| 1 | /* |
| 2 | * Copyright (c) 2010 Mans Rullgard <mans@mansr.com> |
| 3 | * |
| 4 | * This file is part of FFmpeg. |
| 5 | * |
| 6 | * FFmpeg is free software; you can redistribute it and/or |
| 7 | * modify it under the terms of the GNU Lesser General Public |
| 8 | * License as published by the Free Software Foundation; either |
| 9 | * version 2.1 of the License, or (at your option) any later version. |
| 10 | * |
| 11 | * FFmpeg is distributed in the hope that it will be useful, |
| 12 | * but WITHOUT ANY WARRANTY; without even the implied warranty of |
| 13 | * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU |
| 14 | * Lesser General Public License for more details. |
| 15 | * |
| 16 | * You should have received a copy of the GNU Lesser General Public |
| 17 | * License along with FFmpeg; if not, write to the Free Software |
| 18 | * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA |
| 19 | */ |
| 20 | |
| 21 | #ifndef AVCODEC_ARM_AAC_H |
| 22 | #define AVCODEC_ARM_AAC_H |
| 23 | |
| 24 | #include "config.h" |
| 25 | |
| 26 | #if HAVE_NEON_INLINE |
| 27 | |
| 28 | #define VMUL2 VMUL2 |
| 29 | static inline float *VMUL2(float *dst, const float *v, unsigned idx, |
| 30 | const float *scale) |
| 31 | { |
| 32 | unsigned v0, v1; |
| 33 | __asm__ ("ubfx %0, %6, #0, #4 \n\t" |
| 34 | "ubfx %1, %6, #4, #4 \n\t" |
| 35 | "ldr %0, [%5, %0, lsl #2] \n\t" |
| 36 | "ldr %1, [%5, %1, lsl #2] \n\t" |
| 37 | "vld1.32 {d1[]}, [%7,:32] \n\t" |
| 38 | "vmov d0, %0, %1 \n\t" |
| 39 | "vmul.f32 d0, d0, d1 \n\t" |
| 40 | "vst1.32 {d0}, [%2,:64]! \n\t" |
| 41 | : "=&r"(v0), "=&r"(v1), "+r"(dst), "=m"(dst[0]), "=m"(dst[1]) |
| 42 | : "r"(v), "r"(idx), "r"(scale) |
| 43 | : "d0", "d1"); |
| 44 | return dst; |
| 45 | } |
| 46 | |
| 47 | #define VMUL4 VMUL4 |
| 48 | static inline float *VMUL4(float *dst, const float *v, unsigned idx, |
| 49 | const float *scale) |
| 50 | { |
| 51 | unsigned v0, v1, v2, v3; |
| 52 | __asm__ ("ubfx %0, %10, #0, #2 \n\t" |
| 53 | "ubfx %1, %10, #2, #2 \n\t" |
| 54 | "ldr %0, [%9, %0, lsl #2] \n\t" |
| 55 | "ubfx %2, %10, #4, #2 \n\t" |
| 56 | "ldr %1, [%9, %1, lsl #2] \n\t" |
| 57 | "ubfx %3, %10, #6, #2 \n\t" |
| 58 | "ldr %2, [%9, %2, lsl #2] \n\t" |
| 59 | "vmov d0, %0, %1 \n\t" |
| 60 | "ldr %3, [%9, %3, lsl #2] \n\t" |
| 61 | "vld1.32 {d2[],d3[]},[%11,:32] \n\t" |
| 62 | "vmov d1, %2, %3 \n\t" |
| 63 | "vmul.f32 q0, q0, q1 \n\t" |
| 64 | "vst1.32 {q0}, [%4,:128]! \n\t" |
| 65 | : "=&r"(v0), "=&r"(v1), "=&r"(v2), "=&r"(v3), "+r"(dst), |
| 66 | "=m"(dst[0]), "=m"(dst[1]), "=m"(dst[2]), "=m"(dst[3]) |
| 67 | : "r"(v), "r"(idx), "r"(scale) |
| 68 | : "d0", "d1", "d2", "d3"); |
