diff options
author | Nils Wallménius <nils@rockbox.org> | 2013-05-20 22:25:57 +0200 |
---|---|---|
committer | Nils Wallménius <nils@rockbox.org> | 2013-08-31 08:30:51 +0200 |
commit | 580b307fd791c0997a8831bc800bba87797bfb7e (patch) | |
tree | 807846056f06fd944a750ce41217a877910ebd59 /lib/rbcodec/codecs/libopus/celt/mathops.h | |
parent | 74761b70acd96cecc0d35450dd56a98ad9ee7d3d (diff) | |
download | rockbox-580b307fd791c0997a8831bc800bba87797bfb7e.tar.gz rockbox-580b307fd791c0997a8831bc800bba87797bfb7e.zip |
Sync opus codec to upstream git
Sync opus codec to upstream commit
02fed471a4568852d6618e041c4f2af0d7730ee2 (August 30 2013)
This brings in a lot of optimizations but also makes the diff
between our codec and the upstream much smaller as most of our
optimizations have been upstreamed or supeceded.
Speedups across the board for CELT mode files:
64kbps 128kbps
H300 9.82MHz 15.48MHz
c200 4.86MHz 9.63MHz
fuze v1 10.32MHz 15.92MHz
For the silk mode test file (16kbps) arm targets get a speedup
of about 2MHz while the H300 is 7.8MHz slower, likely because it's
now using the pseudostack more rather than the real stack which
is in iram. Patches to get around that are upcomming.
Change-Id: Ifecf963e461c51ac42e09dac1e91bc4bc3b12fa3
Diffstat (limited to 'lib/rbcodec/codecs/libopus/celt/mathops.h')
-rw-r--r-- | lib/rbcodec/codecs/libopus/celt/mathops.h | 66 |
1 files changed, 43 insertions, 23 deletions
diff --git a/lib/rbcodec/codecs/libopus/celt/mathops.h b/lib/rbcodec/codecs/libopus/celt/mathops.h index 4e97795606..44fa97c697 100644 --- a/lib/rbcodec/codecs/libopus/celt/mathops.h +++ b/lib/rbcodec/codecs/libopus/celt/mathops.h | |||
@@ -43,6 +43,41 @@ | |||
43 | 43 | ||
44 | unsigned isqrt32(opus_uint32 _val); | 44 | unsigned isqrt32(opus_uint32 _val); |
45 | 45 | ||
46 | #ifndef OVERRIDE_CELT_MAXABS16 | ||
47 | static inline opus_val32 celt_maxabs16(const opus_val16 *x, int len) | ||
48 | { | ||
49 | int i; | ||
50 | opus_val16 maxval = 0; | ||
51 | opus_val16 minval = 0; | ||
52 | for (i=0;i<len;i++) | ||
53 | { | ||
54 | maxval = MAX16(maxval, x[i]); | ||
55 | minval = MIN16(minval, x[i]); | ||
56 | } | ||
57 | return MAX32(EXTEND32(maxval),-EXTEND32(minval)); | ||
58 | } | ||
59 | #endif | ||
60 | |||
61 | #ifndef OVERRIDE_CELT_MAXABS32 | ||
62 | #ifdef FIXED_POINT | ||
63 | static inline opus_val32 celt_maxabs32(const opus_val32 *x, int len) | ||
64 | { | ||
65 | int i; | ||
66 | opus_val32 maxval = 0; | ||
67 | opus_val32 minval = 0; | ||
68 | for (i=0;i<len;i++) | ||
69 | { | ||
70 | maxval = MAX32(maxval, x[i]); | ||
71 | minval = MIN32(minval, x[i]); | ||
72 | } | ||
73 | return MAX32(maxval, -minval); | ||
74 | } | ||
75 | #else | ||
76 | #define celt_maxabs32(x,len) celt_maxabs16(x,len) | ||
77 | #endif | ||
78 | #endif | ||
79 | |||
80 | |||
46 | #ifndef FIXED_POINT | 81 | #ifndef FIXED_POINT |
47 | 82 | ||
48 | #define PI 3.141592653f | 83 | #define PI 3.141592653f |
@@ -117,27 +152,6 @@ static inline opus_int16 celt_ilog2(opus_int32 x) | |||
117 | } | 152 | } |
118 | #endif | 153 | #endif |
119 | 154 | ||
120 | #ifndef OVERRIDE_CELT_MAXABS16 | ||
121 | static inline opus_val16 celt_maxabs16(opus_val16 *x, int len) | ||
122 | { | ||
123 | int i; | ||
124 | opus_val16 maxval = 0; | ||
125 | for (i=0;i<len;i++) | ||
126 | maxval = MAX16(maxval, ABS16(x[i])); | ||
127 | return maxval; | ||
128 | } | ||
129 | #endif | ||
130 | |||
131 | #ifndef OVERRIDE_CELT_MAXABS32 | ||
132 | static inline opus_val32 celt_maxabs32(opus_val32 *x, int len) | ||
133 | { | ||
134 | int i; | ||
135 | opus_val32 maxval = 0; | ||
136 | for (i=0;i<len;i++) | ||
137 | maxval = MAX32(maxval, ABS32(x[i])); | ||
138 | return maxval; | ||
139 | } | ||
140 | #endif | ||
141 | 155 | ||
142 | /** Integer log in base2. Defined for zero, but not for negative numbers */ | 156 | /** Integer log in base2. Defined for zero, but not for negative numbers */ |
143 | static inline opus_int16 celt_zlog2(opus_val32 x) | 157 | static inline opus_int16 celt_zlog2(opus_val32 x) |
@@ -176,6 +190,13 @@ static inline opus_val16 celt_log2(opus_val32 x) | |||
176 | #define D1 22804 | 190 | #define D1 22804 |
177 | #define D2 14819 | 191 | #define D2 14819 |
178 | #define D3 10204 | 192 | #define D3 10204 |
193 | |||
194 | static inline opus_val32 celt_exp2_frac(opus_val16 x) | ||
195 | { | ||
196 | opus_val16 frac; | ||
197 | frac = SHL16(x, 4); | ||
198 | return ADD16(D0, MULT16_16_Q15(frac, ADD16(D1, MULT16_16_Q15(frac, ADD16(D2 , MULT16_16_Q15(D3,frac)))))); | ||
199 | } | ||
179 | /** Base-2 exponential approximation (2^x). (Q10 input, Q16 output) */ | 200 | /** Base-2 exponential approximation (2^x). (Q10 input, Q16 output) */ |
180 | static inline opus_val32 celt_exp2(opus_val16 x) | 201 | static inline opus_val32 celt_exp2(opus_val16 x) |
181 | { | 202 | { |
@@ -186,8 +207,7 @@ static inline opus_val32 celt_exp2(opus_val16 x) | |||
186 | return 0x7f000000; | 207 | return 0x7f000000; |
187 | else if (integer < -15) | 208 | else if (integer < -15) |
188 | return 0; | 209 | return 0; |
189 | frac = SHL16(x-SHL16(integer,10),4); | 210 | frac = celt_exp2_frac(x-SHL16(integer,10)); |
190 | frac = ADD16(D0, MULT16_16_Q15(frac, ADD16(D1, MULT16_16_Q15(frac, ADD16(D2 , MULT16_16_Q15(D3,frac)))))); | ||
191 | return VSHR32(EXTEND32(frac), -integer-2); | 211 | return VSHR32(EXTEND32(frac), -integer-2); |
192 | } | 212 | } |
193 | 213 | ||