xref: /freebsd/contrib/wireguard-tools/curve25519-fiat32.h (revision 2e3507c25e42292b45a5482e116d278f5515d04d)
1 // SPDX-License-Identifier: GPL-2.0 OR MIT
2 /*
3  * Copyright (C) 2015-2016 The fiat-crypto Authors.
4  * Copyright (C) 2018-2020 Jason A. Donenfeld <Jason@zx2c4.com>. All Rights Reserved.
5  *
6  * This is a machine-generated formally verified implementation of Curve25519
7  * ECDH from: <https://github.com/mit-plv/fiat-crypto>. Though originally
8  * machine generated, it has been tweaked to be suitable for use in the kernel.
9  * It is optimized for 32-bit machines and machines that cannot work efficiently
10  * with 128-bit integer types.
11  */
12 
13 /* fe means field element. Here the field is \Z/(2^255-19). An element t,
14  * entries t[0]...t[9], represents the integer t[0]+2^26 t[1]+2^51 t[2]+2^77
15  * t[3]+2^102 t[4]+...+2^230 t[9].
16  * fe limbs are bounded by 1.125*2^26,1.125*2^25,1.125*2^26,1.125*2^25,etc.
17  * Multiplication and carrying produce fe from fe_loose.
18  */
19 typedef struct fe { u32 v[10]; } fe;
20 
21 /* fe_loose limbs are bounded by 3.375*2^26,3.375*2^25,3.375*2^26,3.375*2^25,etc
22  * Addition and subtraction produce fe_loose from (fe, fe).
23  */
24 typedef struct fe_loose { u32 v[10]; } fe_loose;
25 
26 static __always_inline void fe_frombytes_impl(u32 h[10], const u8 *s)
27 {
28 	/* Ignores top bit of s. */
29 	u32 a0 = get_unaligned_le32(s);
30 	u32 a1 = get_unaligned_le32(s+4);
31 	u32 a2 = get_unaligned_le32(s+8);
32 	u32 a3 = get_unaligned_le32(s+12);
33 	u32 a4 = get_unaligned_le32(s+16);
34 	u32 a5 = get_unaligned_le32(s+20);
35 	u32 a6 = get_unaligned_le32(s+24);
36 	u32 a7 = get_unaligned_le32(s+28);
37 	h[0] = a0&((1<<26)-1);                    /* 26 used, 32-26 left.   26 */
38 	h[1] = (a0>>26) | ((a1&((1<<19)-1))<< 6); /* (32-26) + 19 =  6+19 = 25 */
39 	h[2] = (a1>>19) | ((a2&((1<<13)-1))<<13); /* (32-19) + 13 = 13+13 = 26 */
40 	h[3] = (a2>>13) | ((a3&((1<< 6)-1))<<19); /* (32-13) +  6 = 19+ 6 = 25 */
41 	h[4] = (a3>> 6);                          /* (32- 6)              = 26 */
42 	h[5] = a4&((1<<25)-1);                    /*                        25 */
43 	h[6] = (a4>>25) | ((a5&((1<<19)-1))<< 7); /* (32-25) + 19 =  7+19 = 26 */
44 	h[7] = (a5>>19) | ((a6&((1<<12)-1))<<13); /* (32-19) + 12 = 13+12 = 25 */
45 	h[8] = (a6>>12) | ((a7&((1<< 6)-1))<<20); /* (32-12) +  6 = 20+ 6 = 26 */
46 	h[9] = (a7>> 6)&((1<<25)-1); /*                                     25 */
47 }
48 
49 static __always_inline void fe_frombytes(fe *h, const u8 *s)
50 {
51 	fe_frombytes_impl(h->v, s);
52 }
53 
54 static __always_inline u8 /*bool*/
55 addcarryx_u25(u8 /*bool*/ c, u32 a, u32 b, u32 *low)
56 {
57 	/* This function extracts 25 bits of result and 1 bit of carry
58 	 * (26 total), so a 32-bit intermediate is sufficient.
59 	 */
60 	u32 x = a + b + c;
61 	*low = x & ((1 << 25) - 1);
62 	return (x >> 25) & 1;
63 }
64 
65 static __always_inline u8 /*bool*/
66 addcarryx_u26(u8 /*bool*/ c, u32 a, u32 b, u32 *low)
67 {
68 	/* This function extracts 26 bits of result and 1 bit of carry
69 	 * (27 total), so a 32-bit intermediate is sufficient.
70 	 */
71 	u32 x = a + b + c;
72 	*low = x & ((1 << 26) - 1);
73 	return (x >> 26) & 1;
74 }
75 
76 static __always_inline u8 /*bool*/
77 subborrow_u25(u8 /*bool*/ c, u32 a, u32 b, u32 *low)
78 {
79 	/* This function extracts 25 bits of result and 1 bit of borrow
80 	 * (26 total), so a 32-bit intermediate is sufficient.
81 	 */
82 	u32 x = a - b - c;
83 	*low = x & ((1 << 25) - 1);
84 	return x >> 31;
85 }
86 
87 static __always_inline u8 /*bool*/
88 subborrow_u26(u8 /*bool*/ c, u32 a, u32 b, u32 *low)
89 {
90 	/* This function extracts 26 bits of result and 1 bit of borrow
91 	 *(27 total), so a 32-bit intermediate is sufficient.
