From 6cfabd35363c3ef5e3b209b867169a500b3ccc3c Mon Sep 17 00:00:00 2001 From: Fraunhofer IIS FDK Date: Mon, 26 Feb 2018 20:17:00 +0100 Subject: Upgrade to FDKv2 Bug: 71430241 Test: CTS DecoderTest and DecoderTestAacDrc original-Change-Id: Iaa20f749b8a04d553b20247cfe1a8930ebbabe30 Apply clang-format also on header files. original-Change-Id: I14de1ef16bbc79ec0283e745f98356a10efeb2e4 Fixes for MPEG-D DRC original-Change-Id: If1de2d74bbbac84b3f67de3b88b83f6a23b8a15c Catch unsupported tw_mdct at an early stage original-Change-Id: Ied9dd00d754162a0e3ca1ae3e6b854315d818afe Fixing PVC transition frames original-Change-Id: Ib75725abe39252806c32d71176308f2c03547a4e Move qmf bands sanity check original-Change-Id: Iab540c3013c174d9490d2ae100a4576f51d8dbc4 Initialize scaling variable original-Change-Id: I3c4087101b70e998c71c1689b122b0d7762e0f9e Add 16 qmf band configuration to getSlotNrgHQ() original-Change-Id: I49a5d30f703a1b126ff163df9656db2540df21f1 Always apply byte alignment at the end of the AudioMuxElement original-Change-Id: I42d560287506d65d4c3de8bfe3eb9a4ebeb4efc7 Setup SBR element only if no parse error exists original-Change-Id: I1915b73704bc80ab882b9173d6bec59cbd073676 Additional array index check in HCR original-Change-Id: I18cc6e501ea683b5009f1bbee26de8ddd04d8267 Fix fade-in index selection in concealment module original-Change-Id: Ibf802ed6ed8c05e9257e1f3b6d0ac1162e9b81c1 Enable explicit backward compatible parser for AAC_LD original-Change-Id: I27e9c678dcb5d40ed760a6d1e06609563d02482d Skip spatial specific config in explicit backward compatible ASC original-Change-Id: Iff7cc365561319e886090cedf30533f562ea4d6e Update flags description in decoder API original-Change-Id: I9a5b4f8da76bb652f5580cbd3ba9760425c43830 Add QMF domain reset function original-Change-Id: I4f89a8a2c0277d18103380134e4ed86996e9d8d6 DRC upgrade v2.1.0 original-Change-Id: I5731c0540139dab220094cd978ef42099fc45b74 Fix integer overflow in sqrtFixp_lookup() original-Change-Id: I429a6f0d19aa2cc957e0f181066f0ca73968c914 Fix integer overflow in invSqrtNorm2() original-Change-Id: I84de5cbf9fb3adeb611db203fe492fabf4eb6155 Fix integer overflow in GenerateRandomVector() original-Change-Id: I3118a641008bd9484d479e5b0b1ee2b5d7d44d74 Fix integer overflow in adjustTimeSlot_EldGrid() original-Change-Id: I29d503c247c5c8282349b79df940416a512fb9d5 Fix integer overflow in FDKsbrEnc_codeEnvelope() original-Change-Id: I6b34b61ebb9d525b0c651ed08de2befc1f801449 Follow-up on: Fix integer overflow in adjustTimeSlot_EldGrid() original-Change-Id: I6f8f578cc7089e5eb7c7b93e580b72ca35ad689a Fix integer overflow in get_pk_v2() original-Change-Id: I63375bed40d45867f6eeaa72b20b1f33e815938c Fix integer overflow in Syn_filt_zero() original-Change-Id: Ie0c02fdfbe03988f9d3b20d10cd9fe4c002d1279 Fix integer overflow in CFac_CalcFacSignal() original-Change-Id: Id2d767c40066c591b51768e978eb8af3b803f0c5 Fix integer overflow in FDKaacEnc_FDKaacEnc_calcPeNoAH() original-Change-Id: Idcbd0f4a51ae2550ed106aa6f3d678d1f9724841 Fix integer overflow in sbrDecoder_calculateGainVec() original-Change-Id: I7081bcbe29c5cede9821b38d93de07c7add2d507 Fix integer overflow in CLpc_SynthesisLattice() original-Change-Id: I4a95ddc18de150102352d4a1845f06094764c881 Fix integer overflow in Pred_Lt4() original-Change-Id: I4dbd012b2de7d07c3e70a47b92e3bfae8dbc750a Fix integer overflow in FDKsbrEnc_InitSbrFastTransientDetector() original-Change-Id: I788cbec1a4a00f44c2f3a72ad7a4afa219807d04 Fix unsigned integer overflow in FDKaacEnc_WriteBitstream() original-Change-Id: I68fc75166e7d2cd5cd45b18dbe3d8c2a92f1822a Fix unsigned integer overflow in FDK_MetadataEnc_Init() original-Change-Id: Ie8d025f9bcdb2442c704bd196e61065c03c10af4 Fix overflow in pseudo random number generators original-Change-Id: I3e2551ee01356297ca14e3788436ede80bd5513c Fix unsigned integer overflow in sbrDecoder_Parse() original-Change-Id: I3f231b2f437e9c37db4d5b964164686710eee971 Fix unsigned integer overflow in longsub() original-Change-Id: I73c2bc50415cac26f1f5a29e125bbe75f9180a6e Fix unsigned integer overflow in CAacDecoder_DecodeFrame() original-Change-Id: Ifce2db4b1454b46fa5f887e9d383f1cc43b291e4 Fix overflow at CLpdChannelStream_Read() original-Change-Id: Idb9d822ce3a4272e4794b643644f5434e2d4bf3f Fix unsigned integer overflow in Hcr_State_BODY_SIGN_ESC__ESC_WORD() original-Change-Id: I1ccf77c0015684b85534c5eb97162740a870b71c Fix unsigned integer overflow in UsacConfig_Parse() original-Change-Id: Ie6d27f84b6ae7eef092ecbff4447941c77864d9f Fix unsigned integer overflow in aacDecoder_drcParse() original-Change-Id: I713f28e883eea3d70b6fa56a7b8f8c22bcf66ca0 Fix unsigned integer overflow in aacDecoder_drcReadCompression() original-Change-Id: Ia34dfeb88c4705c558bce34314f584965cafcf7a Fix unsigned integer overflow in CDataStreamElement_Read() original-Change-Id: Iae896cc1d11f0a893d21be6aa90bd3e60a2c25f0 Fix unsigned integer overflow in transportDec_AdjustEndOfAccessUnit() original-Change-Id: I64cf29a153ee784bb4a16fdc088baabebc0007dc Fix unsigned integer overflow in transportDec_GetAuBitsRemaining() original-Change-Id: I975b3420faa9c16a041874ba0db82e92035962e4 Fix unsigned integer overflow in extractExtendedData() original-Change-Id: I2a59eb09e2053cfb58dfb75fcecfad6b85a80a8f Fix signed integer overflow in CAacDecoder_ExtPayloadParse() original-Change-Id: I4ad5ca4e3b83b5d964f1c2f8c5e7b17c477c7929 Fix unsigned integer overflow in CAacDecoder_DecodeFrame() original-Change-Id: I29a39df77d45c52a0c9c5c83c1ba81f8d0f25090 Follow-up on: Fix integer overflow in CLpc_SynthesisLattice() original-Change-Id: I8fb194ffc073a3432a380845be71036a272d388f Fix signed integer overflow in _interpolateDrcGain() original-Change-Id: I879ec9ab14005069a7c47faf80e8bc6e03d22e60 Fix unsigned integer overflow in FDKreadBits() original-Change-Id: I1f47a6a8037ff70375aa8844947d5681bb4287ad Fix unsigned integer overflow in FDKbyteAlign() original-Change-Id: Id5f3a11a0c9e50fc6f76ed6c572dbd4e9f2af766 Fix unsigned integer overflow in FDK_get32() original-Change-Id: I9d33b8e97e3d38cbb80629cb859266ca0acdce96 Fix unsigned integer overflow in FDK_pushBack() original-Change-Id: Ic87f899bc8c6acf7a377a8ca7f3ba74c3a1e1c19 Fix unsigned integer overflow in FDK_pushForward() original-Change-Id: I3b754382f6776a34be1602e66694ede8e0b8effc Fix unsigned integer overflow in ReadPsData() original-Change-Id: I25361664ba8139e32bbbef2ca8c106a606ce9c37 Fix signed integer overflow in E_UTIL_residu() original-Change-Id: I8c3abd1f437ee869caa8fb5903ce7d3d641b6aad REVERT: Follow-up on: Integer overflow in CLpc_SynthesisLattice(). original-Change-Id: I3d340099acb0414795c8dfbe6362bc0a8f045f9b Follow-up on: Fix integer overflow in CLpc_SynthesisLattice() original-Change-Id: I4aedb8b3a187064e9f4d985175aa55bb99cc7590 Follow-up on: Fix unsigned integer overflow in aacDecoder_drcParse() original-Change-Id: I2aa2e13916213bf52a67e8b0518e7bf7e57fb37d Fix integer overflow in acelp original-Change-Id: Ie6390c136d84055f8b728aefbe4ebef6e029dc77 Fix unsigned integer overflow in aacDecoder_UpdateBitStreamCounters() original-Change-Id: I391ffd97ddb0b2c184cba76139bfb356a3b4d2e2 Adjust concealment default settings original-Change-Id: I6a95db935a327c47df348030bcceafcb29f54b21 Saturate estimatedStartPos original-Change-Id: I27be2085e0ae83ec9501409f65e003f6bcba1ab6 Negative shift exponent in _interpolateDrcGain() original-Change-Id: I18edb26b26d002aafd5e633d4914960f7a359c29 Negative shift exponent in calculateICC() original-Change-Id: I3dcd2ae98d2eb70ee0d59750863cbb2a6f4f8aba Too large shift exponent in FDK_put() original-Change-Id: Ib7d9aaa434d2d8de4a13b720ca0464b31ca9b671 Too large shift exponent in CalcInvLdData() original-Change-Id: I43e6e78d4cd12daeb1dcd5d82d1798bdc2550262 Member access within null pointer of type SBR_CHANNEL original-Change-Id: Idc5e4ea8997810376d2f36bbdf628923b135b097 Member access within null pointer of type CpePersistentData original-Change-Id: Ib6c91cb0d37882768e5baf63324e429589de0d9d Member access within null pointer FDKaacEnc_psyMain() original-Change-Id: I7729b7f4479970531d9dc823abff63ca52e01997 Member access within null pointer FDKaacEnc_GetPnsParam() original-Change-Id: I9aa3b9f3456ae2e0f7483dbd5b3dde95fc62da39 Member access within null pointer FDKsbrEnc_EnvEncodeFrame() original-Change-Id: I67936f90ea714e90b3e81bc0dd1472cc713eb23a Add HCR sanity check original-Change-Id: I6c1d9732ebcf6af12f50b7641400752f74be39f7 Fix memory issue for HBE edge case with 8:3 SBR original-Change-Id: I11ea58a61e69fbe8bf75034b640baee3011e63e9 Additional SBR parametrization sanity check for ELD original-Change-Id: Ie26026fbfe174c2c7b3691f6218b5ce63e322140 Add MPEG-D DRC channel layout check original-Change-Id: Iea70a74f171b227cce636a9eac4ba662777a2f72 Additional out-of-bounds checks in MPEG-D DRC original-Change-Id: Ife4a8c3452c6fde8a0a09e941154a39a769777d4 Change-Id: Ic63cb2f628720f54fe9b572b0cb528e2599c624e --- libFDK/src/arm/autocorr2nd.cpp | 90 ----- libFDK/src/arm/dct_arm.cpp | 678 ++++++++++++++++++++--------------- libFDK/src/arm/fft_rad2_arm.cpp | 452 ++++++++++++------------ libFDK/src/arm/qmf_arm.cpp | 766 ---------------------------------------- libFDK/src/arm/scale_arm.cpp | 155 ++++---- 5 files changed, 701 insertions(+), 1440 deletions(-) delete mode 100644 libFDK/src/arm/autocorr2nd.cpp delete mode 100644 libFDK/src/arm/qmf_arm.cpp (limited to 'libFDK/src/arm') diff --git a/libFDK/src/arm/autocorr2nd.cpp b/libFDK/src/arm/autocorr2nd.cpp deleted file mode 100644 index cc76e30..0000000 --- a/libFDK/src/arm/autocorr2nd.cpp +++ /dev/null @@ -1,90 +0,0 @@ - -/* ----------------------------------------------------------------------------------------------------------- -Software License for The Fraunhofer FDK AAC Codec Library for Android - -© Copyright 1995 - 2013 Fraunhofer-Gesellschaft zur Förderung der angewandten Forschung e.V. - All rights reserved. - - 1. INTRODUCTION -The Fraunhofer FDK AAC Codec Library for Android ("FDK AAC Codec") is software that implements -the MPEG Advanced Audio Coding ("AAC") encoding and decoding scheme for digital audio. -This FDK AAC Codec software is intended to be used on a wide variety of Android devices. - -AAC's HE-AAC and HE-AAC v2 versions are regarded as today's most efficient general perceptual -audio codecs. AAC-ELD is considered the best-performing full-bandwidth communications codec by -independent studies and is widely deployed. AAC has been standardized by ISO and IEC as part -of the MPEG specifications. - -Patent licenses for necessary patent claims for the FDK AAC Codec (including those of Fraunhofer) -may be obtained through Via Licensing (www.vialicensing.com) or through the respective patent owners -individually for the purpose of encoding or decoding bit streams in products that are compliant with -the ISO/IEC MPEG audio standards. Please note that most manufacturers of Android devices already license -these patent claims through Via Licensing or directly from the patent owners, and therefore FDK AAC Codec -software may already be covered under those patent licenses when it is used for those licensed purposes only. - -Commercially-licensed AAC software libraries, including floating-point versions with enhanced sound quality, -are also available from Fraunhofer. Users are encouraged to check the Fraunhofer website for additional -applications information and documentation. - -2. COPYRIGHT LICENSE - -Redistribution and use in source and binary forms, with or without modification, are permitted without -payment of copyright license fees provided that you satisfy the following conditions: - -You must retain the complete text of this software license in redistributions of the FDK AAC Codec or -your modifications thereto in source code form. - -You must retain the complete text of this software license in the documentation