1/* ===-------- Intrin.h ---------------------------------------------------===
2 *
3 * Permission is hereby granted, free of charge, to any person obtaining a copy
4 * of this software and associated documentation files (the "Software"), to deal
5 * in the Software without restriction, including without limitation the rights
6 * to use, copy, modify, merge, publish, distribute, sublicense, and/or sell
7 * copies of the Software, and to permit persons to whom the Software is
8 * furnished to do so, subject to the following conditions:
9 *
10 * The above copyright notice and this permission notice shall be included in
11 * all copies or substantial portions of the Software.
12 *
13 * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
14 * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
15 * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
16 * AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
17 * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
18 * OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN
19 * THE SOFTWARE.
20 *
21 *===-----------------------------------------------------------------------===
22 */
23
24/* Only include this if we're compiling for the windows platform. */
25#ifndef _MSC_VER
26#include_next <Intrin.h>
27#else
28
29#ifndef __INTRIN_H
30#define __INTRIN_H
31
32/* First include the standard intrinsics. */
33#if defined(__i386__) || defined(__x86_64__)
34#include <x86intrin.h>
35#endif
36
37/* For the definition of jmp_buf. */
38#if __STDC_HOSTED__
39#include <setjmp.h>
40#endif
41
42#ifdef __cplusplus
43extern "C" {
44#endif
45
46#if defined(__MMX__)
47/* And the random ones that aren't in those files. */
48__m64 _m_from_float(float);
49__m64 _m_from_int(int _l);
50void _m_prefetch(void *);
51float _m_to_float(__m64);
52int _m_to_int(__m64 _M);
53#endif
54
55/* Other assorted instruction intrinsics. */
56void __addfsbyte(unsigned long, unsigned char);
57void __addfsdword(unsigned long, unsigned long);
58void __addfsword(unsigned long, unsigned short);
59void __code_seg(const char *);
60static __inline__
61void __cpuid(int[4], int);
62static __inline__
63void __cpuidex(int[4], int, int);
64void __debugbreak(void);
65__int64 __emul(int, int);
66unsigned __int64 __emulu(unsigned int, unsigned int);
67void __cdecl __fastfail(unsigned int);
68unsigned int __getcallerseflags(void);
69static __inline__
70void __halt(void);
71unsigned char __inbyte(unsigned short);
72void __inbytestring(unsigned short, unsigned char *, unsigned long);
73void __incfsbyte(unsigned long);
74void __incfsdword(unsigned long);
75void __incfsword(unsigned long);
76unsigned long __indword(unsigned short);
77void __indwordstring(unsigned short, unsigned long *, unsigned long);
78void __int2c(void);
79void __invlpg(void *);
80unsigned short __inword(unsigned short);
81void __inwordstring(unsigned short, unsigned short *, unsigned long);
82void __lidt(void *);
83unsigned __int64 __ll_lshift(unsigned __int64, int);
84__int64 __ll_rshift(__int64, int);
85void __llwpcb(void *);
86unsigned char __lwpins32(unsigned int, unsigned int, unsigned int);
87void __lwpval32(unsigned int, unsigned int, unsigned int);
88unsigned int __lzcnt(unsigned int);
89unsigned short __lzcnt16(unsigned short);
90static __inline__
91void __movsb(unsigned char *, unsigned char const *, size_t);
92static __inline__
93void __movsd(unsigned long *, unsigned long const *, size_t);
94static __inline__
95void __movsw(unsigned short *, unsigned short const *, size_t);
96void __nop(void);
97void __nvreg_restore_fence(void);
98void __nvreg_save_fence(void);
99void __outbyte(unsigned short, unsigned char);
100void __outbytestring(unsigned short, unsigned char *, unsigned long);
101void __outdword(unsigned short, unsigned long);
102void __outdwordstring(unsigned short, unsigned long *, unsigned long);
103void __outword(unsigned short, unsigned short);
104void __outwordstring(unsigned short, unsigned short *, unsigned long);
105static __inline__
106unsigned int __popcnt(unsigned int);
107static __inline__
108unsigned short __popcnt16(unsigned short);
109unsigned long __readcr0(void);
110unsigned long __readcr2(void);
111static __inline__
112unsigned long __readcr3(void);
113unsigned long __readcr4(void);
114unsigned long __readcr8(void);
115unsigned int __readdr(unsigned int);
116#ifdef __i386__
117static __inline__
118unsigned char __readfsbyte(unsigned long);
119static __inline__
120unsigned long __readfsdword(unsigned long);
121static __inline__
122unsigned __int64 __readfsqword(unsigned long);
123static __inline__
124unsigned short __readfsword(unsigned long);
125#endif
126static __inline__
127unsigned __int64 __readmsr(unsigned long);
128unsigned __int64 __readpmc(unsigned long);
129unsigned long __segmentlimit(unsigned long);
130void __sidt(void *);
131void *__slwpcb(void);
132static __inline__
133void __stosb(unsigned char *, unsigned char, size_t);
134static __inline__
135void __stosd(unsigned long *, unsigned long, size_t);
136static __inline__
137void __stosw(unsigned short *, unsigned short, size_t);
138void __svm_clgi(void);
139void __svm_invlpga(void *, int);
140void __svm_skinit(int);
141void __svm_stgi(void);
142void __svm_vmload(size_t);
143void __svm_vmrun(size_t);
144void __svm_vmsave(size_t);
145void __ud2(void);
146unsigned __int64 __ull_rshift(unsigned __int64, int);
147void __vmx_off(void);
148void __vmx_vmptrst(unsigned __int64 *);
149void __wbinvd(void);
150void __writecr0(unsigned int);
151static __inline__
152void __writecr3(unsigned int);
153void __writecr4(unsigned int);
154void __writecr8(unsigned int);
155void __writedr(unsigned int, unsigned int);
