1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
352
353
354
355
356
357
358
359
360
361
362
363
364
365
366
367
368
369
370
371
372
373
374
375
376
377
378
379
380
381
382
383
384
385
386
387
388
389
390
391
392
393
394
395
396
397
398
399
400
401
402
403
404
405
406
407
408
409
410
411
412
413
414
415
416
417
418
419
420
421
422
423
424
425
426
427
428
429
430
431
432
433
434
435
436
437
438
439
440
441
442
443
444
445
446
447
448
449
450
451
452
453
454
455
456
457
458
459
460
461
462
463
464
465
466
467
468
469
470
471
472
473
474
475
476
477
478
479
480
481
482
483
484
485
486
487
488
489
490
491
492
493
494
495
496
497
498
499
500
501
502
503
504
505
506
507
508
509
510
511
512
513
514
515
516
517
518
519
520
521
522
523
524
525
526
527
528
529
530
531
532
533
534
535
536
537
538
539
540
541
542
543
544
545
546
547
548
549
550
551
552
553
554
555
556
557
558
559
560
561
562
563
564
565
566
567
568
569
570
571
572
573
574
575
576
577
578
579
580
581
582
583
584
585
586
587
588
589
590
591
592
593
594
595
596
597
598
599
600
601
602
603
604
605
606
607
608
609
610
611
612
613
614
615
616
617
618
619
620
621
622
623
624
625
626
627
628
629
630
631
632
633
634
635
636
637
638
639
640
641
|
/* ----> DO NOT REMOVE THE FOLLOWING NOTICE <----
Copyright (c) 2014-2015 Datalight, Inc.
All Rights Reserved Worldwide.
This program is free software; you can redistribute it and/or modify
it under the terms of the GNU General Public License as published by
the Free Software Foundation; use version 2 of the License.
This program is distributed in the hope that it will be useful,
but "AS-IS," WITHOUT ANY WARRANTY; without even the implied warranty
of MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
GNU General Public License for more details.
You should have received a copy of the GNU General Public License along
with this program; if not, write to the Free Software Foundation, Inc.,
51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA.
*/
/* Businesses and individuals that for commercial or other reasons cannot
comply with the terms of the GPLv2 license may obtain a commercial license
before incorporating Reliance Edge into proprietary software for
distribution in any form. Visit http://www.datalight.com/reliance-edge for
more information.
*/
/** @file
@brief Implements routines for certain 64-bit math operations and simulated
floating point.
RedUint64DivMod32() and RedUint64DivMod64() are derived from code at
http://www.hackersdelight.org. This web site states explicitly that "You
are free to use, copy, and distribute any of the code on this web site,
whether modified by you or not. You need not give attribution."
*/
#include <redfs.h>
#include <redtestutils.h>
static uint32_t nlz64(uint64_t ullValue);
/** @brief Return a ratio value formatted as a floating point string accurate to
the specified number of decimal places.
The function exists to provide floating point style output without using
any actual floating point types.
This function may scale the numbers down to avoid overflow at the high end.
Likewise, potential divide-by-zero errors are internally avoided. Here are
some examples:
Dividend | Divisor | DecPlaces | Result
-------- | ------- | --------- | ------
12133 | 28545 | 2 | "0.42"
1539 | 506 | 2 | "3.04"
To get a number formatted as a percentage, take the take the portion of the
total (normally the smaller part), multiply it by 100, and pass it to this
function as the Dividend, pass the "total" value to this function as the
Divisor, and specify the desired number of decimal places.
For example, if you have a disk format overhead value of N blocks out of a
total of Y blocks on the disk, and you want to display the format overhead
as a percentage, you would use a function call
similar to:
~~~{.c}
RedRatio(szBuffer, sizeof(szBuffer), N*100U, Y, 2U);
~~~
If N=145, Y=4096, and decimal places is 2, the resulting output would be
"3.54".
The string returned will always be null-terminated, even if it means
stomping on the least significant decimal digit.
If either the dividend or divisor values are zero, the string "0.0" will be
returned, with the prescribed number of decimal places.
