MEMMOVE_OR_MEMCPY: unroll word-copying loops in the hope of a small speedup.
[valgrind.git] / cachegrind / cg_branchpred.c
blob24c8fb2d4df64eeb6bb3d1aedb0f443733aaf291
2 /*--------------------------------------------------------------------*/
3 /*--- Branch predictor simulation cg_branchpred.c ---*/
4 /*--------------------------------------------------------------------*/
6 /*
7 This file is part of Cachegrind, a Valgrind tool for cache
8 profiling programs.
10 Copyright (C) 2002-2017 Nicholas Nethercote
11 njn@valgrind.org
13 This program is free software; you can redistribute it and/or
14 modify it under the terms of the GNU General Public License as
15 published by the Free Software Foundation; either version 2 of the
16 License, or (at your option) any later version.
18 This program is distributed in the hope that it will be useful, but
19 WITHOUT ANY WARRANTY; without even the implied warranty of
20 MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
21 General Public License for more details.
23 You should have received a copy of the GNU General Public License
24 along with this program; if not, write to the Free Software
25 Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA
26 02111-1307, USA.
28 The GNU General Public License is contained in the file COPYING.
32 /* This file contains the actual branch predictor simulator and its
33 associated state. As with cg_sim.c it is #included directly into
34 cg_main.c. It provides:
36 - a taken/not-taken predictor for conditional branches
37 - a branch target address predictor for indirect branches
39 Function return-address prediction is not modelled, on the basis
40 that return stack predictors almost always predict correctly, and
41 also that it is difficult for Valgrind to robustly identify
42 function calls and returns.
45 /* How many bits at the bottom of an instruction address are
46 guaranteed to be zero? */
47 #if defined(VGA_ppc32) || defined(VGA_ppc64be) || defined(VGA_ppc64le) \
48 || defined(VGA_mips32) || defined(VGA_mips64) || defined(VGA_arm64)
49 # define N_IADDR_LO_ZERO_BITS 2
50 #elif defined(VGA_x86) || defined(VGA_amd64)
51 # define N_IADDR_LO_ZERO_BITS 0
52 #elif defined(VGA_s390x) || defined(VGA_arm)
53 # define N_IADDR_LO_ZERO_BITS 1
54 #else
55 # error "Unsupported architecture"
56 #endif
59 /* Get a taken/not-taken prediction for the instruction (presumably a
60 conditional branch) at instr_addr. Once that's done, update the
61 predictor state based on whether or not it was actually taken, as
62 indicated by 'taken'. Finally, return 1 for a mispredict and 0 for
63 a successful predict.
65 The predictor is an array of 16k (== 2^14) 2-bit saturating
66 counters. Given the address of the branch instruction, the array
67 index to use is computed both from the low order bits of the branch
68 instruction's address, and the global history - that is, from the
69 taken/not-taken behaviour of the most recent few branches. This
70 makes the predictor able to correlate this branch's behaviour with
71 that of other branches.
73 TODO: use predictor written by someone who understands this stuff.
74 Perhaps it would be better to move to a standard GShare predictor
75 and/or tournament predictor.
77 /* The index is composed of N_HIST bits at the top and N_IADD bits at
78 the bottom. These numbers chosen somewhat arbitrarily, but note
79 that making N_IADD_BITS too small (eg 4) can cause large amounts of
80 aliasing, and hence misprediction, particularly if the history bits
81 are mostly unchanging. */
82 #define N_HIST_BITS 7
83 #define N_IADD_BITS 7
85 #define N_BITS (N_HIST_BITS + N_IADD_BITS)
86 #define N_COUNTERS (1 << N_BITS)
88 static UWord shift_register = 0; /* Contains global history */
89 static UChar counters[N_COUNTERS]; /* Counter array; presumably auto-zeroed */
92 static ULong do_cond_branch_predict ( Addr instr_addr, Word takenW )
94 UWord indx;
95 Bool predicted_taken, actually_taken, mispredict;
97 const UWord hist_mask = (1 << N_HIST_BITS) - 1;
98 const UWord iadd_mask = (1 << N_IADD_BITS) - 1;
99 UWord hist_bits = shift_register & hist_mask;
100 UWord iadd_bits = (instr_addr >> N_IADDR_LO_ZERO_BITS)
101 & iadd_mask;
103 tl_assert(hist_bits <= hist_mask);
104 tl_assert(iadd_bits <= iadd_mask);
105 indx = (hist_bits << N_IADD_BITS) | iadd_bits;
106 tl_assert(indx < N_COUNTERS);
107 if (0) VG_(printf)("index = %d\n", (Int)indx);
109 tl_assert(takenW <= 1);
110 predicted_taken = counters[ indx ] >= 2;
111 actually_taken = takenW > 0;
113 mispredict = (actually_taken && (!predicted_taken))
114 || ((!actually_taken) && predicted_taken);
116 shift_register <<= 1;
117 shift_register |= (actually_taken ? 1 : 0);
119 if (actually_taken) {
120 if (counters[indx] < 3)
121 counters[indx]++;
122 } else {
123 if (counters[indx] > 0)
124 counters[indx]--;
127 tl_assert(counters[indx] <= 3);
129 return mispredict ? 1 : 0;
133 /* A very simple indirect branch predictor. Use the branch's address
134 to index a table which records the previous target address for this
135 branch (or whatever aliased with it) and use that as the
136 prediction. */
137 #define N_BTAC_BITS 9
138 #define N_BTAC (1 << N_BTAC_BITS)
139 static Addr btac[N_BTAC]; /* BTAC; presumably auto-zeroed */
141 static ULong do_ind_branch_predict ( Addr instr_addr, Addr actual )
143 Bool mispredict;
144 const UWord mask = (1 << N_BTAC_BITS) - 1;
145 UWord indx = (instr_addr >> N_IADDR_LO_ZERO_BITS)
146 & mask;
147 tl_assert(indx < N_BTAC);
148 mispredict = btac[indx] != actual;
149 btac[indx] = actual;
150 return mispredict ? 1 : 0;
154 /*--------------------------------------------------------------------*/
155 /*--- end cg_branchpred.c ---*/
156 /*--------------------------------------------------------------------*/