mirror_zfs/module/zfs/vdev_raidz_math.c
Brian Behlendorf 07ca433973 Reduce fletcher4 and raidz benchmark times
During module load time all of the available fetcher4 and raidz
implementations are benchmarked for a fixed amount of time to
determine the fastest available.  Manual testing has shown that this
time can be significantly reduced with negligible effect on the final
results.

This commit changes the benchmark time to 1ms which can reduce the
module load time by over a second on x86_64.  On an x86_64 system
with sse3, ssse3, and avx2 instructions the benchmark times are:

    Fletcher4    603ms   -> 15ms
    RAIDZ        1,322ms -> 64ms

Reviewed-by: Matthew Macy <mmacy@freebsd.org>
Reviewed-by: George Melikov <mail@gmelikov.ru>
Reviewed-by: Ryan Moeller <ryan@iXsystems.com>
Signed-off-by: Brian Behlendorf <behlendorf1@llnl.gov>
Closes #11282
2020-12-23 14:34:59 -08:00

667 lines
17 KiB
C

/*
* CDDL HEADER START
*
* The contents of this file are subject to the terms of the
* Common Development and Distribution License (the "License").
* You may not use this file except in compliance with the License.
*
* You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE
* or http://www.opensolaris.org/os/licensing.
* See the License for the specific language governing permissions
* and limitations under the License.
*
* When distributing Covered Code, include this CDDL HEADER in each
* file and include the License file at usr/src/OPENSOLARIS.LICENSE.
* If applicable, add the following below this CDDL HEADER, with the
* fields enclosed by brackets "[]" replaced with your own identifying
* information: Portions Copyright [yyyy] [name of copyright owner]
*
* CDDL HEADER END
*/
/*
* Copyright (C) 2016 Gvozden Nešković. All rights reserved.
*/
#include <sys/zfs_context.h>
#include <sys/types.h>
#include <sys/zio.h>
#include <sys/debug.h>
#include <sys/zfs_debug.h>
#include <sys/vdev_raidz.h>
#include <sys/vdev_raidz_impl.h>
#include <sys/simd.h>
/* Opaque implementation with NULL methods to represent original methods */
static const raidz_impl_ops_t vdev_raidz_original_impl = {
.name = "original",
.is_supported = raidz_will_scalar_work,
};
/* RAIDZ parity op that contain the fastest methods */
static raidz_impl_ops_t vdev_raidz_fastest_impl = {
.name = "fastest"
};
/* All compiled in implementations */
const raidz_impl_ops_t *raidz_all_maths[] = {
&vdev_raidz_original_impl,
&vdev_raidz_scalar_impl,
#if defined(__x86_64) && defined(HAVE_SSE2) /* only x86_64 for now */
&vdev_raidz_sse2_impl,
#endif
#if defined(__x86_64) && defined(HAVE_SSSE3) /* only x86_64 for now */
&vdev_raidz_ssse3_impl,
#endif
#if defined(__x86_64) && defined(HAVE_AVX2) /* only x86_64 for now */
&vdev_raidz_avx2_impl,
#endif
#if defined(__x86_64) && defined(HAVE_AVX512F) /* only x86_64 for now */
&vdev_raidz_avx512f_impl,
#endif
#if defined(__x86_64) && defined(HAVE_AVX512BW) /* only x86_64 for now */
&vdev_raidz_avx512bw_impl,
#endif
#if defined(__aarch64__) && !defined(__FreeBSD__)
&vdev_raidz_aarch64_neon_impl,
&vdev_raidz_aarch64_neonx2_impl,
#endif
#if defined(__powerpc__) && defined(__altivec__)
&vdev_raidz_powerpc_altivec_impl,
#endif
};
/* Indicate that benchmark has been completed */
static boolean_t raidz_math_initialized = B_FALSE;
/* Select raidz implementation */
#define IMPL_FASTEST (UINT32_MAX)
#define IMPL_CYCLE (UINT32_MAX - 1)
#define IMPL_ORIGINAL (0)
#define IMPL_SCALAR (1)
#define RAIDZ_IMPL_READ(i) (*(volatile uint32_t *) &(i))
static uint32_t zfs_vdev_raidz_impl = IMPL_SCALAR;
static uint32_t user_sel_impl = IMPL_FASTEST;
/* Hold all supported implementations */
static size_t raidz_supp_impl_cnt = 0;
static raidz_impl_ops_t *raidz_supp_impl[ARRAY_SIZE(raidz_all_maths)];
#if defined(_KERNEL)
/*
* kstats values for supported implementations
* Values represent per disk throughput of 8 disk+parity raidz vdev [B/s]
*/
static raidz_impl_kstat_t raidz_impl_kstats[ARRAY_SIZE(raidz_all_maths) + 1];
/* kstat for benchmarked implementations */
static kstat_t *raidz_math_kstat = NULL;
#endif
/*
* Returns the RAIDZ operations for raidz_map() parity calculations. When
* a SIMD implementation is not allowed in the current context, then fallback
* to the fastest generic implementation.
