| 1 | /* Macro for explicit loop unrolling. | 
|---|
| 2 | Copyright (C) 2019-2020 Free Software Foundation, Inc. | 
|---|
| 3 | This file is part of the GNU C Library. | 
|---|
| 4 |  | 
|---|
| 5 | The GNU C Library is free software; you can redistribute it and/or | 
|---|
| 6 | modify it under the terms of the GNU Lesser General Public | 
|---|
| 7 | License as published by the Free Software Foundation; either | 
|---|
| 8 | version 2.1 of the License, or (at your option) any later version. | 
|---|
| 9 |  | 
|---|
| 10 | The GNU C Library is distributed in the hope that it will be useful, | 
|---|
| 11 | but WITHOUT ANY WARRANTY; without even the implied warranty of | 
|---|
| 12 | MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU | 
|---|
| 13 | Lesser General Public License for more details. | 
|---|
| 14 |  | 
|---|
| 15 | You should have received a copy of the GNU Lesser General Public | 
|---|
| 16 | License along with the GNU C Library; if not, see | 
|---|
| 17 | <https://www.gnu.org/licenses/>.  */ | 
|---|
| 18 |  | 
|---|
| 19 | #ifndef _LOOP_UNROLL_H | 
|---|
| 20 | #define _LOOP_UNROLL_H | 
|---|
| 21 |  | 
|---|
| 22 | /* Loop unroll macro to be used for explicit force loop unrolling with a | 
|---|
| 23 | configurable number of iterations.  The idea is to make the loop unrolling | 
|---|
| 24 | independent of whether the compiler is able to unroll through specific | 
|---|
| 25 | optimizations options (-funroll-loops or -funroll-all-loops). | 
|---|
| 26 |  | 
|---|
| 27 | For instance, to implement strcpy with SRC being the source input and | 
|---|
| 28 | DEST the destination buffer, it is expected the macro to be used in this | 
|---|
| 29 | way: | 
|---|
| 30 |  | 
|---|
| 31 | #define ITERATION(index)	\ | 
|---|
| 32 | ({ char c = *str++; *dest++ = c; c != '\0' }) | 
|---|
| 33 |  | 
|---|
| 34 | while (1) | 
|---|
| 35 | UNROLL_REPEAT (4, ITERATION) | 
|---|
| 36 |  | 
|---|
| 37 | The loop will be manually unrolled 4 times.  Another option is to do | 
|---|
| 38 | the index update after the tests: | 
|---|
| 39 |  | 
|---|
| 40 | #define ITERATION(index)	\ | 
|---|
| 41 | ({ char c = *(str + index); *(dest + index) = c; c != '\0' }) | 
|---|
| 42 | #define UPDATE(n)		\ | 
|---|
| 43 | str += n; dst += n | 
|---|
| 44 |  | 
|---|
| 45 | while (1) | 
|---|
| 46 | UNROLL_REPEAT_UPDATE (4, ITERATION, UPDATE) | 
|---|
| 47 |  | 
|---|
| 48 | The loop will be manually unrolled 4 times and the SRC and DEST pointers | 
|---|
| 49 | will be updated only after the last iteration. | 
|---|
| 50 |  | 
|---|
| 51 | Currently, both macros unroll the loop 8 times at maximum.  */ | 
|---|
| 52 |  | 
|---|
| 53 | #define UNROLL_REPEAT_1(X)    if (!X(0)) break; | 
|---|
| 54 | #define UNROLL_REPEAT_2(X)    UNROLL_REPEAT_1 (X) if (!X (1)) break; | 
|---|
| 55 | #define UNROLL_REPEAT_3(X)    UNROLL_REPEAT_2 (X) if (!X (2)) break; | 
|---|
| 56 | #define UNROLL_REPEAT_4(X)    UNROLL_REPEAT_3 (X) if (!X (3)) break; | 
|---|
| 57 | #define UNROLL_REPEAT_5(X)    UNROLL_REPEAT_4 (X) if (!X (4)) break; | 
|---|
| 58 | #define UNROLL_REPEAT_6(X)    UNROLL_REPEAT_5 (X) if (!X (5)) break; | 
|---|
| 59 | #define UNROLL_REPEAT_7(X)    UNROLL_REPEAT_6 (X) if (!X (6)) break; | 
|---|
| 60 | #define UNROLL_REPEAT_8(X)    UNROLL_REPEAT_7 (X) if (!X (7)) break; | 
|---|
| 61 |  | 
|---|
| 62 | #define UNROLL_EXPAND(...)    __VA_ARGS__ | 
|---|
| 63 |  | 
|---|
| 64 | #define UNROLL_REPEAT__(N, X) UNROLL_EXPAND(UNROLL_REPEAT_ ## N) (X) | 
|---|
| 65 | #define UNROLL_REPEAT_(N, X)  UNROLL_REPEAT__ (N, X) | 
|---|
| 66 |  | 
|---|
| 67 | #define UNROLL_REPEAT(N, X)                \ | 
|---|
| 68 | (void) ({                                \ | 
|---|
| 69 | UNROLL_REPEAT_ (UNROLL_EXPAND(N), X);  \ | 
|---|
| 70 | }) | 
|---|
| 71 |  | 
|---|
| 72 | #define UNROLL_REPEAT_UPDATE(N, X, U)      \ | 
|---|
| 73 | (void) ({                                \ | 
|---|
| 74 | UNROLL_REPEAT_ (UNROLL_EXPAND(N), X);  \ | 
|---|
| 75 | UPDATE (N);                            \ | 
|---|
| 76 | }) | 
|---|
| 77 |  | 
|---|
| 78 | #endif | 
|---|
| 79 |  | 
|---|