3 // Copyright (c) 2010 ARM Limited
6 // The license below extends only to copyright in the software and shall
7 // not be construed as granting a license to any other intellectual
8 // property including but not limited to intellectual property relating
9 // to a hardware implementation of the functionality of the software
10 // licensed hereunder. You may use the software subject to the license
11 // terms below provided that you ensure that this notice is replicated
12 // unmodified and in its entirety in all distributions of the software,
13 // modified or unmodified, in source code or in binary form.
15 // Copyright (c) 2007-2008 The Florida State University
16 // All rights reserved.
18 // Redistribution and use in source and binary forms, with or without
19 // modification, are permitted provided that the following conditions are
20 // met: redistributions of source code must retain the above copyright
21 // notice, this list of conditions and the following disclaimer;
22 // redistributions in binary form must reproduce the above copyright
23 // notice, this list of conditions and the following disclaimer in the
24 // documentation and/or other materials provided with the distribution;
25 // neither the name of the copyright holders nor the names of its
26 // contributors may be used to endorse or promote products derived from
27 // this software without specific prior written permission.
29 // THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
30 // "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
31 // LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR
32 // A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT
33 // OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,
34 // SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT
35 // LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
36 // DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
37 // THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
38 // (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
39 // OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
41 // Authors: Stephen Hines
43 ////////////////////////////////////////////////////////////////////
45 // Floating Point operate instructions
48 def template FPAExecute {{
49 Fault %(class_name)s::execute(%(CPU_exec_context)s *xc, Trace::InstRecord *traceData) const
51 Fault fault = NoFault;
58 if (%(predicate_test)s) {
60 if (fault == NoFault) {
69 def template FloatDoubleDecode {{
71 ArmStaticInst *i = NULL;
72 switch (OPCODE_19 << 1 | OPCODE_7)
75 i = (ArmStaticInst *)new %(class_name)sS(machInst);
78 i = (ArmStaticInst *)new %(class_name)sD(machInst);
83 panic("Cannot decode float/double nature of the instruction");
89 // Primary format for float point operate instructions:
90 def format FloatOp(code, *flags) {{
94 iop = InstObjParams(name, Name, 'PredOp',
96 "predicate_test": predicateTest},
98 header_output = BasicDeclare.subst(iop)
99 decoder_output = BasicConstructor.subst(iop)
100 exec_output = FPAExecute.subst(iop)
103 sng_iop = InstObjParams(name, Name+'S', 'PredOp',
105 "predicate_test": predicateTest},
107 header_output += BasicDeclare.subst(sng_iop)
108 decoder_output += BasicConstructor.subst(sng_iop)
109 exec_output += FPAExecute.subst(sng_iop)
111 dbl_code = re.sub(r'\.sf', '.df', orig_code)
114 dbl_iop = InstObjParams(name, Name+'D', 'PredOp',
116 "predicate_test": predicateTest},
118 header_output += BasicDeclare.subst(dbl_iop)
119 decoder_output += BasicConstructor.subst(dbl_iop)
120 exec_output += FPAExecute.subst(dbl_iop)
122 decode_block = FloatDoubleDecode.subst(iop)
127 uint16_t _in, _iz, _ic, _iv;
129 _in = %(fReg1)s < %(fReg2)s;
130 _iz = %(fReg1)s == %(fReg2)s;
131 _ic = %(fReg1)s >= %(fReg2)s;
132 _iv = (isnan(%(fReg1)s) || isnan(%(fReg2)s)) & 1;
134 CondCodes = _in << 31 | _iz << 30 | _ic << 29 | _iv << 28 |
135 (CondCodes & 0x0FFFFFFF);
139 def format FloatCmp(fReg1, fReg2, *flags) {{
140 code = calcFPCcCode % vars()
141 iop = InstObjParams(name, Name, 'PredOp',
143 "predicate_test": predicateTest},
145 header_output = BasicDeclare.subst(iop)
146 decoder_output = BasicConstructor.subst(iop)
147 decode_block = BasicDecode.subst(iop)
148 exec_output = FPAExecute.subst(iop)
154 decodeExtensionRegLoadStore(ExtMachInst machInst);
158 decodeExtensionRegLoadStore(ExtMachInst machInst)
160 const uint32_t opcode = bits(machInst, 24, 20);
161 const uint32_t offset = bits(machInst, 7, 0);
162 const bool single = (bits(machInst, 8) == 0);
163 const IntRegIndex rn = (IntRegIndex)(uint32_t)bits(machInst, 19, 16);
166 vd = (RegIndex)(uint32_t)((bits(machInst, 15, 12) << 1) |
169 vd = (RegIndex)(uint32_t)((bits(machInst, 15, 12) << 1) |
170 (bits(machInst, 22) << 5));
172 switch (bits(opcode, 4, 3)) {
174 if (bits(opcode, 4, 1) == 0x2 &&
175 !(machInst.thumb == 1 && bits(machInst, 28) == 1) &&
176 !(machInst.thumb == 0 && machInst.condCode == 0xf)) {
177 if ((bits(machInst, 7, 4) & 0xd) != 1) {
180 const IntRegIndex rt =
181 (IntRegIndex)(uint32_t)bits(machInst, 15, 12);
182 const IntRegIndex rt2 =
183 (IntRegIndex)(uint32_t)bits(machInst, 19, 16);
184 const bool op = bits(machInst, 20);
187 vm = (bits(machInst, 3, 0) << 1) | bits(machInst, 5);
189 vm = (bits(machInst, 3, 0) << 1) |
190 (bits(machInst, 5) << 5);
193 return new Vmov2Core2Reg(machInst, rt, rt2,
196 return new Vmov2Reg2Core(machInst, (IntRegIndex)vm,
202 switch (bits(opcode, 1, 0)) {
204 return new VLdmStm(machInst, rn, vd, single,
205 true, false, false, offset);
207 return new VLdmStm(machInst, rn, vd, single,
208 true, false, true, offset);
210 return new VLdmStm(machInst, rn, vd, single,
211 true, true, false, offset);
213 // If rn == sp, then this is called vpop.
