From 36344762396ca868d6076c41a84bda25f1ed9d3c Mon Sep 17 00:00:00 2001
From: "David S. Miller" <davem@sunset.davemloft.net>
Date: Sat, 25 Feb 2006 17:16:29 -0800
Subject: [PATCH] [SPARC64]: Niagara optimized XOR functions for RAID.
Signed-off-by: David S. Miller <davem@davemloft.net>
---
arch/sparc64/kernel/sparc64_ksyms.c | 13 ++
arch/sparc64/lib/xor.S | 300 +++++++++++++++++++++++++++-
include/asm-sparc64/xor.h | 34 +++-
3 files changed, 342 insertions(+), 5 deletions(-)
diff --git a/arch/sparc64/kernel/sparc64_ksyms.c b/arch/sparc64/kernel/sparc64_ksyms.c
index 801fc0ce4844b6..e87fe7dfc7de27 100644
--- a/arch/sparc64/kernel/sparc64_ksyms.c
+++ b/arch/sparc64/kernel/sparc64_ksyms.c
@@ -108,6 +108,14 @@ extern void xor_vis_4(unsigned long, unsigned long *, unsigned long *,
extern void xor_vis_5(unsigned long, unsigned long *, unsigned long *,
unsigned long *, unsigned long *, unsigned long *);
+extern void xor_niagara_2(unsigned long, unsigned long *, unsigned long *);
+extern void xor_niagara_3(unsigned long, unsigned long *, unsigned long *,
+ unsigned long *);
+extern void xor_niagara_4(unsigned long, unsigned long *, unsigned long *,
+ unsigned long *, unsigned long *);
+extern void xor_niagara_5(unsigned long, unsigned long *, unsigned long *,
+ unsigned long *, unsigned long *, unsigned long *);
+
/* Per-CPU information table */
EXPORT_PER_CPU_SYMBOL(__cpu_data);
@@ -388,4 +396,9 @@ EXPORT_SYMBOL(xor_vis_3);
EXPORT_SYMBOL(xor_vis_4);
EXPORT_SYMBOL(xor_vis_5);
+EXPORT_SYMBOL(xor_niagara_2);
+EXPORT_SYMBOL(xor_niagara_3);
+EXPORT_SYMBOL(xor_niagara_4);
+EXPORT_SYMBOL(xor_niagara_5);
+
EXPORT_SYMBOL(prom_palette);
diff --git a/arch/sparc64/lib/xor.S b/arch/sparc64/lib/xor.S
index 4cd5d2be1ae1f1..a79c8888170d93 100644
--- a/arch/sparc64/lib/xor.S
+++ b/arch/sparc64/lib/xor.S
@@ -2,9 +2,10 @@
* arch/sparc64/lib/xor.S
*
* High speed xor_block operation for RAID4/5 utilizing the
- * UltraSparc Visual Instruction Set.
+ * UltraSparc Visual Instruction Set and Niagara store-init/twin-load.
*
* Copyright (C) 1997, 1999 Jakub Jelinek (jj@ultra.linux.cz)
+ * Copyright (C) 2006 David S. Miller <davem@davemloft.net>
*/
#include <asm/visasm.h>
@@ -19,6 +20,8 @@
*/
.text
.align 32
+
+ /* VIS versions. */
.globl xor_vis_2
.type xor_vis_2,#function
xor_vis_2:
@@ -352,3 +355,298 @@ xor_vis_5:
ret
restore
.size xor_vis_5, .-xor_vis_5
+
+ /* Niagara versions. */
+ .globl xor_niagara_2
+ .type xor_niagara_2,#function
+xor_niagara_2: /* %o0=bytes, %o1=dest, %o2=src */
+ save %sp, -192, %sp
+ prefetch [%i1], #n_writes
+ prefetch [%i2], #one_read
+ rd %asi, %g7
