Crypto++ 8.7
Free C++ class library of cryptographic schemes
ecp.cpp
1// ecp.cpp - originally written and placed in the public domain by Wei Dai
2
3#include "pch.h"
4
5#ifndef CRYPTOPP_IMPORTS
6
7#include "ecp.h"
8#include "asn.h"
9#include "integer.h"
10#include "nbtheory.h"
11#include "modarith.h"
12#include "filters.h"
13#include "algebra.cpp"
14
15ANONYMOUS_NAMESPACE_BEGIN
16
17using CryptoPP::ECP;
18using CryptoPP::Integer;
19using CryptoPP::ModularArithmetic;
20
21#if defined(HAVE_GCC_INIT_PRIORITY)
22 #define INIT_ATTRIBUTE __attribute__ ((init_priority (CRYPTOPP_INIT_PRIORITY + 50)))
23 const ECP::Point g_identity INIT_ATTRIBUTE = ECP::Point();
24#elif defined(HAVE_MSC_INIT_PRIORITY)
25 #pragma warning(disable: 4075)
26 #pragma init_seg(".CRT$XCU")
27 const ECP::Point g_identity;
28 #pragma warning(default: 4075)
29#elif defined(HAVE_XLC_INIT_PRIORITY)
30 #pragma priority(290)
31 const ECP::Point g_identity;
32#endif
33
34inline ECP::Point ToMontgomery(const ModularArithmetic &mr, const ECP::Point &P)
35{
36 return P.identity ? P : ECP::Point(mr.ConvertIn(P.x), mr.ConvertIn(P.y));
37}
38
39inline ECP::Point FromMontgomery(const ModularArithmetic &mr, const ECP::Point &P)
40{
41 return P.identity ? P : ECP::Point(mr.ConvertOut(P.x), mr.ConvertOut(P.y));
42}
43
44inline Integer IdentityToInteger(bool val)
45{
46 return val ? Integer::One() : Integer::Zero();
47}
48
49struct ProjectivePoint
50{
51 ProjectivePoint() {}
52 ProjectivePoint(const Integer &x, const Integer &y, const Integer &z)
53 : x(x), y(y), z(z) {}
54
55 Integer x, y, z;
56};
57
58ANONYMOUS_NAMESPACE_END
59
60NAMESPACE_BEGIN(CryptoPP)
61
62ECP::ECP(const ECP &ecp, bool convertToMontgomeryRepresentation)
63{
64 if (convertToMontgomeryRepresentation && !ecp.GetField().IsMontgomeryRepresentation())
65 {
66 m_fieldPtr.reset(new MontgomeryRepresentation(ecp.GetField().GetModulus()));
67 m_a = GetField().ConvertIn(ecp.m_a);
68 m_b = GetField().ConvertIn(ecp.m_b);
69 }
70 else
71 operator=(ecp);
72}
73
75 : m_fieldPtr(new Field(bt))
76{
77 BERSequenceDecoder seq(bt);
78 GetField().BERDecodeElement(seq, m_a);
79 GetField().BERDecodeElement(seq, m_b);
80 // skip optional seed
81 if (!seq.EndReached())
82 {
83 SecByteBlock seed;
84 unsigned int unused;
85 BERDecodeBitString(seq, seed, unused);
86 }
87 seq.MessageEnd();
88}
89
91{
92 GetField().DEREncode(bt);
93 DERSequenceEncoder seq(bt);
94 GetField().DEREncodeElement(seq, m_a);
95 GetField().DEREncodeElement(seq, m_b);
96 seq.MessageEnd();
97}
98
99bool ECP::DecodePoint(ECP::Point &P, const byte *encodedPoint, size_t encodedPointLen) const
100{
101 StringStore store(encodedPoint, encodedPointLen);
102 return DecodePoint(P, store, encodedPointLen);
103}
104
105bool ECP::DecodePoint(ECP::Point &P, BufferedTransformation &bt, size_t encodedPointLen) const
106{
107 byte type;
108 if (encodedPointLen < 1 || !bt.Get(type))
109 return false;
110
111 switch (type)
112 {
113 case 0:
114 P.identity = true;
115 return true;
116 case 2:
117 case 3:
118 {
119 if (encodedPointLen != EncodedPointSize(true))
120 return false;
121
122 Integer p = FieldSize();
123
124 P.identity = false;
125 P.x.Decode(bt, GetField().MaxElementByteLength());
126 P.y = ((P.x*P.x+m_a)*P.x+m_b) % p;
127
128 if (Jacobi(P.y, p) !=1)
129 return false;
130
131 P.y = ModularSquareRoot(P.y, p);
132
133 if ((type & 1) != P.y.GetBit(0))
