yading@10: /* yading@10: * Copyright (c) 2012 yading@10: * MIPS Technologies, Inc., California. yading@10: * yading@10: * Redistribution and use in source and binary forms, with or without yading@10: * modification, are permitted provided that the following conditions yading@10: * are met: yading@10: * 1. Redistributions of source code must retain the above copyright yading@10: * notice, this list of conditions and the following disclaimer. yading@10: * 2. Redistributions in binary form must reproduce the above copyright yading@10: * notice, this list of conditions and the following disclaimer in the yading@10: * documentation and/or other materials provided with the distribution. yading@10: * 3. Neither the name of the MIPS Technologies, Inc., nor the names of its yading@10: * contributors may be used to endorse or promote products derived from yading@10: * this software without specific prior written permission. yading@10: * yading@10: * THIS SOFTWARE IS PROVIDED BY THE MIPS TECHNOLOGIES, INC. ``AS IS'' AND yading@10: * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE yading@10: * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE yading@10: * ARE DISCLAIMED. IN NO EVENT SHALL THE MIPS TECHNOLOGIES, INC. BE LIABLE yading@10: * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL yading@10: * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS yading@10: * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) yading@10: * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT yading@10: * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY yading@10: * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF yading@10: * SUCH DAMAGE. yading@10: * yading@10: * Author: Nedeljko Babic (nbabic@mips.com) yading@10: * yading@10: * This file is part of FFmpeg. yading@10: * yading@10: * FFmpeg is free software; you can redistribute it and/or yading@10: * modify it under the terms of the GNU Lesser General Public yading@10: * License as published by the Free Software Foundation; either yading@10: * version 2.1 of the License, or (at your option) any later version. yading@10: * yading@10: * FFmpeg is distributed in the hope that it will be useful, yading@10: * but WITHOUT ANY WARRANTY; without even the implied warranty of yading@10: * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU yading@10: * Lesser General Public License for more details. yading@10: * yading@10: * You should have received a copy of the GNU Lesser General Public yading@10: * License along with FFmpeg; if not, write to the Free Software yading@10: * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA yading@10: */ yading@10: yading@10: /** yading@10: * @file yading@10: * Reference: libavcodec/amrwbdec.c yading@10: */ yading@10: #include "libavutil/avutil.h" yading@10: #include "libavcodec/amrwbdata.h" yading@10: #include "amrwbdec_mips.h" yading@10: yading@10: #if HAVE_INLINE_ASM yading@10: void hb_fir_filter_mips(float *out, const float fir_coef[HB_FIR_SIZE + 1], yading@10: float mem[HB_FIR_SIZE], const float *in) yading@10: { yading@10: int i; yading@10: float data[AMRWB_SFR_SIZE_16k + HB_FIR_SIZE]; // past and current samples yading@10: yading@10: memcpy(data, mem, HB_FIR_SIZE * sizeof(float)); yading@10: memcpy(data + HB_FIR_SIZE, in, AMRWB_SFR_SIZE_16k * sizeof(float)); yading@10: yading@10: for (i = 0; i < AMRWB_SFR_SIZE_16k; i++) { yading@10: float output; yading@10: float * p_data = (data+i); yading@10: yading@10: /** yading@10: * inner loop is entirely unrolled and instructions are scheduled yading@10: * to minimize pipeline stall yading@10: */ yading@10: __asm__ volatile( yading@10: "mtc1 $zero, %[output] \n\t" yading@10: "lwc1 $f0, 0(%[p_data]) \n\t" yading@10: "lwc1 $f1, 0(%[fir_coef]) \n\t" yading@10: "lwc1 $f2, 4(%[p_data]) \n\t" yading@10: "madd.s %[output], %[output], $f0, $f1 \n\t" yading@10: "lwc1 $f3, 4(%[fir_coef]) \n\t" yading@10: "lwc1 $f4, 8(%[p_data]) \n\t" yading@10: "madd.s %[output], %[output], $f2, $f3 \n\t" yading@10: "lwc1 $f5, 8(%[fir_coef]) \n\t" yading@10: yading@10: "lwc1 $f0, 12(%[p_data]) \n\t" yading@10: "lwc1 $f1, 12(%[fir_coef]) \n\t" yading@10: "madd.s %[output], %[output], $f4, $f5 \n\t" yading@10: "lwc1 $f2, 16(%[p_data]) \n\t" yading@10: "madd.s %[output], %[output], $f0, $f1 \n\t" yading@10: "lwc1 $f3, 16(%[fir_coef]) \n\t" yading@10: "lwc1 $f4, 20(%[p_data]) \n\t" yading@10: "lwc1 $f5, 20(%[fir_coef]) \n\t" yading@10: "madd.s %[output], %[output], $f2, $f3 \n\t" yading@10: yading@10: "lwc1 $f0, 24(%[p_data]) \n\t" yading@10: "lwc1 $f1, 24(%[fir_coef]) \n\t" yading@10: "lwc1 $f2, 28(%[p_data]) \n\t" yading@10: "madd.s %[output], %[output], $f4, $f5 \n\t" yading@10: "lwc1 $f3, 28(%[fir_coef]) \n\t" yading@10: "madd.s %[output], %[output], $f0, $f1 \n\t" yading@10: "lwc1 $f4, 32(%[p_data]) \n\t" yading@10: "madd.s %[output], %[output], $f2, $f3 \n\t" yading@10: "lwc1 $f5, 32(%[fir_coef]) \n\t" yading@10: "madd.s %[output], %[output], $f4, $f5 \n\t" yading@10: yading@10: "lwc1 $f0, 36(%[p_data]) \n\t" yading@10: "lwc1 $f1, 36(%[fir_coef]) \n\t" yading@10: "lwc1 $f2, 40(%[p_data]) \n\t" yading@10: "lwc1 $f3, 40(%[fir_coef]) \n\t" yading@10: "madd.s %[output], %[output], $f0, $f1 \n\t" yading@10: "lwc1 $f4, 44(%[p_data]) \n\t" yading@10: "lwc1 $f5, 44(%[fir_coef]) \n\t" yading@10: "madd.s %[output], %[output], $f2, $f3 \n\t" yading@10: yading@10: "lwc1 $f0, 48(%[p_data]) \n\t" yading@10: "lwc1 $f1, 48(%[fir_coef]) \n\t" yading@10: "lwc1 $f2, 52(%[p_data]) \n\t" yading@10: "madd.s %[output], %[output], $f4, $f5 \n\t" yading@10: "lwc1 $f3, 52(%[fir_coef]) \n\t" yading@10: "lwc1 $f4, 56(%[p_data]) \n\t" yading@10: "madd.s %[output], %[output], $f0, $f1 \n\t" yading@10: "lwc1 $f5, 56(%[fir_coef]) \n\t" yading@10: "madd.s %[output], %[output], $f2, $f3 \n\t" yading@10: yading@10: "lwc1 $f0, 60(%[p_data]) \n\t" yading@10: "lwc1 $f1, 60(%[fir_coef]) \n\t" yading@10: "lwc1 $f2, 64(%[p_data]) \n\t" yading@10: "madd.s %[output], %[output], $f4, $f5 \n\t" yading@10: "lwc1 $f3, 64(%[fir_coef]) \n\t" yading@10: "madd.s %[output], %[output], $f0, $f1 \n\t" yading@10: "lwc1 $f4, 68(%[p_data]) \n\t" yading@10: "madd.s %[output], %[output], $f2, $f3 \n\t" yading@10: "lwc1 $f5, 68(%[fir_coef]) \n\t" yading@10: "madd.s %[output], %[output], $f4, $f5 \n\t" yading@10: yading@10: "lwc1 $f0, 72(%[p_data]) \n\t" yading@10: "lwc1 $f1, 72(%[fir_coef]) \n\t" yading@10: "lwc1 $f2, 76(%[p_data]) \n\t" yading@10: "lwc1 $f3, 76(%[fir_coef]) \n\t" yading@10: "madd.s %[output], %[output], $f0, $f1 \n\t" yading@10: "lwc1 $f4, 80(%[p_data]) \n\t" yading@10: "lwc1 $f5, 80(%[fir_coef]) \n\t" yading@10: "madd.s %[output], %[output], $f2, $f3 \n\t" yading@10: yading@10: "lwc1 $f0, 84(%[p_data]) \n\t" yading@10: "lwc1 $f1, 84(%[fir_coef]) \n\t" yading@10: "lwc1 $f2, 88(%[p_data]) \n\t" yading@10: "madd.s %[output], %[output], $f4, $f5 \n\t" yading@10: "lwc1 $f3, 88(%[fir_coef]) \n\t" yading@10: "lwc1 $f4, 92(%[p_data]) \n\t" yading@10: "madd.s %[output], %[output], $f0, $f1 \n\t" yading@10: "lwc1 $f5, 92(%[fir_coef]) \n\t" yading@10: "madd.s %[output], %[output], $f2, $f3 \n\t" yading@10: yading@10: "lwc1 $f0, 96(%[p_data]) \n\t" yading@10: "lwc1 $f1, 96(%[fir_coef]) \n\t" yading@10: "lwc1 $f2, 100(%[p_data]) \n\t" yading@10: "madd.s %[output], %[output], $f4, $f5 \n\t" yading@10: "lwc1 $f3, 100(%[fir_coef]) \n\t" yading@10: "lwc1 $f4, 104(%[p_data]) \n\t" yading@10: "madd.s %[output], %[output], $f0, $f1 \n\t" yading@10: "lwc1 $f5, 104(%[fir_coef]) \n\t" yading@10: "madd.s %[output], %[output], $f2, $f3 \n\t" yading@10: yading@10: "lwc1 $f0, 108(%[p_data]) \n\t" yading@10: "lwc1 $f1, 108(%[fir_coef]) \n\t" yading@10: "madd.s %[output], %[output], $f4, $f5 \n\t" yading@10: "lwc1 $f2, 112(%[p_data]) \n\t" yading@10: "lwc1 $f3, 112(%[fir_coef]) \n\t" yading@10: "madd.s %[output], %[output], $f0, $f1 \n\t" yading@10: "lwc1 $f4, 116(%[p_data]) \n\t" yading@10: "lwc1 $f5, 116(%[fir_coef]) \n\t" yading@10: "lwc1 $f0, 120(%[p_data]) \n\t" yading@10: "madd.s %[output], %[output], $f2, $f3 \n\t" yading@10: "lwc1 $f1, 120(%[fir_coef]) \n\t" yading@10: "madd.s %[output], %[output], $f4, $f5 \n\t" yading@10: "madd.s %[output], %[output], $f0, $f1 \n\t" yading@10: yading@10: : [output]"=&f"(output) yading@10: : [fir_coef]"r"(fir_coef), [p_data]"r"(p_data) yading@10: : "$f0", "$f1", "$f2", "$f3", "$f4", "$f5", "memory" yading@10: ); yading@10: out[i] = output; yading@10: } yading@10: memcpy(mem, data + AMRWB_SFR_SIZE_16k, HB_FIR_SIZE * sizeof(float)); yading@10: } yading@10: #endif /* HAVE_INLINE_ASM */