3 * Testing optimized 128-bit multipliers
5 * (c) 2017 Straylight/Edgeware
8 /*----- Licensing notice --------------------------------------------------*
10 * This file is part of Catacomb.
12 * Catacomb is free software; you can redistribute it and/or modify
13 * it under the terms of the GNU Library General Public License as
14 * published by the Free Software Foundation; either version 2 of the
15 * License, or (at your option) any later version.
17 * Catacomb is distributed in the hope that it will be useful,
18 * but WITHOUT ANY WARRANTY; without even the implied warranty of
19 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
20 * GNU Library General Public License for more details.
22 * You should have received a copy of the GNU Library General Public
23 * License along with Catacomb; if not, write to the Free
24 * Software Foundation, Inc., 59 Temple Place - Suite 330, Boston,
28 /*----- Header files ------------------------------------------------------*/
37 #include <mLib/macros.h>
38 #include <mLib/report.h>
39 #include <mLib/testrig.h>
46 /*----- CPU feature detection ---------------------------------------------*/
49 # define VARIANT _x86_sse2
51 static int cpu_features_p(void) { return (cpu_feature_p(CPUFEAT_X86_SSE2
)); }
55 # define VARIANT _amd64_sse2
57 static int cpu_features_p(void) { return (cpu_feature_p(CPUFEAT_X86_SSE2
)); }
61 # error "Unsupported CPU family."
71 typedef struct { mpw w
[4]; } p128
;
72 typedef struct { mpw w
[8]; } x128
;
73 typedef struct { mpd w
[6]; } carry
;
76 /*----- Test operation table ----------------------------------------------*/
79 /* a c u x v y z' y' c' */ \
80 _(dmul4, NIL, CARRY, P128, P128, P128, P128, P128, NIL, CARRY) \
81 _(dmla4, P128, CARRY, P128, P128, P128, P128, P128, NIL, CARRY) \
82 _(mul4, NIL, CARRY, NIL, P128, NIL, P128, P128, NIL, CARRY) \
83 _(mul4zc,NIL, NIL, NIL, P128, NIL, P128, P128, NIL, CARRY) \
84 _(mla4, P128, CARRY, NIL, P128, NIL, P128, P128, NIL, CARRY) \
85 _(mla4zc,P128, NIL, NIL, P128, NIL, P128, P128, NIL, CARRY) \
86 _(mmul4, NIL, NIL, P128, P128, P128, P128, P128, X128, CARRY) \
87 _(mmla4, P128, NIL, P128, P128, P128, P128, P128, X128, CARRY) \
88 _(mont4, P128, NIL, NIL, P128, NIL, P128, P128, X128, CARRY)
90 /*----- Assembler test interface ------------------------------------------*/
94 #define PARAM(v, ty) ty *v,
95 #define PARAM_NIL(v, q)
96 #define PARAM_P128(v, q) PARAM(v, q p128)
97 #define PARAM_X128(v, q) PARAM(v, q x128)
98 #define PARAM_CARRY(v, q) PARAM(v, q carry)
100 #define DECLSTUB(fn, tya, tyc, tyu, tyx, tyv, tyy, tyzz, tyyy, tycc) \
101 extern void test_##fn(PARAM_##tyzz(zz, EMPTY) PARAM_##tycc(cc, EMPTY) \
102 PARAM_##tyyy(yy, EMPTY) \
103 PARAM_##tyu(u, const) PARAM_##tyx(x, const) \