@note This function has "reasonable" limits which meet the needs of the
various supplemental utilities which use this function. Extremely
large ratios, or using many decimal places may not function as
desired.
Parameters:
@param pBuffer A pointer to the buffer in which to store the null
terminated results.
@param ulBufferLen The length of the output buffer.
@param ullDividend The "total" value to divide.
@param ullDivisor The portion of ullDividend for which to calculate the
ratio (may be greater than ulDividend).
@param ulDecPlaces The number of decimal places to use, from 0 to 9.
@return @p pBuffer.
*/
char *RedRatio(
char *pBuffer,
uint32_t ulBufferLen,
uint64_t ullDividend,
uint64_t ullDivisor,
uint32_t ulDecPlaces)
{
REDASSERT(pBuffer != NULL);
REDASSERT(ulBufferLen > 0U);
REDASSERT(ulDecPlaces <= 9U); /* arbitrary */
if((ullDivisor > 0U) && (ullDividend > 0U))
{
uint32_t ii;
uint32_t ulFactor = 1U;
uint64_t ullDecimal;
uint64_t ullTemp;
for(ii = 1U; ii <= ulDecPlaces; ii++)
{
ulFactor *= 10U;
}
ullDecimal = RedMulDiv64(ullDividend, ulFactor, ullDivisor);
/* Shouldn't really be calling this function in a situation where we
can overflow at this point...
*/
REDASSERT(ullDecimal != UINT64_MAX);
if(ullDivisor <= ullDividend)
{
uint32_t ulDecimal;
(void)RedUint64DivMod32(ullDecimal, ulFactor, &ulDecimal);
ullDecimal = ulDecimal;
}
ullTemp = RedUint64DivMod64(ullDividend, ullDivisor, NULL);
if(ulDecPlaces > 0U)
{
RedSNPrintf(pBuffer, ulBufferLen, "%llu.%0*llu", (unsigned long long)ullTemp,
(unsigned)ulDecPlaces, (unsigned long long)ullDecimal);
}
else
{
RedSNPrintf(pBuffer, ulBufferLen, "%llu", (unsigned long long)ullTemp);
}
}
else
{
/* If either the dividend or divisor is zero, then just output a "0.0"
string with the prescribed number of decimal places.
*/
if(ulDecPlaces > 0U)
{
RedSNPrintf(pBuffer, ulBufferLen, "0.%0*u", (unsigned)ulDecPlaces, 0U);
}
else
{
RedStrNCpy(pBuffer, "0", ulBufferLen);
}
}
/* Ensure the returned buffer is always null-terminated
*/
pBuffer[ulBufferLen - 1U] = '\0';
return pBuffer;
}
/** @brief Multiply 64-bit and 32-bit numbers, and divide by a 64-bit number,
returning a 64-bit result.
@note This function may return an approximate value if multiplying
@p ullBase and @p ulMultplier results in a number larger than 64-bits
_and_ this cannot be avoided by scaling.
@param ullBase The base 64-bit number number.
@param ulMultiplier The 32-bit number by which to multiply.
@param ullDivisor The 64-bit number by which to divide.
@return The 64-bit unsigned integer result. Always returns zero if either
@p ullBase or @p ulMultiplier are zero (regardless what
@p ullDivisor is). Returns UINT64_MAX if an overflow condition
occurred, or if @p ullDivisor is zero.
*/
uint64_t RedMulDiv64(
uint64_t ullBase,
uint32_t ulMultiplier,
uint64_t ullDivisor)
{
uint64_t ullTemp;
/* Result would always be zero if either of these are zero. Specifically
test this case before looking for a zero divisor.
*/
if((ullBase == 0U) || (ulMultiplier == 0U))
{
return 0U;
}
if(ullDivisor == 0U)
{
return UINT64_MAX;
}
/* Since we don't have the ability (yet) to use 128-bit numbers, we jump
through the following hoops (in order) to try to determine the proper
results without losing precision:
1) Shift the divisor and one of the multiplicands as many times as is
necessary to reduce the scale -- only if it can be done without
losing precision.