*/
const raidz_impl_ops_t *
vdev_raidz_math_get_ops(void)
{
if (!kfpu_allowed())
return (&vdev_raidz_scalar_impl);
raidz_impl_ops_t *ops = NULL;
const uint32_t impl = RAIDZ_IMPL_READ(zfs_vdev_raidz_impl);
switch (impl) {
case IMPL_FASTEST:
ASSERT(raidz_math_initialized);
ops = &vdev_raidz_fastest_impl;
break;
case IMPL_CYCLE:
/* Cycle through all supported implementations */
ASSERT(raidz_math_initialized);
ASSERT3U(raidz_supp_impl_cnt, >, 0);
static size_t cycle_impl_idx = 0;
size_t idx = (++cycle_impl_idx) % raidz_supp_impl_cnt;
ops = raidz_supp_impl[idx];
break;
case IMPL_ORIGINAL:
ops = (raidz_impl_ops_t *)&vdev_raidz_original_impl;
break;
case IMPL_SCALAR:
ops = (raidz_impl_ops_t *)&vdev_raidz_scalar_impl;
break;
default:
ASSERT3U(impl, <, raidz_supp_impl_cnt);
ASSERT3U(raidz_supp_impl_cnt, >, 0);
if (impl < ARRAY_SIZE(raidz_all_maths))
ops = raidz_supp_impl[impl];
break;
}
ASSERT3P(ops, !=, NULL);
return (ops);
}
/*
* Select parity generation method for raidz_map
*/
int
vdev_raidz_math_generate(raidz_map_t *rm)
{
raidz_gen_f gen_parity = NULL;
switch (raidz_parity(rm)) {
case 1:
gen_parity = rm->rm_ops->gen[RAIDZ_GEN_P];
break;
case 2:
gen_parity = rm->rm_ops->gen[RAIDZ_GEN_PQ];
break;
case 3:
gen_parity = rm->rm_ops->gen[RAIDZ_GEN_PQR];
break;
default:
gen_parity = NULL;
cmn_err(CE_PANIC, "invalid RAID-Z configuration %d",
raidz_parity(rm));
break;
}
/* if method is NULL execute the original implementation */
if (gen_parity == NULL)
return (RAIDZ_ORIGINAL_IMPL);
gen_parity(rm);
return (0);
}
static raidz_rec_f
reconstruct_fun_p_sel(raidz_map_t *rm, const int *parity_valid,
const int nbaddata)
{
if (nbaddata == 1 && parity_valid[CODE_P]) {
return (rm->rm_ops->rec[RAIDZ_REC_P]);
}
return ((raidz_rec_f) NULL);
}
static raidz_rec_f
reconstruct_fun_pq_sel(raidz_map_t *rm, const int *parity_valid,
const int nbaddata)
{
if (nbaddata == 1) {
if (parity_valid[CODE_P]) {
return (rm->rm_ops->rec[RAIDZ_REC_P]);
} else if (parity_valid[CODE_Q]) {
return (rm->rm_ops->rec[RAIDZ_REC_Q]);
}
} else if (nbaddata == 2 &&
parity_valid[CODE_P] && parity_valid[CODE_Q]) {
return (rm->rm_ops->rec[RAIDZ_REC_PQ]);
}
return ((raidz_rec_f) NULL);
}
static raidz_rec_f
reconstruct_fun_pqr_sel(raidz_map_t *rm, const int *parity_valid,
const int nbaddata)
{
if (nbaddata == 1) {
if (parity_valid[CODE_P]) {
return (rm->rm_ops->rec[RAIDZ_REC_P]);
} else if (parity_valid[CODE_Q]) {
return (rm->rm_ops->rec[RAIDZ_REC_Q]);
} else if (parity_valid[CODE_R]) {
return (rm->rm_ops->rec[RAIDZ_REC_R]);
}
} else if (nbaddata == 2) {
if (parity_valid[CODE_P] && parity_valid[CODE_Q]) {
return (rm->rm_ops->rec[RAIDZ_REC_PQ]);