214 return new VLdmStm(machInst, rn, vd, single,
215 true, true, true, offset);
218 if (bits(opcode, 1, 0) == 0x2) {
219 // If rn == sp, then this is called vpush.
220 return new VLdmStm(machInst, rn, vd, single,
221 false, true, false, offset);
222 } else if (bits(opcode, 1, 0) == 0x3) {
223 return new VLdmStm(machInst, rn, vd, single,
224 false, true, true, offset);
226 // Fall through on purpose
228 const bool up = (bits(machInst, 23) == 1);
229 const uint32_t imm = bits(machInst, 7, 0) << 2;
232 vd = (RegIndex)(uint32_t)((bits(machInst, 15, 12) << 1) |
233 (bits(machInst, 22)));
235 vd = (RegIndex)(uint32_t)((bits(machInst, 15, 12) << 1) |
236 (bits(machInst, 22) << 5));
238 if (bits(opcode, 1, 0) == 0x0) {
241 return new %(vstr_us)s(machInst, vd, rn, up, imm);
243 return new %(vstr_s)s(machInst, vd, rn, up, imm);
247 return new %(vstr_ud)s(machInst, vd, vd + 1,
250 return new %(vstr_d)s(machInst, vd, vd + 1,
254 } else if (bits(opcode, 1, 0) == 0x1) {
257 return new %(vldr_us)s(machInst, vd, rn, up, imm);
259 return new %(vldr_s)s(machInst, vd, rn, up, imm);
263 return new %(vldr_ud)s(machInst, vd, vd + 1,
266 return new %(vldr_d)s(machInst, vd, vd + 1,
272 return new Unknown(machInst);
275 "vldr_us" : "VLDR_" + loadImmClassName(False, True, False),
276 "vldr_s" : "VLDR_" + loadImmClassName(False, False, False),
277 "vldr_ud" : "VLDR_" + loadDoubleImmClassName(False, True, False),
278 "vldr_d" : "VLDR_" + loadDoubleImmClassName(False, False, False),
279 "vstr_us" : "VSTR_" + storeImmClassName(False, True, False),
280 "vstr_s" : "VSTR_" + storeImmClassName(False, False, False),
281 "vstr_ud" : "VSTR_" + storeDoubleImmClassName(False, True, False),
282 "vstr_d" : "VSTR_" + storeDoubleImmClassName(False, False, False)
286 def format ExtensionRegLoadStore() {{
288 return decodeExtensionRegLoadStore(machInst);
295 decodeShortFpTransfer(ExtMachInst machInst);
299 decodeShortFpTransfer(ExtMachInst machInst)
301 const uint32_t l = bits(machInst, 20);
302 const uint32_t c = bits(machInst, 8);
303 const uint32_t a = bits(machInst, 23, 21);
304 const uint32_t b = bits(machInst, 6, 5);
305 if ((machInst.thumb == 1 && bits(machInst, 28) == 1) ||
306 (machInst.thumb == 0 && machInst.condCode == 0xf)) {
307 return new Unknown(machInst);
309 if (l == 0 && c == 0) {
311 const uint32_t vn = (bits(machInst, 19, 16) << 1) |
313 const IntRegIndex rt =
314 (IntRegIndex)(uint32_t)bits(machInst, 15, 12);
315 if (bits(machInst, 20) == 1) {
316 return new VmovRegCoreW(machInst, rt, (IntRegIndex)vn);
318 return new VmovCoreRegW(machInst, (IntRegIndex)vn, rt);
320 } else if (a == 0x7) {
321 const IntRegIndex rt =
322 (IntRegIndex)(uint32_t)bits(machInst, 15, 12);
323 uint32_t specReg = bits(machInst, 19, 16);
326 specReg = MISCREG_FPSID;
329 specReg = MISCREG_FPSCR;
332 specReg = MISCREG_FPEXC;
335 return new Unknown(machInst);
337 return new Vmsr(machInst, (IntRegIndex)specReg, rt);
339 } else if (l == 0 && c == 1) {
340 if (bits(a, 2) == 0) {
341 uint32_t vd = (bits(machInst, 7) << 5) |
342 (bits(machInst, 19, 16) << 1);
343 uint32_t index, size;
344 const IntRegIndex rt =