+ wr %g0, ASI_BLK_INIT_QUAD_LDD_P, %asi
+ srlx %i0, 6, %g1
+ mov %i1, %i0
+ mov %i2, %i1
+1: ldda [%i1 + 0x00] %asi, %i2 /* %i2/%i3 = src + 0x00 */
+ ldda [%i1 + 0x10] %asi, %i4 /* %i4/%i5 = src + 0x10 */
+ ldda [%i1 + 0x20] %asi, %g2 /* %g2/%g3 = src + 0x20 */
+ ldda [%i1 + 0x30] %asi, %l0 /* %l0/%l1 = src + 0x30 */
+ prefetch [%i1 + 0x40], #one_read
+ ldda [%i0 + 0x00] %asi, %o0 /* %o0/%o1 = dest + 0x00 */
+ ldda [%i0 + 0x10] %asi, %o2 /* %o2/%o3 = dest + 0x10 */
+ ldda [%i0 + 0x20] %asi, %o4 /* %o4/%o5 = dest + 0x20 */
+ ldda [%i0 + 0x30] %asi, %l2 /* %l2/%l3 = dest + 0x30 */
+ prefetch [%i0 + 0x40], #n_writes
+ xor %o0, %i2, %o0
+ xor %o1, %i3, %o1
+ stxa %o0, [%i0 + 0x00] %asi
+ stxa %o1, [%i0 + 0x08] %asi
+ xor %o2, %i4, %o2
+ xor %o3, %i5, %o3
+ stxa %o2, [%i0 + 0x10] %asi
+ stxa %o3, [%i0 + 0x18] %asi
+ xor %o4, %g2, %o4
+ xor %o5, %g3, %o5
+ stxa %o4, [%i0 + 0x20] %asi
+ stxa %o5, [%i0 + 0x28] %asi
+ xor %l2, %l0, %l2
+ xor %l3, %l1, %l3
+ stxa %l2, [%i0 + 0x30] %asi
+ stxa %l3, [%i0 + 0x38] %asi
+ add %i0, 0x40, %i0
+ subcc %g1, 1, %g1
+ bne,pt %xcc, 1b
+ add %i1, 0x40, %i1
+ membar #Sync
+ wr %g7, 0x0, %asi
+ ret
+ restore
+ .size xor_niagara_2, .-xor_niagara_2
+
+ .globl xor_niagara_3
+ .type xor_niagara_3,#function
+xor_niagara_3: /* %o0=bytes, %o1=dest, %o2=src1, %o3=src2 */
+ save %sp, -192, %sp
+ prefetch [%i1], #n_writes
+ prefetch [%i2], #one_read
+ prefetch [%i3], #one_read
+ rd %asi, %g7
+ wr %g0, ASI_BLK_INIT_QUAD_LDD_P, %asi
+ srlx %i0, 6, %g1
+ mov %i1, %i0
+ mov %i2, %i1
+ mov %i3, %l7
+1: ldda [%i1 + 0x00] %asi, %i2 /* %i2/%i3 = src1 + 0x00 */
+ ldda [%i1 + 0x10] %asi, %i4 /* %i4/%i5 = src1 + 0x10 */
+ ldda [%l7 + 0x00] %asi, %g2 /* %g2/%g3 = src2 + 0x00 */
+ ldda [%l7 + 0x10] %asi, %l0 /* %l0/%l1 = src2 + 0x10 */
+ ldda [%i0 + 0x00] %asi, %o0 /* %o0/%o1 = dest + 0x00 */
+ ldda [%i0 + 0x10] %asi, %o2 /* %o2/%o3 = dest + 0x10 */
+ xor %g2, %i2, %g2
+ xor %g3, %i3, %g3
+ xor %o0, %g2, %o0
+ xor %o1, %g3, %o1
+ stxa %o0, [%i0 + 0x00] %asi
+ stxa %o1, [%i0 + 0x08] %asi
+ ldda [%i1 + 0x20] %asi, %i2 /* %i2/%i3 = src1 + 0x20 */
+ ldda [%l7 + 0x20] %asi, %g2 /* %g2/%g3 = src2 + 0x20 */
+ ldda [%i0 + 0x20] %asi, %o0 /* %o0/%o1 = dest + 0x20 */
+ xor %l0, %i4, %l0
+ xor %l1, %i5, %l1
+ xor %o2, %l0, %o2
+ xor %o3, %l1, %o3
+ stxa %o2, [%i0 + 0x10] %asi
+ stxa %o3, [%i0 + 0x18] %asi
+ ldda [%i1 + 0x30] %asi, %i4 /* %i4/%i5 = src1 + 0x30 */
+ ldda [%l7 + 0x30] %asi, %l0 /* %l0/%l1 = src2 + 0x30 */
+ ldda [%i0 + 0x30] %asi, %o2 /* %o2/%o3 = dest + 0x30 */
+ prefetch [%i1 + 0x40], #one_read
+ prefetch [%l7 + 0x40], #one_read