} else if (parity_valid[CODE_P] && parity_valid[CODE_R]) {
return (rm->rm_ops->rec[RAIDZ_REC_PR]);
} else if (parity_valid[CODE_Q] && parity_valid[CODE_R]) {
return (rm->rm_ops->rec[RAIDZ_REC_QR]);
}
} else if (nbaddata == 3 &&
parity_valid[CODE_P] && parity_valid[CODE_Q] &&
parity_valid[CODE_R]) {
return (rm->rm_ops->rec[RAIDZ_REC_PQR]);
}
return ((raidz_rec_f) NULL);
}
/*
* Select data reconstruction method for raidz_map
* @parity_valid - Parity validity flag
* @dt - Failed data index array
* @nbaddata - Number of failed data columns
*/
int
vdev_raidz_math_reconstruct(raidz_map_t *rm, const int *parity_valid,
const int *dt, const int nbaddata)
{
raidz_rec_f rec_fn = NULL;
switch (raidz_parity(rm)) {
case PARITY_P:
rec_fn = reconstruct_fun_p_sel(rm, parity_valid, nbaddata);
break;
case PARITY_PQ:
rec_fn = reconstruct_fun_pq_sel(rm, parity_valid, nbaddata);
break;
case PARITY_PQR:
rec_fn = reconstruct_fun_pqr_sel(rm, parity_valid, nbaddata);
break;
default:
cmn_err(CE_PANIC, "invalid RAID-Z configuration %d",
raidz_parity(rm));
break;
}
if (rec_fn == NULL)
return (RAIDZ_ORIGINAL_IMPL);
else
return (rec_fn(rm, dt));
}
const char *raidz_gen_name[] = {
"gen_p", "gen_pq", "gen_pqr"
};
const char *raidz_rec_name[] = {
"rec_p", "rec_q", "rec_r",
"rec_pq", "rec_pr", "rec_qr", "rec_pqr"
};
#if defined(_KERNEL)
#define RAIDZ_KSTAT_LINE_LEN (17 + 10*12 + 1)
static int
raidz_math_kstat_headers(char *buf, size_t size)
{
int i;
ssize_t off;
ASSERT3U(size, >=, RAIDZ_KSTAT_LINE_LEN);
off = snprintf(buf, size, "%-17s", "implementation");
for (i = 0; i < ARRAY_SIZE(raidz_gen_name); i++)
off += snprintf(buf + off, size - off, "%-16s",
raidz_gen_name[i]);
for (i = 0; i < ARRAY_SIZE(raidz_rec_name); i++)
off += snprintf(buf + off, size - off, "%-16s",
raidz_rec_name[i]);
(void) snprintf(buf + off, size - off, "\n");
return (0);
}
static int
raidz_math_kstat_data(char *buf, size_t size, void *data)
{
raidz_impl_kstat_t *fstat = &raidz_impl_kstats[raidz_supp_impl_cnt];
raidz_impl_kstat_t *cstat = (raidz_impl_kstat_t *)data;
ssize_t off = 0;
int i;
ASSERT3U(size, >=, RAIDZ_KSTAT_LINE_LEN);
if (cstat == fstat) {
off += snprintf(buf + off, size - off, "%-17s", "fastest");
for (i = 0; i < ARRAY_SIZE(raidz_gen_name); i++) {
int id = fstat->gen[i];
off += snprintf(buf + off, size - off, "%-16s",
raidz_supp_impl[id]->name);
}
for (i = 0; i < ARRAY_SIZE(raidz_rec_name); i++) {
int id = fstat->rec[i];
off += snprintf(buf + off, size - off, "%-16s",
raidz_supp_impl[id]->name);
}
} else {
ptrdiff_t id = cstat - raidz_impl_kstats;
off += snprintf(buf + off, size - off, "%-17s",
raidz_supp_impl[id]->name);
