diff options
author | Ken Steele <ken@tilera.com> | 2013-08-07 12:39:56 -0400 |
---|---|---|
committer | NeilBrown <neilb@suse.de> | 2013-08-27 02:05:50 -0400 |
commit | ae77cbc1e7b90473a2b0963bce0e1eb163873214 (patch) | |
tree | 0d412c9c4932c59c8b495aa6214f98404849af60 /lib/raid6/tilegx.uc | |
parent | 275c51c4e34ed776d40a99dd97c1deee50303b07 (diff) |
RAID: add tilegx SIMD implementation of raid6
This change adds TILE-Gx SIMD instructions to the software raid
(md), modeling the Altivec implementation. This is only for Syndrome
generation; there is more that could be done to improve recovery,
as in the recent Intel SSE3 recovery implementation.
The code unrolls 8 times; this turns out to be the best on tilegx
hardware among the set 1, 2, 4, 8 or 16. The code reads one
cache-line of data from each disk, stores P and Q then goes to the
next cache-line.
The test code in sys/linux/lib/raid6/test reports 2008 MB/s data
read rate for syndrome generation using 18 disks (16 data and 2
parity). It was 1512 MB/s before this SIMD optimizations. This is
running on 1 core with all the data in cache.
This is based on the paper The Mathematics of RAID-6.
(http://kernel.org/pub/linux/kernel/people/hpa/raid6.pdf).
Signed-off-by: Ken Steele <ken@tilera.com>
Signed-off-by: Chris Metcalf <cmetcalf@tilera.com>
Signed-off-by: NeilBrown <neilb@suse.de>
Diffstat (limited to 'lib/raid6/tilegx.uc')
-rw-r--r-- | lib/raid6/tilegx.uc | 86 |
1 files changed, 86 insertions, 0 deletions
diff --git a/lib/raid6/tilegx.uc b/lib/raid6/tilegx.uc new file mode 100644 index 000000000000..e7c29459cbcd --- /dev/null +++ b/lib/raid6/tilegx.uc | |||
@@ -0,0 +1,86 @@ | |||
1 | /* -*- linux-c -*- ------------------------------------------------------- * | ||
2 | * | ||
3 | * Copyright 2002 H. Peter Anvin - All Rights Reserved | ||
4 | * Copyright 2012 Tilera Corporation - All Rights Reserved | ||
5 | * | ||
6 | * This program is free software; you can redistribute it and/or modify | ||
7 | * it under the terms of the GNU General Public License as published by | ||
8 | * the Free Software Foundation, Inc., 53 Temple Place Ste 330, | ||
9 | * Boston MA 02111-1307, USA; either version 2 of the License, or | ||
10 | * (at your option) any later version; incorporated herein by reference. | ||
11 | * | ||
12 | * ----------------------------------------------------------------------- */ | ||
13 | |||
14 | /* | ||
15 | * tilegx$#.c | ||
16 | * | ||
17 | * $#-way unrolled TILE-Gx SIMD for RAID-6 math. | ||
18 | * | ||
19 | * This file is postprocessed using unroll.awk. | ||
20 | * | ||
21 | */ | ||
22 | |||
23 | #include <linux/raid/pq.h> | ||
24 | |||
25 | /* Create 8 byte copies of constant byte */ | ||
26 | # define NBYTES(x) (__insn_v1addi(0, x)) | ||
27 | # define NSIZE 8 | ||
28 | |||
29 | /* | ||
30 | * The SHLBYTE() operation shifts each byte left by 1, *not* | ||
31 | * rolling over into the next byte | ||
32 | */ | ||
33 | static inline __attribute_const__ u64 SHLBYTE(u64 v) | ||
34 | { | ||
35 | /* Vector One Byte Shift Left Immediate. */ | ||
36 | return __insn_v1shli(v, 1); | ||
37 | } | ||
38 | |||
39 | /* | ||
40 | * The MASK() operation returns 0xFF in any byte for which the high | ||
41 | * bit is 1, 0x00 for any byte for which the high bit is 0. | ||
42 | */ | ||
43 | static inline __attribute_const__ u64 MASK(u64 v) | ||
44 | { | ||
45 | /* Vector One Byte Shift Right Signed Immediate. */ | ||
46 | return __insn_v1shrsi(v, 7); | ||
47 | } | ||
48 | |||
49 | |||
50 | void raid6_tilegx$#_gen_syndrome(int disks, size_t bytes, void **ptrs) | ||
51 | { | ||
52 | u8 **dptr = (u8 **)ptrs; | ||
53 | u64 *p, *q; | ||
54 | int d, z, z0; | ||
55 | |||
56 | u64 wd$$, wq$$, wp$$, w1$$, w2$$; | ||
57 | u64 x1d = NBYTES(0x1d); | ||
58 | u64 * z0ptr; | ||
59 | |||
60 | z0 = disks - 3; /* Highest data disk */ | ||
61 | p = (u64 *)dptr[z0+1]; /* XOR parity */ | ||
62 | q = (u64 *)dptr[z0+2]; /* RS syndrome */ | ||
63 | |||
64 | z0ptr = (u64 *)&dptr[z0][0]; | ||
65 | for ( d = 0 ; d < bytes ; d += NSIZE*$# ) { | ||
66 | wq$$ = wp$$ = *z0ptr++; | ||
67 | for ( z = z0-1 ; z >= 0 ; z-- ) { | ||
68 | wd$$ = *(u64 *)&dptr[z][d+$$*NSIZE]; | ||
69 | wp$$ = wp$$ ^ wd$$; | ||
70 | w2$$ = MASK(wq$$); | ||
71 | w1$$ = SHLBYTE(wq$$); | ||
72 | w2$$ = w2$$ & x1d; | ||
73 | w1$$ = w1$$ ^ w2$$; | ||
74 | wq$$ = w1$$ ^ wd$$; | ||
75 | } | ||
76 | *p++ = wp$$; | ||
77 | *q++ = wq$$; | ||
78 | } | ||
79 | } | ||
80 | |||
81 | const struct raid6_calls raid6_tilegx$# = { | ||
82 | raid6_tilegx$#_gen_syndrome, | ||
83 | NULL, | ||
84 | "tilegx$#", | ||
85 | 0 | ||
86 | }; | ||