2 * Copyright (C) 2012, 2013, 2014
5 * Permission is hereby granted, free of charge, to any person obtaining a copy of
6 * this software and associated documentation files (the "Software"), to deal in
7 * the Software without restriction, including without limitation the rights to
8 * use, copy, modify, merge, publish, distribute, sublicense, and/or sell copies
9 * of the Software, and to permit persons to whom the Software is furnished to do
10 * so, subject to the following conditions:
12 * The above copyright notice and this permission notice shall be included in all
13 * copies or substantial portions of the Software.
15 * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
16 * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
17 * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
18 * AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
19 * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
20 * OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
29 #define FOLD_STRING_UNTRANSLATE_HTSIZE 1024
30 #define FOLD_STRING_DOTRANSLATE_HTSIZE 1024
32 /* The options to use for inexact and arithmetic exceptions */
33 #define FOLD_ROUNDING SFLOAT_ROUND_NEAREST_EVEN
34 #define FOLD_TINYNESS SFLOAT_TBEFORE
37 * The constant folder is also responsible for validating if the constant
38 * expressions produce valid results. We cannot trust the FPU control
39 * unit for these exceptions because setting FPU control words might not
40 * work. Systems can set and enforce FPU modes of operation. It's also valid
41 * for libc's to simply ignore FPU exceptions. For instance ARM CPUs in
42 * glibc. We implement some trivial and IEE 754 conformant functions which
43 * emulate those operations. This is an entierly optional compiler feature
44 * which shouldn't be enabled for anything other than performing strict
45 * passes on constant expressions since it's quite slow.
47 typedef uint32_t sfloat_t;
59 SFLOAT_UNDERFLOW = 16,
61 } sfloat_exceptionflags_t;
64 SFLOAT_ROUND_NEAREST_EVEN,
68 } sfloat_roundingmode_t;
76 sfloat_roundingmode_t roundingmode;
77 sfloat_exceptionflags_t exceptionflags;
78 sfloat_tdetect_t tiny;
81 /* Count of leading zero bits before the most-significand 1 bit. */
83 /* MSVC has an intrinsic for this */
84 static GMQCC_INLINE uint32_t sfloat_clz(uint32_t x) {
86 _BitScanForward(&r, x);
89 # define SFLOAT_CLZ(X, SUB) \
90 (sfloat_clz((X)) - (SUB))
91 #elif defined(__GNUC__) || defined(__CLANG__)
92 /* Clang and GCC have a builtin for this */
93 # define SFLOAT_CLZ(X, SUB) \
94 (__builtin_clz((X)) - (SUB))
97 static GMQCC_INLINE uint32_t sfloat_popcnt(uint32_t x) {
98 x -= ((x >> 1) & 0x55555555);
99 x = (((x >> 2) & 0x33333333) + (x & 0x33333333));
100 x = (((x >> 4) + x) & 0x0F0F0F0F);
103 return x & 0x0000003F;
105 static GMQCC_INLINE uint32_t sfloat_clz(uint32_t x) {
111 return 32 - sfloat_popcnt(x);
113 # define SFLOAT_CLZ(X, SUB) \
114 (sfloat_clz((X) - (SUB)))
117 /* The value of a NaN */
118 #define SFLOAT_NAN 0xFFC00000
120 #define SFLOAT_ISNAN(A) \
121 (0xFF000000 < (uint32_t)((A) << 1))
122 /* Test if signaling NaN */
123 #define SFLOAT_ISSNAN(A) \
124 (((((A) >> 22) & 0x1FF) == 0x1FE) && ((A) & 0x003FFFFF))
125 /* Raise exception */
126 #define SFLOAT_RAISE(STATE, FLAGS) \
127 ((STATE)->exceptionflags = (sfloat_exceptionflags_t)((STATE)->exceptionflags | (FLAGS)))
129 * Shifts `A' right `COUNT' bits. Non-zero bits are stored in LSB. Size
130 * sets the arbitrarly-large limit.
132 #define SFLOAT_SHIFT(SIZE, A, COUNT, Z) \
133 *(Z) = ((COUNT) == 0) \
135 : (((COUNT) < (SIZE)) \
136 ? ((A) >> (COUNT)) | (((A) << ((-(COUNT)) & ((SIZE) - 1))) != 0) \
138 /* Extract fractional component */
139 #define SFLOAT_EXTRACT_FRAC(X) \
140 ((uint32_t)((X) & 0x007FFFFF))
141 /* Extract exponent component */
142 #define SFLOAT_EXTRACT_EXP(X) \
143 ((int16_t)((X) >> 23) & 0xFF)
144 /* Extract sign bit */
145 #define SFLOAT_EXTRACT_SIGN(X) \
147 /* Normalize a subnormal */
148 #define SFLOAT_SUBNORMALIZE(SA, Z, SZ) \
149 (void)(*(SZ) = (SA) << SFLOAT_CLZ((SA), 8), *(SZ) = 1 - SFLOAT_CLZ((SA), 8))
151 * Pack sign, exponent and significand and produce a float.
153 * Integer portions of the significand are added to the exponent. The
154 * exponent input should be one less than the result exponent whenever
155 * the significand is normalized since normalized significand will
156 * always have an integer portion of value one.
158 #define SFLOAT_PACK(SIGN, EXP, SIG) \
159 (sfloat_t)((((uint32_t)(SIGN)) << 31) + (((uint32_t)(EXP)) << 23) + (SIG))
161 /* Calculate NaN. If either operands are signaling then raise invalid */
162 static sfloat_t sfloat_propagate_nan(sfloat_state_t *state, sfloat_t a, sfloat_t b) {
163 bool isnan_a = SFLOAT_ISNAN(a);
164 bool issnan_a = SFLOAT_ISSNAN(a);
165 bool isnan_b = SFLOAT_ISNAN(b);
166 bool issnan_b = SFLOAT_ISSNAN(b);
171 if (issnan_a | issnan_b)
172 SFLOAT_RAISE(state, SFLOAT_INEXACT);
176 return isnan_b ? b : a;
177 } else if (isnan_a) {
178 if (issnan_b | !isnan_b)
181 if ((uint32_t)(a << 1) < (uint32_t)(b << 1)) return b;
182 if ((uint32_t)(b << 1) < (uint32_t)(a << 1)) return a;
183 return (a < b) ? a : b;
189 static sfloat_t SFLOAT_PACK_round(sfloat_state_t *state, bool sign_z, int16_t exp_z, uint32_t sig_z) {
190 sfloat_roundingmode_t mode = state->roundingmode;
191 bool even = !!(mode == SFLOAT_ROUND_NEAREST_EVEN);
192 unsigned char increment = 0x40;
193 unsigned char bits = sig_z & 0x7F;
196 if (mode == SFLOAT_ROUND_TO_ZERO)
201 if (mode == SFLOAT_ROUND_UP)
204 if (mode == SFLOAT_ROUND_DOWN)
210 if (0xFD <= (uint16_t)exp_z) {
211 if ((0xFD < exp_z) || ((exp_z == 0xFD) && ((int32_t)(sig_z + increment) < 0))) {
212 SFLOAT_RAISE(state, SFLOAT_OVERFLOW | SFLOAT_INEXACT);
213 return SFLOAT_PACK(sign_z, 0xFF, 0) - (increment == 0);
216 /* Check for underflow */
217 bool tiny = (state->tiny == SFLOAT_TBEFORE) || (exp_z < -1) || (sig_z + increment < 0x80000000);
218 SFLOAT_SHIFT(32, sig_z, -exp_z, &sig_z);
222 SFLOAT_RAISE(state, SFLOAT_UNDERFLOW);
227 * Significand has point between bits 30 and 29, 7 bits to the left of
228 * the usual place. This shifted significand has to be normalized
229 * or smaller, if it isn't the exponent must be zero, in which case
230 * no rounding occurs since the result will be a subnormal.
