fastmath2_dlib_asm.S
12.1 KB
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
//===----------------------Hexagon builtin routine ------------------------===//
//
// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions.
// See https://llvm.org/LICENSE.txt for license information.
// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
//
//===----------------------------------------------------------------------===//
/* ==================================================================== */
/* FUNCTIONS Optimized double floating point operators */
/* ==================================================================== */
/* c = dadd_asm(a, b) */
/* ==================================================================== *
fast2_QDOUBLE fast2_dadd(fast2_QDOUBLE a,fast2_QDOUBLE b) {
fast2_QDOUBLE c;
lint manta = a & MANTMASK;
int expa = Q6_R_sxth_R(a) ;
lint mantb = b & MANTMASK;
int expb = Q6_R_sxth_R(b) ;
int exp, expdiff, j, k, hi, lo, cn;
lint mant;
expdiff = (int) Q6_P_vabsdiffh_PP(a, b);
expdiff = Q6_R_sxth_R(expdiff) ;
if (expdiff > 63) { expdiff = 62;}
if (expa > expb) {
exp = expa + 1;
expa = 1;
expb = expdiff + 1;
} else {
exp = expb + 1;
expb = 1;
expa = expdiff + 1;
}
mant = (manta>>expa) + (mantb>>expb);
hi = (int) (mant>>32);
lo = (int) (mant);
k = Q6_R_normamt_R(hi);
if(hi == 0 || hi == -1) k = 31+Q6_R_normamt_R(lo);
mant = (mant << k);
cn = (mant == 0x8000000000000000LL);
exp = exp - k + cn;
if (mant == 0 || mant == -1) exp = 0x8001;
c = (mant & MANTMASK) | (((lint) exp) & EXP_MASK);
return(c);
}
* ==================================================================== */
.text
.global fast2_dadd_asm
.type fast2_dadd_asm, @function
fast2_dadd_asm:
#define manta R0
#define mantexpa R1:0
#define lmanta R1:0
#define mantb R2
#define mantexpb R3:2
#define lmantb R3:2
#define expa R4
#define expb R5
#define mantexpd R7:6
#define expd R6
#define exp R8
#define c63 R9
#define lmant R1:0
#define manth R1
#define mantl R0
#define minmin R11:10 // exactly 0x000000000000008001LL
#define minminl R10
#define k R4
#define ce P0
.falign
{
mantexpd = VABSDIFFH(mantexpa, mantexpb) //represented as 0x08001LL
c63 = #62
expa = SXTH(manta)
expb = SXTH(mantb)
} {
expd = SXTH(expd)
ce = CMP.GT(expa, expb);
if ( ce.new) exp = add(expa, #1)
if (!ce.new) exp = add(expb, #1)
} {
if ( ce) expa = #1
if (!ce) expb = #1
manta.L = #0
expd = MIN(expd, c63)
} {
if (!ce) expa = add(expd, #1)
if ( ce) expb = add(expd, #1)
mantb.L = #0
minmin = #0
} {
lmanta = ASR(lmanta, expa)
lmantb = ASR(lmantb, expb)
} {
lmant = add(lmanta, lmantb)
minminl.L = #0x8001
} {
k = clb(lmant)
c63 = #58
} {
k = add(k, #-1)
p0 = cmp.gt(k, c63)
} {
mantexpa = ASL(lmant, k)
exp = SUB(exp, k)
if(p0) jump .Ldenorma
} {
manta = insert(exp, #16, #0)
jumpr r31
}
.Ldenorma:
{
mantexpa = minmin
jumpr r31
}
/* =================================================================== *
fast2_QDOUBLE fast2_dsub(fast2_QDOUBLE a,fast2_QDOUBLE b) {
fast2_QDOUBLE c;
lint manta = a & MANTMASK;
int expa = Q6_R_sxth_R(a) ;
lint mantb = b & MANTMASK;
int expb = Q6_R_sxth_R(b) ;
int exp, expdiff, j, k;
lint mant;
expdiff = (int) Q6_P_vabsdiffh_PP(a, b);
