aboutsummaryrefslogtreecommitdiffstats
path: root/libavutil/x86/float_dsp.asm
blob: 004e6cf1fe5b8a9b53b4e3527f760ebb799e4537 (plain) (blame)
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
;*****************************************************************************
;* x86-optimized Float DSP functions
;*
;* Copyright 2006 Loren Merritt
;*
;* 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 "x86util.asm"

SECTION .text

;-----------------------------------------------------------------------------
; void vector_fmul(float *dst, const float *src0, const float *src1, int len)
;-----------------------------------------------------------------------------
%macro VECTOR_FMUL 0
cglobal vector_fmul, 4,4,2, dst, src0, src1, len
    lea       lenq, [lend*4 - 2*mmsize]
ALIGN 16
.loop:
    mova      m0,   [src0q + lenq]
    mova      m1,   [src0q + lenq + mmsize]
    mulps     m0, m0, [src1q + lenq]
    mulps     m1, m1, [src1q + lenq + mmsize]
    mova      [dstq + lenq], m0
    mova      [dstq + lenq + mmsize], m1

    sub       lenq, 2*mmsize
    jge       .loop
    REP_RET
%endmacro

INIT_XMM sse
VECTOR_FMUL
%if HAVE_AVX_EXTERNAL
INIT_YMM avx
VECTOR_FMUL
%endif

;------------------------------------------------------------------------------
; void ff_vector_fmac_scalar(float *dst, const float *src, float mul, int len)
;------------------------------------------------------------------------------

%macro VECTOR_FMAC_SCALAR 0
%if UNIX64
cglobal vector_fmac_scalar, 3,3,3, dst, src, len
%else
cglobal vector_fmac_scalar, 4,4,3, dst, src, mul, len
%endif
%if ARCH_X86_32
    VBROADCASTSS m0, mulm
%else
%if WIN64
    mova       xmm0, xmm2
%endif
    shufps     xmm0, xmm0, 0
%if cpuflag(avx)
    vinsertf128  m0, m0, xmm0, 1
%endif
%endif
    lea    lenq, [lend*4-2*mmsize]
.loop:
    mulps    m1, m0, [srcq+lenq       ]
    mulps    m2, m0, [srcq+lenq+mmsize]
    addps    m1, m1, [dstq+lenq       ]
    addps    m2, m2, [dstq+lenq+mmsize]
    mova  [dstq+lenq       ], m1
    mova  [dstq+lenq+mmsize], m2
    sub    lenq, 2*mmsize
    jge .loop
    REP_RET
%endmacro

INIT_XMM sse
VECTOR_FMAC_SCALAR
%if HAVE_AVX_EXTERNAL
INIT_YMM avx
VECTOR_FMAC_SCALAR
%endif

;------------------------------------------------------------------------------
; void ff_vector_fmul_scalar(float *dst, const float *src, float mul, int len)
;------------------------------------------------------------------------------

%macro VECTOR_FMUL_SCALAR 0
%if UNIX64
cglobal vector_fmul_scalar, 3,3,2, dst, src, len
%else
cglobal vector_fmul_scalar, 4,4,3, dst, src, mul, len
%endif
%if ARCH_X86_32
    movss    m0, mulm
%elif WIN64
    SWAP 0, 2
%endif
    shufps   m0, m0, 0
    lea    lenq, [lend*4-mmsize]
.loop:
    mova     m1, [srcq+lenq]
    mulps    m1, m0
    mova  [dstq+lenq], m1
    sub    lenq, mmsize
    jge .loop
    REP_RET
%endmacro

INIT_XMM sse
VECTOR_FMUL_SCALAR

;------------------------------------------------------------------------------
; void ff_vector_dmul_scalar(double *dst, const double *src, double mul,
;                            int len)
;------------------------------------------------------------------------------

