cannam@154: /* Copyright (c) 2013 Jean-Marc Valin and John Ridges
cannam@154:    Copyright (c) 2014, Cisco Systems, INC MingXiang WeiZhou MinPeng YanWang*/
cannam@154: /**
cannam@154:    @file pitch_sse.h
cannam@154:    @brief Pitch analysis
cannam@154:  */
cannam@154: 
cannam@154: /*
cannam@154:    Redistribution and use in source and binary forms, with or without
cannam@154:    modification, are permitted provided that the following conditions
cannam@154:    are met:
cannam@154: 
cannam@154:    - Redistributions of source code must retain the above copyright
cannam@154:    notice, this list of conditions and the following disclaimer.
cannam@154: 
cannam@154:    - Redistributions in binary form must reproduce the above copyright
cannam@154:    notice, this list of conditions and the following disclaimer in the
cannam@154:    documentation and/or other materials provided with the distribution.
cannam@154: 
cannam@154:    THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
cannam@154:    ``AS IS'' AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
cannam@154:    LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR
cannam@154:    A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER
cannam@154:    OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL,
cannam@154:    EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO,
cannam@154:    PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR
cannam@154:    PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF
cannam@154:    LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING
cannam@154:    NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS
cannam@154:    SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
cannam@154: */
cannam@154: 
cannam@154: #ifndef PITCH_SSE_H
cannam@154: #define PITCH_SSE_H
cannam@154: 
cannam@154: #if defined(HAVE_CONFIG_H)
cannam@154: #include "config.h"
cannam@154: #endif
cannam@154: 
cannam@154: #if defined(OPUS_X86_MAY_HAVE_SSE4_1) && defined(FIXED_POINT)
cannam@154: void xcorr_kernel_sse4_1(
cannam@154:                     const opus_int16 *x,
cannam@154:                     const opus_int16 *y,
cannam@154:                     opus_val32       sum[4],
cannam@154:                     int              len);
cannam@154: #endif
cannam@154: 
cannam@154: #if defined(OPUS_X86_MAY_HAVE_SSE) && !defined(FIXED_POINT)
cannam@154: void xcorr_kernel_sse(
cannam@154:                     const opus_val16 *x,
cannam@154:                     const opus_val16 *y,
cannam@154:                     opus_val32       sum[4],
cannam@154:                     int              len);
cannam@154: #endif
cannam@154: 
cannam@154: #if defined(OPUS_X86_PRESUME_SSE4_1) && defined(FIXED_POINT)
cannam@154: #define OVERRIDE_XCORR_KERNEL
cannam@154: #define xcorr_kernel(x, y, sum, len, arch) \
cannam@154:     ((void)arch, xcorr_kernel_sse4_1(x, y, sum, len))
cannam@154: 
cannam@154: #elif defined(OPUS_X86_PRESUME_SSE) && !defined(FIXED_POINT)
cannam@154: #define OVERRIDE_XCORR_KERNEL
cannam@154: #define xcorr_kernel(x, y, sum, len, arch) \
cannam@154:     ((void)arch, xcorr_kernel_sse(x, y, sum, len))
cannam@154: 
cannam@154: #elif (defined(OPUS_X86_MAY_HAVE_SSE4_1) && defined(FIXED_POINT)) || (defined(OPUS_X86_MAY_HAVE_SSE) && !defined(FIXED_POINT))
cannam@154: 
cannam@154: extern void (*const XCORR_KERNEL_IMPL[OPUS_ARCHMASK + 1])(
cannam@154:                     const opus_val16 *x,
cannam@154:                     const opus_val16 *y,
cannam@154:                     opus_val32       sum[4],
cannam@154:                     int              len);
cannam@154: 
cannam@154: #define OVERRIDE_XCORR_KERNEL
cannam@154: #define xcorr_kernel(x, y, sum, len, arch) \
cannam@154:     ((*XCORR_KERNEL_IMPL[(arch) & OPUS_ARCHMASK])(x, y, sum, len))
cannam@154: 
cannam@154: #endif
cannam@154: 
cannam@154: #if defined(OPUS_X86_MAY_HAVE_SSE4_1) && defined(FIXED_POINT)
cannam@154: opus_val32 celt_inner_prod_sse4_1(
cannam@154:     const opus_int16 *x,
cannam@154:     const opus_int16 *y,
cannam@154:     int               N);
cannam@154: #endif
cannam@154: 
cannam@154: #if defined(OPUS_X86_MAY_HAVE_SSE2) && defined(FIXED_POINT)
cannam@154: opus_val32 celt_inner_prod_sse2(
cannam@154:     const opus_int16 *x,
cannam@154:     const opus_int16 *y,
