summaryrefslogtreecommitdiff
path: root/pfinet/linux-src/arch/sparc64/lib/checksum.S
diff options
context:
space:
mode:
Diffstat (limited to 'pfinet/linux-src/arch/sparc64/lib/checksum.S')
-rw-r--r--pfinet/linux-src/arch/sparc64/lib/checksum.S278
1 files changed, 278 insertions, 0 deletions
diff --git a/pfinet/linux-src/arch/sparc64/lib/checksum.S b/pfinet/linux-src/arch/sparc64/lib/checksum.S
new file mode 100644
index 00000000..ea732b36
--- /dev/null
+++ b/pfinet/linux-src/arch/sparc64/lib/checksum.S
@@ -0,0 +1,278 @@
+/* checksum.S: Sparc V9 optimized checksum code.
+ *
+ * Copyright(C) 1995 Linus Torvalds
+ * Copyright(C) 1995 Miguel de Icaza
+ * Copyright(C) 1996 David S. Miller
+ * Copyright(C) 1997 Jakub Jelinek
+ *
+ * derived from:
+ * Linux/Alpha checksum c-code
+ * Linux/ix86 inline checksum assembly
+ * RFC1071 Computing the Internet Checksum (esp. Jacobsons m68k code)
+ * David Mosberger-Tang for optimized reference c-code
+ * BSD4.4 portable checksum routine
+ */
+
+#include <asm/errno.h>
+#include <asm/head.h>
+#include <asm/ptrace.h>
+#include <asm/asi.h>
+#include <asm/page.h>
+
+ /* The problem with the "add with carry" instructions on Ultra
+ * are two fold. Firstly, they cannot pair with jack shit,
+ * and also they only add in the 32-bit carry condition bit
+ * into the accumulated sum. The following is much better.
+ * For larger chunks we use VIS code, which is faster ;)
+ */
+
+#define src o0
+#define dst o1
+#define len o2
+#define sum o3
+
+ .text
+ /* I think I have an erection... Once _AGAIN_ the SunSoft
+ * engineers are caught asleep at the keyboard, tsk tsk...
+ */
+
+#define CSUMCOPY_LASTCHUNK(off, t0, t1) \
+ ldxa [%src - off - 0x08] %asi, t0; \
+ ldxa [%src - off - 0x00] %asi, t1; \
+ nop; nop; \
+ addcc t0, %sum, %sum; \
+ stw t0, [%dst - off - 0x04]; \
+ srlx t0, 32, t0; \
+ bcc,pt %xcc, 51f; \
+ stw t0, [%dst - off - 0x08]; \
+ add %sum, 1, %sum; \
+51: addcc t1, %sum, %sum; \
+ stw t1, [%dst - off + 0x04]; \
+ srlx t1, 32, t1; \
+ bcc,pt %xcc, 52f; \
+ stw t1, [%dst - off - 0x00]; \
+ add %sum, 1, %sum; \
+52:
+
+cpc_start:
+cc_end_cruft:
+ andcc %g7, 8, %g0 ! IEU1 Group
+ be,pn %icc, 1f ! CTI
+ and %g7, 4, %g5 ! IEU0
+ ldxa [%src + 0x00] %asi, %g2 ! Load Group
+ add %dst, 8, %dst ! IEU0
+ add %src, 8, %src ! IEU1
+ addcc %g2, %sum, %sum ! IEU1 Group + 2 bubbles
+ stw %g2, [%dst - 0x04] ! Store
+ srlx %g2, 32, %g2 ! IEU0
+ bcc,pt %xcc, 1f ! CTI Group
+ stw %g2, [%dst - 0x08] ! Store
+ add %sum, 1, %sum ! IEU0
+1: brz,pt %g5, 1f ! CTI Group
+ clr %g2 ! IEU0
+ lduwa [%src + 0x00] %asi, %g2 ! Load
+ add %dst, 4, %dst ! IEU0 Group
+ add %src, 4, %src ! IEU1
+ stw %g2, [%dst - 0x04] ! Store Group + 2 bubbles
+ sllx %g2, 32, %g2 ! IEU0
