linux-stable/lib/raid6/mmx.c

145 lines
3.9 KiB
C
Raw Normal View History

/* -*- linux-c -*- ------------------------------------------------------- *
*
* Copyright 2002 H. Peter Anvin - All Rights Reserved
*
* This program 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, Inc., 53 Temple Place Ste 330,
* Boston MA 02111-1307, USA; either version 2 of the License, or
* (at your option) any later version; incorporated herein by reference.
*
* ----------------------------------------------------------------------- */
/*
* raid6/mmx.c
*
* MMX implementation of RAID-6 syndrome functions
*/
#ifdef CONFIG_X86_32
#include <linux/raid/pq.h>
#include "x86.h"
/* Shared with raid6/sse1.c */
const struct raid6_mmx_constants {
u64 x1d;
} raid6_mmx_constants = {
0x1d1d1d1d1d1d1d1dULL,
};
static int raid6_have_mmx(void)
{
/* Not really "boot_cpu" but "all_cpus" */
return boot_cpu_has(X86_FEATURE_MMX);
}
/*
* Plain MMX implementation
*/
static void raid6_mmx1_gen_syndrome(int disks, size_t bytes, void **ptrs)
{
u8 **dptr = (u8 **)ptrs;
u8 *p, *q;
int d, z, z0;
z0 = disks - 3; /* Highest data disk */
p = dptr[z0+1]; /* XOR parity */
q = dptr[z0+2]; /* RS syndrome */
kernel_fpu_begin();
asm volatile("movq %0,%%mm0" : : "m" (raid6_mmx_constants.x1d));
asm volatile("pxor %mm5,%mm5"); /* Zero temp */
for ( d = 0 ; d < bytes ; d += 8 ) {
asm volatile("movq %0,%%mm2" : : "m" (dptr[z0][d])); /* P[0] */
asm volatile("movq %mm2,%mm4"); /* Q[0] */
for ( z = z0-1 ; z >= 0 ; z-- ) {
asm volatile("movq %0,%%mm6" : : "m" (dptr[z][d]));
asm volatile("pcmpgtb %mm4,%mm5");
asm volatile("paddb %mm4,%mm4");
asm volatile("pand %mm0,%mm5");
asm volatile("pxor %mm5,%mm4");
asm volatile("pxor %mm5,%mm5");
asm volatile("pxor %mm6,%mm2");
asm volatile("pxor %mm6,%mm4");
}
asm volatile("movq %%mm2,%0" : "=m" (p[d]));
asm volatile("pxor %mm2,%mm2");
asm volatile("movq %%mm4,%0" : "=m" (q[d]));
asm volatile("pxor %mm4,%mm4");
}
kernel_fpu_end();
}
const struct raid6_calls raid6_mmxx1 = {
raid6_mmx1_gen_syndrome,
md/raid6 algorithms: delta syndrome functions v3: s-o-b comment, explanation of performance and descision for the start/stop implementation Implementing rmw functionality for RAID6 requires optimized syndrome calculation. Up to now we can only generate a complete syndrome. The target P/Q pages are always overwritten. With this patch we provide a framework for inplace P/Q modification. In the first place simply fill those functions with NULL values. xor_syndrome() has two additional parameters: start & stop. These will indicate the first and last page that are changing during a rmw run. That makes it possible to avoid several unneccessary loops and speed up calculation. The caller needs to implement the following logic to make the functions work. 1) xor_syndrome(disks, start, stop, ...): "Remove" all data of source blocks inside P/Q between (and including) start and end. 2) modify any block with start <= block <= stop 3) xor_syndrome(disks, start, stop, ...): "Reinsert" all data of source blocks into P/Q between (and including) start and end. Pages between start and stop that won't be changed should be filled with a pointer to the kernel zero page. The reasons for not taking NULL pages are: 1) Algorithms cross the whole source data line by line. Thus avoid additional branches. 2) Having a NULL page avoids calculating the XOR P parity but still need calulation steps for the Q parity. Depending on the algorithm unrolling that might be only a difference of 2 instructions per loop. The benchmark numbers of the gen_syndrome() functions are displayed in the kernel log. Do the same for the xor_syndrome() functions. This will help to analyze performance problems and give an rough estimate how well the algorithm works. The choice of the fastest algorithm will still depend on the gen_syndrome() performance. With the start/stop page implementation the speed can vary a lot in real life. E.g. a change of page 0 & page 15 on a stripe will be harder to compute than the case where page 0 & page 1 are XOR candidates. To be not to enthusiatic about the expected speeds we will run a worse case test that simulates a change on the upper half of the stripe. So we do: 1) calculation of P/Q for the upper pages 2) continuation of Q for the lower (empty) pages Signed-off-by: Markus Stockhausen <stockhausen@collogia.de> Signed-off-by: NeilBrown <neilb@suse.de>
