diff options
Diffstat (limited to 'media/libstagefright/codecs/amrnb/common/include/basic_op_c_equivalent.h')
-rw-r--r-- | media/libstagefright/codecs/amrnb/common/include/basic_op_c_equivalent.h | 505 |
1 files changed, 505 insertions, 0 deletions
diff --git a/media/libstagefright/codecs/amrnb/common/include/basic_op_c_equivalent.h b/media/libstagefright/codecs/amrnb/common/include/basic_op_c_equivalent.h new file mode 100644 index 0000000..35638e3 --- /dev/null +++ b/media/libstagefright/codecs/amrnb/common/include/basic_op_c_equivalent.h @@ -0,0 +1,505 @@ +/* ------------------------------------------------------------------ + * Copyright (C) 1998-2009 PacketVideo + * + * Licensed under the Apache License, Version 2.0 (the "License"); + * you may not use this file except in compliance with the License. + * You may obtain a copy of the License at + * + * http://www.apache.org/licenses/LICENSE-2.0 + * + * Unless required by applicable law or agreed to in writing, software + * distributed under the License is distributed on an "AS IS" BASIS, + * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either + * express or implied. + * See the License for the specific language governing permissions + * and limitations under the License. + * ------------------------------------------------------------------- + */ +/**************************************************************************************** +Portions of this file are derived from the following 3GPP standard: + + 3GPP TS 26.073 + ANSI-C code for the Adaptive Multi-Rate (AMR) speech codec + Available from http://www.3gpp.org + +(C) 2004, 3GPP Organizational Partners (ARIB, ATIS, CCSA, ETSI, TTA, TTC) +Permission to distribute, modify and use this file under the standard license +terms listed above has been obtained from the copyright holder. +****************************************************************************************/ +/* + + Pathname: ./include/basic_op_c_equivalent.h + +------------------------------------------------------------------------------ + REVISION HISTORY + + Who: Date: + Description: + +------------------------------------------------------------------------------ + INCLUDE DESCRIPTION + + This file includes all the C-Equivalent basicop.c functions. + +------------------------------------------------------------------------------ +*/ + +/*---------------------------------------------------------------------------- +; CONTINUE ONLY IF NOT ALREADY DEFINED +----------------------------------------------------------------------------*/ +#ifndef BASIC_OP_C_EQUIVALENT_H +#define BASIC_OP_C_EQUIVALENT_H + +/*---------------------------------------------------------------------------- +; INCLUDES +----------------------------------------------------------------------------*/ +#include "basicop_malloc.h" + +/*--------------------------------------------------------------------------*/ +#ifdef __cplusplus +extern "C" +{ +#endif + + /*---------------------------------------------------------------------------- + ; MACROS + ; Define module specific macros here + ----------------------------------------------------------------------------*/ + + /*---------------------------------------------------------------------------- + ; DEFINES + ; Include all pre-processor statements here. + ----------------------------------------------------------------------------*/ + + /*---------------------------------------------------------------------------- + ; EXTERNAL VARIABLES REFERENCES + ; Declare variables used in this module but defined elsewhere + ----------------------------------------------------------------------------*/ + + /*---------------------------------------------------------------------------- + ; SIMPLE TYPEDEF'S + ----------------------------------------------------------------------------*/ + + /*---------------------------------------------------------------------------- + ; ENUMERATED TYPEDEF'S + ----------------------------------------------------------------------------*/ + + /*---------------------------------------------------------------------------- + ; STRUCTURES TYPEDEF'S + ----------------------------------------------------------------------------*/ + + /*---------------------------------------------------------------------------- + ; GLOBAL FUNCTION DEFINITIONS + ; Function Prototype declaration + ----------------------------------------------------------------------------*/ + + + /* + ------------------------------------------------------------------------------ + FUNCTION NAME: L_add + ------------------------------------------------------------------------------ + INPUT AND OUTPUT DEFINITIONS + + Inputs: + L_var1 = 32 bit long signed integer (Word32) whose value falls + in the range : 0x8000 0000 <= L_var1 <= 0x7fff ffff. + + L_var2 = 