Files
littlefs/runners/bench_runner.h
T
Christopher Haster c94b5f4767 Redesigned the inlined topology of files, now using geoxylic btrees
As a part of the general redesign of files, all files, not just small
files, can inline some data directly in the metadata log. Originally,
this was a single piece of inlined data or an inlined tree (shrub) that
effectively acted as an overlay over the block/btree data.

This is now changed so that when we have a block/btree, the root of the
btree is inlined. In effect making a full btree a sort of extended
shrub.

I'm currently calling this a "geoxylic btree", since that seems to be a
somewhat related botanical term. Geoxylic btrees have, at least on
paper, a number of benefits:

- There is a single lookup path instead of two, this simplifies code a
  bit and decreases lookup costs.

- One data structure instead of two also means lfsr_file_t requires
  less RAM, since all of the on-disk variants can go into one big union.
  Though I'm not sure this is very significant vs stack/buffer costs.

- The write path is much simpler and has less duplication (it was
  difficult to deduplicate the shrub/btree code because of how the
  shrub goes through the mdir).

  In this redesign, lfsr_btree_commit_ leaves root attrs uncommitted,
  allowing lfsr_bshrub_commit to finish the job via lfsr_mdir_commit.

- We don't need to maintain a shrub estimate, we just lazily evict trees
  during mdir compaction. This has a side-effect of allowing shrubs to
  temporarily grow larger than shrub_size before eviction.

  NOTE THIS (fundamentally?) DOESN'T WORK

- There is no awkwardly high overhead for small btrees. The btree root
  for two-block files should be able to comfortably fit in the shrub
  portion of the btree, for example.

- It may be possible to also make the mtree geoxylic, which should
  reduce storage overhead of small mtrees and make better use of the
  mroot.

All of this being said, things aren't working yet. Shrub eviction during
compaction runs into a problem with a single pcache -- how do we write
the new btrees without dropping the compaction pcache? We can't evict
btrees in a separate pass becauce their number is unbounded...
2023-11-20 23:23:58 -06:00

