You can not select more than 25 topics
Topics must start with a letter or number, can include dashes ('-') and can be up to 35 characters long.
262 lines
15 KiB
262 lines
15 KiB
3 years ago
|
;* ======================================================================== *;
|
||
|
;* TEXAS INSTRUMENTS, INC. *;
|
||
|
;* *;
|
||
|
;* DSPLIB DSP Signal Processing Library *;
|
||
|
;* *;
|
||
|
;* Release: Revision 1.04b *;
|
||
|
;* CVS Revision: 1.8 Sun Sep 29 03:32:17 2002 (UTC) *;
|
||
|
;* Snapshot date: 23-Oct-2003 *;
|
||
|
;* *;
|
||
|
;* This library contains proprietary intellectual property of Texas *;
|
||
|
;* Instruments, Inc. The library and its source code are protected by *;
|
||
|
;* various copyrights, and portions may also be protected by patents or *;
|
||
|
;* other legal protections. *;
|
||
|
;* *;
|
||
|
;* This software is licensed for use with Texas Instruments TMS320 *;
|
||
|
;* family DSPs. This license was provided to you prior to installing *;
|
||
|
;* the software. You may review this license by consulting the file *;
|
||
|
;* TI_license.PDF which accompanies the files in this library. *;
|
||
|
;* ------------------------------------------------------------------------ *;
|
||
|
;* Copyright (C) 2003 Texas Instruments, Incorporated. *;
|
||
|
;* All Rights Reserved. *;
|
||
|
;* ======================================================================== *;
|
||
|
|
||
|
|
||
|
;* ======================================================================== *;
|
||
|
;* Assembler compatibility shim for assembling 4.30 and later code on *;
|
||
|
;* tools prior to 4.30. *;
|
||
|
;* ======================================================================== *;
|
||
|
|
||
|
.if $isdefed(".ASSEMBLER_VERSION")
|
||
|
.asg .ASSEMBLER_VERSION, $asmver
|
||
|
.else
|
||
|
.asg 0, $asmver
|
||
|
.endif
|
||
|
|
||
|
.if ($asmver < 430)
|
||
|
|
||
|
.asg B, CALL ; Function Call
|
||
|
.asg B, RET ; Return from a Function
|
||
|
.asg B, CALLRET ; Function call with Call / Ret chaining.
|
||
|
|
||
|
.if .TMS320C6400
|
||
|
.asg BNOP, CALLNOP ; C64x BNOP as a Fn. Call
|
||
|
.asg BNOP, RETNOP ; C64x BNOP as a Fn. Return
|
||
|
.asg BNOP, CRNOP ; C64x Fn call w/, Call/Ret chaining via BNOP.
|
||
|
.endif
|
||
|
|
||
|
.asg , .asmfunc ; .func equivalent for hand-assembly code
|
||
|
.asg , .endasmfunc ; .endfunc equivalent for hand-assembly code
|
||
|
|
||
|
.endif
|
||
|
|
||
|
;* ======================================================================== *;
|
||
|
;* End of assembler compatibility shim. *;
|
||
|
;* ======================================================================== *;
|
||
|
|
||
|
|
||
|
* ========================================================================= *
|
||
|
* NAME *
|
||
|
* DSP_bexp -- Block exponent of a vector *
|
||
|
* *
|
||
|
* REVISION DATE *
|
||
|
* 12-Jul-2001 *
|
||
|
* *
|
||
|
* USAGE *
|
||
|
* This routine has the following C prototype: *
|
||
|
* *
|
||
|
* int DSP_bexp *
|
||
|
* ( *
|
||
|
* const int *x, /* Input Data */ *
|
||
|
* unsigned nx /* Number of elements */ *
|
||
|
* ); *
|
||
|
* *
|
||
|
* The DSP_bexp routine accepts a list of 'num' input points and *
|
||
|
* performs the determination of the block exponent. It returns *
|
||
|
