This patch set is replacement of patch "libavutil: Make changes in softfloat needed for fixed point aac decoder." The original patch is split so it is easier to do a review. All the changes that are not absolutely needed for implementation of fixed point aac decoder are removed. -Nedeljko
Exponent usage and calculation in softfloat adjusted to the format used in implementation of fixed point aac decoder. Signed-off-by: Nedeljko Babic <nedeljko.babic@imgtec.com> --- libavutil/softfloat.h | 8 ++++---- 1 file changed, 4 insertions(+), 4 deletions(-) diff --git a/libavutil/softfloat.h b/libavutil/softfloat.h index 2e85765..5f4ac26 100644 --- a/libavutil/softfloat.h +++ b/libavutil/softfloat.h @@ -83,7 +83,7 @@ static inline av_const SoftFloat av_mul_sf(SoftFloat a, SoftFloat b){ a.exp += b.exp; av_assert2((int32_t)((a.mant * (int64_t)b.mant) >> ONE_BITS) == (a.mant * (int64_t)b.mant) >> ONE_BITS); a.mant = (a.mant * (int64_t)b.mant) >> ONE_BITS; - return av_normalize1_sf(a); + return av_normalize1_sf((SoftFloat){a.mant, a.exp - 1}); } /** @@ -91,7 +91,7 @@ static inline av_const SoftFloat av_mul_sf(SoftFloat a, SoftFloat b){ * @return Will not be more denormalized than a. */ static av_const SoftFloat av_div_sf(SoftFloat a, SoftFloat b){ - a.exp -= b.exp+1; + a.exp -= b.exp; a.mant = ((int64_t)a.mant<<(ONE_BITS+1)) / b.mant; return av_normalize1_sf(a); } @@ -121,14 +121,14 @@ static inline av_const SoftFloat av_sub_sf(SoftFloat a, SoftFloat b){ * @returns a SoftFloat with value v * 2^frac_bits */ static inline av_const SoftFloat av_int2sf(int v, int frac_bits){ - return av_normalize_sf((SoftFloat){v, ONE_BITS-frac_bits}); + return av_normalize_sf((SoftFloat){v, ONE_BITS + 1 - frac_bits}); } /** * Rounding is to -inf. */ static inline av_const int av_sf2int(SoftFloat v, int frac_bits){ - v.exp += frac_bits - ONE_BITS; + v.exp += frac_bits - (ONE_BITS + 1); if(v.exp >= 0) return v.mant << v.exp ; else return v.mant >>(-v.exp); } -- 1.8.2.1
On Thu, Apr 30, 2015 at 01:51:35PM +0200, Nedeljko Babic wrote:
Exponent usage and calculation in softfloat adjusted to the format used in implementation of fixed point aac decoder.
Signed-off-by: Nedeljko Babic <nedeljko.babic@imgtec.com> --- libavutil/softfloat.h | 8 ++++---- 1 file changed, 4 insertions(+), 4 deletions(-)
applied thanks [...] -- Michael GnuPG fingerprint: 9FF2128B147EF6730BADF133611EC787040B0FAB Why not whip the teacher when the pupil misbehaves? -- Diogenes of Sinope
I submitted 2 more patches with this one, but for some reason they are all over the list (probably because I used the same Subject message as before)... Should I resend them all again so they are together on the list? Thanks, -Nedeljko ________________________________________ Od: ffmpeg-devel-bounces@ffmpeg.org [ffmpeg-devel-bounces@ffmpeg.org] u ime korisnika Michael Niedermayer [michaelni@gmx.at] Poslato: 30. april 2015 14:31 Za: FFmpeg development discussions and patches Tema: Re: [FFmpeg-devel] [PATCH 1/3] libavutil/softfloat: exponent adjusted for aac fixed point dec On Thu, Apr 30, 2015 at 01:51:35PM +0200, Nedeljko Babic wrote:
Exponent usage and calculation in softfloat adjusted to the format used in implementation of fixed point aac decoder.
Signed-off-by: Nedeljko Babic <nedeljko.babic@imgtec.com> --- libavutil/softfloat.h | 8 ++++---- 1 file changed, 4 insertions(+), 4 deletions(-)
applied thanks [...] -- Michael GnuPG fingerprint: 9FF2128B147EF6730BADF133611EC787040B0FAB Why not whip the teacher when the pupil misbehaves? -- Diogenes of Sinope
On Thu, Apr 30, 2015 at 12:50:46PM +0000, Nedeljko Babic wrote:
I submitted 2 more patches with this one, but for some reason they are all over the list (probably because I used the same Subject message as before)...
Should I resend them all again so they are together on the list?
i think ive found and applied all 3 [...] -- Michael GnuPG fingerprint: 9FF2128B147EF6730BADF133611EC787040B0FAB Old school: Use the lowest level language in which you can solve the problem conveniently. New school: Use the highest level language in which the latest supercomputer can solve the problem without the user falling asleep waiting.