104 PARAM_##tyv(v, const) PARAM_##tyy(y, const) \
105 unsigned n, unsigned long long *cyv);
108 /*----- Conversion functions ----------------------------------------------*/
110 static mp
*combine_mpw(mp
*d
, const mpw
*v
, size_t n
, unsigned off
)
118 for (i
= 0, o
= 0; i
< n
; i
++, o
+= off
) {
119 w
[0] = v
[i
]; mp_build(&m
, w
, w
+ 1);
120 t
= mp_lsl(t
, &m
, o
); d
= mp_add(d
, d
, t
);
122 mp_drop(t
); return (d
);
125 static mp
*combine_mpd(mp
*d
, const mpd
*v
, size_t n
, unsigned off
)
133 for (i
= 0, o
= 0; i
< n
; i
++, o
+= off
) {
134 w
[0] = MPW(v
[i
]); w
[1] = MPW(v
[i
] >> MPW_BITS
); mp_build(&m
, w
, w
+ 2);
135 t
= mp_lsl(t
, &m
, o
); d
= mp_add(d
, d
, t
);
137 mp_drop(t
); return (d
);
140 #define DEFTYPE(ty, ld, st, nby, combfn, off) \
142 static void cvt_##ty(const char *buf, dstr *d) \
144 dstr dd = DSTR_INIT; \
149 type_hex.cvt(buf, &dd); \
150 if (dd.len != N(x->w)*nby) die(1, "invalid length for " #ty); \
151 dstr_ensure(d, sizeof(*x)); d->len = sizeof(*x); \
152 x = (ty *)d->buf; p = (const octet *)dd.buf; \
153 for (i = 0; i < N(x->w); i++) { x->w[i] = ld(p); p += nby; } \
157 static void dump_##ty(dstr *d, FILE *fp) \
159 dstr dd = DSTR_INIT; \
161 const ty *x = (const ty *)d->buf; \
162 mp *xx = combfn(MP_NEW, x->w, N(x->w), off); \
165 dstr_ensure(&dd, N(x->w)*nby); p = (octet *)dd.buf; \
166 for (i = 0; i < N(x->w); i++) { st(p, x->w[i]); p += nby; } \
167 dd.len = N(x->w)*nby; \
168 type_hex.dump(&dd, fp); \
171 fputs(" = 0x", fp); mp_writefile(xx, fp, 16); \
172 fputs(" = ", fp); mp_writefile(xx, fp, 10); \
176 static int eq_##ty(const ty *x, const ty *y) \
178 mp *xx = combfn(MP_NEW, x->w, N(x->w), off), \
179 *yy = combfn(MP_NEW, y->w, N(y->w), off); \
180 int rc = MP_EQ(xx, yy); \
181 MP_DROP(xx); MP_DROP(yy); \
185 static const struct test_type type_##ty = { cvt_##ty, dump_##ty };
187 DEFTYPE(p128
, LDW
, STW
, NWBY
, combine_mpw
, MPW_BITS
)
188 DEFTYPE(x128
, LDW
, STW
, NWBY
, combine_mpw
, MPW_BITS
/2)
189 DEFTYPE(carry
, LDD
, STD
, NDBY
, combine_mpd
, MPW_BITS
/2)
191 /*----- Test functions ----------------------------------------------------*/
193 #define DECL_IN(v, ty) \
194 dstr *d_##v = dp++; const ty *v = (const ty *)d_##v->buf;
195 #define DECL_IN_NIL(v)
196 #define DECL_IN_P128(v) DECL_IN(v, p128)
197 #define DECL_IN_X128(v) DECL_IN(v, x128)
198 #define DECL_IN_CARRY(v) DECL_IN(v, carry)
200 #define DECL_OUT(v, ty) \
201 dstr dd_##v = DSTR_INIT, *d_##v = &dd_##v; ty *v;
202 #define DECL_OUT_NIL(v)
203 #define DECL_OUT_P128(v) DECL_OUT(v, p128)
204 #define DECL_OUT_X128(v) DECL_OUT(v, x128)
205 #define DECL_OUT_CARRY(v) DECL_OUT(v, carry)
207 #define INIT_OUT(v, ty) \
208 dstr_ensure(d_##v, sizeof(ty)); v = (ty *)d_##v->buf;
209 #define INIT_OUT_NIL(v)