156void __writefsbyte(unsigned long, unsigned char);
157void __writefsdword(unsigned long, unsigned long);
158void __writefsqword(unsigned long, unsigned __int64);
159void __writefsword(unsigned long, unsigned short);
160void __writemsr(unsigned long, unsigned __int64);
161static __inline__
162void *_AddressOfReturnAddress(void);
163unsigned int _andn_u32(unsigned int, unsigned int);
164unsigned int _bextr_u32(unsigned int, unsigned int, unsigned int);
165unsigned int _bextri_u32(unsigned int, unsigned int);
166static __inline__
167unsigned char _BitScanForward(unsigned long *_Index, unsigned long _Mask);
168static __inline__
169unsigned char _BitScanReverse(unsigned long *_Index, unsigned long _Mask);
170static __inline__
171unsigned char _bittest(long const *, long);
172static __inline__
173unsigned char _bittestandcomplement(long *, long);
174static __inline__
175unsigned char _bittestandreset(long *, long);
176static __inline__
177unsigned char _bittestandset(long *, long);
178unsigned int _blcfill_u32(unsigned int);
179unsigned int _blci_u32(unsigned int);
180unsigned int _blcic_u32(unsigned int);
181unsigned int _blcmsk_u32(unsigned int);
182unsigned int _blcs_u32(unsigned int);
183unsigned int _blsfill_u32(unsigned int);
184unsigned int _blsi_u32(unsigned int);
185unsigned int _blsic_u32(unsigned int);
186unsigned int _blsmsk_u32(unsigned int);
187unsigned int _blsr_u32(unsigned int);
188unsigned __int64 __cdecl _byteswap_uint64(unsigned __int64);
189unsigned long __cdecl _byteswap_ulong(unsigned long);
190unsigned short __cdecl _byteswap_ushort(unsigned short);
191unsigned _bzhi_u32(unsigned int, unsigned int);
192void __cdecl _disable(void);
193void __cdecl _enable(void);
194void __cdecl _fxrstor(void const *);
195void __cdecl _fxsave(void *);
196long _InterlockedAddLargeStatistic(__int64 volatile *_Addend, long _Value);
197static __inline__
198long _InterlockedAnd(long volatile *_Value, long _Mask);
199static __inline__
200short _InterlockedAnd16(short volatile *_Value, short _Mask);
201static __inline__
202char _InterlockedAnd8(char volatile *_Value, char _Mask);
203unsigned char _interlockedbittestandreset(long volatile *, long);
204static __inline__
205unsigned char _interlockedbittestandset(long volatile *, long);
206static __inline__
207long __cdecl _InterlockedCompareExchange(long volatile *_Destination,
208                                         long _Exchange, long _Comparand);
209long _InterlockedCompareExchange_HLEAcquire(long volatile *, long, long);
210long _InterlockedCompareExchange_HLERelease(long volatile *, long, long);
211static __inline__
212short _InterlockedCompareExchange16(short volatile *_Destination,
213                                    short _Exchange, short _Comparand);
214static __inline__
215__int64 _InterlockedCompareExchange64(__int64 volatile *_Destination,
216                                      __int64 _Exchange, __int64 _Comparand);
217__int64 _InterlockedcompareExchange64_HLEAcquire(__int64 volatile *, __int64,
218                                                 __int64);
219__int64 _InterlockedCompareExchange64_HLERelease(__int64 volatile *, __int64,
220                                                 __int64);
221static __inline__
222char _InterlockedCompareExchange8(char volatile *_Destination, char _Exchange,
223                                  char _Comparand);
224void *_InterlockedCompareExchangePointer_HLEAcquire(void *volatile *, void *,
225                                                    void *);
226void *_InterlockedCompareExchangePointer_HLERelease(void *volatile *, void *,
227                                                    void *);
228static __inline__
229long __cdecl _InterlockedDecrement(long volatile *_Addend);
230static __inline__
231short _InterlockedDecrement16(short volatile *_Addend);
232long _InterlockedExchange(long volatile *_Target, long _Value);
233static __inline__
234short _InterlockedExchange16(short volatile *_Target, short _Value);
235static __inline__
236char _InterlockedExchange8(char volatile *_Target, char _Value);
237static __inline__
238long __cdecl _InterlockedExchangeAdd(long volatile *_Addend, long _Value);
239long _InterlockedExchangeAdd_HLEAcquire(long volatile *, long);
240long _InterlockedExchangeAdd_HLERelease(long volatile *, long);
241static __inline__
242short _InterlockedExchangeAdd16(short volatile *_Addend, short _Value);
243__int64 _InterlockedExchangeAdd64_HLEAcquire(__int64 volatile *, __int64);
244__int64 _InterlockedExchangeAdd64_HLERelease(__int64 volatile *, __int64);
245static __inline__
246char _InterlockedExchangeAdd8(char volatile *_Addend, char _Value);
247static __inline__
248long __cdecl _InterlockedIncrement(long volatile *_Addend);
249static __inline__
250short _InterlockedIncrement16(short volatile *_Addend);
251static __inline__
252long _InterlockedOr(long volatile *_Value, long _Mask);
253static __inline__
254short _InterlockedOr16(short volatile *_Value, short _Mask);
255static __inline__
256char _InterlockedOr8(char volatile *_Value, char _Mask);
257static __inline__
258long _InterlockedXor(long volatile *_Value, long _Mask);
259static __inline__
260short _InterlockedXor16(short volatile *_Value, short _Mask);
261static __inline__
262char _InterlockedXor8(char volatile *_Value, char _Mask);
263void __cdecl _invpcid(unsigned int, void *);
264static __inline__
265unsigned long __cdecl _lrotl(unsigned long, int);
266static __inline__
267unsigned long __cdecl _lrotr(unsigned long, int);
268static __inline__
269unsigned int _lzcnt_u32(unsigned int);
270static __inline__
271void _ReadBarrier(void);
272static __inline__
273void _ReadWriteBarrier(void);
274static __inline__
