Merge branch 'upstream'
[linux-2.6] / arch / sparc / math-emu / sfp-util.h
1 #include <linux/kernel.h>
2 #include <linux/sched.h>
3 #include <linux/types.h>
4 #include <asm/byteorder.h>
5
6 #define add_ssaaaa(sh, sl, ah, al, bh, bl)                              \
7   __asm__ ("addcc %r4,%5,%1\n\t"                                                \
8            "addx %r2,%3,%0\n"                                           \
9            : "=r" ((USItype)(sh)),                                      \
10              "=&r" ((USItype)(sl))                                      \
11            : "%rJ" ((USItype)(ah)),                                     \
12              "rI" ((USItype)(bh)),                                      \
13              "%rJ" ((USItype)(al)),                                     \
14              "rI" ((USItype)(bl))                                       \
15            : "cc")
16 #define sub_ddmmss(sh, sl, ah, al, bh, bl)                              \
17   __asm__ ("subcc %r4,%5,%1\n\t"                                                \
18            "subx %r2,%3,%0\n"                                           \
19            : "=r" ((USItype)(sh)),                                      \
20              "=&r" ((USItype)(sl))                                      \
21            : "rJ" ((USItype)(ah)),                                      \
22              "rI" ((USItype)(bh)),                                      \
23              "rJ" ((USItype)(al)),                                      \
24              "rI" ((USItype)(bl))                                       \
25            : "cc")
26
27 #define umul_ppmm(w1, w0, u, v) \
28   __asm__ ("! Inlined umul_ppmm\n\t"                                    \
29         "wr     %%g0,%2,%%y     ! SPARC has 0-3 delay insn after a wr\n\t" \
30         "sra    %3,31,%%g2      ! Don't move this insn\n\t"             \
31         "and    %2,%%g2,%%g2    ! Don't move this insn\n\t"             \
32         "andcc  %%g0,0,%%g1     ! Don't move this insn\n\t"             \
33         "mulscc %%g1,%3,%%g1\n\t"                                       \
34         "mulscc %%g1,%3,%%g1\n\t"                                       \
35         "mulscc %%g1,%3,%%g1\n\t"                                       \
36         "mulscc %%g1,%3,%%g1\n\t"                                       \
37         "mulscc %%g1,%3,%%g1\n\t"                                       \
38         "mulscc %%g1,%3,%%g1\n\t"                                       \
39         "mulscc %%g1,%3,%%g1\n\t"                                       \
40         "mulscc %%g1,%3,%%g1\n\t"                                       \
41         "mulscc %%g1,%3,%%g1\n\t"                                       \
42         "mulscc %%g1,%3,%%g1\n\t"                                       \
43         "mulscc %%g1,%3,%%g1\n\t"                                       \
44         "mulscc %%g1,%3,%%g1\n\t"                                       \
45         "mulscc %%g1,%3,%%g1\n\t"                                       \
46         "mulscc %%g1,%3,%%g1\n\t"                                       \
47         "mulscc %%g1,%3,%%g1\n\t"                                       \
48         "mulscc %%g1,%3,%%g1\n\t"                                       \
49         "mulscc %%g1,%3,%%g1\n\t"                                       \
50         "mulscc %%g1,%3,%%g1\n\t"                                       \
51         "mulscc %%g1,%3,%%g1\n\t"                                       \
52         "mulscc %%g1,%3,%%g1\n\t"                                       \
53         "mulscc %%g1,%3,%%g1\n\t"                                       \
54         "mulscc %%g1,%3,%%g1\n\t"                                       \
55         "mulscc %%g1,%3,%%g1\n\t"                                       \
56         "mulscc %%g1,%3,%%g1\n\t"                                       \
57         "mulscc %%g1,%3,%%g1\n\t"                                       \
58         "mulscc %%g1,%3,%%g1\n\t"                                       \
59         "mulscc %%g1,%3,%%g1\n\t"                                       \
60         "mulscc %%g1,%3,%%g1\n\t"                                       \
61         "mulscc %%g1,%3,%%g1\n\t"                                       \
62         "mulscc %%g1,%3,%%g1\n\t"                                       \
63         "mulscc %%g1,%3,%%g1\n\t"                                       \
64         "mulscc %%g1,%3,%%g1\n\t"                                       \
65         "mulscc %%g1,0,%%g1\n\t"                                        \
66         "add    %%g1,%%g2,%0\n\t"                                       \
67         "rd     %%y,%1\n"                                               \
68            : "=r" ((USItype)(w1)),                                      \
69              "=r" ((USItype)(w0))                                       \
70            : "%rI" ((USItype)(u)),                                      \
71              "r" ((USItype)(v))                                         \
72            : "%g1", "%g2", "cc")
73
74 /* It's quite necessary to add this much assembler for the sparc.
75    The default udiv_qrnnd (in C) is more than 10 times slower!  */
76 #define udiv_qrnnd(q, r, n1, n0, d) \
77   __asm__ ("! Inlined udiv_qrnnd\n\t"                                   \
78            "mov 32,%%g1\n\t"                                            \
79            "subcc       %1,%2,%%g0\n\t"                                 \
80            "1:  bcs     5f\n\t"                                         \
81            "addxcc %0,%0,%0     ! shift n1n0 and a q-bit in lsb\n\t"    \
82            "sub %1,%2,%1        ! this kills msb of n\n\t"              \
83            "addx        %1,%1,%1        ! so this can't give carry\n\t" \
84            "subcc       %%g1,1,%%g1\n\t"                                \
85            "2:  bne     1b\n\t"                                         \
86            "subcc       %1,%2,%%g0\n\t"                                 \
87            "bcs 3f\n\t"                                                 \
88            "addxcc %0,%0,%0     ! shift n1n0 and a q-bit in lsb\n\t"    \
89            "b           3f\n\t"                                         \
90            "sub %1,%2,%1        ! this kills msb of n\n\t"              \
91            "4:  sub     %1,%2,%1\n\t"                                   \
92            "5:  addxcc  %1,%1,%1\n\t"                                   \
93            "bcc 2b\n\t"                                                 \
94            "subcc       %%g1,1,%%g1\n\t"                                \
95            "! Got carry from n.  Subtract next step to cancel this carry.\n\t" \
96            "bne 4b\n\t"                                                 \
97            "addcc       %0,%0,%0        ! shift n1n0 and a 0-bit in lsb\n\t" \
98            "sub %1,%2,%1\n\t"                                           \
99            "3:  xnor    %0,0,%0\n\t"                                    \
100            "! End of inline udiv_qrnnd\n"                               \
101            : "=&r" ((USItype)(q)),                                      \
102              "=&r" ((USItype)(r))                                       \
103            : "r" ((USItype)(d)),                                        \
104              "1" ((USItype)(n1)),                                       \
105              "0" ((USItype)(n0)) : "%g1", "cc")
106 #define UDIV_NEEDS_NORMALIZATION 0
107
108 #define abort()                                                         \
109         return 0
110
111 #ifdef __BIG_ENDIAN
112 #define __BYTE_ORDER __BIG_ENDIAN
113 #else
114 #define __BYTE_ORDER __LITTLE_ENDIAN
115 #endif