for (i = 0; i < ARRAY_SIZE(raidz_gen_name); i++)
off += snprintf(buf + off, size - off, "%-16llu",
(u_longlong_t)cstat->gen[i]);
for (i = 0; i < ARRAY_SIZE(raidz_rec_name); i++)
off += snprintf(buf + off, size - off, "%-16llu",
(u_longlong_t)cstat->rec[i]);
}
(void) snprintf(buf + off, size - off, "\n");
return (0);
}
static void *
raidz_math_kstat_addr(kstat_t *ksp, loff_t n)
{
if (n <= raidz_supp_impl_cnt)
ksp->ks_private = (void *) (raidz_impl_kstats + n);
else
ksp->ks_private = NULL;
return (ksp->ks_private);
}
#define BENCH_D_COLS (8ULL)
#define BENCH_COLS (BENCH_D_COLS + PARITY_PQR)
#define BENCH_ZIO_SIZE (1ULL << SPA_OLD_MAXBLOCKSHIFT) /* 128 kiB */
#define BENCH_NS MSEC2NSEC(1) /* 1ms */
typedef void (*benchmark_fn)(raidz_map_t *rm, const int fn);
static void
benchmark_gen_impl(raidz_map_t *rm, const int fn)
{
(void) fn;
vdev_raidz_generate_parity(rm);
}
static void
benchmark_rec_impl(raidz_map_t *rm, const int fn)
{
static const int rec_tgt[7][3] = {
{1, 2, 3}, /* rec_p: bad QR & D[0] */
{0, 2, 3}, /* rec_q: bad PR & D[0] */
{0, 1, 3}, /* rec_r: bad PQ & D[0] */
{2, 3, 4}, /* rec_pq: bad R & D[0][1] */
{1, 3, 4}, /* rec_pr: bad Q & D[0][1] */
{0, 3, 4}, /* rec_qr: bad P & D[0][1] */
{3, 4, 5} /* rec_pqr: bad & D[0][1][2] */
};
vdev_raidz_reconstruct(rm, rec_tgt[fn], 3);
}
/*
* Benchmarking of all supported implementations (raidz_supp_impl_cnt)
* is performed by setting the rm_ops pointer and calling the top level
* generate/reconstruct methods of bench_rm.
*/
static void
benchmark_raidz_impl(raidz_map_t *bench_rm, const int fn, benchmark_fn bench_fn)
{
uint64_t run_cnt, speed, best_speed = 0;
hrtime_t t_start, t_diff;
raidz_impl_ops_t *curr_impl;
raidz_impl_kstat_t *fstat = &raidz_impl_kstats[raidz_supp_impl_cnt];
int impl, i;
for (impl = 0; impl < raidz_supp_impl_cnt; impl++) {
/* set an implementation to benchmark */
curr_impl = raidz_supp_impl[impl];
bench_rm->rm_ops = curr_impl;
run_cnt = 0;
t_start = gethrtime();
do {
for (i = 0; i < 5; i++, run_cnt++)
bench_fn(bench_rm, fn);
t_diff = gethrtime() - t_start;
} while (t_diff < BENCH_NS);
speed = run_cnt * BENCH_ZIO_SIZE * NANOSEC;
speed /= (t_diff * BENCH_COLS);
if (bench_fn == benchmark_gen_impl)
raidz_impl_kstats[impl].gen[fn] = speed;
else
raidz_impl_kstats[impl].rec[fn] = speed;
/* Update fastest implementation method */
if (speed > best_speed) {
best_speed = speed;
if (bench_fn == benchmark_gen_impl) {
fstat->gen[fn] = impl;
vdev_raidz_fastest_impl.gen[fn] =
curr_impl->gen[fn];
} else {
fstat->rec[fn] = impl;
vdev_raidz_fastest_impl.rec[fn] =
curr_impl->rec[fn];
}
}
}
}
#endif
/*
* Initialize and benchmark all supported implementations.
