field.mx raw

   1  // Copyright (c) 2013-2014 The btcsuite developers
   2  // Copyright (c) 2015-2023 The Decred developers
   3  // Copyright (c) 2013-2023 Dave Collins
   4  // Use of this source code is governed by an ISC
   5  // license that can be found in the LICENSE file.
   6  
   7  package secp256k1
   8  
   9  // References:
  10  //   [HAC]: Handbook of Applied Cryptography Menezes, van Oorschot, Vanstone.
  11  //     http://cacr.uwaterloo.ca/hac/
  12  
  13  // All elliptic curve operations for secp256k1 are done in a finite field
  14  // characterized by a 256-bit prime.  Given this precision is larger than the
  15  // biggest available native type, obviously some form of bignum math is needed.
  16  // This package implements specialized fixed-precision field arithmetic rather
  17  // than relying on an arbitrary-precision arithmetic package such as math/big
  18  // for dealing with the field math since the size is known.  As a result, rather
  19  // large performance gains are achieved by taking advantage of many
  20  // optimizations not available to arbitrary-precision arithmetic and generic
  21  // modular arithmetic algorithms.
  22  //
  23  // There are various ways to internally represent each finite field element.
  24  // For example, the most obvious representation would be to use an array of 4
  25  // uint64s (64 bits * 4 = 256 bits).  However, that representation suffers from
  26  // a couple of issues.  First, there is no native Go type large enough to handle
  27  // the intermediate results while adding or multiplying two 64-bit numbers, and
  28  // second there is no space left for overflows when performing the intermediate
  29  // arithmetic between each array element which would lead to expensive carry
  30  // propagation.
  31  //
  32  // Given the above, this implementation represents the field elements as
  33  // 10 uint32s with each word (array entry) treated as base 2^26.  This was
  34  // chosen for the following reasons:
  35  // 1) Most systems at the current time are 64-bit (or at least have 64-bit
  36  //    registers available for specialized purposes such as MMX) so the
  37  //    intermediate results can typically be done using a native register (and
  38  //    using uint64s to avoid the need for additional half-word arithmetic)
  39  // 2) In order to allow addition of the internal words without having to
  40  //    propagate the carry, the max normalized value for each register must
  41  //    be less than the number of bits available in the register
  42  // 3) Since we're dealing with 32-bit values, 64-bits of overflow is a
  43  //    reasonable choice for #2
  44  // 4) Given the need for 256-bits of precision and the properties stated in #1,
  45  //    #2, and #3, the representation which best accommodates this is 10 uint32s
  46  //    with base 2^26 (26 bits * 10 = 260 bits, so the final word only needs 22
  47  //    bits) which leaves the desired 64 bits (32 * 10 = 320, 320 - 256 = 64) for
  48  //    overflow
  49  //
  50  // Since it is so important that the field arithmetic is extremely fast for high
  51  // performance crypto, this type does not perform any validation where it
  52  // ordinarily would.  See the documentation for FieldVal for more details.
  53  
  54  import (
  55  	"smesh.lol/pkg/nostr/hex"
  56  )
  57  
  58  // Constants used to make the code more readable.
  59  const (
  60  	twoBitsMask   = 0x3
  61  	fourBitsMask  = 0xf
  62  	sixBitsMask   = 0x3f
  63  	eightBitsMask = 0xff
  64  )
  65  
  66  // Constants related to the field representation.
  67  const (
  68  	// fieldWords is the number of words used to internally represent the
  69  	// 256-bit value.
  70  	fieldWords = 10
  71  	// fieldBase is the exponent used to form the numeric base of each word.
  72  	// 2^(fieldBase*i) where i is the word position.
  73  	fieldBase = 26
  74  	// fieldBaseMask is the mask for the bits in each word needed to
  75  	// represent the numeric base of each word (except the most significant
  76  	// word).
  77  	fieldBaseMask = (1 << fieldBase) - 1
  78  	// fieldMSBBits is the number of bits in the most significant word used
  79  	// to represent the value.
  80  	fieldMSBBits = 256 - (fieldBase * (fieldWords - 1))
  81  	// fieldMSBMask is the mask for the bits in the most significant word
  82  	// needed to represent the value.
  83  	fieldMSBMask = (1 << fieldMSBBits) - 1
  84  	// These fields provide convenient access to each of the words of the
  85  	// secp256k1 prime in the internal field representation to improve code
  86  	// readability.
  87  	fieldPrimeWordZero  = 0x03fffc2f
  88  	fieldPrimeWordOne   = 0x03ffffbf
  89  	fieldPrimeWordTwo   = 0x03ffffff
  90  	fieldPrimeWordThree = 0x03ffffff
  91  	fieldPrimeWordFour  = 0x03ffffff
  92  	fieldPrimeWordFive  = 0x03ffffff
  93  	fieldPrimeWordSix   = 0x03ffffff
  94  	fieldPrimeWordSeven = 0x03ffffff
  95  	fieldPrimeWordEight = 0x03ffffff
  96  	fieldPrimeWordNine  = 0x003fffff
  97  )
  98  
  99  // FieldVal implements optimized fixed-precision arithmetic over the
 100  // secp256k1 finite field.  This means all arithmetic is performed modulo
 101  //
 102  //	0xfffffffffffffffffffffffffffffffffffffffffffffffffffffffefffffc2f.
 103  //
 104  // WARNING: Since it is so important for the field arithmetic to be extremely
 105  // fast for high performance crypto, this type does not perform any validation
 106  // of documented preconditions where it ordinarily would.  As a result, it is
 107  // IMPERATIVE for callers to understand some key concepts that are described
 108  // below and ensure the methods are called with the necessary preconditions that
 109  // each method is documented with.  For example, some methods only give the
 110  // correct result if the field value is normalized and others require the field
 111  // values involved to have a maximum magnitude and THERE ARE NO EXPLICIT CHECKS
 112  // TO ENSURE THOSE PRECONDITIONS ARE SATISFIED.  This does, unfortunately, make
 113  // the type more difficult to use correctly and while I typically prefer to
 114  // ensure all state and input is valid for most code, this is a bit of an
 115  // exception because those extra checks really add up in what ends up being
 116  // critical hot paths.
 117  //
 118  // The first key concept when working with this type is normalization.  In order
 119  // to avoid the need to propagate a ton of carries, the internal representation
 120  // provides additional overflow bits for each word of the overall 256-bit value.
 121  // This means that there are multiple internal representations for the same
 122  // value and, as a result, any methods that rely on comparison of the value,
 123  // such as equality and oddness determination, require the caller to provide a
 124  // normalized value.
 125  //
 126  // The second key concept when working with this type is magnitude.  As
 127  // previously mentioned, the internal representation provides additional
 128  // overflow bits which means that the more math operations that are performed on
 129  // the field value between normalizations, the more those overflow bits
 130  // accumulate.  The magnitude is effectively that maximum possible number of
 131  // those overflow bits that could possibly be required as a result of a given
 132  // operation.  Since there are only a limited number of overflow bits available,
 133  // this implies that the max possible magnitude MUST be tracked by the caller
 134  // and the caller MUST normalize the field value if a given operation would
 135  // cause the magnitude of the result to exceed the max allowed value.
 136  //
 137  // IMPORTANT: The max allowed magnitude of a field value is 64.
 138  type FieldVal struct {
 139  	// Each 256-bit value is represented as 10 32-bit integers in base 2^26.
 140  	// This provides 6 bits of overflow in each word (10 bits in the most
 141  	// significant word) for a total of 64 bits of overflow (9*6 + 10 = 64).  It
 142  	// only implements the arithmetic needed for elliptic curve operations.
 143  	//
 144  	// The following depicts the internal representation:
 145  	// 	 -----------------------------------------------------------------
 146  	// 	|        n[9]       |        n[8]       | ... |        n[0]       |
 147  	// 	| 32 bits available | 32 bits available | ... | 32 bits available |
 148  	// 	| 22 bits for value | 26 bits for value | ... | 26 bits for value |
 149  	// 	| 10 bits overflow  |  6 bits overflow  | ... |  6 bits overflow  |
 150  	// 	| Mult: 2^(26*9)    | Mult: 2^(26*8)    | ... | Mult: 2^(26*0)    |
 151  	// 	 -----------------------------------------------------------------
 152  	//
 153  	// For example, consider the number 2^49 + 1.  It would be represented as:
 154  	// 	n[0] = 1
 155  	// 	n[1] = 2^23
 156  	// 	n[2..9] = 0
 157  	//
 158  	// The full 256-bit value is then calculated by looping i from 9..0 and
 159  	// doing sum(n[i] * 2^(26i)) like so:
 160  	// 	n[9] * 2^(26*9) = 0    * 2^234 = 0
 161  	// 	n[8] * 2^(26*8) = 0    * 2^208 = 0
 162  	// 	...
 163  	// 	n[1] * 2^(26*1) = 2^23 * 2^26  = 2^49
 164  	// 	n[0] * 2^(26*0) = 1    * 2^0   = 1
 165  	// 	Sum: 0 + 0 + ... + 2^49 + 1 = 2^49 + 1
 166  	n [10]uint32
 167  }
 168  
 169  // String returns the field value as a normalized human-readable hex string.
 170  //
 171  //	Preconditions: None
 172  //	Output Normalized: Field is not modified -- same as input value
 173  //	Output Max Magnitude: Field is not modified -- same as input value
 174  func (f FieldVal) String() string {
 175  	// f is a copy, so it's safe to normalize it without mutating the original.
 176  	f.Normalize()
 177  	return hex.Enc(f.Bytes()[:])
 178  }
 179  
 180  // Zero sets the field value to zero in constant time.  A newly created field
 181  // value is already set to zero.  This function can be useful to clear an
 182  // existing field value for reuse.
 183  //
 184  //	Preconditions: None
 185  //	Output Normalized: Yes
 186  //	Output Max Magnitude: 1
 187  func (f *FieldVal) Zero() {
 188  	f.n[0] = 0
 189  	f.n[1] = 0
 190  	f.n[2] = 0
 191  	f.n[3] = 0
 192  	f.n[4] = 0
 193  	f.n[5] = 0
 194  	f.n[6] = 0
 195  	f.n[7] = 0
 196  	f.n[8] = 0
 197  	f.n[9] = 0
 198  }
 199  
 200  // Set sets the field value equal to the passed value in constant time.  The
 201  // normalization and magnitude of the two fields will be identical.
 202  //
 203  // The field value is returned to support chaining.  This enables syntax like:
 204  // f := new(FieldVal).Set(f2).Add(1) so that f = f2 + 1 where f2 is not
 205  // modified.
 206  //
 207  //	Preconditions: None
 208  //	Output Normalized: Same as input value
 209  //	Output Max Magnitude: Same as input value
 210  func (f *FieldVal) Set(val *FieldVal) *FieldVal {
 211  	*f = *val
 212  	return f
 213  }
 214  
 215  // SetInt sets the field value to the passed integer in constant time.  This is
 216  // a convenience function since it is fairly common to perform some arithmetic
 217  // with small native integers.
 218  //
 219  // The field value is returned to support chaining.  This enables syntax such
 220  // as f := new(FieldVal).SetInt(2).Mul(f2) so that f = 2 * f2.
 221  //
 222  //	Preconditions: None
 223  //	Output Normalized: Yes
 224  //	Output Max Magnitude: 1
 225  func (f *FieldVal) SetInt(ui uint16) *FieldVal {
 226  	f.Zero()
 227  	f.n[0] = uint32(ui)
 228  	return f
 229  }
 230  
 231  // SetBytes packs the passed 32-byte big-endian value into the internal field
 232  // value representation in constant time.  SetBytes interprets the provided
 233  // array as a 256-bit big-endian unsigned integer, packs it into the internal
 234  // field value representation, and returns either 1 if it is greater than or
 235  // equal to the field prime (aka it overflowed) or 0 otherwise in constant time.
