| // Copyright 2024 The Go Authors. All rights reserved. |
| // Use of this source code is governed by a BSD-style |
| // license that can be found in the LICENSE file. |
| |
| package mlkem |
| |
| import ( |
| "crypto/internal/fips140/sha3" |
| "crypto/internal/fips140deps/byteorder" |
| "errors" |
| ) |
| |
| // fieldElement is an integer modulo q, an element of ℤ_q. It is always reduced. |
| type fieldElement uint16 |
| |
| // fieldCheckReduced checks that a value a is < q. |
| func fieldCheckReduced(a uint16) (fieldElement, error) { |
| if a >= q { |
| return 0, errors.New("unreduced field element") |
| } |
| return fieldElement(a), nil |
| } |
| |
| // fieldReduceOnce reduces a value a < 2q. |
| func fieldReduceOnce(a uint16) fieldElement { |
| x := a - q |
| // If x underflowed, then x >= 2¹⁶ - q > 2¹⁵, so the top bit is set. |
| x += (x >> 15) * q |
| return fieldElement(x) |
| } |
| |
| func fieldAdd(a, b fieldElement) fieldElement { |
| x := uint16(a + b) |
| return fieldReduceOnce(x) |
| } |
| |
| func fieldSub(a, b fieldElement) fieldElement { |
| x := uint16(a - b + q) |
| return fieldReduceOnce(x) |
| } |
| |
| const ( |
| barrettMultiplier = 5039 // 2¹² * 2¹² / q |
| barrettShift = 24 // log₂(2¹² * 2¹²) |
| ) |
| |
| // fieldReduce reduces a value a < 2q² using Barrett reduction, to avoid |
| // potentially variable-time division. |
| func fieldReduce(a uint32) fieldElement { |
| quotient := uint32((uint64(a) * barrettMultiplier) >> barrettShift) |
| return fieldReduceOnce(uint16(a - quotient*q)) |
| } |
| |
| func fieldMul(a, b fieldElement) fieldElement { |
| x := uint32(a) * uint32(b) |
| return fieldReduce(x) |
| } |
| |
| // fieldMulSub returns a * (b - c). This operation is fused to save a |
| // fieldReduceOnce after the subtraction. |
| func fieldMulSub(a, b, c fieldElement) fieldElement { |
| x := uint32(a) * uint32(b-c+q) |
| return fieldReduce(x) |
| } |
| |
| // fieldAddMul returns a * b + c * d. This operation is fused to save a |
| // fieldReduceOnce and a fieldReduce. |
| func fieldAddMul(a, b, c, d fieldElement) fieldElement { |
| x := uint32(a) * uint32(b) |
| x += uint32(c) * uint32(d) |
| return fieldReduce(x) |
| } |
| |
| // compress maps a field element uniformly to the range 0 to 2ᵈ-1, according to |
| // FIPS 203, Definition 4.7. |
| func compress(x fieldElement, d uint8) uint16 { |
| // We want to compute (x * 2ᵈ) / q, rounded to nearest integer, with 1/2 |
| // rounding up (see FIPS 203, Section 2.3). |
| |
| // Barrett reduction produces a quotient and a remainder in the range [0, 2q), |
| // such that dividend = quotient * q + remainder. |
| dividend := uint32(x) << d // x * 2ᵈ |
| quotient := uint32(uint64(dividend) * barrettMultiplier >> barrettShift) |
| remainder := dividend - quotient*q |
| |
| // Since the remainder is in the range [0, 2q), not [0, q), we need to |
| // portion it into three spans for rounding. |
| // |
| // [ 0, q/2 ) -> round to 0 |
| // [ q/2, q + q/2 ) -> round to 1 |
| // [ q + q/2, 2q ) -> round to 2 |
| // |
| // We can convert that to the following logic: add 1 if remainder > q/2, |
| // then add 1 again if remainder > q + q/2. |
| // |
| // Note that if remainder > x, then ⌊x⌋ - remainder underflows, and the top |
| // bit of the difference will be set. |
| quotient += (q/2 - remainder) >> 31 & 1 |
| quotient += (q + q/2 - remainder) >> 31 & 1 |