%macro VECTOR_DMUL_SCALAR 0
%if ARCH_X86_32
cglobal vector_dmul_scalar, 3,4,3, dst, src, mul, len, lenaddr
    mov          lenq, lenaddrm
%elif UNIX64
cglobal vector_dmul_scalar, 3,3,3, dst, src, len
%else
cglobal vector_dmul_scalar, 4,4,3, dst, src, mul, len
%endif
%if ARCH_X86_32
    VBROADCASTSD   m0, mulm
%else
%if WIN64
    movlhps      xmm2, xmm2
%if cpuflag(avx)
    vinsertf128  ymm2, ymm2, xmm2, 1
%endif
    SWAP 0, 2
%else
    movlhps      xmm0, xmm0
%if cpuflag(avx)
    vinsertf128  ymm0, ymm0, xmm0, 1
%endif
%endif
%endif
    lea          lenq, [lend*8-2*mmsize]
.loop:
    mulpd          m1, m0, [srcq+lenq       ]
    mulpd          m2, m0, [srcq+lenq+mmsize]
    mova   [dstq+lenq       ], m1
    mova   [dstq+lenq+mmsize], m2
    sub          lenq, 2*mmsize
    jge .loop
    REP_RET
%endmacro

INIT_XMM sse2
VECTOR_DMUL_SCALAR
%if HAVE_AVX_EXTERNAL
INIT_YMM avx
VECTOR_DMUL_SCALAR
%endif

;-----------------------------------------------------------------------------
; vector_fmul_add(float *dst, const float *src0, const float *src1,
;                 const float *src2, int len)
;-----------------------------------------------------------------------------
%macro VECTOR_FMUL_ADD 0
cglobal vector_fmul_add, 5,5,2, dst, src0, src1, src2, len
    lea       lenq, [lend*4 - 2*mmsize]
ALIGN 16
.loop:
    mova    m0,   [src0q + lenq]
    mova    m1,   [src0q + lenq + mmsize]
    mulps   m0, m0, [src1q + lenq]
    mulps   m1, m1, [src1q + lenq + mmsize]
    addps   m0, m0, [src2q + lenq]
    addps   m1, m1, [src2q + lenq + mmsize]
    mova    [dstq + lenq], m0
    mova    [dstq + lenq + mmsize], m1

    sub     lenq,   2*mmsize
    jge     .loop
    REP_RET
%endmacro

INIT_XMM sse
VECTOR_FMUL_ADD
%if HAVE_AVX_EXTERNAL
INIT_YMM avx
VECTOR_FMUL_ADD
%endif

;-----------------------------------------------------------------------------
; void vector_fmul_reverse(float *dst, const float *src0, const float *src1,
;                          int len)
;-----------------------------------------------------------------------------
%macro VECTOR_FMUL_REVERSE 0
cglobal vector_fmul_reverse, 4,4,2, dst, src0, src1, len
    lea       lenq, [lend*4 - 2*mmsize]
ALIGN 16
.loop:
%if cpuflag(avx)
    vmovaps     xmm0, [src1q + 16]
    vinsertf128 m0, m0, [src1q], 1
    vshufps     m0, m0, m0, q0123
    vmovaps     xmm1, [src1q + mmsize + 16]
    vinsertf128 m1, m1, [src1q + mmsize], 1
    vshufps     m1, m1, m1, q0123
%else
    mova    m0, [src1q]
    mova    m1, [src1q + mmsize]
    shufps  m0, m0, q0123
    shufps  m1, m1, q0123
%endif
    mulps   m0, m0, [src0q + lenq + mmsize]
    mulps   m1, m1, [src0q + lenq]
    mova    [dstq + lenq + mmsize], m0
    mova    [dstq + lenq], m1
    add     src1q, 2*mmsize
    sub     lenq,  2*mmsize
    jge     .loop
    REP_RET
%endmacro

INIT_XMM sse
VECTOR_FMUL_REVERSE
%if HAVE_AVX_EXTERNAL
INIT_YMM avx
VECTOR_FMUL_REVERSE
%endif

; float scalarproduct_float_sse(const float *v1, const float *v2, int len)
INIT_XMM sse
cglobal scalarproduct_float, 3,3,2, v1, v2, offset
    neg   offsetq
    shl   offsetq, 2
    sub       v1q, offsetq
    sub       v2q, offsetq
    xorps    xmm0, xmm0
.loop:
    movaps   xmm1, [v1q+offsetq]
    mulps    xmm1, [v2q+offsetq]
    addps    xmm0, xmm1
    add   offsetq, 16
    js .loop
    movhlps  xmm1, xmm0
    addps    xmm0, xmm1
    movss    xmm1, xmm0
    shufps   xmm0, xmm0, 1
    addss    xmm0, xmm1
%if ARCH_X86_64 == 0
    movss     r0m,  xmm0
    fld dword r0m
%endif
    RET