275void *_ReturnAddress(void);
276unsigned int _rorx_u32(unsigned int, const unsigned int);
277int __cdecl _rdrand16_step(unsigned short *);
278int __cdecl _rdrand32_step(unsigned int *);
279static __inline__
280unsigned int __cdecl _rotl(unsigned int _Value, int _Shift);
281static __inline__
282unsigned short _rotl16(unsigned short _Value, unsigned char _Shift);
283static __inline__
284unsigned __int64 __cdecl _rotl64(unsigned __int64 _Value, int _Shift);
285static __inline__
286unsigned char _rotl8(unsigned char _Value, unsigned char _Shift);
287static __inline__
288unsigned int __cdecl _rotr(unsigned int _Value, int _Shift);
289static __inline__
290unsigned short _rotr16(unsigned short _Value, unsigned char _Shift);
291static __inline__
292unsigned __int64 __cdecl _rotr64(unsigned __int64 _Value, int _Shift);
293static __inline__
294unsigned char _rotr8(unsigned char _Value, unsigned char _Shift);
295int _sarx_i32(int, unsigned int);
296#if __STDC_HOSTED__
297int __cdecl _setjmp(jmp_buf);
298#endif
299unsigned int _shlx_u32(unsigned int, unsigned int);
300unsigned int _shrx_u32(unsigned int, unsigned int);
301void _Store_HLERelease(long volatile *, long);
302void _Store64_HLERelease(__int64 volatile *, __int64);
303void _StorePointer_HLERelease(void *volatile *, void *);
304unsigned int _t1mskc_u32(unsigned int);
305unsigned int _tzcnt_u32(unsigned int);
306unsigned int _tzmsk_u32(unsigned int);
307static __inline__
308void _WriteBarrier(void);
309void _xabort(const unsigned int imm);
310unsigned __int32 xbegin(void);
311void _xend(void);
312static __inline__
313unsigned __int64 __cdecl _xgetbv(unsigned int);
314void __cdecl _xrstor(void const *, unsigned __int64);
315void __cdecl _xsave(void *, unsigned __int64);
316void __cdecl _xsaveopt(void *, unsigned __int64);
317void __cdecl _xsetbv(unsigned int, unsigned __int64);
318unsigned char _xtest(void);
319
320/* These additional intrinsics are turned on in x64/amd64/x86_64 mode. */
321#ifdef __x86_64__
322void __addgsbyte(unsigned long, unsigned char);
323void __addgsdword(unsigned long, unsigned long);
324void __addgsqword(unsigned long, unsigned __int64);
325void __addgsword(unsigned long, unsigned short);
326static __inline__
327void __faststorefence(void);
328void __incgsbyte(unsigned long);
329void __incgsdword(unsigned long);
330void __incgsqword(unsigned long);
331void __incgsword(unsigned long);
332unsigned char __lwpins64(unsigned __int64, unsigned int, unsigned int);
333void __lwpval64(unsigned __int64, unsigned int, unsigned int);
334unsigned __int64 __lzcnt64(unsigned __int64);
335static __inline__
336void __movsq(unsigned long long *, unsigned long long const *, size_t);
337__int64 __mulh(__int64, __int64);
338static __inline__
339unsigned __int64 __popcnt64(unsigned __int64);
340static __inline__
341unsigned char __readgsbyte(unsigned long);
342static __inline__
343unsigned long __readgsdword(unsigned long);
344static __inline__
345unsigned __int64 __readgsqword(unsigned long);
346unsigned short __readgsword(unsigned long);
347unsigned __int64 __shiftleft128(unsigned __int64 _LowPart,
348                                unsigned __int64 _HighPart,
349                                unsigned char _Shift);
350unsigned __int64 __shiftright128(unsigned __int64 _LowPart,
351                                 unsigned __int64 _HighPart,
352                                 unsigned char _Shift);
353static __inline__
354void __stosq(unsigned __int64 *, unsigned __int64, size_t);
355unsigned __int64 __umulh(unsigned __int64, unsigned __int64);
356unsigned char __vmx_on(unsigned __int64 *);
357unsigned char __vmx_vmclear(unsigned __int64 *);
358unsigned char __vmx_vmlaunch(void);
359unsigned char __vmx_vmptrld(unsigned __int64 *);
360unsigned char __vmx_vmread(size_t, size_t *);
361unsigned char __vmx_vmresume(void);
362unsigned char __vmx_vmwrite(size_t, size_t);
363void __writegsbyte(unsigned long, unsigned char);
364void __writegsdword(unsigned long, unsigned long);
365void __writegsqword(unsigned long, unsigned __int64);
366void __writegsword(unsigned long, unsigned short);
367unsigned __int64 _andn_u64(unsigned __int64, unsigned __int64);
368unsigned __int64 _bextr_u64(unsigned __int64, unsigned int, unsigned int);
369unsigned __int64 _bextri_u64(unsigned __int64, unsigned int);
370static __inline__
371unsigned char _BitScanForward64(unsigned long *_Index, unsigned __int64 _Mask);
372static __inline__
373unsigned char _BitScanReverse64(unsigned long *_Index, unsigned __int64 _Mask);
374static __inline__
375unsigned char _bittest64(__int64 const *, __int64);
376static __inline__
377unsigned char _bittestandcomplement64(__int64 *, __int64);
378static __inline__
379unsigned char _bittestandreset64(__int64 *, __int64);
380static __inline__
381unsigned char _bittestandset64(__int64 *, __int64);
382unsigned __int64 _blcfill_u64(unsigned __int64);
383unsigned __int64 _blci_u64(unsigned __int64);
384unsigned __int64 _blcic_u64(unsigned __int64);
385unsigned __int64 _blcmsk_u64(unsigned __int64);
386unsigned __int64 _blcs_u64(unsigned __int64);
387unsigned __int64 _blsfill_u64(unsigned __int64);
388unsigned __int64 _blsi_u64(unsigned __int64);
389unsigned __int64 _blsic_u64(unsigned __int64);
390unsigned __int64 _blsmsk_u64(unsigned __int64);
391unsigned __int64 _blsr_u64(unsigned __int64);
392unsigned __int64 __cdecl _byteswap_uint64(unsigned __int64);
393unsigned __int64 _bzhi_u64(unsigned __int64, unsigned int);
394void __cdecl _fxrstor64(void const *);
395void __cdecl _fxsave64(void *);
396long _InterlockedAnd_np(long volatile *_Value, long _Mask);
397short _InterlockedAnd16_np(short volatile *_Value, short _Mask);
