This adds optimized memset/bzero/page-clear routines for Niagara-4.
We basically can do what powerpc has been able to do for a decade (via
the "dcbz" instruction), which is use cache line clearing stores for
bzero and memsets with a 'c' argument of zero.
As long as we make the cache initializing store to each 32-byte
subblock of the L2 cache line, it works.
As with other Niagara-4 optimized routines, the key is to make sure to
avoid any usage of the %asi register, as reads and writes to it cost
at least 50 cycles.
For the user clear cases, we don't use these new routines, we use the
Niagara-1 variants instead. Those have to use %asi in an unavoidable
way.
A Niagara-4 8K page clear costs just under 600 cycles.
Add definitions of the MRU variants of the cache initializing store
ASIs. By default, cache initializing stores install the line as Least
Recently Used. If we know we're going to use the data immediately
(which is true for page copies and clears) we can use the Most
Recently Used variant, to decrease the likelyhood of the lines being
evicted before they get used.
Signed-off-by: David S. Miller <davem@davemloft.net>
#define ASI_BLK_INIT_QUAD_LDD_P 0xe2 /* (NG) init-store, twin load,
* primary, implicit
*/
+#define ASI_BLK_INIT_QUAD_LDD_S 0xe3 /* (NG) init-store, twin load,
+ * secondary, implicit
+ */
#define ASI_BLK_P 0xf0 /* Primary, blk ld/st */
#define ASI_BLK_S 0xf1 /* Secondary, blk ld/st */
+#define ASI_ST_BLKINIT_MRU_P 0xf2 /* (NG4) init-store, twin load,
+ * Most-Recently-Used, primary,
+ * implicit
+ */
+#define ASI_ST_BLKINIT_MRU_S 0xf2 /* (NG4) init-store, twin load,
+ * Most-Recently-Used, secondary,
+ * implicit
+ */
#define ASI_BLK_PL 0xf8 /* Primary, blk ld/st, little */
#define ASI_BLK_SL 0xf9 /* Secondary, blk ld/st, little */
+#define ASI_ST_BLKINIT_MRU_PL 0xfa /* (NG4) init-store, twin load,
+ * Most-Recently-Used, primary,
+ * implicit, little-endian
+ */
+#define ASI_ST_BLKINIT_MRU_SL 0xfb /* (NG4) init-store, twin load,
+ * Most-Recently-Used, secondary,
+ * implicit, little-endian
+ */
#endif /* _SPARC_ASI_H */
niagara4_patch:
call niagara4_patch_copyops
nop
- call niagara_patch_bzero
+ call niagara4_patch_bzero
nop
call niagara4_patch_pageops
nop
lib-$(CONFIG_SPARC64) += NG2patch.o
lib-$(CONFIG_SPARC64) += NG4memcpy.o NG4copy_from_user.o NG4copy_to_user.o
-lib-$(CONFIG_SPARC64) += NG4patch.o NG4copy_page.o
+lib-$(CONFIG_SPARC64) += NG4patch.o NG4copy_page.o NG4clear_page.o NG4memset.o
lib-$(CONFIG_SPARC64) += GENmemcpy.o GENcopy_from_user.o GENcopy_to_user.o
lib-$(CONFIG_SPARC64) += GENpatch.o GENpage.o GENbzero.o
--- /dev/null
+/* NG4copy_page.S: Niagara-4 optimized clear page.
