mirror_zfs/module/zfs/vdev_raidz_math_impl.h
Romain Dolbeau 62a65a654e Add parity generation/rebuild using 128-bits NEON for Aarch64
This re-use the framework established for SSE2, SSSE3 and
AVX2. However, GCC is using FP registers on Aarch64, so
unlike SSE/AVX2 we can't rely on the registers being left alone
between ASM statements. So instead, the NEON code uses
C variables and GCC extended ASM syntax. Note that since
the kernel explicitly disable vector registers, they
have to be locally re-enabled explicitly.

As we use the variable's number to define the symbolic
name, and GCC won't allow duplicate symbolic names,
numbers have to be unique. Even when the code is not
going to be used (e.g. the case for 4 registers when
using the macro with only 2). Only the actually used
variables should be declared, otherwise the build
will fails in debug mode.

This requires the replacement of the XOR(X,X) syntax
by a new ZERO(X) macro, which does the same thing but
without repeating the argument. And perhaps someday
there will be a machine where there is a more efficient
way to zero a register than XOR with itself. This affects
scalar, SSE2, SSSE3 and AVX2 as they need the new macro.

It's possible to write faster implementations (different
scheduling, different unrolling, interleaving NEON and
scalar, ...) for various cores, but this one has the
advantage of fitting in the current state of the code,
and thus is likely easier to review/check/merge.

The only difference between aarch64-neon and aarch64-neonx2
is that aarch64-neonx2 unroll some functions some more.

Reviewed-by: Gvozden Neskovic <neskovic@gmail.com>
Reviewed-by: Brian Behlendorf <behlendorf1@llnl.gov>
Signed-off-by: Romain Dolbeau <romain.dolbeau@atos.net>
Closes #4801
2016-10-03 09:44:00 -07:00

1200 lines
30 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.
*/
#ifndef _VDEV_RAIDZ_MATH_IMPL_H
#define _VDEV_RAIDZ_MATH_IMPL_H
#include <sys/types.h>
#define raidz_inline inline __attribute__((always_inline))
#ifndef noinline
#define noinline __attribute__((noinline))
#endif
/* Calculate data offset in raidz column, offset is in bytes */
#define COL_OFF(col, off) ((v_t *)(((char *)(col)->rc_data) + (off)))
/*
* PARITY CALCULATION
* An optimized function is called for a full length of data columns
* If RAIDZ map contains remainder columns (shorter columns) the same function
* is called for reminder of full columns.
*
* GEN_[P|PQ|PQR]_BLOCK() functions are designed to be efficiently in-lined by
* the compiler. This removes a lot of conditionals from the inside loop which
* makes the code faster, especially for vectorized code.
* They are also highly parametrized, allowing for each implementation to define
* most optimal stride, and register allocation.
*/
static raidz_inline void
GEN_P_BLOCK(raidz_map_t * const rm, const size_t off, const size_t end,
const int ncols)
{
int c;
size_t ioff;
raidz_col_t * const pcol = raidz_col_p(rm, CODE_P);
raidz_col_t *col;
GEN_P_DEFINE();
for (ioff = off; ioff < end; ioff += (GEN_P_STRIDE * sizeof (v_t))) {
LOAD(COL_OFF(&(rm->rm_col[1]), ioff), GEN_P_P);
for (c = 2; c < ncols; c++) {
col = &rm->rm_col[c];
XOR_ACC(COL_OFF(col, ioff), GEN_P_P);
}
STORE(COL_OFF(pcol, ioff), GEN_P_P);
}
}
/*
* Generate P parity (RAIDZ1)
*
* @rm RAIDZ map
*/
static raidz_inline void
raidz_generate_p_impl(raidz_map_t * const rm)
{
const int ncols = raidz_ncols(rm);
const size_t psize = raidz_big_size(rm);
const size_t short_size = raidz_short_size(rm);
raidz_math_begin();
/* short_size */
GEN_P_BLOCK(rm, 0, short_size, ncols);
/* fullcols */
GEN_P_BLOCK(rm, short_size, psize, raidz_nbigcols(rm));
raidz_math_end();
}
static raidz_inline void
GEN_PQ_BLOCK(raidz_map_t * const rm, const size_t off, const size_t end,
const int ncols, const int nbigcols)
{
int c;
size_t ioff;
raidz_col_t * const pcol = raidz_col_p(rm, CODE_P);
raidz_col_t * const qcol = raidz_col_p(rm, CODE_Q);
raidz_col_t *col;
GEN_PQ_DEFINE();
MUL2_SETUP();
for (ioff = off; ioff < end; ioff += (GEN_PQ_STRIDE * sizeof (v_t))) {
LOAD(COL_OFF(&rm->rm_col[2], ioff), GEN_PQ_P);
COPY(GEN_PQ_P, GEN_PQ_Q);