 236  //
 237  // Note that a bool is not used here because it is not possible in Go to convert
 238  // from a bool to numeric value in constant time and many constant-time
 239  // operations require a numeric value.
 240  //
 241  //	Preconditions: None
 242  //	Output Normalized: Yes if no overflow, no otherwise
 243  //	Output Max Magnitude: 1
 244  func (f *FieldVal) SetBytes(b *[32]byte) uint32 {
 245  	// Pack the 256 total bits across the 10 uint32 words with a max of
 246  	// 26-bits per word.  This could be done with a couple of for loops,
 247  	// but this unrolled version is significantly faster.  Benchmarks show
 248  	// this is about 34 times faster than the variant which uses loops.
 249  	f.n[0] = uint32(b[31]) | uint32(b[30])<<8 | uint32(b[29])<<16 |
 250  		(uint32(b[28])&twoBitsMask)<<24
 251  	f.n[1] = uint32(b[28])>>2 | uint32(b[27])<<6 | uint32(b[26])<<14 |
 252  		(uint32(b[25])&fourBitsMask)<<22
 253  	f.n[2] = uint32(b[25])>>4 | uint32(b[24])<<4 | uint32(b[23])<<12 |
 254  		(uint32(b[22])&sixBitsMask)<<20
 255  	f.n[3] = uint32(b[22])>>6 | uint32(b[21])<<2 | uint32(b[20])<<10 |
 256  		uint32(b[19])<<18
 257  	f.n[4] = uint32(b[18]) | uint32(b[17])<<8 | uint32(b[16])<<16 |
 258  		(uint32(b[15])&twoBitsMask)<<24
 259  	f.n[5] = uint32(b[15])>>2 | uint32(b[14])<<6 | uint32(b[13])<<14 |
 260  		(uint32(b[12])&fourBitsMask)<<22
 261  	f.n[6] = uint32(b[12])>>4 | uint32(b[11])<<4 | uint32(b[10])<<12 |
 262  		(uint32(b[9])&sixBitsMask)<<20
 263  	f.n[7] = uint32(b[9])>>6 | uint32(b[8])<<2 | uint32(b[7])<<10 |
 264  		uint32(b[6])<<18
 265  	f.n[8] = uint32(b[5]) | uint32(b[4])<<8 | uint32(b[3])<<16 |
 266  		(uint32(b[2])&twoBitsMask)<<24
 267  	f.n[9] = uint32(b[2])>>2 | uint32(b[1])<<6 | uint32(b[0])<<14
 268  	// The intuition here is that the field value is greater than the prime if
 269  	// one of the higher individual words is greater than corresponding word of
 270  	// the prime and all higher words in the field value are equal to their
 271  	// corresponding word of the prime.  Since this type is modulo the prime,
 272  	// being equal is also an overflow back to 0.
 273  	//
 274  	// Note that because the input is 32 bytes and it was just packed into the
 275  	// field representation, the only words that can possibly be greater are
 276  	// zero and one, because ceil(log_2(2^256 - 1 - P)) = 33 bits max and the
 277  	// internal field representation encodes 26 bits with each word.
 278  	//
 279  	// Thus, there is no need to test if the upper words of the field value
 280  	// exceeds them, hence, only equality is checked for them.
 281  	highWordsEq := constantTimeEq(f.n[9], fieldPrimeWordNine)
 282  	highWordsEq &= constantTimeEq(f.n[8], fieldPrimeWordEight)
 283  	highWordsEq &= constantTimeEq(f.n[7], fieldPrimeWordSeven)
 284  	highWordsEq &= constantTimeEq(f.n[6], fieldPrimeWordSix)
 285  	highWordsEq &= constantTimeEq(f.n[5], fieldPrimeWordFive)
 286  	highWordsEq &= constantTimeEq(f.n[4], fieldPrimeWordFour)
 287  	highWordsEq &= constantTimeEq(f.n[3], fieldPrimeWordThree)
 288  	highWordsEq &= constantTimeEq(f.n[2], fieldPrimeWordTwo)
 289  	overflow := highWordsEq & constantTimeGreater(f.n[1], fieldPrimeWordOne)
 290  	highWordsEq &= constantTimeEq(f.n[1], fieldPrimeWordOne)
 291  	overflow |= highWordsEq & constantTimeGreaterOrEq(
 292  		f.n[0],
 293  		fieldPrimeWordZero,
 294  	)
 295  	return overflow
 296  }
 297  
 298  // SetByteSlice interprets the provided slice as a 256-bit big-endian unsigned
 299  // integer (meaning it is truncated to the first 32 bytes), packs it into the
 300  // internal field value representation, and returns whether or not the resulting
 301  // truncated 256-bit integer is greater than or equal to the field prime (aka it
 302  // overflowed) in constant time.
 303  //
 304  // Note that since passing a slice with more than 32 bytes is truncated, it is
 305  // possible that the truncated value is less than the field prime and hence it
 306  // will not be reported as having overflowed in that case.  It is up to the
 307  // caller to decide whether it needs to provide numbers of the appropriate size
 308  // or it if is acceptable to use this function with the described truncation and
 309  // overflow behavior.
 310  //
 311  //	Preconditions: None
 312  //	Output Normalized: Yes if no overflow, no otherwise
 313  //	Output Max Magnitude: 1
 314  func (f *FieldVal) SetByteSlice(b []byte) bool {
 315  	var b32 [32]byte
 316  	b = b[:constantTimeMin(uint32(len(b)), 32)]
 317  	copy(b32[:], b32[:32-len(b)])
 318  	copy(b32[32-len(b):], b)
 319  	result := f.SetBytes(&b32)
 320  	zeroArray32(&b32)
 321  	return result != 0
 322  }
 323  
 324  // Normalize normalizes the internal field words into the desired range and
 325  // performs fast modular reduction over the secp256k1 prime by making use of the
 326  // special form of the prime in constant time.
 327  //
 328  //	Preconditions: None
 329  //	Output Normalized: Yes
 330  //	Output Max Magnitude: 1
 331  func (f *FieldVal) Normalize() *FieldVal {
 332  	// The field representation leaves 6 bits of overflow in each word so
 333  	// intermediate calculations can be performed without needing to
 334  	// propagate the carry to each higher word during the calculations.  In
 335  	// order to normalize, we need to "compact" the full 256-bit value to
 336  	// the right while propagating any carries through to the high order
 337  	// word.
 338  	//
 339  	// Since this field is doing arithmetic modulo the secp256k1 prime, we
 340  	// also need to perform modular reduction over the prime.
 341  	//
 342  	// Per [HAC] section 14.3.4: Reduction method of moduli of special form,
 343  	// when the modulus is of the special form m = b^t - c, highly efficient
 344  	// reduction can be achieved.
 345  	//
 346  	// The secp256k1 prime is equivalent to 2^256 - 4294968273, so it fits
 347  	// this criteria.
 348  	//
 349  	// 4294968273 in field representation (base 2^26) is:
 350  	// n[0] = 977
 351  	// n[1] = 64
 352  	// That is to say (2^26 * 64) + 977 = 4294968273
 353  	//
 354  	// The algorithm presented in the referenced section typically repeats
 355  	// until the quotient is zero.  However, due to our field representation
 356  	// we already know to within one reduction how many times we would need
 357  	// to repeat as it's the uppermost bits of the high order word.  Thus we
 358  	// can simply multiply the magnitude by the field representation of the
 359  	// prime and do a single iteration.  After this step there might be an
 360  	// additional carry to bit 256 (bit 22 of the high order word).
 361  	t9 := f.n[9]
 362  	m := t9 >> fieldMSBBits
 363  	t9 = t9 & fieldMSBMask
 364  	t0 := f.n[0] + m*977
 365  	t1 := (t0 >> fieldBase) + f.n[1] + (m << 6)
 366  	t0 = t0 & fieldBaseMask
 367  	t2 := (t1 >> fieldBase) + f.n[2]
 368  	t1 = t1 & fieldBaseMask
 369  	t3 := (t2 >> fieldBase) + f.n[3]
 370  	t2 = t2 & fieldBaseMask
 371  	t4 := (t3 >> fieldBase) + f.n[4]
 372  	t3 = t3 & fieldBaseMask
 373  	t5 := (t4 >> fieldBase) + f.n[5]
 374  	t4 = t4 & fieldBaseMask
 375  	t6 := (t5 >> fieldBase) + f.n[6]
 376  	t5 = t5 & fieldBaseMask
 377  	t7 := (t6 >> fieldBase) + f.n[7]
 378  	t6 = t6 & fieldBaseMask
 379  	t8 := (t7 >> fieldBase) + f.n[8]
 380  	t7 = t7 & fieldBaseMask
 381  	t9 = (t8 >> fieldBase) + t9
 382  	t8 = t8 & fieldBaseMask
 383  	// At this point, the magnitude is guaranteed to be one, however, the
 384  	// value could still be greater than the prime if there was either a
 385  	// carry through to bit 256 (bit 22 of the higher order word) or the
 386  	// value is greater than or equal to the field characteristic.  The
 387  	// following determines if either or these conditions are true and does
 388  	// the final reduction in constant time.
 389  	//
 390  	// Also note that 'm' will be zero when neither of the aforementioned
 391  	// conditions are true and the value will not be changed when 'm' is zero.
 392  	m = constantTimeEq(t9, fieldMSBMask)
 393  	m &= constantTimeEq(t8&t7&t6&t5&t4&t3&t2, fieldBaseMask)
 394  	m &= constantTimeGreater(t1+64+((t0+977)>>fieldBase), fieldBaseMask)
 395  	m |= t9 >> fieldMSBBits
 396  	t0 = t0 + m*977
 397  	t1 = (t0 >> fieldBase) + t1 + (m << 6)
 398  	t0 = t0 & fieldBaseMask
 399  	t2 = (t1 >> fieldBase) + t2
 400  	t1 = t1 & fieldBaseMask
 401  	t3 = (t2 >> fieldBase) + t3
 402  	t2 = t2 & fieldBaseMask
 403  	t4 = (t3 >> fieldBase) + t4
 404  	t3 = t3 & fieldBaseMask
 405  	t5 = (t4 >> fieldBase) + t5
 406  	t4 = t4 & fieldBaseMask
 407  	t6 = (t5 >> fieldBase) + t6
 408  	t5 = t5 & fieldBaseMask
 409  	t7 = (t6 >> fieldBase) + t7
 410  	t6 = t6 & fieldBaseMask
 411  	t8 = (t7 >> fieldBase) + t8
 412  	t7 = t7 & fieldBaseMask
 413  	t9 = (t8 >> fieldBase) + t9
 414  	t8 = t8 & fieldBaseMask
 415  	t9 = t9 & fieldMSBMask // Remove potential multiple of 2^256.
 416  	// Finally, set the normalized and reduced words.
 417  	f.n[0] = t0
 418  	f.n[1] = t1
 419  	f.n[2] = t2
 420  	f.n[3] = t3
 421  	f.n[4] = t4
 422  	f.n[5] = t5
 423  	f.n[6] = t6
 424  	f.n[7] = t7
 425  	f.n[8] = t8
 426  	f.n[9] = t9
 427  	return f
 428  }
 429  
 430  // PutBytesUnchecked unpacks the field value to a 32-byte big-endian value
 431  // directly into the passed byte slice in constant time.  The target slice must
 432  // have at least 32 bytes available or it will panic.
 433  //
 434  // There is a similar function, PutBytes, which unpacks the field value into a
 435  // 32-byte array directly.  This version is provided since it can be useful
 436  // to write directly into part of a larger buffer without needing a separate
 437  // allocation.
 438  //
 439  //	Preconditions:
 440  //	  - The field value MUST be normalized
 441  //	  - The target slice MUST have at least 32 bytes available
 442  func (f *FieldVal) PutBytesUnchecked(b []byte) {
 443  	// Unpack the 256 total bits from the 10 uint32 words with a max of
 444  	// 26-bits per word.  This could be done with a couple of for loops,
 445  	// but this unrolled version is a bit faster.  Benchmarks show this is
 446  	// about 10 times faster than the variant which uses loops.
