[deb_ffmpeg.git] / ffmpeg / libavutil / aarch64 / float_dsp_neon.S

/*
 * ARM NEON optimised Float DSP functions
 * Copyright (c) 2008 Mans Rullgard <mans@mansr.com>
 * Copyright (c) 2014 Janne Grunau <janne-libav@jannau.net>
 *
 * 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
 */

#include "config.h"
#include "asm.S"

function ff_vector_fmul_neon, export=1
1:      subs            w3,  w3,  #16
        ld1             {v0.4S, v1.4S}, [x1], #32
        ld1             {v2.4S, v3.4S}, [x1], #32
        ld1             {v4.4S, v5.4S}, [x2], #32
        ld1             {v6.4S, v7.4S}, [x2], #32
        fmul            v16.4S, v0.4S,  v4.4S
        fmul            v17.4S, v1.4S,  v5.4S
        fmul            v18.4S, v2.4S,  v6.4S
        fmul            v19.4S, v3.4S,  v7.4S
        st1             {v16.4S, v17.4S}, [x0], #32
        st1             {v18.4S, v19.4S}, [x0], #32
        b.ne            1b
        ret
endfunc

function ff_vector_fmac_scalar_neon, export=1
        mov             x3,  #-32
1:      subs            w2,  w2,  #16
        ld1             {v16.4S, v17.4S}, [x0], #32
        ld1             {v18.4S, v19.4S}, [x0], x3
        ld1             {v4.4S,  v5.4S},  [x1], #32
        ld1             {v6.4S,  v7.4S},  [x1], #32
        fmla            v16.4S, v4.4S,  v0.S[0]
        fmla            v17.4S, v5.4S,  v0.S[0]
        fmla            v18.4S, v6.4S,  v0.S[0]
        fmla            v19.4S, v7.4S,  v0.S[0]
        st1             {v16.4S, v17.4S}, [x0], #32
        st1             {v18.4S, v19.4S}, [x0], #32
        b.ne            1b
        ret
endfunc

function ff_vector_fmul_scalar_neon, export=1
        mov             w4,  #15
        bics            w3,  w2,  w4
        dup             v16.4S, v0.S[0]
        b.eq            3f
        ld1             {v0.4S, v1.4S}, [x1], #32
1:      subs            w3,  w3,  #16
        fmul            v0.4S,  v0.4S,  v16.4S
        ld1             {v2.4S, v3.4S}, [x1], #32
        fmul            v1.4S,  v1.4S,  v16.4S
        fmul            v2.4S,  v2.4S,  v16.4S
        st1             {v0.4S, v1.4S}, [x0], #32
        fmul            v3.4S,  v3.4S,  v16.4S
        b.eq            2f
        ld1             {v0.4S, v1.4S}, [x1], #32
        st1             {v2.4S, v3.4S}, [x0], #32
        b               1b
2:      ands            w2,  w2,  #15
        st1             {v2.4S, v3.4S}, [x0], #32
        b.eq            4f
3:      ld1             {v0.4S}, [x1], #16
        fmul            v0.4S,  v0.4S,  v16.4S
        st1             {v0.4S}, [x0], #16
        subs            w2,  w2,  #4
        b.gt            3b
4:      ret
endfunc

function ff_vector_dmul_scalar_neon, export=1
        dup             v16.2D, v0.D[0]
        ld1             {v0.2D, v1.2D}, [x1], #32
1:      subs            w2,  w2,  #8
        fmul            v0.2D,  v0.2D,  v16.2D
        ld1             {v2.2D, v3.2D}, [x1], #32
        fmul            v1.2D,  v1.2D,  v16.2D
        fmul            v2.2D,  v2.2D,  v16.2D
        st1             {v0.2D, v1.2D}, [x0], #32
        fmul            v3.2D,  v3.2D,  v16.2D
        ld1             {v0.2D, v1.2D}, [x1], #32
        st1             {v2.2D, v3.2D}, [x0], #32
        b.gt            1b
        ret
endfunc