| |
| // quotient might have overflowed at this point, so reduce it by masking. |
| var mask uint32 = (1 << d) - 1 |
| return uint16(quotient & mask) |
| } |
| |
| // decompress maps a number x between 0 and 2ᵈ-1 uniformly to the full range of |
| // field elements, according to FIPS 203, Definition 4.8. |
| func decompress(y uint16, d uint8) fieldElement { |
| // We want to compute (y * q) / 2ᵈ, rounded to nearest integer, with 1/2 |
| // rounding up (see FIPS 203, Section 2.3). |
| |
| dividend := uint32(y) * q |
| quotient := dividend >> d // (y * q) / 2ᵈ |
| |
| // The d'th least-significant bit of the dividend (the most significant bit |
| // of the remainder) is 1 for the top half of the values that divide to the |
| // same quotient, which are the ones that round up. |
| quotient += dividend >> (d - 1) & 1 |
| |
| // quotient is at most (2¹¹-1) * q / 2¹¹ + 1 = 3328, so it didn't overflow. |
| return fieldElement(quotient) |
| } |
| |
| // ringElement is a polynomial, an element of R_q, represented as an array |
| // according to FIPS 203, Section 2.4.4. |
| type ringElement [n]fieldElement |
| |
| // polyAdd adds two ringElements or nttElements. |
| func polyAdd[T ~[n]fieldElement](a, b T) (s T) { |
| for i := range s { |
| s[i] = fieldAdd(a[i], b[i]) |
| } |
| return s |
| } |
| |
| // polySub subtracts two ringElements or nttElements. |
| func polySub[T ~[n]fieldElement](a, b T) (s T) { |
| for i := range s { |
| s[i] = fieldSub(a[i], b[i]) |
| } |
| return s |
| } |
| |
| // polyByteEncode appends the 384-byte encoding of f to b. |
| // |
| // It implements ByteEncode₁₂, according to FIPS 203, Algorithm 5. |
| func polyByteEncode[T ~[n]fieldElement](b []byte, f T) []byte { |
| out, B := sliceForAppend(b, encodingSize12) |
| for i := 0; i < n; i += 2 { |
| x := uint32(f[i]) | uint32(f[i+1])<<12 |
| B[0] = uint8(x) |
| B[1] = uint8(x >> 8) |
| B[2] = uint8(x >> 16) |
| B = B[3:] |
| } |
| return out |
| } |
| |
| // polyByteDecode decodes the 384-byte encoding of a polynomial, checking that |
| // all the coefficients are properly reduced. This fulfills the "Modulus check" |
| // step of ML-KEM Encapsulation. |
| // |
| // It implements ByteDecode₁₂, according to FIPS 203, Algorithm 6. |
| func polyByteDecode[T ~[n]fieldElement](b []byte) (T, error) { |
| if len(b) != encodingSize12 { |
| return T{}, errors.New("mlkem: invalid encoding length") |
| } |
| var f T |
| for i := 0; i < n; i += 2 { |
| d := uint32(b[0]) | uint32(b[1])<<8 | uint32(b[2])<<16 |
| const mask12 = 0b1111_1111_1111 |
| var err error |
| if f[i], err = fieldCheckReduced(uint16(d & mask12)); err != nil { |
| return T{}, errors.New("mlkem: invalid polynomial encoding") |
| } |
| if f[i+1], err = fieldCheckReduced(uint16(d >> 12)); err != nil { |
| return T{}, errors.New("mlkem: invalid polynomial encoding") |
| } |
| b = b[3:] |
| } |
| return f, nil |
| } |
| |
| // sliceForAppend takes a slice and a requested number of bytes. It returns a |
| // slice with the contents of the given slice followed by that many bytes and a |
| // second slice that aliases into it and contains only the extra bytes. If the |
| // original slice has sufficient capacity then no allocation is performed. |
| func sliceForAppend(in []byte, n int) (head, tail []byte) { |
| if total := len(in) + n; cap(in) >= total { |
| head = in[:total] |
| } else { |