+1: andcc %g7, 2, %g0 ! IEU1
+ be,pn %icc, 1f ! CTI Group
+ clr %o4 ! IEU1
+ lduha [%src + 0x00] %asi, %o4 ! Load
+ add %src, 2, %src ! IEU0 Group
+ add %dst, 2, %dst ! IEU1
+ sth %o4, [%dst - 0x2] ! Store Group + 2 bubbles
+ sll %o4, 16, %o4 ! IEU0
+1: andcc %g7, 1, %g0 ! IEU1
+ be,pn %icc, 1f ! CTI Group
+ clr %o5 ! IEU0
+ lduba [%src + 0x00] %asi, %o5 ! Load
+ stb %o5, [%dst + 0x00] ! Store Group + 2 bubbles
+ sll %o5, 8, %o5 ! IEU0
+1: or %g2, %o4, %o4 ! IEU1
+ or %o5, %o4, %o4 ! IEU0 Group
+ addcc %o4, %sum, %sum ! IEU1
+ bcc,pt %xcc, ccfold ! CTI
+ sethi %uhi(PAGE_OFFSET), %g4 ! IEU0 Group
+ b,pt %xcc, ccfold ! CTI
+ add %sum, 1, %sum ! IEU1
+
+cc_fixit:
+ cmp %len, 6 ! IEU1 Group
+ bl,a,pn %icc, ccte ! CTI
+ andcc %len, 0xf, %g7 ! IEU1 Group
+ andcc %src, 2, %g0 ! IEU1 Group
+ be,pn %icc, 1f ! CTI
+ andcc %src, 0x4, %g0 ! IEU1 Group
+ lduha [%src + 0x00] %asi, %g4 ! Load
+ sub %len, 2, %len ! IEU0
+ add %src, 2, %src ! IEU0 Group
+ add %dst, 2, %dst ! IEU1
+ sll %g4, 16, %g3 ! IEU0 Group + 1 bubble
+ addcc %g3, %sum, %sum ! IEU1
+ bcc,pt %xcc, 0f ! CTI
+ srl %sum, 16, %g3 ! IEU0 Group
+ add %g3, 1, %g3 ! IEU0 4 clocks (mispredict)
+0: andcc %src, 0x4, %g0 ! IEU1 Group
+ sth %g4, [%dst - 0x2] ! Store
+ sll %sum, 16, %sum ! IEU0
+ sll %g3, 16, %g3 ! IEU0 Group
+ srl %sum, 16, %sum ! IEU0 Group
+ or %g3, %sum, %sum ! IEU0 Group (regdep)
+1: be,pt %icc, ccmerge ! CTI
+ andcc %len, 0xf0, %g1 ! IEU1
+ lduwa [%src + 0x00] %asi, %g4 ! Load Group
+ sub %len, 4, %len ! IEU0
+ add %src, 4, %src ! IEU1
+ add %dst, 4, %dst ! IEU0 Group
+ addcc %g4, %sum, %sum ! IEU1 Group + 1 bubble
+ stw %g4, [%dst - 0x4] ! Store
+ bcc,pt %xcc, ccmerge ! CTI
+ andcc %len, 0xf0, %g1 ! IEU1 Group
+ b,pt %xcc, ccmerge ! CTI 4 clocks (mispredict)
+ add %sum, 1, %sum ! IEU0
+
+ .align 32
+ .globl csum_partial_copy_sparc64
+csum_partial_copy_sparc64: /* %o0=src, %o1=dest, %o2=len, %o3=sum */
+ xorcc %src, %dst, %o4 ! IEU1 Group
+ srl %sum, 0, %sum ! IEU0
+ andcc %o4, 3, %g0 ! IEU1 Group
+ srl %len, 0, %len ! IEU0
+ bne,pn %icc, ccslow ! CTI
+ andcc %src, 1, %g0 ! IEU1 Group
+ bne,pn %icc, ccslow ! CTI
+ cmp %len, 256 ! IEU1 Group
+ bgeu,pt %icc, csum_partial_copy_vis ! CTI
+ andcc %src, 7, %g0 ! IEU1 Group
+ bne,pn %icc, cc_fixit ! CTI
+ andcc %len, 0xf0, %g1 ! IEU1 Group
+ccmerge:be,pn %icc, ccte ! CTI
+ andcc %len, 0xf, %g7 ! IEU1 Group
+ sll %g1, 2, %o4 ! IEU0
+13: sethi %hi(12f), %o5 ! IEU0 Group
+ add %src, %g1, %src ! IEU1
+ sub %o5, %o4, %o5 ! IEU0 Group
+ jmpl %o5 + %lo(12f), %g0 ! CTI Group brk forced
+ add %dst, %g1, %dst ! IEU0 Group
+cctbl: CSUMCOPY_LASTCHUNK(0xe8,%g2,%g3)
+ CSUMCOPY_LASTCHUNK(0xd8,%g2,%g3)
+ CSUMCOPY_LASTCHUNK(0xc8,%g2,%g3)