and/or other materials -provided with redistributions of the FDK AAC Codec or your modifications thereto in binary form. -You must make available free of charge copies of the complete source code of the FDK AAC Codec and your -modifications thereto to recipients of copies in binary form. - -The name of Fraunhofer may not be used to endorse or promote products derived from this library without -prior written permission. - -You may not charge copyright license fees for anyone to use, copy or distribute the FDK AAC Codec -software or your modifications thereto. - -Your modified versions of the FDK AAC Codec must carry prominent notices stating that you changed the software -and the date of any change. For modified versions of the FDK AAC Codec, the term -"Fraunhofer FDK AAC Codec Library for Android" must be replaced by the term -"Third-Party Modified Version of the Fraunhofer FDK AAC Codec Library for Android." - -3. NO PATENT LICENSE - -NO EXPRESS OR IMPLIED LICENSES TO ANY PATENT CLAIMS, including without limitation the patents of Fraunhofer, -ARE GRANTED BY THIS SOFTWARE LICENSE. Fraunhofer provides no warranty of patent non-infringement with -respect to this software. - -You may use this FDK AAC Codec software or modifications thereto only for purposes that are authorized -by appropriate patent licenses. - -4. DISCLAIMER - -This FDK AAC Codec software is provided by Fraunhofer on behalf of the copyright holders and contributors -"AS IS" and WITHOUT ANY EXPRESS OR IMPLIED WARRANTIES, including but not limited to the implied warranties -of merchantability and fitness for a particular purpose. IN NO EVENT SHALL THE COPYRIGHT HOLDER OR -CONTRIBUTORS BE LIABLE for any direct, indirect, incidental, special, exemplary, or consequential damages, -including but not limited to procurement of substitute goods or services; loss of use, data, or profits, -or business interruption, however caused and on any theory of liability, whether in contract, strict -liability, or tort (including negligence), arising in any way out of the use of this software, even if -advised of the possibility of such damage. - -5. CONTACT INFORMATION - -Fraunhofer Institute for Integrated Circuits IIS -Attention: Audio and Multimedia Departments - FDK AAC LL -Am Wolfsmantel 33 -91058 Erlangen, Germany - -www.iis.fraunhofer.de/amm -amm-info@iis.fraunhofer.de ------------------------------------------------------------------------------------------------------------ */ - -/*! - * - * \brief Calculate second order autocorrelation - * - */ - - diff --git a/libFDK/src/arm/dct_arm.cpp b/libFDK/src/arm/dct_arm.cpp index 59b773e..dd66109 100644 --- a/libFDK/src/arm/dct_arm.cpp +++ b/libFDK/src/arm/dct_arm.cpp @@ -1,74 +1,85 @@ - -/* ----------------------------------------------------------------------------------------------------------- +/* ----------------------------------------------------------------------------- Software License for The Fraunhofer FDK AAC Codec Library for Android -© Copyright 1995 - 2013 Fraunhofer-Gesellschaft zur Förderung der angewandten Forschung e.V. - All rights reserved. +© Copyright 1995 - 2018 Fraunhofer-Gesellschaft zur Förderung der angewandten +Forschung e.V. All rights reserved. 1. INTRODUCTION -The Fraunhofer FDK AAC Codec Library for Android ("FDK AAC Codec") is software that implements -the MPEG Advanced Audio Coding ("AAC") encoding and decoding scheme for digital audio. -This FDK AAC Codec software is intended to be used on a wide variety of Android devices. - -AAC's HE-AAC and HE-AAC v2 versions are regarded as today's most efficient general perceptual -audio codecs. AAC-ELD is considered the best-performing full-bandwidth communications codec by -independent studies and is widely deployed. AAC has been standardized by ISO and IEC as part -of the MPEG specifications. - -Patent licenses for necessary patent claims for the FDK AAC Codec (including those of Fraunhofer) -may be obtained through Via Licensing (www.vialicensing.com) or through the respective patent owners -individually for the purpose of encoding or decoding bit streams in products that are compliant with -the ISO/IEC MPEG audio standards. Please note that most manufacturers of Android devices already license -these patent claims through Via Licensing or directly from the patent owners, and therefore FDK AAC Codec -software may already be covered under those patent licenses when it is used for those licensed purposes only. - -Commercially-licensed AAC software libraries, including floating-point versions with enhanced sound quality, -are also available from Fraunhofer. Users are encouraged to check the Fraunhofer website for additional -applications information and documentation. +The Fraunhofer FDK AAC Codec Library for Android ("FDK AAC Codec") is software +that implements the MPEG Advanced Audio Coding ("AAC") encoding and decoding +scheme for digital audio. This FDK AAC Codec software is intended to be used on +a wide variety of Android devices. + +AAC's HE-AAC and HE-AAC v2 versions are regarded as today's most efficient +general perceptual audio codecs. AAC-ELD is considered the best-performing +full-bandwidth communications codec by independent studies and is widely +deployed. AAC has been standardized by ISO and IEC as part of the MPEG +specifications. + +Patent licenses for necessary patent claims for the FDK AAC Codec (including +those of Fraunhofer) may be obtained through Via Licensing +(www.vialicensing.com) or through the respective patent owners individually for +the purpose of encoding or decoding bit streams in products that are compliant +with the ISO/IEC MPEG audio standards. Please note that most manufacturers of +Android devices already license these patent claims through Via Licensing or +directly from the patent owners, and therefore FDK AAC Codec software may +already be covered under those patent licenses when it is used for those +licensed purposes only. + +Commercially-licensed AAC software libraries, including floating-point versions +with enhanced sound quality, are also available from Fraunhofer. Users are +encouraged to check the Fraunhofer website for additional applications +information and documentation. 2. COPYRIGHT LICENSE -Redistribution and use in source and binary forms, with or without modification, are permitted without -payment of copyright license fees provided that you satisfy the following conditions: +Redistribution and use in source and binary forms, with or without modification, +are permitted without payment of copyright license fees provided that you +satisfy the following conditions: -You must retain the complete text of this software license in redistributions of the FDK AAC Codec or -your modifications thereto in source code form. +You must retain the complete text of this software license in redistributions of +the FDK AAC Codec or your modifications thereto in source code form. -You must retain the complete text of this software license in the documentation and/or other materials -provided with redistributions of the FDK AAC Codec or your modifications thereto in binary form. -You must make available free of charge copies of the complete source code of the FDK AAC Codec and your +You must retain the complete text of this software license in the documentation +and/or other materials provided with redistributions of the FDK AAC Codec or +your modifications thereto in binary form. You must make available free of +charge copies of the complete source code of the FDK AAC Codec and your modifications thereto to recipients of copies in binary form. -The name of Fraunhofer may not be used to endorse or promote products derived from this library without -prior written permission. +The name of Fraunhofer may not be used to endorse or promote products derived +from this library without prior written permission. -You may not charge copyright license fees for anyone to use, copy or distribute the FDK AAC Codec -software or your modifications thereto. +You may not charge copyright license fees for anyone to use, copy or distribute +the FDK AAC Codec software or your modifications thereto. -Your modified versions of the FDK AAC Codec must carry prominent notices stating that you changed the software -and the date of any change. For modified versions of the FDK AAC Codec, the term -"Fraunhofer FDK AAC Codec Library for Android" must be replaced by the term -"Third-Party Modified Version of the Fraunhofer FDK AAC Codec Library for Android." +Your modified versions of the FDK AAC Codec must carry prominent notices stating +that you changed the software and the date of any change. For modified versions +of the FDK AAC Codec, the term "Fraunhofer FDK AAC Codec Library for Android" +must be replaced by the term "Third-Party Modified Version of the Fraunhofer FDK +AAC Codec Library for Android." 3. NO PATENT LICENSE -NO EXPRESS OR IMPLIED LICENSES TO ANY PATENT CLAIMS, including without limitation the patents of Fraunhofer, -ARE GRANTED BY THIS SOFTWARE LICENSE. Fraunhofer provides no warranty of patent non-infringement with -respect to this software. +NO EXPRESS OR IMPLIED LICENSES TO ANY PATENT CLAIMS, including without +limitation the patents of Fraunhofer, ARE GRANTED BY THIS SOFTWARE LICENSE. +Fraunhofer provides no warranty of patent non-infringement with respect to this +software. -You may use this FDK AAC Codec software or modifications thereto only for purposes that are authorized -by appropriate patent licenses. +You may use this FDK AAC Codec software or modifications thereto only for +purposes that are authorized by appropriate patent licenses. 4. DISCLAIMER -This FDK AAC Codec software is provided by Fraunhofer on behalf of the copyright holders and contributors -"AS IS" and WITHOUT ANY EXPRESS OR IMPLIED WARRANTIES, including but not limited to the implied warranties -of merchantability and fitness for a particular purpose. IN NO EVENT SHALL THE COPYRIGHT HOLDER OR -CONTRIBUTORS BE LIABLE for any direct, indirect, incidental, special, exemplary, or consequential damages, -including but not limited to procurement of substitute goods or services; loss of use, data, or profits, -or business interruption, however caused and on any theory of liability, whether in contract, strict -liability, or tort (including negligence), arising in any way out of the use of this software, even if -advised of the possibility of such damage. +This FDK AAC Codec software is provided by Fraunhofer on behalf of the copyright +holders and contributors "AS IS" and WITHOUT ANY EXPRESS OR IMPLIED WARRANTIES, +including but not limited to the implied warranties of merchantability and +fitness for a particular purpose. IN NO EVENT SHALL THE COPYRIGHT HOLDER OR +CONTRIBUTORS BE LIABLE for any direct, indirect, incidental, special, exemplary, +or consequential damages, including but not limited to procurement of substitute +goods or services; loss of use, data, or profits, or business interruption, +however caused and on any theory of liability, whether in contract, strict +liability, or tort (including negligence), arising in any way out of the use of +this software, even if advised of the possibility of such damage. 