92 	 */
93 	u32 x = a - b - c;
94 	*low = x & ((1 << 26) - 1);
95 	return x >> 31;
96 }
97 
98 static __always_inline u32 cmovznz32(u32 t, u32 z, u32 nz)
99 {
100 	t = -!!t; /* all set if nonzero, 0 if 0 */
101 	return (t&nz) | ((~t)&z);
102 }
103 
104 static __always_inline void fe_freeze(u32 out[10], const u32 in1[10])
105 {
106 	{ const u32 x17 = in1[9];
107 	{ const u32 x18 = in1[8];
108 	{ const u32 x16 = in1[7];
109 	{ const u32 x14 = in1[6];
110 	{ const u32 x12 = in1[5];
111 	{ const u32 x10 = in1[4];
112 	{ const u32 x8 = in1[3];
113 	{ const u32 x6 = in1[2];
114 	{ const u32 x4 = in1[1];
115 	{ const u32 x2 = in1[0];
116 	{ u32 x20; u8/*bool*/ x21 = subborrow_u26(0x0, x2, 0x3ffffed, &x20);
117 	{ u32 x23; u8/*bool*/ x24 = subborrow_u25(x21, x4, 0x1ffffff, &x23);
118 	{ u32 x26; u8/*bool*/ x27 = subborrow_u26(x24, x6, 0x3ffffff, &x26);
119 	{ u32 x29; u8/*bool*/ x30 = subborrow_u25(x27, x8, 0x1ffffff, &x29);
120 	{ u32 x32; u8/*bool*/ x33 = subborrow_u26(x30, x10, 0x3ffffff, &x32);
121 	{ u32 x35; u8/*bool*/ x36 = subborrow_u25(x33, x12, 0x1ffffff, &x35);
122 	{ u32 x38; u8/*bool*/ x39 = subborrow_u26(x36, x14, 0x3ffffff, &x38);
123 	{ u32 x41; u8/*bool*/ x42 = subborrow_u25(x39, x16, 0x1ffffff, &x41);
124 	{ u32 x44; u8/*bool*/ x45 = subborrow_u26(x42, x18, 0x3ffffff, &x44);
125 	{ u32 x47; u8/*bool*/ x48 = subborrow_u25(x45, x17, 0x1ffffff, &x47);
126 	{ u32 x49 = cmovznz32(x48, 0x0, 0xffffffff);
127 	{ u32 x50 = (x49 & 0x3ffffed);
128 	{ u32 x52; u8/*bool*/ x53 = addcarryx_u26(0x0, x20, x50, &x52);
129 	{ u32 x54 = (x49 & 0x1ffffff);
130 	{ u32 x56; u8/*bool*/ x57 = addcarryx_u25(x53, x23, x54, &x56);
131 	{ u32 x58 = (x49 & 0x3ffffff);
132 	{ u32 x60; u8/*bool*/ x61 = addcarryx_u26(x57, x26, x58, &x60);
133 	{ u32 x62 = (x49 & 0x1ffffff);
134 	{ u32 x64; u8/*bool*/ x65 = addcarryx_u25(x61, x29, x62, &x64);
135 	{ u32 x66 = (x49 & 0x3ffffff);
136 	{ u32 x68; u8/*bool*/ x69 = addcarryx_u26(x65, x32, x66, &x68);
137 	{ u32 x70 = (x49 & 0x1ffffff);
138 	{ u32 x72; u8/*bool*/ x73 = addcarryx_u25(x69, x35, x70, &x72);
139 	{ u32 x74 = (x49 & 0x3ffffff);
140 	{ u32 x76; u8/*bool*/ x77 = addcarryx_u26(x73, x38, x74, &x76);
141 	{ u32 x78 = (x49 & 0x1ffffff);
142 	{ u32 x80; u8/*bool*/ x81 = addcarryx_u25(x77, x41, x78, &x80);
143 	{ u32 x82 = (x49 & 0x3ffffff);
144 	{ u32 x84; u8/*bool*/ x85 = addcarryx_u26(x81, x44, x82, &x84);
145 	{ u32 x86 = (x49 & 0x1ffffff);
146 	{ u32 x88; addcarryx_u25(x85, x47, x86, &x88);
147 	out[0] = x52;
148 	out[1] = x56;
149 	out[2] = x60;
150 	out[3] = x64;
151 	out[4] = x68;
152 	out[5] = x72;
153 	out[6] = x76;
154 	out[7] = x80;
155 	out[8] = x84;
156 	out[9] = x88;
157 	}}}}}}}}}}}}}}}}}}}}}}}}}}}}}}}}}}}}}}}}}
158 }
159 
160 static __always_inline void fe_tobytes(u8 s[32], const fe *f)
161 {
162 	u32 h[10];
163 	fe_freeze(h, f->v);
164 	s[0] = h[0] >> 0;
165 	s[1] = h[0] >> 8;
166 	s[2] = h[0] >> 16;
167 	s[3] = (h[0] >> 24) | (h[1] << 2);
168 	s[4] = h[1] >> 6;
169 	s[5] = h[1] >> 14;
170 	s[6] = (h[1] >> 22) | (h[2] << 3);
171 	s[7] = h[2] >> 5;
172 	s[8] = h[2] >> 13;
173 	s[9] = (h[2] >> 21) | (h[3] << 5);
174 	s[10] = h[3] >> 3;
175 	s[11] = h[3] >> 11;
176 	s[12] = (h[3] >> 19) | (h[4] << 6);
177 	s[13] = h[4] >> 2;
178 	s[14] = h[4] >> 10;
179 	s[15] = h[4] >> 18;
180 	s[16] = h[5] >> 0;
181 	s[17] = h[5] >> 8;
182 	s[18] = h[5] >> 16;
183 	s[19] = (h[5] >> 24) | (h[6] << 1);
184 	s[20] = h[6] >> 7;
185 	s[21] = h[6] >> 15;
186 	s[22] = (h[6] >> 23) | (h[7] << 3);
187 	s[23] = h[7] >> 5;
188 	s[24] = h[7] >> 13;
189 	s[25] = (h[7] >> 21) | (h[8] << 4);
190 	s[26] = h[8] >> 4;
191 	s[27] = h[8] >> 12;
192 	s[28] = (h[8] >> 20) | (h[9] << 6);
193 	s[29] = h[9] >> 2;
194 	s[30] = h[9] >> 10;
195 	s[31] = h[9] >> 18;
196 }
197 
198 /* h = f */
199 static __always_inline void fe_copy(fe *h, const fe *f)
200 {
201 	memmove(h, f, sizeof(u32) * 10);
202 }
203 
204 static __always_inline void fe_copy_lt(fe_loose *h, const fe *f)
205 {
206 	memmove(h, f, sizeof(u32) * 10);
207 }
208 
209 /* h = 0 */
210 static __always_inline void fe_0(fe *h)
211 {
212 	memset(h, 0, sizeof(u32) * 10);
213 }
214 
215 /* h = 1 */
216 static __always_inline void fe_1(fe *h)
217 {
218 	memset(h, 0, sizeof(u32) * 10);
219 	h->v[0] = 1;
220 }
221 
222 static void fe_add_impl(u32 out[10], const u32 in1[10], const u32 in2[10])
223 {
224 	{ const u32 x20 = in1[9];
225 	{ const u32 x21 = in1[8];
226 	{ const u32 x19 = in1[7];
227 	{ const u32 x17 = in1[6];
228 	{ const u32 x15 = in1[5];
229 	{ const u32 x13 = in1[4];
230 	{ const u32 x11 = in1[3];
231 	{ const u32 x9 = in1[2];
232 	{ const u32 x7 = in1[1];
233 	{ const u32 x5 = in1[0];
234 	{ const u32 x38 = in2[9];
235 	{ const u32 x39 = in2[8];
236 	{ const u32 x37 = in2[7];
237 	{ const u32 x35 = in2[6];
238 	{ const u32 x33 = in2[5];
239 	{ const u32 x31 = in2[4];
240 	{ const u32 x29 = in2[3];
241 	{ const u32 x27 = in2[2];
242 	{ const u32 x25 = in2[1];
243 	{ const u32 x23 = in2[0];
244 	out[0] = (x5 + x23);
245 	out[1] = (x7 + x25);
246 	out[2] = (x9 + x27);
247 	out[3] = (x11 + x29);
248 	out[4] = (x13 + x31);
249 	out[5] = (x15 + x33);
250 	out[6] = (x17 + x35);
251 	out[7] = (x19 + x37);
252 	out[8] = (x21 + x39);
253 	out[9] = (x20 + x38);
254 	}}}}}}}}}}}}}}}}}}}}
255 }
256 
257 /* h = f + g
258  * Can overlap h with f or g.