+ prefetch [%i0 + 0x40], #n_writes
+ xor %g2, %i2, %g2
+ xor %g3, %i3, %g3
+ xor %o0, %g2, %o0
+ xor %o1, %g3, %o1
+ stxa %o0, [%i0 + 0x20] %asi
+ stxa %o1, [%i0 + 0x28] %asi
+ xor %l0, %i4, %l0
+ xor %l1, %i5, %l1
+ xor %o2, %l0, %o2
+ xor %o3, %l1, %o3
+ stxa %o2, [%i0 + 0x30] %asi
+ stxa %o3, [%i0 + 0x38] %asi
+ add %i0, 0x40, %i0
+ add %i1, 0x40, %i1
+ subcc %g1, 1, %g1
+ bne,pt %xcc, 1b
+ add %l7, 0x40, %l7
+ membar #Sync
+ wr %g7, 0x0, %asi
+ ret
+ restore
+ .size xor_niagara_3, .-xor_niagara_3
+
+ .globl xor_niagara_4
+ .type xor_niagara_4,#function
+xor_niagara_4: /* %o0=bytes, %o1=dest, %o2=src1, %o3=src2, %o4=src3 */
+ save %sp, -192, %sp
+ prefetch [%i1], #n_writes
+ prefetch [%i2], #one_read
+ prefetch [%i3], #one_read
+ prefetch [%i4], #one_read
+ rd %asi, %g7
+ wr %g0, ASI_BLK_INIT_QUAD_LDD_P, %asi
+ srlx %i0, 6, %g1
+ mov %i1, %i0
+ mov %i2, %i1
+ mov %i3, %l7
+ mov %i4, %l6
+1: ldda [%i1 + 0x00] %asi, %i2 /* %i2/%i3 = src1 + 0x00 */
+ ldda [%l7 + 0x00] %asi, %i4 /* %i4/%i5 = src2 + 0x00 */
+ ldda [%l6 + 0x00] %asi, %g2 /* %g2/%g3 = src3 + 0x00 */
+ ldda [%i0 + 0x00] %asi, %l0 /* %l0/%l1 = dest + 0x00 */
+ xor %i4, %i2, %i4
+ xor %i5, %i3, %i5
+ ldda [%i1 + 0x10] %asi, %i2 /* %i2/%i3 = src1 + 0x10 */
+ xor %g2, %i4, %g2
+ xor %g3, %i5, %g3
+ ldda [%i7 + 0x10] %asi, %i4 /* %i4/%i5 = src2 + 0x10 */
+ xor %l0, %g2, %l0
+ xor %l1, %g3, %l1
+ stxa %l0, [%i0 + 0x00] %asi
+ stxa %l1, [%i0 + 0x08] %asi
+ ldda [%i6 + 0x10] %asi, %g2 /* %g2/%g3 = src3 + 0x10 */
+ ldda [%i0 + 0x10] %asi, %l0 /* %l0/%l1 = dest + 0x10 */
+
+ xor %i4, %i2, %i4
+ xor %i5, %i3, %i5
+ ldda [%i1 + 0x20] %asi, %i2 /* %i2/%i3 = src1 + 0x20 */
+ xor %g2, %i4, %g2
+ xor %g3, %i5, %g3
+ ldda [%i7 + 0x20] %asi, %i4 /* %i4/%i5 = src2 + 0x20 */
+ xor %l0, %g2, %l0
+ xor %l1, %g3, %l1
+ stxa %l0, [%i0 + 0x10] %asi
+ stxa %l1, [%i0 + 0x18] %asi
+ ldda [%i6 + 0x20] %asi, %g2 /* %g2/%g3 = src3 + 0x20 */
+ ldda [%i0 + 0x20] %asi, %l0 /* %l0/%l1 = dest + 0x20 */
+
+ xor %i4, %i2, %i4
+ xor %i5, %i3, %i5
+ ldda [%i1 + 0x30] %asi, %i2 /* %i2/%i3 = src1 + 0x30 */
+ xor %g2, %i4, %g2
+ xor %g3, %i5, %g3
+ ldda [%i7 + 0x30] %asi, %i4 /* %i4/%i5 = src2 + 0x30 */
+ xor %l0, %g2, %l0
+ xor %l1, %g3, %l1
+ stxa %l0, [%i0 + 0x20] %asi
+ stxa %l1, [%i0 + 0x28] %asi
+ ldda [%i6 + 0x30] %asi, %g2 /* %g2/%g3 = src3 + 0x30 */
+ ldda [%i0 + 0x30] %asi, %l0 /* %l0/%l1 = dest + 0x30 */
+
+ prefetch [%i1 + 0x40], #one_read
+ prefetch [%l7 + 0x40], #one_read
+ prefetch [%l6 + 0x40], #one_read
+ prefetch [%i0 + 0x40], #n_writes
+
+ xor %i4, %i2, %i4
+ xor %i5, %i3, %i5
+ xor %g2, %i4, %g2
+ xor %g3, %i5, %g3
