summaryrefslogtreecommitdiffstats
path: root/arch/sparc64/lib/VIScopy.S
diff options
context:
space:
mode:
Diffstat (limited to 'arch/sparc64/lib/VIScopy.S')
-rw-r--r--arch/sparc64/lib/VIScopy.S115
1 files changed, 66 insertions, 49 deletions
diff --git a/arch/sparc64/lib/VIScopy.S b/arch/sparc64/lib/VIScopy.S
index 40b781e73..7f2f497cd 100644
--- a/arch/sparc64/lib/VIScopy.S
+++ b/arch/sparc64/lib/VIScopy.S
@@ -1,9 +1,9 @@
-/* $Id: VIScopy.S,v 1.14 1997/08/22 15:54:53 jj Exp $
+/* $Id: VIScopy.S,v 1.18 1998/06/12 14:53:55 jj Exp $
* VIScopy.S: High speed copy operations utilizing the UltraSparc
* Visual Instruction Set.
*
* Copyright (C) 1997 David S. Miller (davem@caip.rutgers.edu)
- * Copyright (C) 1996, 1997 Jakub Jelinek (jj@sunsite.mff.cuni.cz)
+ * Copyright (C) 1996, 1997, 1998 Jakub Jelinek (jj@ultra.linux.cz)
*/
#include "VIS.h"
@@ -24,12 +24,15 @@
*/
#ifdef __KERNEL__
+
+#include <asm/visasm.h>
+
#define FPU_CLEAN_RETL \
- wr %g0, 0, %fprs; \
+ VISExit \
retl; \
clr %o0;
#define FPU_RETL \
- wr %g0, 0, %fprs; \
+ VISExit \
retl; \
clr %o0;
#define NORMAL_RETL \
@@ -40,7 +43,7 @@
.section .fixup; \
.align 4; \
99: ba VIScopyfixup_ret; \
- a, b, %o0; \
+ a, b, %o1; \
.section __ex_table; \
.align 4; \
.word 98b, 99b; \
@@ -52,7 +55,7 @@
.align 4; \
99: c, d, e; \
ba VIScopyfixup_ret; \
- a, b, %o0; \
+ a, b, %o1; \
.section __ex_table; \
.align 4; \
.word 98b, 99b; \
@@ -298,10 +301,6 @@
.globl __memcpy_entry
.type __memcpy_entry,@function
-
- .globl copy_page
- .type copy_page,@function
-
memcpy_private:
__memcpy:
memcpy: mov ASI_BLK_P, asi_src ! IEU0 Group
@@ -310,12 +309,6 @@ memcpy: mov ASI_BLK_P, asi_src ! IEU0 Group
retl
clr %o0
-copy_page: wr %g0, FPRS_FEF, %fprs ! FPU Group
- sethi %hi(8192), %o2 ! IEU0 Group
- mov ASI_BLK_P, asi_src ! IEU1
- b,pt %xcc, dest_is_64byte_aligned ! CTI
- mov ASI_BLK_P, asi_dest ! IEU0 Group
-
.align 32
.globl __copy_from_user
.type __copy_from_user,@function
@@ -355,7 +348,11 @@ __memcpy_384plus:
#endif
VIS_enter:
be,pt %xcc, dest_is_8byte_aligned ! CTI
+#ifdef __KERNEL__
+ nop ! IEU0 Group
+#else
andcc %o0, 0x38, %g5 ! IEU1 Group
+#endif
do_dest_8byte_align:
mov 8, %g1 ! IEU0
sub %g1, %g2, %g2 ! IEU0 Group
@@ -377,7 +374,8 @@ do_dest_8byte_align:
EX(LDUB [%o1] ASINORMAL, %o5,
add %o2, %g2) ! Load Group
add %o0, 2, %o0 ! IEU0
- EX(LDUB [%o1 + 1] ASINORMAL, %g3,
+ EX2(LDUB [%o1 + 1] ASINORMAL, %g3,
+ sub %o0, 2, %o0,
add %o2, %g2) ! Load Group
ASI_SETDST_NOBLK ! LSU Group