 447  	b[31] = byte(f.n[0] & eightBitsMask)
 448  	b[30] = byte((f.n[0] >> 8) & eightBitsMask)
 449  	b[29] = byte((f.n[0] >> 16) & eightBitsMask)
 450  	b[28] = byte((f.n[0]>>24)&twoBitsMask | (f.n[1]&sixBitsMask)<<2)
 451  	b[27] = byte((f.n[1] >> 6) & eightBitsMask)
 452  	b[26] = byte((f.n[1] >> 14) & eightBitsMask)
 453  	b[25] = byte((f.n[1]>>22)&fourBitsMask | (f.n[2]&fourBitsMask)<<4)
 454  	b[24] = byte((f.n[2] >> 4) & eightBitsMask)
 455  	b[23] = byte((f.n[2] >> 12) & eightBitsMask)
 456  	b[22] = byte((f.n[2]>>20)&sixBitsMask | (f.n[3]&twoBitsMask)<<6)
 457  	b[21] = byte((f.n[3] >> 2) & eightBitsMask)
 458  	b[20] = byte((f.n[3] >> 10) & eightBitsMask)
 459  	b[19] = byte((f.n[3] >> 18) & eightBitsMask)
 460  	b[18] = byte(f.n[4] & eightBitsMask)
 461  	b[17] = byte((f.n[4] >> 8) & eightBitsMask)
 462  	b[16] = byte((f.n[4] >> 16) & eightBitsMask)
 463  	b[15] = byte((f.n[4]>>24)&twoBitsMask | (f.n[5]&sixBitsMask)<<2)
 464  	b[14] = byte((f.n[5] >> 6) & eightBitsMask)
 465  	b[13] = byte((f.n[5] >> 14) & eightBitsMask)
 466  	b[12] = byte((f.n[5]>>22)&fourBitsMask | (f.n[6]&fourBitsMask)<<4)
 467  	b[11] = byte((f.n[6] >> 4) & eightBitsMask)
 468  	b[10] = byte((f.n[6] >> 12) & eightBitsMask)
 469  	b[9] = byte((f.n[6]>>20)&sixBitsMask | (f.n[7]&twoBitsMask)<<6)
 470  	b[8] = byte((f.n[7] >> 2) & eightBitsMask)
 471  	b[7] = byte((f.n[7] >> 10) & eightBitsMask)
 472  	b[6] = byte((f.n[7] >> 18) & eightBitsMask)
 473  	b[5] = byte(f.n[8] & eightBitsMask)
 474  	b[4] = byte((f.n[8] >> 8) & eightBitsMask)
 475  	b[3] = byte((f.n[8] >> 16) & eightBitsMask)
 476  	b[2] = byte((f.n[8]>>24)&twoBitsMask | (f.n[9]&sixBitsMask)<<2)
 477  	b[1] = byte((f.n[9] >> 6) & eightBitsMask)
 478  	b[0] = byte((f.n[9] >> 14) & eightBitsMask)
 479  }
 480  
 481  // PutBytes unpacks the field value to a 32-byte big-endian value using the
 482  // passed byte array in constant time.
 483  //
 484  // There is a similar function, PutBytesUnchecked, which unpacks the field value
 485  // into a slice that must have at least 32 bytes available.  This version is
 486  // provided since it can be useful to write directly into an array that is type
 487  // checked.
 488  //
 489  // Alternatively, there is also Bytes, which unpacks the field value into a new
 490  // array and returns that which can sometimes be more ergonomic in applications
 491  // that aren't concerned about an additional copy.
 492  //
 493  //	Preconditions:
 494  //	  - The field value MUST be normalized
 495  func (f *FieldVal) PutBytes(b *[32]byte) { f.PutBytesUnchecked(b[:]) }
 496  
 497  // Bytes unpacks the field value to a 32-byte big-endian value in constant time.
 498  //
 499  // See PutBytes and PutBytesUnchecked for variants that allow an array or slice
 500  // to be passed which can be useful to cut down on the number of allocations by
 501  // allowing the caller to reuse a buffer or write directly into part of a larger
 502  // buffer.
 503  //
 504  //	Preconditions:
 505  //	  - The field value MUST be normalized
 506  func (f *FieldVal) Bytes() *[32]byte {
 507  	b := &[32]byte{}
 508  	f.PutBytesUnchecked(b[:])
 509  	return b
 510  }
 511  
 512  // IsZeroBit returns 1 when the field value is equal to zero or 0 otherwise in
 513  // constant time.
 514  //
 515  // Note that a bool is not used here because it is not possible in Go to convert
 516  // from a bool to numeric value in constant time and many constant-time
 517  // operations require a numeric value.  See IsZero for the version that returns
 518  // a bool.
 519  //
 520  //	Preconditions:
 521  //	  - The field value MUST be normalized
 522  func (f *FieldVal) IsZeroBit() uint32 {
 523  	// The value can only be zero if no bits are set in any of the words.
 524  	// This is a constant time implementation.
 525  	bits := f.n[0] | f.n[1] | f.n[2] | f.n[3] | f.n[4] |
 526  		f.n[5] | f.n[6] | f.n[7] | f.n[8] | f.n[9]
 527  	return constantTimeEq(bits, 0)
 528  }
 529  
 530  // IsZero returns whether or not the field value is equal to zero in constant
 531  // time.
 532  //
 533  //	Preconditions:
 534  //	  - The field value MUST be normalized
 535  func (f *FieldVal) IsZero() bool {
 536  	// The value can only be zero if no bits are set in any of the words.
 537  	// This is a constant time implementation.
 538  	bits := f.n[0] | f.n[1] | f.n[2] | f.n[3] | f.n[4] |
 539  		f.n[5] | f.n[6] | f.n[7] | f.n[8] | f.n[9]
 540  	return bits == 0
 541  }
 542  
 543  // IsOneBit returns 1 when the field value is equal to one or 0 otherwise in
 544  // constant time.
 545  //
 546  // Note that a bool is not used here because it is not possible in Go to convert
 547  // from a bool to numeric value in constant time and many constant-time
 548  // operations require a numeric value.  See IsOne for the version that returns a
 549  // bool.
 550  //
 551  //	Preconditions:
 552  //	   - The field value MUST be normalized
 553  func (f *FieldVal) IsOneBit() uint32 {
 554  	// The value can only be one if the single lowest significant bit is set in
 555  	// the first word and no other bits are set in any of the other words.
 556  	// This is a constant time implementation.
 557  	bits := (f.n[0] ^ 1) | f.n[1] | f.n[2] | f.n[3] | f.n[4] | f.n[5] |
 558  		f.n[6] | f.n[7] | f.n[8] | f.n[9]
 559  
 560  	return constantTimeEq(bits, 0)
 561  }
 562  
 563  // IsOne returns whether the field value is equal to one in constant
 564  // time.
 565  //
 566  //	Preconditions:
 567  //	  - The field value MUST be normalized
 568  func (f *FieldVal) IsOne() bool {
 569  	// The value can only be one if the single lowest significant bit is set in
 570  	// the first word and no other bits are set in any of the other words.
 571  	// This is a constant time implementation.
 572  	bits := (f.n[0] ^ 1) | f.n[1] | f.n[2] | f.n[3] | f.n[4] | f.n[5] |
 573  		f.n[6] | f.n[7] | f.n[8] | f.n[9]
 574  
 575  	return bits == 0
 576  }
 577  
 578  // IsOddBit returns 1 when the field value is an odd number or 0 otherwise in
 579  // constant time.
 580  //
 581  // Note that a bool is not used here because it is not possible in Go to convert
 582  // from a bool to numeric value in constant time and many constant-time
 583  // operations require a numeric value.  See IsOdd for the version that returns a
 584  // bool.
 585  //
 586  //	Preconditions:
 587  //	  - The field value MUST be normalized
 588  func (f *FieldVal) IsOddBit() uint32 {
 589  	// Only odd numbers have the bottom bit set.
 590  	return f.n[0] & 1
 591  }
 592  
 593  // IsOdd returns whether the field value is an odd number in constant
 594  // time.
 595  //
 596  //	Preconditions:
 597  //	  - The field value MUST be normalized
 598  func (f *FieldVal) IsOdd() bool {
 599  	// Only odd numbers have the bottom bit set.
 600  	return f.n[0]&1 == 1
 601  }
 602  
 603  // Equals returns whether or not the two field values are the same in constant
 604  // time.
 605  //
 606  //	Preconditions:
 607  //	  - Both field values being compared MUST be normalized
 608  func (f *FieldVal) Equals(val *FieldVal) bool {
 609  	// Xor only sets bits when they are different, so the two field values
 610  	// can only be the same if no bits are set after xoring each word.
 611  	// This is a constant time implementation.
 612  	bits := (f.n[0] ^ val.n[0]) | (f.n[1] ^ val.n[1]) | (f.n[2] ^ val.n[2]) |
 613  		(f.n[3] ^ val.n[3]) | (f.n[4] ^ val.n[4]) | (f.n[5] ^ val.n[5]) |
 614  		(f.n[6] ^ val.n[6]) | (f.n[7] ^ val.n[7]) | (f.n[8] ^ val.n[8]) |
 615  		(f.n[9] ^ val.n[9])
 616  
 617  	return bits == 0
 618  }
 619  
 620  // NegateVal negates the passed value and stores the result in f in constant
 621  // time.  The caller must provide the magnitude of the passed value for a
 622  // correct result.
 623  //
 624  // The field value is returned to support chaining.  This enables syntax like:
 625  // f.NegateVal(f2).AddInt(1) so that f = -f2 + 1.
 626  //
 627  //	Preconditions:
 628  //	  - The max magnitude MUST be 63
 629  //	Output Normalized: No
 630  //	Output Max Magnitude: Input magnitude + 1
 631  func (f *FieldVal) NegateVal(val *FieldVal, magnitude uint32) *FieldVal {
 632  	// Negation in the field is just the prime minus the value.  However,
 633  	// in order to allow negation against a field value without having to
 634  	// normalize/reduce it first, multiply by the magnitude (that is how
 635  	// "far" away it is from the normalized value) to adjust.  Also, since
 636  	// negating a value pushes it one more order of magnitude away from the
 637  	// normalized range, add 1 to compensate.
 638  	//
 639  	// For some intuition here, imagine you're performing mod 12 arithmetic
 640  	// (picture a clock) and you are negating the number 7.  So you start at
 641  	// 12 (which is of course 0 under mod 12) and count backwards (left on
 642  	// the clock) 7 times to arrive at 5.  Notice this is just 12-7 = 5.
 643  	// Now, assume you're starting with 19, which is a number that is
 644  	// already larger than the modulus and congruent to 7 (mod 12).  When a
 645  	// value is already in the desired range, its magnitude is 1.  Since 19
 646  	// is an additional "step", its magnitude (mod 12) is 2.  Since any
 647  	// multiple of the modulus is congruent to zero (mod m), the answer can
 648  	// be shortcut by simply multiplying the magnitude by the modulus and
 649  	// subtracting.  Keeping with the example, this would be (2*12)-19 = 5.
 650  	f.n[0] = (magnitude+1)*fieldPrimeWordZero - val.n[0]
 651  	f.n[1] = (magnitude+1)*fieldPrimeWordOne - val.n[1]
 652  	f.n[2] = (magnitude+1)*fieldBaseMask - val.n[2]
 653  	f.n[3] = (magnitude+1)*fieldBaseMask - val.n[3]
 654  	f.n[4] = (magnitude+1)*fieldBaseMask - val.n[4]
 655  	f.n[5] = (magnitude+1)*fieldBaseMask - val.n[5]
 656  	f.n[6] = (magnitude+1)*fieldBaseMask - val.n[6]
 657  	f.n[7] = (magnitude+1)*fieldBaseMask - val.n[7]
 658  	f.n[8] = (magnitude+1)*fieldBaseMask - val.n[8]
 659  	f.n[9] = (magnitude+1)*fieldMSBMask - val.n[9]
 660  	return f
 661  }
 662  
 663  // Negate negates the field value in constant time.  The existing field value is
 664  // modified.  The caller must provide the magnitude of the field value for a
 665  // correct result.