for (c = 3; c < nbigcols; c++) {
col = &rm->rm_col[c];
LOAD(COL_OFF(col, ioff), GEN_PQ_D);
MUL2(GEN_PQ_Q);
XOR(GEN_PQ_D, GEN_PQ_P);
XOR(GEN_PQ_D, GEN_PQ_Q);
}
STORE(COL_OFF(pcol, ioff), GEN_PQ_P);
for (; c < ncols; c++)
MUL2(GEN_PQ_Q);
STORE(COL_OFF(qcol, ioff), GEN_PQ_Q);
}
}
/*
* Generate PQ parity (RAIDZ2)
*
* @rm RAIDZ map
*/
static raidz_inline void
raidz_generate_pq_impl(raidz_map_t * const rm)
{
const int ncols = raidz_ncols(rm);
const size_t psize = raidz_big_size(rm);
const size_t short_size = raidz_short_size(rm);
raidz_math_begin();
/* short_size */
GEN_PQ_BLOCK(rm, 0, short_size, ncols, ncols);
/* fullcols */
GEN_PQ_BLOCK(rm, short_size, psize, ncols, raidz_nbigcols(rm));
raidz_math_end();
}
static raidz_inline void
GEN_PQR_BLOCK(raidz_map_t * const rm, const size_t off, const size_t end,
const int ncols, const int nbigcols)
{
int c;
size_t ioff;
raidz_col_t *col;
raidz_col_t * const pcol = raidz_col_p(rm, CODE_P);
raidz_col_t * const qcol = raidz_col_p(rm, CODE_Q);
raidz_col_t * const rcol = raidz_col_p(rm, CODE_R);
GEN_PQR_DEFINE();
MUL2_SETUP();
for (ioff = off; ioff < end; ioff += (GEN_PQR_STRIDE * sizeof (v_t))) {
LOAD(COL_OFF(&rm->rm_col[3], ioff), GEN_PQR_P);
COPY(GEN_PQR_P, GEN_PQR_Q);
COPY(GEN_PQR_P, GEN_PQR_R);
for (c = 4; c < nbigcols; c++) {
col = &rm->rm_col[c];
LOAD(COL_OFF(col, ioff), GEN_PQR_D);
MUL2(GEN_PQR_Q);
MUL4(GEN_PQR_R);
XOR(GEN_PQR_D, GEN_PQR_P);
XOR(GEN_PQR_D, GEN_PQR_Q);
XOR(GEN_PQR_D, GEN_PQR_R);
}
STORE(COL_OFF(pcol, ioff), GEN_PQR_P);
for (; c < ncols; c++) {
MUL2(GEN_PQR_Q);
MUL4(GEN_PQR_R);
}
STORE(COL_OFF(qcol, ioff), GEN_PQR_Q);
STORE(COL_OFF(rcol, ioff), GEN_PQR_R);
}
}
/*
* Generate PQR parity (RAIDZ3)
*
* @rm RAIDZ map
*/
static raidz_inline void
raidz_generate_pqr_impl(raidz_map_t * const rm)
{
const int ncols = raidz_ncols(rm);
const size_t psize = raidz_big_size(rm);
const size_t short_size = raidz_short_size(rm);
raidz_math_begin();
/* short_size */
GEN_PQR_BLOCK(rm, 0, short_size, ncols, ncols);
/* fullcols */
GEN_PQR_BLOCK(rm, short_size, psize, ncols, raidz_nbigcols(rm));
raidz_math_end();
}
/*
* DATA RECONSTRUCTION
*
* Data reconstruction process consists of two phases:
* - Syndrome calculation
* - Data reconstruction
*
* Syndrome is calculated by generating parity using available data columns
* and zeros in places of erasure. Existing parity is added to corresponding
* syndrome value to obtain the [P|Q|R]syn values from equation:
* P = Psyn + Dx + Dy + Dz
* Q = Qsyn + 2^x * Dx + 2^y * Dy + 2^z * Dz
* R = Rsyn + 4^x * Dx + 4^y * Dy + 4^z * Dz
*
* For data reconstruction phase, the corresponding equations are solved
* for missing data (Dx, Dy, Dz). This generally involves multiplying known
* symbols by an coefficient and adding them together. The multiplication
* constant coefficients are calculated ahead of the operation in
* raidz_rec_[q|r|pq|pq|qr|pqr]_coeff() functions.
*
* IMPLEMENTATION NOTE: RAID-Z block can have complex geometry, with "big"
* and "short" columns.
* For this reason, reconstruction is performed in minimum of
* two steps. First, from offset 0 to short_size, then from short_size to
* short_size. Calculation functions REC_[*]_BLOCK() are implemented to work
* over both ranges. The split also enables removal of conditional expressions
* from loop bodies, improving throughput of SIMD implementations.
* For the best performance, all functions marked with raidz_inline attribute
* must be inlined by compiler.
*
* parity data
* columns columns
* <----------> <------------------>
* x y <----+ missing columns (x, y)
* | |
* +---+---+---+---+-v-+---+-v-+---+ ^ 0
* | | | | | | | | | |
* | | | | | | | | | |
* | P | Q | R | D | D | D | D | D | |
* | | | | 0 | 1 | 2 | 3 | 4 | |
* | | | | | | | | | v
* | | | | | +---+---+---+ ^ short_size
* | | | | | | |
* +---+---+---+---+---+ v big_size
* <------------------> <---------->
* big columns short columns
*
*/
/*
* Functions calculate multiplication constants for data reconstruction.
* Coefficients depend on RAIDZ geometry, indexes of failed child vdevs, and
* used parity columns for reconstruction.