This will normalize sums for which mantissa is smaller than the lower boundary (needed for implementation of fixed point aac decoder). Signed-off-by: Nedeljko Babic <nedeljko.babic@imgtec.com> --- libavutil/softfloat.h | 4 ++-- 1 file changed, 2 insertions(+), 2 deletions(-) diff --git a/libavutil/softfloat.h b/libavutil/softfloat.h index 5f4ac26..42b3c3e 100644 --- a/libavutil/softfloat.h +++ b/libavutil/softfloat.h @@ -105,8 +105,8 @@ static inline av_const int av_cmp_sf(SoftFloat a, SoftFloat b){ static inline av_const SoftFloat av_add_sf(SoftFloat a, SoftFloat b){ int t= a.exp - b.exp; if (t <-31) return b; - else if (t < 0) return av_normalize1_sf((SoftFloat){b.mant + (a.mant >> (-t)), b.exp}); - else if (t < 32) return av_normalize1_sf((SoftFloat){a.mant + (b.mant >> t ), a.exp}); + else if (t < 0) return av_normalize_sf(av_normalize1_sf((SoftFloat){ b.mant + (a.mant >> (-t)), b.exp})); + else if (t < 32) return av_normalize_sf(av_normalize1_sf((SoftFloat){ a.mant + (b.mant >> t ), a.exp})); else return a; } -- 1.8.2.1
On Thu, Apr 30, 2015 at 01:51:36PM +0200, Nedeljko Babic wrote:
This will normalize sums for which mantissa is smaller than the lower boundary (needed for implementation of fixed point aac decoder).
Signed-off-by: Nedeljko Babic <nedeljko.babic@imgtec.com> --- libavutil/softfloat.h | 4 ++-- 1 file changed, 2 insertions(+), 2 deletions(-)
applied please add a testcase for this to the self tests thanks [...] -- Michael GnuPG fingerprint: 9FF2128B147EF6730BADF133611EC787040B0FAB Dictatorship naturally arises out of democracy, and the most aggravated form of tyranny and slavery out of the most extreme liberty. -- Plato
ok thanks, -Nedeljko ________________________________________ Od: ffmpeg-devel-bounces@ffmpeg.org [ffmpeg-devel-bounces@ffmpeg.org] u ime korisnika Michael Niedermayer [michaelni@gmx.at] Poslato: 30. april 2015 14:52 Za: FFmpeg development discussions and patches Tema: Re: [FFmpeg-devel] [PATCH 2/3] libavutil/softfloat: Added av_normalize_sf in av_add_sf On Thu, Apr 30, 2015 at 01:51:36PM +0200, Nedeljko Babic wrote:
This will normalize sums for which mantissa is smaller than the lower boundary (needed for implementation of fixed point aac decoder).
Signed-off-by: Nedeljko Babic <nedeljko.babic@imgtec.com> --- libavutil/softfloat.h | 4 ++-- 1 file changed, 2 insertions(+), 2 deletions(-)
applied please add a testcase for this to the self tests thanks [...] -- Michael GnuPG fingerprint: 9FF2128B147EF6730BADF133611EC787040B0FAB Dictatorship naturally arises out of democracy, and the most aggravated form of tyranny and slavery out of the most extreme liberty. -- Plato
Functions av_gt_sf, av_sqrt_sf and av_sincos_sf added to softfloat Signed-off-by: Nedeljko Babic <nedeljko.babic@imgtec.com> --- libavutil/softfloat.h | 93 +++++++++++++++- libavutil/softfloat_tables.h | 260 +++++++++++++++++++++++++++++++++++++++++++ 2 files changed, 352 insertions(+), 1 deletion(-) create mode 100644 libavutil/softfloat_tables.h diff --git a/libavutil/softfloat.h b/libavutil/softfloat.h index 42b3c3e..d6cfc3c 100644 --- a/libavutil/softfloat.h +++ b/libavutil/softfloat.h @@ -25,6 +25,7 @@ #include "common.h" #include "avassert.h" +#include "softfloat_tables.h" #define MIN_EXP -126 #define MAX_EXP 126 @@ -102,6 +103,13 @@ static inline av_const int av_cmp_sf(SoftFloat a, SoftFloat b){ else return a.mant - (b.mant >> t); } +static inline av_const int av_gt_sf(SoftFloat a, SoftFloat b) +{ + int t= a.exp - b.exp; + if(t<0) return (a.mant >> (-t)) > b.mant ; + else return a.mant > (b.mant >> t); +} + static inline av_const SoftFloat av_add_sf(SoftFloat a, SoftFloat b){ int t= a.exp - b.exp; if (t <-31) return b; @@ -114,7 +122,7 @@ static inline av_const SoftFloat av_sub_sf(SoftFloat a, SoftFloat b){ return av_add_sf(a, (SoftFloat){ -b.mant, b.exp}); } -//FIXME sqrt, log, exp, pow, sin, cos +//FIXME log, exp, pow /** * Converts a mantisse and exponent to a SoftFloat @@ -133,4 +141,87 @@ static inline av_const int av_sf2int(SoftFloat v, int frac_bits){ else return v.mant >>(-v.exp); } +/** + * Rounding-to-nearest used. + */ +static av_always_inline SoftFloat av_sqrt_sf(SoftFloat val) +{ + int tabIndex, rem; + + if (val.mant == 0) + val.exp = 0; + else + { + tabIndex = (val.mant - 0x20000000) >> 20; + + rem = val.mant & 0xFFFFF; + val.mant = (int)(((int64_t)av_sqrttbl_sf[tabIndex] * (0x100000 - rem) + + (int64_t)av_sqrttbl_sf[tabIndex + 1] * rem + + 0x80000) >> 20); + val.mant = (int)(((int64_t)av_sqr_exp_multbl_sf[val.exp & 1] * val.mant + + 0x10000000) >> 29); + + if (val.mant < 0x40000000) + val.exp -= 2; + else + val.mant >>= 1; + + val.exp = (val.exp >> 1) + 1; + } + + return val; +} + +/** + * Rounding-to-nearest used. + */ +static av_always_inline void av_sincos_sf(int a, int *s, int *c) +{ + int idx, sign; + int sv, cv; + int st, ct; + + idx = a >> 26; + sign = (idx << 27) >> 31; + cv = av_costbl_1_sf[idx & 0xf]; + cv = (cv ^ sign) - sign; + + idx -= 8; + sign = (idx << 27) >> 31; + sv = av_costbl_1_sf[idx & 0xf]; + sv = (sv ^ sign) - sign; + + idx = a >> 21; + ct = av_costbl_2_sf[idx & 0x1f]; + st = av_sintbl_2_sf[idx & 0x1f]; + + idx = (int)(((int64_t)cv * ct - (int64_t)sv * st + 0x20000000) >> 30); + + sv = (int)(((int64_t)cv * st + (int64_t)sv * ct + 0x20000000) >> 30); + + cv = idx; + + idx = a >> 16; + ct = av_costbl_3_sf[idx & 0x1f]; + st = av_sintbl_3_sf[idx & 0x1f]; + + idx = (int)(((int64_t)cv * ct - (int64_t)sv * st + 0x20000000) >> 30); + + sv = (int)(((int64_t)cv * st + (int64_t)sv * ct + 0x20000000) >> 30); + cv = idx; + + idx = a >> 11; + + ct = (int)(((int64_t)av_costbl_4_sf[idx & 0x1f] * (0x800 - (a & 0x7ff)) + + (int64_t)av_costbl_4_sf[(idx & 0x1f)+1]*(a & 0x7ff) + + 0x400) >> 11); + st = (int)(((int64_t)av_sintbl_4_sf[idx & 0x1f] * (0x800 - (a & 0x7ff)) + + (int64_t)av_sintbl_4_sf[(idx & 0x1f) + 1] * (a & 0x7ff) + + 0x400) >> 11); + + *c = (int)(((int64_t)cv * ct + (int64_t)sv * st + 0x20000000) >> 30); + + *s = (int)(((int64_t)cv * st + (int64_t)sv * ct + 0x20000000) >> 30); +} + #endif /* AVUTIL_SOFTFLOAT_H */ diff --git a/libavutil/softfloat_tables.h b/libavutil/softfloat_tables.h new file mode 100644 index 0000000..9d3c808 --- /dev/null +++ b/libavutil/softfloat_tables.h @@ -0,0 +1,260 @@ +/* + * Copyright (c) 2012 + * MIPS Technologies, Inc., California. + * + * Redistribution and use in source and binary forms, with or without + * modification, are permitted provided that the following conditions + * are met: + * 1. Redistributions of source code must retain the above copyright + * notice, this list of conditions and the following disclaimer. + * 2. Redistributions in binary form must reproduce the above copyright + * notice, this list of conditions and the following disclaimer in the + * documentation and/or other materials provided with the distribution. + * 3. Neither the name of the MIPS Technologies, Inc., nor the names of is + * contributors may be used to endorse or promote products derived from + * this software without specific prior written permission. + * + * THIS SOFTWARE IS PROVIDED BY THE MIPS TECHNOLOGIES, INC. ``AS IS'' AND + * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE + * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE + * ARE DISCLAIMED. IN NO EVENT SHALL THE MIPS TECHNOLOGIES, INC. BE LIABLE + * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL + * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS + * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) + * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT + * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY + * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF + * SUCH DAMAGE. + * + * Author: Stanislav Ocovaj (stanislav.ocovaj imgtec com) + * + * This file is part of FFmpeg. + * + * FFmpeg is free software; you can redistribute it and/or + * modify it under the terms of the GNU Lesser General Public + * License as published by the Free Software Foundation; either + * version 2.1 of the License, or (at your option) any later version. + * + * FFmpeg is distributed in the hope that it will be useful, + * but WITHOUT ANY WARRANTY; without even the implied warranty of + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU + * Lesser General Public License for more details. + * + * You should have received a copy of the GNU Lesser General Public + * License along with FFmpeg; if not, write to the Free Software + * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA + */ +#ifndef AVUTIL_SOFTFLOAT_TABLES_H +#define AVUTIL_SOFTFLOAT_TABLES_H + +static const int32_t av_sqrttbl_sf[512+1] = { /* sqrt(x), 0.5<=x<1 */ + 0x2d413ccd,0x2d4c8bb3,0x2d57d7c6,0x2d63210a, + 0x2d6e677f,0x2d79ab2a,0x2d84ec0b,0x2d902a23, + 0x2d9b6578,0x2da69e08,0x2db1d3d6,0x2dbd06e6, + 0x2dc83738,0x2dd364ce,0x2dde8fac,0x2de9b7d2, + 0x2df4dd43,0x2e000000,0x2e0b200c,0x2e163d68, + 0x2e215816,0x2e2c701a,0x2e378573,0x2e429824, + 0x2e4da830,0x2e58b598,0x2e63c05d,0x2e6ec883, + 0x2e79ce0a,0x2e84d0f5,0x2e8fd144,0x2e9acefb, + 0x2ea5ca1b,0x2eb0c2a7,0x2ebbb89e,0x2ec6ac04, + 0x2ed19cda,0x2edc8b23,0x2ee776df,0x2ef26012, + 0x2efd46bb,0x2f082add,0x2f130c7b,0x2f1deb95, + 0x2f28c82e,0x2f33a246,0x2f3e79e1,0x2f494eff, + 0x2f5421a3,0x2f5ef1ce,0x2f69bf81,0x2f748abe, + 0x2f7f5388,0x2f8a19e0,0x2f94ddc7,0x2f9f9f3e, + 0x2faa5e48,0x2fb51ae8,0x2fbfd51c,0x2fca8ce9, + 