 666  //
 667  // The field value is returned to support chaining.  This enables syntax like:
 668  // f.Negate().AddInt(1) so that f = -f + 1.
 669  //
 670  //	Preconditions:
 671  //	  - The max magnitude MUST be 63
 672  //	Output Normalized: No
 673  //	Output Max Magnitude: Input magnitude + 1
 674  func (f *FieldVal) Negate(magnitude uint32) *FieldVal {
 675  	return f.NegateVal(f, magnitude)
 676  }
 677  
 678  // AddInt adds the passed integer to the existing field value and stores the
 679  // result in f in constant time.  This is a convenience function since it is
 680  // fairly common to perform some arithmetic with small native integers.
 681  //
 682  // The field value is returned to support chaining.  This enables syntax like:
 683  // f.AddInt(1).Add(f2) so that f = f + 1 + f2.
 684  //
 685  //	Preconditions:
 686  //	  - The field value MUST have a max magnitude of 63
 687  //	Output Normalized: No
 688  //	Output Max Magnitude: Existing field magnitude + 1
 689  func (f *FieldVal) AddInt(ui uint16) *FieldVal {
 690  	// Since the field representation intentionally provides overflow bits,
 691  	// it's ok to use carryless addition as the carry bit is safely part of
 692  	// the word and will be normalized out.
 693  	f.n[0] += uint32(ui)
 694  	return f
 695  }
 696  
 697  // Add adds the passed value to the existing field value and stores the result
 698  // in f in constant time.
 699  //
 700  // The field value is returned to support chaining.  This enables syntax like:
 701  // f.Add(f2).AddInt(1) so that f = f + f2 + 1.
 702  //
 703  //	Preconditions:
 704  //	  - The sum of the magnitudes of the two field values MUST be a max of 64
 705  //	Output Normalized: No
 706  //	Output Max Magnitude: Sum of the magnitude of the two individual field values
 707  func (f *FieldVal) Add(val *FieldVal) *FieldVal {
 708  	// Since the field representation intentionally provides overflow bits,
 709  	// it's ok to use carryless addition as the carry bit is safely part of
 710  	// each word and will be normalized out.  This could obviously be done
 711  	// in a loop, but the unrolled version is faster.
 712  	f.n[0] += val.n[0]
 713  	f.n[1] += val.n[1]
 714  	f.n[2] += val.n[2]
 715  	f.n[3] += val.n[3]
 716  	f.n[4] += val.n[4]
 717  	f.n[5] += val.n[5]
 718  	f.n[6] += val.n[6]
 719  	f.n[7] += val.n[7]
 720  	f.n[8] += val.n[8]
 721  	f.n[9] += val.n[9]
 722  	return f
 723  }
 724  
 725  // Add2 adds the passed two field values together and stores the result in f in
 726  // constant time.
 727  //
 728  // The field value is returned to support chaining.  This enables syntax like:
 729  // f3.Add2(f, f2).AddInt(1) so that f3 = f + f2 + 1.
 730  //
 731  //	Preconditions:
 732  //	  - The sum of the magnitudes of the two field values MUST be a max of 64
 733  //	Output Normalized: No
 734  //	Output Max Magnitude: Sum of the magnitude of the two field values
 735  func (f *FieldVal) Add2(val *FieldVal, val2 *FieldVal) *FieldVal {
 736  	// Since the field representation intentionally provides overflow bits,
 737  	// it's ok to use carryless addition as the carry bit is safely part of
 738  	// each word and will be normalized out.  This could obviously be done
 739  	// in a loop, but the unrolled version is faster.
 740  	f.n[0] = val.n[0] + val2.n[0]
 741  	f.n[1] = val.n[1] + val2.n[1]
 742  	f.n[2] = val.n[2] + val2.n[2]
 743  	f.n[3] = val.n[3] + val2.n[3]
 744  	f.n[4] = val.n[4] + val2.n[4]
 745  	f.n[5] = val.n[5] + val2.n[5]
 746  	f.n[6] = val.n[6] + val2.n[6]
 747  	f.n[7] = val.n[7] + val2.n[7]
 748  	f.n[8] = val.n[8] + val2.n[8]
 749  	f.n[9] = val.n[9] + val2.n[9]
 750  	return f
 751  }
 752  
 753  // MulInt multiplies the field value by the passed int and stores the result in
 754  // f in constant time.  Note that this function can overflow if multiplying the
 755  // value by any of the individual words exceeds a max uint32.  Therefore it is
 756  // important that the caller ensures no overflows will occur before using this
 757  // function.
 758  //
 759  // The field value is returned to support chaining.  This enables syntax like:
 760  // f.MulInt(2).Add(f2) so that f = 2 * f + f2.
 761  //
 762  //	Preconditions:
 763  //	  - The field value magnitude multiplied by given val MUST be a max of 64
 764  //	Output Normalized: No
 765  //	Output Max Magnitude: Existing field magnitude times the provided integer val
 766  func (f *FieldVal) MulInt(val uint8) *FieldVal {
 767  	// Since each word of the field representation can hold up to
 768  	// 32 - fieldBase extra bits which will be normalized out, it's safe
 769  	// to multiply each word without using a larger type or carry
 770  	// propagation so long as the values won't overflow a uint32.  This
 771  	// could obviously be done in a loop, but the unrolled version is
 772  	// faster.
 773  	ui := uint32(val)
 774  	f.n[0] *= ui
 775  	f.n[1] *= ui
 776  	f.n[2] *= ui
 777  	f.n[3] *= ui
 778  	f.n[4] *= ui
 779  	f.n[5] *= ui
 780  	f.n[6] *= ui
 781  	f.n[7] *= ui
 782  	f.n[8] *= ui
 783  	f.n[9] *= ui
 784  	return f
 785  }
 786  
 787  // Mul multiplies the passed value to the existing field value and stores the
 788  // result in f in constant time.  Note that this function can overflow if
 789  // multiplying any of the individual words exceeds a max uint32.  In practice,
 790  // this means the magnitude of either value involved in the multiplication must
 791  // be a max of 8.
 792  //
 793  // The field value is returned to support chaining.  This enables syntax like:
 794  // f.Mul(f2).AddInt(1) so that f = (f * f2) + 1.
 795  //
 796  //	Preconditions:
 797  //	  - Both field values MUST have a max magnitude of 8
 798  //	Output Normalized: No
 799  //	Output Max Magnitude: 1
 800  func (f *FieldVal) Mul(val *FieldVal) *FieldVal {
 801  	return f.Mul2(f, val)
 802  }
 803  
 804  // Mul2 multiplies the passed two field values together and stores the result in
 805  // f in constant time.  Note that this function can overflow if multiplying any
 806  // of the individual words exceeds a max uint32.  In practice, this means the
 807  // magnitude of either value involved in the multiplication must be a max of 8.
 808  //
 809  // The field value is returned to support chaining.  This enables syntax like:
 810  // f3.Mul2(f, f2).AddInt(1) so that f3 = (f * f2) + 1.
 811  //
 812  //	Preconditions:
 813  //	  - Both input field values MUST have a max magnitude of 8
 814  //	Output Normalized: No
 815  //	Output Max Magnitude: 1
 816  func (f *FieldVal) Mul2(val *FieldVal, val2 *FieldVal) *FieldVal {
 817  	// This could be done with a couple of for loops and an array to store
 818  	// the intermediate terms, but this unrolled version is significantly
 819  	// faster.
 820  	// Terms for 2^(fieldBase*0).
 821  	m := uint64(val.n[0]) * uint64(val2.n[0])
 822  	t0 := m & fieldBaseMask
 823  	// Terms for 2^(fieldBase*1).
 824  	m = (m >> fieldBase) +
 825  		uint64(val.n[0])*uint64(val2.n[1]) +
 826  		uint64(val.n[1])*uint64(val2.n[0])
 827  	t1 := m & fieldBaseMask
 828  	// Terms for 2^(fieldBase*2).
 829  	m = (m >> fieldBase) +
 830  		uint64(val.n[0])*uint64(val2.n[2]) +
 831  		uint64(val.n[1])*uint64(val2.n[1]) +
 832  		uint64(val.n[2])*uint64(val2.n[0])
 833  	t2 := m & fieldBaseMask
 834  	// Terms for 2^(fieldBase*3).
 835  	m = (m >> fieldBase) +
 836  		uint64(val.n[0])*uint64(val2.n[3]) +
 837  		uint64(val.n[1])*uint64(val2.n[2]) +
 838  		uint64(val.n[2])*uint64(val2.n[1]) +
 839  		uint64(val.n[3])*uint64(val2.n[0])
 840  	t3 := m & fieldBaseMask
 841  	// Terms for 2^(fieldBase*4).
 842  	m = (m >> fieldBase) +
 843  		uint64(val.n[0])*uint64(val2.n[4]) +
 844  		uint64(val.n[1])*uint64(val2.n[3]) +
 845  		uint64(val.n[2])*uint64(val2.n[2]) +
 846  		uint64(val.n[3])*uint64(val2.n[1]) +
 847  		uint64(val.n[4])*uint64(val2.n[0])
 848  	t4 := m & fieldBaseMask
 849  	// Terms for 2^(fieldBase*5).
 850  	m = (m >> fieldBase) +
 851  		uint64(val.n[0])*uint64(val2.n[5]) +
 852  		uint64(val.n[1])*uint64(val2.n[4]) +
 853  		uint64(val.n[2])*uint64(val2.n[3]) +
 854  		uint64(val.n[3])*uint64(val2.n[2]) +
 855  		uint64(val.n[4])*uint64(val2.n[1]) +
 856  		uint64(val.n[5])*uint64(val2.n[0])
 857  	t5 := m & fieldBaseMask
 858  	// Terms for 2^(fieldBase*6).
 859  	m = (m >> fieldBase) +
 860  		uint64(val.n[0])*uint64(val2.n[6]) +
 861  		uint64(val.n[1])*uint64(val2.n[5]) +
 862  		uint64(val.n[2])*uint64(val2.n[4]) +
 863  		uint64(val.n[3])*uint64(val2.n[3]) +
 864  		uint64(val.n[4])*uint64(val2.n[2]) +
 865  		uint64(val.n[5])*uint64(val2.n[1]) +
 866  		uint64(val.n[6])*uint64(val2.n[0])
 867  	t6 := m & fieldBaseMask
 868  	// Terms for 2^(fieldBase*7).
 869  	m = (m >> fieldBase) +
 870  		uint64(val.n[0])*uint64(val2.n[7]) +
 871  		uint64(val.n[1])*uint64(val2.n[6]) +
 872  		uint64(val.n[2])*uint64(val2.n[5]) +
 873  		uint64(val.n[3])*uint64(val2.n[4]) +
 874  		uint64(val.n[4])*uint64(val2.n[3]) +
 875  		uint64(val.n[5])*uint64(val2.n[2]) +
 876  		uint64(val.n[6])*uint64(val2.n[1]) +
 877  		uint64(val.n[7])*uint64(val2.n[0])
 878  	t7 := m & fieldBaseMask
 879  	// Terms for 2^(fieldBase*8).
 880  	m = (m >> fieldBase) +
 881  		uint64(val.n[0])*uint64(val2.n[8]) +
 882  		uint64(val.n[1])*uint64(val2.n[7]) +
 883  		uint64(val.n[2])*uint64(val2.n[6]) +
 884  		uint64(val.n[3])*uint64(val2.n[5]) +
 885  		uint64(val.n[4])*uint64(val2.n[4]) +
 886  		uint64(val.n[5])*uint64(val2.n[3]) +
 887  		uint64(val.n[6])*uint64(val2.n[2]) +
 888  		uint64(val.n[7])*uint64(val2.n[1]) +
 889  		uint64(val.n[8])*uint64(val2.n[0])
 890  	t8 := m & fieldBaseMask
 891  	// Terms for 2^(fieldBase*9).
