unroll.h

 1/* SPDX-License-Identifier: GPL-2.0 */
 2
 3/*
 4 * Copyright (C) 2023 Google LLC.
 5 */
 6
 7#ifndef __UNROLL_H
 8#define __UNROLL_H
 9
10#include <linux/args.h>
11
12#ifdef CONFIG_CC_IS_CLANG
13#define __pick_unrolled(x, y)		_Pragma(#x)
14#else
15#define __pick_unrolled(x, y)		_Pragma(#y)
16#endif
17
18/**
19 * unrolled - loop attributes to ask the compiler to unroll it
20 *
21 * Usage:
22 *
23 * #define BATCH 8
24 *
25 *	unrolled_count(BATCH)
26 *	for (u32 i = 0; i < BATCH; i++)
27 *		// loop body without cross-iteration dependencies
28 *
29 * This is only a hint and the compiler is free to disable unrolling if it
30 * thinks the count is suboptimal and may hurt performance and/or hugely
31 * increase object code size.
32 * Not having any cross-iteration dependencies (i.e. when iter x + 1 depends
33 * on what iter x will do with variables) is not a strict requirement, but
34 * provides best performance and object code size.
35 * Available only on Clang and GCC 8.x onwards.
36 */
37
38/* Ask the compiler to pick an optimal unroll count, Clang only */
39#define unrolled							\
40	__pick_unrolled(clang loop unroll(enable), /* nothing */)
41
42/* Unroll each @n iterations of the loop */
43#define unrolled_count(n)						\
44	__pick_unrolled(clang loop unroll_count(n), GCC unroll n)
45
46/* Unroll the whole loop */
47#define unrolled_full							\
48	__pick_unrolled(clang loop unroll(full), GCC unroll 65534)
49
50/* Never unroll the loop */
51#define unrolled_none							\
52	__pick_unrolled(clang loop unroll(disable), GCC unroll 1)
53
54#define UNROLL(N, MACRO, args...) CONCATENATE(__UNROLL_, N)(MACRO, args)
55
56#define __UNROLL_0(MACRO, args...)
57#define __UNROLL_1(MACRO, args...)  __UNROLL_0(MACRO, args)  MACRO(0, args)
58#define __UNROLL_2(MACRO, args...)  __UNROLL_1(MACRO, args)  MACRO(1, args)
59#define __UNROLL_3(MACRO, args...)  __UNROLL_2(MACRO, args)  MACRO(2, args)
60#define __UNROLL_4(MACRO, args...)  __UNROLL_3(MACRO, args)  MACRO(3, args)
61#define __UNROLL_5(MACRO, args...)  __UNROLL_4(MACRO, args)  MACRO(4, args)
62#define __UNROLL_6(MACRO, args...)  __UNROLL_5(MACRO, args)  MACRO(5, args)
63#define __UNROLL_7(MACRO, args...)  __UNROLL_6(MACRO, args)  MACRO(6, args)
64#define __UNROLL_8(MACRO, args...)  __UNROLL_7(MACRO, args)  MACRO(7, args)
65#define __UNROLL_9(MACRO, args...)  __UNROLL_8(MACRO, args)  MACRO(8, args)
66#define __UNROLL_10(MACRO, args...) __UNROLL_9(MACRO, args)  MACRO(9, args)
67#define __UNROLL_11(MACRO, args...) __UNROLL_10(MACRO, args) MACRO(10, args)
68#define __UNROLL_12(MACRO, args...) __UNROLL_11(MACRO, args) MACRO(11, args)
69#define __UNROLL_13(MACRO, args...) __UNROLL_12(MACRO, args) MACRO(12, args)
70#define __UNROLL_14(MACRO, args...) __UNROLL_13(MACRO, args) MACRO(13, args)
71#define __UNROLL_15(MACRO, args...) __UNROLL_14(MACRO, args) MACRO(14, args)
72#define __UNROLL_16(MACRO, args...) __UNROLL_15(MACRO, args) MACRO(15, args)
73#define __UNROLL_17(MACRO, args...) __UNROLL_16(MACRO, args) MACRO(16, args)
74#define __UNROLL_18(MACRO, args...) __UNROLL_17(MACRO, args) MACRO(17, args)
75#define __UNROLL_19(MACRO, args...) __UNROLL_18(MACRO, args) MACRO(18, args)
76#define __UNROLL_20(MACRO, args...) __UNROLL_19(MACRO, args) MACRO(19, args)
77
78#endif /* __UNROLL_H */