+ xor %l0, %g2, %l0
+ xor %l1, %g3, %l1
+ stxa %l0, [%i0 + 0x30] %asi
+ stxa %l1, [%i0 + 0x38] %asi
+
+ add %i0, 0x40, %i0
+ add %i1, 0x40, %i1
+ add %l7, 0x40, %l7
+ subcc %g1, 1, %g1
+ bne,pt %xcc, 1b
+ add %l6, 0x40, %l6
+ membar #Sync
+ wr %g7, 0x0, %asi
+ ret
+ restore
+ .size xor_niagara_4, .-xor_niagara_4
+
+ .globl xor_niagara_5
+ .type xor_niagara_5,#function
+xor_niagara_5: /* %o0=bytes, %o1=dest, %o2=src1, %o3=src2, %o4=src3, %o5=src4 */
+ save %sp, -192, %sp
+ prefetch [%i1], #n_writes
+ prefetch [%i2], #one_read
+ prefetch [%i3], #one_read
+ prefetch [%i4], #one_read
+ prefetch [%i5], #one_read
+ rd %asi, %g7
+ wr %g0, ASI_BLK_INIT_QUAD_LDD_P, %asi
+ srlx %i0, 6, %g1
+ mov %i1, %i0
+ mov %i2, %i1
+ mov %i3, %l7
+ mov %i4, %l6
+ mov %i5, %l5
+1: ldda [%i1 + 0x00] %asi, %i2 /* %i2/%i3 = src1 + 0x00 */
+ ldda [%l7 + 0x00] %asi, %i4 /* %i4/%i5 = src2 + 0x00 */
+ ldda [%l6 + 0x00] %asi, %g2 /* %g2/%g3 = src3 + 0x00 */
+ ldda [%l5 + 0x00] %asi, %l0 /* %l0/%l1 = src4 + 0x00 */
+ ldda [%i0 + 0x00] %asi, %l2 /* %l2/%l3 = dest + 0x00 */
+ xor %i4, %i2, %i4
+ xor %i5, %i3, %i5
+ ldda [%i1 + 0x10] %asi, %i2 /* %i2/%i3 = src1 + 0x10 */
+ xor %g2, %i4, %g2
+ xor %g3, %i5, %g3
+ ldda [%l7 + 0x10] %asi, %i4 /* %i4/%i5 = src2 + 0x10 */
+ xor %l0, %g2, %l0
+ xor %l1, %g3, %l1
+ ldda [%l6 + 0x10] %asi, %g2 /* %g2/%g3 = src3 + 0x10 */
+ xor %l2, %l0, %l2
+ xor %l3, %l1, %l3
+ stxa %l2, [%i0 + 0x00] %asi
+ stxa %l3, [%i0 + 0x08] %asi
+ ldda [%l5 + 0x10] %asi, %l0 /* %l0/%l1 = src4 + 0x10 */
+ ldda [%i0 + 0x10] %asi, %l2 /* %l2/%l3 = dest + 0x10 */
+
+ xor %i4, %i2, %i4
+ xor %i5, %i3, %i5
+ ldda [%i1 + 0x20] %asi, %i2 /* %i2/%i3 = src1 + 0x20 */
+ xor %g2, %i4, %g2
+ xor %g3, %i5, %g3
+ ldda [%l7 + 0x20] %asi, %i4 /* %i4/%i5 = src2 + 0x20 */
+ xor %l0, %g2, %l0
+ xor %l1, %g3, %l1
+ ldda [%l6 + 0x20] %asi, %g2 /* %g2/%g3 = src3 + 0x20 */
+ xor %l2, %l0, %l2
+ xor %l3, %l1, %l3
+ stxa %l2, [%i0 + 0x10] %asi
+ stxa %l3, [%i0 + 0x18] %asi
+ ldda [%l5 + 0x20] %asi, %l0 /* %l0/%l1 = src4 + 0x20 */
+ ldda [%i0 + 0x20] %asi, %l2 /* %l2/%l3 = dest + 0x20 */
+
+ xor %i4, %i2, %i4
+ xor %i5, %i3, %i5
+ ldda [%i1 + 0x30] %asi, %i2 /* %i2/%i3 = src1 + 0x30 */
+ xor %g2, %i4, %g2
+ xor %g3, %i5, %g3
+ ldda [%l7 + 0x30] %asi, %i4 /* %i4/%i5 = src2 + 0x30 */
+ xor %l0, %g2, %l0
+ xor %l1, %g3, %l1
+ ldda [%l6 + 0x30] %asi, %g2 /* %g2/%g3 = src3 + 0x30 */
+ xor %l2, %l0, %l2
+ xor %l3, %l1, %l3
+ stxa %l2, [%i0 + 0x20] %asi
+ stxa %l3, [%i0 + 0x28] %asi
+ ldda [%l5 + 0x30] %asi, %l0 /* %l0/%l1 = src4 + 0x30 */
+ ldda [%i0 + 0x30] %asi, %l2 /* %l2/%l3 = dest + 0x30 */