345 (IntRegIndex)(uint32_t)bits(machInst, 15, 12);
346 if (bits(machInst, 22) == 1) {
348 index = (bits(machInst, 21) << 2) |
349 bits(machInst, 6, 5);
350 } else if (bits(machInst, 5) == 1) {
352 index = (bits(machInst, 21) << 1) |
354 } else if (bits(machInst, 6) == 0) {
356 index = bits(machInst, 21);
358 return new Unknown(machInst);
360 if (index >= (32 / size)) {
361 index -= (32 / size);
366 return new VmovCoreRegB(machInst, (IntRegIndex)vd,
369 return new VmovCoreRegH(machInst, (IntRegIndex)vd,
372 return new VmovCoreRegW(machInst, (IntRegIndex)vd, rt);
374 } else if (bits(b, 1) == 0) {
376 return new WarnUnimplemented("vdup", machInst);
378 } else if (l == 1 && c == 0) {
380 const uint32_t vn = (bits(machInst, 19, 16) << 1) |
382 const IntRegIndex rt =
383 (IntRegIndex)(uint32_t)bits(machInst, 15, 12);
384 if (bits(machInst, 20) == 1) {
385 return new VmovRegCoreW(machInst, rt, (IntRegIndex)vn);
387 return new VmovCoreRegW(machInst, (IntRegIndex)vn, rt);
390 const IntRegIndex rt =
391 (IntRegIndex)(uint32_t)bits(machInst, 15, 12);
392 uint32_t specReg = bits(machInst, 19, 16);
395 specReg = MISCREG_FPSID;
398 specReg = MISCREG_FPSCR;
401 specReg = MISCREG_MVFR1;
404 specReg = MISCREG_MVFR0;
407 specReg = MISCREG_FPEXC;
410 return new Unknown(machInst);
412 return new Vmrs(machInst, rt, (IntRegIndex)specReg);
415 uint32_t vd = (bits(machInst, 7) << 5) |
416 (bits(machInst, 19, 16) << 1);
417 uint32_t index, size;
418 const IntRegIndex rt =
419 (IntRegIndex)(uint32_t)bits(machInst, 15, 12);
420 const bool u = (bits(machInst, 23) == 1);
421 if (bits(machInst, 22) == 1) {
423 index = (bits(machInst, 21) << 2) |
424 bits(machInst, 6, 5);
425 } else if (bits(machInst, 5) == 1) {
427 index = (bits(machInst, 21) << 1) |
429 } else if (bits(machInst, 6) == 0 && !u) {
431 index = bits(machInst, 21);
433 return new Unknown(machInst);
435 if (index >= (32 / size)) {
436 index -= (32 / size);
442 return new VmovRegCoreUB(machInst, rt,
443 (IntRegIndex)vd, index);
445 return new VmovRegCoreSB(machInst, rt,
446 (IntRegIndex)vd, index);
450 return new VmovRegCoreUH(machInst, rt,
451 (IntRegIndex)vd, index);
453 return new VmovRegCoreSH(machInst, rt,
454 (IntRegIndex)vd, index);
457 return new VmovRegCoreW(machInst, rt, (IntRegIndex)vd);
460 return new Unknown(machInst);
465 def format ShortFpTransfer() {{
467 return decodeShortFpTransfer(machInst);
474 decodeVfpData(ExtMachInst machInst);
478 decodeVfpData(ExtMachInst machInst)
480 const uint32_t opc1 = bits(machInst, 23, 20);
481 const uint32_t opc2 = bits(machInst, 19, 16);
482 const uint32_t opc3 = bits(machInst, 7, 6);
483 //const uint32_t opc4 = bits(machInst, 3, 0);
484 const bool single = (bits(machInst, 8) == 0);
489 vd = (IntRegIndex)(bits(machInst, 22) |
490 (bits(machInst, 15, 12) << 1));
491 vm = (IntRegIndex)(bits(machInst, 5) |
492 (bits(machInst, 3, 0) << 1));
493 vn = (IntRegIndex)(bits(machInst, 7) |
494 (bits(machInst, 19, 16) << 1));
496 vd = (IntRegIndex)((bits(machInst, 22) << 5) |
497 (bits(machInst, 15, 12) << 1));
498 vm = (IntRegIndex)((bits(machInst, 5) << 5) |
499 (bits(machInst, 3, 0) << 1));