*/
static void
benchmark_raidz(void)
{
raidz_impl_ops_t *curr_impl;
int i, c;
/* Move supported impl into raidz_supp_impl */
for (i = 0, c = 0; i < ARRAY_SIZE(raidz_all_maths); i++) {
curr_impl = (raidz_impl_ops_t *)raidz_all_maths[i];
if (curr_impl->init)
curr_impl->init();
if (curr_impl->is_supported())
raidz_supp_impl[c++] = (raidz_impl_ops_t *)curr_impl;
}
membar_producer(); /* complete raidz_supp_impl[] init */
raidz_supp_impl_cnt = c; /* number of supported impl */
#if defined(_KERNEL)
zio_t *bench_zio = NULL;
raidz_map_t *bench_rm = NULL;
uint64_t bench_parity;
/* Fake a zio and run the benchmark on a warmed up buffer */
bench_zio = kmem_zalloc(sizeof (zio_t), KM_SLEEP);
bench_zio->io_offset = 0;
bench_zio->io_size = BENCH_ZIO_SIZE; /* only data columns */
bench_zio->io_abd = abd_alloc_linear(BENCH_ZIO_SIZE, B_TRUE);
memset(abd_to_buf(bench_zio->io_abd), 0xAA, BENCH_ZIO_SIZE);
/* Benchmark parity generation methods */
for (int fn = 0; fn < RAIDZ_GEN_NUM; fn++) {
bench_parity = fn + 1;
/* New raidz_map is needed for each generate_p/q/r */
bench_rm = vdev_raidz_map_alloc(bench_zio, SPA_MINBLOCKSHIFT,
BENCH_D_COLS + bench_parity, bench_parity);
benchmark_raidz_impl(bench_rm, fn, benchmark_gen_impl);
vdev_raidz_map_free(bench_rm);
}
/* Benchmark data reconstruction methods */
bench_rm = vdev_raidz_map_alloc(bench_zio, SPA_MINBLOCKSHIFT,
BENCH_COLS, PARITY_PQR);
for (int fn = 0; fn < RAIDZ_REC_NUM; fn++)
benchmark_raidz_impl(bench_rm, fn, benchmark_rec_impl);
vdev_raidz_map_free(bench_rm);
/* cleanup the bench zio */
abd_free(bench_zio->io_abd);
kmem_free(bench_zio, sizeof (zio_t));
#else
/*
* Skip the benchmark in user space to avoid impacting libzpool
* consumers (zdb, zhack, zinject, ztest). The last implementation
* is assumed to be the fastest and used by default.