+ *
+ * Copyright (C) 2012 (davem@davemloft.net)
+ */
+
+#include <asm/asi.h>
+#include <asm/page.h>
+
+ .text
+
+ .register %g3, #scratch
+
+ .align 32
+ .globl NG4clear_page
+ .globl NG4clear_user_page
+NG4clear_page: /* %o0=dest */
+NG4clear_user_page: /* %o0=dest, %o1=vaddr */
+ set PAGE_SIZE, %g7
+ mov 0x20, %g3
+1: stxa %g0, [%o0 + %g0] ASI_ST_BLKINIT_MRU_P
+ subcc %g7, 0x40, %g7
+ stxa %g0, [%o0 + %g3] ASI_ST_BLKINIT_MRU_P
+ bne,pt %xcc, 1b
+ add %o0, 0x40, %o0
+ membar #StoreLoad|#StoreStore
+ retl
+ nop
+ .size NG4clear_page,.-NG4clear_page
+ .size NG4clear_user_page,.-NG4clear_user_page
\ No newline at end of file
ldx [%o1 + 0x10], %o4
ldx [%o1 + 0x18], %o5
ldx [%o1 + 0x20], %g1
- stxa %o2, [%o0] ASI_BLK_INIT_QUAD_LDD_P
+ stxa %o2, [%o0] ASI_ST_BLKINIT_MRU_P
add %o0, 0x08, %o0
ldx [%o1 + 0x28], %g2
- stxa %o3, [%o0] ASI_BLK_INIT_QUAD_LDD_P
+ stxa %o3, [%o0] ASI_ST_BLKINIT_MRU_P
add %o0, 0x08, %o0
ldx [%o1 + 0x30], %g3
- stxa %o4, [%o0] ASI_BLK_INIT_QUAD_LDD_P
+ stxa %o4, [%o0] ASI_ST_BLKINIT_MRU_P
add %o0, 0x08, %o0
ldx [%o1 + 0x38], %o2
add %o1, 0x40, %o1
- stxa %o5, [%o0] ASI_BLK_INIT_QUAD_LDD_P
+ stxa %o5, [%o0] ASI_ST_BLKINIT_MRU_P
add %o0, 0x08, %o0
- stxa %g1, [%o0] ASI_BLK_INIT_QUAD_LDD_P
+ stxa %g1, [%o0] ASI_ST_BLKINIT_MRU_P
add %o0, 0x08, %o0
- stxa %g2, [%o0] ASI_BLK_INIT_QUAD_LDD_P
+ stxa %g2, [%o0] ASI_ST_BLKINIT_MRU_P
add %o0, 0x08, %o0
- stxa %g3, [%o0] ASI_BLK_INIT_QUAD_LDD_P
+ stxa %g3, [%o0] ASI_ST_BLKINIT_MRU_P
add %o0, 0x08, %o0
- stxa %o2, [%o0] ASI_BLK_INIT_QUAD_LDD_P
+ stxa %o2, [%o0] ASI_ST_BLKINIT_MRU_P
add %o0, 0x08, %o0
bne,pt %icc, 1b
prefetch [%o1 + 0x200], #n_reads_strong
--- /dev/null
+/* NG4memset.S: Niagara-4 optimized memset/bzero.
+ *
+ * Copyright (C) 2012 David S. Miller (davem@davemloft.net)
+ */
+
+#include <asm/asi.h>
+
+ .register %g2, #scratch
+ .register %g3, #scratch
+
+ .text
+ .align 32
+ .globl NG4memset
+NG4memset:
+ andcc %o1, 0xff, %o4
+ be,pt %icc, 1f
+ mov %o2, %o1
+ sllx %o4, 8, %g1
+ or %g1, %o4, %o2
+ sllx %o2, 16, %g1
+ or %g1, %o2, %o2
+ sllx %o2, 32, %g1
+ ba,pt %icc, 1f
+ or %g1, %o2, %o4
+ .size NG4memset,.-NG4memset
+
+ .align 32
+ .globl NG4bzero
+NG4bzero:
+ clr %o4
+1: cmp %o1, 16
+ ble %icc, .Ltiny
+ mov %o0, %o3
+ sub %g0, %o0, %g1
+ and %g1, 0x7, %g1
+ brz,pt %g1, .Laligned8
+ sub %o1, %g1, %o1
+1: stb %o4, [%o0 + 0x00]
+ subcc %g1, 1, %g1
+ bne,pt %icc, 1b