398__int64 _InterlockedAnd64_np(__int64 volatile *_Value, __int64 _Mask);
399char _InterlockedAnd8_np(char volatile *_Value, char _Mask);
400unsigned char _interlockedbittestandreset64(__int64 volatile *, __int64);
401static __inline__
402unsigned char _interlockedbittestandset64(__int64 volatile *, __int64);
403long _InterlockedCompareExchange_np(long volatile *_Destination, long _Exchange,
404                                    long _Comparand);
405unsigned char _InterlockedCompareExchange128(__int64 volatile *_Destination,
406                                             __int64 _ExchangeHigh,
407                                             __int64 _ExchangeLow,
408                                             __int64 *_CompareandResult);
409unsigned char _InterlockedCompareExchange128_np(__int64 volatile *_Destination,
410                                                __int64 _ExchangeHigh,
411                                                __int64 _ExchangeLow,
412                                                __int64 *_ComparandResult);
413short _InterlockedCompareExchange16_np(short volatile *_Destination,
414                                       short _Exchange, short _Comparand);
415__int64 _InterlockedCompareExchange64_HLEAcquire(__int64 volatile *, __int64,
416                                                 __int64);
417__int64 _InterlockedCompareExchange64_HLERelease(__int64 volatile *, __int64,
418                                                 __int64);
419__int64 _InterlockedCompareExchange64_np(__int64 volatile *_Destination,
420                                         __int64 _Exchange, __int64 _Comparand);
421void *_InterlockedCompareExchangePointer(void *volatile *_Destination,
422                                         void *_Exchange, void *_Comparand);
423void *_InterlockedCompareExchangePointer_np(void *volatile *_Destination,
424                                            void *_Exchange, void *_Comparand);
425static __inline__
426__int64 _InterlockedDecrement64(__int64 volatile *_Addend);
427static __inline__
428__int64 _InterlockedExchange64(__int64 volatile *_Target, __int64 _Value);
429static __inline__
430__int64 _InterlockedExchangeAdd64(__int64 volatile *_Addend, __int64 _Value);
431void *_InterlockedExchangePointer(void *volatile *_Target, void *_Value);
432static __inline__
433__int64 _InterlockedIncrement64(__int64 volatile *_Addend);
434long _InterlockedOr_np(long volatile *_Value, long _Mask);
435short _InterlockedOr16_np(short volatile *_Value, short _Mask);
436static __inline__
437__int64 _InterlockedOr64(__int64 volatile *_Value, __int64 _Mask);
438__int64 _InterlockedOr64_np(__int64 volatile *_Value, __int64 _Mask);
439char _InterlockedOr8_np(char volatile *_Value, char _Mask);
440long _InterlockedXor_np(long volatile *_Value, long _Mask);
441short _InterlockedXor16_np(short volatile *_Value, short _Mask);
442static __inline__
443__int64 _InterlockedXor64(__int64 volatile *_Value, __int64 _Mask);
444__int64 _InterlockedXor64_np(__int64 volatile *_Value, __int64 _Mask);
445char _InterlockedXor8_np(char volatile *_Value, char _Mask);
446static __inline__
447unsigned __int64 _lzcnt_u64(unsigned __int64);
448__int64 _mul128(__int64 _Multiplier, __int64 _Multiplicand,
449                __int64 *_HighProduct);
450unsigned int __cdecl _readfsbase_u32(void);
451unsigned __int64 __cdecl _readfsbase_u64(void);
452unsigned int __cdecl _readgsbase_u32(void);
453unsigned __int64 __cdecl _readgsbase_u64(void);
454unsigned __int64 _rorx_u64(unsigned __int64, const unsigned int);
455__int64 _sarx_i64(__int64, unsigned int);
456#if __STDC_HOSTED__
457int __cdecl _setjmpex(jmp_buf);
458#endif
459unsigned __int64 _shlx_u64(unsigned __int64, unsigned int);
460unsigned __int64 shrx_u64(unsigned __int64, unsigned int);
461unsigned __int64 _tzcnt_u64(unsigned __int64);
462unsigned __int64 _tzmsk_u64(unsigned __int64);
463unsigned __int64 _umul128(unsigned __int64 _Multiplier,
464                          unsigned __int64 _Multiplicand,
465                          unsigned __int64 *_HighProduct);
466void __cdecl _writefsbase_u32(unsigned int);
467void _cdecl _writefsbase_u64(unsigned __int64);
468void __cdecl _writegsbase_u32(unsigned int);
469void __cdecl _writegsbase_u64(unsigned __int64);
470void __cdecl _xrstor64(void const *, unsigned __int64);
471void __cdecl _xsave64(void *, unsigned __int64);
472void __cdecl _xsaveopt64(void *, unsigned __int64);
473
474#endif /* __x86_64__ */
475
476/*----------------------------------------------------------------------------*\
477|* Bit Twiddling
478\*----------------------------------------------------------------------------*/
479static __inline__ unsigned char __attribute__((__always_inline__, __nodebug__))
480_rotl8(unsigned char _Value, unsigned char _Shift) {
481  _Shift &= 0x7;
482  return _Shift ? (_Value << _Shift) | (_Value >> (8 - _Shift)) : _Value;
483}
484static __inline__ unsigned char __attribute__((__always_inline__, __nodebug__))
485_rotr8(unsigned char _Value, unsigned char _Shift) {
486  _Shift &= 0x7;
487  return _Shift ? (_Value >> _Shift) | (_Value << (8 - _Shift)) : _Value;
488}
489static __inline__ unsigned short __attribute__((__always_inline__, __nodebug__))
490_rotl16(unsigned short _Value, unsigned char _Shift) {
491  _Shift &= 0xf;
492  return _Shift ? (_Value << _Shift) | (_Value >> (16 - _Shift)) : _Value;
493}
494static __inline__ unsigned short __attribute__((__always_inline__, __nodebug__))
495_rotr16(unsigned short _Value, unsigned char _Shift) {
496  _Shift &= 0xf;
497  return _Shift ? (_Value >> _Shift) | (_Value << (16 - _Shift)) : _Value;
498}