 892  	m = (m >> fieldBase) +
 893  		uint64(val.n[0])*uint64(val2.n[9]) +
 894  		uint64(val.n[1])*uint64(val2.n[8]) +
 895  		uint64(val.n[2])*uint64(val2.n[7]) +
 896  		uint64(val.n[3])*uint64(val2.n[6]) +
 897  		uint64(val.n[4])*uint64(val2.n[5]) +
 898  		uint64(val.n[5])*uint64(val2.n[4]) +
 899  		uint64(val.n[6])*uint64(val2.n[3]) +
 900  		uint64(val.n[7])*uint64(val2.n[2]) +
 901  		uint64(val.n[8])*uint64(val2.n[1]) +
 902  		uint64(val.n[9])*uint64(val2.n[0])
 903  	t9 := m & fieldBaseMask
 904  	// Terms for 2^(fieldBase*10).
 905  	m = (m >> fieldBase) +
 906  		uint64(val.n[1])*uint64(val2.n[9]) +
 907  		uint64(val.n[2])*uint64(val2.n[8]) +
 908  		uint64(val.n[3])*uint64(val2.n[7]) +
 909  		uint64(val.n[4])*uint64(val2.n[6]) +
 910  		uint64(val.n[5])*uint64(val2.n[5]) +
 911  		uint64(val.n[6])*uint64(val2.n[4]) +
 912  		uint64(val.n[7])*uint64(val2.n[3]) +
 913  		uint64(val.n[8])*uint64(val2.n[2]) +
 914  		uint64(val.n[9])*uint64(val2.n[1])
 915  	t10 := m & fieldBaseMask
 916  	// Terms for 2^(fieldBase*11).
 917  	m = (m >> fieldBase) +
 918  		uint64(val.n[2])*uint64(val2.n[9]) +
 919  		uint64(val.n[3])*uint64(val2.n[8]) +
 920  		uint64(val.n[4])*uint64(val2.n[7]) +
 921  		uint64(val.n[5])*uint64(val2.n[6]) +
 922  		uint64(val.n[6])*uint64(val2.n[5]) +
 923  		uint64(val.n[7])*uint64(val2.n[4]) +
 924  		uint64(val.n[8])*uint64(val2.n[3]) +
 925  		uint64(val.n[9])*uint64(val2.n[2])
 926  	t11 := m & fieldBaseMask
 927  	// Terms for 2^(fieldBase*12).
 928  	m = (m >> fieldBase) +
 929  		uint64(val.n[3])*uint64(val2.n[9]) +
 930  		uint64(val.n[4])*uint64(val2.n[8]) +
 931  		uint64(val.n[5])*uint64(val2.n[7]) +
 932  		uint64(val.n[6])*uint64(val2.n[6]) +
 933  		uint64(val.n[7])*uint64(val2.n[5]) +
 934  		uint64(val.n[8])*uint64(val2.n[4]) +
 935  		uint64(val.n[9])*uint64(val2.n[3])
 936  	t12 := m & fieldBaseMask
 937  	// Terms for 2^(fieldBase*13).
 938  	m = (m >> fieldBase) +
 939  		uint64(val.n[4])*uint64(val2.n[9]) +
 940  		uint64(val.n[5])*uint64(val2.n[8]) +
 941  		uint64(val.n[6])*uint64(val2.n[7]) +
 942  		uint64(val.n[7])*uint64(val2.n[6]) +
 943  		uint64(val.n[8])*uint64(val2.n[5]) +
 944  		uint64(val.n[9])*uint64(val2.n[4])
 945  	t13 := m & fieldBaseMask
 946  	// Terms for 2^(fieldBase*14).
 947  	m = (m >> fieldBase) +
 948  		uint64(val.n[5])*uint64(val2.n[9]) +
 949  		uint64(val.n[6])*uint64(val2.n[8]) +
 950  		uint64(val.n[7])*uint64(val2.n[7]) +
 951  		uint64(val.n[8])*uint64(val2.n[6]) +
 952  		uint64(val.n[9])*uint64(val2.n[5])
 953  	t14 := m & fieldBaseMask
 954  	// Terms for 2^(fieldBase*15).
 955  	m = (m >> fieldBase) +
 956  		uint64(val.n[6])*uint64(val2.n[9]) +
 957  		uint64(val.n[7])*uint64(val2.n[8]) +
 958  		uint64(val.n[8])*uint64(val2.n[7]) +
 959  		uint64(val.n[9])*uint64(val2.n[6])
 960  	t15 := m & fieldBaseMask
 961  	// Terms for 2^(fieldBase*16).
 962  	m = (m >> fieldBase) +
 963  		uint64(val.n[7])*uint64(val2.n[9]) +
 964  		uint64(val.n[8])*uint64(val2.n[8]) +
 965  		uint64(val.n[9])*uint64(val2.n[7])
 966  	t16 := m & fieldBaseMask
 967  	// Terms for 2^(fieldBase*17).
 968  	m = (m >> fieldBase) +
 969  		uint64(val.n[8])*uint64(val2.n[9]) +
 970  		uint64(val.n[9])*uint64(val2.n[8])
 971  	t17 := m & fieldBaseMask
 972  	// Terms for 2^(fieldBase*18).
 973  	m = (m >> fieldBase) + uint64(val.n[9])*uint64(val2.n[9])
 974  	t18 := m & fieldBaseMask
 975  	// What's left is for 2^(fieldBase*19).
 976  	t19 := m >> fieldBase
 977  	// At this point, all of the terms are grouped into their respective
 978  	// base.
 979  	//
 980  	// Per [HAC] section 14.3.4: Reduction method of moduli of special form,
 981  	// when the modulus is of the special form m = b^t - c, highly efficient
 982  	// reduction can be achieved per the provided algorithm.
 983  	//
 984  	// The secp256k1 prime is equivalent to 2^256 - 4294968273, so it fits
 985  	// this criteria.
 986  	//
 987  	// 4294968273 in field representation (base 2^26) is:
 988  	// n[0] = 977
 989  	// n[1] = 64
 990  	// That is to say (2^26 * 64) + 977 = 4294968273
 991  	//
 992  	// Since each word is in base 26, the upper terms (t10 and up) start
 993  	// at 260 bits (versus the final desired range of 256 bits), so the
 994  	// field representation of 'c' from above needs to be adjusted for the
 995  	// extra 4 bits by multiplying it by 2^4 = 16.  4294968273 * 16 =
 996  	// 68719492368.  Thus, the adjusted field representation of 'c' is:
 997  	// n[0] = 977 * 16 = 15632
 998  	// n[1] = 64 * 16 = 1024
 999  	// That is to say (2^26 * 1024) + 15632 = 68719492368
1000  	//
1001  	// To reduce the final term, t19, the entire 'c' value is needed instead
1002  	// of only n[0] because there are no more terms left to handle n[1].
1003  	// This means there might be some magnitude left in the upper bits that
1004  	// is handled below.
1005  	m = t0 + t10*15632
1006  	t0 = m & fieldBaseMask
1007  	m = (m >> fieldBase) + t1 + t10*1024 + t11*15632
1008  	t1 = m & fieldBaseMask
1009  	m = (m >> fieldBase) + t2 + t11*1024 + t12*15632
1010  	t2 = m & fieldBaseMask
1011  	m = (m >> fieldBase) + t3 + t12*1024 + t13*15632
1012  	t3 = m & fieldBaseMask
1013  	m = (m >> fieldBase) + t4 + t13*1024 + t14*15632
1014  	t4 = m & fieldBaseMask
1015  	m = (m >> fieldBase) + t5 + t14*1024 + t15*15632
1016  	t5 = m & fieldBaseMask
1017  	m = (m >> fieldBase) + t6 + t15*1024 + t16*15632
1018  	t6 = m & fieldBaseMask
1019  	m = (m >> fieldBase) + t7 + t16*1024 + t17*15632
1020  	t7 = m & fieldBaseMask
1021  	m = (m >> fieldBase) + t8 + t17*1024 + t18*15632
1022  	t8 = m & fieldBaseMask
1023  	m = (m >> fieldBase) + t9 + t18*1024 + t19*68719492368
1024  	t9 = m & fieldMSBMask
1025  	m = m >> fieldMSBBits
1026  	// At this point, if the magnitude is greater than 0, the overall value
1027  	// is greater than the max possible 256-bit value.  In particular, it is
1028  	// "how many times larger" than the max value it is.
1029  	//
1030  	// The algorithm presented in [HAC] section 14.3.4 repeats until the
1031  	// quotient is zero.  However, due to the above, we already know at
1032  	// least how many times we would need to repeat as it's the value
1033  	// currently in m.  Thus we can simply multiply the magnitude by the
1034  	// field representation of the prime and do a single iteration.  Notice
1035  	// that nothing will be changed when the magnitude is zero, so we could
1036  	// skip this in that case, however always running regardless allows it
1037  	// to run in constant time.  The final result will be in the range
1038  	// 0 <= result <= prime + (2^64 - c), so it is guaranteed to have a
1039  	// magnitude of 1, but it is denormalized.
1040  	d := t0 + m*977
1041  	f.n[0] = uint32(d & fieldBaseMask)
1042  	d = (d >> fieldBase) + t1 + m*64
1043  	f.n[1] = uint32(d & fieldBaseMask)
1044  	f.n[2] = uint32((d >> fieldBase) + t2)
1045  	f.n[3] = uint32(t3)
1046  	f.n[4] = uint32(t4)
1047  	f.n[5] = uint32(t5)
1048  	f.n[6] = uint32(t6)
1049  	f.n[7] = uint32(t7)
1050  	f.n[8] = uint32(t8)
1051  	f.n[9] = uint32(t9)
1052  	return f
1053  }
1054  
1055  // SquareRootVal either calculates the square root of the passed value when it
1056  // exists or the square root of the negation of the value when it does not exist
1057  // and stores the result in f in constant time.  The return flag is true when
1058  // the calculated square root is for the passed value itself and false when it
1059  // is for its negation.
1060  //
1061  // Note that this function can overflow if multiplying any of the individual
1062  // words exceeds a max uint32.  In practice, this means the magnitude of the
1063  // field must be a max of 8 to prevent overflow.  The magnitude of the result
1064  // will be 1.
1065  //
1066  //	Preconditions:
1067  //	  - The input field value MUST have a max magnitude of 8
1068  //	Output Normalized: No
1069  //	Output Max Magnitude: 1
1070  func (f *FieldVal) SquareRootVal(val *FieldVal) bool {
1071  	// This uses the Tonelli-Shanks method for calculating the square root of
1072  	// the value when it exists.  The key principles of the method follow.
1073  	//
1074  	// Fermat's little theorem states that for a nonzero number 'a' and prime
1075  	// 'p', a^(p-1) ≡ 1 (mod p).
1076  	//
1077  	// Further, Euler's criterion states that an integer 'a' has a square root
1078  	// (aka is a quadratic residue) modulo a prime if a^((p-1)/2) ≡ 1 (mod p)
1079  	// and, conversely, when it does NOT have a square root (aka 'a' is a
1080  	// non-residue) a^((p-1)/2) ≡ -1 (mod p).
1081  	//
1082  	// This can be seen by considering that Fermat's little theorem can be
1083  	// written as (a^((p-1)/2) - 1)(a^((p-1)/2) + 1) ≡ 0 (mod p).  Therefore,
1084  	// one of the two factors must be 0.  Then, when a ≡ x^2 (aka 'a' is a
1085  	// quadratic residue), (x^2)^((p-1)/2) ≡ x^(p-1) ≡ 1 (mod p) which implies
1086  	// the first factor must be zero.  Finally, per Lagrange's theorem, the
1087  	// non-residues are the only remaining possible solutions and thus must make
1088  	// the second factor zero to satisfy Fermat's little theorem implying that
1089  	// a^((p-1)/2) ≡ -1 (mod p) for that case.