2) Divide one of the multiplicands by the divisor first, but only if it
divides evenly, preserving precision.
3) Same as #2, but try it for the other multiplicand.
4) Last ditch, divide the larger multiplicand by the divisor first, then
do the multiply. This <WILL> lose precision.
These solutions are identified as CODE-PATHs #1-4 which are used to
identify the matching tests in dltmain.c.
Note that execution might partially include CODE-PATH #1 up until
shifting can no longer be done without losing precision. In that case,
one of the three remaining options will be used.
*/
ullTemp = RedUint64DivMod32(UINT64_MAX, ulMultiplier, NULL);
while(ullBase > ullTemp)
{
uint64_t ullMod;
uint64_t ullBaseTemp;
uint64_t ullWideMultiplier;
/* CODE-PATH #1
*/
/* So long as ulDivisor, and at least one of the other numbers, are
evenly divisible by 2, we can scale the numbers so the result does
not overflow the intermediate 64-bit value.
*/
if((ullDivisor & 1U) == 0U)
{
if((ullBase & 1U) == 0U)
{
/* CODE-PATH #1a
*/
ullDivisor >>= 1U;
ullBase >>= 1U;
continue;
}
if(((ulMultiplier & 1U) == 0U) && ((ullTemp & UINT64_SUFFIX(0x8000000000000000)) == 0U))
{
/* CODE-PATH #1b
*/
ullDivisor >>= 1U;
ulMultiplier >>= 1U;
ullTemp <<= 1U;
continue;
}
}
/* If we get to this point, the above method (#1) cannot be used
because not enough of the numbers are even long enough to scale the
operands down. We'll see if either multiplicand is evenly divisble
by ulDivisor, and if so, do the divide first, then the multiply.
(Note that once we get to this point, we will never exercise the
while{} loop anymore.)
*/
/* CODE-PATH #2
*/
ullBaseTemp = RedUint64DivMod64(ullBase, ullDivisor, &ullMod);
if(ullMod == 0U)
{
/* Evenly divides, so check that we won't overflow, and finish up.
*/
ullBase = ullBaseTemp;
if(ullBase > ullTemp)
{
return UINT64_MAX;
}
else
{
/* We've validated that this will not overflow.
*/
ullBase *= ulMultiplier;
return ullBase;
}
}
/* CODE-PATH #3
*/
ullWideMultiplier = RedUint64DivMod64(ulMultiplier, ullDivisor, &ullMod);
if(ullMod == 0U)
{
/* Evenly divides, so check that we won't overflow, and finish up.
*/
/* Must recalculate ullTemp relative to ullBase
*/
ullTemp = RedUint64DivMod64(UINT64_MAX, ullBase, NULL);
if(ullWideMultiplier > ullTemp)
{
return UINT64_MAX;
}
else
{
uint32_t ulNarrowMultiplier = (uint32_t)ullWideMultiplier;
/* We've validated that this will not overflow.
*/
ullBase *= ulNarrowMultiplier;
return ullBase;
}
}
/* CODE-PATH #4
Neither of the multipliers is evenly divisible by the divisor, so
just punt and divide the larger number first, then do the final
multiply.
All the other attempts above would preserve precision -- this is the
only case where precision may be lost.
*/
/* If necessary reverse the ullBase and ulMultiplier operands so that
ullBase contains the larger of the two values.
*/
if(ullBase < ulMultiplier)
{
uint32_t ulTemp = ulMultiplier;
ulMultiplier = (uint32_t)ullBase;
ullBase = ulTemp;
}
ullBase = RedUint64DivMod64(ullBase, ullDivisor, NULL);
ullTemp = RedUint64DivMod32(UINT64_MAX, ulMultiplier, NULL);
if(ullBase > ullTemp)
{
return UINT64_MAX;
}
else
{
ullBase *= ulMultiplier;
return ullBase;
}
}
/* We only get to this point if either there was never any chance of
overflow, or if the pure shifting mechanism succeeded in reducing
the scale so overflow is not a problem.