259  */
260 static __always_inline void fe_add(fe_loose *h, const fe *f, const fe *g)
261 {
262 	fe_add_impl(h->v, f->v, g->v);
263 }
264 
265 static void fe_sub_impl(u32 out[10], const u32 in1[10], const u32 in2[10])
266 {
267 	{ const u32 x20 = in1[9];
268 	{ const u32 x21 = in1[8];
269 	{ const u32 x19 = in1[7];
270 	{ const u32 x17 = in1[6];
271 	{ const u32 x15 = in1[5];
272 	{ const u32 x13 = in1[4];
273 	{ const u32 x11 = in1[3];
274 	{ const u32 x9 = in1[2];
275 	{ const u32 x7 = in1[1];
276 	{ const u32 x5 = in1[0];
277 	{ const u32 x38 = in2[9];
278 	{ const u32 x39 = in2[8];
279 	{ const u32 x37 = in2[7];
280 	{ const u32 x35 = in2[6];
281 	{ const u32 x33 = in2[5];
282 	{ const u32 x31 = in2[4];
283 	{ const u32 x29 = in2[3];
284 	{ const u32 x27 = in2[2];
285 	{ const u32 x25 = in2[1];
286 	{ const u32 x23 = in2[0];
287 	out[0] = ((0x7ffffda + x5) - x23);
288 	out[1] = ((0x3fffffe + x7) - x25);
289 	out[2] = ((0x7fffffe + x9) - x27);
290 	out[3] = ((0x3fffffe + x11) - x29);
291 	out[4] = ((0x7fffffe + x13) - x31);
292 	out[5] = ((0x3fffffe + x15) - x33);
293 	out[6] = ((0x7fffffe + x17) - x35);
294 	out[7] = ((0x3fffffe + x19) - x37);
295 	out[8] = ((0x7fffffe + x21) - x39);
296 	out[9] = ((0x3fffffe + x20) - x38);
297 	}}}}}}}}}}}}}}}}}}}}
298 }
299 
300 /* h = f - g
301  * Can overlap h with f or g.
302  */
303 static __always_inline void fe_sub(fe_loose *h, const fe *f, const fe *g)
304 {
305 	fe_sub_impl(h->v, f->v, g->v);
306 }
307 
308 static void fe_mul_impl(u32 out[10], const u32 in1[10], const u32 in2[10])
309 {
310 	{ const u32 x20 = in1[9];
311 	{ const u32 x21 = in1[8];
312 	{ const u32 x19 = in1[7];
313 	{ const u32 x17 = in1[6];
314 	{ const u32 x15 = in1[5];
315 	{ const u32 x13 = in1[4];
316 	{ const u32 x11 = in1[3];
317 	{ const u32 x9 = in1[2];
318 	{ const u32 x7 = in1[1];
319 	{ const u32 x5 = in1[0];
320 	{ const u32 x38 = in2[9];
321 	{ const u32 x39 = in2[8];
322 	{ const u32 x37 = in2[7];
323 	{ const u32 x35 = in2[6];
324 	{ const u32 x33 = in2[5];
325 	{ const u32 x31 = in2[4];
326 	{ const u32 x29 = in2[3];
327 	{ const u32 x27 = in2[2];
328 	{ const u32 x25 = in2[1];
329 	{ const u32 x23 = in2[0];
330 	{ u64 x40 = ((u64)x23 * x5);
331 	{ u64 x41 = (((u64)x23 * x7) + ((u64)x25 * x5));
332 	{ u64 x42 = ((((u64)(0x2 * x25) * x7) + ((u64)x23 * x9)) + ((u64)x27 * x5));
333 	{ u64 x43 = (((((u64)x25 * x9) + ((u64)x27 * x7)) + ((u64)x23 * x11)) + ((u64)x29 * x5));
334 	{ u64 x44 = (((((u64)x27 * x9) + (0x2 * (((u64)x25 * x11) + ((u64)x29 * x7)))) + ((u64)x23 * x13)) + ((u64)x31 * x5));
335 	{ u64 x45 = (((((((u64)x27 * x11) + ((u64)x29 * x9)) + ((u64)x25 * x13)) + ((u64)x31 * x7)) + ((u64)x23 * x15)) + ((u64)x33 * x5));
336 	{ u64 x46 = (((((0x2 * ((((u64)x29 * x11) + ((u64)x25 * x15)) + ((u64)x33 * x7))) + ((u64)x27 * x13)) + ((u64)x31 * x9)) + ((u64)x23 * x17)) + ((u64)x35 * x5));
337 	{ u64 x47 = (((((((((u64)x29 * x13) + ((u64)x31 * x11)) + ((u64)x27 * x15)) + ((u64)x33 * x9)) + ((u64)x25 * x17)) + ((u64)x35 * x7)) + ((u64)x23 * x19)) + ((u64)x37 * x5));
338 	{ u64 x48 = (((((((u64)x31 * x13) + (0x2 * (((((u64)x29 * x15) + ((u64)x33 * x11)) + ((u64)x25 * x19)) + ((u64)x37 * x7)))) + ((u64)x27 * x17)) + ((u64)x35 * x9)) + ((u64)x23 * x21)) + ((u64)x39 * x5));
339 	{ u64 x49 = (((((((((((u64)x31 * x15) + ((u64)x33 * x13)) + ((u64)x29 * x17)) + ((u64)x35 * x11)) + ((u64)x27 * x19)) + ((u64)x37 * x9)) + ((u64)x25 * x21)) + ((u64)x39 * x7)) + ((u64)x23 * x20)) + ((u64)x38 * x5));
340 	{ u64 x50 = (((((0x2 * ((((((u64)x33 * x15) + ((u64)x29 * x19)) + ((u64)x37 * x11)) + ((u64)x25 * x20)) + ((u64)x38 * x7))) + ((u64)x31 * x17)) + ((u64)x35 * x13)) + ((u64)x27 * x21)) + ((u64)x39 * x9));
341 	{ u64 x51 = (((((((((u64)x33 * x17) + ((u64)x35 * x15)) + ((u64)x31 * x19)) + ((u64)x37 * x13)) + ((u64)x29 * x21)) + ((u64)x39 * x11)) + ((u64)x27 * x20)) + ((u64)x38 * x9));
342 	{ u64 x52 = (((((u64)x35 * x17) + (0x2 * (((((u64)x33 * x19) + ((u64)x37 * x15)) + ((u64)x29 * x20)) + ((u64)x38 * x11)))) + ((u64)x31 * x21)) + ((u64)x39 * x13));
343 	{ u64 x53 = (((((((u64)x35 * x19) + ((u64)x37 * x17)) + ((u64)x33 * x21)) + ((u64)x39 * x15)) + ((u64)x31 * x20)) + ((u64)x38 * x13));