1090  	//
1091  	// The Tonelli-Shanks method uses these facts along with factoring out
1092  	// powers of two to solve a congruence that results in either the solution
1093  	// when the square root exists or the square root of the negation of the
1094  	// value when it does not.  In the case of primes that are ≡ 3 (mod 4), the
1095  	// possible solutions are r = ±a^((p+1)/4) (mod p).  Therefore, either r^2 ≡
1096  	// a (mod p) is true in which case ±r are the two solutions, or r^2 ≡ -a
1097  	// (mod p) in which case 'a' is a non-residue and there are no solutions.
1098  	//
1099  	// The secp256k1 prime is ≡ 3 (mod 4), so this result applies.
1100  	//
1101  	// In other words, calculate a^((p+1)/4) and then square it and check it
1102  	// against the original value to determine if it is actually the square
1103  	// root.
1104  	//
1105  	// In order to efficiently compute a^((p+1)/4), (p+1)/4 needs to be split
1106  	// into a sequence of squares and multiplications that minimizes the number
1107  	// of multiplications needed (since they are more costly than squarings).
1108  	//
1109  	// The secp256k1 prime + 1 / 4 is 2^254 - 2^30 - 244.  In binary, that is:
1110  	//
1111  	// 00111111 11111111 11111111 11111111
1112  	// 11111111 11111111 11111111 11111111
1113  	// 11111111 11111111 11111111 11111111
1114  	// 11111111 11111111 11111111 11111111
1115  	// 11111111 11111111 11111111 11111111
1116  	// 11111111 11111111 11111111 11111111
1117  	// 11111111 11111111 11111111 11111111
1118  	// 10111111 11111111 11111111 00001100
1119  	//
1120  	// Notice that can be broken up into three windows of consecutive 1s (in
1121  	// order of least to most significant) as:
1122  	//
1123  	//   6-bit window with two bits set (bits 4, 5, 6, 7 unset)
1124  	//   23-bit window with 22 bits set (bit 30 unset)
1125  	//   223-bit window with all 223 bits set
1126  	//
1127  	// Thus, the groups of 1 bits in each window forms the set:
1128  	// S = {2, 22, 223}.
1129  	//
1130  	// The strategy is to calculate a^(2^n - 1) for each grouping via an
1131  	// addition chain with a sliding window.
1132  	//
1133  	// The addition chain used is (credits to Peter Dettman):
1134  	// (0,0),(1,0),(2,2),(3,2),(4,1),(5,5),(6,6),(7,7),(8,8),(9,7),(10,2)
1135  	// => 2^1 2^[2] 2^3 2^6 2^9 2^11 2^[22] 2^44 2^88 2^176 2^220 2^[223]
1136  	//
1137  	// This has a cost of 254 field squarings and 13 field multiplications.
1138  	var a, a2, a3, a6, a9, a11, a22, a44, a88, a176, a220, a223 FieldVal
1139  	a.Set(val)
1140  	a2.SquareVal(&a).Mul(&a)                                  // a2 = a^(2^2 - 1)
1141  	a3.SquareVal(&a2).Mul(&a)                                 // a3 = a^(2^3 - 1)
1142  	a6.SquareVal(&a3).Square().Square()                       // a6 = a^(2^6 - 2^3)
1143  	a6.Mul(&a3)                                               // a6 = a^(2^6 - 1)
1144  	a9.SquareVal(&a6).Square().Square()                       // a9 = a^(2^9 - 2^3)
1145  	a9.Mul(&a3)                                               // a9 = a^(2^9 - 1)
1146  	a11.SquareVal(&a9).Square()                               // a11 = a^(2^11 - 2^2)
1147  	a11.Mul(&a2)                                              // a11 = a^(2^11 - 1)
1148  	a22.SquareVal(&a11).Square().Square().Square().Square()   // a22 = a^(2^16 - 2^5)
1149  	a22.Square().Square().Square().Square().Square()          // a22 = a^(2^21 - 2^10)
1150  	a22.Square()                                              // a22 = a^(2^22 - 2^11)
1151  	a22.Mul(&a11)                                             // a22 = a^(2^22 - 1)
1152  	a44.SquareVal(&a22).Square().Square().Square().Square()   // a44 = a^(2^27 - 2^5)
1153  	a44.Square().Square().Square().Square().Square()          // a44 = a^(2^32 - 2^10)
1154  	a44.Square().Square().Square().Square().Square()          // a44 = a^(2^37 - 2^15)
1155  	a44.Square().Square().Square().Square().Square()          // a44 = a^(2^42 - 2^20)
1156  	a44.Square().Square()                                     // a44 = a^(2^44 - 2^22)
1157  	a44.Mul(&a22)                                             // a44 = a^(2^44 - 1)
1158  	a88.SquareVal(&a44).Square().Square().Square().Square()   // a88 = a^(2^49 - 2^5)
1159  	a88.Square().Square().Square().Square().Square()          // a88 = a^(2^54 - 2^10)
1160  	a88.Square().Square().Square().Square().Square()          // a88 = a^(2^59 - 2^15)
1161  	a88.Square().Square().Square().Square().Square()          // a88 = a^(2^64 - 2^20)
1162  	a88.Square().Square().Square().Square().Square()          // a88 = a^(2^69 - 2^25)
1163  	a88.Square().Square().Square().Square().Square()          // a88 = a^(2^74 - 2^30)
1164  	a88.Square().Square().Square().Square().Square()          // a88 = a^(2^79 - 2^35)
1165  	a88.Square().Square().Square().Square().Square()          // a88 = a^(2^84 - 2^40)
1166  	a88.Square().Square().Square().Square()                   // a88 = a^(2^88 - 2^44)
1167  	a88.Mul(&a44)                                             // a88 = a^(2^88 - 1)
1168  	a176.SquareVal(&a88).Square().Square().Square().Square()  // a176 = a^(2^93 - 2^5)
1169  	a176.Square().Square().Square().Square().Square()         // a176 = a^(2^98 - 2^10)
1170  	a176.Square().Square().Square().Square().Square()         // a176 = a^(2^103 - 2^15)
1171  	a176.Square().Square().Square().Square().Square()         // a176 = a^(2^108 - 2^20)
1172  	a176.Square().Square().Square().Square().Square()         // a176 = a^(2^113 - 2^25)
1173  	a176.Square().Square().Square().Square().Square()         // a176 = a^(2^118 - 2^30)
1174  	a176.Square().Square().Square().Square().Square()         // a176 = a^(2^123 - 2^35)
1175  	a176.Square().Square().Square().Square().Square()         // a176 = a^(2^128 - 2^40)
1176  	a176.Square().Square().Square().Square().Square()         // a176 = a^(2^133 - 2^45)
1177  	a176.Square().Square().Square().Square().Square()         // a176 = a^(2^138 - 2^50)
1178  	a176.Square().Square().Square().Square().Square()         // a176 = a^(2^143 - 2^55)
1179  	a176.Square().Square().Square().Square().Square()         // a176 = a^(2^148 - 2^60)
1180  	a176.Square().Square().Square().Square().Square()         // a176 = a^(2^153 - 2^65)
1181  	a176.Square().Square().Square().Square().Square()         // a176 = a^(2^158 - 2^70)
1182  	a176.Square().Square().Square().Square().Square()         // a176 = a^(2^163 - 2^75)
1183  	a176.Square().Square().Square().Square().Square()         // a176 = a^(2^168 - 2^80)
1184  	a176.Square().Square().Square().Square().Square()         // a176 = a^(2^173 - 2^85)
1185  	a176.Square().Square().Square()                           // a176 = a^(2^176 - 2^88)
1186  	a176.Mul(&a88)                                            // a176 = a^(2^176 - 1)
1187  	a220.SquareVal(&a176).Square().Square().Square().Square() // a220 = a^(2^181 - 2^5)
1188  	a220.Square().Square().Square().Square().Square()         // a220 = a^(2^186 - 2^10)
1189  	a220.Square().Square().Square().Square().Square()         // a220 = a^(2^191 - 2^15)
1190  	a220.Square().Square().Square().Square().Square()         // a220 = a^(2^196 - 2^20)
1191  	a220.Square().Square().Square().Square().Square()         // a220 = a^(2^201 - 2^25)
1192  	a220.Square().Square().Square().Square().Square()         // a220 = a^(2^206 - 2^30)
1193  	a220.Square().Square().Square().Square().Square()         // a220 = a^(2^211 - 2^35)
1194  	a220.Square().Square().Square().Square().Square()         // a220 = a^(2^216 - 2^40)
1195  	a220.Square().Square().Square().Square()                  // a220 = a^(2^220 - 2^44)
1196  	a220.Mul(&a44)                                            // a220 = a^(2^220 - 1)
1197  	a223.SquareVal(&a220).Square().Square()                   // a223 = a^(2^223 - 2^3)
1198  	a223.Mul(&a3)                                             // a223 = a^(2^223 - 1)
1199  
1200  	f.SquareVal(&a223).Square().Square().Square().Square() // f = a^(2^228 - 2^5)
1201  	f.Square().Square().Square().Square().Square()         // f = a^(2^233 - 2^10)
1202  	f.Square().Square().Square().Square().Square()         // f = a^(2^238 - 2^15)
1203  	f.Square().Square().Square().Square().Square()         // f = a^(2^243 - 2^20)
1204  	f.Square().Square().Square()                           // f = a^(2^246 - 2^23)
1205  	f.Mul(&a22)                                            // f = a^(2^246 - 2^22 - 1)
1206  	f.Square().Square().Square().Square().Square()         // f = a^(2^251 - 2^27 - 2^5)
1207  	f.Square()                                             // f = a^(2^252 - 2^28 - 2^6)
1208  	f.Mul(&a2)                                             // f = a^(2^252 - 2^28 - 2^6 - 2^1 - 1)
1209  	f.Square().Square()                                    // f = a^(2^254 - 2^30 - 2^8 - 2^3 - 2^2)
1210  	//                                                     //   = a^(2^254 - 2^30 - 244)
1211  	//                                                     //   = a^((p+1)/4)
1212  	// Ensure the calculated result is actually the square root by squaring it
1213  	// and checking against the original value.
1214  	var sqr FieldVal
1215  	return sqr.SquareVal(f).Normalize().Equals(val.Normalize())
1216  }
1217  
1218  // Square squares the field value in constant time.  The existing field value is
1219  // modified.  Note that this function can overflow if multiplying any of the
1220  // individual words exceeds a max uint32.  In practice, this means the magnitude
1221  // of the field must be a max of 8 to prevent overflow.
1222  //
1223  // The field value is returned to support chaining.  This enables syntax like:
1224  // f.Square().Mul(f2) so that f = f^2 * f2.
1225  //
1226  //	Preconditions:
1227  //	  - The field value MUST have a max magnitude of 8
1228  //	Output Normalized: No
1229  //	Output Max Magnitude: 1
1230  func (f *FieldVal) Square() *FieldVal { return f.SquareVal(f) }
1231  
1232  // SquareVal squares the passed value and stores the result in f in constant
1233  // time.  Note that this function can overflow if multiplying any of the
1234  // individual words exceeds a max uint32.  In practice, this means the magnitude
1235  // of the field being squared must be a max of 8 to prevent overflow.
1236  //
1237  // The field value is returned to support chaining.  This enables syntax like:
1238  // f3.SquareVal(f).Mul(f) so that f3 = f^2 * f = f^3.
1239  //
1240  //	Preconditions:
1241  //	  - The input field value MUST have a max magnitude of 8
1242  //	Output Normalized: No
1243  //	Output Max Magnitude: 1
1244  func (f *FieldVal) SquareVal(val *FieldVal) *FieldVal {
1245  	// This could be done with a couple of for loops and an array to store
1246  	// the intermediate terms, but this unrolled version is significantly
1247  	// faster.
1248  	//
1249  	// Terms for 2^(fieldBase*0).
