
---|
793 | %define cDivisor r8
|
---|
794 | %define uQhat rcx
|
---|
795 | mov r8d, edx ; cDivisor
|
---|
796 | %define uMulCarry r11
|
---|
797 | %else
|
---|
798 | %define pauDividendJ rcx
|
---|
799 | %define pauDivisor r10
|
---|
800 | %define cDivisor r8
|
---|
801 | %define uQhat r9
|
---|
802 | mov r10, rdx ; pauDivisor
|
---|
803 | mov r8d, r8d ; cDivisor - paranoia
|
---|
804 | %define uMulCarry r11
|
---|
805 | %endif
|
---|
806 |
|
---|
807 | %elifdef RT_ARCH_X86
|
---|
808 | push edi
|
---|
809 | push esi
|
---|
810 | push ebx
|
---|
811 | %define pauDividendJ edi
|
---|
812 | mov pauDividendJ, [ebp + 08h]
|
---|
813 | %define pauDivisor esi
|
---|
814 | mov pauDivisor, [ebp + 0ch]
|
---|
815 | %define cDivisor ecx
|
---|
816 | mov cDivisor, [ebp + 10h]
|
---|
817 | %define uQhat dword [ebp + 14h]
|
---|
818 | %define uMulCarry ebx
|
---|
819 | %else
|
---|
820 | %error "Unsupported arch."
|
---|
821 | %endif
|
---|
822 |
|
---|
823 | %ifdef RT_STRICT
|
---|
824 | ;
|
---|
825 | ; Some sanity checks.
|
---|
826 | ;
|
---|
827 | cmp cDivisor, 0
|
---|
828 | jne .cDivisor_not_zero
|
---|
829 | int3
|
---|
830 | .cDivisor_not_zero:
|
---|
831 | %endif
|
---|
832 |
|
---|
833 | ;
|
---|
834 | ; Initialize the loop.
|
---|
835 | ;
|
---|
836 | xor uMulCarry, uMulCarry
|
---|
837 |
|
---|
838 | ;
|
---|
839 | ; do ... while (cDivisor-- > 0);
|
---|
840 | ;
|
---|
841 | .the_loop:
|
---|
842 | ; RTUInt128MulU64ByU64(&uSub, uQhat, pauDivisor[i]);
|
---|
843 | mov xAX, uQhat
|
---|
844 | mul RTBIGNUM_ELEMENT_PRE [pauDivisor]
|
---|
845 | ; RTUInt128AssignAddU64(&uSub, uMulCarry);
|
---|
846 | add xAX, uMulCarry
|
---|
847 | adc xDX, 0
|
---|
848 | mov uMulCarry, xDX
|
---|
849 | ; Subtract uSub.s.Lo+fCarry from pauDividendJ[i]
|
---|
850 | sub [pauDividendJ], xAX
|
---|
851 | adc uMulCarry, 0
|
---|
852 | %ifdef RT_STRICT
|
---|
853 | jnc .uMulCarry_did_not_overflow
|
---|
854 | int3
|
---|
855 | .uMulCarry_did_not_overflow
|
---|
856 | %endif
|
---|
857 |
|
---|
858 | ; Advance.
|
---|
859 | add pauDividendJ, RTBIGNUM_ELEMENT_SIZE
|
---|
860 | add pauDivisor, RTBIGNUM_ELEMENT_SIZE
|
---|
861 | dec cDivisor
|
---|
862 | jnz .the_loop
|
---|
863 |
|
---|
864 | ;
|
---|
865 | ; Final dividend element (no corresponding divisor element).
|
---|
866 | ;
|
---|
867 | sub [pauDividendJ], uMulCarry
|
---|
868 | sbb eax, eax
|
---|
869 | and eax, 1
|
---|
870 |
|
---|
871 | .done:
|
---|
872 | %ifdef RT_ARCH_AMD64
|
---|
873 | %elifdef RT_ARCH_X86
|
---|
874 | pop ebx
|
---|
875 | pop esi
|
---|
876 | pop edi
|
---|
877 | %endif
|
---|
878 | leave
|
---|
879 | ret
|
---|
880 | ENDPROC rtBigNumKnuthD4_MulSub
|
---|
881 |
|
---|