expdiff = Q6_R_sxth_R(expdiff) ;
if (expdiff > 63) { expdiff = 62;}
if (expa > expb) {
exp = expa + 1;
expa = 1;
expb = expdiff + 1;
} else {
exp = expb + 1;
expb = 1;
expa = expdiff + 1;
}
mant = (manta>>expa) - (mantb>>expb);
k = Q6_R_clb_P(mant)-1;
mant = (mant << k);
exp = exp - k;
if (mant == 0 || mant == -1) exp = 0x8001;
c = (mant & MANTMASK) | (((lint) exp) & EXP_MASK);
return(c);
}
* ==================================================================== */
.text
.global fast2_dsub_asm
.type fast2_dsub_asm, @function
fast2_dsub_asm:
#define manta R0
#define mantexpa R1:0
#define lmanta R1:0
#define mantb R2
#define mantexpb R3:2
#define lmantb R3:2
#define expa R4
#define expb R5
#define mantexpd R7:6
#define expd R6
#define exp R8
#define c63 R9
#define lmant R1:0
#define manth R1
#define mantl R0
#define minmin R11:10 // exactly 0x000000000000008001LL
#define minminl R10
#define k R4
#define ce P0
.falign
{
mantexpd = VABSDIFFH(mantexpa, mantexpb) //represented as 0x08001LL
c63 = #62
expa = SXTH(manta)
expb = SXTH(mantb)
} {
expd = SXTH(expd)
ce = CMP.GT(expa, expb);
if ( ce.new) exp = add(expa, #1)
if (!ce.new) exp = add(expb, #1)
} {
if ( ce) expa = #1
if (!ce) expb = #1
manta.L = #0
expd = MIN(expd, c63)
} {
if (!ce) expa = add(expd, #1)
if ( ce) expb = add(expd, #1)
mantb.L = #0
minmin = #0
} {
lmanta = ASR(lmanta, expa)
lmantb = ASR(lmantb, expb)
} {
lmant = sub(lmanta, lmantb)
minminl.L = #0x8001
} {
k = clb(lmant)
c63 = #58
} {
k = add(k, #-1)
p0 = cmp.gt(k, c63)
} {
mantexpa = ASL(lmant, k)
exp = SUB(exp, k)
if(p0) jump .Ldenorm
} {
manta = insert(exp, #16, #0)
jumpr r31
}
.Ldenorm:
{
mantexpa = minmin
jumpr r31
}
/* ==================================================================== *
fast2_QDOUBLE fast2_dmpy(fast2_QDOUBLE a,fast2_QDOUBLE b) {
fast2_QDOUBLE c;
lint manta = a & MANTMASK;
int expa = Q6_R_sxth_R(a) ;
lint mantb = b & MANTMASK;
int expb = Q6_R_sxth_R(b) ;
int exp, k;
lint mant;
int hia, hib, hi, lo;
unsigned int loa, lob;
hia = (int)(a >> 32);
loa = Q6_R_extractu_RII((int)manta, 31, 1);
hib = (int)(b >> 32);
lob = Q6_R_extractu_RII((int)mantb, 31, 1);
mant = Q6_P_mpy_RR(hia, lob);
mant = Q6_P_mpyacc_RR(mant,hib, loa);
mant = (mant >> 30) + (Q6_P_mpy_RR(hia, hib)<<1);
hi = (int) (mant>>32);
k = Q6_R_normamt_R(hi);
mant = mant << k;
exp = expa + expb - k;
if (mant == 0 || mant == -1) exp = 0x8001;
c = (mant & MANTMASK) | (((lint) exp) & EXP_MASK);
return(c);
}
* ==================================================================== */
.text
.global fast2_dmpy_asm
.type fast2_dmpy_asm, @function
fast2_dmpy_asm:
#define mantal R0
#define mantah R1
#define mantexpa R1:0
#define mantbl R2
#define mantbh R3
#define mantexpb R3:2
#define expa R4
#define expb R5
#define c8001 R12
#define mantexpd R7:6
#define mantdh R7
#define exp R8
#define lmantc R11:10
#define kb R9
#define guard R11
#define mantal_ R12
#define mantbl_ R13
#define min R15:14
#define minh R15
.falign
{
mantbl_= lsr(mantbl, #16)
expb = sxth(mantbl)
expa = sxth(mantal)
mantal_= lsr(mantal, #16)
}
{
lmantc = mpy(mantah, mantbh)
mantexpd = mpy(mantah, mantbl_)
mantal.L = #0x0
min = #0
}
{
lmantc = add(lmantc, lmantc)