344 	{ u64 x54 = (((0x2 * ((((u64)x37 * x19) + ((u64)x33 * x20)) + ((u64)x38 * x15))) + ((u64)x35 * x21)) + ((u64)x39 * x17));
345 	{ u64 x55 = (((((u64)x37 * x21) + ((u64)x39 * x19)) + ((u64)x35 * x20)) + ((u64)x38 * x17));
346 	{ u64 x56 = (((u64)x39 * x21) + (0x2 * (((u64)x37 * x20) + ((u64)x38 * x19))));
347 	{ u64 x57 = (((u64)x39 * x20) + ((u64)x38 * x21));
348 	{ u64 x58 = ((u64)(0x2 * x38) * x20);
349 	{ u64 x59 = (x48 + (x58 << 0x4));
350 	{ u64 x60 = (x59 + (x58 << 0x1));
351 	{ u64 x61 = (x60 + x58);
352 	{ u64 x62 = (x47 + (x57 << 0x4));
353 	{ u64 x63 = (x62 + (x57 << 0x1));
354 	{ u64 x64 = (x63 + x57);
355 	{ u64 x65 = (x46 + (x56 << 0x4));
356 	{ u64 x66 = (x65 + (x56 << 0x1));
357 	{ u64 x67 = (x66 + x56);
358 	{ u64 x68 = (x45 + (x55 << 0x4));
359 	{ u64 x69 = (x68 + (x55 << 0x1));
360 	{ u64 x70 = (x69 + x55);
361 	{ u64 x71 = (x44 + (x54 << 0x4));
362 	{ u64 x72 = (x71 + (x54 << 0x1));
363 	{ u64 x73 = (x72 + x54);
364 	{ u64 x74 = (x43 + (x53 << 0x4));
365 	{ u64 x75 = (x74 + (x53 << 0x1));
366 	{ u64 x76 = (x75 + x53);
367 	{ u64 x77 = (x42 + (x52 << 0x4));
368 	{ u64 x78 = (x77 + (x52 << 0x1));
369 	{ u64 x79 = (x78 + x52);
370 	{ u64 x80 = (x41 + (x51 << 0x4));
371 	{ u64 x81 = (x80 + (x51 << 0x1));
372 	{ u64 x82 = (x81 + x51);
373 	{ u64 x83 = (x40 + (x50 << 0x4));
374 	{ u64 x84 = (x83 + (x50 << 0x1));
375 	{ u64 x85 = (x84 + x50);
376 	{ u64 x86 = (x85 >> 0x1a);
377 	{ u32 x87 = ((u32)x85 & 0x3ffffff);
378 	{ u64 x88 = (x86 + x82);
379 	{ u64 x89 = (x88 >> 0x19);
380 	{ u32 x90 = ((u32)x88 & 0x1ffffff);
381 	{ u64 x91 = (x89 + x79);
382 	{ u64 x92 = (x91 >> 0x1a);
383 	{ u32 x93 = ((u32)x91 & 0x3ffffff);
384 	{ u64 x94 = (x92 + x76);
385 	{ u64 x95 = (x94 >> 0x19);
386 	{ u32 x96 = ((u32)x94 & 0x1ffffff);
387 	{ u64 x97 = (x95 + x73);
388 	{ u64 x98 = (x97 >> 0x1a);
389 	{ u32 x99 = ((u32)x97 & 0x3ffffff);
390 	{ u64 x100 = (x98 + x70);
391 	{ u64 x101 = (x100 >> 0x19);
392 	{ u32 x102 = ((u32)x100 & 0x1ffffff);
393 	{ u64 x103 = (x101 + x67);
394 	{ u64 x104 = (x103 >> 0x1a);
395 	{ u32 x105 = ((u32)x103 & 0x3ffffff);
396 	{ u64 x106 = (x104 + x64);
397 	{ u64 x107 = (x106 >> 0x19);
398 	{ u32 x108 = ((u32)x106 & 0x1ffffff);
399 	{ u64 x109 = (x107 + x61);
400 	{ u64 x110 = (x109 >> 0x1a);
401 	{ u32 x111 = ((u32)x109 & 0x3ffffff);
402 	{ u64 x112 = (x110 + x49);
403 	{ u64 x113 = (x112 >> 0x19);
404 	{ u32 x114 = ((u32)x112 & 0x1ffffff);
405 	{ u64 x115 = (x87 + (0x13 * x113));
406 	{ u32 x116 = (u32) (x115 >> 0x1a);
407 	{ u32 x117 = ((u32)x115 & 0x3ffffff);
408 	{ u32 x118 = (x116 + x90);
409 	{ u32 x119 = (x118 >> 0x19);
410 	{ u32 x120 = (x118 & 0x1ffffff);
411 	out[0] = x117;
412 	out[1] = x120;
413 	out[2] = (x119 + x93);
414 	out[3] = x96;
415 	out[4] = x99;
416 	out[5] = x102;
417 	out[6] = x105;
418 	out[7] = x108;
419 	out[8] = x111;
420 	out[9] = x114;
421 	}}}}}}}}}}}}}}}}}}}}}}}}}}}}}}}}}}}}}}}}}}}}}}}}}}}}}}}}}}}}}}}}}}}}}}}}}}}}}}}}}}}}}}}}}}}}}}}}}}}}}
422 }
423 
424 static __always_inline void fe_mul_ttt(fe *h, const fe *f, const fe *g)
425 {
426 	fe_mul_impl(h->v, f->v, g->v);
427 }
428 
429 static __always_inline void fe_mul_tlt(fe *h, const fe_loose *f, const fe *g)
430 {
431 	fe_mul_impl(h->v, f->v, g->v);
432 }
433 
434 static __always_inline void
435 fe_mul_tll(fe *h, const fe_loose *f, const fe_loose *g)
436 {
437 	fe_mul_impl(h->v, f->v, g->v);
438 }
439 
440 static void fe_sqr_impl(u32 out[10], const u32 in1[10])
441 {
442 	{ const u32 x17 = in1[9];
443 	{ const u32 x18 = in1[8];
444 	{ const u32 x16 = in1[7];
445 	{ const u32 x14 = in1[6];
446 	{ const u32 x12 = in1[5];
447 	{ const u32 x10 = in1[4];
448 	{ const u32 x8 = in1[3];
449 	{ const u32 x6 = in1[2];
450 	{ const u32 x4 = in1[1];
451 	{ const u32 x2 = in1[0];
452 	{ u64 x19 = ((u64)x2 * x2);
453 	{ u64 x20 = ((u64)(0x2 * x2) * x4);
454 	{ u64 x21 = (0x2 * (((u64)x4 * x4) + ((u64)x2 * x6)));
455 	{ u64 x22 = (0x2 * (((u64)x4 * x6) + ((u64)x2 * x8)));
456 	{ u64 x23 = ((((u64)x6 * x6) + ((u64)(0x4 * x4) * x8)) + ((u64)(0x2 * x2) * x10));
457 	{ u64 x24 = (0x2 * ((((u64)x6 * x8) + ((u64)x4 * x10)) + ((u64)x2 * x12)));
458 	{ u64 x25 = (0x2 * (((((u64)x8 * x8) + ((u64)x6 * x10)) + ((u64)x2 * x14)) + ((u64)(0x2 * x4) * x12)));