32 bit long signed integer (Word32) whose value falls + in the range : 0x8000 0000 <= L_var1 <= 0x7fff ffff. + + pOverflow = pointer to overflow (Flag) + + Outputs: + pOverflow -> 1 if the 32 bit add operation resulted in overflow + + Returns: + L_sum = 32-bit sum of L_var1 and L_var2 (Word32) + */ + static inline Word32 L_add(register Word32 L_var1, register Word32 L_var2, Flag *pOverflow) + { + Word32 L_sum; + + L_sum = L_var1 + L_var2; + + if ((L_var1 ^ L_var2) >= 0) + { + if ((L_sum ^ L_var1) < 0) + { + L_sum = (L_var1 < 0) ? MIN_32 : MAX_32; + *pOverflow = 1; + } + } + + return (L_sum); + } + + /* + ------------------------------------------------------------------------------ + FUNCTION NAME: L_sub + ------------------------------------------------------------------------------ + INPUT AND OUTPUT DEFINITIONS + + Inputs: + L_var1 = 32 bit long signed integer (Word32) whose value falls + in the range : 0x8000 0000 <= L_var1 <= 0x7fff ffff. + + L_var2 = 32 bit long signed integer (Word32) whose value falls + in the range : 0x8000 0000 <= L_var1 <= 0x7fff ffff. + + pOverflow = pointer to overflow (Flag) + + Outputs: + pOverflow -> 1 if the 32 bit add operation resulted in overflow + + Returns: + L_diff = 32-bit difference of L_var1 and L_var2 (Word32) + */ + static inline Word32 L_sub(register Word32 L_var1, register Word32 L_var2, + register Flag *pOverflow) + { + Word32 L_diff; + + L_diff = L_var1 - L_var2; + + if ((L_var1 ^ L_var2) < 0) + { + if ((L_diff ^ L_var1) & MIN_32) + { + L_diff = (L_var1 < 0L) ? MIN_32 : MAX_32; + *pOverflow = 1; + } + } + + return (L_diff); + } + + + /* + ------------------------------------------------------------------------------ + FUNCTION NAME: L_mac + ------------------------------------------------------------------------------ + INPUT AND OUTPUT DEFINITIONS + + Inputs: + L_var3 = 32 bit long signed integer (Word32) whose value falls + in the range : 0x8000 0000 <= L_var3 <= 0x7fff ffff. + var1 = 16 bit short signed integer (Word16) whose value falls in + the range : 0xffff 8000 <= var1 <= 0x0000 7fff. + var2 = 16 bit short signed integer (Word16) whose value falls in + the range : 0xffff 8000 <= var2 <= 0x0000 7fff. + + pOverflow = pointer to overflow (Flag) + + Outputs: + pOverflow -> 1 if the 32 bit add operation resulted in overflow + + Returns: + result = 32-bit result of L_var3 + (var1 * var2)(Word32) + */ + __inline Word32 L_mac(Word32 L_var3, Word16 var1, Word16 var2, Flag *pOverflow) + { + Word32 result; + Word32 L_sum; + result = (Word32) var1 * var2; + if (result != (Word32) 0x40000000L) + { + L_sum = (result << 1) + L_var3; + + /* Check if L_sum and L_var_3 share the same sign */ + if ((L_var3 ^ result) > 0) + { + if ((L_sum ^ L_var3) < 0) + { + L_sum = (L_var3 < 0) ? MIN_32 : MAX_32; + *pOverflow = 1; + } + } + } + else + { + *pOverflow = 1; + L_sum = MAX_32; + } + return (L_sum); + } + + /* + ------------------------------------------------------------------------------ + FUNCTION NAME: L_mult + ------------------------------------------------------------------------------ + INPUT AND OUTPUT DEFINITIONS + + Inputs: + L_var1 = 16 bit short signed integer (Word16) whose value falls in + the range : 0xffff 8000 <= var1 <= 0x0000 7fff. + + L_var2 = 16 bit short signed integer (Word16) whose value falls in + the range : 0xffff 8000 <= var1 <= 0x0000 7fff. + + pOverflow = pointer to overflow (Flag) + + Outputs: + pOverflow -> 1 if the 32 bit add operation resulted in overflow + + Returns: + L_product = 32-bit product of L_var1 and L_var2 (Word32) + */ + static inline Word32 L_mult(Word16 var1, Word16 var2, Flag *pOverflow) + { + register Word32 L_product; + + L_product = (Word32) var1 * var2; + + if (L_product != (Word32) 0x40000000L) + { + L_product <<= 1; /* Multiply by 2 */ + } + else + { + *pOverflow = 1; + L_product = MAX_32; + } + + return (L_product); + } + + + /* + ------------------------------------------------------------------------------ + FUNCTION NAME: L_msu + ------------------------------------------------------------------------------ + INPUT AND OUTPUT DEFINITIONS + + Inputs: + L_var3 = 32 bit long signed integer (Word32) whose value falls + in the range : 0x8000 0000 <= L_var3 <= 0x7fff ffff. + + var1 = 16 bit short signed integer (Word16) whose value falls in + the range : 0xffff 8000 <= var1 <= 0x0000 7fff. + var2 = 16 bit short signed integer (Word16) whose value falls in + the range : 0xffff 8000 <= var2 <= 0x0000 7fff. + + pOverflow = pointer to overflow (Flag) + + Outputs: + pOverflow -> 1 if the 32 bit operation resulted in overflow + + Returns: + result = 32-bit result of L_var3 - (var1 * var2) + */ + + static inline Word32 L_msu(Word32 L_var3, Word16 var1, Word16 var2, Flag *pOverflow) + { + Word32 result; + + result = L_mult(var1, var2, pOverflow); + result = L_sub(L_var3, result, pOverflow); + + return (result); + } + + /* + ------------------------------------------------------------------------------ + FUNCTION NAME: Mpy_32 + ------------------------------------------------------------------------------ + INPUT AND OUTPUT DEFINITIONS + + Inputs: + L_var1_hi = most significant word of first input (Word16). + L_var1_lo = least significant word of first input (Word16). + L_var2_hi = most significant word of second input (Word16). + L_var2_lo = least significant word of second input (Word16). + + pOverflow = pointer to overflow (Flag) + + Outputs: + pOverflow -> 1 if the 32 bit multiply operation resulted in overflow + + Returns: + L_product = 32-bit product of L_var1 and L_var2 (Word32) + */ + __inline Word32 Mpy_32(Word16 L_var1_hi, + Word16 L_var1_lo, + Word16 L_var2_hi, + Word16 L_var2_lo, + Flag *pOverflow) + { + Word32 L_product; + Word32 L_sum; + Word32 product32; + + OSCL_UNUSED_ARG(pOverflow); + L_product = (Word32) L_var1_hi * L_var2_hi; + + if (L_product != (Word32) 0x40000000L) + { + L_product <<= 1; + } + else + { + L_product = MAX_32; + } + + /* result = mult (L_var1_hi, L_var2_lo, pOverflow); */ + product32 = ((Word32) L_var1_hi * L_var2_lo) >> 15; + + /* L_product = L_mac (L_product, result, 1, pOverflow); */ + L_sum = L_product + (product32 << 1); + + if ((L_product ^ product32) > 0) + { + if ((L_sum ^ L_product) < 0) + { + L_sum = (L_product < 0) ? MIN_32 : MAX_32; + } + } + + L_product = L_sum; + + /* result = mult (L_var1_lo, L_var2_hi, pOverflow); */ + product32 = ((Word32) L_var1_lo * L_var2_hi) >> 15; + + /* L_product = L_mac (L_product, result, 1, pOverflow); */ + L_sum = L_product + (product32 << 1); + + if ((L_product ^ product32) > 0) + { + if ((L_sum ^ L_product) < 0) + { + L_sum = (L_product < 0) ? MIN_32 : MAX_32; + } + } + return (L_sum); + } + + /* + ------------------------------------------------------------------------------ + FUNCTION NAME: Mpy_32_16 + ------------------------------------------------------------------------------ + INPUT AND OUTPUT DEFINITIONS + + Inputs: + L_var1_hi = most significant 16 bits of 32-bit input (Word16). + L_var1_lo = least significant 16 bits of 32-bit input (Word16). + var2 = 16-bit signed integer (Word16). + + pOverflow = pointer to overflow (Flag) + + Outputs: + pOverflow -> 1 if the 32 bit product operation resulted in overflow + + Returns: + product = 32-bit product of the 32-bit L_var1 and 16-bit var1 (Word32) + */ + + __inline Word32 Mpy_32_16(Word16 L_var1_hi, + Word16 L_var1_lo, + Word16 var2, + Flag *pOverflow) + { + + Word32 L_product; + Word32 L_sum; + Word32 result; + L_product = (Word32) L_var1_hi * var2; + + if (L_product != (Word32) 0x40000000L) + { + L_product <<= 1; + } + else + { + *pOverflow = 1; + L_product = MAX_32; + } + + result = ((Word32)L_var1_lo * var2) >> 15; + + L_sum = L_product + (result << 1); + + if ((L_product ^ result) > 0) + { + if ((L_sum ^ L_product) < 0) + { + L_sum = (L_product < 0) ? MIN_32 : MAX_32; + *pOverflow = 1; + } + } + return (L_sum); + + } + + /* + ------------------------------------------------------------------------------ + FUNCTION NAME: mult + ------------------------------------------------------------------------------ + INPUT AND OUTPUT DEFINITIONS + + Inputs: + var1 = 16 bit short signed integer (Word16) whose value falls in + the range : 0xffff 8000 <= var1 <= 0x0000 7fff. + + var2 = 16 bit short signed integer (Word16) whose value falls in + the range : 0xffff 8000 <= var2 <= 0x0000 7fff. + + pOverflow = pointer to overflow (Flag) + + Outputs: + pOverflow -> 1 if the add operation resulted in overflow + + Returns: + product = 16-bit limited product of var1 and var2 (Word16) + */ + static inline Word16 mult(Word16 var1, Word16 var2, Flag *pOverflow) + { + register Word32 product; + + product = ((Word32) var1 * var2) >> 15; + + /* Saturate result (if necessary). */ + /* var1 * var2 >0x00007fff is the only case */ + /* that saturation occurs. */ + + if (product > 0x00007fffL) + { + *pOverflow = 1; + product = (Word32) MAX_16; + } + + + /* Return the product as a 16 bit value by type casting Word32 to Word16 */ + + return ((Word16) product); + } + + + static inline Word32 amrnb_fxp_mac_16_by_16bb(Word32 L_var1, Word32 L_var2, Word32 L_var3) + { + Word32 result; + + result = L_var3 + L_var1 * L_var2; + + return result; + } + + static inline Word32 amrnb_fxp_msu_16_by_16bb(Word32 L_var1, Word32 L_var2, Word32 L_var3) + { + Word32 result; + + result = L_var3 - L_var1 * L_var2; + + return result; + } + + + /*---------------------------------------------------------------------------- + ; END + ----------------------------------------------------------------------------*/ +#ifdef __cplusplus +} +#endif + +#endif /* BASIC_OP_C_EQUIVALENT_H */ + + + |