134 P.y = p-P.y;
135
136 return true;
137 }
138 case 4:
139 {
140 if (encodedPointLen != EncodedPointSize(false))
141 return false;
142
143 unsigned int len = GetField().MaxElementByteLength();
144 P.identity = false;
145 P.x.Decode(bt, len);
146 P.y.Decode(bt, len);
147 return true;
148 }
149 default:
150 return false;
151 }
152}
153
154void ECP::EncodePoint(BufferedTransformation &bt, const Point &P, bool compressed) const
155{
156 if (P.identity)
157 NullStore().TransferTo(bt, EncodedPointSize(compressed));
158 else if (compressed)
159 {
160 bt.Put((byte)(2U + P.y.GetBit(0)));
161 P.x.Encode(bt, GetField().MaxElementByteLength());
162 }
163 else
164 {
165 unsigned int len = GetField().MaxElementByteLength();
166 bt.Put(4U); // uncompressed
167 P.x.Encode(bt, len);
168 P.y.Encode(bt, len);
169 }
170}
171
172void ECP::EncodePoint(byte *encodedPoint, const Point &P, bool compressed) const
173{
174 ArraySink sink(encodedPoint, EncodedPointSize(compressed));
175 EncodePoint(sink, P, compressed);
176 CRYPTOPP_ASSERT(sink.TotalPutLength() == EncodedPointSize(compressed));
177}
178
180{
181 SecByteBlock str;
182 BERDecodeOctetString(bt, str);
183 Point P;
184 if (!DecodePoint(P, str, str.size()))
186 return P;
187}
188
189void ECP::DEREncodePoint(BufferedTransformation &bt, const Point &P, bool compressed) const
190{
191 SecByteBlock str(EncodedPointSize(compressed));
192 EncodePoint(str, P, compressed);
193 DEREncodeOctetString(bt, str);
194}
195
196bool ECP::ValidateParameters(RandomNumberGenerator &rng, unsigned int level) const
197{
198 Integer p = FieldSize();
199
200 bool pass = p.IsOdd();
201 pass = pass && !m_a.IsNegative() && m_a<p && !m_b.IsNegative() && m_b<p;
202
203 if (level >= 1)
204 pass = pass && ((4*m_a*m_a*m_a+27*m_b*m_b)%p).IsPositive();
205
206 if (level >= 2)
207 pass = pass && VerifyPrime(rng, p);
208
209 return pass;
210}
211
212bool ECP::VerifyPoint(const Point &P) const
213{
214 const FieldElement &x = P.x, &y = P.y;
215 Integer p = FieldSize();
216 return P.identity ||
217 (!x.IsNegative() && x<p && !y.IsNegative() && y<p
218 && !(((x*x+m_a)*x+m_b-y*y)%p));
219}
220
221bool ECP::Equal(const Point &P, const Point &Q) const
222{
223 if (P.identity && Q.identity)
224 return true;
225
226 if (P.identity && !Q.identity)
227 return false;
228
229 if (!P.identity && Q.identity)
230 return false;
231
232 return (GetField().Equal(P.x,Q.x) && GetField().Equal(P.y,Q.y));
233}
234
235const ECP::Point& ECP::Identity() const
236{
237#if defined(HAVE_GCC_INIT_PRIORITY) || defined(HAVE_MSC_INIT_PRIORITY) || defined(HAVE_XLC_INIT_PRIORITY)
238 return g_identity;
239#elif defined(CRYPTOPP_CXX11_STATIC_INIT)
240 static const ECP::Point g_identity;
241 return g_identity;
242#else
243 return Singleton<Point>().Ref();
244#endif
245}
246
247const ECP::Point& ECP::Inverse(const Point &P) const
248{
249 if (P.identity)
250 return P;
251 else
252 {
253 m_R.identity = false;
254 m_R.x = P.x;
255 m_R.y = GetField().Inverse(P.y);
256 return m_R;
257 }
258}
259
260const ECP::Point& ECP::Add(const Point &P, const Point &Q) const
261{
262 if (P.identity) return Q;
263 if (Q.identity) return P;
264 if (GetField().Equal(P.x, Q.x))
265 return GetField().Equal(P.y, Q.y) ? Double(P) : Identity();
266
267 FieldElement t = GetField().Subtract(Q.y, P.y);
268 t = GetField().Divide(t, GetField().Subtract(Q.x, P.x));