1250  	m := uint64(val.n[0]) * uint64(val.n[0])
1251  	t0 := m & fieldBaseMask
1252  	// Terms for 2^(fieldBase*1).
1253  	m = (m >> fieldBase) + 2*uint64(val.n[0])*uint64(val.n[1])
1254  	t1 := m & fieldBaseMask
1255  	// Terms for 2^(fieldBase*2).
1256  	m = (m >> fieldBase) +
1257  		2*uint64(val.n[0])*uint64(val.n[2]) +
1258  		uint64(val.n[1])*uint64(val.n[1])
1259  	t2 := m & fieldBaseMask
1260  	// Terms for 2^(fieldBase*3).
1261  	m = (m >> fieldBase) +
1262  		2*uint64(val.n[0])*uint64(val.n[3]) +
1263  		2*uint64(val.n[1])*uint64(val.n[2])
1264  	t3 := m & fieldBaseMask
1265  	// Terms for 2^(fieldBase*4).
1266  	m = (m >> fieldBase) +
1267  		2*uint64(val.n[0])*uint64(val.n[4]) +
1268  		2*uint64(val.n[1])*uint64(val.n[3]) +
1269  		uint64(val.n[2])*uint64(val.n[2])
1270  	t4 := m & fieldBaseMask
1271  	// Terms for 2^(fieldBase*5).
1272  	m = (m >> fieldBase) +
1273  		2*uint64(val.n[0])*uint64(val.n[5]) +
1274  		2*uint64(val.n[1])*uint64(val.n[4]) +
1275  		2*uint64(val.n[2])*uint64(val.n[3])
1276  	t5 := m & fieldBaseMask
1277  	// Terms for 2^(fieldBase*6).
1278  	m = (m >> fieldBase) +
1279  		2*uint64(val.n[0])*uint64(val.n[6]) +
1280  		2*uint64(val.n[1])*uint64(val.n[5]) +
1281  		2*uint64(val.n[2])*uint64(val.n[4]) +
1282  		uint64(val.n[3])*uint64(val.n[3])
1283  	t6 := m & fieldBaseMask
1284  	// Terms for 2^(fieldBase*7).
1285  	m = (m >> fieldBase) +
1286  		2*uint64(val.n[0])*uint64(val.n[7]) +
1287  		2*uint64(val.n[1])*uint64(val.n[6]) +
1288  		2*uint64(val.n[2])*uint64(val.n[5]) +
1289  		2*uint64(val.n[3])*uint64(val.n[4])
1290  	t7 := m & fieldBaseMask
1291  	// Terms for 2^(fieldBase*8).
1292  	m = (m >> fieldBase) +
1293  		2*uint64(val.n[0])*uint64(val.n[8]) +
1294  		2*uint64(val.n[1])*uint64(val.n[7]) +
1295  		2*uint64(val.n[2])*uint64(val.n[6]) +
1296  		2*uint64(val.n[3])*uint64(val.n[5]) +
1297  		uint64(val.n[4])*uint64(val.n[4])
1298  	t8 := m & fieldBaseMask
1299  	// Terms for 2^(fieldBase*9).
1300  	m = (m >> fieldBase) +
1301  		2*uint64(val.n[0])*uint64(val.n[9]) +
1302  		2*uint64(val.n[1])*uint64(val.n[8]) +
1303  		2*uint64(val.n[2])*uint64(val.n[7]) +
1304  		2*uint64(val.n[3])*uint64(val.n[6]) +
1305  		2*uint64(val.n[4])*uint64(val.n[5])
1306  	t9 := m & fieldBaseMask
1307  	// Terms for 2^(fieldBase*10).
1308  	m = (m >> fieldBase) +
1309  		2*uint64(val.n[1])*uint64(val.n[9]) +
1310  		2*uint64(val.n[2])*uint64(val.n[8]) +
1311  		2*uint64(val.n[3])*uint64(val.n[7]) +
1312  		2*uint64(val.n[4])*uint64(val.n[6]) +
1313  		uint64(val.n[5])*uint64(val.n[5])
1314  	t10 := m & fieldBaseMask
1315  	// Terms for 2^(fieldBase*11).
1316  	m = (m >> fieldBase) +
1317  		2*uint64(val.n[2])*uint64(val.n[9]) +
1318  		2*uint64(val.n[3])*uint64(val.n[8]) +
1319  		2*uint64(val.n[4])*uint64(val.n[7]) +
1320  		2*uint64(val.n[5])*uint64(val.n[6])
1321  	t11 := m & fieldBaseMask
1322  	// Terms for 2^(fieldBase*12).
1323  	m = (m >> fieldBase) +
1324  		2*uint64(val.n[3])*uint64(val.n[9]) +
1325  		2*uint64(val.n[4])*uint64(val.n[8]) +
1326  		2*uint64(val.n[5])*uint64(val.n[7]) +
1327  		uint64(val.n[6])*uint64(val.n[6])
1328  	t12 := m & fieldBaseMask
1329  	// Terms for 2^(fieldBase*13).
1330  	m = (m >> fieldBase) +
1331  		2*uint64(val.n[4])*uint64(val.n[9]) +
1332  		2*uint64(val.n[5])*uint64(val.n[8]) +
1333  		2*uint64(val.n[6])*uint64(val.n[7])
1334  	t13 := m & fieldBaseMask
1335  	// Terms for 2^(fieldBase*14).
1336  	m = (m >> fieldBase) +
1337  		2*uint64(val.n[5])*uint64(val.n[9]) +
1338  		2*uint64(val.n[6])*uint64(val.n[8]) +
1339  		uint64(val.n[7])*uint64(val.n[7])
1340  	t14 := m & fieldBaseMask
1341  	// Terms for 2^(fieldBase*15).
1342  	m = (m >> fieldBase) +
1343  		2*uint64(val.n[6])*uint64(val.n[9]) +
1344  		2*uint64(val.n[7])*uint64(val.n[8])
1345  	t15 := m & fieldBaseMask
1346  	// Terms for 2^(fieldBase*16).
1347  	m = (m >> fieldBase) +
1348  		2*uint64(val.n[7])*uint64(val.n[9]) +
1349  		uint64(val.n[8])*uint64(val.n[8])
1350  	t16 := m & fieldBaseMask
1351  	// Terms for 2^(fieldBase*17).
1352  	m = (m >> fieldBase) + 2*uint64(val.n[8])*uint64(val.n[9])
1353  	t17 := m & fieldBaseMask
1354  	// Terms for 2^(fieldBase*18).
1355  	m = (m >> fieldBase) + uint64(val.n[9])*uint64(val.n[9])
1356  	t18 := m & fieldBaseMask
1357  	// What's left is for 2^(fieldBase*19).
1358  	t19 := m >> fieldBase
1359  	// At this point, all of the terms are grouped into their respective
1360  	// base.
1361  	//
1362  	// Per [HAC] section 14.3.4: Reduction method of moduli of special form,
1363  	// when the modulus is of the special form m = b^t - c, highly efficient
1364  	// reduction can be achieved per the provided algorithm.
1365  	//
1366  	// The secp256k1 prime is equivalent to 2^256 - 4294968273, so it fits
1367  	// this criteria.
1368  	//
1369  	// 4294968273 in field representation (base 2^26) is:
1370  	// n[0] = 977
1371  	// n[1] = 64
1372  	// That is to say (2^26 * 64) + 977 = 4294968273
1373  	//
1374  	// Since each word is in base 26, the upper terms (t10 and up) start
1375  	// at 260 bits (versus the final desired range of 256 bits), so the
1376  	// field representation of 'c' from above needs to be adjusted for the
1377  	// extra 4 bits by multiplying it by 2^4 = 16.  4294968273 * 16 =
1378  	// 68719492368.  Thus, the adjusted field representation of 'c' is:
1379  	// n[0] = 977 * 16 = 15632
1380  	// n[1] = 64 * 16 = 1024
1381  	// That is to say (2^26 * 1024) + 15632 = 68719492368
1382  	//
1383  	// To reduce the final term, t19, the entire 'c' value is needed instead
1384  	// of only n[0] because there are no more terms left to handle n[1].
1385  	// This means there might be some magnitude left in the upper bits that
1386  	// is handled below.
1387  	m = t0 + t10*15632
1388  	t0 = m & fieldBaseMask
1389  	m = (m >> fieldBase) + t1 + t10*1024 + t11*15632
1390  	t1 = m & fieldBaseMask
1391  	m = (m >> fieldBase) + t2 + t11*1024 + t12*15632
1392  	t2 = m & fieldBaseMask
1393  	m = (m >> fieldBase) + t3 + t12*1024 + t13*15632
1394  	t3 = m & fieldBaseMask
1395  	m = (m >> fieldBase) + t4 + t13*1024 + t14*15632
1396  	t4 = m & fieldBaseMask
1397  	m = (m >> fieldBase) + t5 + t14*1024 + t15*15632
1398  	t5 = m & fieldBaseMask
1399  	m = (m >> fieldBase) + t6 + t15*1024 + t16*15632
1400  	t6 = m & fieldBaseMask
1401  	m = (m >> fieldBase) + t7 + t16*1024 + t17*15632
1402  	t7 = m & fieldBaseMask
1403  	m = (m >> fieldBase) + t8 + t17*1024 + t18*15632
1404  	t8 = m & fieldBaseMask
1405  	m = (m >> fieldBase) + t9 + t18*1024 + t19*68719492368
1406  	t9 = m & fieldMSBMask
1407  	m = m >> fieldMSBBits
1408  	// At this point, if the magnitude is greater than 0, the overall value
1409  	// is greater than the max possible 256-bit value.  In particular, it is
1410  	// "how many times larger" than the max value it is.
1411  	//
1412  	// The algorithm presented in [HAC] section 14.3.4 repeats until the
1413  	// quotient is zero.  However, due to the above, we already know at
1414  	// least how many times we would need to repeat as it's the value
1415  	// currently in m.  Thus we can simply multiply the magnitude by the
1416  	// field representation of the prime and do a single iteration.  Notice
1417  	// that nothing will be changed when the magnitude is zero, so we could
1418  	// skip this in that case, however always running regardless allows it
1419  	// to run in constant time.  The final result will be in the range
1420  	// 0 <= result <= prime + (2^64 - c), so it is guaranteed to have a
1421  	// magnitude of 1, but it is denormalized.
1422  	n := t0 + m*977
1423  	f.n[0] = uint32(n & fieldBaseMask)
1424  	n = (n >> fieldBase) + t1 + m*64
1425  	f.n[1] = uint32(n & fieldBaseMask)
1426  	f.n[2] = uint32((n >> fieldBase) + t2)
1427  	f.n[3] = uint32(t3)
1428  	f.n[4] = uint32(t4)
1429  	f.n[5] = uint32(t5)
1430  	f.n[6] = uint32(t6)
1431  	f.n[7] = uint32(t7)
1432  	f.n[8] = uint32(t8)
1433  	f.n[9] = uint32(t9)
1434  	return f
1435  }
1436  
1437  // Inverse finds the modular multiplicative inverse of the field value in
1438  // constant time.  The existing field value is modified.
1439  //
1440  // The field value is returned to support chaining.  This enables syntax like:
1441  // f.Inverse().Mul(f2) so that f = f^-1 * f2.
1442  //
1443  //	Preconditions:
1444  //	  - The field value MUST have a max magnitude of 8
1445  //	Output Normalized: No
1446  //	Output Max Magnitude: 1
1447  func (f *FieldVal) Inverse() *FieldVal {
1448  	// Fermat's little theorem states that for a nonzero number a and prime
1449  	// p, a^(p-1) ≡ 1 (mod p).  Since the multiplicative inverse is
1450  	// a*b ≡ 1 (mod p), it follows that b ≡ a*a^(p-2) ≡ a^(p-1) ≡ 1 (mod p).
1451  	// Thus, a^(p-2) is the multiplicative inverse.