459 	{ u64 x26 = (0x2 * (((((u64)x8 * x10) + ((u64)x6 * x12)) + ((u64)x4 * x14)) + ((u64)x2 * x16)));
460 	{ u64 x27 = (((u64)x10 * x10) + (0x2 * ((((u64)x6 * x14) + ((u64)x2 * x18)) + (0x2 * (((u64)x4 * x16) + ((u64)x8 * x12))))));
461 	{ u64 x28 = (0x2 * ((((((u64)x10 * x12) + ((u64)x8 * x14)) + ((u64)x6 * x16)) + ((u64)x4 * x18)) + ((u64)x2 * x17)));
462 	{ u64 x29 = (0x2 * (((((u64)x12 * x12) + ((u64)x10 * x14)) + ((u64)x6 * x18)) + (0x2 * (((u64)x8 * x16) + ((u64)x4 * x17)))));
463 	{ u64 x30 = (0x2 * (((((u64)x12 * x14) + ((u64)x10 * x16)) + ((u64)x8 * x18)) + ((u64)x6 * x17)));
464 	{ u64 x31 = (((u64)x14 * x14) + (0x2 * (((u64)x10 * x18) + (0x2 * (((u64)x12 * x16) + ((u64)x8 * x17))))));
465 	{ u64 x32 = (0x2 * ((((u64)x14 * x16) + ((u64)x12 * x18)) + ((u64)x10 * x17)));
466 	{ u64 x33 = (0x2 * ((((u64)x16 * x16) + ((u64)x14 * x18)) + ((u64)(0x2 * x12) * x17)));
467 	{ u64 x34 = (0x2 * (((u64)x16 * x18) + ((u64)x14 * x17)));
468 	{ u64 x35 = (((u64)x18 * x18) + ((u64)(0x4 * x16) * x17));
469 	{ u64 x36 = ((u64)(0x2 * x18) * x17);
470 	{ u64 x37 = ((u64)(0x2 * x17) * x17);
471 	{ u64 x38 = (x27 + (x37 << 0x4));
472 	{ u64 x39 = (x38 + (x37 << 0x1));
473 	{ u64 x40 = (x39 + x37);
474 	{ u64 x41 = (x26 + (x36 << 0x4));
475 	{ u64 x42 = (x41 + (x36 << 0x1));
476 	{ u64 x43 = (x42 + x36);
477 	{ u64 x44 = (x25 + (x35 << 0x4));
478 	{ u64 x45 = (x44 + (x35 << 0x1));
479 	{ u64 x46 = (x45 + x35);
480 	{ u64 x47 = (x24 + (x34 << 0x4));
481 	{ u64 x48 = (x47 + (x34 << 0x1));
482 	{ u64 x49 = (x48 + x34);
483 	{ u64 x50 = (x23 + (x33 << 0x4));
484 	{ u64 x51 = (x50 + (x33 << 0x1));
485 	{ u64 x52 = (x51 + x33);
486 	{ u64 x53 = (x22 + (x32 << 0x4));
487 	{ u64 x54 = (x53 + (x32 << 0x1));
488 	{ u64 x55 = (x54 + x32);
489 	{ u64 x56 = (x21 + (x31 << 0x4));
490 	{ u64 x57 = (x56 + (x31 << 0x1));
491 	{ u64 x58 = (x57 + x31);
492 	{ u64 x59 = (x20 + (x30 << 0x4));
493 	{ u64 x60 = (x59 + (x30 << 0x1));
494 	{ u64 x61 = (x60 + x30);
495 	{ u64 x62 = (x19 + (x29 << 0x4));
496 	{ u64 x63 = (x62 + (x29 << 0x1));
497 	{ u64 x64 = (x63 + x29);
498 	{ u64 x65 = (x64 >> 0x1a);
499 	{ u32 x66 = ((u32)x64 & 0x3ffffff);
500 	{ u64 x67 = (x65 + x61);
501 	{ u64 x68 = (x67 >> 0x19);
502 	{ u32 x69 = ((u32)x67 & 0x1ffffff);
503 	{ u64 x70 = (x68 + x58);
504 	{ u64 x71 = (x70 >> 0x1a);
505 	{ u32 x72 = ((u32)x70 & 0x3ffffff);
506 	{ u64 x73 = (x71 + x55);
507 	{ u64 x74 = (x73 >> 0x19);
508 	{ u32 x75 = ((u32)x73 & 0x1ffffff);
509 	{ u64 x76 = (x74 + x52);
510 	{ u64 x77 = (x76 >> 0x1a);
511 	{ u32 x78 = ((u32)x76 & 0x3ffffff);
512 	{ u64 x79 = (x77 + x49);
513 	{ u64 x80 = (x79 >> 0x19);
514 	{ u32 x81 = ((u32)x79 & 0x1ffffff);
515 	{ u64 x82 = (x80 + x46);
516 	{ u64 x83 = (x82 >> 0x1a);
517 	{ u32 x84 = ((u32)x82 & 0x3ffffff);
518 	{ u64 x85 = (x83 + x43);
519 	{ u64 x86 = (x85 >> 0x19);
520 	{ u32 x87 = ((u32)x85 & 0x1ffffff);
521 	{ u64 x88 = (x86 + x40);
522 	{ u64 x89 = (x88 >> 0x1a);
523 	{ u32 x90 = ((u32)x88 & 0x3ffffff);
524 	{ u64 x91 = (x89 + x28);
525 	{ u64 x92 = (x91 >> 0x19);
526 	{ u32 x93 = ((u32)x91 & 0x1ffffff);
527 	{ u64 x94 = (x66 + (0x13 * x92));
528 	{ u32 x95 = (u32) (x94 >> 0x1a);
529 	{ u32 x96 = ((u32)x94 & 0x3ffffff);
530 	{ u32 x97 = (x95 + x69);
531 	{ u32 x98 = (x97 >> 0x19);
532 	{ u32 x99 = (x97 & 0x1ffffff);
533 	out[0] = x96;
534 	out[1] = x99;
535 	out[2] = (x98 + x72);
536 	out[3] = x75;
537 	out[4] = x78;
538 	out[5] = x81;
539 	out[6] = x84;
540 	out[7] = x87;
541 	out[8] = x90;
542 	out[9] = x93;
543 	}}}}}}}}}}}}}}}}}}}}}}}}}}}}}}}}}}}}}}}}}}}}}}}}}}}}}}}}}}}}}}}}}}}}}}}}}}}}}}}}}}}}}}}}}}}
544 }
545 
546 static __always_inline void fe_sq_tl(fe *h, const fe_loose *f)
547 {
548 	fe_sqr_impl(h->v, f->v);
549 }
550 
551 static __always_inline void fe_sq_tt(fe *h, const fe *f)
552 {
553 	fe_sqr_impl(h->v, f->v);
554 }
555 
556 static __always_inline void fe_loose_invert(fe *out, const fe_loose *z)
557 {
558 	fe t0;
559 	fe t1;
560 	fe t2;
561 	fe t3;
562 	int i;
563 
564 	fe_sq_tl(&t0, z);
565 	fe_sq_tt(&t1, &t0);
566 	for (i = 1; i < 2; ++i)
567 		fe_sq_tt(&t1, &t1);
568 	fe_mul_tlt(&t1, z, &t1);
569 	fe_mul_ttt(&t0, &t0, &t1);
570 	fe_sq_tt(&t2, &t0);