269 FieldElement x = GetField().Subtract(GetField().Subtract(GetField().Square(t), P.x), Q.x);
270 m_R.y = GetField().Subtract(GetField().Multiply(t, GetField().Subtract(P.x, x)), P.y);
271
272 m_R.x.swap(x);
273 m_R.identity = false;
274 return m_R;
275}
276
277const ECP::Point& ECP::Double(const Point &P) const
278{
279 if (P.identity || P.y==GetField().Identity()) return Identity();
280
281 FieldElement t = GetField().Square(P.x);
282 t = GetField().Add(GetField().Add(GetField().Double(t), t), m_a);
283 t = GetField().Divide(t, GetField().Double(P.y));
284 FieldElement x = GetField().Subtract(GetField().Subtract(GetField().Square(t), P.x), P.x);
285 m_R.y = GetField().Subtract(GetField().Multiply(t, GetField().Subtract(P.x, x)), P.y);
286
287 m_R.x.swap(x);
288 m_R.identity = false;
289 return m_R;
290}
291
292template <class T, class Iterator> void ParallelInvert(const AbstractRing<T> &ring, Iterator begin, Iterator end)
293{
294 size_t n = end-begin;
295 if (n == 1)
296 *begin = ring.MultiplicativeInverse(*begin);
297 else if (n > 1)
298 {
299 std::vector<T> vec((n+1)/2);
300 unsigned int i;
301 Iterator it;
302
303 for (i=0, it=begin; i<n/2; i++, it+=2)
304 vec[i] = ring.Multiply(*it, *(it+1));
305 if (n%2 == 1)
306 vec[n/2] = *it;
307
308 ParallelInvert(ring, vec.begin(), vec.end());
309
310 for (i=0, it=begin; i<n/2; i++, it+=2)
311 {
312 if (!vec[i])
313 {
314 *it = ring.MultiplicativeInverse(*it);
315 *(it+1) = ring.MultiplicativeInverse(*(it+1));
316 }
317 else
318 {
319 std::swap(*it, *(it+1));
320 *it = ring.Multiply(*it, vec[i]);
321 *(it+1) = ring.Multiply(*(it+1), vec[i]);
322 }
323 }
324 if (n%2 == 1)
325 *it = vec[n/2];
326 }
327}
328
329class ProjectiveDoubling
330{
331public:
332 ProjectiveDoubling(const ModularArithmetic &m_mr, const Integer &m_a, const Integer &m_b, const ECPPoint &Q)
333 : mr(m_mr)
334 {
335 CRYPTOPP_UNUSED(m_b);
336 if (Q.identity)
337 {
338 sixteenY4 = P.x = P.y = mr.MultiplicativeIdentity();
339 aZ4 = P.z = mr.Identity();
340 }
341 else
342 {
343 P.x = Q.x;
344 P.y = Q.y;
345 sixteenY4 = P.z = mr.MultiplicativeIdentity();
346 aZ4 = m_a;
347 }
348 }
349
350 void Double()
351 {
352 twoY = mr.Double(P.y);
353 P.z = mr.Multiply(P.z, twoY);
354 fourY2 = mr.Square(twoY);
355 S = mr.Multiply(fourY2, P.x);
356 aZ4 = mr.Multiply(aZ4, sixteenY4);
357 M = mr.Square(P.x);
358 M = mr.Add(mr.Add(mr.Double(M), M), aZ4);
359 P.x = mr.Square(M);
360 mr.Reduce(P.x, S);
361 mr.Reduce(P.x, S);
362 mr.Reduce(S, P.x);
363 P.y = mr.Multiply(M, S);
364 sixteenY4 = mr.Square(fourY2);
365 mr.Reduce(P.y, mr.Half(sixteenY4));
366 }
367
368 const ModularArithmetic &mr;
369 ProjectivePoint P;
370 Integer sixteenY4, aZ4, twoY, fourY2, S, M;
371};
372
373struct ZIterator
374{
375 ZIterator() {}
376 ZIterator(std::vector<ProjectivePoint>::iterator it) : it(it) {}
377 Integer& operator*() {return it->z;}
378 int operator-(ZIterator it2) {return int(it-it2.it);}
379 ZIterator operator+(int i) {return ZIterator(it+i);}
380 ZIterator& operator+=(int i) {it+=i; return *this;}
381 std::vector<ProjectivePoint>::iterator it;
382};
383
384ECP::Point ECP::ScalarMultiply(const Point &P, const Integer &k) const
385{
386 Element result;
387 if (k.BitCount() <= 5)
389 else
390 ECP::SimultaneousMultiply(&result, P, &k, 1);
391 return result;
392}
393
394void ECP::SimultaneousMultiply(ECP::Point *results, const ECP::Point &P, const Integer *expBegin, unsigned int expCount) const