0x2fd5424e,0x2fdff54e,0x2feaa5eb,0x2ff55426, + 0x30000000,0x300aa97b,0x3015509a,0x301ff55c, + 0x302a97c5,0x303537d5,0x303fd58e,0x304a70f2, + 0x30550a01,0x305fa0be,0x306a352a,0x3074c747, + 0x307f5716,0x3089e499,0x30946fd2,0x309ef8c0, + 0x30a97f67,0x30b403c7,0x30be85e2,0x30c905bb, + 0x30d38351,0x30ddfea6,0x30e877bc,0x30f2ee96, + 0x30fd6332,0x3107d594,0x311245bc,0x311cb3ad, + 0x31271f67,0x313188ec,0x313bf03d,0x3146555c, + 0x3150b84a,0x315b1909,0x31657798,0x316fd3fc, + 0x317a2e34,0x31848642,0x318edc28,0x31992fe5, + 0x31a3817d,0x31add0f0,0x31b81e40,0x31c2696e, + 0x31ccb27b,0x31d6f969,0x31e13e38,0x31eb80eb, + 0x31f5c182,0x32000000,0x320a3c65,0x321476b1, + 0x321eaee8,0x3228e50a,0x32331917,0x323d4b13, + 0x32477afc,0x3251a8d6,0x325bd4a2,0x3265fe5f, + 0x32702611,0x327a4bb8,0x32846f55,0x328e90e9, + 0x3298b076,0x32a2cdfd,0x32ace97e,0x32b702fd, + 0x32c11a79,0x32cb2ff3,0x32d5436d,0x32df54e9, + 0x32e96466,0x32f371e8,0x32fd7d6d,0x330786f9, + 0x33118e8c,0x331b9426,0x332597cb,0x332f9979, + 0x33399933,0x334396fa,0x334d92cf,0x33578cb2, + 0x336184a6,0x336b7aab,0x33756ec3,0x337f60ed, + 0x3389512d,0x33933f83,0x339d2bef,0x33a71672, + 0x33b0ff10,0x33bae5c7,0x33c4ca99,0x33cead88, + 0x33d88e95,0x33e26dbf,0x33ec4b09,0x33f62673, + 0x34000000,0x3409d7af,0x3413ad82,0x341d817a, + 0x34275397,0x343123db,0x343af248,0x3444bedd, + 0x344e899d,0x34585288,0x3462199f,0x346bdee3, + 0x3475a254,0x347f63f5,0x348923c6,0x3492e1c9, + 0x349c9dfe,0x34a65865,0x34b01101,0x34b9c7d2, + 0x34c37cda,0x34cd3018,0x34d6e18f,0x34e0913f, + 0x34ea3f29,0x34f3eb4d,0x34fd95ae,0x35073e4c, + 0x3510e528,0x351a8a43,0x35242d9d,0x352dcf39, + 0x35376f16,0x35410d36,0x354aa99a,0x35544442, + 0x355ddd2f,0x35677463,0x357109df,0x357a9da2, + 0x35842fb0,0x358dc007,0x35974ea9,0x35a0db98, + 0x35aa66d3,0x35b3f05c,0x35bd7833,0x35c6fe5a, + 0x35d082d3,0x35da059c,0x35e386b7,0x35ed0626, + 0x35f683e8,0x36000000,0x36097a6e,0x3612f331, + 0x361c6a4d,0x3625dfc1,0x362f538f,0x3638c5b7, + 0x36423639,0x364ba518,0x36551252,0x365e7deb, + 0x3667e7e2,0x36715039,0x367ab6f0,0x36841c07, + 0x368d7f81,0x3696e15d,0x36a0419d,0x36a9a040, + 0x36b2fd49,0x36bc58b8,0x36c5b28e,0x36cf0acb, + 0x36d86170,0x36e1b680,0x36eb09f8,0x36f45bdc, + 0x36fdac2b,0x3706fae7,0x37104810,0x371993a7, + 0x3722ddad,0x372c2622,0x37356d08,0x373eb25f, + 0x3747f629,0x37513865,0x375a7914,0x3763b838, + 0x376cf5d0,0x377631e0,0x377f6c64,0x3788a561, + 0x3791dcd6,0x379b12c4,0x37a4472c,0x37ad7a0e, + 0x37b6ab6a,0x37bfdb44,0x37c90999,0x37d2366d, + 0x37db61be,0x37e48b8e,0x37edb3de,0x37f6daae, + 0x38000000,0x380923d3,0x3812462a,0x381b6703, + 0x38248660,0x382da442,0x3836c0aa,0x383fdb97, + 0x3848f50c,0x38520d09,0x385b238d,0x3864389b, + 0x386d4c33,0x38765e55,0x387f6f01,0x38887e3b, + 0x38918c00,0x389a9853,0x38a3a334,0x38acaca3, + 0x38b5b4a3,0x38bebb32,0x38c7c051,0x38d0c402, + 0x38d9c645,0x38e2c71b,0x38ebc685,0x38f4c482, + 0x38fdc114,0x3906bc3c,0x390fb5fa,0x3918ae4f, + 0x3921a53a,0x392a9abe,0x39338edb,0x393c8192, + 0x394572e2,0x394e62ce,0x39575155,0x39603e77, + 0x39692a36,0x39721494,0x397afd8f,0x3983e527, + 0x398ccb60,0x3995b039,0x399e93b2,0x39a775cc, + 0x39b05689,0x39b935e8,0x39c213e9,0x39caf08e, + 0x39d3cbd9,0x39dca5c7,0x39e57e5b,0x39ee5596, + 