571 	fe_mul_ttt(&t1, &t1, &t2);
572 	fe_sq_tt(&t2, &t1);
573 	for (i = 1; i < 5; ++i)
574 		fe_sq_tt(&t2, &t2);
575 	fe_mul_ttt(&t1, &t2, &t1);
576 	fe_sq_tt(&t2, &t1);
577 	for (i = 1; i < 10; ++i)
578 		fe_sq_tt(&t2, &t2);
579 	fe_mul_ttt(&t2, &t2, &t1);
580 	fe_sq_tt(&t3, &t2);
581 	for (i = 1; i < 20; ++i)
582 		fe_sq_tt(&t3, &t3);
583 	fe_mul_ttt(&t2, &t3, &t2);
584 	fe_sq_tt(&t2, &t2);
585 	for (i = 1; i < 10; ++i)
586 		fe_sq_tt(&t2, &t2);
587 	fe_mul_ttt(&t1, &t2, &t1);
588 	fe_sq_tt(&t2, &t1);
589 	for (i = 1; i < 50; ++i)
590 		fe_sq_tt(&t2, &t2);
591 	fe_mul_ttt(&t2, &t2, &t1);
592 	fe_sq_tt(&t3, &t2);
593 	for (i = 1; i < 100; ++i)
594 		fe_sq_tt(&t3, &t3);
595 	fe_mul_ttt(&t2, &t3, &t2);
596 	fe_sq_tt(&t2, &t2);
597 	for (i = 1; i < 50; ++i)
598 		fe_sq_tt(&t2, &t2);
599 	fe_mul_ttt(&t1, &t2, &t1);
600 	fe_sq_tt(&t1, &t1);
601 	for (i = 1; i < 5; ++i)
602 		fe_sq_tt(&t1, &t1);
603 	fe_mul_ttt(out, &t1, &t0);
604 }
605 
606 static __always_inline void fe_invert(fe *out, const fe *z)
607 {
608 	fe_loose l;
609 	fe_copy_lt(&l, z);
610 	fe_loose_invert(out, &l);
611 }
612 
613 /* Replace (f,g) with (g,f) if b == 1;
614  * replace (f,g) with (f,g) if b == 0.
615  *
616  * Preconditions: b in {0,1}
617  */
618 static __always_inline void fe_cswap(fe *f, fe *g, unsigned int b)
619 {
620 	unsigned i;
621 	b = 0 - b;
622 	for (i = 0; i < 10; i++) {
623 		u32 x = f->v[i] ^ g->v[i];
624 		x &= b;
625 		f->v[i] ^= x;
626 		g->v[i] ^= x;
627 	}
628 }
629 
630 /* NOTE: based on fiat-crypto fe_mul, edited for in2=121666, 0, 0.*/
631 static __always_inline void fe_mul_121666_impl(u32 out[10], const u32 in1[10])
632 {
633 	{ const u32 x20 = in1[9];
634 	{ const u32 x21 = in1[8];
635 	{ const u32 x19 = in1[7];
636 	{ const u32 x17 = in1[6];
637 	{ const u32 x15 = in1[5];
638 	{ const u32 x13 = in1[4];
639 	{ const u32 x11 = in1[3];
640 	{ const u32 x9 = in1[2];
641 	{ const u32 x7 = in1[1];
642 	{ const u32 x5 = in1[0];
643 	{ const u32 x38 = 0;
644 	{ const u32 x39 = 0;
645 	{ const u32 x37 = 0;
646 	{ const u32 x35 = 0;
647 	{ const u32 x33 = 0;
648 	{ const u32 x31 = 0;
649 	{ const u32 x29 = 0;
650 	{ const u32 x27 = 0;
651 	{ const u32 x25 = 0;
652 	{ const u32 x23 = 121666;
653 	{ u64 x40 = ((u64)x23 * x5);
654 	{ u64 x41 = (((u64)x23 * x7) + ((u64)x25 * x5));
655 	{ u64 x42 = ((((u64)(0x2 * x25) * x7) + ((u64)x23 * x9)) + ((u64)x27 * x5));
656 	{ u64 x43 = (((((u64)x25 * x9) + ((u64)x27 * x7)) + ((u64)x23 * x11)) + ((u64)x29 * x5));
657 	{ u64 x44 = (((((u64)x27 * x9) + (0x2 * (((u64)x25 * x11) + ((u64)x29 * x7)))) + ((u64)x23 * x13)) + ((u64)x31 * x5));
658 	{ u64 x45 = (((((((u64)x27 * x11) + ((u64)x29 * x9)) + ((u64)x25 * x13)) + ((u64)x31 * x7)) + ((u64)x23 * x15)) + ((u64)x33 * x5));
659 	{ u64 x46 = (((((0x2 * ((((u64)x29 * x11) + ((u64)x25 * x15)) + ((u64)x33 * x7))) + ((u64)x27 * x13)) + ((u64)x31 * x9)) + ((u64)x23 * x17)) + ((u64)x35 * x5));
660 	{ u64 x47 = (((((((((u64)x29 * x13) + ((u64)x31 * x11)) + ((u64)x27 * x15)) + ((u64)x33 * x9)) + ((u64)x25 * x17)) + ((u64)x35 * x7)) + ((u64)x23 * x19)) + ((u64)x37 * x5));
661 	{ u64 x48 = (((((((u64)x31 * x13) + (0x2 * (((((u64)x29 * x15) + ((u64)x33 * x11)) + ((u64)x25 * x19)) + ((u64)x37 * x7)))) + ((u64)x27 * x17)) + ((u64)x35 * x9)) + ((u64)x23 * x21)) + ((u64)x39 * x5));
662 	{ u64 x49 = (((((((((((u64)x31 * x15) + ((u64)x33 * x13)) + ((u64)x29 * x17)) + ((u64)x35 * x11)) + ((u64)x27 * x19)) + ((u64)x37 * x9)) + ((u64)x25 * x21)) + ((u64)x39 * x7)) + ((u64)x23 * x20)) + ((u64)x38 * x5));
663 	{ u64 x50 = (((((0x2 * ((((((u64)x33 * x15) + ((u64)x29 * x19)) + ((u64)x37 * x11)) + ((u64)x25 * x20)) + ((u64)x38 * x7))) + ((u64)x31 * x17)) + ((u64)x35 * x13)) + ((u64)x27 * x21)) + ((u64)x39 * x9));
664 	{ u64 x51 = (((((((((u64)x33 * x17) + ((u64)x35 * x15)) + ((u64)x31 * x19)) + ((u64)x37 * x13)) + ((u64)x29 * x21)) + ((u64)x39 * x11)) + ((u64)x27 * x20)) + ((u64)x38 * x9));
665 	{ u64 x52 = (((((u64)x35 * x17) + (0x2 * (((((u64)x33 * x19) + ((u64)x37 * x15)) + ((u64)x29 * x20)) + ((u64)x38 * x11)))) + ((u64)x31 * x21)) + ((u64)x39 * x13));
666 	{ u64 x53 = (((((((u64)x35 * x19) + ((u64)x37 * x17)) + ((u64)x33 * x21)) + ((u64)x39 * x15)) + ((u64)x31 * x20)) + ((u64)x38 * x13));