* @rm RAIDZ map
* @tgtidx array of missing data indexes
* @coeff output array of coefficients. Array must be user
* provided and must hold minimum MUL_CNT values
*/
static noinline void
raidz_rec_q_coeff(const raidz_map_t *rm, const int *tgtidx, unsigned *coeff)
{
const unsigned ncols = raidz_ncols(rm);
const unsigned x = tgtidx[TARGET_X];
coeff[MUL_Q_X] = gf_exp2(255 - (ncols - x - 1));
}
static noinline void
raidz_rec_r_coeff(const raidz_map_t *rm, const int *tgtidx, unsigned *coeff)
{
const unsigned ncols = raidz_ncols(rm);
const unsigned x = tgtidx[TARGET_X];
coeff[MUL_R_X] = gf_exp4(255 - (ncols - x - 1));
}
static noinline void
raidz_rec_pq_coeff(const raidz_map_t *rm, const int *tgtidx, unsigned *coeff)
{
const unsigned ncols = raidz_ncols(rm);
const unsigned x = tgtidx[TARGET_X];
const unsigned y = tgtidx[TARGET_Y];
gf_t a, b, e;
a = gf_exp2(x + 255 - y);
b = gf_exp2(255 - (ncols - x - 1));
e = a ^ 0x01;
coeff[MUL_PQ_X] = gf_div(a, e);
coeff[MUL_PQ_Y] = gf_div(b, e);
}
static noinline void
raidz_rec_pr_coeff(const raidz_map_t *rm, const int *tgtidx, unsigned *coeff)
{
const unsigned ncols = raidz_ncols(rm);
const unsigned x = tgtidx[TARGET_X];
const unsigned y = tgtidx[TARGET_Y];
gf_t a, b, e;
a = gf_exp4(x + 255 - y);
b = gf_exp4(255 - (ncols - x - 1));
e = a ^ 0x01;
coeff[MUL_PR_X] = gf_div(a, e);
coeff[MUL_PR_Y] = gf_div(b, e);
}
static noinline void
raidz_rec_qr_coeff(const raidz_map_t *rm, const int *tgtidx, unsigned *coeff)
{
const unsigned ncols = raidz_ncols(rm);
const unsigned x = tgtidx[TARGET_X];
const unsigned y = tgtidx[TARGET_Y];
gf_t nx, ny, nxxy, nxyy, d;
nx = gf_exp2(ncols - x - 1);
ny = gf_exp2(ncols - y - 1);
nxxy = gf_mul(gf_mul(nx, nx), ny);
nxyy = gf_mul(gf_mul(nx, ny), ny);
d = nxxy ^ nxyy;
coeff[MUL_QR_XQ] = ny;
coeff[MUL_QR_X] = gf_div(ny, d);
coeff[MUL_QR_YQ] = nx;
coeff[MUL_QR_Y] = gf_div(nx, d);
}
static noinline void
raidz_rec_pqr_coeff(const raidz_map_t *rm, const int *tgtidx, unsigned *coeff)
{
const unsigned ncols = raidz_ncols(rm);
const unsigned x = tgtidx[TARGET_X];
const unsigned y = tgtidx[TARGET_Y];
const unsigned z = tgtidx[TARGET_Z];
gf_t nx, ny, nz, nxx, nyy, nzz, nyyz, nyzz, xd, yd;
nx = gf_exp2(ncols - x - 1);
ny = gf_exp2(ncols - y - 1);
nz = gf_exp2(ncols - z - 1);
nxx = gf_exp4(ncols - x - 1);
nyy = gf_exp4(ncols - y - 1);
nzz = gf_exp4(ncols - z - 1);
nyyz = gf_mul(gf_mul(ny, nz), ny);
nyzz = gf_mul(nzz, ny);
xd = gf_mul(nxx, ny) ^ gf_mul(nx, nyy) ^ nyyz ^
gf_mul(nxx, nz) ^ gf_mul(nzz, nx) ^ nyzz;
yd = gf_inv(ny ^ nz);
coeff[MUL_PQR_XP] = gf_div(nyyz ^ nyzz, xd);
coeff[MUL_PQR_XQ] = gf_div(nyy ^ nzz, xd);
coeff[MUL_PQR_XR] = gf_div(ny ^ nz, xd);
coeff[MUL_PQR_YU] = nx;
coeff[MUL_PQR_YP] = gf_mul(nz, yd);
coeff[MUL_PQR_YQ] = yd;
}
/*
* Reconstruction using P parity
* @rm RAIDZ map
* @off starting offset
* @end ending offset
* @x missing data column
* @ncols number of column
*/
static raidz_inline void
REC_P_BLOCK(raidz_map_t * const rm, const size_t off, const size_t end,
const int x, const int ncols)
{
int c;
size_t ioff;
const size_t firstdc = raidz_parity(rm);
raidz_col_t * const pcol = raidz_col_p(rm, CODE_P);
raidz_col_t * const xcol = raidz_col_p(rm, x);
raidz_col_t *col;
REC_P_DEFINE();
for (ioff = off; ioff < end; ioff += (REC_P_STRIDE * sizeof (v_t))) {
LOAD(COL_OFF(pcol, ioff), REC_P_X);
for (c = firstdc; c < x; c++) {
col = &rm->rm_col[c];
XOR_ACC(COL_OFF(col, ioff), REC_P_X);
}
for (c++; c < ncols; c++) {
col = &rm->rm_col[c];
XOR_ACC(COL_OFF(col, ioff), REC_P_X);
}