499static __inline__ unsigned int __attribute__((__always_inline__, __nodebug__))
500_rotl(unsigned int _Value, int _Shift) {
501  _Shift &= 0x1f;
502  return _Shift ? (_Value << _Shift) | (_Value >> (32 - _Shift)) : _Value;
503}
504static __inline__ unsigned int __attribute__((__always_inline__, __nodebug__))
505_rotr(unsigned int _Value, int _Shift) {
506  _Shift &= 0x1f;
507  return _Shift ? (_Value >> _Shift) | (_Value << (32 - _Shift)) : _Value;
508}
509static __inline__ unsigned long __attribute__((__always_inline__, __nodebug__))
510_lrotl(unsigned long _Value, int _Shift) {
511  _Shift &= 0x1f;
512  return _Shift ? (_Value << _Shift) | (_Value >> (32 - _Shift)) : _Value;
513}
514static __inline__ unsigned long __attribute__((__always_inline__, __nodebug__))
515_lrotr(unsigned long _Value, int _Shift) {
516  _Shift &= 0x1f;
517  return _Shift ? (_Value >> _Shift) | (_Value << (32 - _Shift)) : _Value;
518}
519static
520__inline__ unsigned __int64 __attribute__((__always_inline__, __nodebug__))
521_rotl64(unsigned __int64 _Value, int _Shift) {
522  _Shift &= 0x3f;
523  return _Shift ? (_Value << _Shift) | (_Value >> (64 - _Shift)) : _Value;
524}
525static
526__inline__ unsigned __int64 __attribute__((__always_inline__, __nodebug__))
527_rotr64(unsigned __int64 _Value, int _Shift) {
528  _Shift &= 0x3f;
529  return _Shift ? (_Value >> _Shift) | (_Value << (64 - _Shift)) : _Value;
530}
531/*----------------------------------------------------------------------------*\
532|* Bit Counting and Testing
533\*----------------------------------------------------------------------------*/
534static __inline__ unsigned char __attribute__((__always_inline__, __nodebug__))
535_BitScanForward(unsigned long *_Index, unsigned long _Mask) {
536  if (!_Mask)
537    return 0;
538  *_Index = __builtin_ctzl(_Mask);
539  return 1;
540}
541static __inline__ unsigned char __attribute__((__always_inline__, __nodebug__))
542_BitScanReverse(unsigned long *_Index, unsigned long _Mask) {
543  if (!_Mask)
544    return 0;
545  *_Index = 31 - __builtin_clzl(_Mask);
546  return 1;
547}
548static __inline__ unsigned int __attribute__((__always_inline__, __nodebug__))
549_lzcnt_u32(unsigned int a) {
550  if (!a)
551    return 32;
552  return __builtin_clzl(a);
553}
554static __inline__ unsigned short __attribute__((__always_inline__, __nodebug__))
555__popcnt16(unsigned short value) {
556  return __builtin_popcount((int)value);
557}
558static __inline__ unsigned int __attribute__((__always_inline__, __nodebug__))
559__popcnt(unsigned int value) {
560  return __builtin_popcount(value);
561}
562static __inline__ unsigned char __attribute__((__always_inline__, __nodebug__))
563_bittest(long const *a, long b) {
564  return (*a >> b) & 1;
565}
566static __inline__ unsigned char __attribute__((__always_inline__, __nodebug__))
567_bittestandcomplement(long *a, long b) {
568  unsigned char x = (*a >> b) & 1;
569  *a = *a ^ (1 << b);
570  return x;
571}
572static __inline__ unsigned char __attribute__((__always_inline__, __nodebug__))
573_bittestandreset(long *a, long b) {
574  unsigned char x = (*a >> b) & 1;
575  *a = *a & ~(1 << b);
576  return x;
577}
578static __inline__ unsigned char __attribute__((__always_inline__, __nodebug__))
579_bittestandset(long *a, long b) {
580  unsigned char x = (*a >> b) & 1;
581  *a = *a | (1 << b);
582  return x;
583}
584#if defined(__i386__) || defined(__x86_64__)
585static __inline__ unsigned char __attribute__((__always_inline__, __nodebug__))
586_interlockedbittestandset(long volatile *__BitBase, long __BitPos) {
587  unsigned char __Res;
588  __asm__ ("xor %0, %0\n"
589           "lock bts %2, %1\n"
590           "setc %0\n"
591           : "=r" (__Res), "+m"(*__BitBase)
592           : "Ir"(__BitPos));
593  return __Res;
594}
595#endif
596#ifdef __x86_64__
597static __inline__ unsigned char __attribute__((__always_inline__, __nodebug__))
598_BitScanForward64(unsigned long *_Index, unsigned __int64 _Mask) {
599  if (!_Mask)
600    return 0;
601  *_Index = __builtin_ctzll(_Mask);
602  return 1;
603}
604static __inline__ unsigned char __attribute__((__always_inline__, __nodebug__))
605_BitScanReverse64(unsigned long *_Index, unsigned __int64 _Mask) {
606  if (!_Mask)
607    return 0;
608  *_Index = 63 - __builtin_clzll(_Mask);
609  return 1;
610}
611static
612__inline__ unsigned __int64 __attribute__((__always_inline__, __nodebug__))
613_lzcnt_u64(unsigned __int64 a) {
614  if (!a)
615    return 64;
616  return __builtin_clzll(a);
617}
618static __inline__
619unsigned __int64 __attribute__((__always_inline__, __nodebug__))
620 __popcnt64(unsigned __int64 value) {
621  return __builtin_popcountll(value);
622}
623static __inline__ unsigned char __attribute__((__always_inline__, __nodebug__))
624_bittest64(__int64 const *a, __int64 b) {
625  return (*a >> b) & 1;
626}
627static __inline__ unsigned char __attribute__((__always_inline__, __nodebug__))
628_bittestandcomplement64(__int64 *a, __int64 b) {
629  unsigned char x = (*a >> b) & 1;
630  *a = *a ^ (1ll << b);
631  return x;
632}
633static __inline__ unsigned char __attribute__((__always_inline__, __nodebug__))
634_bittestandreset64(__int64 *a, __int64 b) {
635  unsigned char x = (*a >> b) & 1;
636  *a = *a & ~(1ll << b);
637  return x;
638}
639static __inline__ unsigned char __attribute__((__always_inline__, __nodebug__))
640_bittestandset64(__int64 *a, __int64 b) {
641  unsigned char x = (*a >> b) & 1;
642  *a = *a | (1ll << b);
643  return x;
644}
645static __inline__ unsigned char __attribute__((__always_inline__, __nodebug__))
646_interlockedbittestandset64(__int64 volatile *__BitBase, __int64 __BitPos) {
647  unsigned char __Res;