667 	{ u64 x54 = (((0x2 * ((((u64)x37 * x19) + ((u64)x33 * x20)) + ((u64)x38 * x15))) + ((u64)x35 * x21)) + ((u64)x39 * x17));
668 	{ u64 x55 = (((((u64)x37 * x21) + ((u64)x39 * x19)) + ((u64)x35 * x20)) + ((u64)x38 * x17));
669 	{ u64 x56 = (((u64)x39 * x21) + (0x2 * (((u64)x37 * x20) + ((u64)x38 * x19))));
670 	{ u64 x57 = (((u64)x39 * x20) + ((u64)x38 * x21));
671 	{ u64 x58 = ((u64)(0x2 * x38) * x20);
672 	{ u64 x59 = (x48 + (x58 << 0x4));
673 	{ u64 x60 = (x59 + (x58 << 0x1));
674 	{ u64 x61 = (x60 + x58);
675 	{ u64 x62 = (x47 + (x57 << 0x4));
676 	{ u64 x63 = (x62 + (x57 << 0x1));
677 	{ u64 x64 = (x63 + x57);
678 	{ u64 x65 = (x46 + (x56 << 0x4));
679 	{ u64 x66 = (x65 + (x56 << 0x1));
680 	{ u64 x67 = (x66 + x56);
681 	{ u64 x68 = (x45 + (x55 << 0x4));
682 	{ u64 x69 = (x68 + (x55 << 0x1));
683 	{ u64 x70 = (x69 + x55);
684 	{ u64 x71 = (x44 + (x54 << 0x4));
685 	{ u64 x72 = (x71 + (x54 << 0x1));
686 	{ u64 x73 = (x72 + x54);
687 	{ u64 x74 = (x43 + (x53 << 0x4));
688 	{ u64 x75 = (x74 + (x53 << 0x1));
689 	{ u64 x76 = (x75 + x53);
690 	{ u64 x77 = (x42 + (x52 << 0x4));
691 	{ u64 x78 = (x77 + (x52 << 0x1));
692 	{ u64 x79 = (x78 + x52);
693 	{ u64 x80 = (x41 + (x51 << 0x4));
694 	{ u64 x81 = (x80 + (x51 << 0x1));
695 	{ u64 x82 = (x81 + x51);
696 	{ u64 x83 = (x40 + (x50 << 0x4));
697 	{ u64 x84 = (x83 + (x50 << 0x1));
698 	{ u64 x85 = (x84 + x50);
699 	{ u64 x86 = (x85 >> 0x1a);
700 	{ u32 x87 = ((u32)x85 & 0x3ffffff);
701 	{ u64 x88 = (x86 + x82);
702 	{ u64 x89 = (x88 >> 0x19);
703 	{ u32 x90 = ((u32)x88 & 0x1ffffff);
704 	{ u64 x91 = (x89 + x79);
705 	{ u64 x92 = (x91 >> 0x1a);
706 	{ u32 x93 = ((u32)x91 & 0x3ffffff);
707 	{ u64 x94 = (x92 + x76);
708 	{ u64 x95 = (x94 >> 0x19);
709 	{ u32 x96 = ((u32)x94 & 0x1ffffff);
710 	{ u64 x97 = (x95 + x73);
711 	{ u64 x98 = (x97 >> 0x1a);
712 	{ u32 x99 = ((u32)x97 & 0x3ffffff);
713 	{ u64 x100 = (x98 + x70);
714 	{ u64 x101 = (x100 >> 0x19);
715 	{ u32 x102 = ((u32)x100 & 0x1ffffff);
716 	{ u64 x103 = (x101 + x67);
717 	{ u64 x104 = (x103 >> 0x1a);
718 	{ u32 x105 = ((u32)x103 & 0x3ffffff);
719 	{ u64 x106 = (x104 + x64);
720 	{ u64 x107 = (x106 >> 0x19);
721 	{ u32 x108 = ((u32)x106 & 0x1ffffff);
722 	{ u64 x109 = (x107 + x61);
723 	{ u64 x110 = (x109 >> 0x1a);
724 	{ u32 x111 = ((u32)x109 & 0x3ffffff);
725 	{ u64 x112 = (x110 + x49);
726 	{ u64 x113 = (x112 >> 0x19);
727 	{ u32 x114 = ((u32)x112 & 0x1ffffff);
728 	{ u64 x115 = (x87 + (0x13 * x113));
729 	{ u32 x116 = (u32) (x115 >> 0x1a);
730 	{ u32 x117 = ((u32)x115 & 0x3ffffff);
731 	{ u32 x118 = (x116 + x90);
732 	{ u32 x119 = (x118 >> 0x19);
733 	{ u32 x120 = (x118 & 0x1ffffff);
734 	out[0] = x117;
735 	out[1] = x120;
736 	out[2] = (x119 + x93);
737 	out[3] = x96;
738 	out[4] = x99;
739 	out[5] = x102;
740 	out[6] = x105;
741 	out[7] = x108;
742 	out[8] = x111;
743 	out[9] = x114;
744 	}}}}}}}}}}}}}}}}}}}}}}}}}}}}}}}}}}}}}}}}}}}}}}}}}}}}}}}}}}}}}}}}}}}}}}}}}}}}}}}}}}}}}}}}}}}}}}}}}}}}}
745 }
746 
747 static __always_inline void fe_mul121666(fe *h, const fe_loose *f)
748 {
749 	fe_mul_121666_impl(h->v, f->v);
750 }
751 
752 static void curve25519_generic(u8 out[CURVE25519_KEY_SIZE],
753 			       const u8 scalar[CURVE25519_KEY_SIZE],
754 			       const u8 point[CURVE25519_KEY_SIZE])
755 {
756 	fe x1, x2, z2, x3, z3;
757 	fe_loose x2l, z2l, x3l;
758 	unsigned swap = 0;
759 	int pos;
760 	u8 e[32];
761 
762 	memcpy(e, scalar, 32);
763 	curve25519_clamp_secret(e);
764 
765 	/* The following implementation was transcribed to Coq and proven to
766 	 * correspond to unary scalar multiplication in affine coordinates given
767 	 * that x1 != 0 is the x coordinate of some point on the curve. It was
768 	 * also checked in Coq that doing a ladderstep with x1 = x3 = 0 gives
769 	 * z2' = z3' = 0, and z2 = z3 = 0 gives z2' = z3' = 0. The statement was
770 	 * quantified over the underlying field, so it applies to Curve25519
771 	 * itself and the quadratic twist of Curve25519. It was not proven in
772 	 * Coq that prime-field arithmetic correctly simulates extension-field
773 	 * arithmetic on prime-field values. The decoding of the byte array
774 	 * representation of e was not considered.