+ add %o0, 1, %o0
+.Laligned8:
+ cmp %o1, 64 + (64 - 8)
+ ble .Lmedium
+ sub %g0, %o0, %g1
+ andcc %g1, (64 - 1), %g1
+ brz,pn %g1, .Laligned64
+ sub %o1, %g1, %o1
+1: stx %o4, [%o0 + 0x00]
+ subcc %g1, 8, %g1
+ bne,pt %icc, 1b
+ add %o0, 0x8, %o0
+.Laligned64:
+ andn %o1, 64 - 1, %g1
+ sub %o1, %g1, %o1
+ brnz,pn %o4, .Lnon_bzero_loop
+ mov 0x20, %g2
+1: stxa %o4, [%o0 + %g0] ASI_BLK_INIT_QUAD_LDD_P
+ subcc %g1, 0x40, %g1
+ stxa %o4, [%o0 + %g2] ASI_BLK_INIT_QUAD_LDD_P
+ bne,pt %icc, 1b
+ add %o0, 0x40, %o0
+.Lpostloop:
+ cmp %o1, 8
+ bl,pn %icc, .Ltiny
+ membar #StoreStore|#StoreLoad
+.Lmedium:
+ andn %o1, 0x7, %g1
+ sub %o1, %g1, %o1
+1: stx %o4, [%o0 + 0x00]
+ subcc %g1, 0x8, %g1
+ bne,pt %icc, 1b
+ add %o0, 0x08, %o0
+ andcc %o1, 0x4, %g1
+ be,pt %icc, .Ltiny
+ sub %o1, %g1, %o1
+ stw %o4, [%o0 + 0x00]
+ add %o0, 0x4, %o0
+.Ltiny:
+ cmp %o1, 0
+ be,pn %icc, .Lexit
+1: subcc %o1, 1, %o1
+ stb %o4, [%o0 + 0x00]
+ bne,pt %icc, 1b
+ add %o0, 1, %o0
+.Lexit:
+ retl
+ mov %o3, %o0
+.Lnon_bzero_loop:
+ mov 0x08, %g3
+ mov 0x28, %o5
+1: stxa %o4, [%o0 + %g0] ASI_BLK_INIT_QUAD_LDD_P
+ subcc %g1, 0x40, %g1
+ stxa %o4, [%o0 + %g2] ASI_BLK_INIT_QUAD_LDD_P
+ stxa %o4, [%o0 + %g3] ASI_BLK_INIT_QUAD_LDD_P
+ stxa %o4, [%o0 + %o5] ASI_BLK_INIT_QUAD_LDD_P
+ add %o0, 0x10, %o0
+ stxa %o4, [%o0 + %g0] ASI_BLK_INIT_QUAD_LDD_P
+ stxa %o4, [%o0 + %g2] ASI_BLK_INIT_QUAD_LDD_P
+ stxa %o4, [%o0 + %g3] ASI_BLK_INIT_QUAD_LDD_P
+ stxa %o4, [%o0 + %o5] ASI_BLK_INIT_QUAD_LDD_P
+ bne,pt %icc, 1b
+ add %o0, 0x30, %o0
+ ba,a,pt %icc, .Lpostloop
+ .size NG4bzero,.-NG4bzero
nop
.size niagara4_patch_copyops,.-niagara4_patch_copyops
+ .globl niagara4_patch_bzero
+ .type niagara4_patch_bzero,#function
+niagara4_patch_bzero:
+ NG_DO_PATCH(memset, NG4memset)
+ NG_DO_PATCH(__bzero, NG4bzero)
+ NG_DO_PATCH(__clear_user, NGclear_user)
+ NG_DO_PATCH(tsb_init, NGtsb_init)
+ retl
+ nop
+ .size niagara4_patch_bzero,.-niagara4_patch_bzero
+
.globl niagara4_patch_pageops
.type niagara4_patch_pageops,#function
niagara4_patch_pageops:
NG_DO_PATCH(copy_user_page, NG4copy_user_page)
- NG_DO_PATCH(_clear_page, NGclear_page)
- NG_DO_PATCH(clear_user_page, NGclear_user_page)
+ NG_DO_PATCH(_clear_page, NG4clear_page)
+ NG_DO_PATCH(clear_user_page, NG4clear_user_page)
retl
nop
.size niagara4_patch_pageops,.-niagara4_patch_pageops
projects / ~emulex / infiniband.git / commitdiff