@//
@//  Copyright (c) 2013 The WebRTC project authors. All Rights Reserved.
@//
@//  Use of this source code is governed by a BSD-style license
@//  that can be found in the LICENSE file in the root of the source
@//  tree. An additional intellectual property rights grant can be found
@//  in the file PATENTS.  All contributing project authors may
@//  be found in the AUTHORS file in the root of the source tree.
@//
@//  This file was originally licensed as follows. It has been
@//  relicensed with permission from the copyright holders.
@//

@// 
@// File Name:  armSP_FFT_CToC_SC32_Radix4_unsafe_s.s
@// OpenMAX DL: v1.0.2
@// Last Modified Revision:   7767
@// Last Modified Date:       Thu, 27 Sep 2007
@// 
@// (c) Copyright 2007-2008 ARM Limited. All Rights Reserved.
@// 
@// 
@//
@// Description:
@// Compute a Radix 4 FFT stage for a N point complex signal
@// 


@// Include standard headers

#include "dl/api/armCOMM_s.h"
#include "dl/api/omxtypes_s.h"
        
        
@// Import symbols required from other files
@// (For example tables)
    
        
@// Set debugging level        
@//DEBUG_ON    SETL {TRUE}


@// Guarding implementation by the processor name
    
    
@// Guarding implementation by the processor name
    
    
@// Import symbols required from other files
@// (For example tables)
    
    
@//Input Registers

#define pSrc		r0
#define pDst		r2
#define pTwiddle	r1
#define subFFTNum	r6
#define subFFTSize	r7


@//Output Registers


@//Local Scratch Registers

#define grpCount	r3
#define pointStep	r4
#define outPointStep	r5
#define stepTwiddle	r12
#define setCount	r14
#define srcStep		r8
#define setStep		r9
#define dstStep		r10
#define twStep		r11
#define t1		r3

@// Neon Registers

#define dW1	D0.S32
#define dW2	D1.S32
#define dW3	D2.S32   

#define dXr0	D4.S32
#define dXi0	D5.S32
#define dXr1	D6.S32
#define dXi1	D7.S32
#define dXr2	D8.S32
#define dXi2	D9.S32
#define dXr3	D10.S32
#define dXi3	D11.S32
#define dYr0	D12.S32
#define dYi0	D13.S32
#define dYr1	D14.S32
#define dYi1	D15.S32
#define dYr2	D16.S32
#define dYi2	D17.S32
#define dYr3	D18.S32
#define dYi3	D19.S32
#define qT0	Q8.S64   
#define qT1	Q9.S64
#define qT2	Q6.S64
#define qT3	Q7.S64

#define dZr0	D20.S32
#define dZi0	D21.S32
#define dZr1	D22.S32
#define dZi1	D23.S32
#define dZr2	D24.S32
#define dZi2	D25.S32
#define dZr3	D26.S32
#define dZi3	D27.S32

#define qY0	Q6.S32
#define qY1	Q7.S32
#define qY2	Q8.S32
#define qY3	Q9.S32   
#define qX0	Q2.S32
#define qZ0	Q10.S32
#define qZ1	Q11.S32
#define qZ2	Q12.S32
#define qZ3	Q13.S32

        
        .MACRO FFTSTAGE scaled, inverse , name
        
        @// Define stack arguments
        
        
        @// Update grpCount and grpSize rightaway inorder to reuse pGrpCount and pGrpSize regs
        
        LSL     grpCount,subFFTSize,#2
        LSR     subFFTNum,subFFTNum,#2  
        MOV     subFFTSize,grpCount
        
        VLD1     dW1,[pTwiddle]                             @//[wi | wr]
        @// pT0+1 increments pT0 by 8 bytes
        @// pT0+pointStep = increment of 8*pointStep bytes = 2*grpSize bytes
        MOV     pointStep,subFFTNum,LSL #1
        
        
        @// pOut0+1 increments pOut0 by 8 bytes
        @// pOut0+outPointStep == increment of 8*outPointStep bytes = 2*size bytes
        
        MOV     stepTwiddle,#0
        VLD1     dW2,[pTwiddle]                             @//[wi | wr]
        SMULBB  outPointStep,grpCount,pointStep  
        LSL     pointStep,pointStep,#2                      @// 2*grpSize    
        
        VLD1     dW3,[pTwiddle]                             @//[wi | wr]
        MOV     srcStep,pointStep,LSL #1                    @// srcStep = 2*pointStep
        ADD     setStep,srcStep,pointStep                   @// setStep = 3*pointStep
        @//RSB     setStep,setStep,#16                         @// setStep = - 3*pointStep+16
        RSB     setStep,setStep,#0                         @// setStep = - 3*pointStep
        SUB     srcStep,srcStep,#16                         @// srcStep = 2*pointStep-16
        
