////////////////////////////////////////////////////////////////////////////////
|
////////////////////////////////////////////////////////////////////////////////
|
//
|
//
|
// Filename: fftgen.cpp
|
// Filename: fftgen.cpp
|
//
|
//
|
// Project: A General Purpose Pipelined FFT Implementation
|
// Project: A General Purpose Pipelined FFT Implementation
|
//
|
//
|
// Purpose: This is the core generator for the project. Every part
|
// Purpose: This is the core generator for the project. Every part
|
// and piece of this project begins and ends in this program.
|
// and piece of this project begins and ends in this program.
|
// Once built, this program will build an FFT (or IFFT) core of arbitrary
|
// Once built, this program will build an FFT (or IFFT) core of arbitrary
|
// width, precision, etc., that will run at two samples per clock.
|
// width, precision, etc., that will run at two samples per clock.
|
// (Incidentally, I didn't pick two samples per clock because it was
|
// (Incidentally, I didn't pick two samples per clock because it was
|
// easier, but rather because there weren't any two-sample per clock
|
// easier, but rather because there weren't any two-sample per clock
|
// FFT's posted on opencores.com. Further, FFT's running at one sample
|
// FFT's posted on opencores.com. Further, FFT's running at one sample
|
// per aren't that hard to find.)
|
// per aren't that hard to find.)
|
//
|
//
|
// You can find the documentation for this program in two places. One is
|
// You can find the documentation for this program in two places. One is
|
// in the usage() function below. The second is in the 'doc'uments
|
// in the usage() function below. The second is in the 'doc'uments
|
// directory that comes with this package, specifically in the spec.pdf
|
// directory that comes with this package, specifically in the spec.pdf
|
// file. If it's not there, type make in the documents directory to
|
// file. If it's not there, type make in the documents directory to
|
// build it.
|
// build it.
|
//
|
//
|
// 20160123 - Thanks to Lesha Birukov, adjusted for MS Visual Studio 2012.
|
// 20160123 - Thanks to Lesha Birukov, adjusted for MS Visual Studio 2012.
|
// (Adjustments are at the top of the file ...)
|
// (Adjustments are at the top of the file ...)
|
//
|
//
|
// Creator: Dan Gisselquist, Ph.D.
|
// Creator: Dan Gisselquist, Ph.D.
|
// Gisselquist Technology, LLC
|
// Gisselquist Technology, LLC
|
//
|
//
|
////////////////////////////////////////////////////////////////////////////////
|
////////////////////////////////////////////////////////////////////////////////
|
//
|
//
|
// Copyright (C) 2015-2018, Gisselquist Technology, LLC
|
// Copyright (C) 2015-2018, Gisselquist Technology, LLC
|
//
|
//
|
// This program is free software (firmware): you can redistribute it and/or
|
// This program is free software (firmware): you can redistribute it and/or
|
// modify it under the terms of the GNU General Public License as published
|
// modify it under the terms of the GNU General Public License as published
|
// by the Free Software Foundation, either version 3 of the License, or (at
|
// by the Free Software Foundation, either version 3 of the License, or (at
|
// your option) any later version.
|
// your option) any later version.
|
//
|
//
|
// This program is distributed in the hope that it will be useful, but WITHOUT
|
// This program is distributed in the hope that it will be useful, but WITHOUT
|
// ANY WARRANTY; without even the implied warranty of MERCHANTIBILITY or
|
// ANY WARRANTY; without even the implied warranty of MERCHANTIBILITY or
|
// FITNESS FOR A PARTICULAR PURPOSE. See the GNU General Public License
|
// FITNESS FOR A PARTICULAR PURPOSE. See the GNU General Public License
|
// for more details.
|
// for more details.
|
//
|
//
|
// You should have received a copy of the GNU General Public License along
|
// You should have received a copy of the GNU General Public License along
|
// with this program. (It's in the $(ROOT)/doc directory, run make with no
|
// with this program. (It's in the $(ROOT)/doc directory. Run make with no
|
// target there if the PDF file isn't present.) If not, see
|
// target there if the PDF file isn't present.) If not, see
|
// <http://www.gnu.org/licenses/> for a copy.
|
// <http://www.gnu.org/licenses/> for a copy.