* the minimum number of "redundant sign bits" observed on the *
|
||
|
* block of data. *
|
||
|
* *
|
||
|
* DESCRIPTION *
|
||
|
* The DSP_bexp function, performs a determination of the block *
|
||
|
* exponent of the vector of elements and returns the maximum *
|
||
|
* exponent. This information can then be used to re-normalize *
|
||
|
* the vector. This operation is useful when auto-scaling is *
|
||
|
* required as in a FFT. The DSP_bexp operation returns the minimum *
|
||
|
* norm of the vector. The minimum norm corresponds to the *
|
||
|
* maximum exponent. *
|
||
|
* *
|
||
|
* C CODE *
|
||
|
* The following is a general C code description of the algorithm *
|
||
|
* without restrictions. This implementation may have restrictions *
|
||
|
* as noted under 'ASSUMPTIONS' below. *
|
||
|
* *
|
||
|
* int DSP_bexp(const int *x, short nx) *
|
||
|
* { *
|
||
|
* int i; *
|
||
|
* int mask, mag; *
|
||
|
* *
|
||
|
* mask = 0; *
|
||
|
* *
|
||
|
* /* ---------------------------------------------------- */ *
|
||
|
* /* Merge the absolute values of all the elements */ *
|
||
|
* /* together. The leftmost bit-change on this mask */ *
|
||
|
* /* gives us the magnitude we desire. */ *
|
||
|
* /* ---------------------------------------------------- */ *
|
||
|
* for (i = 0; i < nx; i++) *
|
||
|
* mask |= x[i] ^ (x[i] >> 31); *
|
||
|
* *
|
||
|
* /* ---------------------------------------------------- */ *
|
||
|
* /* Find the magnitude of the mask. */ *
|
||
|
* /* ---------------------------------------------------- */ *
|
||
|
* for (mag = 0; (1 << mag) < mask; mag++) *
|
||
|
* ; *
|
||
|
* *
|
||
|
* return 31 - mag; *
|
||
|
* } *
|
||
|
* *
|
||
|
* ASSUMPTIONS *
|
||
|
* nx >= 8. *
|
||
|
* nx % 8 == 0. *
|
||
|
* *
|
||
|
* The array 'x' is double-word aligned. *
|
||
|
* *
|
||
|
* NOTES *
|
||
|
* This code is interrupt tolerant, but not interruptible. It masks *
|
||
|
* interrupts for its entire duration with branch delay slots. *
|
||
|
* *
|
||
|
* This code is ENDIAN NEUTRAL. *
|
||
|
* *
|
||
|
* SOURCE *
|
||
|
* DSPLIB. *
|
||
|
* *
|
||
|
* CYCLES *
|
||
|
* cycles = nx/2 + 21, including 6 cycle function call overhead. *
|
||
|
* For nx = 32, cycles = 37. *
|
||
|
* *
|
||
|
* CODESIZE *
|
||
|
* 236 bytes. *
|
||
|
* *
|
||
|
* ------------------------------------------------------------------------- *
|
||
|
* Copyright (c) 2003 Texas Instruments, Incorporated. *
|
||
|
* All Rights Reserved. *
|
||
|
* ========================================================================= *
|
||
|
|
||
|
|
||
|
|
||
|
.sect ".text:_bexp"
|
||
|
.global _DSP_bexp
|
||
|
_DSP_bexp:
|
||
|
* ===================== SYMBOLIC REGISTER ASSIGNMENTS ===================== *
|
||
|
.asg A4, A_x
|
||
|
.asg B4, B_nx
|
||
|
.asg B3, B_ret_addr
|
||
|
.asg A4, A_ret_val
|
||
|
.asg B0, B_i
|
||
|
.asg A0, A_p
|
||
|
.asg B17, B_x
|
||
|
.asg A9, A_x7
|
||
|
.asg A8, A_x6
|
||
|
.asg B9, B_x5
|
||
|
.asg B8, B_x4
|
||
|
.asg A7, A_x3
|
||
|
.asg A6, A_x2
|
||
|
.asg B7, B_x1
|
||
|
.asg B6, B_x0
|
||
|
.asg A16, A_k31
|
||
|
.asg A18, A_a7
|
||
|
.asg A17, A_a6
|
||
|
.asg B16, B_k31
|
||
|
.asg B19, B_a5
|
||
|
.asg B18, B_a4
|
||
|
.asg A18, A_a3
|
||
|
.asg A17, A_a2
|
||
|
.asg B19, B_a1
|
||
|
.asg B18, B_a0
|
||