subcc %g2, 2, %g2 ! IEU1 Group
@@ -389,17 +387,17 @@ do_dest_8byte_align:
EX2(STB %g3, [%o0 - 1] ASINORMAL,
add %g2, 1, %g2,
add %o2, %g2) ! Store
-3: andcc %o0, 0x38, %g5 ! IEU1 Group
-dest_is_8byte_aligned:
- be,pt %icc, dest_is_64byte_aligned ! CTI
#ifdef __KERNEL__
- wr %g0, FPRS_FEF, %fprs ! FPU Group
-do_dest_64byte_align:
- mov 64, %g1 ! IEU0 Group
+3:
+dest_is_8byte_aligned:
+ VISEntry
+ andcc %o0, 0x38, %g5 ! IEU1 Group
#else
- mov 64, %g1 ! IEU0 Group
-do_dest_64byte_align:
+3: andcc %o0, 0x38, %g5 ! IEU1 Group
+dest_is_8byte_aligned:
#endif
+ be,pt %icc, dest_is_64byte_aligned ! CTI
+ mov 64, %g1 ! IEU0
fmovd %f0, %f2 ! FPU
sub %g1, %g5, %g5 ! IEU0 Group
ASI_SETSRC_NOBLK ! LSU Group
@@ -646,7 +644,9 @@ __memcpy_short:
2: ASI_SETSRC_NOBLK ! LSU Group
EXO2(LDUB [%o1] ASINORMAL, %g5) ! LOAD Group
add %o0, 2, %o0 ! IEU0
- EXO2(LDUB [%o1 + 1] ASINORMAL, %o5) ! LOAD Group
+ EX2(LDUB [%o1 + 1] ASINORMAL, %o5,
+ sub %o0, 2, %o0,
+ add %o2, %g0) ! LOAD Group
add %o1, 2, %o1 ! IEU0
ASI_SETDST_NOBLK ! LSU Group
subcc %o2, 2, %o2 ! IEU1 Group
@@ -866,9 +866,9 @@ normal_retl:
ASI_SETSRC_NOBLK ! LSU Group
EX(LDX [%o1] ASINORMAL, %g2,
and %o2, 0xf) ! Load Group
- add %o1, 8, %o1 ! IEU0
+ add %o0, 8, %o0 ! IEU0
ASI_SETDST_NOBLK ! LSU Group
- add %o0, 8, %o0 ! IEU0 Group
+ add %o1, 8, %o1 ! IEU0 Group
EX(STX %g2, [%o0 - 0x8] ASINORMAL,
and %o2, 0xf) ! Store
85: be,pt %xcc, 1f ! CTI
@@ -876,9 +876,9 @@ normal_retl:
ASI_SETSRC_NOBLK ! LSU Group
EX(LDUW [%o1] ASINORMAL, %g2,
and %o2, 0x7) ! Load Group
- add %o1, 4, %o1 ! IEU0
+ add %o0, 4, %o0 ! IEU0
ASI_SETDST_NOBLK ! LSU Group
- add %o0, 4, %o0 ! IEU0 Group
+ add %o1, 4, %o1 ! IEU0 Group
EX(STW %g2, [%o0 - 0x4] ASINORMAL,
and %o2, 0x7) ! Store
1: be,pt %xcc, 1f ! CTI
@@ -886,9 +886,9 @@ normal_retl:
ASI_SETSRC_NOBLK ! LSU Group
EX(LDUH [%o1] ASINORMAL, %g2,
and %o2, 0x3) ! Load Group
- add %o1, 2, %o1 ! IEU0
+ add %o0, 2, %o0 ! IEU0
ASI_SETDST_NOBLK ! LSU Group
- add %o0, 2, %o0 ! IEU0 Group
+ add %o1, 2, %o1 ! IEU0 Group
EX(STH %g2, [%o0 - 0x2] ASINORMAL,
and %o2, 0x3) ! Store
1: be,pt %xcc, 1f ! CTI
@@ -920,7 +920,7 @@ memcpy_noVIS_misaligned:
add %o2, 1) ! Store
2:
#ifdef __KERNEL__
- wr %g0, FPRS_FEF, %fprs ! FPU Group
+ VISEntry
#endif
andn %o2, 7, %g5 ! IEU0 Group
and %o2, 7, %o2 ! IEU1
@@ -976,16 +976,31 @@ fpu_retl:
.section .fixup
.align 4
VIScopyfixup_reto2:
- mov %o2, %o0
+ mov %o2, %o1
VIScopyfixup_ret:
+ /* If this is copy_from_user(), zero out the rest of the
+ * kernel buffer.