5. CONTACT INFORMATION @@ -79,9 +90,15 @@ Am Wolfsmantel 33 www.iis.fraunhofer.de/amm amm-info@iis.fraunhofer.de ------------------------------------------------------------------------------------------------------------ */ +----------------------------------------------------------------------------- */ + +/******************* Library for basic calculation routines ******************** + + Author(s): + Description: +*******************************************************************************/ #ifdef FUNCTION_dct_IV_func1 @@ -91,119 +108,168 @@ amm-info@iis.fraunhofer.de With this version, we save 2 cycles per loop iteration. */ -__asm void dct_IV_func1( - int i, - const FIXP_SPK *twiddle, - FIXP_DBL *RESTRICT pDat_0, - FIXP_DBL *RESTRICT pDat_1) -{ - /* Register map: - r0 i - r1 twiddle - r2 pDat_0 - r3 pDat_1 - r4 accu1 - r5 accu2 - r6 accu3 - r7 accu4 - r8 val_tw - r9 accuX - */ - PUSH {r4-r9} - - /* 44 cycles for 2 iterations = 22 cycles/iteration */ -dct_IV_loop1_start -/* First iteration */ - LDR r8, [r1], #4 // val_tw = *twiddle++; - LDR r5, [r2, #0] // accu2 = pDat_0[0] - LDR r4, [r3, #0] // accu1 = pDat_1[0] - - SMULWT r9, r5, r8 // accuX = accu2*val_tw.l - SMULWB r5, r5, r8 // accu2 = accu2*val_tw.h - RSB r9, r9, #0 // accuX =-accu2*val_tw.l - SMLAWT r5, r4, r8, r5 // accu2 = accu2*val_tw.h + accu1*val_tw.l - SMLAWB r4, r4, r8, r9 // accu1 = accu1*val_tw.h - accu2*val_tw.l - - LDR r8, [r1], #4 // val_tw = *twiddle++; - LDR r7, [r3, #-4] // accu4 = pDat_1[-1] - LDR r6, [r2, #4] // accu3 = pDat_0[1] - - SMULWB r9, r7, r8 // accuX = accu4*val_tw.h - SMULWT r7, r7, r8 // accu4 = accu4*val_tw.l - RSB r9, r9, #0 // accuX =-accu4*val_tw.h - SMLAWB r7, r6, r8, r7 // accu4 = accu4*val_tw.l+accu3*val_tw.h - SMLAWT r6, r6, r8, r9 // accu3 = accu3*val_tw.l-accu4*val_tw.h - - STR r5, [r2], #4 // *pDat_0++ = accu2 - STR r4, [r2], #4 // *pDat_0++ = accu1 - STR r6, [r3], #-4 // *pDat_1-- = accu3 - STR r7, [r3], #-4 // *pDat_1-- = accu4 - -/* Second iteration */ - LDR r8, [r1], #4 // val_tw = *twiddle++; - LDR r5, [r2, #0] // accu2 = pDat_0[0] - LDR r4, [r3, #0] // accu1 = pDat_1[0] - - SMULWT r9, r5, r8 // accuX = accu2*val_tw.l - SMULWB r5, r5, r8 // accu2 = accu2*val_tw.h - RSB r9, r9, #0 // accuX =-accu2*val_tw.l - SMLAWT r5, r4, r8, r5 // accu2 = accu2*val_tw.h + accu1*val_tw.l - SMLAWB r4, r4, r8, r9 // accu1 = accu1*val_tw.h - accu2*val_tw.l - - LDR r8, [r1], #4 // val_tw = *twiddle++; - LDR r7, [r3, #-4] // accu4 = pDat_1[-1] - LDR r6, [r2, #4] // accu3 = pDat_0[1] - - SMULWB r9, r7, r8 // accuX = accu4*val_tw.h - SMULWT r7, r7, r8 // accu4 = accu4*val_tw.l - RSB r9, r9, #0 // accuX =-accu4*val_tw.h - SMLAWB r7, r6, r8, r7 // accu4 = accu4*val_tw.l+accu3*val_tw.h - SMLAWT r6, r6, r8, r9 // accu3 = accu3*val_tw.l-accu4*val_tw.h - - STR r5, [r2], #4 // *pDat_0++ = accu2 - STR r4, [r2], #4 // *pDat_0++ = accu1 - STR r6, [r3], #-4 // *pDat_1-- = accu3 - STR r7, [r3], #-4 // *pDat_1-- = accu4 - - SUBS r0, r0, #1 - BNE dct_IV_loop1_start - - POP {r4-r9} - - BX lr +__asm void dct_IV_func1(int i, const FIXP_SPK *twiddle, + FIXP_DBL *RESTRICT pDat_0, FIXP_DBL *RESTRICT pDat_1) { + /* Register map: + r0 i + r1 twiddle + r2 pDat_0 + r3 pDat_1 + r4 accu1 + r5 accu2 + r6 accu3 + r7 accu4 + r8 val_tw + r9 accuX + */ + PUSH{r4 - r9} + + /* 44 cycles for 2 iterations = 22 cycles/iteration */ + dct_IV_loop1_start + /* First iteration */ + LDR r8, + [r1], +# 4 // val_tw = *twiddle++; + LDR r5, + [ r2, #0 ] // accu2 = pDat_0[0] + LDR r4, + [ r3, #0 ] // accu1 = pDat_1[0] + + SMULWT r9, + r5, + r8 // accuX = accu2*val_tw.l + SMULWB r5, + r5, + r8 // accu2 = accu2*val_tw.h + RSB r9, + r9, +# 0 // accuX =-accu2*val_tw.l + SMLAWT r5, r4, r8, + r5 // accu2 = accu2*val_tw.h + accu1*val_tw.l + SMLAWB r4, + r4, r8, + r9 // accu1 = accu1*val_tw.h - accu2*val_tw.l + + LDR r8, + [r1], +# 4 // val_tw = *twiddle++; + LDR r7, + [ r3, # - 4 ] // accu4 = pDat_1[-1] + LDR r6, + [ r2, #4 ] // accu3 = pDat_0[1] + + SMULWB r9, + r7, + r8 // accuX = accu4*val_tw.h + SMULWT r7, + r7, + r8 // accu4 = accu4*val_tw.l + RSB r9, + r9, +# 0 // accuX =-accu4*val_tw.h + SMLAWB r7, r6, r8, + r7 // accu4 = accu4*val_tw.l+accu3*val_tw.h + SMLAWT r6, + r6, r8, + r9 // accu3 = accu3*val_tw.l-accu4*val_tw.h + + STR r5, + [r2], +# 4 // *pDat_0++ = accu2 + STR r4, [r2], +# 4 // *pDat_0++ = accu1 + STR r6, [r3], +#- 4 // *pDat_1-- = accu3 + STR r7, [r3], +#- 4 // *pDat_1-- = accu4 + + /* Second iteration */ + LDR r8, [r1], +# 4 // val_tw = *twiddle++; + LDR r5, + [ r2, #0 ] // accu2 = pDat_0[0] + LDR r4, + [ r3, #0 ] // accu1 = pDat_1[0] + + SMULWT r9, + r5, + r8 // accuX = accu2*val_tw.l + SMULWB r5, + r5, + r8 // accu2 = accu2*val_tw.h + RSB r9, + r9, +# 0 // accuX =-accu2*val_tw.l + SMLAWT r5, r4, r8, + r5 // accu2 = accu2*val_tw.h + accu1*val_tw.l + SMLAWB r4, + r4, r8, + r9 // accu1 = accu1*val_tw.h - accu2*val_tw.l + + LDR r8, + [r1], +# 4 // val_tw = *twiddle++; + LDR r7, + [ r3, # - 4 ] // accu4 = pDat_1[-1] + LDR r6, + [ r2, #4 ] // accu3 = pDat_0[1] + + SMULWB r9, + r7, + r8 // accuX = accu4*val_tw.h + SMULWT r7, + r7, + r8 // accu4 = accu4*val_tw.l + RSB r9, + r9, +# 0 // accuX =-accu4*val_tw.h + SMLAWB r7, r6, r8, + r7 // accu4 = accu4*val_tw.l+accu3*val_tw.h + SMLAWT r6, + r6, r8, + r9 // accu3 = accu3*val_tw.l-accu4*val_tw.h + + STR r5, + [r2], +# 4 // *pDat_0++ = accu2 + STR r4, [r2], +# 4 // *pDat_0++ = accu1 + STR r6, [r3], +#- 4 // *pDat_1-- = accu3 + STR r7, [r3], +#- 4 // *pDat_1-- = accu4 + + SUBS r0, r0, +# 1 BNE dct_IV_loop1_start + + POP { r4 - r9 } + + BX lr } #endif /* FUNCTION_dct_IV_func1 */ - #ifdef FUNCTION_dct_IV_func2 -FDK_INLINE /* __attribute__((noinline)) */ -static void dct_IV_func2( - int i, - const FIXP_SPK *twiddle, - FIXP_DBL *pDat_0, - FIXP_DBL *pDat_1, - int inc) -{ +static inline void dct_IV_func2(int i, const FIXP_SPK *twiddle, + FIXP_DBL *pDat_0, FIXP_DBL *pDat_1, int inc) { FIXP_DBL accu1, accu2, accu3, accu4, accuX; LONG val_tw; accu1 = pDat_1[-2]; accu2 = pDat_1[-1]; - *--pDat_1 = -(pDat_0[1]>>1); - *pDat_0++ = (pDat_0[0]>>1); + *--pDat_1 = -(pDat_0[1] >> 1); + *pDat_0++ = (pDat_0[0] >> 1); twiddle += inc; -__asm - { - LDR val_tw, [twiddle], inc, LSL #2 // val_tw = *twiddle; twiddle += inc + __asm { + LDR val_tw, [twiddle], inc, LSL #2 // val_tw = *twiddle; twiddle += inc B dct_IV_loop2_2nd_part - /* 42 cycles for 2 iterations = 21 cycles/iteration */ + /* 42 cycles for 2 iterations = 21 cycles/iteration */ dct_IV_loop2: SMULWT accuX, accu2, val_tw SMULWB accu2, accu2, val_tw @@ -224,7 +290,7 @@ dct_IV_loop2: LDR accu1, [pDat_1, #-8] LDR accu2, [pDat_1, #-4] - LDR val_tw, [twiddle], inc, LSL #2 // val_tw = *twiddle; twiddle += inc + LDR val_tw, [twiddle], inc, LSL #2 // val_tw = *twiddle; twiddle += inc STR accuX, [pDat_1, #-4] ! STR accu4, [pDat_0], #4 @@ -252,7 +318,7 @@ dct_IV_loop2_2nd_part: STR accuX, [pDat_1, #-4] ! STR accu4, [pDat_0], #4 - LDR val_tw, [twiddle], inc, LSL #2 // val_tw = *twiddle; twiddle += inc + LDR val_tw, [twiddle], inc, LSL #2 // val_tw = *twiddle; twiddle += inc SUBS i, i, #1 BNE dct_IV_loop2 @@ -267,96 +333,148 @@ dct_IV_loop2_2nd_part: } #endif /* FUNCTION_dct_IV_func2 */ - #ifdef FUNCTION_dst_IV_func1 -__asm void dst_IV_func1( - int i, - const FIXP_SPK *twiddle, - FIXP_DBL *pDat_0, - FIXP_DBL *pDat_1) -{ - /* Register map: - r0 i - r1 twiddle - r2 pDat_0 - r3 pDat_1 - r4 accu1 - r5 accu2 - r6 accu3 - r7 accu4 - r8 val_tw - r9 accuX - */ - PUSH {r4-r9} - -dst_IV_loop1 - LDR r8, [r1], #4 // val_tw = *twiddle++ - LDR r5, [r2] // accu2 = pDat_0[0] - LDR r6, [r2, #4] // accu3 = pDat_0[1] - RSB r5, r5, #0 // accu2 = -accu2 - SMULWT r9, r5, r8 // accuX = (-accu2)*val_tw.l - LDR r4, [r3, #-4] // accu1 = pDat_1[-1] - RSB r9, r9, #0 // accuX = -(-accu2)*val_tw.l - SMLAWB r9, r4, r8, r9 // accuX = accu1*val_tw.h-(-accu2)*val_tw.l - SMULWT r4, r4, r8 // accu1 = accu1*val_tw.l - LDR r7, [r3, #-8] // accu4 = pDat_1[-2] - SMLAWB r5, r5, r8, r4 // accu2 = (-accu2)*val_tw.t+accu1*val_tw.l - LDR r8, [r1], #4 // val_tw = *twiddle++ - STR r5, [r2], #4 // *pDat_0++ = accu2 - STR r9, [r2], #4 // *pDat_0++ = accu1 (accuX) - RSB r7, r7, #0 // accu4 = -accu4 - SMULWB r5, r7, r8 // accu2 = (-accu4)*val_tw.h - SMULWB r4, r6, r8 // accu1 = (-accu4)*val_tw.l - RSB r5, r5, #0 // accu2 = -(-accu4)*val_tw.h - SMLAWT r6, r6, r8, r5 // accu3 = (-accu4)*val_tw.l-(-accu3)*val_tw.h - SMLAWT r7, r7, r8, r4 // accu4 = (-accu3)*val_tw.l+(-accu4)*val_tw.h - STR r6, [r3, #-4] ! // *--pDat_1 = accu3 - STR r7, [r3, #-4] ! // *--pDat_1 = accu4 - - LDR r8, [r1], #4 // val_tw = *twiddle++ - LDR r5, [r2] // accu2 = pDat_0[0] - LDR r6, [r2, #4] // accu3 = pDat_0[1] - RSB r5, r5, #0 // accu2 = -accu2 - SMULWT r9, r5, r8 // accuX = (-accu2)*val_tw.l - LDR r4, [r3, #-4] // accu1 = pDat_1[-1] - RSB r9, r9, #0 // accuX = -(-accu2)*val_tw.l - SMLAWB r9, r4, r8, r9 // accuX = accu1*val_tw.h-(-accu2)*val_tw.l - SMULWT r4, r4, r8 // accu1 = accu1*val_tw.l - LDR r7, [r3, #-8] // accu4 = pDat_1[-2] - SMLAWB r5, r5, r8, r4 // accu2 = (-accu2)*val_tw.t+accu1*val_tw.l - LDR r8, [r1], #4 // val_tw = *twiddle++ - STR r5, [r2], #4 // *pDat_0++ = accu2 - STR r9, [r2], #4 // *pDat_0++ = accu1 (accuX) - RSB r7, r7, #0 // accu4 = -accu4 - SMULWB r5, r7, r8 // accu2 = (-accu4)*val_tw.h - SMULWB r4, r6, r8 // accu1 = (-accu4)*val_tw.l - RSB r5, r5, #0 // accu2 = -(-accu4)*val_tw.h - SMLAWT r6, r6, r8, r5 // accu3 = (-accu4)*val_tw.l-(-accu3)*val_tw.h - SMLAWT r7, r7, r8, r4 // accu4 = (-accu3)*val_tw.l+(-accu4)*val_tw.h - STR r6, [r3, #-4] ! // *--pDat_1 = accu3 - STR r7, [r3, #-4] ! // *--pDat_1 = accu4 - - SUBS r0, r0, #4 // i-= 4 - BNE dst_IV_loop1 - - POP {r4-r9} - BX lr +__asm void dst_IV_func1(int i, const FIXP_SPK *twiddle, FIXP_DBL *pDat_0, + FIXP_DBL *pDat_1) { + /* Register map: + r0 i + r1 twiddle + r2 pDat_0 + r3 pDat_1 + r4 accu1 + r5 accu2 + r6 accu3 + r7 accu4 + r8 val_tw + r9 accuX + */ + PUSH{r4 - r9} + + dst_IV_loop1 LDR r8, + [r1], +# 4 // val_tw = *twiddle++ + LDR r5, + [r2] // accu2 = pDat_0[0] + LDR r6, + [ r2, #4 ] // accu3 = pDat_0[1] + RSB r5, + r5, +# 0 // accu2 = -accu2 + SMULWT r9, r5, + r8 // accuX = (-accu2)*val_tw.l + LDR r4, + [ r3, # - 4 ] // accu1 = pDat_1[-1] + RSB r9, + r9, +# 0 // accuX = -(-accu2)*val_tw.l + SMLAWB r9, r4, r8, + r9 // accuX = accu1*val_tw.h-(-accu2)*val_tw.l + SMULWT r4, + r4, + r8 // accu1 = accu1*val_tw.l + LDR r7, + [ r3, # - 8 ] // accu4 = pDat_1[-2] + SMLAWB r5, + r5, r8, + r4 // accu2 = (-accu2)*val_tw.t+accu1*val_tw.l + LDR r8, + [r1], +# 4 // val_tw = *twiddle++ + STR r5, [r2], +# 4 // *pDat_0++ = accu2 + STR r9, [r2], +# 4 // *pDat_0++ = accu1 (accuX) + RSB r7, r7, +# 0 // accu4 = -accu4 + SMULWB r5, r7, + r8 // accu2 = (-accu4)*val_tw.h + SMULWB r4, + r6, + r8 // accu1 = (-accu4)*val_tw.l + RSB r5, + r5, +# 0 // accu2 = -(-accu4)*val_tw.h + SMLAWT r6, r6, r8, + r5 // accu3 = (-accu4)*val_tw.l-(-accu3)*val_tw.h + SMLAWT r7, + r7, r8, + r4 // accu4 = (-accu3)*val_tw.l+(-accu4)*val_tw.h + STR r6, + [ r3, # - 4 ] ! // *--pDat_1 = accu3 + STR r7, + [ r3, # - 4 ] ! // *--pDat_1 = accu4 + + LDR r8, + [r1], +# 4 // val_tw = *twiddle++ + LDR r5, + [r2] // accu2 = pDat_0[0] + LDR r6, + [ r2, #4 ] // accu3 = pDat_0[1] + RSB r5, + r5, +# 0 // accu2 = -accu2 + SMULWT r9, r5, + r8 // accuX = (-accu2)*val_tw.l + LDR r4, + [ r3, # - 4 ] // accu1 = pDat_1[-1] + RSB r9, + r9, +# 0 // accuX = -(-accu2)*val_tw.l + SMLAWB r9, r4, r8, + r9 // accuX = accu1*val_tw.h-(-accu2)*val_tw.l + SMULWT r4, + r4, + r8 // accu1 = accu1*val_tw.l + LDR r7, + [ r3, # - 8 ] // accu4 = pDat_1[-2] + SMLAWB r5, + r5, r8, + r4 // accu2 = (-accu2)*val_tw.t+accu1*val_tw.l + LDR r8, + [r1], +# 4 // val_tw = *twiddle++ + STR r5, [r2], +# 4 // *pDat_0++ = accu2 + STR r9, [r2], +# 4 // *pDat_0++ = accu1 (accuX) + RSB r7, r7, +# 0 // accu4 = -accu4 + SMULWB r5, r7, + r8 // accu2 = (-accu4)*val_tw.h + SMULWB r4, + r6, + r8 // accu1 = (-accu4)*val_tw.l + RSB r5, + r5, +# 0 // accu2 = -(-accu4)*val_tw.h + SMLAWT r6, r6, r8, + r5 // accu3 = (-accu4)*val_tw.l-(-accu3)*val_tw.h + SMLAWT r7, + r7, r8, + r4 // accu4 = (-accu3)*val_tw.l+(-accu4)*val_tw.h + STR r6, + [ r3, # - 4 ] ! // *--pDat_1 = accu3 + STR r7, + [ r3, # - 4 ] ! // *--pDat_1 = accu4 + + SUBS r0, + r0, +# 4 // i-= 4 + BNE dst_IV_loop1 + + POP{r4 - r9} BX lr } #endif /* FUNCTION_dst_IV_func1 */ #ifdef FUNCTION_dst_IV_func2 -FDK_INLINE /* __attribute__((noinline)) */ -static void dst_IV_func2( - int i, - const