|
.asg A9, A_b7
|
||
|
.asg A17, A_b6
|
||
|
.asg B9, B_b5
|
||
|
.asg B18, B_b4
|
||
|
.asg A7, A_b3
|
||
|
.asg A17, A_b2
|
||
|
.asg B7, B_b1
|
||
|
.asg B18, B_b0
|
||
|
.asg A5, A_m
|
||
|
.asg B5, B_m
|
||
|
* ========================================================================= *
|
||
|
* =========================== PIPE LOOP PROLOG ============================ *
|
||
|
B .S1 loop
|
||
|
|| MVK .L1 2, A_p
|
||
|
|| ADD .L2X A_x, 8, B_x
|
||
|
|| SHR .S2 B_nx, 3, B_i
|
||
|
|| ZERO .D1 A_m
|
||
|
|| ZERO .D2 B_m
|
||
|
|
||
|
LDDW .D1T2 *A_x[2], B_x5:B_x4 ;[ 1,1]
|
||
|
|| LDDW .D2T1 *B_x[2], A_x7:A_x6 ;[ 1,1]
|
||
|
|| MVK .S1 31, A_k31
|
||
|
|| MVK .S2 31, B_k31
|
||
|
|| SUB .L2 B_i, 1, B_i
|
||
|
; ===== 2 stages of prolog collapsed
|
||
|
* =========================== PIPE LOOP KERNEL ============================ *
|
||
|
loop:
|
||
|
[!A_p]OR .S2 B_b4, B_m, B_m ;[10,1]
|
||
|
||[!A_p]OR .S1 A_b6, A_m, A_m ;[10,1]
|
||
|
|| XOR .D2 B_x1, B_a1, B_b1 ;[10,1]
|
||
|
|| XOR .D1 A_x3, A_a3, A_b3 ;[10,1]
|
||
|
|| SSHVR .M2 B_x5, B_k31, B_a5 ;[ 6,2]
|
||
|
|| SSHVR .M1 A_x7, A_k31, A_a7 ;[ 6,2]
|
||
|
|
||
|
[!A_p]OR .L2 B_b1, B_m, B_m ;[11,1]
|
||
|
||[!A_p]OR .S1 A_b3, A_m, A_m ;[11,1]
|
||
|
|| XOR .S2 B_x0, B_a0, B_b0 ;[11,1]
|
||
|
|| XOR .L1 A_x2, A_a2, A_b2 ;[11,1]
|
||
|
|| SSHVR .M2 B_x4, B_k31, B_a4 ;[ 7,2]
|
||
|
|| SSHVR .M1 A_x6, A_k31, A_a6 ;[ 7,2]
|
||
|
|| LDDW .D1T2 *A_x++[4], B_x1:B_x0 ;[ 3,3]
|
||
|
|| LDDW .D2T1 *B_x++[4], A_x3:A_x2 ;[ 3,3]
|
||
|
|
||
|
[!A_p]OR .L2 B_b0, B_m, B_m ;[12,1]
|
||
|
||[!A_p]OR .D1 A_b2, A_m, A_m ;[12,1]
|
||
|
||[ B_i]BDEC .S2 loop, B_i ;[ 8,2]
|
||
|
||[ A_p]SUB .L1 A_p, 1, A_p ;[ 8,2]
|
||
|
|| XOR .D2 B_x5, B_a5, B_b5 ;[ 8,2]
|
||
|
|| XOR .S1 A_x7, A_a7, A_b7 ;[ 8,2]
|
||
|
|| SSHVR .M2 B_x1, B_k31, B_a1 ;[ 8,2]
|
||
|
|| SSHVR .M1 A_x3, A_k31, A_a3 ;[ 8,2]
|
||
|
|
||
|
[!A_p]OR .S2 B_b5, B_m, B_m ;[ 9,2]
|
||
|
||[!A_p]OR .L1 A_b7, A_m, A_m ;[ 9,2]
|
||
|
|| XOR .L2 B_x4, B_a4, B_b4 ;[ 9,2]
|
||
|
|| XOR .S1 A_x6, A_a6, A_b6 ;[ 9,2]
|
||
|
|| SSHVR .M2 B_x0, B_k31, B_a0 ;[ 9,2]
|
||
|
|| SSHVR .M1 A_x2, A_k31, A_a2 ;[ 9,2]
|
||
|
|| LDDW .D1T2 *A_x[2], B_x5:B_x4 ;[ 1,4]
|
||
|
|| LDDW .D2T1 *B_x[2], A_x7:A_x6 ;[ 1,4]
|
||
|
|
||
|
* =========================== PIPE LOOP EPILOG ============================ *
|
||
|
; ===== 2 stages of epilog collasped
|
||
|
OR .L2 B_b4, B_m, B_m ;[10,4]
|
||
|
|| OR .S1 A_b6, A_m, A_m ;[10,4]
|
||
|
|| XOR .D2 B_x1, B_a1, B_b1 ;[10,4]
|
||
|
|| XOR .D1 A_x3, A_a3, A_b3 ;[10,4]
|
||
|
|| RET .S2 B_ret_addr
|
||
|
|
||
|
OR .L2 B_b1, B_m, B_m ;[11,4]
|
||
|
|| OR .S1 A_b3, A_m, A_m ;[11,4]
|
||
|
|| XOR .S2 B_x0, B_a0, B_b0 ;[11,4]
|
||
|
|| XOR .L1 A_x2, A_a2, A_b2 ;[11,4]
|
||
|
|
||
|
OR .L2 B_b0, B_m, B_m ;[12,4]
|
||
|
|
||
|
OR .D1 A_b2, A_m, A_m ;[12,4]
|
||
|
|
||
|
OR .L1X A_m, B_m, A_m
|
||
|
|
||
|
NORM .L1 A_m, A_ret_val
|
||
|
* ========================================================================= *
|
||
|
* End of file: dsp_bexp.asm *
|
||
|
* ------------------------------------------------------------------------- *
|
||
|
* Copyright (c) 2003 Texas Instruments, Incorporated. *
|
||
|
* All Rights Reserved. *
|
||
|
* ========================================================================= *
|