function ff_vector_fmul_window_neon, export=1
        sxtw            x4,  w4                 // len
        sub             x2,  x2,  #8
        sub             x5,  x4,  #2
        add             x2,  x2,  x5, lsl #2    // src1 + 4 * (len - 4)
        add             x6,  x3,  x5, lsl #3    // win  + 8 * (len - 2)
        add             x5,  x0,  x5, lsl #3    // dst  + 8 * (len - 2)
        mov             x7,  #-16
        ld1             {v0.4S},  [x1], #16     // s0
        ld1             {v2.4S},  [x3], #16     // wi
        ld1             {v1.4S},  [x2], x7      // s1
1:      ld1             {v3.4S},  [x6], x7      // wj
        subs            x4,  x4,  #4
        fmul            v17.4S, v0.4S,  v2.4S   // s0 * wi
        rev64           v4.4S,  v1.4S
        rev64           v5.4S,  v3.4S
        rev64           v17.4S, v17.4S
        ext             v4.16B,  v4.16B,  v4.16B,  #8 // s1_r
        ext             v5.16B,  v5.16B,  v5.16B,  #8 // wj_r
        ext             v17.16B, v17.16B, v17.16B, #8 // (s0 * wi)_rev
        fmul            v16.4S, v0.4S,  v5.4S  // s0 * wj_r
        fmla            v17.4S, v1.4S,  v3.4S  // (s0 * wi)_rev + s1 * wj
        b.eq            2f
        ld1             {v0.4S},  [x1], #16
        fmls            v16.4S, v4.4S,  v2.4S  // s0 * wj_r - s1_r * wi
        st1             {v17.4S}, [x5], x7
        ld1             {v2.4S},  [x3], #16
        ld1             {v1.4S},  [x2], x7
        st1             {v16.4S}, [x0], #16
        b               1b
2:
        fmls            v16.4S, v4.4S,  v2.4S  // s0 * wj_r - s1_r * wi
        st1             {v17.4S}, [x5], x7
        st1             {v16.4S}, [x0], #16
        ret
endfunc

function ff_vector_fmul_add_neon, export=1
        ld1             {v0.4S, v1.4S},  [x1], #32
        ld1             {v2.4S, v3.4S},  [x2], #32
        ld1             {v4.4S, v5.4S},  [x3], #32
1:      subs            w4,  w4,  #8
        fmla            v4.4S,  v0.4S,  v2.4S
        fmla            v5.4S,  v1.4S,  v3.4S
        b.eq            2f
        ld1             {v0.4S, v1.4S},  [x1], #32
        ld1             {v2.4S, v3.4S},  [x2], #32
        st1             {v4.4S, v5.4S},  [x0], #32
        ld1             {v4.4S, v5.4S},  [x3], #32
        b               1b
2:      st1             {v4.4S, v5.4S},  [x0], #32
        ret
endfunc

function ff_vector_fmul_reverse_neon, export=1
        sxtw            x3,  w3
        add             x2,  x2,  x3,  lsl #2
        sub             x2,  x2,  #32
        mov             x4, #-32
        ld1             {v2.4S, v3.4S},  [x2], x4
        ld1             {v0.4S, v1.4S},  [x1], #32
1:      subs            x3,  x3,  #8
        rev64           v3.4S,  v3.4S
        rev64           v2.4S,  v2.4S
        ext             v3.16B, v3.16B, v3.16B,  #8
        ext             v2.16B, v2.16B, v2.16B,  #8
        fmul            v16.4S, v0.4S,  v3.4S
        fmul            v17.4S, v1.4S,  v2.4S
        b.eq            2f
        ld1             {v2.4S, v3.4S},  [x2], x4
        ld1             {v0.4S, v1.4S},  [x1], #32
        st1             {v16.4S, v17.4S},  [x0], #32
        b               1b
2:      st1             {v16.4S, v17.4S},  [x0], #32
        ret
endfunc

function ff_butterflies_float_neon, export=1
1:      ld1             {v0.4S}, [x0]
        ld1             {v1.4S}, [x1]
        subs            w2,  w2,  #4
        fsub            v2.4S,   v0.4S,  v1.4S
        fadd            v3.4S,   v0.4S,  v1.4S
        st1             {v2.4S}, [x1],   #16
        st1             {v3.4S}, [x0],   #16
        b.gt            1b
        ret
endfunc