FIXP_SPK *twiddle, - FIXP_DBL *RESTRICT pDat_0, - FIXP_DBL *RESTRICT pDat_1, - int inc) -{ - FIXP_DBL accu1,accu2,accu3,accu4; +static inline void dst_IV_func2(int i, const FIXP_SPK *twiddle, + FIXP_DBL *RESTRICT pDat_0, + FIXP_DBL *RESTRICT pDat_1, int inc) { + FIXP_DBL accu1, accu2, accu3, accu4; LONG val_tw; accu4 = pDat_0[0]; @@ -371,76 +489,74 @@ static void dst_IV_func2( *pDat_0++ = accu3; *pDat_1-- = accu4; - - __asm - { + __asm { B dst_IV_loop2_2nd_part - /* 50 cycles for 2 iterations = 25 cycles/iteration */ + /* 50 cycles for 2 iterations = 25 cycles/iteration */ dst_IV_loop2: - LDR val_tw, [twiddle], inc, LSL #2 // val_tw = *twiddle; twiddle += inc + LDR val_tw, [twiddle], inc, LSL #2 // val_tw = *twiddle; twiddle += inc - RSB accu2, accu2, #0 // accu2 = -accu2 - RSB accu1, accu1, #0 // accu1 = -accu1 - SMULWT accu3, accu2, val_tw // accu3 = (-accu2)*val_tw.l - SMULWT accu4, accu1, val_tw // accu4 = (-accu1)*val_tw.l - RSB accu3, accu3, #0 // accu3 = -accu2*val_tw.l - SMLAWB accu1, accu1, val_tw, accu3 // accu1 = -accu1*val_tw.h-(-accu2)*val_tw.l - SMLAWB accu2, accu2, val_tw, accu4 // accu2 = (-accu1)*val_tw.l+(-accu2)*val_tw.h - STR accu1, [pDat_1], #-4 // *pDat_1-- = accu1 - STR accu2, [pDat_0], #4 // *pDat_0++ = accu2 + RSB accu2, accu2, #0 // accu2 = -accu2 + RSB accu1, accu1, #0 // accu1 = -accu1 + SMULWT accu3, accu2, val_tw // accu3 = (-accu2)*val_tw.l + SMULWT accu4, accu1, val_tw // accu4 = (-accu1)*val_tw.l + RSB accu3, accu3, #0 // accu3 = -accu2*val_tw.l + SMLAWB accu1, accu1, val_tw, accu3 // accu1 = -accu1*val_tw.h-(-accu2)*val_tw.l + SMLAWB accu2, accu2, val_tw, accu4 // accu2 = (-accu1)*val_tw.l+(-accu2)*val_tw.h + STR accu1, [pDat_1], #-4 // *pDat_1-- = accu1 + STR accu2, [pDat_0], #4 // *pDat_0++ = accu2 - LDR accu4, [pDat_0] // accu4 = pDat_0[0] - LDR accu3, [pDat_0, #4] // accu3 = pDat_0[1] + LDR accu4, [pDat_0] // accu4 = pDat_0[0] + LDR accu3, [pDat_0, #4] // accu3 = pDat_0[1] - RSB accu4, accu4, #0 // accu4 = -accu4 - RSB accu3, accu3, #0 // accu3 = -accu3 + RSB accu4, accu4, #0 // accu4 = -accu4 + RSB accu3, accu3, #0 // accu3 = -accu3 - SMULWB accu1, accu3, val_tw // accu1 = (-accu3)*val_tw.h - SMULWT accu2, accu3, val_tw // accu2 = (-accu3)*val_tw.l - RSB accu1, accu1, #0 // accu1 = -(-accu3)*val_tw.h - SMLAWT accu3, accu4, val_tw, accu1 // accu3 = (-accu4)*val_tw.l-(-accu3)*val_tw.h - SMLAWB accu4, accu4, val_tw, accu2 // accu4 = (-accu3)*val_tw.l+(-accu4)*val_tw.h + SMULWB accu1, accu3, val_tw // accu1 = (-accu3)*val_tw.h + SMULWT accu2, accu3, val_tw // accu2 = (-accu3)*val_tw.l + RSB accu1, accu1, #0 // accu1 = -(-accu3)*val_tw.h + SMLAWT accu3, accu4, val_tw, accu1 // accu3 = (-accu4)*val_tw.l-(-accu3)*val_tw.h + SMLAWB accu4, accu4, val_tw, accu2 // accu4 = (-accu3)*val_tw.l+(-accu4)*val_tw.h - LDR accu1, [pDat_1, #-4] // accu1 = pDat_1[-1] - LDR accu2, [pDat_1] // accu2 = pDat_1[0] + LDR accu1, [pDat_1, #-4] // accu1 = pDat_1[-1] + LDR accu2, [pDat_1] // accu2 = pDat_1[0] - STR accu3, [pDat_0], #4 // *pDat_0++ = accu3 - STR accu4, [pDat_1], #-4 // *pDat_1-- = accu4 + STR accu3, [pDat_0], #4 // *pDat_0++ = accu3 + STR accu4, [pDat_1], #-4 // *pDat_1-- = accu4 dst_IV_loop2_2nd_part: - LDR val_tw, [twiddle], inc, LSL #2 // val_tw = *twiddle; twiddle += inc + LDR val_tw, [twiddle], inc, LSL #2 // val_tw = *twiddle; twiddle += inc - RSB accu2, accu2, #0 // accu2 = -accu2 - RSB accu1, accu1, #0 // accu1 = -accu1 - SMULWT accu3, accu2, val_tw // accu3 = (-accu2)*val_tw.l - SMULWT accu4, accu1, val_tw // accu4 = (-accu1)*val_tw.l - RSB accu3, accu3, #0 // accu3 = -accu2*val_tw.l - SMLAWB accu1, accu1, val_tw, accu3 // accu1 = -accu1*val_tw.h-(-accu2)*val_tw.l - SMLAWB accu2, accu2, val_tw, accu4 // accu2 = (-accu1)*val_tw.l+(-accu2)*val_tw.h - STR accu1, [pDat_1], #-4 // *pDat_1-- = accu1 - STR accu2, [pDat_0], #4 // *pDat_0++ = accu2 + RSB accu2, accu2, #0 // accu2 = -accu2 + RSB accu1, accu1, #0 // accu1 = -accu1 + SMULWT accu3, accu2, val_tw // accu3 = (-accu2)*val_tw.l + SMULWT accu4, accu1, val_tw // accu4 = (-accu1)*val_tw.l + RSB accu3, accu3, #0 // accu3 = -accu2*val_tw.l + SMLAWB accu1, accu1, val_tw, accu3 // accu1 = -accu1*val_tw.h-(-accu2)*val_tw.l + SMLAWB accu2, accu2, val_tw, accu4 // accu2 = (-accu1)*val_tw.l+(-accu2)*val_tw.h + STR accu1, [pDat_1], #-4 // *pDat_1-- = accu1 + STR accu2, [pDat_0], #4 // *pDat_0++ = accu2 - LDR accu4, [pDat_0] // accu4 = pDat_0[0] - LDR accu3, [pDat_0, #4] // accu3 = pDat_0[1] + LDR accu4, [pDat_0] // accu4 = pDat_0[0] + LDR accu3, [pDat_0, #4] // accu3 = pDat_0[1] - RSB accu4, accu4, #0 // accu4 = -accu4 - RSB accu3, accu3, #0 // accu3 = -accu3 + RSB accu4, accu4, #0 // accu4 = -accu4 + RSB accu3, accu3, #0 // accu3 = -accu3 - SMULWB accu1, accu3, val_tw // accu1 = (-accu3)*val_tw.h - SMULWT accu2, accu3, val_tw // accu2 = (-accu3)*val_tw.l - RSB accu1, accu1, #0 // accu1 = -(-accu3)*val_tw.h - SMLAWT accu3, accu4, val_tw, accu1 // accu3 = (-accu4)*val_tw.l-(-accu3)*val_tw.h - SMLAWB accu4, accu4, val_tw, accu2 // accu4 = (-accu3)*val_tw.l+(-accu4)*val_tw.h + SMULWB accu1, accu3, val_tw // accu1 = (-accu3)*val_tw.h + SMULWT accu2, accu3, val_tw // accu2 = (-accu3)*val_tw.l + RSB accu1, accu1, #0 // accu1 = -(-accu3)*val_tw.h + SMLAWT accu3, accu4, val_tw, accu1 // accu3 = (-accu4)*val_tw.l-(-accu3)*val_tw.h + SMLAWB accu4, accu4, val_tw, accu2 // accu4 = (-accu3)*val_tw.l+(-accu4)*val_tw.h - LDR accu1, [pDat_1, #-4] // accu1 = pDat_1[-1] - LDR accu2, [pDat_1] // accu2 = pDat_1[0] + LDR accu1, [pDat_1, #-4] // accu1 = pDat_1[-1] + LDR accu2, [pDat_1] // accu2 = pDat_1[0] - STR accu3, [pDat_0], #4 // *pDat_0++ = accu3 - STR accu4, [pDat_1], #-4 // *pDat_1-- = accu4 + STR accu3, [pDat_0], #4 // *pDat_0++ = accu3 + STR accu4, [pDat_1], #-4 // *pDat_1-- = accu4 SUBS i, i, #1 BNE dst_IV_loop2 diff --git a/libFDK/src/arm/fft_rad2_arm.cpp b/libFDK/src/arm/fft_rad2_arm.cpp index a6e2e42..2c03b11 100644 --- a/libFDK/src/arm/fft_rad2_arm.cpp +++ b/libFDK/src/arm/fft_rad2_arm.cpp @@ -1,74 +1,85 @@ - -/* ----------------------------------------------------------------------------------------------------------- +/* ----------------------------------------------------------------------------- Software License for The Fraunhofer FDK AAC Codec Library for Android -© Copyright 1995 - 2013 Fraunhofer-Gesellschaft zur Förderung der angewandten Forschung e.V. - All rights reserved. +© Copyright 1995 - 2018 Fraunhofer-Gesellschaft zur Förderung der angewandten +Forschung e.V. All rights reserved. 1. INTRODUCTION -The Fraunhofer FDK AAC Codec Library for Android ("FDK AAC Codec") is software that implements -the MPEG Advanced Audio Coding ("AAC") encoding and decoding scheme for digital audio. -This FDK AAC Codec software is intended to be used on a wide variety of Android devices. - -AAC's HE-AAC and HE-AAC v2 versions are regarded as today's most efficient general perceptual -audio codecs. AAC-ELD is considered the best-performing full-bandwidth communications codec by -independent studies and is widely deployed. AAC has been standardized by ISO and IEC as part -of the MPEG specifications. - -Patent licenses for necessary patent claims for the FDK AAC Codec (including those of Fraunhofer) -may be obtained through Via Licensing (www.vialicensing.com) or through the respective patent owners -individually for the purpose of encoding or decoding bit streams in products that are compliant with -the ISO/IEC MPEG audio standards. Please note that most manufacturers of Android devices already license -these patent claims through Via Licensing or directly from the patent owners, and therefore FDK AAC Codec -software may already be covered under those patent licenses when it is used for those licensed purposes only. - -Commercially-licensed AAC software libraries, including floating-point versions with enhanced sound quality, -are also available from Fraunhofer. Users are encouraged to check the Fraunhofer website for additional -applications information and documentation. +The Fraunhofer FDK AAC Codec Library for Android ("FDK AAC Codec") is software +that implements the MPEG Advanced Audio Coding ("AAC") encoding and decoding +scheme for digital audio. This FDK AAC Codec software is intended to be used on +a wide variety of Android devices. + +AAC's HE-AAC and HE-AAC v2 versions are regarded as today's most efficient +general perceptual audio codecs. AAC-ELD is considered the best-performing +full-bandwidth communications codec by independent studies and is widely +deployed. AAC has been standardized by ISO and IEC as part of the MPEG +specifications. + +Patent licenses for necessary patent claims for the FDK AAC Codec (including +those of Fraunhofer) may be obtained through Via Licensing +(www.vialicensing.com) or through the respective patent owners individually for +the purpose of encoding or decoding bit streams in products that are compliant +with the ISO/IEC MPEG audio standards. Please note that most manufacturers of +Android devices already license these patent claims through Via Licensing or +directly from the patent owners, and therefore FDK AAC Codec software may +already be covered under those patent licenses when it is used for those +licensed purposes only. + +Commercially-licensed AAC software libraries, including floating-point versions +with enhanced sound quality, are also available from Fraunhofer. Users are +encouraged to check the Fraunhofer website for additional applications +information and documentation. 2. COPYRIGHT LICENSE -Redistribution and use in source and binary forms, with or without modification, are permitted without -payment of copyright license fees provided that you satisfy the following conditions: +Redistribution and use in source and binary forms, with or without modification, +are permitted without payment of copyright license fees provided that you +satisfy the following conditions: -You must retain the complete text of this software license in redistributions of the FDK AAC Codec or -your modifications thereto in source code form. +You must retain the complete text of this software license in redistributions of +the FDK AAC Codec or your modifications thereto in source code form. -You must retain the complete text of this software license in the documentation and/or other materials -provided with redistributions of the FDK AAC Codec or your modifications thereto in binary form. -You must make available free of charge copies of the complete source code of the FDK AAC Codec and your +You must retain the complete text of this software license in the documentation +and/or other materials provided with redistributions of the FDK AAC Codec or +your modifications thereto in binary form. You must make available free of +charge copies of the complete source code of the FDK AAC Codec and your modifications thereto to recipients of copies in binary form. -The name of Fraunhofer may not be used to endorse or promote products derived from this library without -prior written permission. +The name of Fraunhofer may not be used to endorse or promote products derived +from this library without prior written permission. -You may not charge copyright license fees for anyone to use, copy or distribute the FDK AAC Codec -software or your modifications thereto. +You may not charge copyright license fees for anyone to use, copy or distribute +the FDK AAC Codec software or your modifications thereto. -Your modified versions of the FDK AAC Codec must carry prominent notices stating that you changed the software -and the date of any change. For modified versions of the FDK AAC Codec, the term -"Fraunhofer FDK AAC Codec Library for Android" must be replaced by the term -"Third-Party Modified Version of the Fraunhofer FDK AAC Codec Library for Android." +Your modified versions of the FDK AAC Codec must carry prominent notices stating +that you changed the software and the date of any change. For modified versions +of the FDK AAC Codec, the term "Fraunhofer FDK AAC Codec Library for Android" +must be replaced by the term "Third-Party Modified Version of the Fraunhofer FDK +AAC Codec Library for Android." 