| 69 | return dst; |
| 70 | } |
| 71 | |
| 72 | #define VMUL2S VMUL2S |
| 73 | static inline float *VMUL2S(float *dst, const float *v, unsigned idx, |
| 74 | unsigned sign, const float *scale) |
| 75 | { |
| 76 | unsigned v0, v1, v2, v3; |
| 77 | __asm__ ("ubfx %0, %8, #0, #4 \n\t" |
| 78 | "ubfx %1, %8, #4, #4 \n\t" |
| 79 | "ldr %0, [%7, %0, lsl #2] \n\t" |
| 80 | "lsl %2, %10, #30 \n\t" |
| 81 | "ldr %1, [%7, %1, lsl #2] \n\t" |
| 82 | "lsl %3, %10, #31 \n\t" |
| 83 | "vmov d0, %0, %1 \n\t" |
| 84 | "bic %2, %2, #1<<30 \n\t" |
| 85 | "vld1.32 {d1[]}, [%9,:32] \n\t" |
| 86 | "vmov d2, %2, %3 \n\t" |
| 87 | "veor d0, d0, d2 \n\t" |
| 88 | "vmul.f32 d0, d0, d1 \n\t" |
| 89 | "vst1.32 {d0}, [%4,:64]! \n\t" |
| 90 | : "=&r"(v0), "=&r"(v1), "=&r"(v2), "=&r"(v3), "+r"(dst), |
| 91 | "=m"(dst[0]), "=m"(dst[1]) |
| 92 | : "r"(v), "r"(idx), "r"(scale), "r"(sign) |
| 93 | : "d0", "d1", "d2"); |
| 94 | return dst; |
| 95 | } |
| 96 | |
| 97 | #define VMUL4S VMUL4S |
| 98 | static inline float *VMUL4S(float *dst, const float *v, unsigned idx, |
| 99 | unsigned sign, const float *scale) |
| 100 | { |
| 101 | unsigned v0, v1, v2, v3, nz; |
| 102 | __asm__ ("vld1.32 {d2[],d3[]},[%13,:32] \n\t" |
| 103 | "ubfx %0, %12, #0, #2 \n\t" |
| 104 | "ubfx %1, %12, #2, #2 \n\t" |
| 105 | "ldr %0, [%11,%0, lsl #2] \n\t" |
| 106 | "ubfx %2, %12, #4, #2 \n\t" |
| 107 | "ldr %1, [%11,%1, lsl #2] \n\t" |
| 108 | "ubfx %3, %12, #6, #2 \n\t" |
| 109 | "ldr %2, [%11,%2, lsl #2] \n\t" |
| 110 | "vmov d0, %0, %1 \n\t" |
| 111 | "ldr %3, [%11,%3, lsl #2] \n\t" |
| 112 | "lsr %6, %12, #12 \n\t" |
| 113 | "rbit %6, %6 \n\t" |
| 114 | "vmov d1, %2, %3 \n\t" |
| 115 | "lsls %6, %6, #1 \n\t" |
| 116 | "and %0, %5, #1<<31 \n\t" |
| 117 | "it cs \n\t" |
| 118 | "lslcs %5, %5, #1 \n\t" |
| 119 | "lsls %6, %6, #1 \n\t" |
| 120 | "and %1, %5, #1<<31 \n\t" |
| 121 | "it cs \n\t" |
| 122 | "lslcs %5, %5, #1 \n\t" |
| 123 | "lsls %6, %6, #1 \n\t" |
| 124 | "and %2, %5, #1<<31 \n\t" |
| 125 | "it cs \n\t" |
| 126 | "lslcs %5, %5, #1 \n\t" |
| 127 | "vmov d4, %0, %1 \n\t" |
| 128 | "and %3, %5, #1<<31 \n\t" |
| 129 | "vmov d5, %2, %3 \n\t" |
| 130 | "veor q0, q0, q2 \n\t" |
| 131 | "vmul.f32 q0, q0, q1 \n\t" |
| 132 | "vst1.32 {q0}, [%4,:128]! \n\t" |
| 133 | : "=&r"(v0), "=&r"(v1), "=&r"(v2), "=&r"(v3), "+r"(dst), |
| 134 | "+r"(sign), "=r"(nz), |
| 135 | "=m"(dst[0]), "=m"(dst[1]), "=m"(dst[2]), "=m"(dst[3]) |
| 136 | : "r"(v), "r"(idx), "r"(scale) |
| 137 | : "cc", "d0", "d1", "d2", "d3", "d4", "d5"); |
| 138 | return dst; |
| 139 | } |
| 140 | |
| 141 | #endif /* HAVE_NEON_INLINE */ |
| 142 | |
| 143 | #endif /* AVCODEC_ARM_AAC_H */ |