1452  	//
1453  	// In order to efficiently compute a^(p-2), p-2 needs to be split into
1454  	// a sequence of squares and multiplications that minimizes the number
1455  	// of multiplications needed (since they are more costly than
1456  	// squarings). Intermediate results are saved and reused as well.
1457  	//
1458  	// The secp256k1 prime - 2 is 2^256 - 4294968275.
1459  	//
1460  	// This has a cost of 258 field squarings and 33 field multiplications.
1461  	var a2, a3, a4, a10, a11, a21, a42, a45, a63, a1019, a1023 FieldVal
1462  	a2.SquareVal(f)
1463  	a3.Mul2(&a2, f)
1464  	a4.SquareVal(&a2)
1465  	a10.SquareVal(&a4).Mul(&a2)
1466  	a11.Mul2(&a10, f)
1467  	a21.Mul2(&a10, &a11)
1468  	a42.SquareVal(&a21)
1469  	a45.Mul2(&a42, &a3)
1470  	a63.Mul2(&a42, &a21)
1471  	a1019.SquareVal(&a63).Square().Square().Square().Mul(&a11)
1472  	a1023.Mul2(&a1019, &a4)
1473  	f.Set(&a63)                                    // f = a^(2^6 - 1)
1474  	f.Square().Square().Square().Square().Square() // f = a^(2^11 - 32)
1475  	f.Square().Square().Square().Square().Square() // f = a^(2^16 - 1024)
1476  	f.Mul(&a1023)                                  // f = a^(2^16 - 1)
1477  	f.Square().Square().Square().Square().Square() // f = a^(2^21 - 32)
1478  	f.Square().Square().Square().Square().Square() // f = a^(2^26 - 1024)
1479  	f.Mul(&a1023)                                  // f = a^(2^26 - 1)
1480  	f.Square().Square().Square().Square().Square() // f = a^(2^31 - 32)
1481  	f.Square().Square().Square().Square().Square() // f = a^(2^36 - 1024)
1482  	f.Mul(&a1023)                                  // f = a^(2^36 - 1)
1483  	f.Square().Square().Square().Square().Square() // f = a^(2^41 - 32)
1484  	f.Square().Square().Square().Square().Square() // f = a^(2^46 - 1024)
1485  	f.Mul(&a1023)                                  // f = a^(2^46 - 1)
1486  	f.Square().Square().Square().Square().Square() // f = a^(2^51 - 32)
1487  	f.Square().Square().Square().Square().Square() // f = a^(2^56 - 1024)
1488  	f.Mul(&a1023)                                  // f = a^(2^56 - 1)
1489  	f.Square().Square().Square().Square().Square() // f = a^(2^61 - 32)
1490  	f.Square().Square().Square().Square().Square() // f = a^(2^66 - 1024)
1491  	f.Mul(&a1023)                                  // f = a^(2^66 - 1)
1492  	f.Square().Square().Square().Square().Square() // f = a^(2^71 - 32)
1493  	f.Square().Square().Square().Square().Square() // f = a^(2^76 - 1024)
1494  	f.Mul(&a1023)                                  // f = a^(2^76 - 1)
1495  	f.Square().Square().Square().Square().Square() // f = a^(2^81 - 32)
1496  	f.Square().Square().Square().Square().Square() // f = a^(2^86 - 1024)
1497  	f.Mul(&a1023)                                  // f = a^(2^86 - 1)
1498  	f.Square().Square().Square().Square().Square() // f = a^(2^91 - 32)
1499  	f.Square().Square().Square().Square().Square() // f = a^(2^96 - 1024)
1500  	f.Mul(&a1023)                                  // f = a^(2^96 - 1)
1501  	f.Square().Square().Square().Square().Square() // f = a^(2^101 - 32)
1502  	f.Square().Square().Square().Square().Square() // f = a^(2^106 - 1024)
1503  	f.Mul(&a1023)                                  // f = a^(2^106 - 1)
1504  	f.Square().Square().Square().Square().Square() // f = a^(2^111 - 32)
1505  	f.Square().Square().Square().Square().Square() // f = a^(2^116 - 1024)
1506  	f.Mul(&a1023)                                  // f = a^(2^116 - 1)
1507  	f.Square().Square().Square().Square().Square() // f = a^(2^121 - 32)
1508  	f.Square().Square().Square().Square().Square() // f = a^(2^126 - 1024)
1509  	f.Mul(&a1023)                                  // f = a^(2^126 - 1)
1510  	f.Square().Square().Square().Square().Square() // f = a^(2^131 - 32)
1511  	f.Square().Square().Square().Square().Square() // f = a^(2^136 - 1024)
1512  	f.Mul(&a1023)                                  // f = a^(2^136 - 1)
1513  	f.Square().Square().Square().Square().Square() // f = a^(2^141 - 32)
1514  	f.Square().Square().Square().Square().Square() // f = a^(2^146 - 1024)
1515  	f.Mul(&a1023)                                  // f = a^(2^146 - 1)
1516  	f.Square().Square().Square().Square().Square() // f = a^(2^151 - 32)
1517  	f.Square().Square().Square().Square().Square() // f = a^(2^156 - 1024)
1518  	f.Mul(&a1023)                                  // f = a^(2^156 - 1)
1519  	f.Square().Square().Square().Square().Square() // f = a^(2^161 - 32)
1520  	f.Square().Square().Square().Square().Square() // f = a^(2^166 - 1024)
1521  	f.Mul(&a1023)                                  // f = a^(2^166 - 1)
1522  	f.Square().Square().Square().Square().Square() // f = a^(2^171 - 32)
1523  	f.Square().Square().Square().Square().Square() // f = a^(2^176 - 1024)
1524  	f.Mul(&a1023)                                  // f = a^(2^176 - 1)
1525  	f.Square().Square().Square().Square().Square() // f = a^(2^181 - 32)
1526  	f.Square().Square().Square().Square().Square() // f = a^(2^186 - 1024)
1527  	f.Mul(&a1023)                                  // f = a^(2^186 - 1)
1528  	f.Square().Square().Square().Square().Square() // f = a^(2^191 - 32)
1529  	f.Square().Square().Square().Square().Square() // f = a^(2^196 - 1024)
1530  	f.Mul(&a1023)                                  // f = a^(2^196 - 1)
1531  	f.Square().Square().Square().Square().Square() // f = a^(2^201 - 32)
1532  	f.Square().Square().Square().Square().Square() // f = a^(2^206 - 1024)
1533  	f.Mul(&a1023)                                  // f = a^(2^206 - 1)
1534  	f.Square().Square().Square().Square().Square() // f = a^(2^211 - 32)
1535  	f.Square().Square().Square().Square().Square() // f = a^(2^216 - 1024)
1536  	f.Mul(&a1023)                                  // f = a^(2^216 - 1)
1537  	f.Square().Square().Square().Square().Square() // f = a^(2^221 - 32)
1538  	f.Square().Square().Square().Square().Square() // f = a^(2^226 - 1024)
1539  	f.Mul(&a1019)                                  // f = a^(2^226 - 5)
1540  	f.Square().Square().Square().Square().Square() // f = a^(2^231 - 160)
1541  	f.Square().Square().Square().Square().Square() // f = a^(2^236 - 5120)
1542  	f.Mul(&a1023)                                  // f = a^(2^236 - 4097)
1543  	f.Square().Square().Square().Square().Square() // f = a^(2^241 - 131104)
1544  	f.Square().Square().Square().Square().Square() // f = a^(2^246 - 4195328)
1545  	f.Mul(&a1023)                                  // f = a^(2^246 - 4194305)
1546  	f.Square().Square().Square().Square().Square() // f = a^(2^251 - 134217760)
1547  	f.Square().Square().Square().Square().Square() // f = a^(2^256 - 4294968320)
1548  	return f.Mul(&a45)                             // f = a^(2^256 - 4294968275) = a^(p-2)
1549  }
1550  
1551  // IsGtOrEqPrimeMinusOrder returns whether the field value exceeds the
1552  // group order divided by 2 in constant time.
1553  //
1554  //	Preconditions:
1555  //	  - The field value MUST be normalized
1556  func (f *FieldVal) IsGtOrEqPrimeMinusOrder() bool {
1557  	// The secp256k1 prime is equivalent to 2^256 - 4294968273 and the group
1558  	// order is 2^256 - 432420386565659656852420866394968145599.  Thus,
1559  	// the prime minus the group order is:
1560  	// 432420386565659656852420866390673177326
1561  	//
1562  	// In hex that is:
1563  	// 0x00000000 00000000 00000000 00000001 45512319 50b75fc4 402da172 2fc9baee
1564  	//
1565  	// Converting that to field representation (base 2^26) is:
1566  	//
1567  	// n[0] = 0x03c9baee
1568  	// n[1] = 0x03685c8b
1569  	// n[2] = 0x01fc4402
1570  	// n[3] = 0x006542dd
1571  	// n[4] = 0x01455123
1572  	//
1573  	// This can be verified with the following test code:
1574  	//   pMinusN := new(big.Int).Sub(curveParams.P, curveParams.N)
1575  	//   var fv FieldVal
1576  	//   fv.SetByteSlice(pMinusN.Bytes())
1577  	//   t.Logf("%x", fv.n)
1578  	//
1579  	//   Outputs: [3c9baee 3685c8b 1fc4402 6542dd 1455123 0 0 0 0 0]
1580  	const (
1581  		pMinusNWordZero  = 0x03c9baee
1582  		pMinusNWordOne   = 0x03685c8b
1583  		pMinusNWordTwo   = 0x01fc4402
1584  		pMinusNWordThree = 0x006542dd
1585  		pMinusNWordFour  = 0x01455123
1586  		pMinusNWordFive  = 0x00000000
1587  		pMinusNWordSix   = 0x00000000
1588  		pMinusNWordSeven = 0x00000000
1589  		pMinusNWordEight = 0x00000000
1590  		pMinusNWordNine  = 0x00000000
1591  	)
1592  	// The intuition here is that the value is greater than field prime minus
1593  	// the group order if one of the higher individual words is greater than the
1594  	// corresponding word and all higher words in the value are equal.
1595  	result := constantTimeGreater(f.n[9], pMinusNWordNine)
1596  	highWordsEqual := constantTimeEq(f.n[9], pMinusNWordNine)
1597  	result |= highWordsEqual & constantTimeGreater(f.n[8], pMinusNWordEight)
1598  	highWordsEqual &= constantTimeEq(f.n[8], pMinusNWordEight)
1599  	result |= highWordsEqual & constantTimeGreater(f.n[7], pMinusNWordSeven)
1600  	highWordsEqual &= constantTimeEq(f.n[7], pMinusNWordSeven)
1601  	result |= highWordsEqual & constantTimeGreater(f.n[6], pMinusNWordSix)
1602  	highWordsEqual &= constantTimeEq(f.n[6], pMinusNWordSix)
1603  	result |= highWordsEqual & constantTimeGreater(f.n[5], pMinusNWordFive)
1604  	highWordsEqual &= constantTimeEq(f.n[5], pMinusNWordFive)
1605  	result |= highWordsEqual & constantTimeGreater(f.n[4], pMinusNWordFour)
1606  	highWordsEqual &= constantTimeEq(f.n[4], pMinusNWordFour)
1607  	result |= highWordsEqual & constantTimeGreater(f.n[3], pMinusNWordThree)
1608  	highWordsEqual &= constantTimeEq(f.n[3], pMinusNWordThree)
1609  	result |= highWordsEqual & constantTimeGreater(f.n[2], pMinusNWordTwo)
1610  	highWordsEqual &= constantTimeEq(f.n[2], pMinusNWordTwo)
1611  	result |= highWordsEqual & constantTimeGreater(f.n[1], pMinusNWordOne)
1612  	highWordsEqual &= constantTimeEq(f.n[1], pMinusNWordOne)
1613  	result |= highWordsEqual & constantTimeGreaterOrEq(f.n[0], pMinusNWordZero)
1614  	return result != 0
1615  }
1616