233 SFLOAT_RAISE(state, SFLOAT_INEXACT);
234 sig_z = (sig_z + increment) >> 7;
235 sig_z &= ~(((bits ^ 0x40) == 0) & even);
238 return SFLOAT_PACK(sign_z, exp_z, sig_z);
241 /* Normalized round and pack */
242 static sfloat_t SFLOAT_PACK_normal(sfloat_state_t *state, bool sign_z, int16_t exp_z, uint32_t sig_z) {
243 unsigned char c = SFLOAT_CLZ(sig_z, 1);
244 return SFLOAT_PACK_round(state, sign_z, exp_z - c, sig_z << c);
247 static sfloat_t sfloat_add_impl(sfloat_state_t *state, sfloat_t a, sfloat_t b, bool sign_z) {
248 int16_t exp_a = SFLOAT_EXTRACT_EXP(a);
249 int16_t exp_b = SFLOAT_EXTRACT_EXP(b);
251 int16_t exp_d = exp_a - exp_b;
252 uint32_t sig_a = SFLOAT_EXTRACT_FRAC(a) << 6;
253 uint32_t sig_b = SFLOAT_EXTRACT_FRAC(b) << 6;
258 return sig_a ? sfloat_propagate_nan(state, a, b) : a;
263 SFLOAT_SHIFT(32, sig_b, exp_d, &sig_b);
265 } else if (exp_d < 0) {
267 return sig_b ? sfloat_propagate_nan(state, a, b) : SFLOAT_PACK(sign_z, 0xFF, 0);
272 SFLOAT_SHIFT(32, sig_a, -exp_d, &sig_a);
276 return (sig_a | sig_b) ? sfloat_propagate_nan(state, a, b) : a;
278 return SFLOAT_PACK(sign_z, 0, (sig_a + sig_b) >> 6);
279 sig_z = 0x40000000 + sig_a + sig_b;
284 sig_z = (sig_a + sig_b) << 1;
286 if ((int32_t)sig_z < 0) {
287 sig_z = sig_a + sig_b;
291 return SFLOAT_PACK_round(state, sign_z, exp_z, sig_z);
294 static sfloat_t sfloat_sub_impl(sfloat_state_t *state, sfloat_t a, sfloat_t b, bool sign_z) {
295 int16_t exp_a = SFLOAT_EXTRACT_EXP(a);
296 int16_t exp_b = SFLOAT_EXTRACT_EXP(b);
298 int16_t exp_d = exp_a - exp_b;
299 uint32_t sig_a = SFLOAT_EXTRACT_FRAC(a) << 7;
300 uint32_t sig_b = SFLOAT_EXTRACT_FRAC(b) << 7;
303 if (0 < exp_d) goto exp_greater_a;
304 if (exp_d < 0) goto exp_greater_b;
308 return sfloat_propagate_nan(state, a, b);
309 SFLOAT_RAISE(state, SFLOAT_INVALID);
316 if (sig_b < sig_a) goto greater_a;
317 if (sig_a < sig_b) goto greater_b;
319 return SFLOAT_PACK(state->roundingmode == SFLOAT_ROUND_DOWN, 0, 0);
323 return (sig_b) ? sfloat_propagate_nan(state, a, b) : SFLOAT_PACK(sign_z ^ 1, 0xFF, 0);
328 SFLOAT_SHIFT(32, sig_a, -exp_d, &sig_a);
331 sig_z = sig_b - sig_a;
338 return (sig_a) ? sfloat_propagate_nan(state, a, b) : a;
343 SFLOAT_SHIFT(32, sig_b, exp_d, &sig_b);
346 sig_z = sig_a - sig_b;
351 return SFLOAT_PACK_normal(state, sign_z, exp_z, sig_z);
354 static GMQCC_INLINE sfloat_t sfloat_add(sfloat_state_t *state, sfloat_t a, sfloat_t b) {
355 bool sign_a = SFLOAT_EXTRACT_SIGN(a);
356 bool sign_b = SFLOAT_EXTRACT_SIGN(b);
357 return (sign_a == sign_b) ? sfloat_add_impl(state, a, b, sign_a)
358 : sfloat_sub_impl(state, a, b, sign_a);
361 static GMQCC_INLINE sfloat_t sfloat_sub(sfloat_state_t *state, sfloat_t a, sfloat_t b) {
362 bool sign_a = SFLOAT_EXTRACT_SIGN(a);
363 bool sign_b = SFLOAT_EXTRACT_SIGN(b);
364 return (sign_a == sign_b) ? sfloat_sub_impl(state, a, b, sign_a)
365 : sfloat_add_impl(state, a, b, sign_a);
368 static sfloat_t sfloat_mul(sfloat_state_t *state, sfloat_t a, sfloat_t b) {
369 int16_t exp_a = SFLOAT_EXTRACT_EXP(a);
370 int16_t exp_b = SFLOAT_EXTRACT_EXP(b);
372 uint32_t sig_a = SFLOAT_EXTRACT_FRAC(a);
373 uint32_t sig_b = SFLOAT_EXTRACT_FRAC(b);
375 uint64_t sig_z64 = 0;
376 bool sign_a = SFLOAT_EXTRACT_SIGN(a);
377 bool sign_b = SFLOAT_EXTRACT_SIGN(b);
378 bool sign_z = sign_a ^ sign_b;
381 if (sig_a || ((exp_b == 0xFF) && sig_b))
382 return sfloat_propagate_nan(state, a, b);
383 if ((exp_b | sig_b) == 0) {
384 SFLOAT_RAISE(state, SFLOAT_INVALID);
387 return SFLOAT_PACK(sign_z, 0xFF, 0);
391 return sfloat_propagate_nan(state, a, b);
392 if ((exp_a | sig_a) == 0) {
393 SFLOAT_RAISE(state, SFLOAT_INVALID);
396 return SFLOAT_PACK(sign_z, 0xFF, 0);
400 return SFLOAT_PACK(sign_z, 0, 0);
401 SFLOAT_SUBNORMALIZE(sig_a, &exp_a, &sig_a);
405 return SFLOAT_PACK(sign_z, 0, 0);
406 SFLOAT_SUBNORMALIZE(sig_b, &exp_b, &sig_b);
408 exp_z = exp_a + exp_b - 0x7F;
409 sig_a = (sig_a | 0x00800000) << 7;
410 sig_b = (sig_b | 0x00800000) << 8;
411 SFLOAT_SHIFT(64, ((uint64_t)sig_a) * sig_b, 32, &sig_z64);
413 if (0 <= (int32_t)(sig_z << 1)) {
417 return SFLOAT_PACK_round(state, sign_z, exp_z, sig_z);
420 static sfloat_t sfloat_div(sfloat_state_t *state, sfloat_t a, sfloat_t b) {
421 int16_t exp_a = SFLOAT_EXTRACT_EXP(a);
422 int16_t exp_b = SFLOAT_EXTRACT_EXP(b);
424 uint32_t sig_a = SFLOAT_EXTRACT_FRAC(a);
425 uint32_t sig_b = SFLOAT_EXTRACT_FRAC(b);
427 bool sign_a = SFLOAT_EXTRACT_SIGN(a);
428 bool sign_b = SFLOAT_EXTRACT_SIGN(b);
429 bool sign_z = sign_a ^ sign_b;
433 return sfloat_propagate_nan(state, a, b);
436 return sfloat_propagate_nan(state, a, b);
437 SFLOAT_RAISE(state, SFLOAT_INVALID);
440 return SFLOAT_PACK(sign_z, 0xFF, 0);
443 return (sig_b) ? sfloat_propagate_nan(state, a, b) : SFLOAT_PACK(sign_z, 0, 0);
446 if ((exp_a | sig_a) == 0) {
447 SFLOAT_RAISE(state, SFLOAT_INVALID);
450 SFLOAT_RAISE(state, SFLOAT_DIVBYZERO);
451 return SFLOAT_PACK(sign_z, 0xFF, 0);
453 SFLOAT_SUBNORMALIZE(sig_b, &exp_b, &sig_b);
457 return SFLOAT_PACK(sign_z, 0, 0);
458 SFLOAT_SUBNORMALIZE(sig_a, &exp_a, &sig_a);
460 exp_z = exp_a - exp_b + 0x7D;
461 sig_a = (sig_a | 0x00800000) << 7;
462 sig_b = (sig_b | 0x00800000) << 8;
463 if (sig_b <= (sig_a + sig_a)) {
467 sig_z = (((uint64_t)sig_a) << 32) / sig_b;
468 if ((sig_z & 0x3F) == 0)
469 sig_z |= ((uint64_t)sig_b * sig_z != ((uint64_t)sig_a) << 32);
470 return SFLOAT_PACK_round(state, sign_z, exp_z, sig_z);
473 static GMQCC_INLINE void sfloat_check(lex_ctx_t ctx, sfloat_state_t *state, const char *vec) {