0x39f72b77,0x3a000000,0x3a08d331,0x3a11a50a, + 0x3a1a758d,0x3a2344ba,0x3a2c1291,0x3a34df13, + 0x3a3daa41,0x3a46741b,0x3a4f3ca3,0x3a5803d7, + 0x3a60c9ba,0x3a698e4b,0x3a72518b,0x3a7b137c, + 0x3a83d41d,0x3a8c936f,0x3a955173,0x3a9e0e29, + 0x3aa6c992,0x3aaf83ae,0x3ab83c7e,0x3ac0f403, + 0x3ac9aa3c,0x3ad25f2c,0x3adb12d1,0x3ae3c52d, + 0x3aec7642,0x3af5260e,0x3afdd492,0x3b0681d0, + 0x3b0f2dc6,0x3b17d878,0x3b2081e4,0x3b292a0c, + 0x3b31d0f0,0x3b3a7690,0x3b431aec,0x3b4bbe06, + 0x3b545fdf,0x3b5d0077,0x3b659fcd,0x3b6e3de4, + 0x3b76daba,0x3b7f7651,0x3b8810aa,0x3b90a9c4, + 0x3b9941a1,0x3ba1d842,0x3baa6da5,0x3bb301cd, + 0x3bbb94b9,0x3bc4266a,0x3bccb6e2,0x3bd5461f, + 0x3bddd423,0x3be660ee,0x3beeec81,0x3bf776dc, + 0x3c000000,0x3c0887ed,0x3c110ea4,0x3c199426, + 0x3c221872,0x3c2a9b8a,0x3c331d6e,0x3c3b9e1d, + 0x3c441d9a,0x3c4c9be5,0x3c5518fd,0x3c5d94e3, + 0x3c660f98,0x3c6e891d,0x3c770172,0x3c7f7898, + 0x3c87ee8e,0x3c906356,0x3c98d6ef,0x3ca1495b, + 0x3ca9ba9a,0x3cb22aac,0x3cba9992,0x3cc3074c, + 0x3ccb73dc,0x3cd3df41,0x3cdc497b,0x3ce4b28c, + 0x3ced1a73,0x3cf58132,0x3cfde6c8,0x3d064b37, + 0x3d0eae7f,0x3d17109f,0x3d1f719a,0x3d27d16e, + 0x3d30301d,0x3d388da8,0x3d40ea0d,0x3d49454f, + 0x3d519f6d,0x3d59f867,0x3d625040,0x3d6aa6f6, + 0x3d72fc8b,0x3d7b50fe,0x3d83a451,0x3d8bf683, + 0x3d944796,0x3d9c9788,0x3da4e65c,0x3dad3412, + 0x3db580a9,0x3dbdcc24,0x3dc61680,0x3dce5fc0, + 0x3dd6a7e4,0x3ddeeeed,0x3de734d9,0x3def79ab, + 0x3df7bd62,0x3e000000,0x3e084184,0x3e1081ee, + 0x3e18c140,0x3e20ff7a,0x3e293c9c,0x3e3178a7, + 0x3e39b39a,0x3e41ed77,0x3e4a263d,0x3e525def, + 0x3e5a948b,0x3e62ca12,0x3e6afe85,0x3e7331e4, + 0x3e7b642f,0x3e839567,0x3e8bc58c,0x3e93f49f, + 0x3e9c22a1,0x3ea44f91,0x3eac7b6f,0x3eb4a63e, + 0x3ebccffb,0x3ec4f8aa,0x3ecd2049,0x3ed546d9, + 0x3edd6c5a,0x3ee590cd,0x3eedb433,0x3ef5d68c, + 0x3efdf7d7,0x3f061816,0x3f0e3749,0x3f165570, + 0x3f1e728c,0x3f268e9d,0x3f2ea9a4,0x3f36c3a0, + 0x3f3edc93,0x3f46f47c,0x3f4f0b5d,0x3f572135, + 0x3f5f3606,0x3f6749cf,0x3f6f5c90,0x3f776e4a, + 0x3f7f7efe,0x3f878eab,0x3f8f9d53,0x3f97aaf6, + 0x3f9fb793,0x3fa7c32c,0x3fafcdc1,0x3fb7d752, + 0x3fbfdfe0,0x3fc7e76b,0x3fcfedf3,0x3fd7f378, + 0x3fdff7fc,0x3fe7fb7f,0x3feffe00,0x3ff7ff80, + 0x3fffffff, +}; + +static const int32_t av_sqr_exp_multbl_sf[2] = { + 0x20000000,0x2d413ccd, +}; + +static const int32_t av_costbl_1_sf[16] = { + 0x40000000,0x3ec52fa0,0x3b20d79e,0x3536cc52, + 0x2d413ccd,0x238e7673,0x187de2a7,0x0c7c5c1e, + 0x00000000,0xf383a3e3,0xe7821d5a,0xdc71898e, + 0xd2bec334,0xcac933af,0xc4df2863,0xc13ad061, +}; + +static const int32_t av_costbl_2_sf[32] = { + 0x40000000,0x3fffb10b,0x3ffec42d,0x3ffd3969, + 0x3ffb10c1,0x3ff84a3c,0x3ff4e5e0,0x3ff0e3b6, + 0x3fec43c7,0x3fe7061f,0x3fe12acb,0x3fdab1d9, + 0x3fd39b5a,0x3fcbe75e,0x3fc395f9,0x3fbaa740, + 0x3fb11b48,0x3fa6f228,0x3f9c2bfb,0x3f90c8da, + 0x3f84c8e2,0x3f782c30,0x3f6af2e3,0x3f5d1d1d, + 0x3f4eaafe,0x3f3f9cab,0x3f2ff24a,0x3f1fabff, + 0x3f0ec9f5,0x3efd4c54,0x3eeb3347,0x3ed87efc, +}; + +static const int32_t av_sintbl_2_sf[32] = { + 0x00000000,0x006487c4,0x00c90e90,0x012d936c, + 0x0192155f,0x01f69373,0x025b0caf,0x02bf801a, + 