| head = make([]byte, total) |
| copy(head, in) |
| } |
| tail = head[len(in):] |
| return |
| } |
| |
| // ringCompressAndEncode1 appends a 32-byte encoding of a ring element to s, |
| // compressing one coefficients per bit. |
| // |
| // It implements Compress₁, according to FIPS 203, Definition 4.7, |
| // followed by ByteEncode₁, according to FIPS 203, Algorithm 5. |
| func ringCompressAndEncode1(s []byte, f ringElement) []byte { |
| s, b := sliceForAppend(s, encodingSize1) |
| clear(b) |
| for i := range f { |
| b[i/8] |= uint8(compress(f[i], 1) << (i % 8)) |
| } |
| return s |
| } |
| |
| // ringDecodeAndDecompress1 decodes a 32-byte slice to a ring element where each |
| // bit is mapped to 0 or ⌈q/2⌋. |
| // |
| // It implements ByteDecode₁, according to FIPS 203, Algorithm 6, |
| // followed by Decompress₁, according to FIPS 203, Definition 4.8. |
| func ringDecodeAndDecompress1(b *[encodingSize1]byte) ringElement { |
| var f ringElement |
| for i := range f { |
| b_i := b[i/8] >> (i % 8) & 1 |
| const halfQ = (q + 1) / 2 // ⌈q/2⌋, rounded up per FIPS 203, Section 2.3 |
| f[i] = fieldElement(b_i) * halfQ // 0 decompresses to 0, and 1 to ⌈q/2⌋ |
| } |
| return f |
| } |
| |
| // ringCompressAndEncode4 appends a 128-byte encoding of a ring element to s, |
| // compressing two coefficients per byte. |
| // |
| // It implements Compress₄, according to FIPS 203, Definition 4.7, |
| // followed by ByteEncode₄, according to FIPS 203, Algorithm 5. |
| func ringCompressAndEncode4(s []byte, f ringElement) []byte { |
| s, b := sliceForAppend(s, encodingSize4) |
| for i := 0; i < n; i += 2 { |
| b[i/2] = uint8(compress(f[i], 4) | compress(f[i+1], 4)<<4) |
| } |
| return s |
| } |
| |
| // ringDecodeAndDecompress4 decodes a 128-byte encoding of a ring element where |
| // each four bits are mapped to an equidistant distribution. |
| // |
| // It implements ByteDecode₄, according to FIPS 203, Algorithm 6, |
| // followed by Decompress₄, according to FIPS 203, Definition 4.8. |
| func ringDecodeAndDecompress4(b *[encodingSize4]byte) ringElement { |
| var f ringElement |
| for i := 0; i < n; i += 2 { |
| f[i] = fieldElement(decompress(uint16(b[i/2]&0b1111), 4)) |
| f[i+1] = fieldElement(decompress(uint16(b[i/2]>>4), 4)) |
| } |
| return f |
| } |
| |
| // ringCompressAndEncode10 appends a 320-byte encoding of a ring element to s, |
| // compressing four coefficients per five bytes. |
| // |
| // It implements Compress₁₀, according to FIPS 203, Definition 4.7, |
| // followed by ByteEncode₁₀, according to FIPS 203, Algorithm 5. |
| func ringCompressAndEncode10(s []byte, f ringElement) []byte { |
| s, b := sliceForAppend(s, encodingSize10) |
| for i := 0; i < n; i += 4 { |
| var x uint64 |
| x |= uint64(compress(f[i], 10)) |
| x |= uint64(compress(f[i+1], 10)) << 10 |
| x |= uint64(compress(f[i+2], 10)) << 20 |
| x |= uint64(compress(f[i+3], 10)) << 30 |
| b[0] = uint8(x) |
| b[1] = uint8(x >> 8) |
| b[2] = uint8(x >> 16) |
| b[3] = uint8(x >> 24) |
| b[4] = uint8(x >> 32) |
| b = b[5:] |
| } |
| return s |
| } |
| |
| // ringDecodeAndDecompress10 decodes a 320-byte encoding of a ring element where |
| // each ten bits are mapped to an equidistant distribution. |
| // |
| // It implements ByteDecode₁₀, according to FIPS 203, Algorithm 6, |
| // followed by Decompress₁₀, according to FIPS 203, Definition 4.8. |