+ */
+ andcc asi_src, 0x1, %g0
+ be,pt %icc, 1f
+ andcc asi_dest, 0x1, %g0
+ bne,pn %icc, 1f
+ VISExit
+ save %sp, -160, %sp
+ mov %i0, %o0
+ call __bzero
+ mov %i1, %o1
+ restore
+1: mov %o1, %o0
retl
- wr %g0, 0, %fprs
+ nop
VIScopyfixup1: subcc %g2, 18, %g2
+ add %o0, 32, %o0
bgeu,a,pt %icc, VIScopyfixup1
sub %g7, 32, %g7
+ sub %o0, 32, %o0
rd %pc, %g5
- add %g2, 18, %g2
- add %g2, 20, %g2
+ add %g2, (18 + 16), %g2
ldub [%g5 + %g2], %g2
ba,a,pt %xcc, 2f
.byte 0, 0, 0, 0, 0, 0, 0, 4, 4, 8, 12, 12, 16, 20, 20, 24, 28, 28
@@ -994,41 +1009,43 @@ VIScopyfixup2: mov (7 * 16), %g7
1: subcc %g2, 10, %g2
bgeu,a,pt %icc, 1b
sub %g7, 16, %g7
+ sub %o0, %g7, %o0
rd %pc, %g5
- add %g2, 10, %g2
- add %g2, 20, %g2
+ add %g2, (10 + 16), %g2
ldub [%g5 + %g2], %g2
ba,a,pt %xcc, 4f
.byte 0, 0, 0, 0, 0, 4, 4, 8, 12, 12
.align 4
VIScopyfixup3: subcc %g2, 10, %g2
+ add %o0, 32, %o0
bgeu,a,pt %icc, VIScopyfixup3
sub %g7, 32, %g7
+ sub %o0, 32, %o0
rd %pc, %g5
- add %g2, 10, %g2
- add %g2, 20, %g2
+ add %g2, (10 + 16), %g2
ldub [%g5 + %g2], %g2
ba,a,pt %xcc, 2f
.byte 0, 0, 0, 0, 0, 0, 0, 8, 16, 24
.align 4
-2: and %g1, 0x7f, %g1
+2: and %o2, 0x7f, %o2
sub %g7, %g2, %g7
ba,pt %xcc, VIScopyfixup_ret
- add %g7, %g1, %o0
+ add %g7, %o2, %o1
VIScopyfixup4: mov (7 * 16), %g7
3: subcc %g2, 6, %g2
bgeu,a,pt %icc, 3b
sub %g7, 16, %g7
+ sub %o0, %g7, %o0
rd %pc, %g5
- add %g2, 6, %g2
- add %g2, 20, %g2
+ add %g2, (6 + 16), %g2
ldub [%g5 + %g2], %g2
ba,a,pt %xcc, 4f
.byte 0, 0, 0, 0, 0, 8
.align 4
-4: and %g1, 7, %g1
+4: and %o2, 0xf, %o2
+ sub %g7, %g2, %g7
ba,pt %xcc, VIScopyfixup_ret
- add %g7, %g1, %o0
+ add %g7, %o2, %o1
VIScopyfixup_vis3:
sub %o2, 0x80, %o2
VIScopyfixup_vis2:
@@ -1038,13 +1055,13 @@ VIScopyfixup_vis0:
VIScopyfixup_vis1:
add %g7, %g3, %g7
ba,pt %xcc, VIScopyfixup_ret
- add %o2, %g7, %o0
+ add %o2, %g7, %o1
VIScopyfixup_vis5:
add %g3, 8, %g3
VIScopyfixup_vis4:
add %g3, 8, %g3
ba,pt %xcc, VIScopyfixup_ret
- add %o2, %g3, %o0
+ add %o2, %g3, %o1
#endif
#ifdef __KERNEL__
generated by cgit v1.2.3 (git 2.25.1) at 2024-09-20 18:46:11 +0000