474 /* Exception comes from vector component */
476 if (state->exceptionflags & SFLOAT_DIVBYZERO)
477 compile_error(ctx, "division by zero in `%s' component", vec);
478 if (state->exceptionflags & SFLOAT_INVALID)
479 compile_error(ctx, "undefined (inf) in `%s' component", vec);
480 if (state->exceptionflags & SFLOAT_OVERFLOW)
481 compile_error(ctx, "arithmetic overflow in `%s' component", vec);
482 if (state->exceptionflags & SFLOAT_UNDERFLOW)
483 compile_error(ctx, "arithmetic underflow in `%s' component", vec);
486 if (state->exceptionflags & SFLOAT_DIVBYZERO)
487 compile_error(ctx, "division by zero");
488 if (state->exceptionflags & SFLOAT_INVALID)
489 compile_error(ctx, "undefined (inf)");
490 if (state->exceptionflags & SFLOAT_OVERFLOW)
491 compile_error(ctx, "arithmetic overflow");
492 if (state->exceptionflags & SFLOAT_UNDERFLOW)
493 compile_error(ctx, "arithmetic underflow");
496 static GMQCC_INLINE void sfloat_init(sfloat_state_t *state) {
497 state->exceptionflags = SFLOAT_NOEXCEPT;
498 state->roundingmode = FOLD_ROUNDING;
499 state->tiny = FOLD_TINYNESS;
503 * There is two stages to constant folding in GMQCC: there is the parse
504 * stage constant folding, where, witht he help of the AST, operator
505 * usages can be constant folded. Then there is the constant folding
506 * in the IR for things like eliding if statements, can occur.
508 * This file is thus, split into two parts.
511 #define isfloat(X) (((ast_expression*)(X))->vtype == TYPE_FLOAT)
512 #define isvector(X) (((ast_expression*)(X))->vtype == TYPE_VECTOR)
513 #define isstring(X) (((ast_expression*)(X))->vtype == TYPE_STRING)
514 #define isarray(X) (((ast_expression*)(X))->vtype == TYPE_ARRAY)
515 #define isfloats(X,Y) (isfloat (X) && isfloat (Y))
518 * Implementation of basic vector math for vec3_t, for trivial constant
521 * TODO: gcc/clang hinting for autovectorization
537 sfloat_state_t state[3];
540 static GMQCC_INLINE vec3_soft_t vec3_soft_convert(vec3_t vec) {
548 static GMQCC_INLINE bool vec3_soft_exception(vec3_soft_state_t *vstate, size_t index) {
549 sfloat_exceptionflags_t flags = vstate->state[index].exceptionflags;
550 if (flags & SFLOAT_DIVBYZERO) return true;
551 if (flags & SFLOAT_INVALID) return true;
552 if (flags & SFLOAT_OVERFLOW) return true;
553 if (flags & SFLOAT_UNDERFLOW) return true;
557 static GMQCC_INLINE void vec3_soft_eval(vec3_soft_state_t *state,
558 sfloat_t (*callback)(sfloat_state_t *, sfloat_t, sfloat_t),
562 vec3_soft_t sa = vec3_soft_convert(a);
563 vec3_soft_t sb = vec3_soft_convert(b);
564 callback(&state->state[0], sa.x.s, sb.x.s);
565 if (vec3_soft_exception(state, 0)) state->faults = (vec3_comp_t)(state->faults | VEC_COMP_X);
566 callback(&state->state[1], sa.y.s, sb.y.s);
567 if (vec3_soft_exception(state, 1)) state->faults = (vec3_comp_t)(state->faults | VEC_COMP_Y);
568 callback(&state->state[2], sa.z.s, sb.z.s);
569 if (vec3_soft_exception(state, 2)) state->faults = (vec3_comp_t)(state->faults | VEC_COMP_Z);
572 static GMQCC_INLINE void vec3_check_except(vec3_t a,
575 sfloat_t (*callback)(sfloat_state_t *, sfloat_t, sfloat_t))
577 vec3_soft_state_t state;
579 if (!OPTS_FLAG(ARITHMETIC_EXCEPTIONS))
582 sfloat_init(&state.state[0]);
583 sfloat_init(&state.state[1]);
584 sfloat_init(&state.state[2]);
586 vec3_soft_eval(&state, callback, a, b);
587 if (state.faults & VEC_COMP_X) sfloat_check(ctx, &state.state[0], "x");
588 if (state.faults & VEC_COMP_Y) sfloat_check(ctx, &state.state[1], "y");
589 if (state.faults & VEC_COMP_Z) sfloat_check(ctx, &state.state[2], "z");
592 static GMQCC_INLINE vec3_t vec3_add(lex_ctx_t ctx, vec3_t a, vec3_t b) {
594 vec3_check_except(a, b, ctx, &sfloat_add);
601 static GMQCC_INLINE vec3_t vec3_sub(lex_ctx_t ctx, vec3_t a, vec3_t b) {
603 vec3_check_except(a, b, ctx, &sfloat_sub);
610 static GMQCC_INLINE vec3_t vec3_neg(vec3_t a) {
618 static GMQCC_INLINE vec3_t vec3_or(vec3_t a, vec3_t b) {
620 out.x = (qcfloat_t)(((qcint_t)a.x) | ((qcint_t)b.x));
621 out.y = (qcfloat_t)(((qcint_t)a.y) | ((qcint_t)b.y));
622 out.z = (qcfloat_t)(((qcint_t)a.z) | ((qcint_t)b.z));
626 static GMQCC_INLINE vec3_t vec3_orvf(vec3_t a, qcfloat_t b) {
628 out.x = (qcfloat_t)(((qcint_t)a.x) | ((qcint_t)b));
629 out.y = (qcfloat_t)(((qcint_t)a.y) | ((qcint_t)b));
630 out.z = (qcfloat_t)(((qcint_t)a.z) | ((qcint_t)b));
634 static GMQCC_INLINE vec3_t vec3_and(vec3_t a, vec3_t b) {
636 out.x = (qcfloat_t)(((qcint_t)a.x) & ((qcint_t)b.x));
637 out.y = (qcfloat_t)(((qcint_t)a.y) & ((qcint_t)b.y));
638 out.z = (qcfloat_t)(((qcint_t)a.z) & ((qcint_t)b.z));
642 static GMQCC_INLINE vec3_t vec3_andvf(vec3_t a, qcfloat_t b) {
644 out.x = (qcfloat_t)(((qcint_t)a.x) & ((qcint_t)b));
645 out.y = (qcfloat_t)(((qcint_t)a.y) & ((qcint_t)b));
646 out.z = (qcfloat_t)(((qcint_t)a.z) & ((qcint_t)b));
650 static GMQCC_INLINE vec3_t vec3_xor(vec3_t a, vec3_t b) {
652 out.x = (qcfloat_t)(((qcint_t)a.x) ^ ((qcint_t)b.x));
653 out.y = (qcfloat_t)(((qcint_t)a.y) ^ ((qcint_t)b.y));
654 out.z = (qcfloat_t)(((qcint_t)a.z) ^ ((qcint_t)b.z));
658 static GMQCC_INLINE vec3_t vec3_xorvf(vec3_t a, qcfloat_t b) {
660 out.x = (qcfloat_t)(((qcint_t)a.x) ^ ((qcint_t)b));
661 out.y = (qcfloat_t)(((qcint_t)a.y) ^ ((qcint_t)b));
662 out.z = (qcfloat_t)(((qcint_t)a.z) ^ ((qcint_t)b));
666 static GMQCC_INLINE vec3_t vec3_not(vec3_t a) {
674 static GMQCC_INLINE qcfloat_t vec3_mulvv(lex_ctx_t ctx, vec3_t a, vec3_t b) {
680 if (!OPTS_FLAG(ARITHMETIC_EXCEPTIONS))
683 sa = vec3_soft_convert(a);