0x0323ecbe,0x038851a2,0x03ecadcf,0x0451004d, + 0x04b54825,0x0519845e,0x057db403,0x05e1d61b, + 0x0645e9af,0x06a9edc9,0x070de172,0x0771c3b3, + 0x07d59396,0x08395024,0x089cf867,0x09008b6a, + 0x09640837,0x09c76dd8,0x0a2abb59,0x0a8defc3, + 0x0af10a22,0x0b540982,0x0bb6ecef,0x0c19b374, +}; + +static const int32_t av_costbl_3_sf[32] = { + 0x40000000,0x3fffffec,0x3fffffb1,0x3fffff4e, + 0x3ffffec4,0x3ffffe13,0x3ffffd39,0x3ffffc39, + 0x3ffffb11,0x3ffff9c1,0x3ffff84a,0x3ffff6ac, + 0x3ffff4e6,0x3ffff2f8,0x3ffff0e3,0x3fffeea7, + 0x3fffec43,0x3fffe9b7,0x3fffe705,0x3fffe42a, + 0x3fffe128,0x3fffddff,0x3fffdaae,0x3fffd736, + 0x3fffd396,0x3fffcfcf,0x3fffcbe0,0x3fffc7ca, + 0x3fffc38c,0x3fffbf27,0x3fffba9b,0x3fffb5e7, +}; + +static const int32_t av_sintbl_3_sf[32] = { + 0x00000000,0x0003243f,0x0006487f,0x00096cbe, + 0x000c90fe,0x000fb53d,0x0012d97c,0x0015fdbb, + 0x001921fb,0x001c463a,0x001f6a79,0x00228eb8, + 0x0025b2f7,0x0028d736,0x002bfb74,0x002f1fb3, + 0x003243f1,0x00356830,0x00388c6e,0x003bb0ac, + 0x003ed4ea,0x0041f928,0x00451d66,0x004841a3, + 0x004b65e1,0x004e8a1e,0x0051ae5b,0x0054d297, + 0x0057f6d4,0x005b1b10,0x005e3f4c,0x00616388, +}; + +static const int32_t av_costbl_4_sf[33] = { + 0x40000000,0x40000000,0x40000000,0x40000000, + 0x40000000,0x40000000,0x3fffffff,0x3fffffff, + 0x3fffffff,0x3ffffffe,0x3ffffffe,0x3ffffffe, + 0x3ffffffd,0x3ffffffd,0x3ffffffc,0x3ffffffc, + 0x3ffffffb,0x3ffffffa,0x3ffffffa,0x3ffffff9, + 0x3ffffff8,0x3ffffff7,0x3ffffff7,0x3ffffff6, + 0x3ffffff5,0x3ffffff4,0x3ffffff3,0x3ffffff2, + 0x3ffffff1,0x3ffffff0,0x3fffffef,0x3fffffed, + 0x3fffffec, +}; + +static const int32_t av_sintbl_4_sf[33] = { + 0x00000000,0x00001922,0x00003244,0x00004b66, + 0x00006488,0x00007daa,0x000096cc,0x0000afee, + 0x0000c910,0x0000e232,0x0000fb54,0x00011476, + 0x00012d98,0x000146ba,0x00015fdc,0x000178fe, + 0x00019220,0x0001ab42,0x0001c464,0x0001dd86, + 0x0001f6a8,0x00020fca,0x000228ec,0x0002420e, + 0x00025b30,0x00027452,0x00028d74,0x0002a696, + 0x0002bfb7,0x0002d8d9,0x0002f1fb,0x00030b1d, + 0x0003243f, +}; +#endif /* AVUTIL_SOFTFLOAT_TABLES_H */ -- 1.8.2.1
On Thu, Apr 30, 2015 at 01:51:37PM +0200, Nedeljko Babic wrote:
Functions av_gt_sf, av_sqrt_sf and av_sincos_sf added to softfloat
Signed-off-by: Nedeljko Babic <nedeljko.babic@imgtec.com> --- libavutil/softfloat.h | 93 +++++++++++++++- libavutil/softfloat_tables.h | 260 +++++++++++++++++++++++++++++++++++++++++++ 2 files changed, 352 insertions(+), 1 deletion(-) create mode 100644 libavutil/softfloat_tables.h
applied thanks [...] -- Michael GnuPG fingerprint: 9FF2128B147EF6730BADF133611EC787040B0FAB If a bugfix only changes things apparently unrelated to the bug with no further explanation, that is a good sign that the bugfix is wrong.
On Thu, Apr 30, 2015 at 01:51:37PM +0200, Nedeljko Babic wrote:
Functions av_gt_sf, av_sqrt_sf and av_sincos_sf added to softfloat
Signed-off-by: Nedeljko Babic <nedeljko.babic@imgtec.com> --- libavutil/softfloat.h | 93 +++++++++++++++- libavutil/softfloat_tables.h | 260 +++++++++++++++++++++++++++++++++++++++++++ 2 files changed, 352 insertions(+), 1 deletion(-) create mode 100644 libavutil/softfloat_tables.h
[...]