STORE(COL_OFF(xcol, ioff), REC_P_X);
}
}
/*
* Reconstruct single data column using P parity
* @rec_method REC_P_BLOCK()
*
* @rm RAIDZ map
* @tgtidx array of missing data indexes
*/
static raidz_inline int
raidz_reconstruct_p_impl(raidz_map_t *rm, const int *tgtidx)
{
const int x = tgtidx[TARGET_X];
const int ncols = raidz_ncols(rm);
const int nbigcols = raidz_nbigcols(rm);
const size_t xsize = raidz_col_size(rm, x);
const size_t short_size = raidz_short_size(rm);
raidz_math_begin();
/* 0 - short_size */
REC_P_BLOCK(rm, 0, short_size, x, ncols);
/* short_size - xsize */
REC_P_BLOCK(rm, short_size, xsize, x, nbigcols);
raidz_math_end();
return (1 << CODE_P);
}
/*
* Reconstruct using Q parity
*/
#define REC_Q_SYN_UPDATE() MUL2(REC_Q_X)
#define REC_Q_INNER_LOOP(c) \
{ \
col = &rm->rm_col[c]; \
REC_Q_SYN_UPDATE(); \
XOR_ACC(COL_OFF(col, ioff), REC_Q_X); \
}
/*
* Reconstruction using Q parity
* @rm RAIDZ map
* @off starting offset
* @end ending offset
* @x missing data column
* @coeff multiplication coefficients
* @ncols number of column
* @nbigcols number of big columns
*/
static raidz_inline void
REC_Q_BLOCK(raidz_map_t * const rm, const size_t off, const size_t end,
const int x, const unsigned *coeff, const int ncols, const int nbigcols)
{
int c;
size_t ioff = 0;
const size_t firstdc = raidz_parity(rm);
raidz_col_t * const qcol = raidz_col_p(rm, CODE_Q);
raidz_col_t * const xcol = raidz_col_p(rm, x);
raidz_col_t *col;
REC_Q_DEFINE();
for (ioff = off; ioff < end; ioff += (REC_Q_STRIDE * sizeof (v_t))) {
MUL2_SETUP();
ZERO(REC_Q_X);
if (ncols == nbigcols) {
for (c = firstdc; c < x; c++)
REC_Q_INNER_LOOP(c);
REC_Q_SYN_UPDATE();
for (c++; c < nbigcols; c++)
REC_Q_INNER_LOOP(c);
} else {
for (c = firstdc; c < nbigcols; c++) {
REC_Q_SYN_UPDATE();
if (x != c) {
col = &rm->rm_col[c];
XOR_ACC(COL_OFF(col, ioff), REC_Q_X);
}
}
for (; c < ncols; c++)
REC_Q_SYN_UPDATE();
}
XOR_ACC(COL_OFF(qcol, ioff), REC_Q_X);
MUL(coeff[MUL_Q_X], REC_Q_X);
STORE(COL_OFF(xcol, ioff), REC_Q_X);
}
}
/*
* Reconstruct single data column using Q parity
* @rec_method REC_Q_BLOCK()
*
* @rm RAIDZ map
* @tgtidx array of missing data indexes
*/
static raidz_inline int
raidz_reconstruct_q_impl(raidz_map_t *rm, const int *tgtidx)
{
const int x = tgtidx[TARGET_X];
const int ncols = raidz_ncols(rm);
const int nbigcols = raidz_nbigcols(rm);
const size_t xsize = raidz_col_size(rm, x);
const size_t short_size = raidz_short_size(rm);
unsigned coeff[MUL_CNT];
raidz_rec_q_coeff(rm, tgtidx, coeff);
raidz_math_begin();
/* 0 - short_size */
REC_Q_BLOCK(rm, 0, short_size, x, coeff, ncols, ncols);
/* short_size - xsize */
REC_Q_BLOCK(rm, short_size, xsize, x, coeff, ncols, nbigcols);
raidz_math_end();
return (1 << CODE_Q);
}
/*
* Reconstruct using R parity
*/
#define REC_R_SYN_UPDATE() MUL4(REC_R_X)
#define REC_R_INNER_LOOP(c) \
{ \
col = &rm->rm_col[c]; \
REC_R_SYN_UPDATE(); \
XOR_ACC(COL_OFF(col, ioff), REC_R_X); \
}
/*
* Reconstruction using R parity
* @rm RAIDZ map
* @off starting offset
* @end ending offset
* @x missing data column
* @coeff multiplication coefficients
* @ncols number of column
* @nbigcols number of big columns
*/
static raidz_inline void
REC_R_BLOCK(raidz_map_t * const rm, const size_t off, const size_t end,
const int x, const unsigned *coeff, const int ncols, const int nbigcols)
{
int c;
size_t ioff = 0;
const size_t firstdc = raidz_parity(rm);
raidz_col_t * const rcol = raidz_col_p(rm, CODE_R);
raidz_col_t * const xcol = raidz_col_p(rm, x);
raidz_col_t *col;
