1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
|
/* Description of target passes for AVR 8-bit microcontrollers.
Copyright (C) 2016-2025 Free Software Foundation, Inc. */
/* This file is part of GCC.
GCC is free software; you can redistribute it and/or modify it under
the terms of the GNU General Public License as published by the Free
Software Foundation; either version 3, or (at your option) any later
version.
GCC is distributed in the hope that it will be useful, but WITHOUT ANY
WARRANTY; without even the implied warranty of MERCHANTABILITY or
FITNESS FOR A PARTICULAR PURPOSE. See the GNU General Public License
for more details.
You should have received a copy of the GNU General Public License
along with GCC; see the file COPYING3. If not see
<http://www.gnu.org/licenses/>. */
/* A post reload optimization pass that fuses PLUS insns with CONST_INT
addend with a load or store insn to get POST_INC or PRE_DEC addressing.
It can also fuse two PLUSes to a single one, which may occur due to
splits from `avr_split_fake_addressing_move'. We do this in an own
pass because it can find more cases than peephole2, for example when
there are unrelated insns between the interesting ones. */
INSERT_PASS_BEFORE (pass_peephole2, 1, avr_pass_fuse_add);
/* There are cases where avr-fuse-add doesn't find POST_INC cases because
the RTL code at that time is too long-winded, and moves registers back and
forth (which seems to be the same reason for why pass auto_inc_dec cannot
find POST_INC, either). Some of that long-windedness is cleaned up very
late in pass cprop_hardreg, which opens up new opportunities to find post
increments. An example is the following function from AVR-LibC's qsort:
void swapfunc (char *a, char *b, int n)
{
do
{
char tmp = *a;
*a++ = *b;
*b++ = tmp;
} while (--n > 0);
}
Hence, run avr-fuse-add twice; the second time after cprop_hardreg. */
INSERT_PASS_AFTER (pass_cprop_hardreg, 1, avr_pass_fuse_add);
/* An analysis pass that runs prior to prologue / epilogue generation.
Computes cfun->machine->gasisr.maybe which is used in prologue and
epilogue generation provided -mgas-isr-prologues is on. */
INSERT_PASS_BEFORE (pass_thread_prologue_and_epilogue, 1, avr_pass_pre_proep);
/* This avr-specific pass (re)computes insn notes, in particular REG_DEAD
notes which are used by `avr.cc::reg_unused_after' and branch offset
computations. These notes must be correct, i.e. there must be no
dangling REG_DEAD notes; otherwise wrong code might result, cf. PR64331.
DF needs (correct) CFG, hence right before free_cfg is the last
opportunity to rectify notes. */
INSERT_PASS_BEFORE (pass_free_cfg, 1, avr_pass_recompute_notes);
/* casesi uses a SImode switch index which is quite costly as most code will
work on HImode or QImode. The following pass runs right after .expand and
tries to fix such situations by operating on the original mode. This
reduces code size and register pressure.
The assertion is that the code generated by casesi is unaltered and
a sign-extend or zero-extend from QImode or HImode precedes the casesi
insns without any insns in between. */
INSERT_PASS_AFTER (pass_expand, 1, avr_pass_casesi);
/* If-else decision trees generated for switch / case may produce sequences
like
SREG = compare (reg, val);
if (SREG == 0) goto label1;
SREG = compare (reg, 1 + val);
if (SREG >= 0) goto label2;
which can be optimized to
SREG = compare (reg, val);
if (SREG == 0) goto label1;
if (SREG >= 0) goto label2;
The optimal place for such a pass would be directly after expand, but
it's not possible for a jump insn to target more than one code label.
Hence, run a mini pass right before split2 which introduces REG_CC. */
INSERT_PASS_BEFORE (pass_split_after_reload, 1, avr_pass_ifelse);
/* A post reload pass that tracks known values held in registers
and performs optimizations based on that knowledge.
It also splits non-memory insns that can be represented in
terms of byte operations.
It runs between the two instances of the RTL peephole pass because
- The RTL peepholer may provide a scratch reg for *reload_in<mode>.
- The RTL peepholer may optimize insns involving lower registers. */
INSERT_PASS_AFTER (pass_peephole2, 1, avr_pass_fuse_move);
/* Run an instance of post-reload split prior to avr-fuse-move.
Purpose is to split the `3op' alternative (which allows 3 operands)
of shift insns into a 3-operand shift with a byte offset, and
a 2-operand residual shift. This additional split pass runs after
the 1st RTL peephole pass but prior to avr-fuse-move.
The respective define_split patterns have a `n_avr_fuse_add_executed'
condition (amongst others) so that split passes that run before
the 1st RTL peephole pass won't split them. Shifts with a constant
offset that is a multiple of 8 are split by avr-fuse-move. */
INSERT_PASS_AFTER (pass_peephole2, 1, avr_pass_split_after_peephole2);
|