| func ringDecodeAndDecompress10(bb *[encodingSize10]byte) ringElement { |
| b := bb[:] |
| var f ringElement |
| for i := 0; i < n; i += 4 { |
| x := uint64(b[0]) | uint64(b[1])<<8 | uint64(b[2])<<16 | uint64(b[3])<<24 | uint64(b[4])<<32 |
| b = b[5:] |
| f[i] = fieldElement(decompress(uint16(x>>0&0b11_1111_1111), 10)) |
| f[i+1] = fieldElement(decompress(uint16(x>>10&0b11_1111_1111), 10)) |
| f[i+2] = fieldElement(decompress(uint16(x>>20&0b11_1111_1111), 10)) |
| f[i+3] = fieldElement(decompress(uint16(x>>30&0b11_1111_1111), 10)) |
| } |
| return f |
| } |
| |
| // ringCompressAndEncode appends an encoding of a ring element to s, |
| // compressing each coefficient to d bits. |
| // |
| // It implements Compress, according to FIPS 203, Definition 4.7, |
| // followed by ByteEncode, according to FIPS 203, Algorithm 5. |
| func ringCompressAndEncode(s []byte, f ringElement, d uint8) []byte { |
| var b byte |
| var bIdx uint8 |
| for i := 0; i < n; i++ { |
| c := compress(f[i], d) |
| var cIdx uint8 |
| for cIdx < d { |
| b |= byte(c>>cIdx) << bIdx |
| bits := min(8-bIdx, d-cIdx) |
| bIdx += bits |
| cIdx += bits |
| if bIdx == 8 { |
| s = append(s, b) |
| b = 0 |
| bIdx = 0 |
| } |
| } |
| } |
| if bIdx != 0 { |
| panic("mlkem: internal error: bitsFilled != 0") |
| } |
| return s |
| } |
| |
| // ringDecodeAndDecompress decodes an encoding of a ring element where |
| // each d bits are mapped to an equidistant distribution. |
| // |
| // It implements ByteDecode, according to FIPS 203, Algorithm 6, |
| // followed by Decompress, according to FIPS 203, Definition 4.8. |
| func ringDecodeAndDecompress(b []byte, d uint8) ringElement { |
| var f ringElement |
| var bIdx uint8 |
| for i := 0; i < n; i++ { |
| var c uint16 |
| var cIdx uint8 |
| for cIdx < d { |
| c |= uint16(b[0]>>bIdx) << cIdx |
| c &= (1 << d) - 1 |
| bits := min(8-bIdx, d-cIdx) |
| bIdx += bits |
| cIdx += bits |
| if bIdx == 8 { |
| b = b[1:] |
| bIdx = 0 |
| } |
| } |
| f[i] = fieldElement(decompress(c, d)) |
| } |
| if len(b) != 0 { |
| panic("mlkem: internal error: leftover bytes") |
| } |
| return f |
| } |
| |
| // ringCompressAndEncode5 appends a 160-byte encoding of a ring element to s, |
| // compressing eight coefficients per five bytes. |
| // |
| // It implements Compress₅, according to FIPS 203, Definition 4.7, |
| // followed by ByteEncode₅, according to FIPS 203, Algorithm 5. |
| func ringCompressAndEncode5(s []byte, f ringElement) []byte { |
| return ringCompressAndEncode(s, f, 5) |
| } |
| |
| // ringDecodeAndDecompress5 decodes a 160-byte encoding of a ring element where |
| // each five bits are mapped to an equidistant distribution. |
| // |
| // It implements ByteDecode₅, according to FIPS 203, Algorithm 6, |
| // followed by Decompress₅, according to FIPS 203, Definition 4.8. |
| func ringDecodeAndDecompress5(bb *[encodingSize5]byte) ringElement { |
| return ringDecodeAndDecompress(bb[:], 5) |
| } |
| |
| // ringCompressAndEncode11 appends a 352-byte encoding of a ring element to s, |
| // compressing eight coefficients per eleven bytes. |
| // |
| // It implements Compress₁₁, according to FIPS 203, Definition 4.7, |
| // followed by ByteEncode₁₁, according to FIPS 203, Algorithm 5. |
| func ringCompressAndEncode11(s []byte, f ringElement) []byte { |
| return ringCompressAndEncode(s, f, 11) |
| } |
| |