395{
396 if (!GetField().IsMontgomeryRepresentation())
397 {
398 ECP ecpmr(*this, true);
399 const ModularArithmetic &mr = ecpmr.GetField();
400 ecpmr.SimultaneousMultiply(results, ToMontgomery(mr, P), expBegin, expCount);
401 for (unsigned int i=0; i<expCount; i++)
402 results[i] = FromMontgomery(mr, results[i]);
403 return;
404 }
405
406 ProjectiveDoubling rd(GetField(), m_a, m_b, P);
407 std::vector<ProjectivePoint> bases;
408 std::vector<WindowSlider> exponents;
409 exponents.reserve(expCount);
410 std::vector<std::vector<word32> > baseIndices(expCount);
411 std::vector<std::vector<bool> > negateBase(expCount);
412 std::vector<std::vector<word32> > exponentWindows(expCount);
413 unsigned int i;
414
415 for (i=0; i<expCount; i++)
416 {
417 CRYPTOPP_ASSERT(expBegin->NotNegative());
418 exponents.push_back(WindowSlider(*expBegin++, InversionIsFast(), 5));
419 exponents[i].FindNextWindow();
420 }
421
422 unsigned int expBitPosition = 0;
423 bool notDone = true;
424
425 while (notDone)
426 {
427 notDone = false;
428 bool baseAdded = false;
429 for (i=0; i<expCount; i++)
430 {
431 if (!exponents[i].finished && expBitPosition == exponents[i].windowBegin)
432 {
433 if (!baseAdded)
434 {
435 bases.push_back(rd.P);
436 baseAdded =true;
437 }
438
439 exponentWindows[i].push_back(exponents[i].expWindow);
440 baseIndices[i].push_back((word32)bases.size()-1);
441 negateBase[i].push_back(exponents[i].negateNext);
442
443 exponents[i].FindNextWindow();
444 }
445 notDone = notDone || !exponents[i].finished;
446 }
447
448 if (notDone)
449 {
450 rd.Double();
451 expBitPosition++;
452 }
453 }
454
455 // convert from projective to affine coordinates
456 ParallelInvert(GetField(), ZIterator(bases.begin()), ZIterator(bases.end()));
457 for (i=0; i<bases.size(); i++)
458 {
459 if (bases[i].z.NotZero())
460 {
461 bases[i].y = GetField().Multiply(bases[i].y, bases[i].z);
462 bases[i].z = GetField().Square(bases[i].z);
463 bases[i].x = GetField().Multiply(bases[i].x, bases[i].z);
464 bases[i].y = GetField().Multiply(bases[i].y, bases[i].z);
465 }
466 }
467
468 std::vector<BaseAndExponent<Point, Integer> > finalCascade;
469 for (i=0; i<expCount; i++)
470 {
471 finalCascade.resize(baseIndices[i].size());
472 for (unsigned int j=0; j<baseIndices[i].size(); j++)
473 {
474 ProjectivePoint &base = bases[baseIndices[i][j]];
475 if (base.z.IsZero())
476 finalCascade[j].base.identity = true;
477 else
478 {
479 finalCascade[j].base.identity = false;
480 finalCascade[j].base.x = base.x;
481 if (negateBase[i][j])
482 finalCascade[j].base.y = GetField().Inverse(base.y);
483 else
484 finalCascade[j].base.y = base.y;
485 }
486 finalCascade[j].exponent = Integer(Integer::POSITIVE, 0, exponentWindows[i][j]);
487 }
488 results[i] = GeneralCascadeMultiplication(*this, finalCascade.begin(), finalCascade.end());
489 }
490}
491
492ECP::Point ECP::CascadeScalarMultiply(const Point &P, const Integer &k1, const Point &Q, const Integer &k2) const
493{
494 if (!GetField().IsMontgomeryRepresentation())
495 {
496 ECP ecpmr(*this, true);
497 const ModularArithmetic &mr = ecpmr.GetField();
498 return FromMontgomery(mr, ecpmr.CascadeScalarMultiply(ToMontgomery(mr, P), k1, ToMontgomery(mr, Q), k2));
499 }
500 else
502}
503
504NAMESPACE_END
505
506#endif
Classes and functions for working with ANS.1 objects.