*/
memcpy(&vdev_raidz_fastest_impl,
raidz_supp_impl[raidz_supp_impl_cnt - 1],
sizeof (vdev_raidz_fastest_impl));
strcpy(vdev_raidz_fastest_impl.name, "fastest");
#endif /* _KERNEL */
}
void
vdev_raidz_math_init(void)
{
/* Determine the fastest available implementation. */
benchmark_raidz();
#if defined(_KERNEL)
/* Install kstats for all implementations */
raidz_math_kstat = kstat_create("zfs", 0, "vdev_raidz_bench", "misc",
KSTAT_TYPE_RAW, 0, KSTAT_FLAG_VIRTUAL);
if (raidz_math_kstat != NULL) {
raidz_math_kstat->ks_data = NULL;
raidz_math_kstat->ks_ndata = UINT32_MAX;
kstat_set_raw_ops(raidz_math_kstat,
raidz_math_kstat_headers,
raidz_math_kstat_data,
raidz_math_kstat_addr);
kstat_install(raidz_math_kstat);
}
#endif
/* Finish initialization */
atomic_swap_32(&zfs_vdev_raidz_impl, user_sel_impl);
raidz_math_initialized = B_TRUE;
}
void
vdev_raidz_math_fini(void)
{
raidz_impl_ops_t const *curr_impl;
#if defined(_KERNEL)
if (raidz_math_kstat != NULL) {
kstat_delete(raidz_math_kstat);
raidz_math_kstat = NULL;
}
#endif
for (int i = 0; i < ARRAY_SIZE(raidz_all_maths); i++) {
curr_impl = raidz_all_maths[i];
if (curr_impl->fini)
curr_impl->fini();
}
}
static const struct {
char *name;
uint32_t sel;
} math_impl_opts[] = {
{ "cycle", IMPL_CYCLE },
{ "fastest", IMPL_FASTEST },
{ "original", IMPL_ORIGINAL },
{ "scalar", IMPL_SCALAR }
};
/*
* Function sets desired raidz implementation.
*
* If we are called before init(), user preference will be saved in
* user_sel_impl, and applied in later init() call. This occurs when module
* parameter is specified on module load. Otherwise, directly update
* zfs_vdev_raidz_impl.
*
* @val Name of raidz implementation to use
* @param Unused.
*/
int
vdev_raidz_impl_set(const char *val)
{
int err = -EINVAL;
char req_name[RAIDZ_IMPL_NAME_MAX];
uint32_t impl = RAIDZ_IMPL_READ(user_sel_impl);
size_t i;
/* sanitize input */
i = strnlen(val, RAIDZ_IMPL_NAME_MAX);
if (i == 0 || i == RAIDZ_IMPL_NAME_MAX)
return (err);
strlcpy(req_name, val, RAIDZ_IMPL_NAME_MAX);
while (i > 0 && !!isspace(req_name[i-1]))
i--;
req_name[i] = '\0';
/* Check mandatory options */
for (i = 0; i < ARRAY_SIZE(math_impl_opts); i++) {
if (strcmp(req_name, math_impl_opts[i].name) == 0) {
impl = math_impl_opts[i].sel;
err = 0;
break;
}
}
/* check all supported impl if init() was already called */
if (err != 0 && raidz_math_initialized) {
/* check all supported implementations */
for (i = 0; i < raidz_supp_impl_cnt; i++) {
if (strcmp(req_name, raidz_supp_impl[i]->name) == 0) {
impl = i;
err = 0;
break;
}
}
}
if (err == 0) {
if (raidz_math_initialized)
atomic_swap_32(&zfs_vdev_raidz_impl, impl);
else
atomic_swap_32(&user_sel_impl, impl);
}
return (err);
}
#if defined(_KERNEL) && defined(__linux__)
static int
zfs_vdev_raidz_impl_set(const char *val, zfs_kernel_param_t *kp)
{
return (vdev_raidz_impl_set(val));
}
static int
zfs_vdev_raidz_impl_get(char *buffer, zfs_kernel_param_t *kp)
{
int i, cnt = 0;
char *fmt;
const uint32_t impl = RAIDZ_IMPL_READ(zfs_vdev_raidz_impl);
ASSERT(raidz_math_initialized);
/* list mandatory options */
for (i = 0; i < ARRAY_SIZE(math_impl_opts) - 2; i++) {
fmt = (impl == math_impl_opts[i].sel) ? "[%s] " : "%s ";
cnt += sprintf(buffer + cnt, fmt, math_impl_opts[i].name);
}
/* list all supported implementations */
for (i = 0; i < raidz_supp_impl_cnt; i++) {
fmt = (i == impl) ? "[%s] " : "%s ";
cnt += sprintf(buffer + cnt, fmt, raidz_supp_impl[i]->name);
}
return (cnt);
}
module_param_call(zfs_vdev_raidz_impl, zfs_vdev_raidz_impl_set,
zfs_vdev_raidz_impl_get, NULL, 0644);
MODULE_PARM_DESC(zfs_vdev_raidz_impl, "Select raidz implementation.");
#endif