648  __asm__ ("xor %0, %0\n"
649           "lock bts %2, %1\n"
650           "setc %0\n"
651           : "=r" (__Res), "+m"(*__BitBase)
652           : "Ir"(__BitPos));
653  return __Res;
654}
655#endif
656/*----------------------------------------------------------------------------*\
657|* Interlocked Exchange Add
658\*----------------------------------------------------------------------------*/
659static __inline__ char __attribute__((__always_inline__, __nodebug__))
660_InterlockedExchangeAdd8(char volatile *_Addend, char _Value) {
661  return __atomic_add_fetch(_Addend, _Value, 0) - _Value;
662}
663static __inline__ short __attribute__((__always_inline__, __nodebug__))
664_InterlockedExchangeAdd16(short volatile *_Addend, short _Value) {
665  return __atomic_add_fetch(_Addend, _Value, 0) - _Value;
666}
667#ifdef __x86_64__
668static __inline__ __int64 __attribute__((__always_inline__, __nodebug__))
669_InterlockedExchangeAdd64(__int64 volatile *_Addend, __int64 _Value) {
670  return __atomic_add_fetch(_Addend, _Value, 0) - _Value;
671}
672#endif
673/*----------------------------------------------------------------------------*\
674|* Interlocked Exchange Sub
675\*----------------------------------------------------------------------------*/
676static __inline__ char __attribute__((__always_inline__, __nodebug__))
677_InterlockedExchangeSub8(char volatile *_Subend, char _Value) {
678  return __atomic_sub_fetch(_Subend, _Value, 0) + _Value;
679}
680static __inline__ short __attribute__((__always_inline__, __nodebug__))
681_InterlockedExchangeSub16(short volatile *_Subend, short _Value) {
682  return __atomic_sub_fetch(_Subend, _Value, 0) + _Value;
683}
684static __inline__ long __attribute__((__always_inline__, __nodebug__))
685_InterlockedExchangeSub(long volatile *_Subend, long _Value) {
686  return __atomic_sub_fetch(_Subend, _Value, 0) + _Value;
687}
688#ifdef __x86_64__
689static __inline__ __int64 __attribute__((__always_inline__, __nodebug__))
690_InterlockedExchangeSub64(__int64 volatile *_Subend, __int64 _Value) {
691  return __atomic_sub_fetch(_Subend, _Value, 0) + _Value;
692}
693#endif
694/*----------------------------------------------------------------------------*\
695|* Interlocked Increment
696\*----------------------------------------------------------------------------*/
697static __inline__ short __attribute__((__always_inline__, __nodebug__))
698_InterlockedIncrement16(short volatile *_Value) {
699  return __atomic_add_fetch(_Value, 1, 0);
700}
701#ifdef __x86_64__
702static __inline__ __int64 __attribute__((__always_inline__, __nodebug__))
703_InterlockedIncrement64(__int64 volatile *_Value) {
704  return __atomic_add_fetch(_Value, 1, 0);
705}
706#endif
707/*----------------------------------------------------------------------------*\
708|* Interlocked Decrement
709\*----------------------------------------------------------------------------*/
710static __inline__ short __attribute__((__always_inline__, __nodebug__))
711_InterlockedDecrement16(short volatile *_Value) {
712  return __atomic_sub_fetch(_Value, 1, 0);
713}
714#ifdef __x86_64__
715static __inline__ __int64 __attribute__((__always_inline__, __nodebug__))
716_InterlockedDecrement64(__int64 volatile *_Value) {
717  return __atomic_sub_fetch(_Value, 1, 0);
718}
719#endif
720/*----------------------------------------------------------------------------*\
721|* Interlocked And
722\*----------------------------------------------------------------------------*/
723static __inline__ char __attribute__((__always_inline__, __nodebug__))
724_InterlockedAnd8(char volatile *_Value, char _Mask) {
725  return __atomic_and_fetch(_Value, _Mask, 0);
726}
727static __inline__ short __attribute__((__always_inline__, __nodebug__))
728_InterlockedAnd16(short volatile *_Value, short _Mask) {
729  return __atomic_and_fetch(_Value, _Mask, 0);
730}
731static __inline__ long __attribute__((__always_inline__, __nodebug__))
732_InterlockedAnd(long volatile *_Value, long _Mask) {
733  return __atomic_and_fetch(_Value, _Mask, 0);
734}
735#ifdef __x86_64__
736static __inline__ __int64 __attribute__((__always_inline__, __nodebug__))
737_InterlockedAnd64(__int64 volatile *_Value, __int64 _Mask) {
738  return __atomic_and_fetch(_Value, _Mask, 0);
739}
740#endif
741/*----------------------------------------------------------------------------*\
742|* Interlocked Or
743\*----------------------------------------------------------------------------*/
744static __inline__ char __attribute__((__always_inline__, __nodebug__))
745_InterlockedOr8(char volatile *_Value, char _Mask) {
746  return __atomic_or_fetch(_Value, _Mask, 0);
747}
748static __inline__ short __attribute__((__always_inline__, __nodebug__))
749_InterlockedOr16(short volatile *_Value, short _Mask) {
750  return __atomic_or_fetch(_Value, _Mask, 0);
751}
752static __inline__ long __attribute__((__always_inline__, __nodebug__))
753_InterlockedOr(long volatile *_Value, long _Mask) {
754  return __atomic_or_fetch(_Value, _Mask, 0);
755}
756#ifdef __x86_64__
757static __inline__ __int64 __attribute__((__always_inline__, __nodebug__))
758_InterlockedOr64(__int64 volatile *_Value, __int64 _Mask) {
759  return __atomic_or_fetch(_Value, _Mask, 0);
760}
761#endif
762/*----------------------------------------------------------------------------*\
763|* Interlocked Xor
764\*----------------------------------------------------------------------------*/
765static __inline__ char __attribute__((__always_inline__, __nodebug__))
766_InterlockedXor8(char volatile *_Value, char _Mask) {
767  return __atomic_xor_fetch(_Value, _Mask, 0);
768}