210 #define INIT_OUT_P128(v) INIT_OUT(v, p128)
211 #define INIT_OUT_X128(v) INIT_OUT(v, x128)
212 #define INIT_OUT_CARRY(v) INIT_OUT(v, carry)
216 #define ARG_P128(v) ARG(v)
217 #define ARG_X128(v) ARG(v)
218 #define ARG_CARRY(v) ARG(v)
220 #define CHECK(v, vv, ty) if (!eq_##ty(v, vv)) ok = 0;
221 #define CHECK_NIL(v, vv)
222 #define CHECK_P128(v, vv) CHECK(v, vv, p128)
223 #define CHECK_X128(v, vv) CHECK(v, vv, x128)
224 #define CHECK_CARRY(v, vv) CHECK(v, vv, carry)
226 #define DUMP(v, ty) \
227 fprintf(stderr, "\n\t%-6s = ", #v); dump_##ty(d_##v, stderr);
229 #define DUMP_P128(v) DUMP(v, p128)
230 #define DUMP_X128(v) DUMP(v, x128)
231 #define DUMP_CARRY(v) DUMP(v, carry)
233 #define COPY(v, vv, ty) *v = *vv;
234 #define COPY_NIL(v, vv)
235 #define COPY_P128(v, vv) COPY(v, vv, p128)
236 #define COPY_X128(v, vv) COPY(v, vv, x128)
237 #define COPY_CARRY(v, vv) COPY(v, vv, carry)
239 #define FREE_OUT(v, ty) dstr_destroy(d_##v);
240 #define FREE_OUT_NIL(v)
241 #define FREE_OUT_P128(v) FREE_OUT(v, p128)
242 #define FREE_OUT_X128(v) FREE_OUT(v, x128)
243 #define FREE_OUT_CARRY(v) FREE_OUT(v, carry)
245 #define DEFTESTFN(fn, tya, tyc, tyu, tyx, tyv, tyy, tyzz, tyyy, tycc) \
247 static int v##fn(dstr dv[]) \
256 DECL_IN_##tyzz(zz_exp) \
257 DECL_IN_##tyyy(yy_exp) \
258 DECL_IN_##tycc(cc_exp) \
259 DECL_OUT_##tyzz(zz_out) \
260 DECL_OUT_##tyyy(yy_out) \
261 DECL_OUT_##tycc(cc_out) \
262 unsigned long long cyv[1]; \
265 INIT_OUT_##tyzz(zz_out) \
266 INIT_OUT_##tyyy(yy_out) \
267 INIT_OUT_##tycc(cc_out) \
269 COPY_##tya(zz_out, a); \
270 COPY_##tyc(cc_out, c); \
271 test_##fn(zz_out ARG_##tycc(cc_out) ARG_##tyyy(yy_out) \
272 ARG_##tyu(u) ARG_##tyx(x) ARG_##tyv(v) ARG_##tyy(y), \
274 CHECK_##tyzz(zz_exp, zz_out) \
275 CHECK_##tyyy(yy_exp, yy_out) \
276 CHECK_##tycc(cc_exp, cc_out) \
279 fputs(#fn " failed", stderr); \
286 DUMP_##tyzz(zz_exp) \
287 DUMP_##tyzz(zz_out) \
288 DUMP_##tyyy(yy_exp) \
289 DUMP_##tyyy(yy_out) \
290 DUMP_##tycc(cc_exp) \
291 DUMP_##tycc(cc_out) \
292 fputc('\n', stderr); \
295 FREE_OUT_##tyzz(zz_out); \
296 FREE_OUT_##tyyy(yy_out); \
297 FREE_OUT_##tycc(cc_out); \
304 /*----- Main code ---------------------------------------------------------*/
307 #define P128 &type_p128,
308 #define X128 &type_x128,
309 #define CARRY &type_carry,
311 static test_chunk tests
[] = {
312 #define DEFCHUNK(fn, tya, tyc, tyu, tyx, tyv, tyy, tyzz, tyyy, tycc) \
313 { #fn, v##fn, { tya tyc tyu tyx tyv tyy tyzz tyyy tycc } },
319 int main(int argc
, char *argv
[])
322 if (!cpu_features_p())
323 { fprintf(stderr
, "required cpu feature not available\n"); exit(77); }
324 test_run(argc
, argv
, tests
, SRCDIR
"/t/mpx-mul4");
328 /*----- That's all, folks -------------------------------------------------*/