REC_R_DEFINE();
for (ioff = off; ioff < end; ioff += (REC_R_STRIDE * sizeof (v_t))) {
MUL2_SETUP();
ZERO(REC_R_X);
if (ncols == nbigcols) {
for (c = firstdc; c < x; c++)
REC_R_INNER_LOOP(c);
REC_R_SYN_UPDATE();
for (c++; c < nbigcols; c++)
REC_R_INNER_LOOP(c);
} else {
for (c = firstdc; c < nbigcols; c++) {
REC_R_SYN_UPDATE();
if (c != x) {
col = &rm->rm_col[c];
XOR_ACC(COL_OFF(col, ioff), REC_R_X);
}
}
for (; c < ncols; c++)
REC_R_SYN_UPDATE();
}
XOR_ACC(COL_OFF(rcol, ioff), REC_R_X);
MUL(coeff[MUL_R_X], REC_R_X);
STORE(COL_OFF(xcol, ioff), REC_R_X);
}
}
/*
* Reconstruct single data column using R parity
* @rec_method REC_R_BLOCK()
*
* @rm RAIDZ map
* @tgtidx array of missing data indexes
*/
static raidz_inline int
raidz_reconstruct_r_impl(raidz_map_t *rm, const int *tgtidx)
{
const int x = tgtidx[TARGET_X];
const int ncols = raidz_ncols(rm);
const int nbigcols = raidz_nbigcols(rm);
const size_t xsize = raidz_col_size(rm, x);
const size_t short_size = raidz_short_size(rm);
unsigned coeff[MUL_CNT];
raidz_rec_r_coeff(rm, tgtidx, coeff);
raidz_math_begin();
/* 0 - short_size */
REC_R_BLOCK(rm, 0, short_size, x, coeff, ncols, ncols);
/* short_size - xsize */
REC_R_BLOCK(rm, short_size, xsize, x, coeff, ncols, nbigcols);
raidz_math_end();
return (1 << CODE_R);
}
/*
* Reconstruct using PQ parity
*/
#define REC_PQ_SYN_UPDATE() MUL2(REC_PQ_Y)
#define REC_PQ_INNER_LOOP(c) \
{ \
col = &rm->rm_col[c]; \
LOAD(COL_OFF(col, ioff), REC_PQ_D); \
REC_PQ_SYN_UPDATE(); \
XOR(REC_PQ_D, REC_PQ_X); \
XOR(REC_PQ_D, REC_PQ_Y); \
}
/*
* Reconstruction using PQ parity
* @rm RAIDZ map
* @off starting offset
* @end ending offset
* @x missing data column
* @y missing data column
* @coeff multiplication coefficients
* @ncols number of column
* @nbigcols number of big columns
* @calcy calculate second data column
*/
static raidz_inline void
REC_PQ_BLOCK(raidz_map_t * const rm, const size_t off, const size_t end,
const int x, const int y, const unsigned *coeff, const int ncols,
const int nbigcols, const boolean_t calcy)
{
int c;
size_t ioff = 0;
const size_t firstdc = raidz_parity(rm);
raidz_col_t * const pcol = raidz_col_p(rm, CODE_P);
raidz_col_t * const qcol = raidz_col_p(rm, CODE_Q);
raidz_col_t * const xcol = raidz_col_p(rm, x);
raidz_col_t * const ycol = raidz_col_p(rm, y);
raidz_col_t *col;
REC_PQ_DEFINE();
for (ioff = off; ioff < end; ioff += (REC_PQ_STRIDE * sizeof (v_t))) {
LOAD(COL_OFF(pcol, ioff), REC_PQ_X);
ZERO(REC_PQ_Y);
MUL2_SETUP();
if (ncols == nbigcols) {
for (c = firstdc; c < x; c++)
REC_PQ_INNER_LOOP(c);
REC_PQ_SYN_UPDATE();
for (c++; c < y; c++)
REC_PQ_INNER_LOOP(c);
REC_PQ_SYN_UPDATE();
for (c++; c < nbigcols; c++)
REC_PQ_INNER_LOOP(c);
} else {
for (c = firstdc; c < nbigcols; c++) {
REC_PQ_SYN_UPDATE();
if (c != x && c != y) {
col = &rm->rm_col[c];
LOAD(COL_OFF(col, ioff), REC_PQ_D);
XOR(REC_PQ_D, REC_PQ_X);
XOR(REC_PQ_D, REC_PQ_Y);
}
}
for (; c < ncols; c++)
REC_PQ_SYN_UPDATE();
}
XOR_ACC(COL_OFF(qcol, ioff), REC_PQ_Y);
/* Save Pxy */
COPY(REC_PQ_X, REC_PQ_D);
/* Calc X */
MUL(coeff[MUL_PQ_X], REC_PQ_X);
MUL(coeff[MUL_PQ_Y], REC_PQ_Y);
XOR(REC_PQ_Y, REC_PQ_X);
STORE(COL_OFF(xcol, ioff), REC_PQ_X);
if (calcy) {
/* Calc Y */
XOR(REC_PQ_D, REC_PQ_X);
STORE(COL_OFF(ycol, ioff), REC_PQ_X);
}
}
}
/*
* Reconstruct two data columns using PQ parity
* @rec_method REC_PQ_BLOCK()
*
* @rm RAIDZ map
* @tgtidx array of missing data indexes
*/
static raidz_inline int
raidz_reconstruct_pq_impl(raidz_map_t *rm, const int *tgtidx)
{
const int x = tgtidx[TARGET_X];