769static __inline__ short __attribute__((__always_inline__, __nodebug__))
770_InterlockedXor16(short volatile *_Value, short _Mask) {
771  return __atomic_xor_fetch(_Value, _Mask, 0);
772}
773static __inline__ long __attribute__((__always_inline__, __nodebug__))
774_InterlockedXor(long volatile *_Value, long _Mask) {
775  return __atomic_xor_fetch(_Value, _Mask, 0);
776}
777#ifdef __x86_64__
778static __inline__ __int64 __attribute__((__always_inline__, __nodebug__))
779_InterlockedXor64(__int64 volatile *_Value, __int64 _Mask) {
780  return __atomic_xor_fetch(_Value, _Mask, 0);
781}
782#endif
783/*----------------------------------------------------------------------------*\
784|* Interlocked Exchange
785\*----------------------------------------------------------------------------*/
786static __inline__ char __attribute__((__always_inline__, __nodebug__))
787_InterlockedExchange8(char volatile *_Target, char _Value) {
788  __atomic_exchange(_Target, &_Value, &_Value, 0);
789  return _Value;
790}
791static __inline__ short __attribute__((__always_inline__, __nodebug__))
792_InterlockedExchange16(short volatile *_Target, short _Value) {
793  __atomic_exchange(_Target, &_Value, &_Value, 0);
794  return _Value;
795}
796#ifdef __x86_64__
797static __inline__ __int64 __attribute__((__always_inline__, __nodebug__))
798_InterlockedExchange64(__int64 volatile *_Target, __int64 _Value) {
799  __atomic_exchange(_Target, &_Value, &_Value, 0);
800  return _Value;
801}
802#endif
803/*----------------------------------------------------------------------------*\
804|* Interlocked Compare Exchange
805\*----------------------------------------------------------------------------*/
806static __inline__ char __attribute__((__always_inline__, __nodebug__))
807_InterlockedCompareExchange8(char volatile *_Destination,
808                             char _Exchange, char _Comparand) {
809  __atomic_compare_exchange(_Destination, &_Comparand, &_Exchange, 0, 0, 0);
810  return _Comparand;
811}
812static __inline__ short __attribute__((__always_inline__, __nodebug__))
813_InterlockedCompareExchange16(short volatile *_Destination,
814                              short _Exchange, short _Comparand) {
815  __atomic_compare_exchange(_Destination, &_Comparand, &_Exchange, 0, 0, 0);
816  return _Comparand;
817}
818static __inline__ __int64 __attribute__((__always_inline__, __nodebug__))
819_InterlockedCompareExchange64(__int64 volatile *_Destination,
820                              __int64 _Exchange, __int64 _Comparand) {
821  __atomic_compare_exchange(_Destination, &_Comparand, &_Exchange, 0, 0, 0);
822  return _Comparand;
823}
824/*----------------------------------------------------------------------------*\
825|* Barriers
826\*----------------------------------------------------------------------------*/
827#if defined(__i386__) || defined(__x86_64__)
828static __inline__ void __attribute__((__always_inline__, __nodebug__))
829__attribute__((deprecated("use other intrinsics or C++11 atomics instead")))
830_ReadWriteBarrier(void) {
831  __asm__ volatile ("" : : : "memory");
832}
833static __inline__ void __attribute__((__always_inline__, __nodebug__))
834__attribute__((deprecated("use other intrinsics or C++11 atomics instead")))
835_ReadBarrier(void) {
836  __asm__ volatile ("" : : : "memory");
837}
838static __inline__ void __attribute__((__always_inline__, __nodebug__))
839__attribute__((deprecated("use other intrinsics or C++11 atomics instead")))
840_WriteBarrier(void) {
841  __asm__ volatile ("" : : : "memory");
842}
843#endif
844#ifdef __x86_64__
845static __inline__ void __attribute__((__always_inline__, __nodebug__))
846__faststorefence(void) {
847  __asm__ volatile("lock orq $0, (%%rsp)" : : : "memory");
848}
849#endif
850/*----------------------------------------------------------------------------*\
851|* readfs, readgs
852|* (Pointers in address space #256 and #257 are relative to the GS and FS
853|* segment registers, respectively.)
854\*----------------------------------------------------------------------------*/
855#define __ptr_to_addr_space(__addr_space_nbr, __type, __offset)              \
856    ((volatile __type __attribute__((__address_space__(__addr_space_nbr)))*) \
857    (__offset))
858
859#ifdef __i386__
860static __inline__ unsigned char __attribute__((__always_inline__, __nodebug__))
861__readfsbyte(unsigned long __offset) {
862  return *__ptr_to_addr_space(257, unsigned char, __offset);
863}
864static __inline__ unsigned long __attribute__((__always_inline__, __nodebug__))
865__readfsdword(unsigned long __offset) {
866  return *__ptr_to_addr_space(257, unsigned long, __offset);
867}
868static __inline__ unsigned __int64 __attribute__((__always_inline__, __nodebug__))
869__readfsqword(unsigned long __offset) {
870  return *__ptr_to_addr_space(257, unsigned __int64, __offset);
871}
872static __inline__ unsigned short __attribute__((__always_inline__, __nodebug__))
873__readfsword(unsigned long __offset) {
874  return *__ptr_to_addr_space(257, unsigned short, __offset);
875}
876#endif
877#ifdef __x86_64__
878static __inline__ unsigned char __attribute__((__always_inline__, __nodebug__))
879__readgsbyte(unsigned long __offset) {
880  return *__ptr_to_addr_space(256, unsigned char, __offset);
881}
882static __inline__ unsigned long __attribute__((__always_inline__, __nodebug__))
883__readgsdword(unsigned long __offset) {
884  return *__ptr_to_addr_space(256, unsigned long, __offset);
885}
886static __inline__ unsigned __int64 __attribute__((__always_inline__, __nodebug__))
887__readgsqword(unsigned long __offset) {
888  return *__ptr_to_addr_space(256, unsigned __int64, __offset);
889}