684 sb = vec3_soft_convert(b);
692 r[0] = sfloat_mul(&s[0], sa.x.s, sb.x.s);
693 r[1] = sfloat_mul(&s[1], sa.y.s, sb.y.s);
694 r[2] = sfloat_mul(&s[2], sa.z.s, sb.z.s);
695 r[3] = sfloat_add(&s[3], r[0], r[1]);
696 r[4] = sfloat_add(&s[4], r[3], r[2]);
698 sfloat_check(ctx, &s[0], NULL);
699 sfloat_check(ctx, &s[1], NULL);
700 sfloat_check(ctx, &s[2], NULL);
701 sfloat_check(ctx, &s[3], NULL);
702 sfloat_check(ctx, &s[4], NULL);
705 return (a.x * b.x + a.y * b.y + a.z * b.z);
708 static GMQCC_INLINE vec3_t vec3_mulvf(lex_ctx_t ctx, vec3_t a, qcfloat_t b) {
714 if (!OPTS_FLAG(ARITHMETIC_EXCEPTIONS))
717 sa = vec3_soft_convert(a);
723 sfloat_mul(&s[0], sa.x.s, sb.s);
724 sfloat_mul(&s[1], sa.y.s, sb.s);
725 sfloat_mul(&s[2], sa.z.s, sb.s);
727 sfloat_check(ctx, &s[0], "x");
728 sfloat_check(ctx, &s[1], "y");
729 sfloat_check(ctx, &s[2], "z");
738 static GMQCC_INLINE bool vec3_cmp(vec3_t a, vec3_t b) {
744 static GMQCC_INLINE vec3_t vec3_create(float x, float y, float z) {
752 static GMQCC_INLINE qcfloat_t vec3_notf(vec3_t a) {
753 return (!a.x && !a.y && !a.z);
756 static GMQCC_INLINE bool vec3_pbool(vec3_t a) {
757 return (a.x || a.y || a.z);
760 static GMQCC_INLINE vec3_t vec3_cross(lex_ctx_t ctx, vec3_t a, vec3_t b) {
767 if (!OPTS_FLAG(ARITHMETIC_EXCEPTIONS))
770 sa = vec3_soft_convert(a);
771 sb = vec3_soft_convert(b);
783 r[0] = sfloat_mul(&s[0], sa.y.s, sb.z.s);
784 r[1] = sfloat_mul(&s[1], sa.z.s, sb.y.s);
785 r[2] = sfloat_mul(&s[2], sa.z.s, sb.x.s);
786 r[3] = sfloat_mul(&s[3], sa.x.s, sb.z.s);
787 r[4] = sfloat_mul(&s[4], sa.x.s, sb.y.s);
788 r[5] = sfloat_mul(&s[5], sa.y.s, sb.x.s);
789 r[6] = sfloat_sub(&s[6], r[0], r[1]);
790 r[7] = sfloat_sub(&s[7], r[2], r[3]);
791 r[8] = sfloat_sub(&s[8], r[4], r[5]);
793 sfloat_check(ctx, &s[0], NULL);
794 sfloat_check(ctx, &s[1], NULL);
795 sfloat_check(ctx, &s[2], NULL);
796 sfloat_check(ctx, &s[3], NULL);
797 sfloat_check(ctx, &s[4], NULL);
798 sfloat_check(ctx, &s[5], NULL);
799 sfloat_check(ctx, &s[6], "x");
800 sfloat_check(ctx, &s[7], "y");
801 sfloat_check(ctx, &s[8], "z");
804 out.x = a.y * b.z - a.z * b.y;
805 out.y = a.z * b.x - a.x * b.z;
806 out.z = a.x * b.y - a.y * b.x;
810 static lex_ctx_t fold_ctx(fold_t *fold) {
812 if (fold->parser->lex)
813 return parser_ctx(fold->parser);
815 memset(&ctx, 0, sizeof(ctx));
819 static GMQCC_INLINE bool fold_immediate_true(fold_t *fold, ast_value *v) {
820 switch (v->expression.vtype) {
822 return !!v->constval.vfloat;
824 return !!v->constval.vint;
826 if (OPTS_FLAG(CORRECT_LOGIC))
827 return vec3_pbool(v->constval.vvec);
828 return !!(v->constval.vvec.x);
830 if (!v->constval.vstring)
832 if (OPTS_FLAG(TRUE_EMPTY_STRINGS))
834 return !!v->constval.vstring[0];
836 compile_error(fold_ctx(fold), "internal error: fold_immediate_true on invalid type");
839 return !!v->constval.vfunc;
842 /* Handy macros to determine if an ast_value can be constant folded. */
843 #define fold_can_1(X) \
844 (ast_istype(((ast_expression*)(X)), ast_value) && (X)->hasvalue && ((X)->cvq == CV_CONST) && \
845 ((ast_expression*)(X))->vtype != TYPE_FUNCTION)
847 #define fold_can_2(X, Y) (fold_can_1(X) && fold_can_1(Y))
849 #define fold_immvalue_float(E) ((E)->constval.vfloat)
850 #define fold_immvalue_vector(E) ((E)->constval.vvec)
851 #define fold_immvalue_string(E) ((E)->constval.vstring)
853 fold_t *fold_init(parser_t *parser) {
854 fold_t *fold = (fold_t*)mem_a(sizeof(fold_t));
855 fold->parser = parser;
856 fold->imm_float = NULL;
857 fold->imm_vector = NULL;
858 fold->imm_string = NULL;
859 fold->imm_string_untranslate = util_htnew(FOLD_STRING_UNTRANSLATE_HTSIZE);
860 fold->imm_string_dotranslate = util_htnew(FOLD_STRING_DOTRANSLATE_HTSIZE);
863 * prime the tables with common constant values at constant
866 (void)fold_constgen_float (fold, 0.0f, false);
867 (void)fold_constgen_float (fold, 1.0f, false);
868 (void)fold_constgen_float (fold, -1.0f, false);
869 (void)fold_constgen_float (fold, 2.0f, false);
871 (void)fold_constgen_vector(fold, vec3_create(0.0f, 0.0f, 0.0f));
872 (void)fold_constgen_vector(fold, vec3_create(-1.0f, -1.0f, -1.0f));
877 bool fold_generate(fold_t *fold, ir_builder *ir) {
878 /* generate globals for immediate folded values */
882 for (i = 0; i < vec_size(fold->imm_float); ++i)
883 if (!ast_global_codegen ((cur = fold->imm_float[i]), ir, false)) goto err;
884 for (i = 0; i < vec_size(fold->imm_vector); ++i)
885 if (!ast_global_codegen((cur = fold->imm_vector[i]), ir, false)) goto err;
886 for (i = 0; i < vec_size(fold->imm_string); ++i)
887 if (!ast_global_codegen((cur = fold->imm_string[i]), ir, false)) goto err;
892 con_out("failed to generate global %s\n", cur->name);
893 ir_builder_delete(ir);
897 void fold_cleanup(fold_t *fold) {
900 for (i = 0; i < vec_size(fold->imm_float); ++i) ast_delete(fold->imm_float[i]);
901 for (i = 0; i < vec_size(fold->imm_vector); ++i) ast_delete(fold->imm_vector[i]);
902 for (i = 0; i < vec_size(fold->imm_string); ++i) ast_delete(fold->imm_string[i]);
904 vec_free(fold->imm_float);
905 vec_free(fold->imm_vector);
906 vec_free(fold->imm_string);
908 util_htdel(fold->imm_string_untranslate);
909 util_htdel(fold->imm_string_dotranslate);
914 ast_expression *fold_constgen_float(fold_t *fold, qcfloat_t value, bool inexact) {
915 ast_value *out = NULL;
918 for (i = 0; i < vec_size(fold->imm_float); i++) {
919 if (!memcmp(&fold->imm_float[i]->constval.vfloat, &value, sizeof(qcfloat_t)))
920 return (ast_expression*)fold->imm_float[i];