500 vn = (IntRegIndex)((bits(machInst, 7) << 5) |
501 (bits(machInst, 19, 16) << 1));
503 switch (opc1 & 0xb /* 1011 */) {
505 if (bits(machInst, 6) == 0) {
507 return new VmlaS(machInst, vd, vn, vm);
509 return new VmlaD(machInst, vd, vn, vm);
513 return new VmlsS(machInst, vd, vn, vm);
515 return new VmlsD(machInst, vd, vn, vm);
519 if (bits(machInst, 6) == 1) {
521 return new VnmlaS(machInst, vd, vn, vm);
523 return new VnmlaD(machInst, vd, vn, vm);
527 return new VnmlsS(machInst, vd, vn, vm);
529 return new VnmlsD(machInst, vd, vn, vm);
533 if ((opc3 & 0x1) == 0) {
535 return new VmulS(machInst, vd, vn, vm);
537 return new VmulD(machInst, vd, vn, vm);
541 return new VnmulS(machInst, vd, vn, vm);
543 return new VnmulD(machInst, vd, vn, vm);
547 if ((opc3 & 0x1) == 0) {
549 return new VaddS(machInst, vd, vn, vm);
551 return new VaddD(machInst, vd, vn, vm);
555 return new VsubS(machInst, vd, vn, vm);
557 return new VsubD(machInst, vd, vn, vm);
561 if ((opc3 & 0x1) == 0) {
563 return new VdivS(machInst, vd, vn, vm);
565 return new VdivD(machInst, vd, vn, vm);
570 if ((opc3 & 0x1) == 0) {
571 const uint32_t baseImm =
572 bits(machInst, 3, 0) | (bits(machInst, 19, 16) << 4);
574 uint32_t imm = vfp_modified_imm(baseImm, false);
575 return new VmovImmS(machInst, vd, imm);
577 uint64_t imm = vfp_modified_imm(baseImm, true);
578 return new VmovImmD(machInst, vd, imm);
585 return new VmovRegS(machInst, vd, vm);
587 return new VmovRegD(machInst, vd, vm);
591 return new VabsS(machInst, vd, vm);
593 return new VabsD(machInst, vd, vm);
599 return new VnegS(machInst, vd, vm);
601 return new VnegD(machInst, vd, vm);
605 return new VsqrtS(machInst, vd, vm);
607 return new VsqrtD(machInst, vd, vm);
612 // Between half and single precision.
613 return new WarnUnimplemented("vcvtb, vcvtt", machInst);
616 return new WarnUnimplemented("vcmp, vcmpe", machInst);
620 vm = (IntRegIndex)(bits(machInst, 5) |
621 (bits(machInst, 3, 0) << 1));
622 return new VcvtFpSFpD(machInst, vd, vm);
624 vd = (IntRegIndex)(bits(machInst, 22) |
625 (bits(machInst, 15, 12) << 1));
626 return new VcvtFpDFpS(machInst, vd, vm);
631 if (bits(machInst, 7) == 0) {
633 return new VcvtUIntFpS(machInst, vd, vm);
635 vm = (IntRegIndex)(bits(machInst, 5) |
636 (bits(machInst, 3, 0) << 1));
637 return new VcvtUIntFpD(machInst, vd, vm);
641 return new VcvtSIntFpS(machInst, vd, vm);
643 vm = (IntRegIndex)(bits(machInst, 5) |
644 (bits(machInst, 3, 0) << 1));
645 return new VcvtSIntFpD(machInst, vd, vm);
650 // Between FP and fixed point.
651 return new WarnUnimplemented("vcvt", machInst);
654 return new VcvtFpUIntS(machInst, vd, vm);
656 vd = (IntRegIndex)(bits(machInst, 22) |
657 (bits(machInst, 15, 12) << 1));
658 return new VcvtFpUIntD(machInst, vd, vm);
662 return new VcvtFpSIntS(machInst, vd, vm);
664 vd = (IntRegIndex)(bits(machInst, 22) |
665 (bits(machInst, 15, 12) << 1));
666 return new VcvtFpSIntD(machInst, vd, vm);
670 // Between FP and fixed point.
671 return new WarnUnimplemented("vcvt", machInst);
675 return new Unknown(machInst);
680 def format VfpData() {{
682 return decodeVfpData(machInst);