        MOV     dstStep,outPointStep,LSL #1
        ADD     dstStep,dstStep,outPointStep                @// dstStep = 3*outPointStep
        RSB     dstStep,dstStep,#16                          @// dstStep = - 3*outPointStep+16
        

grpLoop\name :	
        
        VLD2    {dXr0,dXi0},[pSrc],pointStep                @//  data[0]
        ADD      stepTwiddle,stepTwiddle,pointStep
        VLD2    {dXr1,dXi1},[pSrc],pointStep                @//  data[1]
        ADD      pTwiddle,pTwiddle,stepTwiddle              @// set pTwiddle to the first point
        VLD2    {dXr2,dXi2},[pSrc],pointStep                @//  data[2]
        MOV      twStep,stepTwiddle,LSL #2
        
        VLD2    {dXr3,dXi3},[pSrc],setStep                  @//  data[3] & update pSrc for the next set
        SUB      twStep,stepTwiddle,twStep                  @// twStep = -3*stepTwiddle
        
        MOV      setCount,pointStep,LSR #3
        ADD     pSrc,pSrc,#16                         @// set pSrc to data[0] of the next set
        ADD     pSrc,pSrc,pointStep                   @// increment to data[1] of the next set
       
        
        @// Loop on the sets

setLoop\name :	
        
        
        SUBS    setCount,setCount,#2                    @// decrement the loop counter
        
        .ifeqs  "\inverse", "TRUE"
            VMULL   qT0,dXr1,dW1[0]
            VMLAL   qT0,dXi1,dW1[1]                       @// real part
            VMULL   qT1,dXi1,dW1[0]
            VMLSL   qT1,dXr1,dW1[1]                       @// imag part
            
        .else
            VMULL   qT0,dXr1,dW1[0]
            VMLSL   qT0,dXi1,dW1[1]                       @// real part
            VMULL   qT1,dXi1,dW1[0]
            VMLAL   qT1,dXr1,dW1[1]                       @// imag part
        
        .endif
        
        VLD2    {dXr1,dXi1},[pSrc],pointStep              @//  data[1] for next iteration
        
        .ifeqs  "\inverse", "TRUE"
            VMULL   qT2,dXr2,dW2[0]
            VMLAL   qT2,dXi2,dW2[1]                       @// real part
            VMULL   qT3,dXi2,dW2[0]
            VMLSL   qT3,dXr2,dW2[1]                       @// imag part
            
        .else
            VMULL   qT2,dXr2,dW2[0]
            VMLSL   qT2,dXi2,dW2[1]                       @// real part
            VMULL   qT3,dXi2,dW2[0]
            VMLAL   qT3,dXr2,dW2[1]                       @// imag part
        
        .endif
        
        VRSHRN  dZr1,qT0,#31
        VRSHRN  dZi1,qT1,#31
        VLD2    {dXr2,dXi2},[pSrc],pointStep              @//  data[2] for next iteration
        
        
        .ifeqs  "\inverse", "TRUE"
            VMULL   qT0,dXr3,dW3[0]
            VMLAL   qT0,dXi3,dW3[1]                       @// real part
            VMULL   qT1,dXi3,dW3[0]
            VMLSL   qT1,dXr3,dW3[1]                       @// imag part
            
        .else
            VMULL   qT0,dXr3,dW3[0]
            VMLSL   qT0,dXi3,dW3[1]                       @// real part
            VMULL   qT1,dXi3,dW3[0]
            VMLAL   qT1,dXr3,dW3[1]                       @// imag part
        
        .endif
        
        VRSHRN  dZr2,qT2,#31
        VRSHRN  dZi2,qT3,#31
        
        
        VRSHRN  dZr3,qT0,#31
        VRSHRN  dZi3,qT1,#31
        VLD2    {dXr3,dXi3},[pSrc],setStep            @//  data[3] & update pSrc to data[0]
        
        .ifeqs "\scaled", "TRUE"
        
            @// finish first stage of 4 point FFT 
            VHADD    qY0,qX0,qZ2
            VHSUB    qY2,qX0,qZ2
                        
            VLD2    {dXr0,dXi0},[pSrc]!          @//  data[0] for next iteration
            VHADD    qY1,qZ1,qZ3
            VHSUB    qY3,qZ1,qZ3
            
            @// finish second stage of 4 point FFT 
            
            VHSUB    qZ0,qY2,qY1
            
            
            .ifeqs  "\inverse", "TRUE"
                