923 out = ast_value_new(fold_ctx(fold), "#IMMEDIATE", TYPE_FLOAT);
925 out->hasvalue = true;
926 out->inexact = inexact;
927 out->constval.vfloat = value;
929 vec_push(fold->imm_float, out);
931 return (ast_expression*)out;
934 ast_expression *fold_constgen_vector(fold_t *fold, vec3_t value) {
938 for (i = 0; i < vec_size(fold->imm_vector); i++) {
939 if (vec3_cmp(fold->imm_vector[i]->constval.vvec, value))
940 return (ast_expression*)fold->imm_vector[i];
943 out = ast_value_new(fold_ctx(fold), "#IMMEDIATE", TYPE_VECTOR);
945 out->hasvalue = true;
946 out->constval.vvec = value;
948 vec_push(fold->imm_vector, out);
950 return (ast_expression*)out;
953 ast_expression *fold_constgen_string(fold_t *fold, const char *str, bool translate) {
954 hash_table_t *table = (translate) ? fold->imm_string_untranslate : fold->imm_string_dotranslate;
955 ast_value *out = NULL;
956 size_t hash = util_hthash(table, str);
958 if ((out = (ast_value*)util_htgeth(table, str, hash)))
959 return (ast_expression*)out;
963 util_snprintf(name, sizeof(name), "dotranslate_%lu", (unsigned long)(fold->parser->translated++));
964 out = ast_value_new(parser_ctx(fold->parser), name, TYPE_STRING);
965 out->expression.flags |= AST_FLAG_INCLUDE_DEF; /* def needs to be included for translatables */
967 out = ast_value_new(fold_ctx(fold), "#IMMEDIATE", TYPE_STRING);
970 out->hasvalue = true;
972 out->constval.vstring = parser_strdup(str);
974 vec_push(fold->imm_string, out);
975 util_htseth(table, str, hash, out);
977 return (ast_expression*)out;
981 static GMQCC_INLINE ast_expression *fold_op_mul_vec(fold_t *fold, vec3_t vec, ast_value *sel, const char *set) {
983 * vector-component constant folding works by matching the component sets
984 * to eliminate expensive operations on whole-vectors (3 components at runtime).
985 * to achive this effect in a clean manner this function generalizes the
986 * values through the use of a set paramater, which is used as an indexing method
987 * for creating the elided ast binary expression.
989 * Consider 'n 0 0' where y, and z need to be tested for 0, and x is
990 * used as the value in a binary operation generating an INSTR_MUL instruction,
991 * to acomplish the indexing of the correct component value we use set[0], set[1], set[2]
992 * as x, y, z, where the values of those operations return 'x', 'y', 'z'. Because
993 * of how ASCII works we can easily deliniate:
994 * vec.z is the same as set[2]-'x' for when set[2] is 'z', 'z'-'x' results in a
995 * literal value of 2, using this 2, we know that taking the address of vec->x (float)
996 * and indxing it with this literal will yeild the immediate address of that component
998 * Of course more work needs to be done to generate the correct index for the ast_member_new
999 * call, which is no problem: set[0]-'x' suffices that job.
1001 qcfloat_t x = (&vec.x)[set[0]-'x'];
1002 qcfloat_t y = (&vec.x)[set[1]-'x'];
1003 qcfloat_t z = (&vec.x)[set[2]-'x'];
1006 ast_expression *out;
1007 ++opts_optimizationcount[OPTIM_VECTOR_COMPONENTS];
1008 out = (ast_expression*)ast_member_new(fold_ctx(fold), (ast_expression*)sel, set[0]-'x', NULL);
1009 out->node.keep = false;
1010 ((ast_member*)out)->rvalue = true;
1012 return (ast_expression*)ast_binary_new(fold_ctx(fold), INSTR_MUL_F, fold_constgen_float(fold, x, false), out);
1018 static GMQCC_INLINE ast_expression *fold_op_neg(fold_t *fold, ast_value *a) {
1021 return fold_constgen_float(fold, -fold_immvalue_float(a), false);
1022 } else if (isvector(a)) {
1024 return fold_constgen_vector(fold, vec3_neg(fold_immvalue_vector(a)));
1029 static GMQCC_INLINE ast_expression *fold_op_not(fold_t *fold, ast_value *a) {
1032 return fold_constgen_float(fold, !fold_immvalue_float(a), false);
1033 } else if (isvector(a)) {
1035 return fold_constgen_float(fold, vec3_notf(fold_immvalue_vector(a)), false);
1036 } else if (isstring(a)) {
1037 if (fold_can_1(a)) {
1038 if (OPTS_FLAG(TRUE_EMPTY_STRINGS))
1039 return fold_constgen_float(fold, !fold_immvalue_string(a), false);
1041 return fold_constgen_float(fold, !fold_immvalue_string(a) || !*fold_immvalue_string(a), false);
1047 static bool fold_check_except_float(sfloat_t (*callback)(sfloat_state_t *, sfloat_t, sfloat_t),
1056 if (!OPTS_FLAG(ARITHMETIC_EXCEPTIONS) && !OPTS_WARN(WARN_INEXACT_COMPARES))
1060 ca.f = fold_immvalue_float(a);
1061 cb.f = fold_immvalue_float(b);
1063 callback(&s, ca.s, cb.s);
1064 if (s.exceptionflags == 0)
1067 if (!OPTS_FLAG(ARITHMETIC_EXCEPTIONS))
1068 goto inexact_possible;
1070 sfloat_check(fold_ctx(fold), &s, NULL);
1073 return s.exceptionflags & SFLOAT_INEXACT;
1076 static bool fold_check_inexact_float(fold_t *fold, ast_value *a, ast_value *b) {
1077 lex_ctx_t ctx = fold_ctx(fold);
1078 if (!OPTS_WARN(WARN_INEXACT_COMPARES))
1080 if (!a->inexact && !b->inexact)
1082 return compile_warning(ctx, WARN_INEXACT_COMPARES, "inexact value in comparison");
1085 static GMQCC_INLINE ast_expression *fold_op_add(fold_t *fold, ast_value *a, ast_value *b) {
1087 if (fold_can_2(a, b)) {
1088 bool inexact = fold_check_except_float(&sfloat_add, fold, a, b);
1089 return fold_constgen_float(fold, fold_immvalue_float(a) + fold_immvalue_float(b), inexact);
1091 } else if (isvector(a)) {
1092 if (fold_can_2(a, b))
1093 return fold_constgen_vector(fold, vec3_add(fold_ctx(fold),
1094 fold_immvalue_vector(a),
1095 fold_immvalue_vector(b)));
1100 static GMQCC_INLINE ast_expression *fold_op_sub(fold_t *fold, ast_value *a, ast_value *b) {