199 lines
6.7 KiB
C

/*
* Runner for littlefs benchmarks
*
* Copyright (c) 2022, The littlefs authors.
* SPDX-License-Identifier: BSD-3-Clause
*/
#ifndef BENCH_RUNNER_H
#define BENCH_RUNNER_H
// override LFS_TRACE
void bench_trace(const char *fmt, ...);
#define LFS_TRACE_(fmt, ...) \
bench_trace("%s:%d:trace: " fmt "%s\n", \
__FILE__, \
__LINE__, \
__VA_ARGS__)
#define LFS_TRACE(...) LFS_TRACE_(__VA_ARGS__, "")
#define LFS_EMUBD_TRACE(...) LFS_TRACE_(__VA_ARGS__, "")
// BENCH_START/BENCH_STOP macros measure readed/proged/erased bytes
// through emubd
void bench_start(const char *meas, uintmax_t iter, uintmax_t size);
void bench_stop(const char *meas);
#define BENCH_START(meas, iter, size) \
bench_start(meas, iter, size)
#define BENCH_STOP(meas) \
bench_stop(meas)
// BENCH_RESULT/BENCH_FRESULT allow for explicit non-io measurements
void bench_result(const char *meas, uintmax_t iter, uintmax_t size,
uintmax_t result);
void bench_fresult(const char *meas, uintmax_t iter, uintmax_t size,
double result);
#define BENCH_RESULT(meas, iter, size, result) \
bench_result(meas, iter, size, result)
#define BENCH_FRESULT(meas, iter, size, result) \
bench_fresult(meas, iter, size, result)
// note these are indirectly included in any generated files
#include "bd/lfs_emubd.h"
#include <stdio.h>
// give source a chance to define feature macros
#undef _FEATURES_H
#undef _STDIO_H
// generated bench configurations
struct lfs_config;
enum bench_flags {
BENCH_INTERNAL = 0x1,
};
typedef uint8_t bench_flags_t;
typedef struct bench_define {
intmax_t (*cb)(void *data, size_t i);
void *data;
size_t permutations;
} bench_define_t;
struct bench_case {
const char *name;
const char *path;
bench_flags_t flags;
const bench_define_t *defines;
size_t permutations;
bool (*if_)(void);
void (*run)(struct lfs_config *cfg);
};
struct bench_suite {
const char *name;
const char *path;
bench_flags_t flags;
const char *const *define_names;
size_t define_count;
const struct bench_case *cases;
size_t case_count;
};
extern const struct bench_suite *const bench_suites[];
extern const size_t bench_suite_count;
// deterministic prng for pseudo-randomness in benches
uint32_t bench_prng(uint32_t *state);
#define BENCH_PRNG(state) bench_prng(state)
// generation of specific permutations of an array for exhaustive benching
size_t bench_factorial(size_t x);
void bench_permutation(size_t i, uint32_t *buffer, size_t size);
#define BENCH_FACTORIAL(x) bench_factorial(x)
#define BENCH_PERMUTATION(i, buffer, size) bench_permutation(i, buffer, size)
// access generated bench defines
intmax_t bench_define(size_t define);
#define BENCH_DEFINE(i) bench_define(i)
// a few preconfigured defines that control how benches run
#define BENCH_IMPLICIT_DEFINE_COUNT 16
#define BENCH_GEOMETRY_DEFINE_COUNT 3
#define READ_SIZE_i 0
#define PROG_SIZE_i 1
#define BLOCK_SIZE_i 2
#define BLOCK_COUNT_i 3
#define DISK_SIZE_i 4
#define CACHE_SIZE_i 5
#define INLINE_SIZE_i 6
#define SHRUB_SIZE_i 7
#define FRAGMENT_SIZE_i 8
#define CRYSTAL_SIZE_i 9
#define LOOKAHEAD_SIZE_i 10
#define BLOCK_CYCLES_i 11
#define ERASE_VALUE_i 12
#define ERASE_CYCLES_i 13
#define BADBLOCK_BEHAVIOR_i 14
#define POWERLOSS_BEHAVIOR_i 15
#define READ_SIZE bench_define(READ_SIZE_i)
#define PROG_SIZE bench_define(PROG_SIZE_i)
#define BLOCK_SIZE bench_define(BLOCK_SIZE_i)
#define BLOCK_COUNT bench_define(BLOCK_COUNT_i)
#define DISK_SIZE bench_define(DISK_SIZE_i)
#define CACHE_SIZE bench_define(CACHE_SIZE_i)
#define INLINE_SIZE bench_define(INLINE_SIZE_i)
#define SHRUB_SIZE bench_define(SHRUB_SIZE_i)
#define FRAGMENT_SIZE bench_define(FRAGMENT_SIZE_i)
#define CRYSTAL_SIZE bench_define(CRYSTAL_SIZE_i)
#define LOOKAHEAD_SIZE bench_define(LOOKAHEAD_SIZE_i)
#define BLOCK_CYCLES bench_define(BLOCK_CYCLES_i)
#define ERASE_VALUE bench_define(ERASE_VALUE_i)
#define ERASE_CYCLES bench_define(ERASE_CYCLES_i)
#define BADBLOCK_BEHAVIOR bench_define(BADBLOCK_BEHAVIOR_i)
#define POWERLOSS_BEHAVIOR bench_define(POWERLOSS_BEHAVIOR_i)
#define BENCH_IMPLICIT_DEFINES \
/* name value (overridable) */ \
BENCH_DEF(READ_SIZE, PROG_SIZE ) \
BENCH_DEF(PROG_SIZE, BLOCK_SIZE ) \
BENCH_DEF(BLOCK_SIZE, 0 ) \
BENCH_DEF(BLOCK_COUNT, DISK_SIZE/BLOCK_SIZE ) \
BENCH_DEF(DISK_SIZE, 1024*1024 ) \
BENCH_DEF(CACHE_SIZE, lfs_max(16, lfs_max(READ_SIZE, PROG_SIZE))) \
BENCH_DEF(INLINE_SIZE, BLOCK_SIZE/8 ) \
BENCH_DEF(SHRUB_SIZE, INLINE_SIZE ) \
BENCH_DEF(FRAGMENT_SIZE, CACHE_SIZE ) \
BENCH_DEF(CRYSTAL_SIZE, BLOCK_SIZE/8 ) \
BENCH_DEF(LOOKAHEAD_SIZE, 16 ) \
BENCH_DEF(BLOCK_CYCLES, -1 ) \
BENCH_DEF(ERASE_VALUE, 0xff ) \
BENCH_DEF(ERASE_CYCLES, 0 ) \
BENCH_DEF(BADBLOCK_BEHAVIOR, LFS_EMUBD_BADBLOCK_PROGERROR ) \
BENCH_DEF(POWERLOSS_BEHAVIOR, LFS_EMUBD_POWERLOSS_NOOP )
#define BENCH_GEOMETRIES \
/* name read_size prog_size block_size */ \
BENCH_GEO("default", 16, 16, 512 ) \
BENCH_GEO("eeprom", 1, 1, 512 ) \
BENCH_GEO("emmc", 512, 512, 512 ) \
BENCH_GEO("nor", 1, 1, 4096 ) \
BENCH_GEO("nand", 4096, 4096, 32768 )
#define BENCH_CFG \
.read_size = READ_SIZE, \
.prog_size = PROG_SIZE, \
.block_size = BLOCK_SIZE, \
.block_count = BLOCK_COUNT, \
.block_cycles = BLOCK_CYCLES, \
.cache_size = CACHE_SIZE, \
.inline_size = INLINE_SIZE, \
.shrub_size = SHRUB_SIZE, \
.fragment_size = FRAGMENT_SIZE, \
.crystal_size = CRYSTAL_SIZE, \
.lookahead_size = LOOKAHEAD_SIZE,
#define BENCH_BDCFG \
.erase_value = ERASE_VALUE, \
.erase_cycles = ERASE_CYCLES, \
.badblock_behavior = BADBLOCK_BEHAVIOR,
#endif