+/** + * Rounding-to-nearest used. + */ +static av_always_inline void av_sincos_sf(int a, int *s, int *c)
this function seems to be buggy the sine values are very accurate but the cosine values are not look: (tested with libavutil/softfloat-test and tightened up thresholds) sincos FAIL 8 0.642788 0.766263 0.642788 0.766044 sincos FAIL 14 0.939693 0.342220 0.939693 0.342020 sincos FAIL 15 0.965926 0.259066 0.965926 0.258819 sincos FAIL 16 0.984808 0.173942 0.984808 0.173648 sincos FAIL 17 0.996195 0.087495 0.996195 0.087156 sincos FAIL 24 0.866025 -0.499779 0.866025 -0.500000 sincos FAIL 25 0.819152 -0.573332 0.819152 -0.573576 sincos FAIL 26 0.766044 -0.642526 0.766044 -0.642788 sincos FAIL 44 -0.642788 -0.766263 -0.642788 -0.766044 sincos FAIL 50 -0.939693 -0.342220 -0.939693 -0.342020 sincos FAIL 51 -0.965926 -0.259066 -0.965926 -0.258819 sincos FAIL 52 -0.984808 -0.173942 -0.984808 -0.173648 sincos FAIL 53 -0.996195 -0.087495 -0.996195 -0.087156 sincos FAIL 60 -0.866025 0.499779 -0.866025 0.500000 sincos FAIL 61 -0.819152 0.573332 -0.819152 0.573576 sincos FAIL 62 -0.766044 0.642526 -0.766044 0.642788 sincos FAIL 80 0.642788 0.766263 0.642788 0.766044 sincos FAIL 86 0.939693 0.342220 0.939693 0.342020 sincos FAIL 87 0.965926 0.259066 0.965926 0.258819 sincos FAIL 88 0.984808 0.173942 0.984808 0.173648 sincos FAIL 89 0.996195 0.087495 0.996195 0.087156 sincos FAIL 96 0.866025 -0.499779 0.866025 -0.500000 sincos FAIL 97 0.819152 -0.573332 0.819152 -0.573576 sincos FAIL 98 0.766044 -0.642526 0.766044 -0.642788 sincos FAIL 116 -0.642788 -0.766263 -0.642788 -0.766044 sincos FAIL 122 -0.939693 -0.342220 -0.939693 -0.342020 sincos FAIL 123 -0.965926 -0.259066 -0.965926 -0.258819 sincos FAIL 124 -0.984808 -0.173942 -0.984808 -0.173648 sincos FAIL 125 -0.996195 -0.087495 -0.996195 -0.087156 sincos FAIL 132 -0.866025 0.499779 -0.866025 0.500000 sincos FAIL 133 -0.819152 0.573332 -0.819152 0.573576 sincos FAIL 134 -0.766044 0.642526 -0.766044 0.642788 [...] -- Michael GnuPG fingerprint: 9FF2128B147EF6730BADF133611EC787040B0FAB The greatest way to live with honor in this world is to be what we pretend to be. -- Socrates
+/** + * Rounding-to-nearest used. + */ +static av_always_inline void av_sincos_sf(int a, int *s, int *c)
this function seems to be buggy the sine values are very accurate but the cosine values are not
look: (tested with libavutil/softfloat-test and tightened up thresholds)
Could you please send me configuration which you used here? I tried with several configurations and didn't have a problem with sincos on this test. Thanks, Nedeljko
On Mon, Jul 27, 2015 at 01:36:42PM +0000, Nedeljko Babic wrote:
+/** + * Rounding-to-nearest used. + */ +static av_always_inline void av_sincos_sf(int a, int *s, int *c)
this function seems to be buggy the sine values are very accurate but the cosine values are not
look: (tested with libavutil/softfloat-test and tightened up thresholds)
Could you please send me configuration which you used here?
I tried with several configurations and didn't have a problem with sincos on this test.
sure first apply: (to use the same threshold for cosine as sine diff --git a/libavutil/softfloat.c b/libavutil/softfloat.c index a9d1b0c..597be04 100644 --- a/libavutil/softfloat.c +++ b/libavutil/softfloat.c @@ -104,7 +104,7 @@ int main(void){ av_sincos_sf(i*(1ULL<<32)/36/4, &s, &c); errs = (double)s/ (1<<30) - sin(i*M_PI/36); errc = (double)c/ (1<<30) - cos(i*M_PI/36); - if (fabs(errs) > 0.00000002 || fabs(errc) >0.001) { + if (fabs(errs) > 0.00000002 || fabs(errc) >0.00000002) { printf("sincos FAIL %d %f %f %f %f\n", i, (float)s/ (1<<30), (float)c/ (1<<30), sin(i*M_PI/36), cos(i*M_PI/36)); } ./configure && make -j12 libavutil/softfloat-test && libavutil/softfloat-test [...] -- Michael GnuPG fingerprint: 9FF2128B147EF6730BADF133611EC787040B0FAB it is not once nor twice but times without number that the same ideas make their appearance in the world. -- Aristotle
participants (3)
-
Michael Niedermayer -
Michael Niedermayer -
Nedeljko Babic