1102 if (fold_can_2(a, b)) {
1103 bool inexact = fold_check_except_float(&sfloat_sub, fold, a, b);
1104 return fold_constgen_float(fold, fold_immvalue_float(a) - fold_immvalue_float(b), inexact);
1106 } else if (isvector(a)) {
1107 if (fold_can_2(a, b))
1108 return fold_constgen_vector(fold, vec3_sub(fold_ctx(fold),
1109 fold_immvalue_vector(a),
1110 fold_immvalue_vector(b)));
1115 static GMQCC_INLINE ast_expression *fold_op_mul(fold_t *fold, ast_value *a, ast_value *b) {
1118 if (fold_can_2(a, b))
1119 return fold_constgen_vector(fold, vec3_mulvf(fold_ctx(fold), fold_immvalue_vector(b), fold_immvalue_float(a)));
1121 if (fold_can_2(a, b)) {
1122 bool inexact = fold_check_except_float(&sfloat_mul, fold, a, b);
1123 return fold_constgen_float(fold, fold_immvalue_float(a) * fold_immvalue_float(b), inexact);
1126 } else if (isvector(a)) {
1128 if (fold_can_2(a, b))
1129 return fold_constgen_vector(fold, vec3_mulvf(fold_ctx(fold), fold_immvalue_vector(a), fold_immvalue_float(b)));
1131 if (fold_can_2(a, b)) {
1132 return fold_constgen_float(fold, vec3_mulvv(fold_ctx(fold), fold_immvalue_vector(a), fold_immvalue_vector(b)), false);
1133 } else if (OPTS_OPTIMIZATION(OPTIM_VECTOR_COMPONENTS) && fold_can_1(a)) {
1134 ast_expression *out;
1135 if ((out = fold_op_mul_vec(fold, fold_immvalue_vector(a), b, "xyz"))) return out;
1136 if ((out = fold_op_mul_vec(fold, fold_immvalue_vector(a), b, "yxz"))) return out;
1137 if ((out = fold_op_mul_vec(fold, fold_immvalue_vector(a), b, "zxy"))) return out;
1138 } else if (OPTS_OPTIMIZATION(OPTIM_VECTOR_COMPONENTS) && fold_can_1(b)) {
1139 ast_expression *out;
1140 if ((out = fold_op_mul_vec(fold, fold_immvalue_vector(b), a, "xyz"))) return out;
1141 if ((out = fold_op_mul_vec(fold, fold_immvalue_vector(b), a, "yxz"))) return out;
1142 if ((out = fold_op_mul_vec(fold, fold_immvalue_vector(b), a, "zxy"))) return out;
1149 static GMQCC_INLINE ast_expression *fold_op_div(fold_t *fold, ast_value *a, ast_value *b) {
1151 if (fold_can_2(a, b)) {
1152 bool inexact = fold_check_except_float(&sfloat_div, fold, a, b);
1153 return fold_constgen_float(fold, fold_immvalue_float(a) / fold_immvalue_float(b), inexact);
1154 } else if (fold_can_1(b)) {
1155 return (ast_expression*)ast_binary_new(
1159 fold_constgen_float(fold, 1.0f / fold_immvalue_float(b), false)
1162 } else if (isvector(a)) {
1163 if (fold_can_2(a, b)) {
1164 return fold_constgen_vector(fold, vec3_mulvf(fold_ctx(fold), fold_immvalue_vector(a), 1.0f / fold_immvalue_float(b)));
1166 return (ast_expression*)ast_binary_new(
1171 ? (ast_expression*)fold_constgen_float(fold, 1.0f / fold_immvalue_float(b), false)
1172 : (ast_expression*)ast_binary_new(
1175 (ast_expression*)fold->imm_float[1],
1184 static GMQCC_INLINE ast_expression *fold_op_mod(fold_t *fold, ast_value *a, ast_value *b) {
1185 return (fold_can_2(a, b))
1186 ? fold_constgen_float(fold, fmod(fold_immvalue_float(a), fold_immvalue_float(b)), false)
1190 static GMQCC_INLINE ast_expression *fold_op_bor(fold_t *fold, ast_value *a, ast_value *b) {
1192 if (fold_can_2(a, b))
1193 return fold_constgen_float(fold, (qcfloat_t)(((qcint_t)fold_immvalue_float(a)) | ((qcint_t)fold_immvalue_float(b))), false);
1196 if (fold_can_2(a, b))
1197 return fold_constgen_vector(fold, vec3_or(fold_immvalue_vector(a), fold_immvalue_vector(b)));
1199 if (fold_can_2(a, b))
1200 return fold_constgen_vector(fold, vec3_orvf(fold_immvalue_vector(a), fold_immvalue_float(b)));
1206 static GMQCC_INLINE ast_expression *fold_op_band(fold_t *fold, ast_value *a, ast_value *b) {
1208 if (fold_can_2(a, b))
1209 return fold_constgen_float(fold, (qcfloat_t)(((qcint_t)fold_immvalue_float(a)) & ((qcint_t)fold_immvalue_float(b))), false);
1212 if (fold_can_2(a, b))
1213 return fold_constgen_vector(fold, vec3_and(fold_immvalue_vector(a), fold_immvalue_vector(b)));
1215 if (fold_can_2(a, b))
1216 return fold_constgen_vector(fold, vec3_andvf(fold_immvalue_vector(a), fold_immvalue_float(b)));
1222 static GMQCC_INLINE ast_expression *fold_op_xor(fold_t *fold, ast_value *a, ast_value *b) {
1224 if (fold_can_2(a, b))
1225 return fold_constgen_float(fold, (qcfloat_t)(((qcint_t)fold_immvalue_float(a)) ^ ((qcint_t)fold_immvalue_float(b))), false);
1227 if (fold_can_2(a, b)) {
1229 return fold_constgen_vector(fold, vec3_xor(fold_immvalue_vector(a), fold_immvalue_vector(b)));
1231 return fold_constgen_vector(fold, vec3_xorvf(fold_immvalue_vector(a), fold_immvalue_float(b)));
1237 static GMQCC_INLINE ast_expression *fold_op_lshift(fold_t *fold, ast_value *a, ast_value *b) {
1238 if (fold_can_2(a, b) && isfloats(a, b))
1239 return fold_constgen_float(fold, (qcfloat_t)floorf(fold_immvalue_float(a) * powf(2.0f, fold_immvalue_float(b))), false);
1243 static GMQCC_INLINE ast_expression *fold_op_rshift(fold_t *fold, ast_value *a, ast_value *b) {
1244 if (fold_can_2(a, b) && isfloats(a, b))
1245 return fold_constgen_float(fold, (qcfloat_t)floorf(fold_immvalue_float(a) / powf(2.0f, fold_immvalue_float(b))), false);
1249 static GMQCC_INLINE ast_expression *fold_op_andor(fold_t *fold, ast_value *a, ast_value *b, float expr) {
1250 if (fold_can_2(a, b)) {
1251 if (OPTS_FLAG(PERL_LOGIC)) {
1253 return (fold_immediate_true(fold, a)) ? (ast_expression*)a : (ast_expression*)b;
1255 return (fold_immediate_true(fold, a)) ? (ast_expression*)b : (ast_expression*)a;
1257 return fold_constgen_float (
1259 ((expr) ? (fold_immediate_true(fold, a) || fold_immediate_true(fold, b))
1260 : (fold_immediate_true(fold, a) && fold_immediate_true(fold, b)))