890static __inline__ unsigned short __attribute__((__always_inline__, __nodebug__))
891__readgsword(unsigned long __offset) {
892  return *__ptr_to_addr_space(256, unsigned short, __offset);
893}
894#endif
895#undef __ptr_to_addr_space
896/*----------------------------------------------------------------------------*\
897|* movs, stos
898\*----------------------------------------------------------------------------*/
899#if defined(__i386__) || defined(__x86_64__)
900static __inline__ void __attribute__((__always_inline__, __nodebug__))
901__movsb(unsigned char *__dst, unsigned char const *__src, size_t __n) {
902  __asm__("rep movsb" : : "D"(__dst), "S"(__src), "c"(__n)
903                        : "%edi", "%esi", "%ecx");
904}
905static __inline__ void __attribute__((__always_inline__, __nodebug__))
906__movsd(unsigned long *__dst, unsigned long const *__src, size_t __n) {
907  __asm__("rep movsl" : : "D"(__dst), "S"(__src), "c"(__n)
908                        : "%edi", "%esi", "%ecx");
909}
910static __inline__ void __attribute__((__always_inline__, __nodebug__))
911__movsw(unsigned short *__dst, unsigned short const *__src, size_t __n) {
912  __asm__("rep movsh" : : "D"(__dst), "S"(__src), "c"(__n)
913                        : "%edi", "%esi", "%ecx");
914}
915static __inline__ void __attribute__((__always_inline__, __nodebug__))
916__stosb(unsigned char *__dst, unsigned char __x, size_t __n) {
917  __asm__("rep stosb" : : "D"(__dst), "a"(__x), "c"(__n)
918                        : "%edi", "%ecx");
919}
920static __inline__ void __attribute__((__always_inline__, __nodebug__))
921__stosd(unsigned long *__dst, unsigned long __x, size_t __n) {
922  __asm__("rep stosl" : : "D"(__dst), "a"(__x), "c"(__n)
923                        : "%edi", "%ecx");
924}
925static __inline__ void __attribute__((__always_inline__, __nodebug__))
926__stosw(unsigned short *__dst, unsigned short __x, size_t __n) {
927  __asm__("rep stosh" : : "D"(__dst), "a"(__x), "c"(__n)
928                        : "%edi", "%ecx");
929}
930#endif
931#ifdef __x86_64__
932static __inline__ void __attribute__((__always_inline__, __nodebug__))
933__movsq(unsigned long long *__dst, unsigned long long const *__src, size_t __n) {
934  __asm__("rep movsq" : : "D"(__dst), "S"(__src), "c"(__n)
935                        : "%edi", "%esi", "%ecx");
936}
937static __inline__ void __attribute__((__always_inline__, __nodebug__))
938__stosq(unsigned __int64 *__dst, unsigned __int64 __x, size_t __n) {
939  __asm__("rep stosq" : : "D"(__dst), "a"(__x), "c"(__n)
940                        : "%edi", "%ecx");
941}
942#endif
943
944/*----------------------------------------------------------------------------*\
945|* Misc
946\*----------------------------------------------------------------------------*/
947static __inline__ void * __attribute__((__always_inline__, __nodebug__))
948_AddressOfReturnAddress(void) {
949  return (void*)((char*)__builtin_frame_address(0) + sizeof(void*));
950}
951static __inline__ void * __attribute__((__always_inline__, __nodebug__))
952_ReturnAddress(void) {
953  return __builtin_return_address(0);
954}
955#if defined(__i386__) || defined(__x86_64__)
956static __inline__ void __attribute__((__always_inline__, __nodebug__))
957__cpuid(int __info[4], int __level) {
958  __asm__ ("cpuid" : "=a"(__info[0]), "=b" (__info[1]), "=c"(__info[2]), "=d"(__info[3])
959                   : "a"(__level));
960}
961static __inline__ void __attribute__((__always_inline__, __nodebug__))
962__cpuidex(int __info[4], int __level, int __ecx) {
963  __asm__ ("cpuid" : "=a"(__info[0]), "=b" (__info[1]), "=c"(__info[2]), "=d"(__info[3])
964                   : "a"(__level), "c"(__ecx));
965}
966static __inline__ unsigned __int64 __cdecl __attribute__((__always_inline__, __nodebug__))
967_xgetbv(unsigned int __xcr_no) {
968  unsigned int __eax, __edx;
969  __asm__ ("xgetbv" : "=a" (__eax), "=d" (__edx) : "c" (__xcr_no));
970  return ((unsigned __int64)__edx << 32) | __eax;
971}
972static __inline__ void __attribute__((__always_inline__, __nodebug__))
973__halt(void) {
974  __asm__ volatile ("hlt");
975}
976#endif
977
978/*----------------------------------------------------------------------------*\
979|* Privileged intrinsics
980\*----------------------------------------------------------------------------*/
981#if defined(__i386__) || defined(__x86_64__)
982static __inline__ unsigned __int64 __attribute__((__always_inline__, __nodebug__))
983__readmsr(unsigned long __register) {
984  // Loads the contents of a 64-bit model specific register (MSR) specified in
985  // the ECX register into registers EDX:EAX. The EDX register is loaded with
986  // the high-order 32 bits of the MSR and the EAX register is loaded with the
987  // low-order 32 bits. If less than 64 bits are implemented in the MSR being
988  // read, the values returned to EDX:EAX in unimplemented bit locations are
989  // undefined.
990  unsigned long __edx;
991  unsigned long __eax;
992  __asm__ ("rdmsr" : "=d"(__edx), "=a"(__eax) : "c"(__register));
993  return (((unsigned __int64)__edx) << 32) | (unsigned __int64)__eax;
994}
995
996static __inline__ unsigned long __attribute__((always_inline, __nodebug__))
997__readcr3(void) {
998  unsigned long __cr3_val;
999  __asm__ __volatile__ ("mov %%cr3, %0" : "=q"(__cr3_val) : : "memory");
1000  return __cr3_val;
1001}
1002
1003static __inline__ void __attribute__((always_inline, __nodebug__))
1004__writecr3(unsigned int __cr3_val) {
1005  __asm__ ("mov %0, %%cr3" : : "q"(__cr3_val) : "memory");
1006}
1007#endif
1008
1009#ifdef __cplusplus
1010}
1011#endif
1012
1013#endif /* __INTRIN_H */
1014#endif /* _MSC_VER */
1015