+ CSUMCOPY_LASTCHUNK(0xb8,%g2,%g3)
+ CSUMCOPY_LASTCHUNK(0xa8,%g2,%g3)
+ CSUMCOPY_LASTCHUNK(0x98,%g2,%g3)
+ CSUMCOPY_LASTCHUNK(0x88,%g2,%g3)
+ CSUMCOPY_LASTCHUNK(0x78,%g2,%g3)
+ CSUMCOPY_LASTCHUNK(0x68,%g2,%g3)
+ CSUMCOPY_LASTCHUNK(0x58,%g2,%g3)
+ CSUMCOPY_LASTCHUNK(0x48,%g2,%g3)
+ CSUMCOPY_LASTCHUNK(0x38,%g2,%g3)
+ CSUMCOPY_LASTCHUNK(0x28,%g2,%g3)
+ CSUMCOPY_LASTCHUNK(0x18,%g2,%g3)
+ CSUMCOPY_LASTCHUNK(0x08,%g2,%g3)
+12:
+ andcc %len, 0xf, %g7 ! IEU1 Group
+ccte: bne,pn %icc, cc_end_cruft ! CTI
+ sethi %uhi(PAGE_OFFSET), %g4 ! IEU0
+ccfold: sllx %sum, 32, %o0 ! IEU0 Group
+ addcc %sum, %o0, %o0 ! IEU1 Group (regdep)
+ srlx %o0, 32, %o0 ! IEU0 Group (regdep)
+ bcs,a,pn %xcc, 1f ! CTI
+ add %o0, 1, %o0 ! IEU1 4 clocks (mispredict)
+1: retl ! CTI Group brk forced
+ sllx %g4, 32, %g4 ! IEU0 Group
+
+ccslow: mov 0, %g5
+ brlez,pn %len, 4f
+ andcc %src, 1, %o5
+ be,a,pt %icc, 1f
+ srl %len, 1, %g7
+ sub %len, 1, %len
+ lduba [%src] %asi, %g5
+ add %src, 1, %src
+ stb %g5, [%dst]
+ srl %len, 1, %g7
+ add %dst, 1, %dst
+1: brz,a,pn %g7, 3f
+ andcc %len, 1, %g0
+ andcc %src, 2, %g0
+ be,a,pt %icc, 1f
+ srl %g7, 1, %g7
+ lduha [%src] %asi, %o4
+ sub %len, 2, %len
+ srl %o4, 8, %g2
+ sub %g7, 1, %g7
+ stb %g2, [%dst]
+ add %o4, %g5, %g5
+ stb %o4, [%dst + 1]
+ add %src, 2, %src
+ srl %g7, 1, %g7
+ add %dst, 2, %dst
+1: brz,a,pn %g7, 2f
+ andcc %len, 2, %g0
+ lduwa [%src] %asi, %o4
+5: srl %o4, 24, %g2
+ srl %o4, 16, %g3
+ stb %g2, [%dst]
+ srl %o4, 8, %g2
+ stb %g3, [%dst + 1]
+ add %src, 4, %src
+ stb %g2, [%dst + 2]
+ addcc %o4, %g5, %g5
+ stb %o4, [%dst + 3]
+ addc %g5, %g0, %g5
+ add %dst, 4, %dst
+ subcc %g7, 1, %g7
+ bne,a,pt %icc, 5b
+ lduwa [%src] %asi, %o4
+ sll %g5, 16, %g2
+ srl %g5, 16, %g5
+ srl %g2, 16, %g2
+ andcc %len, 2, %g0
+ add %g2, %g5, %g5
+2: be,a,pt %icc, 3f
+ andcc %len, 1, %g0
+ lduha [%src] %asi, %o4
+ andcc %len, 1, %g0
+ srl %o4, 8, %g2
+ add %src, 2, %src
+ stb %g2, [%dst]
+ add %g5, %o4, %g5
+ stb %o4, [%dst + 1]
+ add %dst, 2, %dst
+3: be,a,pt %icc, 1f
+ sll %g5, 16, %o4
+ lduba [%src] %asi, %g2
+ sll %g2, 8, %o4
+ stb %g2, [%dst]
+ add %g5, %o4, %g5
+ sll %g5, 16, %o4
+1: addcc %o4, %g5, %g5
+ srl %g5, 16, %o4
+ addc %g0, %o4, %g5
+ brz,pt %o5, 4f
+ srl %g5, 8, %o4
+ and %g5, 0xff, %g2
+ and %o4, 0xff, %o4
+ sll %g2, 8, %g2
+ or %g2, %o4, %g5
+4: addcc %sum, %g5, %sum
+ addc %g0, %sum, %o0
+ retl
+ srl %o0, 0, %o0
+cpc_end:
+
+ .globl cpc_handler
+cpc_handler:
+ ldx [%sp + 0x7ff + 128], %g1
+ sub %g0, EFAULT, %g2
+ brnz,a,pt %g1, 1f
+ st %g2, [%g1]
+1: sethi %uhi(PAGE_OFFSET), %g4
+ retl
+ sllx %g4, 32, %g4
+
+ .section __ex_table
+ .align 4
+ .word cpc_start, 0, cpc_end, cpc_handler
+
generated by cgit v1.2.3 (git 2.39.1) at 2024-10-24 03:23:55 +0000