1270 static GMQCC_INLINE ast_expression *fold_op_tern(fold_t *fold, ast_value *a, ast_value *b, ast_value *c) {
1271 if (fold_can_1(a)) {
1272 return fold_immediate_true(fold, a)
1273 ? (ast_expression*)b
1274 : (ast_expression*)c;
1279 static GMQCC_INLINE ast_expression *fold_op_exp(fold_t *fold, ast_value *a, ast_value *b) {
1280 if (fold_can_2(a, b))
1281 return fold_constgen_float(fold, (qcfloat_t)powf(fold_immvalue_float(a), fold_immvalue_float(b)), false);
1285 static GMQCC_INLINE ast_expression *fold_op_lteqgt(fold_t *fold, ast_value *a, ast_value *b) {
1286 if (fold_can_2(a,b)) {
1287 fold_check_inexact_float(fold, a, b);
1288 if (fold_immvalue_float(a) < fold_immvalue_float(b)) return (ast_expression*)fold->imm_float[2];
1289 if (fold_immvalue_float(a) == fold_immvalue_float(b)) return (ast_expression*)fold->imm_float[0];
1290 if (fold_immvalue_float(a) > fold_immvalue_float(b)) return (ast_expression*)fold->imm_float[1];
1295 static GMQCC_INLINE ast_expression *fold_op_ltgt(fold_t *fold, ast_value *a, ast_value *b, bool lt) {
1296 if (fold_can_2(a, b)) {
1297 fold_check_inexact_float(fold, a, b);
1298 return (lt) ? (ast_expression*)fold->imm_float[!!(fold_immvalue_float(a) < fold_immvalue_float(b))]
1299 : (ast_expression*)fold->imm_float[!!(fold_immvalue_float(a) > fold_immvalue_float(b))];
1304 static GMQCC_INLINE ast_expression *fold_op_cmp(fold_t *fold, ast_value *a, ast_value *b, bool ne) {
1305 if (fold_can_2(a, b)) {
1306 if (isfloat(a) && isfloat(b)) {
1307 float la = fold_immvalue_float(a);
1308 float lb = fold_immvalue_float(b);
1309 fold_check_inexact_float(fold, a, b);
1310 return (ast_expression*)fold->imm_float[!(ne ? la == lb : la != lb)];
1311 } if (isvector(a) && isvector(b)) {
1312 vec3_t la = fold_immvalue_vector(a);
1313 vec3_t lb = fold_immvalue_vector(b);
1314 return (ast_expression*)fold->imm_float[!(ne ? vec3_cmp(la, lb) : !vec3_cmp(la, lb))];
1320 static GMQCC_INLINE ast_expression *fold_op_bnot(fold_t *fold, ast_value *a) {
1323 return fold_constgen_float(fold, -1-fold_immvalue_float(a), false);
1327 return fold_constgen_vector(fold, vec3_not(fold_immvalue_vector(a)));
1333 static GMQCC_INLINE ast_expression *fold_op_cross(fold_t *fold, ast_value *a, ast_value *b) {
1334 if (fold_can_2(a, b))
1335 return fold_constgen_vector(fold, vec3_cross(fold_ctx(fold),
1336 fold_immvalue_vector(a),
1337 fold_immvalue_vector(b)));
1341 static GMQCC_INLINE ast_expression *fold_op_length(fold_t *fold, ast_value *a) {
1342 if (fold_can_1(a) && isstring(a))
1343 return fold_constgen_float(fold, strlen(fold_immvalue_string(a)), false);
1345 return fold_constgen_float(fold, vec_size(a->initlist), false);
1349 ast_expression *fold_op(fold_t *fold, const oper_info *info, ast_expression **opexprs) {
1350 ast_value *a = (ast_value*)opexprs[0];
1351 ast_value *b = (ast_value*)opexprs[1];
1352 ast_value *c = (ast_value*)opexprs[2];
1353 ast_expression *e = NULL;
1355 /* can a fold operation be applied to this operator usage? */
1359 switch(info->operands) {
1360 case 3: if(!c) return NULL;
1361 case 2: if(!b) return NULL;
1364 compile_error(fold_ctx(fold), "internal error: fold_op no operands to fold\n");
1370 * we could use a boolean and default case but ironically gcc produces
1371 * invalid broken assembly from that operation. clang/tcc get it right,
1372 * but interestingly ignore compiling this to a jump-table when I do that,
1373 * this happens to be the most efficent method, since you have per-level
1374 * granularity on the pointer check happening only for the case you check
1375 * it in. Opposed to the default method which would involve a boolean and
1376 * pointer check after wards.
1378 #define fold_op_case(ARGS, ARGS_OPID, OP, ARGS_FOLD) \
1379 case opid##ARGS ARGS_OPID: \
1380 if ((e = fold_op_##OP ARGS_FOLD)) { \
1381 ++opts_optimizationcount[OPTIM_CONST_FOLD]; \
1386 fold_op_case(2, ('-', 'P'), neg, (fold, a));
1387 fold_op_case(2, ('!', 'P'), not, (fold, a));
1388 fold_op_case(1, ('+'), add, (fold, a, b));
1389 fold_op_case(1, ('-'), sub, (fold, a, b));
1390 fold_op_case(1, ('*'), mul, (fold, a, b));
1391 fold_op_case(1, ('/'), div, (fold, a, b));
1392 fold_op_case(1, ('%'), mod, (fold, a, b));
1393 fold_op_case(1, ('|'), bor, (fold, a, b));
1394 fold_op_case(1, ('&'), band, (fold, a, b));
1395 fold_op_case(1, ('^'), xor, (fold, a, b));
1396 fold_op_case(1, ('<'), ltgt, (fold, a, b, true));
1397 fold_op_case(1, ('>'), ltgt, (fold, a, b, false));
1398 fold_op_case(2, ('<', '<'), lshift, (fold, a, b));
1399 fold_op_case(2, ('>', '>'), rshift, (fold, a, b));
1400 fold_op_case(2, ('|', '|'), andor, (fold, a, b, true));
1401 fold_op_case(2, ('&', '&'), andor, (fold, a, b, false));
1402 fold_op_case(2, ('?', ':'), tern, (fold, a, b, c));
1403 fold_op_case(2, ('*', '*'), exp, (fold, a, b));
1404 fold_op_case(3, ('<','=','>'), lteqgt, (fold, a, b));
1405 fold_op_case(2, ('!', '='), cmp, (fold, a, b, true));
1406 fold_op_case(2, ('=', '='), cmp, (fold, a, b, false));
1407 fold_op_case(2, ('~', 'P'), bnot, (fold, a));
1408 fold_op_case(2, ('>', '<'), cross, (fold, a, b));
1409 fold_op_case(3, ('l', 'e', 'n'), length, (fold, a));
1412 compile_error(fold_ctx(fold), "internal error: attempted to constant-fold for unsupported operator");
1417 * Constant folding for compiler intrinsics, simaler approach to operator
1418 * folding, primarly: individual functions for each intrinsics to fold,
1419 * and a generic selection function.