3. NO PATENT LICENSE -NO EXPRESS OR IMPLIED LICENSES TO ANY PATENT CLAIMS, including without limitation the patents of Fraunhofer, -ARE GRANTED BY THIS SOFTWARE LICENSE. Fraunhofer provides no warranty of patent non-infringement with -respect to this software. +NO EXPRESS OR IMPLIED LICENSES TO ANY PATENT CLAIMS, including without +limitation the patents of Fraunhofer, ARE GRANTED BY THIS SOFTWARE LICENSE. +Fraunhofer provides no warranty of patent non-infringement with respect to this +software. -You may use this FDK AAC Codec software or modifications thereto only for purposes that are authorized -by appropriate patent licenses. +You may use this FDK AAC Codec software or modifications thereto only for +purposes that are authorized by appropriate patent licenses. 4. DISCLAIMER -This FDK AAC Codec software is provided by Fraunhofer on behalf of the copyright holders and contributors -"AS IS" and WITHOUT ANY EXPRESS OR IMPLIED WARRANTIES, including but not limited to the implied warranties -of merchantability and fitness for a particular purpose. IN NO EVENT SHALL THE COPYRIGHT HOLDER OR -CONTRIBUTORS BE LIABLE for any direct, indirect, incidental, special, exemplary, or consequential damages, -including but not limited to procurement of substitute goods or services; loss of use, data, or profits, -or business interruption, however caused and on any theory of liability, whether in contract, strict -liability, or tort (including negligence), arising in any way out of the use of this software, even if -advised of the possibility of such damage. +This FDK AAC Codec software is provided by Fraunhofer on behalf of the copyright +holders and contributors "AS IS" and WITHOUT ANY EXPRESS OR IMPLIED WARRANTIES, +including but not limited to the implied warranties of merchantability and +fitness for a particular purpose. IN NO EVENT SHALL THE COPYRIGHT HOLDER OR +CONTRIBUTORS BE LIABLE for any direct, indirect, incidental, special, exemplary, +or consequential damages, including but not limited to procurement of substitute +goods or services; loss of use, data, or profits, or business interruption, +however caused and on any theory of liability, whether in contract, strict +liability, or tort (including negligence), arising in any way out of the use of +this software, even if advised of the possibility of such damage. 5. CONTACT INFORMATION @@ -79,239 +90,228 @@ Am Wolfsmantel 33 www.iis.fraunhofer.de/amm amm-info@iis.fraunhofer.de ------------------------------------------------------------------------------------------------------------ */ +----------------------------------------------------------------------------- */ -/*************************** Fraunhofer IIS FDK Tools ********************** +/******************* Library for basic calculation routines ******************** Author(s): + Description: dit_fft ARM assembler replacements. -******************************************************************************/ +*******************************************************************************/ -/* NEON optimized FFT currently builds only with RVCT toolchain */ +#ifndef __FFT_RAD2_CPP__ +#error \ + "Do not compile this file separately. It is included on demand from fft_rad2.cpp" +#endif #ifndef FUNCTION_dit_fft - -/* If dit_fft was not yet defined by ARM-Cortex ... */ - #if defined(SINETABLE_16BIT) #define FUNCTION_dit_fft - -/***************************************************************************** - - date: 28.07.2005 srl - - Contents/description: dit-tukey-FFT-algorithm - -******************************************************************************/ - #if defined(FUNCTION_dit_fft) - -void dit_fft(FIXP_DBL *x, const INT ldn, const FIXP_STP *trigdata, const INT trigDataSize) -{ - const INT n=1<> 1; /* Re A + Re B */ + a10 = (x[i + 4] + x[i + 6]) >> 1; /* Re C + Re D */ + a20 = (x[i + 1] + x[i + 3]) >> 1; /* Im A + Im B */ + a30 = (x[i + 5] + x[i + 7]) >> 1; /* Im C + Im D */ + + x[i + 0] = a00 + a10; /* Re A' = Re A + Re B + Re C + Re D */ + x[i + 4] = a00 - a10; /* Re C' = Re A + Re B - Re C - Re D */ + x[i + 1] = a20 + a30; /* Im A' = Im A + Im B + Im C + Im D */ + x[i + 5] = a20 - a30; /* Im C' = Im A + Im B - Im C - Im D */ + + a00 = a00 - x[i + 2]; /* Re A - Re B */ + a10 = a10 - x[i + 6]; /* Re C - Re D */ + a20 = a20 - x[i + 3]; /* Im A - Im B */ + a30 = a30 - x[i + 7]; /* Im C - Im D */ + + x[i + 2] = a00 + a30; /* Re B' = Re A - Re B + Im C - Im D */ + x[i + 6] = a00 - a30; /* Re D' = Re A - Re B - Im C + Im D */ + x[i + 3] = a20 - a10; /* Im B' = Im A - Im B - Re C + Re D */ + x[i + 7] = a20 + a10; /* Im D' = Im A - Im B + Re C - Re D */ + } + + INT mh = 1 << 1; + INT ldm = ldn - 2; + INT trigstep = trigDataSize; + + do { + const FIXP_STP *pTrigData = trigdata; + INT j; + + mh <<= 1; + trigstep >>= 1; + + FDK_ASSERT(trigstep > 0); + + /* Do first iteration with c=1.0 and s=0.0 separately to avoid loosing to + much precision. Beware: The impact on the overal FFT precision is rather + large. */ { - FIXP_DBL a00, a10, a20, a30; - a00 = (x[i + 0] + x[i + 2])>>1; /* Re A + Re B */ - a10 = (x[i + 4] + x[i + 6])>>1; /* Re C + Re D */ - a20 = (x[i + 1] + x[i + 3])>>1; /* Im A + Im B */ - a30 = (x[i + 5] + x[i + 7])>>1; /* Im C + Im D */ - - x[i + 0] = a00 + a10; /* Re A' = Re A + Re B + Re C + Re D */ - x[i + 4] = a00 - a10; /* Re C' = Re A + Re B - Re C - Re D */ - x[i + 1] = a20 + a30; /* Im A' = Im A + Im B + Im C + Im D */ - x[i + 5] = a20 - a30; /* Im C' = Im A + Im B - Im C - Im D */ - - a00 = a00 - x[i + 2]; /* Re A - Re B */ - a10 = a10 - x[i + 6]; /* Re C - Re D */ - a20 = a20 - x[i + 3]; /* Im A - Im B */ - a30 = a30 - x[i + 7]; /* Im C - Im D */ - - x[i + 2] = a00 + a30; /* Re B' = Re A - Re B + Im C - Im D */ - x[i + 6] = a00 - a30; /* Re D' = Re A - Re B - Im C + Im D */ - x[i + 3] = a20 - a10; /* Im B' = Im A - Im B - Re C + Re D */ - x[i + 7] = a20 + a10; /* Im D' = Im A - Im B + Re C - Re D */ - } - - INT mh = 1 << 1; - INT ldm = ldn - 2; - INT trigstep = trigDataSize; - - do - { - const FIXP_STP *pTrigData = trigdata; - INT j; - - mh <<= 1; - trigstep >>= 1; - - FDK_ASSERT(trigstep > 0); + FIXP_DBL *xt1 = x; + int r = n; - /* Do first iteration with c=1.0 and s=0.0 separately to avoid loosing to much precision. - Beware: The impact on the overal FFT precision is rather large. */ - { - FIXP_DBL *xt1 = x; - int r = n; + do { + FIXP_DBL *xt2 = xt1 + (mh << 1); + /* + FIXP_DBL *xt1 = x+ ((r)<<1); + FIXP_DBL *xt2 = xt1 + (mh<<1); + */ + FIXP_DBL vr, vi, ur, ui; - do { - FIXP_DBL *xt2 = xt1 + (mh<<1); - /* - FIXP_DBL *xt1 = x+ ((r)<<1); - FIXP_DBL *xt2 = xt1 + (mh<<1); - */ - FIXP_DBL vr,vi,ur,ui; + // cplxMultDiv2(&vi, &vr, x[t2+1], x[t2], (FIXP_SGL)1.0, (FIXP_SGL)0.0); + vi = xt2[1] >> 1; + vr = xt2[0] >> 1; - //cplxMultDiv2(&vi, &vr, x[t2+1], x[t2], (FIXP_SGL)1.0, (FIXP_SGL)0.0); - vi = xt2[1]>>1; - vr = xt2[0]>>1; + ur = xt1[0] >> 1; + ui = xt1[1] >> 1; - ur = xt1[0]>>1; - ui = xt1[1]>>1; + xt1[0] = ur + vr; + xt1[1] = ui + vi; - xt1[0] = ur+vr; - xt1[1] = ui+vi; + xt2[0] = ur - vr; + xt2[1] = ui - vi; - xt2[0] = ur-vr; - xt2[1] = ui-vi; + xt1 += mh; + xt2 += mh; - xt1 += mh; - xt2 += mh; + // cplxMultDiv2(&vr, &vi, x[t2+1], x[t2], (FIXP_SGL)1.0, (FIXP_SGL)0.0); + vr = xt2[1] >> 1; + vi = xt2[0] >> 1; - //cplxMultDiv2(&vr, &vi, x[t2+1], x[t2], (FIXP_SGL)1.0, (FIXP_SGL)0.0); - vr = xt2[1]>>1; - vi = xt2[0]>>1; + ur = xt1[0] >> 1; + ui = xt1[1] >> 1; - ur = xt1[0]>>1; - ui = xt1[1]>>1; + xt1[0] = ur + vr; + xt1[1] = ui - vi; - xt1[0] = ur+vr; - xt1[1] = ui-vi; + xt2[0] = ur - vr; + xt2[1] = ui + vi; - xt2[0] = ur-vr; - xt2[1] = ui+vi; - - xt1 = xt2 + mh; - } while ((r=r-(mh<<1)) != 0); - } - for(j=4; j>1); - FIXP_SPK cs; - int r = n; + xt1 = xt2 + mh; + } while ((r = r - (mh << 1)) != 0); + } + for (j = 4; j < mh; j += 4) { + FIXP_DBL *xt1 = x + (j >> 1); + FIXP_SPK cs; + int r = n; - pTrigData += trigstep; - cs = *pTrigData; + pTrigData += trigstep; + cs = *pTrigData; - do - { - FIXP_DBL *xt2 = xt1 + (mh<<1); - FIXP_DBL vr,vi,ur,ui; + do { + FIXP_DBL *xt2 = xt1 + (mh << 1); + FIXP_DBL vr, vi, ur, ui; - cplxMultDiv2(&vi, &vr, xt2[1], xt2[0], cs); + cplxMultDiv2(&vi, &vr, xt2[1], xt2[0], cs); - ur = xt1[0]>>1; - ui = xt1[1]>>1; + ur = xt1[0] >> 1; + ui = xt1[1] >> 1; - xt1[0] = ur+vr; - xt1[1] = ui+vi; + xt1[0] = ur + vr; + xt1[1] = ui + vi; - xt2[0] = ur-vr; - xt2[1] = ui-vi; + xt2[0] = ur - vr; + xt2[1] = ui - vi; - xt1 += mh; - xt2 += mh; + xt1 += mh; + xt2 += mh; - cplxMultDiv2(&vr, &vi, xt2[1], xt2[0], cs); + cplxMultDiv2(&vr, &vi, xt2[1], xt2[0], cs); - ur = xt1[0]>>1; - ui = xt1[1]>>1; + ur = xt1[0] >> 1; + ui = xt1[1] >> 1; - xt1[0] = ur+vr; - xt1[1] = ui-vi; + xt1[0] = ur + vr; + xt1[1] = ui - vi; - xt2[0] = ur-vr; - xt2[1] = ui+vi; + xt2[0] = ur - vr; + xt2[1] = ui + vi; - /* Same as above but for t1,t2 with j>mh/4 and thus cs swapped */ - xt1 = xt1 - (j); - xt2 = xt1 + (mh<<1); + /* Same as above but for t1,t2 with j>mh/4 and thus cs swapped */ + xt1 = xt1 - (j); + xt2 = xt1 + (mh << 1); - cplxMultDiv2(&vi, &vr, xt2[0], xt2[1], cs); + cplxMultDiv2(&vi, &vr, xt2[0], xt2[1], cs); - ur = xt1[0]>>1; - ui = xt1[1]>>1; + ur = xt1[0] >> 1; + ui = xt1[1] >> 1; - xt1[0] = ur+vr; - xt1[1] = ui-vi; + xt1[0] = ur + vr; + xt1[1] = ui - vi; - xt2[0] = ur-vr; - xt2[1] = ui+vi; + xt2[0] = ur - vr; + xt2[1] = ui + vi; - xt1 += mh; - xt2 += mh; + xt1 += mh; + xt2 += mh; - cplxMultDiv2(&vr, &vi, xt2[0], xt2[1], cs); + cplxMultDiv2(&vr, &vi, xt2[0], xt2[1], cs); - ur = xt1[0]>>1; - ui = xt1[1]>>1; + ur = xt1[0] >> 1; + ui = xt1[1] >> 1; - xt1[0] = ur-vr; - xt1[1] = ui-vi; + xt1[0] = ur - vr; + xt1[1] = ui - vi; - xt2[0] = ur+vr; - xt2[1] = ui+vi; + xt2[0] = ur + vr; + xt2[1] = ui + vi; - xt1 = xt2 + (j); - } while ((r=r-(mh<<1)) != 0); - } - { - FIXP_DBL *xt1 = x + (mh>>1); - int r = n; + xt1 = xt2 + (j); + } while ((r = r - (mh << 1)) != 0); + } + { + FIXP_DBL *xt1 = x + (mh >> 1); + int r = n; - do - { - FIXP_DBL *xt2 = xt1 + (mh<<1); - FIXP_DBL vr,vi,ur,ui; + do { + FIXP_DBL *xt2 = xt1 + (mh << 1); + FIXP_DBL vr, vi, ur, ui; - cplxMultDiv2(&vi, &vr, xt2[1], xt2[0], STC(0x5a82799a), STC(0x5a82799a)); + cplxMultDiv2(&vi, &vr, xt2[1], xt2[0], STC(0x5a82799a), + STC(0x5a82799a)); - ur = xt1[0]>>1; - ui = xt1[1]>>1; + ur = xt1[0] >> 1; + ui = xt1[1] >> 1; - xt1[0] = ur+vr; - xt1[1] = ui+vi; + xt1[0] = ur + vr; + xt1[1] = ui + vi; - xt2[0] = ur-vr; - xt2[1] = ui-vi; + xt2[0] = ur - vr; + xt2[1] = ui - vi; - xt1 += mh; - xt2 += mh; + xt1 += mh; + xt2 += mh; - cplxMultDiv2(&vr, &vi, xt2[1], xt2[0], STC(0x5a82799a), STC(0x5a82799a)); + cplxMultDiv2(&vr, &vi, xt2[1], xt2[0], STC(0x5a82799a), + STC(0x5a82799a)); - ur = xt1[0]>>1; - ui = xt1[1]>>1; + ur = xt1[0] >> 1; + ui = xt1[1] >> 1; - xt1[0] = ur+vr; - xt1[1] = ui-vi; + xt1[0] = ur + vr; + xt1[1] = ui - vi; - xt2[0] = ur-vr; - xt2[1] = ui+vi; + xt2[0] = ur - vr; + xt2[1] = ui + vi; - xt1 = xt2 + mh; - } while ((r=r-(mh<<1)) != 0); - } - } while (--ldm != 0); + xt1 = xt2 + mh; + } while ((r = r - (mh << 1)) != 0); + } + } while (--ldm != 0); } #endif /* if defined(FUNCTION_dit_fft) */ diff --git a/libFDK/src/arm/qmf_arm.cpp b/libFDK/src/arm/qmf_arm.cpp deleted file mode 100644 index fbeebfb..0000000 --- a/libFDK/src/arm/qmf_arm.cpp +++ /dev/null @@ -1,766 +0,0 @@ - -/* ----------------------------------------------------------------------------------------------------------- -Software License for The Fraunhofer FDK AAC Codec Library for Android - -© Copyright 1995 - 2015 Fraunhofer-Gesellschaft