Mercurial > vec

comparison utils/gendouble.c @ 41:c6e0df09b86f

Find changesets by keywords (author, files, the commit message), revision number or hash, or revset expression.
*: performance improvements with old GCC, reimplement altivec
author Paper <paper@tflc.us>
date Mon, 28 Apr 2025 16:31:59 -0400
parents
children
comparison
equal deleted inserted replaced
40:55cadb1fac4b 41:c6e0df09b86f
1 /**
2 * vec - a tiny SIMD vector library in C99
3 *
4 * Copyright (c) 2024-2025 Paper
5 *
6 * Permission is hereby granted, free of charge, to any person obtaining a copy
7 * of this software and associated documentation files (the "Software"), to deal
8 * in the Software without restriction, including without limitation the rights
9 * to use, copy, modify, merge, publish, distribute, sublicense, and/or sell
10 * copies of the Software, and to permit persons to whom the Software is
11 * furnished to do so, subject to the following conditions:
12 *
13 * The above copyright notice and this permission notice shall be included in all
14 * copies or substantial portions of the Software.
15 *
16 * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
17 * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
18 * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
19 * AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
20 * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
21 * OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
22 * SOFTWARE.
23 **/
24
25 /* Use this file to generate include/vec/impl/double.h !!
26 *
27 * `gcc -o gendouble gendouble.c` */
28
29 #include <stdio.h>
30 #include <stdlib.h>
31 #include <string.h>
32 #include <ctype.h>
33
34 #define ARRAY_SIZE(x) (sizeof(x)/sizeof((x)[0]))
35
36 /* XXX: would it be faster to unroll literally everything instead of defining everything,
37 * and then unpacking it all? */
38 static const char *header =
39 "/**\n"
40 " * vec - a tiny SIMD vector library in C99\n"
41 " * \n"
42 " * Copyright (c) 2024-2025 Paper\n"
43 " * \n"
44 " * Permission is hereby granted, free of charge, to any person obtaining a copy\n"
45 " * of this software and associated documentation files (the \"Software\"), to deal\n"
46 " * in the Software without restriction, including without limitation the rights\n"
47 " * to use, copy, modify, merge, publish, distribute, sublicense, and/or sell\n"
48 " * copies of the Software, and to permit persons to whom the Software is\n"
49 " * furnished to do so, subject to the following conditions:\n"
50 " * \n"
51 " * The above copyright notice and this permission notice shall be included in all\n"
52 " * copies or substantial portions of the Software.\n"
53 " * \n"
54 " * THE SOFTWARE IS PROVIDED \"AS IS\", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR\n"
55 " * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,\n"
56 " * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE\n"
57 " * AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER\n"
58 " * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,\n"
59 " * OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE\n"
60 " * SOFTWARE.\n"
61 "**/\n"
62 "\n"
63 "/* This file is automatically generated! Do not edit it directly!\n"
64 " * Edit the code that generates it in utils/gengeneric.c --paper */\n"
65 "\n"
66 "#ifndef VEC_IMPL_DOUBLE_H_\n"
67 "#define VEC_IMPL_DOUBLE_H_\n"
68 "\n"
69 "#define VEC_DOUBLE_SPLAT(sign, bits, size, halfsize) \\\n"
70 " VEC_FUNC_IMPL v##sign##int##bits##x##size v##sign##int##bits##x##size##_splat(vec_##sign##int##bits x) \\\n"
71 " { \\\n"
72 " v##sign##int##bits##x##size vec; \\\n"
73 " \\\n"