const int y = tgtidx[TARGET_Y];
const int ncols = raidz_ncols(rm);
const int nbigcols = raidz_nbigcols(rm);
const size_t xsize = raidz_col_size(rm, x);
const size_t ysize = raidz_col_size(rm, y);
const size_t short_size = raidz_short_size(rm);
unsigned coeff[MUL_CNT];
raidz_rec_pq_coeff(rm, tgtidx, coeff);
raidz_math_begin();
/* 0 - short_size */
REC_PQ_BLOCK(rm, 0, short_size, x, y, coeff, ncols, ncols, B_TRUE);
/* short_size - xsize */
REC_PQ_BLOCK(rm, short_size, xsize, x, y, coeff, ncols, nbigcols,
xsize == ysize);
raidz_math_end();
return ((1 << CODE_P) | (1 << CODE_Q));
}
/*
* Reconstruct using PR parity
*/
#define REC_PR_SYN_UPDATE() MUL4(REC_PR_Y)
#define REC_PR_INNER_LOOP(c) \
{ \
col = &rm->rm_col[c]; \
LOAD(COL_OFF(col, ioff), REC_PR_D); \
REC_PR_SYN_UPDATE(); \
XOR(REC_PR_D, REC_PR_X); \
XOR(REC_PR_D, REC_PR_Y); \
}
/*
* Reconstruction using PR parity
* @rm RAIDZ map
* @off starting offset
* @end ending offset
* @x missing data column
* @y missing data column
* @coeff multiplication coefficients
* @ncols number of column
* @nbigcols number of big columns
* @calcy calculate second data column
*/
static raidz_inline void
REC_PR_BLOCK(raidz_map_t * const rm, const size_t off, const size_t end,
const int x, const int y, const unsigned *coeff, const int ncols,
const int nbigcols, const boolean_t calcy)
{
int c;
size_t ioff;
const size_t firstdc = raidz_parity(rm);
raidz_col_t * const pcol = raidz_col_p(rm, CODE_P);
raidz_col_t * const rcol = raidz_col_p(rm, CODE_R);
raidz_col_t * const xcol = raidz_col_p(rm, x);
raidz_col_t * const ycol = raidz_col_p(rm, y);
raidz_col_t *col;
REC_PR_DEFINE();
for (ioff = off; ioff < end; ioff += (REC_PR_STRIDE * sizeof (v_t))) {
LOAD(COL_OFF(pcol, ioff), REC_PR_X);
ZERO(REC_PR_Y);
MUL2_SETUP();
if (ncols == nbigcols) {
for (c = firstdc; c < x; c++)
REC_PR_INNER_LOOP(c);
REC_PR_SYN_UPDATE();
for (c++; c < y; c++)
REC_PR_INNER_LOOP(c);
REC_PR_SYN_UPDATE();
for (c++; c < nbigcols; c++)
REC_PR_INNER_LOOP(c);
} else {
for (c = firstdc; c < nbigcols; c++) {
REC_PR_SYN_UPDATE();
if (c != x && c != y) {
col = &rm->rm_col[c];
LOAD(COL_OFF(col, ioff), REC_PR_D);
XOR(REC_PR_D, REC_PR_X);
XOR(REC_PR_D, REC_PR_Y);
}
}
for (; c < ncols; c++)
REC_PR_SYN_UPDATE();
}
XOR_ACC(COL_OFF(rcol, ioff), REC_PR_Y);
/* Save Pxy */
COPY(REC_PR_X, REC_PR_D);
/* Calc X */
MUL(coeff[MUL_PR_X], REC_PR_X);
MUL(coeff[MUL_PR_Y], REC_PR_Y);
XOR(REC_PR_Y, REC_PR_X);
STORE(COL_OFF(xcol, ioff), REC_PR_X);
if (calcy) {
/* Calc Y */
XOR(REC_PR_D, REC_PR_X);
STORE(COL_OFF(ycol, ioff), REC_PR_X);
}
}
}
/*
* Reconstruct two data columns using PR parity
* @rec_method REC_PR_BLOCK()
*
* @rm RAIDZ map
* @tgtidx array of missing data indexes
*/
static raidz_inline int
raidz_reconstruct_pr_impl(raidz_map_t *rm, const int *tgtidx)
{
const int x = tgtidx[TARGET_X];
const int y = tgtidx[TARGET_Y];
const int ncols = raidz_ncols(rm);
const int nbigcols = raidz_nbigcols(rm);
const size_t xsize = raidz_col_size(rm, x);
const size_t ysize = raidz_col_size(rm, y);
const size_t short_size = raidz_short_size(rm);
unsigned coeff[MUL_CNT];
raidz_rec_pr_coeff(rm, tgtidx, coeff);
raidz_math_begin();
/* 0 - short_size */
REC_PR_BLOCK(rm, 0, short_size, x, y, coeff, ncols, ncols, B_TRUE);
/* short_size - xsize */
REC_PR_BLOCK(rm, short_size, xsize, x, y, coeff, ncols, nbigcols,
xsize == ysize);
raidz_math_end();
return ((1 << CODE_P) | (1 << CODE_R));
}
/*
* Reconstruct using QR parity
*/
#define REC_QR_SYN_UPDATE() \
{ \
MUL2(REC_QR_X); \
MUL4(REC_QR_Y); \