                VHADD    dZr3,dYr0,dYi3
                VST2    {dZr0,dZi0},[pDst :128],outPointStep
                VHSUB    dZi3,dYi0,dYr3
                
                VHADD    qZ2,qY2,qY1
                VST2    {dZr3,dZi3},[pDst :128],outPointStep
            
                VHSUB    dZr1,dYr0,dYi3
                VST2    {dZr2,dZi2},[pDst :128],outPointStep
                VHADD    dZi1,dYi0,dYr3
            
                VST2    {dZr1,dZi1},[pDst :128],dstStep
                
                
            .else
                
                VHSUB    dZr1,dYr0,dYi3
                VST2    {dZr0,dZi0},[pDst :128],outPointStep
                VHADD    dZi1,dYi0,dYr3
            
                VHADD    qZ2,qY2,qY1
                VST2    {dZr1,dZi1},[pDst :128],outPointStep
            
                VHADD    dZr3,dYr0,dYi3
                VST2    {dZr2,dZi2},[pDst :128],outPointStep
                VHSUB    dZi3,dYi0,dYr3
            
                VST2    {dZr3,dZi3},[pDst :128],dstStep

            
            .endif
        
        
        .else
        
            @// finish first stage of 4 point FFT 
            VADD    qY0,qX0,qZ2
            VSUB    qY2,qX0,qZ2
                        
            VLD2    {dXr0,dXi0},[pSrc :128]!          @//  data[0] for next iteration
            VADD    qY1,qZ1,qZ3
            VSUB    qY3,qZ1,qZ3
            
            @// finish second stage of 4 point FFT 
            
            VSUB    qZ0,qY2,qY1
            
            
            .ifeqs  "\inverse", "TRUE"
                
                VADD    dZr3,dYr0,dYi3
                VST2    {dZr0,dZi0},[pDst :128],outPointStep
                VSUB    dZi3,dYi0,dYr3
                
                VADD    qZ2,qY2,qY1
                VST2    {dZr3,dZi3},[pDst :128],outPointStep
            
                VSUB    dZr1,dYr0,dYi3
                VST2    {dZr2,dZi2},[pDst :128],outPointStep
                VADD    dZi1,dYi0,dYr3
            
                VST2    {dZr1,dZi1},[pDst :128],dstStep
                
                
            .else
                
                VSUB    dZr1,dYr0,dYi3
                VST2    {dZr0,dZi0},[pDst :128],outPointStep
                VADD    dZi1,dYi0,dYr3
            
                VADD    qZ2,qY2,qY1
                VST2    {dZr1,dZi1},[pDst :128],outPointStep
            
                VADD    dZr3,dYr0,dYi3
                VST2    {dZr2,dZi2},[pDst :128],outPointStep
                VSUB    dZi3,dYi0,dYr3
            
                VST2    {dZr3,dZi3},[pDst :128],dstStep

            
            .endif
            
        .endif
        
        ADD     pSrc,pSrc,pointStep                         @// increment to data[1] of the next set              
        BGT     setLoop\name
        
        
        VLD1     dW1,[pTwiddle :64],stepTwiddle                  @//[wi | wr]
        SUBS    grpCount,grpCount,#4                    @// subtract 4 since grpCount multiplied by 4               
        VLD1     dW2,[pTwiddle :64],stepTwiddle                  @//[wi | wr]
        ADD     pSrc,pSrc,srcStep                       @// increment pSrc for the next grp
        VLD1     dW3,[pTwiddle :64],twStep                       @//[wi | wr]
        BGT     grpLoop\name

                
        @// Reset and Swap pSrc and pDst for the next stage
        MOV     t1,pDst
        SUB     pDst,pSrc,outPointStep,LSL #2                  @// pDst -= 2*size; pSrc -= 8*size bytes           
        SUB     pSrc,t1,outPointStep    
        
        
        .endm
        
        
        M_START armSP_FFTFwd_CToC_SC32_Radix4_OutOfPlace_unsafe,r4
            FFTSTAGE "FALSE","FALSE",FWD
        M_END

        
        M_START armSP_FFTInv_CToC_SC32_Radix4_OutOfPlace_unsafe,r4
            FFTSTAGE "FALSE","TRUE",INV
        M_END
 
        
        M_START armSP_FFTFwd_CToC_SC32_Sfs_Radix4_OutOfPlace_unsafe,r4
            FFTSTAGE "TRUE","FALSE",FWDSFS
        M_END

        
        M_START armSP_FFTInv_CToC_SC32_Sfs_Radix4_OutOfPlace_unsafe,r4
            FFTSTAGE "TRUE","TRUE",INVSFS
        M_END

        
	.end