CRYPTOPP_DLL size_t BERDecodeBitString(BufferedTransformation &bt, SecByteBlock &str, unsigned int &unusedBits)
DER decode bit string.
OID operator+(const OID &lhs, unsigned long rhs)
Append a value to an OID.
CRYPTOPP_DLL size_t DEREncodeOctetString(BufferedTransformation &bt, const byte *str, size_t strLen)
DER encode octet string.
CRYPTOPP_DLL size_t BERDecodeOctetString(BufferedTransformation &bt, SecByteBlock &str)
BER decode octet string.
void BERDecodeError()
Raises a BERDecodeErr.
Definition: asn.h:104
virtual Element CascadeScalarMultiply(const Element &x, const Integer &e1, const Element &y, const Integer &e2) const
TODO.
Definition: algebra.cpp:97
virtual void SimultaneousMultiply(Element *results, const Element &base, const Integer *exponents, unsigned int exponentsCount) const
Multiplies a base to multiple exponents in a group.
Definition: algebra.cpp:256
virtual const Element & Subtract(const Element &a, const Element &b) const
Subtracts elements in the group.
Definition: algebra.cpp:20
Abstract ring.
Definition: algebra.h:119
virtual const Element & Multiply(const Element &a, const Element &b) const =0
Multiplies elements in the group.
virtual const Element & MultiplicativeInverse(const Element &a) const =0
Calculate the multiplicative inverse of an element in the group.
Copy input to a memory buffer.
Definition: filters.h:1200
BER Sequence Decoder.
Definition: asn.h:525
Interface for buffered transformations.
Definition: cryptlib.h:1652
virtual size_t Get(byte &outByte)
Retrieve a 8-bit byte.
lword TransferTo(BufferedTransformation &target, lword transferMax=LWORD_MAX, const std::string &channel=DEFAULT_CHANNEL)
move transferMax bytes of the buffered output to target as input
Definition: cryptlib.h:1991
size_t Put(byte inByte, bool blocking=true)
Input a byte for processing.
Definition: cryptlib.h:1673
DER Sequence Encoder.
Definition: asn.h:557
Elliptic Curve over GF(p), where p is prime.
Definition: ecp.h:27
bool InversionIsFast() const
Determine if inversion is fast.
Definition: ecp.h:75
const Point & Double(const Point &P) const
Doubles an element in the group.
Point ScalarMultiply(const Point &P, const Integer &k) const
Performs a scalar multiplication.
void EncodePoint(byte *encodedPoint, const Point &P, bool compressed) const
Encodes an elliptic curve point.
ECP()
Construct an ECP.
Definition: ecp.h:36
bool Equal(const Point &P, const Point &Q) const
Compare two points.
void DEREncodePoint(BufferedTransformation &bt, const Point &P, bool compressed) const
DER Encodes an elliptic curve point.
bool VerifyPoint(const Point &P) const
Verifies points on elliptic curve.
Point CascadeScalarMultiply(const Point &P, const Integer &k1, const Point &Q, const Integer &k2) const
TODO.
const Point & Inverse(const Point &P) const
Inverts the element in the group.
const Point & Identity() const
Provides the Identity element.
Point BERDecodePoint(BufferedTransformation &bt) const
BER Decodes an elliptic curve point.
unsigned int EncodedPointSize(bool compressed=false) const
Determines encoded point size.
Definition: ecp.h:90
bool DecodePoint(Point &P, BufferedTransformation &bt, size_t len) const
Decodes an elliptic curve point.
void DEREncode(BufferedTransformation &bt) const
DER Encode.
const Point & Add(const Point &P, const Point &Q) const
Adds elements in the group.
void SimultaneousMultiply(Point *results, const Point &base, const Integer *exponents, unsigned int exponentsCount) const
Multiplies a base to multiple exponents in a group.