74 " vec.generic[0] = v##sign##int##bits##x##halfsize##_splat(x); \\\n"
75 " vec.generic[1] = v##sign##int##bits##x##halfsize##_splat(x); \\\n"
76 " \\\n"
77 " return vec; \\\n"
78 " }\n"
79 "\n"
80 "#define VEC_DOUBLE_LOAD_EX(name, sign, bits, size, halfsize) \\\n"
81 " VEC_FUNC_IMPL v##sign##int##bits##x##size v##sign##int##bits##x##size##_##name(const vec_##sign##int##bits x[size]) \\\n"
82 " { \\\n"
83 " v##sign##int##bits##x##size vec; \\\n"
84 " \\\n"
85 " vec.generic[0] = v##sign##int##bits##x##halfsize##_##name(x); \\\n"
86 " vec.generic[1] = v##sign##int##bits##x##halfsize##_##name(x + halfsize); \\\n"
87 " \\\n"
88 " return vec; \\\n"
89 " }\n"
90 "\n"
91 "#define VEC_DOUBLE_LOAD(sign, bits, size, halfsize) VEC_DOUBLE_LOAD_EX(load, sign, bits, size, halfsize)\n"
92 "#define VEC_DOUBLE_LOAD_ALIGNED(sign, bits, size, halfsize) VEC_DOUBLE_LOAD_EX(load_aligned, sign, bits, size, halfsize)\n"
93 "\n"
94 "#define VEC_DOUBLE_STORE_EX(name, sign, bits, size, halfsize) \\\n"
95 " VEC_FUNC_IMPL void v##sign##int##bits##x##size##_##name(v##sign##int##bits##x##size vec, vec_##sign##int##bits x[size]) \\\n"
96 " { \\\n"
97 " v##sign##int##bits##x##halfsize##_##name(vec.generic[0], x); \\\n"
98 " v##sign##int##bits##x##halfsize##_##name(vec.generic[1], x + halfsize); \\\n"
99 " }\n"
100 "\n"
101 "#define VEC_DOUBLE_STORE(sign, bits, size, halfsize) VEC_DOUBLE_STORE_EX(store, sign, bits, size, halfsize)\n"
102 "#define VEC_DOUBLE_STORE_ALIGNED(sign, bits, size, halfsize) VEC_DOUBLE_STORE_EX(store_aligned, sign, bits, size, halfsize)\n"
103 "\n"
104 "#define VEC_DOUBLE_OP(name, sign, bits, size, halfsize, secondsign) \\\n"
105 " VEC_FUNC_IMPL v##sign##int##bits##x##size v##sign##int##bits##x##size##_##name(v##sign##int##bits##x##size vec1, v##secondsign##int##bits##x##size vec2) \\\n"
106 " { \\\n"
107 " vec1.generic[0] = v##sign##int##bits##x##halfsize##_##name(vec1.generic[0], vec2.generic[0]); \\\n"
108 " vec1.generic[1] = v##sign##int##bits##x##halfsize##_##name(vec1.generic[1], vec2.generic[1]); \\\n"
109 " \\\n"
110 " return vec1; \\\n"
111 " }\n"
112 "\n"
113 "#define VEC_DOUBLE_ADD(sign, bits, size, halfsize) VEC_DOUBLE_OP(add, sign, bits, size, halfsize, sign)\n"
114 "#define VEC_DOUBLE_SUB(sign, bits, size, halfsize) VEC_DOUBLE_OP(sub, sign, bits, size, halfsize, sign)\n"
115 "#define VEC_DOUBLE_MUL(sign, bits, size, halfsize) VEC_DOUBLE_OP(mul, sign, bits, size, halfsize, sign)\n"
116 "#define VEC_DOUBLE_DIV(sign, bits, size, halfsize) VEC_DOUBLE_OP(div, sign, bits, size, halfsize, sign)\n"
117 "#define VEC_DOUBLE_MOD(sign, bits, size, halfsize) VEC_DOUBLE_OP(mod, sign, bits, size, halfsize, sign)\n"
118 "#define VEC_DOUBLE_AVG(sign, bits, size, halfsize) VEC_DOUBLE_OP(avg, sign, bits, size, halfsize, sign)\n"
119 "#define VEC_DOUBLE_LSHIFT(sign, bits, size, halfsize) VEC_DOUBLE_OP(lshift, sign, bits, size, halfsize, u)\n"
120 "#define VEC_DOUBLE_RSHIFT(sign, bits, size, halfsize) VEC_DOUBLE_OP(rshift, sign, bits, size, halfsize, u)\n"
121 "#define VEC_DOUBLE_LRSHIFT(sign, bits, size, halfsize) VEC_DOUBLE_OP(lrshift, sign, bits, size, halfsize, u)\n"
122 "#define VEC_DOUBLE_AND(sign, bits, size, halfsize) VEC_DOUBLE_OP(and, sign, bits, size, halfsize, sign)\n"
123 "#define VEC_DOUBLE_OR(sign, bits, size, halfsize) VEC_DOUBLE_OP(or, sign, bits, size, halfsize, sign)\n"
124 "#define VEC_DOUBLE_XOR(sign, bits, size, halfsize) VEC_DOUBLE_OP(xor, sign, bits, size, halfsize, sign)\n"
125 "#define VEC_DOUBLE_MIN(sign, bits, size, halfsize) VEC_DOUBLE_OP(min, sign, bits, size, halfsize, sign)\n"
126 "#define VEC_DOUBLE_MAX(sign, bits, size, halfsize) VEC_DOUBLE_OP(max, sign, bits, size, halfsize, sign)\n"