}
#define REC_QR_INNER_LOOP(c) \
{ \
col = &rm->rm_col[c]; \
LOAD(COL_OFF(col, ioff), REC_QR_D); \
REC_QR_SYN_UPDATE(); \
XOR(REC_QR_D, REC_QR_X); \
XOR(REC_QR_D, REC_QR_Y); \
}
/*
* Reconstruction using QR parity
* @rm RAIDZ map
* @off starting offset
* @end ending offset
* @x missing data column
* @y missing data column
* @coeff multiplication coefficients
* @ncols number of column
* @nbigcols number of big columns
* @calcy calculate second data column
*/
static raidz_inline void
REC_QR_BLOCK(raidz_map_t * const rm, const size_t off, const size_t end,
const int x, const int y, const unsigned *coeff, const int ncols,
const int nbigcols, const boolean_t calcy)
{
int c;
size_t ioff;
const size_t firstdc = raidz_parity(rm);
raidz_col_t * const qcol = raidz_col_p(rm, CODE_Q);
raidz_col_t * const rcol = raidz_col_p(rm, CODE_R);
raidz_col_t * const xcol = raidz_col_p(rm, x);
raidz_col_t * const ycol = raidz_col_p(rm, y);
raidz_col_t *col;
REC_QR_DEFINE();
for (ioff = off; ioff < end; ioff += (REC_QR_STRIDE * sizeof (v_t))) {
MUL2_SETUP();
ZERO(REC_QR_X);
ZERO(REC_QR_Y);
if (ncols == nbigcols) {
for (c = firstdc; c < x; c++)
REC_QR_INNER_LOOP(c);
REC_QR_SYN_UPDATE();
for (c++; c < y; c++)
REC_QR_INNER_LOOP(c);
REC_QR_SYN_UPDATE();
for (c++; c < nbigcols; c++)
REC_QR_INNER_LOOP(c);
} else {
for (c = firstdc; c < nbigcols; c++) {
REC_QR_SYN_UPDATE();
if (c != x && c != y) {
col = &rm->rm_col[c];
LOAD(COL_OFF(col, ioff), REC_QR_D);
XOR(REC_QR_D, REC_QR_X);
XOR(REC_QR_D, REC_QR_Y);
}
}
for (; c < ncols; c++)
REC_QR_SYN_UPDATE();
}
XOR_ACC(COL_OFF(qcol, ioff), REC_QR_X);
XOR_ACC(COL_OFF(rcol, ioff), REC_QR_Y);
/* Save Qxy */
COPY(REC_QR_X, REC_QR_D);
/* Calc X */
MUL(coeff[MUL_QR_XQ], REC_QR_X); /* X = Q * xqm */
XOR(REC_QR_Y, REC_QR_X); /* X = R ^ X */
MUL(coeff[MUL_QR_X], REC_QR_X); /* X = X * xm */
STORE(COL_OFF(xcol, ioff), REC_QR_X);
if (calcy) {
/* Calc Y */
MUL(coeff[MUL_QR_YQ], REC_QR_D); /* X = Q * xqm */
XOR(REC_QR_Y, REC_QR_D); /* X = R ^ X */
MUL(coeff[MUL_QR_Y], REC_QR_D); /* X = X * xm */
STORE(COL_OFF(ycol, ioff), REC_QR_D);
}
}
}
/*
* Reconstruct two data columns using QR parity
* @rec_method REC_QR_BLOCK()
*
* @rm RAIDZ map
* @tgtidx array of missing data indexes
*/
static raidz_inline int
raidz_reconstruct_qr_impl(raidz_map_t *rm, const int *tgtidx)
{
const int x = tgtidx[TARGET_X];
const int y = tgtidx[TARGET_Y];
const int ncols = raidz_ncols(rm);
const int nbigcols = raidz_nbigcols(rm);
const size_t xsize = raidz_col_size(rm, x);
const size_t ysize = raidz_col_size(rm, y);
const size_t short_size = raidz_short_size(rm);
unsigned coeff[MUL_CNT];
raidz_rec_qr_coeff(rm, tgtidx, coeff);
raidz_math_begin();
/* 0 - short_size */
REC_QR_BLOCK(rm, 0, short_size, x, y, coeff, ncols, ncols, B_TRUE);
/* short_size - xsize */
REC_QR_BLOCK(rm, short_size, xsize, x, y, coeff, ncols, nbigcols,
xsize == ysize);
raidz_math_end();
return ((1 << CODE_Q) | (1 << CODE_R));
}
/*
* Reconstruct using PQR parity
*/
#define REC_PQR_SYN_UPDATE() \
{ \
MUL2(REC_PQR_Y); \
MUL4(REC_PQR_Z); \
}
#define REC_PQR_INNER_LOOP(c) \
{ \
col = &rm->rm_col[(c)]; \
LOAD(COL_OFF(col, ioff), REC_PQR_D); \
REC_PQR_SYN_UPDATE(); \
XOR(REC_PQR_D, REC_PQR_X); \
XOR(REC_PQR_D, REC_PQR_Y); \
XOR(REC_PQR_D, REC_PQR_Z); \
}
/*
* Reconstruction using PQR parity
* @rm RAIDZ map
* @off starting offset
* @end ending offset
* @x missing data column
* @y missing data column
* @z missing data column
* @coeff multiplication coefficients
* @ncols number of column
* @nbigcols number of big columns