| // ringDecodeAndDecompress11 decodes a 352-byte encoding of a ring element where |
| // each eleven bits are mapped to an equidistant distribution. |
| // |
| // It implements ByteDecode₁₁, according to FIPS 203, Algorithm 6, |
| // followed by Decompress₁₁, according to FIPS 203, Definition 4.8. |
| func ringDecodeAndDecompress11(bb *[encodingSize11]byte) ringElement { |
| return ringDecodeAndDecompress(bb[:], 11) |
| } |
| |
| // samplePolyCBD draws a ringElement from the special Dη distribution given a |
| // stream of random bytes generated by the PRF function, according to FIPS 203, |
| // Algorithm 8 and Definition 4.3. |
| func samplePolyCBD(s []byte, b byte) ringElement { |
| prf := sha3.NewShake256() |
| prf.Write(s) |
| prf.Write([]byte{b}) |
| B := make([]byte, 64*2) // η = 2 |
| prf.Read(B) |
| |
| // SamplePolyCBD simply draws four (2η) bits for each coefficient, and adds |
| // the first two and subtracts the last two. |
| |
| var f ringElement |
| for i := 0; i < n; i += 2 { |
| b := B[i/2] |
| b_7, b_6, b_5, b_4 := b>>7, b>>6&1, b>>5&1, b>>4&1 |
| b_3, b_2, b_1, b_0 := b>>3&1, b>>2&1, b>>1&1, b&1 |
| f[i] = fieldSub(fieldElement(b_0+b_1), fieldElement(b_2+b_3)) |
| f[i+1] = fieldSub(fieldElement(b_4+b_5), fieldElement(b_6+b_7)) |
| } |
| return f |
| } |
| |
| // nttElement is an NTT representation, an element of T_q, represented as an |
| // array according to FIPS 203, Section 2.4.4. |
| type nttElement [n]fieldElement |
| |
| // gammas are the values ζ^2BitRev7(i)+1 mod q for each index i, according to |
| // FIPS 203, Appendix A (with negative values reduced to positive). |
| var gammas = [128]fieldElement{17, 3312, 2761, 568, 583, 2746, 2649, 680, 1637, 1692, 723, 2606, 2288, 1041, 1100, 2229, 1409, 1920, 2662, 667, 3281, 48, 233, 3096, 756, 2573, 2156, 1173, 3015, 314, 3050, 279, 1703, 1626, 1651, 1678, 2789, 540, 1789, 1540, 1847, 1482, 952, 2377, 1461, 1868, 2687, 642, 939, 2390, 2308, 1021, 2437, 892, 2388, 941, 733, 2596, 2337, 992, 268, 3061, 641, 2688, 1584, 1745, 2298, 1031, 2037, 1292, 3220, 109, 375, 2954, 2549, 780, 2090, 1239, 1645, 1684, 1063, 2266, 319, 3010, 2773, 556, 757, 2572, 2099, 1230, 561, 2768, 2466, 863, 2594, 735, 2804, 525, 1092, 2237, 403, 2926, 1026, 2303, 1143, 2186, 2150, 1179, 2775, 554, 886, 2443, 1722, 1607, 1212, 2117, 1874, 1455, 1029, 2300, 2110, 1219, 2935, 394, 885, 2444, 2154, 1175} |
| |
| // nttMul multiplies two nttElements. |
| // |
| // It implements MultiplyNTTs, according to FIPS 203, Algorithm 11. |
| func nttMul(f, g nttElement) nttElement { |
| var h nttElement |
| // We use i += 2 for bounds check elimination. See https://go.dev/issue/66826. |
| for i := 0; i < 256; i += 2 { |
| a0, a1 := f[i], f[i+1] |
| b0, b1 := g[i], g[i+1] |
| h[i] = fieldAddMul(a0, b0, fieldMul(a1, b1), gammas[i/2]) |
| h[i+1] = fieldAddMul(a0, b1, a1, b0) |
| } |
| return h |
| } |
| |
| // zetas are the values ζ^BitRev7(k) mod q for each index k, according to FIPS |
| // 203, Appendix A. |
| var zetas = [128]fieldElement{1, 1729, 2580, 3289, 2642, 630, 1897, 848, 1062, 1919, 193, 797, 2786, 3260, 569, 1746, 296, 2447, 1339, 1476, 3046, 56, 2240, 1333, 1426, 2094, 535, 2882, 2393, 2879, 1974, 821, 289, 331, 3253, 1756, 1197, 2304, 2277, 2055, 650, 1977, 2513, 632, 2865, 33, 1320, 1915, 2319, 1435, 807, 452, 1438, 2868, 1534, 2402, 2647, 2617, 1481, 648, 2474, 3110, 1227, 910, 17, 2761, 583, 2649, 1637, 723, 2288, 1100, 1409, 2662, 3281, 233, 756, 2156, 3015, 3050, 1703, 1651, 2789, 1789, 1847, 952, 1461, 2687, 939, 2308, 2437, 2388, 733, 2337, 268, 641, 1584, 2298, 2037, 3220, 375, 2549, 2090, 1645, 1063, 319, 2773, 757, 2099, 561, 2466, 2594, 2804, 1092, 403, 1026, 1143, 2150, 2775, 886, 1722, 1212, 1874, 1029, 2110, 2935, 885, 2154} |