127 "#define VEC_DOUBLE_CMPLT(sign, bits, size, halfsize) VEC_DOUBLE_OP(cmplt, sign, bits, size, halfsize, sign)\n"
128 "#define VEC_DOUBLE_CMPLE(sign, bits, size, halfsize) VEC_DOUBLE_OP(cmple, sign, bits, size, halfsize, sign)\n"
129 "#define VEC_DOUBLE_CMPEQ(sign, bits, size, halfsize) VEC_DOUBLE_OP(cmpeq, sign, bits, size, halfsize, sign)\n"
130 "#define VEC_DOUBLE_CMPGE(sign, bits, size, halfsize) VEC_DOUBLE_OP(cmpge, sign, bits, size, halfsize, sign)\n"
131 "#define VEC_DOUBLE_CMPGT(sign, bits, size, halfsize) VEC_DOUBLE_OP(cmpgt, sign, bits, size, halfsize, sign)\n"
132 "\n"
133 "#define VEC_DOUBLE_NOT(sign, bits, size, halfsize) \\\n"
134 " VEC_FUNC_IMPL v##sign##int##bits##x##size v##sign##int##bits##x##size##_not(v##sign##int##bits##x##size vec) \\\n"
135 " { \\\n"
136 " vec.generic[0] = v##sign##int##bits##x##halfsize##_not(vec.generic[0]); \\\n"
137 " vec.generic[1] = v##sign##int##bits##x##halfsize##_not(vec.generic[1]); \\\n"
138 " \\\n"
139 " return vec; \\\n"
140 " }\n"
141 "\n"
142 "#endif /* VEC_IMPL_DOUBLE_H_ */ \n"
143 "\n"
144 "/* ------------------------------------------------------------------------ */\n"
145 "/* PREPROCESSOR HELL INCOMING */\n"
146 "";
147
148 static const char *footer =
149 "" /* nothing */;
150
151 /* ------------------------------------------------------------------------ */
152
153 static void print_generic_dbl_op(const char *op, int is_signed, int bits, int size)
154 {
155 printf(
156 "#if !defined(V%sINT%dx%d_%s_DEFINED) && defined(V%sINT%dx%d_%s_DEFINED)\n"
157 "VEC_DOUBLE_%s(%s, %d, %d, %d)\n"
158 "# define V%sINT%dx%d_%s_DEFINED\n"
159 "#endif\n\n",
160 (is_signed ? "" : "U"), bits, size, op, (is_signed ? "" : "U"), bits, size / 2, op,
161 op, (is_signed ? "/* nothing */" : "u"), bits, size, size / 2,
162 (is_signed ? "" : "U"), bits, size, op);
163 }
164
165 typedef void (*print_op_spec)(const char *op, int is_signed, int bits, int size);
166
167 static inline void print_ops(int is_signed, int bits, int size, print_op_spec print_op)
168 {
169 /* all supported operations here */
170 static const char *ops[] = {
171 "SPLAT",
172 "LOAD_ALIGNED",
173 "LOAD",
174 "STORE_ALIGNED",
175 "STORE",
176 "ADD",
177 "SUB",
178 "MUL",
179 "DIV",
180 "MOD",
181 "AVG",
182 "AND",
183 "OR",
184 "XOR",
185 "NOT",
186 "CMPLT",
187 "CMPEQ",
188 "CMPGT",
189 "CMPLE", /* these two must be after CMPLT and CMPGT respectfully, */
190 "CMPGE", /* because their definitions call those functions */
191 "MIN",
192 "MAX",
193 "RSHIFT",
194 "LRSHIFT",
195 "LSHIFT",
196 NULL,
197 };
198 int i;
199
200 printf("\n\n/* v%sint%dx%d */\n\n", (is_signed ? "u" : ""), bits, size);
201
202 for (i = 0; ops[i]; i++)
203 print_op(ops[i], is_signed, bits, size);
204 }
205
206 int main(void)
207 {
208 static struct {
209 int bits, size;
210 print_op_spec print_op;
211 } defs[] = {
212 /* -- 8-bit */
213 {8, 4, print_generic_dbl_op},
214 {8, 8, print_generic_dbl_op},
215 {8, 16, print_generic_dbl_op},
216 {8, 32, print_generic_dbl_op},
217 {8, 64, print_generic_dbl_op},
218
219 /* -- 16-bit */
220 {16, 4, print_generic_dbl_op},
221 {16, 8, print_generic_dbl_op},
222 {16, 16, print_generic_dbl_op},
223 {16, 32, print_generic_dbl_op},
224
225 /* -- 32-bit */
226 {32, 4, print_generic_dbl_op},
227 {32, 8, print_generic_dbl_op},
228 {32, 16, print_generic_dbl_op},
229
230 /* -- 64-bit */
231 {64, 4, print_generic_dbl_op},
232 {64, 8, print_generic_dbl_op},
233 };
234 int i;
235
236 puts(header);
237
238 for (i = 0; i < ARRAY_SIZE(defs); i++) {
239 print_ops(1, defs[i].bits, defs[i].size, defs[i].print_op);
240 print_ops(0, defs[i].bits, defs[i].size, defs[i].print_op);
241 }
242
243 puts(footer);
244 }