1421 static GMQCC_INLINE ast_expression *fold_intrin_isfinite(fold_t *fold, ast_value *a) {
1422 return fold_constgen_float(fold, isfinite(fold_immvalue_float(a)), false);
1424 static GMQCC_INLINE ast_expression *fold_intrin_isinf(fold_t *fold, ast_value *a) {
1425 return fold_constgen_float(fold, isinf(fold_immvalue_float(a)), false);
1427 static GMQCC_INLINE ast_expression *fold_intrin_isnan(fold_t *fold, ast_value *a) {
1428 return fold_constgen_float(fold, isnan(fold_immvalue_float(a)), false);
1430 static GMQCC_INLINE ast_expression *fold_intrin_isnormal(fold_t *fold, ast_value *a) {
1431 return fold_constgen_float(fold, isnormal(fold_immvalue_float(a)), false);
1433 static GMQCC_INLINE ast_expression *fold_intrin_signbit(fold_t *fold, ast_value *a) {
1434 return fold_constgen_float(fold, signbit(fold_immvalue_float(a)), false);
1436 static GMQCC_INLINE ast_expression *fold_intirn_acosh(fold_t *fold, ast_value *a) {
1437 return fold_constgen_float(fold, acoshf(fold_immvalue_float(a)), false);
1439 static GMQCC_INLINE ast_expression *fold_intrin_asinh(fold_t *fold, ast_value *a) {
1440 return fold_constgen_float(fold, asinhf(fold_immvalue_float(a)), false);
1442 static GMQCC_INLINE ast_expression *fold_intrin_atanh(fold_t *fold, ast_value *a) {
1443 return fold_constgen_float(fold, (float)atanh(fold_immvalue_float(a)), false);
1445 static GMQCC_INLINE ast_expression *fold_intrin_exp(fold_t *fold, ast_value *a) {
1446 return fold_constgen_float(fold, expf(fold_immvalue_float(a)), false);
1448 static GMQCC_INLINE ast_expression *fold_intrin_exp2(fold_t *fold, ast_value *a) {
1449 return fold_constgen_float(fold, exp2f(fold_immvalue_float(a)), false);
1451 static GMQCC_INLINE ast_expression *fold_intrin_expm1(fold_t *fold, ast_value *a) {
1452 return fold_constgen_float(fold, expm1f(fold_immvalue_float(a)), false);
1454 static GMQCC_INLINE ast_expression *fold_intrin_mod(fold_t *fold, ast_value *lhs, ast_value *rhs) {
1455 return fold_constgen_float(fold, fmodf(fold_immvalue_float(lhs), fold_immvalue_float(rhs)), false);
1457 static GMQCC_INLINE ast_expression *fold_intrin_pow(fold_t *fold, ast_value *lhs, ast_value *rhs) {
1458 return fold_constgen_float(fold, powf(fold_immvalue_float(lhs), fold_immvalue_float(rhs)), false);
1460 static GMQCC_INLINE ast_expression *fold_intrin_fabs(fold_t *fold, ast_value *a) {
1461 return fold_constgen_float(fold, fabsf(fold_immvalue_float(a)), false);
1465 ast_expression *fold_intrin(fold_t *fold, const char *intrin, ast_expression **arg) {
1466 ast_expression *ret = NULL;
1467 ast_value *a = (ast_value*)arg[0];
1468 ast_value *b = (ast_value*)arg[1];
1470 if (!strcmp(intrin, "isfinite")) ret = fold_intrin_isfinite(fold, a);
1471 if (!strcmp(intrin, "isinf")) ret = fold_intrin_isinf(fold, a);
1472 if (!strcmp(intrin, "isnan")) ret = fold_intrin_isnan(fold, a);
1473 if (!strcmp(intrin, "isnormal")) ret = fold_intrin_isnormal(fold, a);
1474 if (!strcmp(intrin, "signbit")) ret = fold_intrin_signbit(fold, a);
1475 if (!strcmp(intrin, "acosh")) ret = fold_intirn_acosh(fold, a);
1476 if (!strcmp(intrin, "asinh")) ret = fold_intrin_asinh(fold, a);
1477 if (!strcmp(intrin, "atanh")) ret = fold_intrin_atanh(fold, a);
1478 if (!strcmp(intrin, "exp")) ret = fold_intrin_exp(fold, a);
1479 if (!strcmp(intrin, "exp2")) ret = fold_intrin_exp2(fold, a);
1480 if (!strcmp(intrin, "expm1")) ret = fold_intrin_expm1(fold, a);
1481 if (!strcmp(intrin, "mod")) ret = fold_intrin_mod(fold, a, b);
1482 if (!strcmp(intrin, "pow")) ret = fold_intrin_pow(fold, a, b);
1483 if (!strcmp(intrin, "fabs")) ret = fold_intrin_fabs(fold, a);
1486 ++opts_optimizationcount[OPTIM_CONST_FOLD];
1492 * These are all the actual constant folding methods that happen in between
1493 * the AST/IR stage of the compiler , i.e eliminating branches for const
1494 * expressions, which is the only supported thing so far. We undefine the
1495 * testing macros here because an ir_value is differant than an ast_value.
1501 #undef fold_immvalue_float
1502 #undef fold_immvalue_string
1503 #undef fold_immvalue_vector
1507 #define isfloat(X) ((X)->vtype == TYPE_FLOAT)
1508 /*#define isstring(X) ((X)->vtype == TYPE_STRING)*/
1509 /*#define isvector(X) ((X)->vtype == TYPE_VECTOR)*/
1510 #define fold_immvalue_float(X) ((X)->constval.vfloat)
1511 #define fold_immvalue_vector(X) ((X)->constval.vvec)
1512 /*#define fold_immvalue_string(X) ((X)->constval.vstring)*/
1513 #define fold_can_1(X) ((X)->hasvalue && (X)->cvq == CV_CONST)
1514 /*#define fold_can_2(X,Y) (fold_can_1(X) && fold_can_1(Y))*/
1516 static ast_expression *fold_superfluous(ast_expression *left, ast_expression *right, int op) {
1517 ast_expression *swapped = NULL; /* using this as bool */
1520 if (!ast_istype(right, ast_value) || !fold_can_1((load = (ast_value*)right))) {
1526 if (!ast_istype(right, ast_value) || !fold_can_1((load = (ast_value*)right)))
1534 if (fold_immvalue_float(load) == 1.0f) {
1535 ++opts_optimizationcount[OPTIM_PEEPHOLE];
1546 if (fold_immvalue_float(load) == 0.0f) {
1547 ++opts_optimizationcount[OPTIM_PEEPHOLE];
1554 if (vec3_cmp(fold_immvalue_vector(load), vec3_create(1, 1, 1))) {
1555 ++opts_optimizationcount[OPTIM_PEEPHOLE];
1565 if (vec3_cmp(fold_immvalue_vector(load), vec3_create(0, 0, 0))) {
1566 ++opts_optimizationcount[OPTIM_PEEPHOLE];
1576 ast_expression *fold_binary(lex_ctx_t ctx, int op, ast_expression *left, ast_expression *right) {
1577 ast_expression *ret = fold_superfluous(left, right, op);
1580 return (ast_expression*)ast_binary_new(ctx, op, left, right);
1583 static GMQCC_INLINE int fold_cond(ir_value *condval, ast_function *func, ast_ifthen *branch) {
1584 if (isfloat(condval) && fold_can_1(condval) && OPTS_OPTIMIZATION(OPTIM_CONST_FOLD_DCE)) {
1585 ast_expression_codegen *cgen;
1588 bool istrue = (fold_immvalue_float(condval) != 0.0f && branch->on_true);
1589 bool isfalse = (fold_immvalue_float(condval) == 0.0f && branch->on_false);
1590 ast_expression *path = (istrue) ? branch->on_true :
1591 (isfalse) ? branch->on_false : NULL;
1594 * no path to take implies that the evaluation is if(0) and there
1595 * is no else block. so eliminate all the code.
1597 ++opts_optimizationcount[OPTIM_CONST_FOLD_DCE];
1601 if (!(elide = ir_function_create_block(ast_ctx(branch), func->ir_func, ast_function_label(func, ((istrue) ? "ontrue" : "onfalse")))))
1603 if (!(*(cgen = path->codegen))((ast_expression*)path, func, false, &dummy))
1605 if (!ir_block_create_jump(func->curblock, ast_ctx(branch), elide))
1608 * now the branch has been eliminated and the correct block for the constant evaluation
1609 * is expanded into the current block for the function.
1611 func->curblock = elide;
1612 ++opts_optimizationcount[OPTIM_CONST_FOLD_DCE];
1615 return -1; /* nothing done */
1618 int fold_cond_ternary(ir_value *condval, ast_function *func, ast_ternary *branch) {
1619 return fold_cond(condval, func, (ast_ifthen*)branch);
1622 int fold_cond_ifthen(ir_value *condval, ast_function *func, ast_ifthen *branch) {
1623 return fold_cond(condval, func, branch);