*/
ullBase *= ulMultiplier;
ullBase = RedUint64DivMod64(ullBase, ullDivisor, NULL);
return ullBase;
}
/** @brief Divide a 64-bit value by a 32-bit value, returning the quotient and
the remainder.
Essentially this function does the following:
~~~{.c}
if(pulRemainder != NULL)
{
*pulRemainder = (uint32_t)(ullDividend % ulDivisor);
}
return ullDividend / ulDivisor;
~~~
However, it does so without ever actually dividing/modulating a 64-bit
value, since such operations are not allowed in all environments.
@param ullDividend The value to divide.
@param ulDivisor The value to divide by.
@param pulRemainder Populated with the remainder; may be NULL.
@return The quotient (result of the division).
*/
uint64_t RedUint64DivMod32(
uint64_t ullDividend,
uint32_t ulDivisor,
uint32_t *pulRemainder)
{
uint64_t ullQuotient;
uint32_t ulResultRemainder;
/* Check for divide by zero.
*/
if(ulDivisor == 0U)
{
REDERROR();
/* Nonsense value if no asserts.
*/
ullQuotient = UINT64_SUFFIX(0xFFFFFFFFFFFFFBAD);
ulResultRemainder = 0xFFFFFBADU;
}
else if(ullDividend <= UINT32_MAX)
{
uint32_t ulDividend = (uint32_t)ullDividend;
ullQuotient = ulDividend / ulDivisor;
ulResultRemainder = ulDividend % ulDivisor;
}
else
{
uint32_t ulResultHi;
uint32_t ulResultLo;
uint32_t ulRemainder;
uint8_t bIndex;
uint32_t ulThisDivision;
uint32_t ulMask;
uint8_t ucNextValue;
uint32_t ulInterimHi, ulInterimLo;
uint32_t ulLowDword = (uint32_t)ullDividend;
uint32_t ulHighDword = (uint32_t)(ullDividend >> 32U);
/* Compute the high part and get the remainder
*/
ulResultHi = ulHighDword / ulDivisor;
ulResultLo = 0U;
ulRemainder = ulHighDword % ulDivisor;
/* Compute the low part
*/
ulMask = 0xFF000000U;
for(bIndex = 0U; bIndex < sizeof(uint32_t); bIndex++)
{
ucNextValue = (uint8_t)((ulLowDword & ulMask) >> ((sizeof(uint32_t) - 1U - bIndex) * 8U));
ulInterimHi = ulRemainder >> 24U;
ulInterimLo = (ulRemainder << 8U) | ucNextValue;
ulThisDivision = 0U;
while(ulInterimHi != 0U)
{
uint64_t ullInterim = ((uint64_t)ulInterimHi << 32U) + ulInterimLo;
ullInterim -= ulDivisor;
ulThisDivision++;
ulInterimHi = (uint32_t)(ullInterim >> 32U);
ulInterimLo = (uint32_t)ullInterim;
}
ulThisDivision += ulInterimLo / ulDivisor;
ulRemainder = ulInterimLo % ulDivisor;
ulResultLo <<= 8U;
ulResultLo += ulThisDivision;
ulMask >>= 8U;
}
ullQuotient = ((uint64_t)ulResultHi << 32U) + ulResultLo;
ulResultRemainder = (uint32_t)(ullDividend - (ullQuotient * ulDivisor));
}
if(pulRemainder != NULL)
{
*pulRemainder = ulResultRemainder;
}
return ullQuotient;
}
/** @brief Divide a 64-bit value by a 64-bit value, returning the quotient and
the remainder.
Essentially this function does the following:
~~~{.c}
if(pullRemainder != NULL)
{
*pullRemainder = ullDividend % ullDivisor;
}
return ullDividend / ullDivisor;
~~~
However, it does so without ever actually dividing/modulating a 64-bit
value, since such operations are not allowed in all environments.
@param ullDividend The value to divide.