zur Förderung der angewandten Forschung e.V. - All rights reserved. - - 1. INTRODUCTION -The Fraunhofer FDK AAC Codec Library for Android ("FDK AAC Codec") is software that implements -the MPEG Advanced Audio Coding ("AAC") encoding and decoding scheme for digital audio. -This FDK AAC Codec software is intended to be used on a wide variety of Android devices. - -AAC's HE-AAC and HE-AAC v2 versions are regarded as today's most efficient general perceptual -audio codecs. AAC-ELD is considered the best-performing full-bandwidth communications codec by -independent studies and is widely deployed. AAC has been standardized by ISO and IEC as part -of the MPEG specifications. - -Patent licenses for necessary patent claims for the FDK AAC Codec (including those of Fraunhofer) -may be obtained through Via Licensing (www.vialicensing.com) or through the respective patent owners -individually for the purpose of encoding or decoding bit streams in products that are compliant with -the ISO/IEC MPEG audio standards. Please note that most manufacturers of Android devices already license -these patent claims through Via Licensing or directly from the patent owners, and therefore FDK AAC Codec -software may already be covered under those patent licenses when it is used for those licensed purposes only. - -Commercially-licensed AAC software libraries, including floating-point versions with enhanced sound quality, -are also available from Fraunhofer. Users are encouraged to check the Fraunhofer website for additional -applications information and documentation. - -2. COPYRIGHT LICENSE - -Redistribution and use in source and binary forms, with or without modification, are permitted without -payment of copyright license fees provided that you satisfy the following conditions: - -You must retain the complete text of this software license in redistributions of the FDK AAC Codec or -your modifications thereto in source code form. - -You must retain the complete text of this software license in the documentation and/or other materials -provided with redistributions of the FDK AAC Codec or your modifications thereto in binary form. -You must make available free of charge copies of the complete source code of the FDK AAC Codec and your -modifications thereto to recipients of copies in binary form. - -The name of Fraunhofer may not be used to endorse or promote products derived from this library without -prior written permission. - -You may not charge copyright license fees for anyone to use, copy or distribute the FDK AAC Codec -software or your modifications thereto. - -Your modified versions of the FDK AAC Codec must carry prominent notices stating that you changed the software -and the date of any change. For modified versions of the FDK AAC Codec, the term -"Fraunhofer FDK AAC Codec Library for Android" must be replaced by the term -"Third-Party Modified Version of the Fraunhofer FDK AAC Codec Library for Android." - -3. NO PATENT LICENSE - -NO EXPRESS OR IMPLIED LICENSES TO ANY PATENT CLAIMS, including without limitation the patents of Fraunhofer, -ARE GRANTED BY THIS SOFTWARE LICENSE. Fraunhofer provides no warranty of patent non-infringement with -respect to this software. - -You may use this FDK AAC Codec software or modifications thereto only for purposes that are authorized -by appropriate patent licenses. - -4. DISCLAIMER - -This FDK AAC Codec software is provided by Fraunhofer on behalf of the copyright holders and contributors -"AS IS" and WITHOUT ANY EXPRESS OR IMPLIED WARRANTIES, including but not limited to the implied warranties -of merchantability and fitness for a particular purpose. IN NO EVENT SHALL THE COPYRIGHT HOLDER OR -CONTRIBUTORS BE LIABLE for any direct, indirect, incidental, special, exemplary, or consequential damages, -including but not limited to procurement of substitute goods or services; loss of use, data, or profits, -or business interruption, however caused and on any theory of liability, whether in contract, strict -liability, or tort (including negligence), arising in any way out of the use of this software, even if -advised of the possibility of such damage. - -5. CONTACT INFORMATION - -Fraunhofer Institute for Integrated Circuits IIS -Attention: Audio and Multimedia Departments - FDK AAC LL -Am Wolfsmantel 33 -91058 Erlangen, Germany - -www.iis.fraunhofer.de/amm -amm-info@iis.fraunhofer.de ------------------------------------------------------------------------------------------------------------ */ - -#if (QMF_NO_POLY==5) - -#define FUNCTION_qmfForwardModulationLP_odd - -#ifdef FUNCTION_qmfForwardModulationLP_odd -static void -qmfForwardModulationLP_odd( HANDLE_QMF_FILTER_BANK anaQmf, /*!< Handle of Qmf Analysis Bank */ - const FIXP_QMF *timeIn, /*!< Time Signal */ - FIXP_QMF *rSubband ) /*!< Real Output */ -{ - int i; - int L = anaQmf->no_channels; - int M = L>>1; - int shift = (anaQmf->no_channels>>6) + 1; - int rSubband_e = 0; - - FIXP_QMF *rSubbandPtr0 = &rSubband[M+0]; /* runs with increment */ - FIXP_QMF *rSubbandPtr1 = &rSubband[M-1]; /* runs with decrement */ - FIXP_QMF *timeIn0 = (FIXP_DBL *) &timeIn[0]; /* runs with increment */ - FIXP_QMF *timeIn1 = (FIXP_DBL *) &timeIn[L]; /* runs with increment */ - FIXP_QMF *timeIn2 = (FIXP_DBL *) &timeIn[L-1]; /* runs with decrement */ - FIXP_QMF *timeIn3 = (FIXP_DBL *) &timeIn[2*L-1]; /* runs with decrement */ - - for (i = 0; i < M; i++) - { - *rSubbandPtr0++ = (*timeIn2-- >> 1) - (*timeIn0++ >> shift); - *rSubbandPtr1-- = (*timeIn1++ >> 1) + (*timeIn3-- >> shift); - } - - dct_IV(rSubband,L, &rSubband_e); -} -#endif /* FUNCTION_qmfForwardModulationLP_odd */ - - -/* NEON optimized QMF currently builts only with RVCT toolchain */ - -#if defined(__ARM_ARCH_6__) || defined(__ARM_ARCH_5TE__) - -#if (SAMPLE_BITS == 16) -#define FUNCTION_qmfAnaPrototypeFirSlot -#endif - -#ifdef FUNCTION_qmfAnaPrototypeFirSlot - -#if defined(__GNUC__) /* cppp replaced: elif */ - -inline INT SMULBB (const SHORT a, const LONG b) -{ - INT result ; - __asm__ ("smulbb %0, %1, %2" - : "=r" (result) - : "r" (a), "r" (b)) ; - return result ; -} -inline INT SMULBT (const SHORT a, const LONG b) -{ - INT result ; - __asm__ ("smulbt %0, %1, %2" - : "=r" (result) - : "r" (a), "r" (b)) ; - return result ; -} - -inline INT SMLABB(const LONG accu, const SHORT a, const LONG b) -{ - INT result ; - __asm__ ("smlabb %0, %1, %2,%3" - : "=r" (result) - : "r" (a), "r" (b), "r" (accu)) ; - return result; -} -inline INT SMLABT(const LONG accu, const SHORT a, const LONG b) -{ - INT result ; - __asm__ ("smlabt %0, %1, %2,%3" - : "=r" (result) - : "r" (a), "r" (b), "r" (accu)) ; - return result; -} -#endif /* compiler selection */ - - -void qmfAnaPrototypeFirSlot( FIXP_QMF *analysisBuffer, - int no_channels, /*!< Number channels of analysis filter */ - const FIXP_PFT *p_filter, - int p_stride, /*!< Stide of analysis filter */ - FIXP_QAS *RESTRICT pFilterStates - ) -{ - LONG *p_flt = (LONG *) p_filter; - LONG flt; - FIXP_QMF *RESTRICT pData_0 = analysisBuffer + 2*no_channels - 1; - FIXP_QMF *RESTRICT pData_1 = analysisBuffer; - - FIXP_QAS *RESTRICT sta_0 = (FIXP_QAS *)pFilterStates; - FIXP_QAS *RESTRICT sta_1 = (FIXP_QAS *)pFilterStates + (2*QMF_NO_POLY*no_channels) - 1; - - FIXP_DBL accu0, accu1; - FIXP_QAS sta0, sta1; - - int staStep1 = no_channels<<1; - int staStep2 = (no_channels<<3) - 1; /* Rewind one less */ - - if (p_stride == 1) - { - /* FIR filter 0 */ - flt = *p_flt++; - sta1 = *sta_1; sta_1 -= staStep1; - accu1 = SMULBB( sta1, flt); - sta1 = *sta_1; sta_1 -= staStep1; - accu1 = SMLABT( accu1, sta1, flt); - - flt = *p_flt++; - sta1 = *sta_1; sta_1 -= staStep1; - accu1 = SMLABB( accu1, sta1, flt); - sta1 = *sta_1; sta_1 -= staStep1; - accu1 = SMLABT( accu1, sta1, flt); - - flt = *p_flt++; - sta1 = *sta_1; sta_1 += staStep2; - accu1 = SMLABB( accu1, sta1, flt); - *pData_1++ = FX_DBL2FX_QMF(accu1<<1); - - /* FIR filters 1..63 127..65 or 1..31 63..33 */ - no_channels >>= 1; - for (; --no_channels; ) - { - sta0 = *sta_0; sta_0 += staStep1; /* 1,3,5, ... 29/61 */ - sta1 = *sta_1; sta_1 -= staStep1; - accu0 = SMULBT( sta0, flt); - accu1 = SMULBT( sta1, flt); - - flt = *p_flt++; - sta0 = *sta_0; sta_0 += staStep1; - sta1 = *sta_1; sta_1 -= staStep1; - accu0 = SMLABB( accu0, sta0, flt); - accu1 = SMLABB( accu1, sta1, flt); - - sta0 = *sta_0; sta_0 += staStep1; - sta1 = *sta_1; sta_1 -= staStep1; - accu0 = SMLABT( accu0, sta0, flt); - accu1 = SMLABT( accu1, sta1, flt); - - flt = *p_flt++; - sta0 = *sta_0; sta_0 += staStep1; - sta1 = *sta_1; sta_1 -= staStep1; - accu0 = SMLABB( accu0, sta0, flt); - accu1 = SMLABB( accu1, sta1, flt); - - sta0 = *sta_0; sta_0 -= staStep2; - sta1 = *sta_1; sta_1 += staStep2; - accu0 = SMLABT( accu0, sta0, flt); - accu1 = SMLABT( accu1, sta1, flt); - - *pData_0-- = FX_DBL2FX_QMF(accu0<<1); - *pData_1++ = FX_DBL2FX_QMF(accu1<<1); - - /* Same sequence as above, but mix B=bottom with T=Top */ - - flt = *p_flt++; - sta0 = *sta_0; sta_0 += staStep1; /* 2,4,6, ... 30/62 */ - sta1 = *sta_1; sta_1 -= staStep1; - accu0 = SMULBB( sta0, flt); - accu1 = SMULBB( sta1, flt); - - sta0 = *sta_0; sta_0 += staStep1; - sta1 = *sta_1; sta_1 -= staStep1; - accu0 = SMLABT( accu0, sta0, flt); - accu1 = SMLABT( accu1, sta1, flt); - - flt = *p_flt++; - sta0 = *sta_0; sta_0 += staStep1; - sta1 = *sta_1; sta_1 -= staStep1; - accu0 = SMLABB( accu0, sta0, flt); - accu1 = SMLABB( accu1, sta1, flt); - - sta0 = *sta_0; sta_0 += staStep1; - sta1 = *sta_1; sta_1 -= staStep1; - accu0 = SMLABT( accu0, sta0, flt); - accu1 = SMLABT( accu1, sta1, flt); - - flt = *p_flt++; - sta0 = *sta_0; sta_0 -= staStep2; - sta1 = *sta_1; sta_1 += staStep2; - accu0 = SMLABB( accu0, sta0, flt); - accu1 = SMLABB( accu1, sta1, flt); - - *pData_0-- = FX_DBL2FX_QMF(accu0<<1); - *pData_1++ = FX_DBL2FX_QMF(accu1<<1); - } - - /* FIR filter 31/63 and 33/65 */ - sta0 = *sta_0; sta_0 += staStep1; - sta1 = *sta_1; sta_1 -= staStep1; - accu0 = SMULBT( sta0, flt); - accu1 = SMULBT( sta1, flt); - - flt = *p_flt++; - sta0 = *sta_0; sta_0 += staStep1; - sta1 = *sta_1; sta_1 -= staStep1; - accu0 = SMLABB( accu0, sta0, flt); - accu1 = SMLABB( accu1, sta1, flt); - - sta0 = *sta_0; sta_0 += staStep1; - sta1 = *sta_1; sta_1 -= staStep1; - accu0 = SMLABT( accu0, sta0, flt); - accu1 = SMLABT( accu1, sta1, flt); - - flt = *p_flt++; - sta0 = *sta_0; sta_0 += staStep1; - sta1 = *sta_1; sta_1 -= staStep1; - accu0 = SMLABB( accu0, sta0, flt); - accu1 = SMLABB( accu1, sta1, flt); - - sta0 = *sta_0; sta_0 -= staStep2; - sta1 = *sta_1; sta_1 += staStep2; - accu0 = SMLABT( accu0, sta0, flt); - accu1 = SMLABT( accu1, sta1, flt); - - *pData_0-- = FX_DBL2FX_QMF(accu0<<1); - *pData_1++ = FX_DBL2FX_QMF(accu1<<1); - - /* FIR filter 32/64 */ - flt = *p_flt++; - sta0 = *sta_0; sta_0 += staStep1; - sta1 = *sta_1; sta_1 -= staStep1; - accu0 = SMULBB( sta0, flt); - accu1 = SMULBB( sta1, flt); - - sta0 = *sta_0; sta_0 += staStep1; - sta1 = *sta_1; sta_1 -= staStep1; - accu0 = SMLABT( accu0, sta0, flt); - accu1 = SMLABT( accu1, sta1, flt); - - flt = *p_flt++; - sta0 = *sta_0; sta_0 += staStep1; - sta1 = *sta_1; sta_1 -= staStep1; - accu0 = SMLABB( accu0, sta0, flt); - accu1 = SMLABB( accu1, sta1, flt); - - sta0 = *sta_0; sta_0 += staStep1; - sta1 = *sta_1; sta_1 -= staStep1; - accu0 = SMLABT( accu0, sta0, flt); - accu1 = SMLABT( accu1, sta1, flt); - - flt = *p_flt; - sta0 = *sta_0; - sta1 = *sta_1; - accu0 = SMLABB( accu0, sta0, flt); - accu1 = SMLABB( accu1, sta1, flt); - - *pData_0-- = FX_DBL2FX_QMF(accu0<<1); - *pData_1++ = FX_DBL2FX_QMF(accu1<<1); - } - else - { - int pfltStep = QMF_NO_POLY * (p_stride-1); - - flt = p_flt[0]; - sta1 = *sta_1; sta_1 -= staStep1; - accu1 = SMULBB( sta1, flt); - sta1 = *sta_1; sta_1 -= staStep1; - accu1 = SMLABT( accu1, sta1, flt); - - flt = p_flt[1]; - sta1 = *sta_1; sta_1 -= staStep1; - accu1 = SMLABB( accu1, sta1, flt); - sta1 = *sta_1; sta_1 -= staStep1; - accu1 = SMLABT( accu1, sta1, flt); - - flt = p_flt[2]; p_flt += pfltStep; - sta1 = *sta_1; sta_1 += staStep2; - accu1 = SMLABB( accu1, sta1, flt); - *pData_1++ = FX_DBL2FX_QMF(accu1<<1); - - /* FIR filters 1..63 127..65 or 1..31 63..33 */ - for (; --no_channels; ) - { - flt = p_flt[0]; - sta0 = *sta_0; sta_0 += staStep1; - sta1 = *sta_1; sta_1 -= staStep1; - accu0 = SMULBB( sta0, flt); - accu1 = SMULBB( sta1, flt); - - sta0 = *sta_0; sta_0 += staStep1; - sta1 = *sta_1; sta_1 -= staStep1; - accu0 = SMLABT( accu0, sta0, flt); - accu1 = SMLABT( accu1, sta1, flt); - - flt = p_flt[1]; - sta0 = *sta_0; sta_0 += staStep1; - sta1 = *sta_1; sta_1 -= staStep1; - accu0 = SMLABB( accu0, sta0, flt); - accu1 = SMLABB( accu1, sta1, flt); - - sta0 = *sta_0; sta_0 += staStep1; - sta1 = *sta_1; sta_1 -= staStep1; - accu0 = SMLABT( accu0, sta0, flt); - accu1 = SMLABT( accu1, sta1, flt); - - flt = p_flt[2]; p_flt += pfltStep; - sta0 = *sta_0; sta_0 -= staStep2; - sta1 = *sta_1; sta_1 += staStep2; - accu0 = SMLABB( accu0, sta0, flt); - accu1 = SMLABB( accu1, sta1, flt); - - *pData_0-- = FX_DBL2FX_QMF(accu0<<1); - *pData_1++ = FX_DBL2FX_QMF(accu1<<1); - } - - /* FIR filter 32/64 */ - flt = p_flt[0]; - sta0 = *sta_0; sta_0 += staStep1; - accu0 = SMULBB( sta0, flt); - sta0 = *sta_0; sta_0 += staStep1; - accu0 = SMLABT( accu0, sta0, flt); - - flt = p_flt[1]; - sta0 = *sta_0; sta_0 += staStep1; - accu0 = SMLABB( accu0, sta0, flt); - sta0 = *sta_0; sta_0 += staStep1; - accu0 = SMLABT( accu0, sta0, flt); - - flt = p_flt[2]; - sta0 = *sta_0; - accu0 = SMLABB( accu0, sta0, flt); - *pData_0-- = FX_DBL2FX_QMF(accu0<<1); - } -} -#endif /* FUNCTION_qmfAnaPrototypeFirSlot */ -#endif /* #if defined(__CC_ARM) && defined(__ARM_ARCH_6__) */ - -#if ( defined(__ARM_ARCH_5TE__) && (SAMPLE_BITS == 16) ) && !defined(QMF_TABLE_FULL) - -#define FUNCTION_qmfSynPrototypeFirSlot - -#if defined(FUNCTION_qmfSynPrototypeFirSlot) - -#if defined(__GNUC__) /* cppp replaced: elif */ - -inline INT SMULWB (const LONG a, const LONG b) -{ - INT result ; - __asm__ ("smulwb %0, %1, %2" - : "=r" (result) - : "r" (a), "r" (b)) ; - - return result ; -} -inline INT SMULWT (const LONG a, const LONG b) -{ - INT result ; - __asm__ ("smulwt %0, %1, %2" - : "=r" (result) - : "r" (a), "r" (b)) ; - - return result ; -} - -inline INT SMLAWB(const LONG accu, const LONG a, const LONG b) -{ - INT result; - asm("smlawb %0, %1, %2, %3 " - : "=r" (result) - : "r" (a), "r" (b), "r" (accu) ); - return result ; -} - -inline INT SMLAWT(const LONG accu, const LONG a, const LONG b) -{ - INT result; - asm("smlawt %0, %1, %2, %3 " - : "=r" (result) - : "r" (a), "r" (b), "r" (accu) ); - return result ; -} - -#endif /* ARM compiler selector */ - - -static void qmfSynPrototypeFirSlot1_filter(FIXP_QMF *RESTRICT realSlot, - FIXP_QMF *RESTRICT imagSlot, - const FIXP_DBL *RESTRICT p_flt, - FIXP_QSS *RESTRICT sta, - FIXP_DBL *pMyTimeOut, - int no_channels) -{ - /* This code was the base for the above listed assembler sequence */ - /* It can be used for debugging purpose or further optimizations */ - const FIXP_DBL *RESTRICT p_fltm = p_flt + 155; - - do - { - FIXP_DBL result; - FIXP_DBL A, B, real, imag, sta0; - - real = *--realSlot; - imag = *--imagSlot; - B = p_flt[4]; /* Bottom=[8] Top=[9] */ - A = p_fltm[3]; /* Bottom=[316] Top=[317] */ - sta0 = sta[0]; /* save state[0] */ - sta[0] = SMLAWT( sta[1], imag, B ); sta++; /* index=9...........319 */ - sta[0] = SMLAWB( sta[1], real, A ); sta++; /* index=316...........6 */ - sta[0] = SMLAWB( sta[1], imag, B ); sta++; /* index=8,18, ...318 */ - B = p_flt[3]; /* Bottom=[6] Top=[7] */ - sta[0] = SMLAWT( sta[1], real, A ); sta++; /* index=317...........7 */ - A = p_fltm[4]; /* Bottom=[318] Top=[319] */ - sta[0] = SMLAWT( sta[1], imag, B ); sta++; /* index=7...........317 */ - sta[0] = SMLAWB( sta[1], real, A ); sta++; /* index=318...........8 */ - sta[0] = SMLAWB( sta[1], imag, B ); sta++; /* index=6...........316 */ - B = p_flt[2]; /* Bottom=[X] Top=[5] */ - sta[0] = SMLAWT( sta[1], real, A ); sta++; /* index=9...........319 */ - A = p_fltm[2]; /* Bottom=[X] Top=[315] */ - sta[0] = SMULWT( imag, B ); sta++; /* index=5,15, ... 315 */ - result = SMLAWT( sta0, real, A ); /* index=315...........5 */ - - pMyTimeOut[0] = result; pMyTimeOut++; - - real = *--realSlot; - imag = *--imagSlot; - A = p_fltm[0]; /* Bottom=[310] Top=[311] */ - B = p_flt[7]; /* Bottom=[14] Top=[15] */ - result = SMLAWB( sta[0], real, A ); /* index=310...........0 */ - sta[0] = SMLAWB( sta[1], imag, B ); sta++; /* index=14..........324 */ - pMyTimeOut[0] = result; pMyTimeOut++; - B = p_flt[6]; /* Bottom=[12] Top=[13] */ - sta[0] = SMLAWT( sta[1], real, A ); sta++; /* index=311...........1 */ - A = p_fltm[1]; /* Bottom=[312] Top=[313] */ - sta[0] = SMLAWT( sta[1], imag, B ); sta++; /* index=13..........323 */ - sta[0] = SMLAWB( sta[1], real, A ); sta++; /* index=312...........2 */ - sta[0] = SMLAWB( sta[1], imag, B ); sta++; /* index=12..........322 */ - sta[0] = SMLAWT( sta[1], real, A ); sta++; /* index=313...........3 */ - A = p_fltm[2]; /* Bottom=[314] Top=[315] */ - B = p_flt[5]; /* Bottom=[10] Top=[11] */ - sta[0] = SMLAWT( sta[1], imag, B ); sta++; /* index=11..........321 */ - sta[0] = SMLAWB( sta[1], real, A ); sta++; /* index=314...........4 */ - sta[0] = SMULWB( imag, B ); sta++; /* index=10..........320 */ - - - p_flt += 5; - p_fltm -= 5; - } - while ((--no_channels) != 0); - -} - - - -INT qmfSynPrototypeFirSlot2( - HANDLE_QMF_FILTER_BANK qmf, - FIXP_QMF *RESTRICT realSlot, /*!< Input: Pointer to real Slot */ - FIXP_QMF *RESTRICT imagSlot, /*!< Input: Pointer to imag Slot */ - INT_PCM *RESTRICT timeOut, /*!< Time domain data */ - INT stride /*!< Time output buffer stride factor*/ - ) -{ - FIXP_QSS *RESTRICT sta = (FIXP_QSS*)qmf->FilterStates; - int no_channels = qmf->no_channels; - int scale = ((DFRACT_BITS-SAMPLE_BITS)-1-qmf->outScalefactor); - - /* We map an arry of 16-bit values upon an array of 2*16-bit values to read 2 values in one shot */ - const FIXP_DBL *RESTRICT p_flt = (FIXP_DBL *) qmf->p_filter; /* low=[0], high=[1] */ - const FIXP_DBL *RESTRICT p_fltm = (FIXP_DBL *) qmf->p_filter + 155; /* low=[310], high=[311] */ - - FDK_ASSERT(SAMPLE_BITS-1-qmf->outScalefactor >= 0); // (DFRACT_BITS-SAMPLE_BITS)-1-qmf->outScalefactor >= 0); - FDK_ASSERT(qmf->p_stride==2 && qmf->no_channels == 32); - - FDK_ASSERT((no_channels&3) == 0); /* should be a multiple of 4 */ - - realSlot += no_channels-1; // ~~"~~ - imagSlot += no_channels-1; // no_channels-1 .. 0 - - FIXP_DBL MyTimeOut[32]; - FIXP_DBL *pMyTimeOut = &MyTimeOut[0]; - - for (no_channels = no_channels; no_channels--;) - { - FIXP_DBL result; - FIXP_DBL A, B, real, imag; - - real = *realSlot--; - imag = *imagSlot--; - A = p_fltm[0]; /* Bottom=[310] Top=[311] */ - B = p_flt[7]; /* Bottom=[14] Top=[15] */ - result = SMLAWB( sta[0], real, A ); /* index=310...........0 */ - sta[0] = SMLAWB( sta[1], imag, B ); sta++; /* index=14..........324 */ - B = p_flt[6]; /* Bottom=[12] Top=[13] */ - sta[0] = SMLAWT( sta[1], real, A ); sta++; /* index=311...........1 */ - A = p_fltm[1]; /* Bottom=[312] Top=[313] */ - sta[0] = SMLAWT( sta[1], imag, B ); sta++; /* index=13..........323 */ - sta[0] = SMLAWB( sta[1], real, A ); sta++; /* index=312...........2 */ - sta[0] = SMLAWB( sta[1], imag, B ); sta++; /* index=12..........322 */ - sta[0] = SMLAWT( sta[1], real, A ); sta++; /* index=313...........3 */ - A = p_fltm[2]; /* Bottom=[314] Top=[315] */ - B = p_flt[5]; /* Bottom=[10] Top=[11] */ - sta[0] = SMLAWT( sta[1], imag, B ); sta++; /* index=11..........321 */ - sta[0] = SMLAWB( sta[1], real, A ); sta++; /* index=314...........4 */ - sta[0] = SMULWB( imag, B ); sta++; /* index=10..........320 */ - - pMyTimeOut[0] = result; pMyTimeOut++; - - p_fltm -= 5; - p_flt += 5; - } - - pMyTimeOut = &MyTimeOut[0]; -#if (SAMPLE_BITS == 16) - const FIXP_DBL max_pos = (FIXP_DBL) 0x00007FFF << scale; - const FIXP_DBL max_neg = (FIXP_DBL) 0xFFFF8001 << scale; -#else - scale = -scale; - const FIXP_DBL max_pos = (FIXP_DBL) 0x7FFFFFFF >> scale; - const FIXP_DBL max_neg = (FIXP_DBL) 0x80000001 >> scale; -#endif - const FIXP_DBL add_neg = (1 << scale) - 1; - - no_channels = qmf->no_channels; - - timeOut += no_channels*stride; - - FDK_ASSERT(scale >= 0); - - if (qmf->outGain != 0x80000000) - { - FIXP_DBL gain = qmf->outGain; - for (no_channels>>=2; no_channels--;) - { - FIXP_DBL result1, result2; - - result1 = pMyTimeOut[0]; pMyTimeOut++; - result2 = pMyTimeOut[0]; pMyTimeOut++; - - result1 = fMult(result1,gain); - timeOut -= stride; - if (result1 < 0) result1 += add_neg; - if (result1 < max_neg) result1 = max_neg; - if (result1 > max_pos) result1 = max_pos; -#if (SAMPLE_BITS == 16) - timeOut[0] = result1 >> scale; -#else - timeOut[0] = result1 << scale; -#endif - - result2 = fMult(result2,gain); - timeOut -= stride; - if (result2 < 0) result2 += add_neg; - if (result2 < max_neg) result2 = max_neg; - if (result2 > max_pos) result2 = max_pos; -#if (SAMPLE_BITS == 16) - timeOut[0] = result2 >> scale; -#else - timeOut[0] = result2 << scale; -#endif - - result1 = pMyTimeOut[0]; pMyTimeOut++; - result2 = pMyTimeOut[0]; pMyTimeOut++; - - result1 = fMult(result1,gain); - timeOut -= stride; - if (result1 < 0) result1 += add_neg; - if (result1 < max_neg) result1 = max_neg; - if (result1 > max_pos) result1 = max_pos; -#if (SAMPLE_BITS == 16) - timeOut[0] = result1 >> scale; -#else - timeOut[0] = result1 << scale; -#endif - - result2 = fMult(result2,gain); - timeOut -= stride; - if (result2 < 0) result2 += add_neg; - if (result2 < max_neg) result2 = max_neg; - if (result2 > max_pos) result2 = max_pos; -#if (SAMPLE_BITS == 16) - timeOut[0] = result2 >> scale; -#else - timeOut[0] = result2 << scale; -#endif - } - } - else - { - for (no_channels>>=2; no_channels--;) - { - FIXP_DBL result1, result2; - result1 = pMyTimeOut[0]; pMyTimeOut++; - result2 = pMyTimeOut[0]; pMyTimeOut++; - timeOut -= stride; - if (result1 < 0) result1 += add_neg; - if (result1 < max_neg) result1 = max_neg; - if (result1 > max_pos) result1 = max_pos; -#if (SAMPLE_BITS == 16) - timeOut[0] = result1 >> scale; -#else - timeOut[0] = result1 << scale; -#endif - - timeOut -= stride; - if (result2 < 0) result2 += add_neg; - if (result2 < max_neg) result2 = max_neg; - if (result2 > max_pos) result2 = max_pos; -#if (SAMPLE_BITS == 16) - timeOut[0] = result2 >> scale; -#else - timeOut[0] = result2 << scale; -#endif - - result1 = pMyTimeOut[0]; pMyTimeOut++; - result2 = pMyTimeOut[0]; pMyTimeOut++; - timeOut -= stride; - if (result1 < 0) result1 += add_neg; - if (result1 < max_neg) result1 = max_neg; - if (result1 > max_pos) result1 = max_pos; -#if (SAMPLE_BITS == 16) - timeOut[0] = result1 >> scale; -#else - timeOut[0] = result1 << scale; -#endif - - timeOut -= stride; - if (result2 < 0) result2 += add_neg; - if (result2 < max_neg) result2 = max_neg; - if (result2 > max_pos) result2 = max_pos; -#if (SAMPLE_BITS == 16) - timeOut[0] = result2 >> scale; -#else - timeOut[0] = result2 << scale; -#endif - } - } - return 0; -} - -static -void qmfSynPrototypeFirSlot_fallback( HANDLE_QMF_FILTER_BANK qmf, - FIXP_DBL *realSlot, /*!< Input: Pointer to real Slot */ - FIXP_DBL *imagSlot, /*!< Input: Pointer to imag Slot */ - INT_PCM *timeOut, /*!< Time domain data */ - const int stride - ); - -/*! - \brief Perform Synthesis Prototype Filtering on a single slot of input data. - - The filter takes 2 * #MAX_SYNTHESIS_CHANNELS of input data and - generates #MAX_SYNTHESIS_CHANNELS time domain output samples. -*/ - -static -void qmfSynPrototypeFirSlot( HANDLE_QMF_FILTER_BANK qmf, - FIXP_DBL *realSlot, /*!< Input: Pointer to real Slot */ - FIXP_DBL *imagSlot, /*!< Input: Pointer to imag Slot */ - INT_PCM *timeOut, /*!