function ff_scalarproduct_float_neon, export=1
        movi            v2.4S,  #0
1:      ld1             {v0.4S}, [x0],   #16
        ld1             {v1.4S}, [x1],   #16
        subs            w2,      w2,     #4
        fmla            v2.4S,   v0.4S,  v1.4S
        b.gt            1b
        faddp           v0.4S,   v2.4S,  v2.4S
        faddp           s0,      v0.2S
        ret
endfunc
Commit	Line	Data
2ba45a60 DM	1	/*
	2	* ARM NEON optimised Float DSP functions
	3	* Copyright (c) 2008 Mans Rullgard <mans@mansr.com>
	4	* Copyright (c) 2014 Janne Grunau <janne-libav@jannau.net>
	5	*
	6	* This file is part of FFmpeg.
	7	*
	8	* FFmpeg is free software; you can redistribute it and/or
	9	* modify it under the terms of the GNU Lesser General Public
	10	* License as published by the Free Software Foundation; either
	11	* version 2.1 of the License, or (at your option) any later version.
	12	*
	13	* FFmpeg is distributed in the hope that it will be useful,
	14	* but WITHOUT ANY WARRANTY; without even the implied warranty of
	15	* MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
	16	* Lesser General Public License for more details.
	17	*
	18	* You should have received a copy of the GNU Lesser General Public
	19	* License along with FFmpeg; if not, write to the Free Software
	20	* Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
	21	*/
	22
	23	#include "config.h"
	24	#include "asm.S"
	25
	26	function ff_vector_fmul_neon, export=1
	27	1: subs w3, w3, #16
	28	ld1 {v0.4S, v1.4S}, [x1], #32
	29	ld1 {v2.4S, v3.4S}, [x1], #32
	30	ld1 {v4.4S, v5.4S}, [x2], #32
	31	ld1 {v6.4S, v7.4S}, [x2], #32
	32	fmul v16.4S, v0.4S, v4.4S
	33	fmul v17.4S, v1.4S, v5.4S
	34	fmul v18.4S, v2.4S, v6.4S
	35	fmul v19.4S, v3.4S, v7.4S
	36	st1 {v16.4S, v17.4S}, [x0], #32
	37	st1 {v18.4S, v19.4S}, [x0], #32
	38	b.ne 1b
	39	ret
	40	endfunc
	41
	42	function ff_vector_fmac_scalar_neon, export=1
	43	mov x3, #-32
	44	1: subs w2, w2, #16
	45	ld1 {v16.4S, v17.4S}, [x0], #32
	46	ld1 {v18.4S, v19.4S}, [x0], x3
	47	ld1 {v4.4S, v5.4S}, [x1], #32
	48	ld1 {v6.4S, v7.4S}, [x1], #32
	49	fmla v16.4S, v4.4S, v0.S[0]
	50	fmla v17.4S, v5.4S, v0.S[0]
	51	fmla v18.4S, v6.4S, v0.S[0]
	52	fmla v19.4S, v7.4S, v0.S[0]
	53	st1 {v16.4S, v17.4S}, [x0], #32
	54	st1 {v18.4S, v19.4S}, [x0], #32
	55	b.ne 1b
	56	ret
	57	endfunc
	58
	59	function ff_vector_fmul_scalar_neon, export=1
	60	mov w4, #15
	61	bics w3, w2, w4
	62	dup v16.4S, v0.S[0]
	63	b.eq 3f
	64	ld1 {v0.4S, v1.4S}, [x1], #32
65	1: subs w3, w3, #16
66	fmul v0.4S, v0.4S, v16.4S
67	ld1 {v2.4S, v3.4S}, [x1], #32
68	fmul v1.4S, v1.4S, v16.4S
69	fmul v2.4S, v2.4S, v16.4S
70	st1 {v0.4S, v1.4S}, [x0], #32
71	fmul v3.4S, v3.4S, v16.4S
72	b.eq 2f
73	ld1 {v0.4S, v1.4S}, [x1], #32
74	st1 {v2.4S, v3.4S}, [x0], #32
75	b 1b
76	2: ands w2, w2, #15
77	st1 {v2.4S, v3.4S}, [x0], #32
78	b.eq 4f
79	3: ld1 {v0.4S}, [x1], #16
80	fmul v0.4S, v0.4S, v16.4S
81	st1 {v0.4S}, [x0], #16
82	subs w2, w2, #4
83	b.gt 3b
84	4: ret
85	endfunc
86
87	function ff_vector_dmul_scalar_neon, export=1
88	dup v16.2D, v0.D[0]