mantexpd+= mpy(mantbh, mantal_)
mantbl.L = #0x0
minh.H = #0x8000
}
{
mantexpd = asr(mantexpd, #15)
c8001.L = #0x8001
p1 = cmp.eq(mantexpa, mantexpb)
}
{
mantexpd = add(mantexpd, lmantc)
exp = add(expa, expb)
p2 = cmp.eq(mantexpa, min)
}
{
kb = clb(mantexpd)
mantexpb = abs(mantexpd)
guard = #58
}
{
p1 = and(p1, p2)
exp = sub(exp, kb)
kb = add(kb, #-1)
p0 = cmp.gt(kb, guard)
}
{
exp = add(exp, #1)
mantexpa = asl(mantexpd, kb)
if(p1) jump .Lsat //rarely happens
}
{
mantal = insert(exp,#16, #0)
if(!p0) jumpr r31
}
{
mantal = insert(c8001,#16, #0)
jumpr r31
}
.Lsat:
{
mantexpa = #-1
}
{
mantexpa = lsr(mantexpa, #1)
}
{
mantal = insert(exp,#16, #0)
jumpr r31
}
/* ==================================================================== *
int fast2_qd2f(fast2_QDOUBLE a) {
int exp;
long long int manta;
int ic, rnd, mantb;
manta = a>>32;
exp = Q6_R_sxth_R(a) ;
ic = 0x80000000 & manta;
manta = Q6_R_abs_R_sat(manta);
mantb = (manta + rnd)>>7;
rnd = 0x40
exp = (exp + 126);
if((manta & 0xff) == rnd) rnd = 0x00;
if((manta & 0x7fffffc0) == 0x7fffffc0) {
manta = 0x0; exp++;
} else {
manta= mantb & 0x007fffff;
}
exp = (exp << 23) & 0x7fffffc0;
ic = Q6_R_addacc_RR(ic, exp, manta);
return (ic);
}
* ==================================================================== */
.text
.global fast2_qd2f_asm
.type fast2_qd2f_asm, @function
fast2_qd2f_asm:
#define mantah R1
#define mantal R0
#define cff R0
#define mant R3
#define expo R4
#define rnd R5
#define mask R6
#define c07f R7
#define c80 R0
#define mantb R2
#define ic R0
.falign
{
mant = abs(mantah):sat
expo = sxth(mantal)
rnd = #0x40
mask.L = #0xffc0
}
{
cff = extractu(mant, #8, #0)
p2 = cmp.gt(expo, #126)
p3 = cmp.ge(expo, #-126)
mask.H = #0x7fff
}
{
p1 = cmp.eq(cff,#0x40)
if(p1.new) rnd = #0
expo = add(expo, #126)
if(!p3) jump .Lmin
}
{
p0 = bitsset(mant, mask)
c80.L = #0x0000
mantb = add(mant, rnd)
c07f = lsr(mask, #8)
}
{
if(p0) expo = add(expo, #1)
if(p0) mant = #0
mantb = lsr(mantb, #7)
c80.H = #0x8000
}
{
ic = and(c80, mantah)
mask &= asl(expo, #23)
if(!p0) mant = and(mantb, c07f)
if(p2) jump .Lmax
}
{
ic += add(mask, mant)
jumpr r31
}
.Lmax:
{
ic.L = #0xffff;
}
{
ic.H = #0x7f7f;
jumpr r31
}
.Lmin:
{
ic = #0x0
jumpr r31
}
/* ==================================================================== *
fast2_QDOUBLE fast2_f2qd(int ia) {
lint exp;
lint mant;
fast2_QDOUBLE c;
mant = ((ia << 7) | 0x40000000)&0x7fffff80 ;
if (ia & 0x80000000) mant = -mant;
exp = ((ia >> 23) & 0xFFLL) - 126;
c = (mant<<32) | Q6_R_zxth_R(exp);;
return(c);
}
* ==================================================================== */
.text
.global fast2_f2qd_asm
.type fast2_f2qd_asm, @function
fast2_f2qd_asm:
#define ia R0
#define mag R3
#define mantr R1
#define expr R0
#define zero R2
#define maxneg R5:4
#define maxnegl R4
.falign
{
mantr = asl(ia, #7)
p0 = tstbit(ia, #31)
maxneg = #0
mag = add(ia,ia)
}
{
mantr = setbit(mantr, #30)
expr= extractu(ia,#8,#23)
maxnegl.L = #0x8001
p1 = cmp.eq(mag, #0)
}
{
mantr= extractu(mantr, #31, #0)
expr= add(expr, #-126)
zero = #0
if(p1) jump .Lminqd
}
{
expr = zxth(expr)
if(p0) mantr= sub(zero, mantr)
jumpr r31
}
.Lminqd:
{
R1:0 = maxneg
jumpr r31
}