2014-12-15 01:57:04 +00:00
NULL, /* XOR not yet implemented */
raid6_have_mmx,
"mmxx1",
0
};
/*
* Unrolled-by-2 MMX implementation
*/
static void raid6_mmx2_gen_syndrome(int disks, size_t bytes, void **ptrs)
{
u8 **dptr = (u8 **)ptrs;
u8 *p, *q;
int d, z, z0;
z0 = disks - 3; /* Highest data disk */
p = dptr[z0+1]; /* XOR parity */
q = dptr[z0+2]; /* RS syndrome */
kernel_fpu_begin();
asm volatile("movq %0,%%mm0" : : "m" (raid6_mmx_constants.x1d));
asm volatile("pxor %mm5,%mm5"); /* Zero temp */
asm volatile("pxor %mm7,%mm7"); /* Zero temp */
for ( d = 0 ; d < bytes ; d += 16 ) {
asm volatile("movq %0,%%mm2" : : "m" (dptr[z0][d])); /* P[0] */
asm volatile("movq %0,%%mm3" : : "m" (dptr[z0][d+8]));
asm volatile("movq %mm2,%mm4"); /* Q[0] */
asm volatile("movq %mm3,%mm6"); /* Q[1] */
for ( z = z0-1 ; z >= 0 ; z-- ) {
asm volatile("pcmpgtb %mm4,%mm5");
asm volatile("pcmpgtb %mm6,%mm7");
asm volatile("paddb %mm4,%mm4");
asm volatile("paddb %mm6,%mm6");
asm volatile("pand %mm0,%mm5");
asm volatile("pand %mm0,%mm7");
asm volatile("pxor %mm5,%mm4");
asm volatile("pxor %mm7,%mm6");
asm volatile("movq %0,%%mm5" : : "m" (dptr[z][d]));
asm volatile("movq %0,%%mm7" : : "m" (dptr[z][d+8]));
asm volatile("pxor %mm5,%mm2");
asm volatile("pxor %mm7,%mm3");
asm volatile("pxor %mm5,%mm4");
asm volatile("pxor %mm7,%mm6");
asm volatile("pxor %mm5,%mm5");
asm volatile("pxor %mm7,%mm7");
}
asm volatile("movq %%mm2,%0" : "=m" (p[d]));
asm volatile("movq %%mm3,%0" : "=m" (p[d+8]));
asm volatile("movq %%mm4,%0" : "=m" (q[d]));
asm volatile("movq %%mm6,%0" : "=m" (q[d+8]));
}
kernel_fpu_end();
}
const struct raid6_calls raid6_mmxx2 = {
raid6_mmx2_gen_syndrome,
md/raid6 algorithms: delta syndrome functions v3: s-o-b comment, explanation of performance and descision for the start/stop implementation Implementing rmw functionality for RAID6 requires optimized syndrome calculation. Up to now we can only generate a complete syndrome. The target P/Q pages are always overwritten. With this patch we provide a framework for inplace P/Q modification. In the first place simply fill those functions with NULL values. xor_syndrome() has two additional parameters: start & stop. These will indicate the first and last page that are changing during a rmw run. That makes it possible to avoid several unneccessary loops and speed up calculation. The caller needs to implement the following logic to make the functions work. 1) xor_syndrome(disks, start, stop, ...): "Remove" all data of source blocks inside P/Q between (and including) start and end. 2) modify any block with start <= block <= stop 3) xor_syndrome(disks, start, stop, ...): "Reinsert" all data of source blocks into P/Q between (and including) start and end. Pages between start and stop that won't be changed should be filled with a pointer to the kernel zero page. The reasons for not taking NULL pages are: 1) Algorithms cross the whole source data line by line. Thus avoid additional branches. 2) Having a NULL page avoids calculating the XOR P parity but still need calulation steps for the Q parity. Depending on the algorithm unrolling that might be only a difference of 2 instructions per loop. The benchmark numbers of the gen_syndrome() functions are displayed in the kernel log. Do the same for the xor_syndrome() functions. This will help to analyze performance problems and give an rough estimate how well the algorithm works. The choice of the fastest algorithm will still depend on the gen_syndrome() performance. With the start/stop page implementation the speed can vary a lot in real life. E.g. a change of page 0 & page 15 on a stripe will be harder to compute than the case where page 0 & page 1 are XOR candidates. To be not to enthusiatic about the expected speeds we will run a worse case test that simulates a change on the upper half of the stripe. So we do: 1) calculation of P/Q for the upper pages 2) continuation of Q for the lower (empty) pages Signed-off-by: Markus Stockhausen <stockhausen@collogia.de> Signed-off-by: NeilBrown <neilb@suse.de>
2014-12-15 01:57:04 +00:00
NULL, /* XOR not yet implemented */
raid6_have_mmx,
"mmxx2",
0
};
#endif