Multiple precision integer with arithmetic operations.
Definition: integer.h:50
unsigned int BitCount() const
Determines the number of bits required to represent the Integer.
bool NotNegative() const
Determines if the Integer is non-negative.
Definition: integer.h:344
void swap(Integer &a)
Swaps this Integer with another Integer.
bool IsNegative() const
Determines if the Integer is negative.
Definition: integer.h:341
@ POSITIVE
the value is positive or 0
Definition: integer.h:75
bool IsOdd() const
Determines if the Integer is odd parity.
Definition: integer.h:356
static const Integer & One()
Integer representing 1.
Ring of congruence classes modulo n.
Definition: modarith.h:44
const Integer & Square(const Integer &a) const
Square an element in the ring.
Definition: modarith.h:197
const Integer & Inverse(const Integer &a) const
Inverts the element in the ring.
unsigned int MaxElementByteLength() const
Provides the maximum byte size of an element in the ring.
Definition: modarith.h:248
void DEREncodeElement(BufferedTransformation &out, const Element &a) const
Encodes element in DER format.
virtual bool IsMontgomeryRepresentation() const
Retrieves the representation.
Definition: modarith.h:108
bool Equal(const Integer &a, const Integer &b) const
Compare two elements for equality.
Definition: modarith.h:135
const Integer & GetModulus() const
Retrieves the modulus.
Definition: modarith.h:99
const Integer & Multiply(const Integer &a, const Integer &b) const
Multiplies elements in the ring.
Definition: modarith.h:190
virtual Integer ConvertOut(const Integer &a) const
Reduces an element in the congruence class.
Definition: modarith.h:123
const Integer & Subtract(const Integer &a, const Integer &b) const
Subtracts elements in the ring.
const Integer & Divide(const Integer &a, const Integer &b) const
Divides elements in the ring.
Definition: modarith.h:218
const Integer & Add(const Integer &a, const Integer &b) const
Adds elements in the ring.
virtual Integer ConvertIn(const Integer &a) const
Reduces an element in the congruence class.
Definition: modarith.h:115
void DEREncode(BufferedTransformation &bt) const
Encodes in DER format.
Performs modular arithmetic in Montgomery representation for increased speed.
Definition: modarith.h:296
Empty store.
Definition: filters.h:1321
Interface for random number generators.
Definition: cryptlib.h:1435
size_type size() const
Provides the count of elements in the SecBlock.
Definition: secblock.h:867
SecBlock<byte> typedef.
Definition: secblock.h:1226
Restricts the instantiation of a class to one static object without locks.
Definition: misc.h:307
const T & Ref(...) const
Return a reference to the inner Singleton object.
Definition: misc.h:327
Square block cipher.
Definition: square.h:25
String-based implementation of Store interface.
Definition: filters.h:1259
unsigned int word32
32-bit unsigned datatype
Definition: config_int.h:62
Classes for Elliptic Curves over prime fields.
Implementation of BufferedTransformation's attachment interface.
Multiple precision integer with arithmetic operations.
inline ::Integer operator-(const ::Integer &a, const ::Integer &b)
Subtraction.
Definition: integer.h:771
inline ::Integer operator*(const ::Integer &a, const ::Integer &b)
Multiplication.
Definition: integer.h:774
Class file for performing modular arithmetic.
Crypto++ library namespace.
Classes and functions for number theoretic operations.
CRYPTOPP_DLL int Jacobi(const Integer &a, const Integer &b)
Calculate the Jacobi symbol.
CRYPTOPP_DLL Integer ModularSquareRoot(const Integer &a, const Integer &p)
Extract a modular square root.
CRYPTOPP_DLL bool VerifyPrime(RandomNumberGenerator &rng, const Integer &p, unsigned int level=1)
Verifies a number is probably prime.
Precompiled header file.
void swap(::SecBlock< T, A > &a, ::SecBlock< T, A > &b)
Swap two SecBlocks.
Definition: secblock.h:1289
Elliptical Curve Point over GF(p), where p is prime.
Definition: ecpoint.h:21
#define CRYPTOPP_ASSERT(exp)
Debugging and diagnostic assertion.
Definition: trap.h:68