775 	 *
776 	 * Specification of Montgomery curves in affine coordinates:
777 	 * <https://github.com/mit-plv/fiat-crypto/blob/2456d821825521f7e03e65882cc3521795b0320f/src/Spec/MontgomeryCurve.v#L27>
778 	 *
779 	 * Proof that these form a group that is isomorphic to a Weierstrass
780 	 * curve:
781 	 * <https://github.com/mit-plv/fiat-crypto/blob/2456d821825521f7e03e65882cc3521795b0320f/src/Curves/Montgomery/AffineProofs.v#L35>
782 	 *
783 	 * Coq transcription and correctness proof of the loop
784 	 * (where scalarbits=255):
785 	 * <https://github.com/mit-plv/fiat-crypto/blob/2456d821825521f7e03e65882cc3521795b0320f/src/Curves/Montgomery/XZ.v#L118>
786 	 * <https://github.com/mit-plv/fiat-crypto/blob/2456d821825521f7e03e65882cc3521795b0320f/src/Curves/Montgomery/XZProofs.v#L278>
787 	 * preconditions: 0 <= e < 2^255 (not necessarily e < order),
788 	 * fe_invert(0) = 0
789 	 */
790 	fe_frombytes(&x1, point);
791 	fe_1(&x2);
792 	fe_0(&z2);
793 	fe_copy(&x3, &x1);
794 	fe_1(&z3);
795 
796 	for (pos = 254; pos >= 0; --pos) {
797 		fe tmp0, tmp1;
798 		fe_loose tmp0l, tmp1l;
799 		/* loop invariant as of right before the test, for the case
800 		 * where x1 != 0:
801 		 *   pos >= -1; if z2 = 0 then x2 is nonzero; if z3 = 0 then x3
802 		 *   is nonzero
803 		 *   let r := e >> (pos+1) in the following equalities of
804 		 *   projective points:
805 		 *   to_xz (r*P)     === if swap then (x3, z3) else (x2, z2)
806 		 *   to_xz ((r+1)*P) === if swap then (x2, z2) else (x3, z3)
807 		 *   x1 is the nonzero x coordinate of the nonzero
808 		 *   point (r*P-(r+1)*P)
809 		 */
810 		unsigned b = 1 & (e[pos / 8] >> (pos & 7));
811 		swap ^= b;
812 		fe_cswap(&x2, &x3, swap);
813 		fe_cswap(&z2, &z3, swap);
814 		swap = b;
815 		/* Coq transcription of ladderstep formula (called from
816 		 * transcribed loop):
817 		 * <https://github.com/mit-plv/fiat-crypto/blob/2456d821825521f7e03e65882cc3521795b0320f/src/Curves/Montgomery/XZ.v#L89>
818 		 * <https://github.com/mit-plv/fiat-crypto/blob/2456d821825521f7e03e65882cc3521795b0320f/src/Curves/Montgomery/XZProofs.v#L131>
819 		 * x1 != 0 <https://github.com/mit-plv/fiat-crypto/blob/2456d821825521f7e03e65882cc3521795b0320f/src/Curves/Montgomery/XZProofs.v#L217>
820 		 * x1  = 0 <https://github.com/mit-plv/fiat-crypto/blob/2456d821825521f7e03e65882cc3521795b0320f/src/Curves/Montgomery/XZProofs.v#L147>
821 		 */
822 		fe_sub(&tmp0l, &x3, &z3);
823 		fe_sub(&tmp1l, &x2, &z2);
824 		fe_add(&x2l, &x2, &z2);
825 		fe_add(&z2l, &x3, &z3);
826 		fe_mul_tll(&z3, &tmp0l, &x2l);
827 		fe_mul_tll(&z2, &z2l, &tmp1l);
828 		fe_sq_tl(&tmp0, &tmp1l);
829 		fe_sq_tl(&tmp1, &x2l);
830 		fe_add(&x3l, &z3, &z2);
831 		fe_sub(&z2l, &z3, &z2);
832 		fe_mul_ttt(&x2, &tmp1, &tmp0);
833 		fe_sub(&tmp1l, &tmp1, &tmp0);
834 		fe_sq_tl(&z2, &z2l);
835 		fe_mul121666(&z3, &tmp1l);
836 		fe_sq_tl(&x3, &x3l);
837 		fe_add(&tmp0l, &tmp0, &z3);
838 		fe_mul_ttt(&z3, &x1, &z2);
839 		fe_mul_tll(&z2, &tmp1l, &tmp0l);
840 	}
841 	/* here pos=-1, so r=e, so to_xz (e*P) === if swap then (x3, z3)
842 	 * else (x2, z2)
843 	 */
844 	fe_cswap(&x2, &x3, swap);
845 	fe_cswap(&z2, &z3, swap);
846 
847 	fe_invert(&z2, &z2);
848 	fe_mul_ttt(&x2, &x2, &z2);
849 	fe_tobytes(out, &x2);
850 
851 	memzero_explicit(&x1, sizeof(x1));
852 	memzero_explicit(&x2, sizeof(x2));
853 	memzero_explicit(&z2, sizeof(z2));
854 	memzero_explicit(&x3, sizeof(x3));
855 	memzero_explicit(&z3, sizeof(z3));
856 	memzero_explicit(&x2l, sizeof(x2l));
857 	memzero_explicit(&z2l, sizeof(z2l));
858 	memzero_explicit(&x3l, sizeof(x3l));
859 	memzero_explicit(&e, sizeof(e));
860 }
861