* @calcy calculate second data column
* @calcz calculate third data column
*/
static raidz_inline void
REC_PQR_BLOCK(raidz_map_t * const rm, const size_t off, const size_t end,
const int x, const int y, const int z, const unsigned *coeff,
const int ncols, const int nbigcols, const boolean_t calcy,
const boolean_t calcz)
{
int c;
size_t ioff;
const size_t firstdc = raidz_parity(rm);
raidz_col_t * const pcol = raidz_col_p(rm, CODE_P);
raidz_col_t * const qcol = raidz_col_p(rm, CODE_Q);
raidz_col_t * const rcol = raidz_col_p(rm, CODE_R);
raidz_col_t * const xcol = raidz_col_p(rm, x);
raidz_col_t * const ycol = raidz_col_p(rm, y);
raidz_col_t * const zcol = raidz_col_p(rm, z);
raidz_col_t *col;
REC_PQR_DEFINE();
for (ioff = off; ioff < end; ioff += (REC_PQR_STRIDE * sizeof (v_t))) {
MUL2_SETUP();
LOAD(COL_OFF(pcol, ioff), REC_PQR_X);
ZERO(REC_PQR_Y);
ZERO(REC_PQR_Z);
if (ncols == nbigcols) {
for (c = firstdc; c < x; c++)
REC_PQR_INNER_LOOP(c);
REC_PQR_SYN_UPDATE();
for (c++; c < y; c++)
REC_PQR_INNER_LOOP(c);
REC_PQR_SYN_UPDATE();
for (c++; c < z; c++)
REC_PQR_INNER_LOOP(c);
REC_PQR_SYN_UPDATE();
for (c++; c < nbigcols; c++)
REC_PQR_INNER_LOOP(c);
} else {
for (c = firstdc; c < nbigcols; c++) {
REC_PQR_SYN_UPDATE();
if (c != x && c != y && c != z) {
col = &rm->rm_col[c];
LOAD(COL_OFF(col, ioff), REC_PQR_D);
XOR(REC_PQR_D, REC_PQR_X);
XOR(REC_PQR_D, REC_PQR_Y);
XOR(REC_PQR_D, REC_PQR_Z);
}
}
for (; c < ncols; c++)
REC_PQR_SYN_UPDATE();
}
XOR_ACC(COL_OFF(qcol, ioff), REC_PQR_Y);
XOR_ACC(COL_OFF(rcol, ioff), REC_PQR_Z);
/* Save Pxyz and Qxyz */
COPY(REC_PQR_X, REC_PQR_XS);
COPY(REC_PQR_Y, REC_PQR_YS);
/* Calc X */
MUL(coeff[MUL_PQR_XP], REC_PQR_X); /* Xp = Pxyz * xp */
MUL(coeff[MUL_PQR_XQ], REC_PQR_Y); /* Xq = Qxyz * xq */
XOR(REC_PQR_Y, REC_PQR_X);
MUL(coeff[MUL_PQR_XR], REC_PQR_Z); /* Xr = Rxyz * xr */
XOR(REC_PQR_Z, REC_PQR_X); /* X = Xp + Xq + Xr */
STORE(COL_OFF(xcol, ioff), REC_PQR_X);
if (calcy) {
/* Calc Y */
XOR(REC_PQR_X, REC_PQR_XS); /* Pyz = Pxyz + X */
MUL(coeff[MUL_PQR_YU], REC_PQR_X); /* Xq = X * upd_q */
XOR(REC_PQR_X, REC_PQR_YS); /* Qyz = Qxyz + Xq */
COPY(REC_PQR_XS, REC_PQR_X); /* restore Pyz */
MUL(coeff[MUL_PQR_YP], REC_PQR_X); /* Yp = Pyz * yp */
MUL(coeff[MUL_PQR_YQ], REC_PQR_YS); /* Yq = Qyz * yq */
XOR(REC_PQR_X, REC_PQR_YS); /* Y = Yp + Yq */
STORE(COL_OFF(ycol, ioff), REC_PQR_YS);
}
if (calcz) {
/* Calc Z */
XOR(REC_PQR_XS, REC_PQR_YS); /* Z = Pz = Pyz + Y */
STORE(COL_OFF(zcol, ioff), REC_PQR_YS);
}
}
}
/*
* Reconstruct three data columns using PQR parity
* @rec_method REC_PQR_BLOCK()
*
* @rm RAIDZ map
* @tgtidx array of missing data indexes
*/
static raidz_inline int
raidz_reconstruct_pqr_impl(raidz_map_t *rm, const int *tgtidx)
{
const int x = tgtidx[TARGET_X];
const int y = tgtidx[TARGET_Y];
const int z = tgtidx[TARGET_Z];
const int ncols = raidz_ncols(rm);
const int nbigcols = raidz_nbigcols(rm);
const size_t xsize = raidz_col_size(rm, x);
const size_t ysize = raidz_col_size(rm, y);
const size_t zsize = raidz_col_size(rm, z);
const size_t short_size = raidz_short_size(rm);
unsigned coeff[MUL_CNT];
raidz_rec_pqr_coeff(rm, tgtidx, coeff);
raidz_math_begin();
/* 0 - short_size */
REC_PQR_BLOCK(rm, 0, short_size, x, y, z, coeff, ncols, ncols,
B_TRUE, B_TRUE);
/* short_size - xsize */
REC_PQR_BLOCK(rm, short_size, xsize, x, y, z, coeff, ncols, nbigcols,
xsize == ysize, xsize == zsize);
raidz_math_end();
return ((1 << CODE_P) | (1 << CODE_Q) | (1 << CODE_R));
}
#endif /* _VDEV_RAIDZ_MATH_IMPL_H */