+
+ prefetch [%i1 + 0x40], #one_read
+ prefetch [%l7 + 0x40], #one_read
+ prefetch [%l6 + 0x40], #one_read
+ prefetch [%l5 + 0x40], #one_read
+ prefetch [%i0 + 0x40], #n_writes
+
+ xor %i4, %i2, %i4
+ xor %i5, %i3, %i5
+ xor %g2, %i4, %g2
+ xor %g3, %i5, %g3
+ xor %l0, %g2, %l0
+ xor %l1, %g3, %l1
+ xor %l2, %l0, %l2
+ xor %l3, %l1, %l3
+ stxa %l2, [%i0 + 0x30] %asi
+ stxa %l3, [%i0 + 0x38] %asi
+
+ add %i0, 0x40, %i0
+ add %i1, 0x40, %i1
+ add %l7, 0x40, %l7
+ add %l6, 0x40, %l6
+ subcc %g1, 1, %g1
+ bne,pt %xcc, 1b
+ add %l5, 0x40, %l5
+ membar #Sync
+ wr %g7, 0x0, %asi
+ ret
+ restore
+ .size xor_niagara_5, .-xor_niagara_5
diff --git a/include/asm-sparc64/xor.h b/include/asm-sparc64/xor.h
index 8b3a7e4b6062b6..8ce3f1813e280b 100644
--- a/include/asm-sparc64/xor.h
+++ b/include/asm-sparc64/xor.h
@@ -2,9 +2,11 @@
* include/asm-sparc64/xor.h
*
* High speed xor_block operation for RAID4/5 utilizing the
- * UltraSparc Visual Instruction Set.
+ * UltraSparc Visual Instruction Set and Niagara block-init
+ * twin-load instructions.
*
* Copyright (C) 1997, 1999 Jakub Jelinek (jj@ultra.linux.cz)
+ * Copyright (C) 2006 David S. Miller <davem@davemloft.net>
*
* 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
@@ -16,8 +18,7 @@
* Software Foundation, Inc., 675 Mass Ave, Cambridge, MA 02139, USA.
*/
-#include <asm/pstate.h>
-#include <asm/asi.h>
+#include <asm/spitfire.h>
extern void xor_vis_2(unsigned long, unsigned long *, unsigned long *);
extern void xor_vis_3(unsigned long, unsigned long *, unsigned long *,
@@ -37,4 +38,29 @@ static struct xor_block_template xor_block_VIS = {
.do_5 = xor_vis_5,
};
-#define XOR_TRY_TEMPLATES xor_speed(&xor_block_VIS)
+extern void xor_niagara_2(unsigned long, unsigned long *, unsigned long *);
+extern void xor_niagara_3(unsigned long, unsigned long *, unsigned long *,
+ unsigned long *);
+extern void xor_niagara_4(unsigned long, unsigned long *, unsigned long *,
+ unsigned long *, unsigned long *);
+extern void xor_niagara_5(unsigned long, unsigned long *, unsigned long *,
+ unsigned long *, unsigned long *, unsigned long *);
+
+static struct xor_block_template xor_block_niagara = {
+ .name = "Niagara",
+ .do_2 = xor_niagara_2,
+ .do_3 = xor_niagara_3,
+ .do_4 = xor_niagara_4,
+ .do_5 = xor_niagara_5,
+};
+
+#undef XOR_TRY_TEMPLATES
+#define XOR_TRY_TEMPLATES \
+ do { \
+ xor_speed(&xor_block_VIS); \
+ xor_speed(&xor_block_niagara); \
+ } while (0)
+
+/* For VIS for everything except Niagara. */
+#define XOR_SELECT_TEMPLATE(FASTEST) \
+ (tlb_type == hypervisor ? &xor_block_niagara : &xor_block_VIS)
--
GitLab