< Time domain data */ - const int stride - ) -{ - INT err = -1; - - switch (qmf->p_stride) { - case 2: - err = qmfSynPrototypeFirSlot2(qmf, realSlot, imagSlot, timeOut, stride); - break; - default: - err = -1; - } - - /* fallback if configuration not available or failed */ - if(err!=0) { - qmfSynPrototypeFirSlot_fallback(qmf, realSlot, imagSlot, timeOut, stride); - } -} -#endif /* FUNCTION_qmfSynPrototypeFirSlot */ - -#endif /* ( defined(__CC_ARM) && defined(__ARM_ARCH_5TE__) && (SAMPLE_BITS == 16) ) && !defined(QMF_TABLE_FULL) */ - - - -/* #####################################################################################*/ - - - -#endif /* (QMF_NO_POLY==5) */ - diff --git a/libFDK/src/arm/scale_arm.cpp b/libFDK/src/arm/scale_arm.cpp index e2837ef..92c9edc 100644 --- a/libFDK/src/arm/scale_arm.cpp +++ b/libFDK/src/arm/scale_arm.cpp @@ -1,74 +1,85 @@ - -/* ----------------------------------------------------------------------------------------------------------- +/* ----------------------------------------------------------------------------- Software License for The Fraunhofer FDK AAC Codec Library for Android -© Copyright 1995 - 2013 Fraunhofer-Gesellschaft zur Förderung der angewandten Forschung e.V. - All rights reserved. +© Copyright 1995 - 2018 Fraunhofer-Gesellschaft zur Förderung der angewandten +Forschung e.V. All rights reserved. 1. INTRODUCTION -The Fraunhofer FDK AAC Codec Library for Android ("FDK AAC Codec") is software that implements -the MPEG Advanced Audio Coding ("AAC") encoding and decoding scheme for digital audio. -This FDK AAC Codec software is intended to be used on a wide variety of Android devices. - -AAC's HE-AAC and HE-AAC v2 versions are regarded as today's most efficient general perceptual -audio codecs. AAC-ELD is considered the best-performing full-bandwidth communications codec by -independent studies and is widely deployed. AAC has been standardized by ISO and IEC as part -of the MPEG specifications. - -Patent licenses for necessary patent claims for the FDK AAC Codec (including those of Fraunhofer) -may be obtained through Via Licensing (www.vialicensing.com) or through the respective patent owners -individually for the purpose of encoding or decoding bit streams in products that are compliant with -the ISO/IEC MPEG audio standards. Please note that most manufacturers of Android devices already license -these patent claims through Via Licensing or directly from the patent owners, and therefore FDK AAC Codec -software may already be covered under those patent licenses when it is used for those licensed purposes only. - -Commercially-licensed AAC software libraries, including floating-point versions with enhanced sound quality, -are also available from Fraunhofer. Users are encouraged to check the Fraunhofer website for additional -applications information and documentation. +The Fraunhofer FDK AAC Codec Library for Android ("FDK AAC Codec") is software +that implements the MPEG Advanced Audio Coding ("AAC") encoding and decoding +scheme for digital audio. This FDK AAC Codec software is intended to be used on +a wide variety of Android devices. + +AAC's HE-AAC and HE-AAC v2 versions are regarded as today's most efficient +general perceptual audio codecs. AAC-ELD is considered the best-performing +full-bandwidth communications codec by independent studies and is widely +deployed. AAC has been standardized by ISO and IEC as part of the MPEG +specifications. + +Patent licenses for necessary patent claims for the FDK AAC Codec (including +those of Fraunhofer) may be obtained through Via Licensing +(www.vialicensing.com) or through the respective patent owners individually for +the purpose of encoding or decoding bit streams in products that are compliant +with the ISO/IEC MPEG audio standards. Please note that most manufacturers of +Android devices already license these patent claims through Via Licensing or +directly from the patent owners, and therefore FDK AAC Codec software may +already be covered under those patent licenses when it is used for those +licensed purposes only. + +Commercially-licensed AAC software libraries, including floating-point versions +with enhanced sound quality, are also available from Fraunhofer. Users are +encouraged to check the Fraunhofer website for additional applications +information and documentation. 2. COPYRIGHT LICENSE -Redistribution and use in source and binary forms, with or without modification, are permitted without -payment of copyright license fees provided that you satisfy the following conditions: +Redistribution and use in source and binary forms, with or without modification, +are permitted without payment of copyright license fees provided that you +satisfy the following conditions: -You must retain the complete text of this software license in redistributions of the FDK AAC Codec or -your modifications thereto in source code form. +You must retain the complete text of this software license in redistributions of +the FDK AAC Codec or your modifications thereto in source code form. -You must retain the complete text of this software license in the documentation and/or other materials -provided with redistributions of the FDK AAC Codec or your modifications thereto in binary form. -You must make available free of charge copies of the complete source code of the FDK AAC Codec and your +You must retain the complete text of this software license in the documentation +and/or other materials provided with redistributions of the FDK AAC Codec or +your modifications thereto in binary form. You must make available free of +charge copies of the complete source code of the FDK AAC Codec and your modifications thereto to recipients of copies in binary form. -The name of Fraunhofer may not be used to endorse or promote products derived from this library without -prior written permission. +The name of Fraunhofer may not be used to endorse or promote products derived +from this library without prior written permission. -You may not charge copyright license fees for anyone to use, copy or distribute the FDK AAC Codec -software or your modifications thereto. +You may not charge copyright license fees for anyone to use, copy or distribute +the FDK AAC Codec software or your modifications thereto. -Your modified versions of the FDK AAC Codec must carry prominent notices stating that you changed the software -and the date of any change. For modified versions of the FDK AAC Codec, the term -"Fraunhofer FDK AAC Codec Library for Android" must be replaced by the term -"Third-Party Modified Version of the Fraunhofer FDK AAC Codec Library for Android." +Your modified versions of the FDK AAC Codec must carry prominent notices stating +that you changed the software and the date of any change. For modified versions +of the FDK AAC Codec, the term "Fraunhofer FDK AAC Codec Library for Android" +must be replaced by the term "Third-Party Modified Version of the Fraunhofer FDK +AAC Codec Library for Android." 3. NO PATENT LICENSE -NO EXPRESS OR IMPLIED LICENSES TO ANY PATENT CLAIMS, including without limitation the patents of Fraunhofer, -ARE GRANTED BY THIS SOFTWARE LICENSE. Fraunhofer provides no warranty of patent non-infringement with -respect to this software. +NO EXPRESS OR IMPLIED LICENSES TO ANY PATENT CLAIMS, including without +limitation the patents of Fraunhofer, ARE GRANTED BY THIS SOFTWARE LICENSE. +Fraunhofer provides no warranty of patent non-infringement with respect to this +software. -You may use this FDK AAC Codec software or modifications thereto only for purposes that are authorized -by appropriate patent licenses. +You may use this FDK AAC Codec software or modifications thereto only for +purposes that are authorized by appropriate patent licenses. 4. DISCLAIMER -This FDK AAC Codec software is provided by Fraunhofer on behalf of the copyright holders and contributors -"AS IS" and WITHOUT ANY EXPRESS OR IMPLIED WARRANTIES, including but not limited to the implied warranties -of merchantability and fitness for a particular purpose. IN NO EVENT SHALL THE COPYRIGHT HOLDER OR -CONTRIBUTORS BE LIABLE for any direct, indirect, incidental, special, exemplary, or consequential damages, -including but not limited to procurement of substitute goods or services; loss of use, data, or profits, -or business interruption, however caused and on any theory of liability, whether in contract, strict -liability, or tort (including negligence), arising in any way out of the use of this software, even if -advised of the possibility of such damage. +This FDK AAC Codec software is provided by Fraunhofer on behalf of the copyright +holders and contributors "AS IS" and WITHOUT ANY EXPRESS OR IMPLIED WARRANTIES, +including but not limited to the implied warranties of merchantability and +fitness for a particular purpose. IN NO EVENT SHALL THE COPYRIGHT HOLDER OR +CONTRIBUTORS BE LIABLE for any direct, indirect, incidental, special, exemplary, +or consequential damages, including but not limited to procurement of substitute +goods or services; loss of use, data, or profits, or business interruption, +however caused and on any theory of liability, whether in contract, strict +liability, or tort (including negligence), arising in any way out of the use of +this software, even if advised of the possibility of such damage. 5. CONTACT INFORMATION @@ -79,40 +90,34 @@ Am Wolfsmantel 33 www.iis.fraunhofer.de/amm amm-info@iis.fraunhofer.de ------------------------------------------------------------------------------------------------------------ */ +----------------------------------------------------------------------------- */ + +/******************* Library for basic calculation routines ******************** -/*************************** Fraunhofer IIS FDK Tools ********************** + Author(s): Arthur Tritthart - Author(s): Arthur Tritthart Description: Scaling operations for ARM -******************************************************************************/ +*******************************************************************************/ /* prevent multiple inclusion with re-definitions */ #ifndef __INCLUDE_SCALE_ARM__ #define __INCLUDE_SCALE_ARM__ +#if !defined(FUNCTION_scaleValuesWithFactor_DBL) #define FUNCTION_scaleValuesWithFactor_DBL - SCALE_INLINE -void scaleValuesWithFactor( - FIXP_DBL *vector, - FIXP_DBL factor, - INT len, - INT scalefactor - ) -{ +void scaleValuesWithFactor(FIXP_DBL *vector, FIXP_DBL factor, INT len, + INT scalefactor) { /* This code combines the fMult with the scaling */ /* It performs a fMultDiv2 and increments shift by 1 */ int shift = scalefactor + 1; FIXP_DBL *mySpec = vector; - shift = fixmin_I(shift,(INT)DFRACT_BITS-1); + shift = fixmin_I(shift, (INT)DFRACT_BITS - 1); - if (shift >= 0) - { - for (int i=0; i<(len>>2); i++) - { + if (shift >= 0) { + for (int i = 0; i < (len >> 2); i++) { FIXP_DBL tmp0 = mySpec[0]; FIXP_DBL tmp1 = mySpec[1]; FIXP_DBL tmp2 = mySpec[2]; @@ -130,19 +135,15 @@ void scaleValuesWithFactor( *mySpec++ = tmp2; *mySpec++ = tmp3; } - for (int i=len&3; i--;) - { + for (int i = len & 3; i--;) { FIXP_DBL tmp0 = mySpec[0]; tmp0 = fMultDiv2(tmp0, factor); tmp0 <<= shift; *mySpec++ = tmp0; } - } - else - { + } else { shift = -shift; - for (int i=0; i<(len>>2); i++) - { + for (int i = 0; i < (len >> 2); i++) { FIXP_DBL tmp0 = mySpec[0]; FIXP_DBL tmp1 = mySpec[1]; FIXP_DBL tmp2 = mySpec[2]; @@ -160,8 +161,7 @@ void scaleValuesWithFactor( *mySpec++ = tmp2; *mySpec++ = tmp3; } - for (int i=len&3; i--;) - { + for (int i = len & 3; i--;) { FIXP_DBL tmp0 = mySpec[0]; tmp0 = fMultDiv2(tmp0, factor); tmp0 >>= shift; @@ -169,5 +169,6 @@ void scaleValuesWithFactor( } } } +#endif /* #if !defined(FUNCTION_scaleValuesWithFactor_DBL) */ #endif /* #ifndef __INCLUDE_SCALE_ARM__ */ -- cgit v1.2.3