@param ullDivisor The value to divide by.
@param pullRemainder Populated with the remainder; may be NULL.
@return The quotient (result of the division).
*/
uint64_t RedUint64DivMod64(
uint64_t ullDividend,
uint64_t ullDivisor,
uint64_t *pullRemainder)
{
/* The variables u0, u1, etc. take on only 32-bit values, but they are
declared uint64_t to avoid some compiler warning messages and to avoid
some unnecessary EXTRs that the compiler would put in, to convert
uint64_ts to ints.
*/
uint64_t u0;
uint64_t u1;
uint64_t q0;
uint64_t q1;
uint64_t ullQuotient;
/* First the procedure takes care of the case in which the divisor is a
32-bit quantity. There are two subcases: (1) If the left half of the
dividend is less than the divisor, one execution of RedUint64DivMod32()
is all that is required (overflow is not possible). (2) Otherwise it
does two divisions, using the grade school method.
*/
if((ullDivisor >> 32U) == 0U)
{
if((ullDividend >> 32U) < ullDivisor)
{
/* If ullDividend/ullDivisor cannot overflow, just do one division.
*/
ullQuotient = RedUint64DivMod32(ullDividend, (uint32_t)ullDivisor, NULL);
}
else
{
uint32_t k;
/* If ullDividend/ullDivisor would overflow:
*/
/* Break ullDividend up into two halves.
*/
u1 = ullDividend >> 32U;
u0 = ullDividend & 0xFFFFFFFFU;
/* First quotient digit and first remainder.
*/
q1 = RedUint64DivMod32(u1, (uint32_t)ullDivisor, &k);
/* 2nd quot. digit.
*/
q0 = RedUint64DivMod32(((uint64_t)k << 32U) + u0, (uint32_t)ullDivisor, NULL);
ullQuotient = (q1 << 32U) + q0;
}
}
else
{
uint64_t n;
uint64_t v1;
n = nlz64(ullDivisor); /* 0 <= n <= 31. */
v1 = (ullDivisor << n) >> 32U; /* Normalize the divisor so its MSB is 1. */
u1 = ullDividend >> 1U; /* To ensure no overflow. */
/* Get quotient from divide unsigned insn.
*/
q1 = RedUint64DivMod32(u1, (uint32_t)v1, NULL);
q0 = (q1 << n) >> 31U; /* Undo normalization and division of ullDividend by 2. */
/* Make q0 correct or too small by 1.
*/
if(q0 != 0U)
{
q0--;
}
if((ullDividend - (q0 * ullDivisor)) >= ullDivisor)
{
q0++; /* Now q0 is correct. */
}
ullQuotient = q0;
}
if(pullRemainder != NULL)
{
*pullRemainder = ullDividend - (ullQuotient * ullDivisor);
}
return ullQuotient;
}
/** @brief Compute the number of leading zeroes in a 64-bit value.
@param ullValue The value for which to compute the NLZ.
@return The number of leading zeroes in @p ullValue.
*/
static uint32_t nlz64(
uint64_t ullValue)
{
uint32_t n;
if(ullValue == 0U)
{
n = 64U;
}
else
{
uint64_t x = ullValue;
n = 0U;
if(x <= UINT64_SUFFIX(0x00000000FFFFFFFF))
{
n += 32U;
x <<= 32U;
}
if(x <= UINT64_SUFFIX(0x0000FFFFFFFFFFFF))
{
n += 16U;
x <<= 16U;
}
if(x <= UINT64_SUFFIX(0x00FFFFFFFFFFFFFF))
{
n += 8U;
x <<= 8U;
}
if(x <= UINT64_SUFFIX(0x0FFFFFFFFFFFFFFF))
{
n += 4U;
x <<= 4U;
}
if(x <= UINT64_SUFFIX(0x3FFFFFFFFFFFFFFF))
{
n += 2U;
x <<= 2U;
}
if(x <= UINT64_SUFFIX(0x7FFFFFFFFFFFFFFF))
{
n += 1;
}
}
return n;
}
|