| |
| // ntt maps a ringElement to its nttElement representation. |
| // |
| // It implements NTT, according to FIPS 203, Algorithm 9. |
| func ntt(f ringElement) nttElement { |
| k := 1 |
| for len := 128; len >= 2; len /= 2 { |
| for start := 0; start < 256; start += 2 * len { |
| zeta := zetas[k] |
| k++ |
| // Bounds check elimination hint. |
| f, flen := f[start:start+len], f[start+len:start+len+len] |
| for j := 0; j < len; j++ { |
| t := fieldMul(zeta, flen[j]) |
| flen[j] = fieldSub(f[j], t) |
| f[j] = fieldAdd(f[j], t) |
| } |
| } |
| } |
| return nttElement(f) |
| } |
| |
| // inverseNTT maps a nttElement back to the ringElement it represents. |
| // |
| // It implements NTT⁻¹, according to FIPS 203, Algorithm 10. |
| func inverseNTT(f nttElement) ringElement { |
| k := 127 |
| for len := 2; len <= 128; len *= 2 { |
| for start := 0; start < 256; start += 2 * len { |
| zeta := zetas[k] |
| k-- |
| // Bounds check elimination hint. |
| f, flen := f[start:start+len], f[start+len:start+len+len] |
| for j := 0; j < len; j++ { |
| t := f[j] |
| f[j] = fieldAdd(t, flen[j]) |
| flen[j] = fieldMulSub(zeta, flen[j], t) |
| } |
| } |
| } |
| for i := range f { |
| f[i] = fieldMul(f[i], 3303) // 3303 = 128⁻¹ mod q |
| } |
| return ringElement(f) |
| } |
| |
| // sampleNTT draws a uniformly random nttElement from a stream of uniformly |
| // random bytes generated by the XOF function, according to FIPS 203, |
| // Algorithm 7. |
| func sampleNTT(rho []byte, ii, jj byte) nttElement { |
| B := sha3.NewShake128() |
| B.Write(rho) |
| B.Write([]byte{ii, jj}) |
| |
| // SampleNTT essentially draws 12 bits at a time from r, interprets them in |
| // little-endian, and rejects values higher than q, until it drew 256 |
| // values. (The rejection rate is approximately 19%.) |
| // |
| // To do this from a bytes stream, it draws three bytes at a time, and |
| // splits them into two uint16 appropriately masked. |
| // |
| // r₀ r₁ r₂ |
| // |- - - - - - - -|- - - - - - - -|- - - - - - - -| |
| // |
| // Uint16(r₀ || r₁) |
| // |- - - - - - - - - - - - - - - -| |
| // |- - - - - - - - - - - -| |
| // d₁ |
| // |
| // Uint16(r₁ || r₂) |
| // |- - - - - - - - - - - - - - - -| |
| // |- - - - - - - - - - - -| |
| // d₂ |
| // |
| // Note that in little-endian, the rightmost bits are the most significant |
| // bits (dropped with a mask) and the leftmost bits are the least |
| // significant bits (dropped with a right shift). |
| |
| var a nttElement |
| var j int // index into a |
| var buf [24]byte // buffered reads from B |
| off := len(buf) // index into buf, starts in a "buffer fully consumed" state |
| for { |
| if off >= len(buf) { |
| B.Read(buf[:]) |
| off = 0 |
| } |
| d1 := byteorder.LEUint16(buf[off:]) & 0b1111_1111_1111 |
| d2 := byteorder.LEUint16(buf[off+1:]) >> 4 |
| off += 3 |
| if d1 < q { |
| a[j] = fieldElement(d1) |
| j++ |
| } |
| if j >= len(a) { |
| break |
| } |
| if d2 < q { |
| a[j] = fieldElement(d2) |
| j++ |
| } |
| if j >= len(a) { |
| break |
| } |
| } |
| return a |
| } |