1*a58d3d2aSXin Li /* Copyright (C) 2013 Xiph.Org Foundation and contributors */
2*a58d3d2aSXin Li /*
3*a58d3d2aSXin Li Redistribution and use in source and binary forms, with or without
4*a58d3d2aSXin Li modification, are permitted provided that the following conditions
5*a58d3d2aSXin Li are met:
6*a58d3d2aSXin Li
7*a58d3d2aSXin Li - Redistributions of source code must retain the above copyright
8*a58d3d2aSXin Li notice, this list of conditions and the following disclaimer.
9*a58d3d2aSXin Li
10*a58d3d2aSXin Li - Redistributions in binary form must reproduce the above copyright
11*a58d3d2aSXin Li notice, this list of conditions and the following disclaimer in the
12*a58d3d2aSXin Li documentation and/or other materials provided with the distribution.
13*a58d3d2aSXin Li
14*a58d3d2aSXin Li THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
15*a58d3d2aSXin Li ``AS IS'' AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
16*a58d3d2aSXin Li LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR
17*a58d3d2aSXin Li A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER
18*a58d3d2aSXin Li OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL,
19*a58d3d2aSXin Li EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO,
20*a58d3d2aSXin Li PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR
21*a58d3d2aSXin Li PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF
22*a58d3d2aSXin Li LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING
23*a58d3d2aSXin Li NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS
24*a58d3d2aSXin Li SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
25*a58d3d2aSXin Li */
26*a58d3d2aSXin Li
27*a58d3d2aSXin Li #ifndef FIXED_ARMv4_H
28*a58d3d2aSXin Li #define FIXED_ARMv4_H
29*a58d3d2aSXin Li
30*a58d3d2aSXin Li /** 16x32 multiplication, followed by a 16-bit shift right. Results fits in 32 bits */
31*a58d3d2aSXin Li #undef MULT16_32_Q16
MULT16_32_Q16_armv4(opus_val16 a,opus_val32 b)32*a58d3d2aSXin Li static OPUS_INLINE opus_val32 MULT16_32_Q16_armv4(opus_val16 a, opus_val32 b)
33*a58d3d2aSXin Li {
34*a58d3d2aSXin Li unsigned rd_lo;
35*a58d3d2aSXin Li int rd_hi;
36*a58d3d2aSXin Li __asm__(
37*a58d3d2aSXin Li "#MULT16_32_Q16\n\t"
38*a58d3d2aSXin Li "smull %0, %1, %2, %3\n\t"
39*a58d3d2aSXin Li : "=&r"(rd_lo), "=&r"(rd_hi)
40*a58d3d2aSXin Li : "%r"(b),"r"(SHL32(a,16))
41*a58d3d2aSXin Li );
42*a58d3d2aSXin Li return rd_hi;
43*a58d3d2aSXin Li }
44*a58d3d2aSXin Li #define MULT16_32_Q16(a, b) (MULT16_32_Q16_armv4(a, b))
45*a58d3d2aSXin Li
46*a58d3d2aSXin Li
47*a58d3d2aSXin Li /** 16x32 multiplication, followed by a 15-bit shift right. Results fits in 32 bits */
48*a58d3d2aSXin Li #undef MULT16_32_Q15
MULT16_32_Q15_armv4(opus_val16 a,opus_val32 b)49*a58d3d2aSXin Li static OPUS_INLINE opus_val32 MULT16_32_Q15_armv4(opus_val16 a, opus_val32 b)
50*a58d3d2aSXin Li {
51*a58d3d2aSXin Li unsigned rd_lo;
52*a58d3d2aSXin Li int rd_hi;
53*a58d3d2aSXin Li __asm__(
54*a58d3d2aSXin Li "#MULT16_32_Q15\n\t"
55*a58d3d2aSXin Li "smull %0, %1, %2, %3\n\t"
56*a58d3d2aSXin Li : "=&r"(rd_lo), "=&r"(rd_hi)
57*a58d3d2aSXin Li : "%r"(b), "r"(SHL32(a,16))
58*a58d3d2aSXin Li );
59*a58d3d2aSXin Li /*We intentionally don't OR in the high bit of rd_lo for speed.*/
60*a58d3d2aSXin Li return SHL32(rd_hi,1);
61*a58d3d2aSXin Li }
62*a58d3d2aSXin Li #define MULT16_32_Q15(a, b) (MULT16_32_Q15_armv4(a, b))
63*a58d3d2aSXin Li
64*a58d3d2aSXin Li
65*a58d3d2aSXin Li /** 16x32 multiply, followed by a 15-bit shift right and 32-bit add.
66*a58d3d2aSXin Li b must fit in 31 bits.
67*a58d3d2aSXin Li Result fits in 32 bits. */
68*a58d3d2aSXin Li #undef MAC16_32_Q15
69*a58d3d2aSXin Li #define MAC16_32_Q15(c, a, b) ADD32(c, MULT16_32_Q15(a, b))
70*a58d3d2aSXin Li
71*a58d3d2aSXin Li /** 16x32 multiply, followed by a 16-bit shift right and 32-bit add.
72*a58d3d2aSXin Li Result fits in 32 bits. */
73*a58d3d2aSXin Li #undef MAC16_32_Q16
74*a58d3d2aSXin Li #define MAC16_32_Q16(c, a, b) ADD32(c, MULT16_32_Q16(a, b))
75*a58d3d2aSXin Li
76*a58d3d2aSXin Li /** 32x32 multiplication, followed by a 31-bit shift right. Results fits in 32 bits */
77*a58d3d2aSXin Li #undef MULT32_32_Q31
78*a58d3d2aSXin Li #define MULT32_32_Q31(a,b) (opus_val32)((((opus_int64)(a)) * ((opus_int64)(b)))>>31)
79*a58d3d2aSXin Li
80*a58d3d2aSXin Li #endif
81