cannam@154:     int               N);
cannam@154: #endif
cannam@154: 
cannam@154: #if defined(OPUS_X86_MAY_HAVE_SSE2) && !defined(FIXED_POINT)
cannam@154: opus_val32 celt_inner_prod_sse(
cannam@154:     const opus_val16 *x,
cannam@154:     const opus_val16 *y,
cannam@154:     int               N);
cannam@154: #endif
cannam@154: 
cannam@154: 
cannam@154: #if defined(OPUS_X86_PRESUME_SSE4_1) && defined(FIXED_POINT)
cannam@154: #define OVERRIDE_CELT_INNER_PROD
cannam@154: #define celt_inner_prod(x, y, N, arch) \
cannam@154:     ((void)arch, celt_inner_prod_sse4_1(x, y, N))
cannam@154: 
cannam@154: #elif defined(OPUS_X86_PRESUME_SSE2) && defined(FIXED_POINT) && !defined(OPUS_X86_MAY_HAVE_SSE4_1)
cannam@154: #define OVERRIDE_CELT_INNER_PROD
cannam@154: #define celt_inner_prod(x, y, N, arch) \
cannam@154:     ((void)arch, celt_inner_prod_sse2(x, y, N))
cannam@154: 
cannam@154: #elif defined(OPUS_X86_PRESUME_SSE) && !defined(FIXED_POINT)
cannam@154: #define OVERRIDE_CELT_INNER_PROD
cannam@154: #define celt_inner_prod(x, y, N, arch) \
cannam@154:     ((void)arch, celt_inner_prod_sse(x, y, N))
cannam@154: 
cannam@154: 
cannam@154: #elif ((defined(OPUS_X86_MAY_HAVE_SSE4_1) || defined(OPUS_X86_MAY_HAVE_SSE2)) && defined(FIXED_POINT)) || \
cannam@154:     (defined(OPUS_X86_MAY_HAVE_SSE) && !defined(FIXED_POINT))
cannam@154: 
cannam@154: extern opus_val32 (*const CELT_INNER_PROD_IMPL[OPUS_ARCHMASK + 1])(
cannam@154:                     const opus_val16 *x,
cannam@154:                     const opus_val16 *y,
cannam@154:                     int               N);
cannam@154: 
cannam@154: #define OVERRIDE_CELT_INNER_PROD
cannam@154: #define celt_inner_prod(x, y, N, arch) \
cannam@154:     ((*CELT_INNER_PROD_IMPL[(arch) & OPUS_ARCHMASK])(x, y, N))
cannam@154: 
cannam@154: #endif
cannam@154: 
cannam@154: #if defined(OPUS_X86_MAY_HAVE_SSE) && !defined(FIXED_POINT)
cannam@154: 
cannam@154: #define OVERRIDE_DUAL_INNER_PROD
cannam@154: #define OVERRIDE_COMB_FILTER_CONST
cannam@154: 
cannam@154: #undef dual_inner_prod
cannam@154: #undef comb_filter_const
cannam@154: 
cannam@154: void dual_inner_prod_sse(const opus_val16 *x,
cannam@154:     const opus_val16 *y01,
cannam@154:     const opus_val16 *y02,
cannam@154:     int               N,
cannam@154:     opus_val32       *xy1,
cannam@154:     opus_val32       *xy2);
cannam@154: 
cannam@154: void comb_filter_const_sse(opus_val32 *y,
cannam@154:     opus_val32 *x,
cannam@154:     int         T,
cannam@154:     int         N,
cannam@154:     opus_val16  g10,
cannam@154:     opus_val16  g11,
cannam@154:     opus_val16  g12);
cannam@154: 
cannam@154: 
cannam@154: #if defined(OPUS_X86_PRESUME_SSE)
cannam@154: # define dual_inner_prod(x, y01, y02, N, xy1, xy2, arch) \
cannam@154:     ((void)(arch),dual_inner_prod_sse(x, y01, y02, N, xy1, xy2))
cannam@154: 
cannam@154: # define comb_filter_const(y, x, T, N, g10, g11, g12, arch) \
cannam@154:     ((void)(arch),comb_filter_const_sse(y, x, T, N, g10, g11, g12))
cannam@154: #else
cannam@154: 
cannam@154: extern void (*const DUAL_INNER_PROD_IMPL[OPUS_ARCHMASK + 1])(
cannam@154:               const opus_val16 *x,
cannam@154:               const opus_val16 *y01,
cannam@154:               const opus_val16 *y02,
cannam@154:               int               N,
cannam@154:               opus_val32       *xy1,
cannam@154:               opus_val32       *xy2);
cannam@154: 
cannam@154: #define dual_inner_prod(x, y01, y02, N, xy1, xy2, arch) \
cannam@154:     ((*DUAL_INNER_PROD_IMPL[(arch) & OPUS_ARCHMASK])(x, y01, y02, N, xy1, xy2))
cannam@154: 
cannam@154: extern void (*const COMB_FILTER_CONST_IMPL[OPUS_ARCHMASK + 1])(
cannam@154:               opus_val32 *y,
cannam@154:               opus_val32 *x,
cannam@154:               int         T,
cannam@154:               int         N,
cannam@154:               opus_val16  g10,
cannam@154:               opus_val16  g11,
cannam@154:               opus_val16  g12);
cannam@154: 
cannam@154: #define comb_filter_const(y, x, T, N, g10, g11, g12, arch) \
cannam@154:     ((*COMB_FILTER_CONST_IMPL[(arch) & OPUS_ARCHMASK])(y, x, T, N, g10, g11, g12))
cannam@154: 
cannam@154: #define NON_STATIC_COMB_FILTER_CONST_C
cannam@154: 
cannam@154: #endif
cannam@154: #endif
cannam@154: 
cannam@154: #endif