89	ld1 {v0.2D, v1.2D}, [x1], #32
90	1: subs w2, w2, #8
91	fmul v0.2D, v0.2D, v16.2D
92	ld1 {v2.2D, v3.2D}, [x1], #32
93	fmul v1.2D, v1.2D, v16.2D
94	fmul v2.2D, v2.2D, v16.2D
95	st1 {v0.2D, v1.2D}, [x0], #32
96	fmul v3.2D, v3.2D, v16.2D
97	ld1 {v0.2D, v1.2D}, [x1], #32
98	st1 {v2.2D, v3.2D}, [x0], #32
99	b.gt 1b
100	ret
101	endfunc
102
103	function ff_vector_fmul_window_neon, export=1
104	sxtw x4, w4 // len
105	sub x2, x2, #8
106	sub x5, x4, #2
107	add x2, x2, x5, lsl #2 // src1 + 4 * (len - 4)
108	add x6, x3, x5, lsl #3 // win + 8 * (len - 2)
109	add x5, x0, x5, lsl #3 // dst + 8 * (len - 2)
110	mov x7, #-16
111	ld1 {v0.4S}, [x1], #16 // s0
112	ld1 {v2.4S}, [x3], #16 // wi
113	ld1 {v1.4S}, [x2], x7 // s1
114	1: ld1 {v3.4S}, [x6], x7 // wj
115	subs x4, x4, #4
116	fmul v17.4S, v0.4S, v2.4S // s0 * wi
117	rev64 v4.4S, v1.4S
118	rev64 v5.4S, v3.4S
119	rev64 v17.4S, v17.4S
120	ext v4.16B, v4.16B, v4.16B, #8 // s1_r
121	ext v5.16B, v5.16B, v5.16B, #8 // wj_r
122	ext v17.16B, v17.16B, v17.16B, #8 // (s0 * wi)_rev
123	fmul v16.4S, v0.4S, v5.4S // s0 * wj_r
124	fmla v17.4S, v1.4S, v3.4S // (s0 * wi)_rev + s1 * wj
125	b.eq 2f
126	ld1 {v0.4S}, [x1], #16
127	fmls v16.4S, v4.4S, v2.4S // s0 * wj_r - s1_r * wi
128	st1 {v17.4S}, [x5], x7
129	ld1 {v2.4S}, [x3], #16
130	ld1 {v1.4S}, [x2], x7
131	st1 {v16.4S}, [x0], #16
132	b 1b
133	2:
134	fmls v16.4S, v4.4S, v2.4S // s0 * wj_r - s1_r * wi
135	st1 {v17.4S}, [x5], x7
136	st1 {v16.4S}, [x0], #16
137	ret
138	endfunc
139
140	function ff_vector_fmul_add_neon, export=1
141	ld1 {v0.4S, v1.4S}, [x1], #32
142	ld1 {v2.4S, v3.4S}, [x2], #32
143	ld1 {v4.4S, v5.4S}, [x3], #32
144	1: subs w4, w4, #8
145	fmla v4.4S, v0.4S, v2.4S
146	fmla v5.4S, v1.4S, v3.4S
147	b.eq 2f
148	ld1 {v0.4S, v1.4S}, [x1], #32
149	ld1 {v2.4S, v3.4S}, [x2], #32
150	st1 {v4.4S, v5.4S}, [x0], #32
151	ld1 {v4.4S, v5.4S}, [x3], #32
152	b 1b
153	2: st1 {v4.4S, v5.4S}, [x0], #32
154	ret
155	endfunc
156
157	function ff_vector_fmul_reverse_neon, export=1
158	sxtw x3, w3
159	add x2, x2, x3, lsl #2
160	sub x2, x2, #32
161	mov x4, #-32
162	ld1 {v2.4S, v3.4S}, [x2], x4
163	ld1 {v0.4S, v1.4S}, [x1], #32
164	1: subs x3, x3, #8
165	rev64 v3.4S, v3.4S
166	rev64 v2.4S, v2.4S
167	ext v3.16B, v3.16B, v3.16B, #8
168	ext v2.16B, v2.16B, v2.16B, #8
169	fmul v16.4S, v0.4S, v3.4S
170	fmul v17.4S, v1.4S, v2.4S
171	b.eq 2f
172	ld1 {v2.4S, v3.4S}, [x2], x4
173	ld1 {v0.4S, v1.4S}, [x1], #32
174	st1 {v16.4S, v17.4S}, [x0], #32
175	b 1b
176	2: st1 {v16.4S, v17.4S}, [x0], #32
177	ret
178	endfunc
179
180	function ff_butterflies_float_neon, export=1
181	1: ld1 {v0.4S}, [x0]
182	ld1 {v1.4S}, [x1]
183	subs w2, w2, #4
184	fsub v2.4S, v0.4S, v1.4S
185	fadd v3.4S, v0.4S, v1.4S
186	st1 {v2.4S}, [x1], #16
187	st1 {v3.4S}, [x0], #16
188	b.gt 1b
189	ret
190	endfunc
191
192	function ff_scalarproduct_float_neon, export=1
193	movi v2.4S, #0
194	1: ld1 {v0.4S}, [x0], #16
195	ld1 {v1.4S}, [x1], #16
196	subs w2, w2, #4
197	fmla v2.4S, v0.4S, v1.4S
198	b.gt 1b
199	faddp v0.4S, v2.4S, v2.4S
200	faddp s0, v0.2S
201	ret
202	endfunc