|
//
|
//
|
// License: GPL, v3, as defined and found on www.gnu.org,
|
// License: GPL, v3, as defined and found on www.gnu.org,
|
// http://www.gnu.org/licenses/gpl.html
|
// http://www.gnu.org/licenses/gpl.html
|
//
|
//
|
//
|
//
|
////////////////////////////////////////////////////////////////////////////////
|
////////////////////////////////////////////////////////////////////////////////
|
//
|
//
|
//
|
//
|
#define _CRT_SECURE_NO_WARNINGS // ms vs 2012 doesn't like fopen
|
#define _CRT_SECURE_NO_WARNINGS // ms vs 2012 doesn't like fopen
|
#include <stdio.h>
|
#include <stdio.h>
|
#include <stdlib.h>
|
#include <stdlib.h>
|
|
|
#ifdef _MSC_VER // added for ms vs compatibility
|
#ifdef _MSC_VER // added for ms vs compatibility
|
|
|
#include <io.h>
|
#include <io.h>
|
#include <direct.h>
|
#include <direct.h>
|
#define _USE_MATH_DEFINES
|
#define _USE_MATH_DEFINES
|
#define R_OK 4 /* Test for read permission. */
|
#define R_OK 4 /* Test for read permission. */
|
#define W_OK 2 /* Test for write permission. */
|
#define W_OK 2 /* Test for write permission. */
|
#define X_OK 0 /* !!!!!! execute permission - unsupported in windows*/
|
#define X_OK 0 /* !!!!!! execute permission - unsupported in windows*/
|
#define F_OK 0 /* Test for existence. */
|
#define F_OK 0 /* Test for existence. */
|
|
|
#if _MSC_VER <= 1700
|
#if _MSC_VER <= 1700
|
|
|
int lstat(const char *filename, struct stat *buf) { return 1; };
|
int lstat(const char *filename, struct stat *buf) { return 1; };
|
#define S_ISDIR(A) 0
|
#define S_ISDIR(A) 0
|
|
|
#else
|
#else
|
|
|
#define lstat _stat
|
#define lstat _stat
|
#define S_ISDIR _S_IFDIR
|
#define S_ISDIR _S_IFDIR
|
|
|
#endif
|
#endif
|
|
|
#define mkdir(A,B) _mkdir(A)
|
#define mkdir(A,B) _mkdir(A)
|
|
|
#define access _access
|
#define access _access
|
|
|
#else
|
#else
|
// And for G++/Linux environment
|
// And for G++/Linux environment
|
|
|
#include <unistd.h> // Defines the R_OK/W_OK/etc. macros
|
#include <unistd.h> // Defines the R_OK/W_OK/etc. macros
|
#include <sys/stat.h>
|
#include <sys/stat.h>
|
#endif
|
#endif
|
|
|
#include <string.h>
|
#include <string.h>
|
#include <string>
|
#include <string>
|
#include <math.h>
|
#include <math.h>
|
#include <ctype.h>
|
#include <ctype.h>
|
#include <assert.h>
|
#include <assert.h>
|
|
|
#include "defaults.h"
|
#include "defaults.h"
|
#include "legal.h"
|
#include "legal.h"
|
#include "rounding.h"
|
#include "rounding.h"
|
#include "fftlib.h"
|
#include "fftlib.h"
|
#include "bldstage.h"
|
#include "bldstage.h"
|
#include "bitreverse.h"
|
#include "bitreverse.h"
|
#include "softmpy.h"
|
#include "softmpy.h"
|
#include "butterfly.h"
|
#include "butterfly.h"
|
|
|
void build_dblquarters(const char *fname, ROUND_T rounding, const bool async_reset=false, const bool dbg=false) {
|
void build_dblquarters(const char *fname, ROUND_T rounding, const bool async_reset=false, const bool dbg=false) {
|
FILE *fp = fopen(fname, "w");
|
FILE *fp = fopen(fname, "w");
|
if (NULL == fp) {
|
if (NULL == fp) {
|
fprintf(stderr, "Could not open \'%s\' for writing\n", fname);
|
fprintf(stderr, "Could not open \'%s\' for writing\n", fname);
|
perror("O/S Err was:");
|
perror("O/S Err was:");
|
return;
|
return;
|
}
|
}
|
const char *rnd_string;
|
const char *rnd_string;
|
if (rounding == RND_TRUNCATE)
|
if (rounding == RND_TRUNCATE)
|
rnd_string = "truncate";
|
rnd_string = "truncate";
|
else if (rounding == RND_FROMZERO)
|
else if (rounding == RND_FROMZERO)
|
rnd_string = "roundfromzero";
|
rnd_string = "roundfromzero";
|
else if (rounding == RND_HALFUP)
|
else if (rounding == RND_HALFUP)
|
rnd_string = "roundhalfup";
|
rnd_string = "roundhalfup";
|
else
|
else
|
rnd_string = "convround";
|
rnd_string = "convround";
|
|
|
|
|
fprintf(fp,
|
fprintf(fp,
|
SLASHLINE
|
SLASHLINE
|
"//\n"
|
"//\n"
|
"// Filename:\tqtrstage%s.v\n"
|
"// Filename:\tqtrstage%s.v\n"
|
"//\n"
|
"//\n"
|
"// Project:\t%s\n"
|
"// Project:\t%s\n"
|
"//\n"
|
"//\n"
|
"// Purpose: This file encapsulates the 4 point stage of a decimation in\n"
|
"// Purpose: This file encapsulates the 4 point stage of a decimation in\n"
|
"// frequency FFT. This particular implementation is optimized\n"
|
"// frequency FFT. This particular implementation is optimized\n"
|
"// so that all of the multiplies are accomplished by additions and\n"
|
"// so that all of the multiplies are accomplished by additions and\n"
|
"// multiplexers only.\n"
|
"// multiplexers only.\n"
|
"//\n"
|
"//\n"
|
"//\n%s"
|
"//\n%s"
|
"//\n",
|
"//\n",
|
(dbg)?"_dbg":"", prjname, creator);
|
(dbg)?"_dbg":"", prjname, creator);
|
fprintf(fp, "%s", cpyleft);
|
fprintf(fp, "%s", cpyleft);
|
fprintf(fp, "//\n//\n`default_nettype\tnone\n//\n");
|
fprintf(fp, "//\n//\n`default_nettype\tnone\n//\n");
|
|
|
std::string resetw("i_reset");
|
std::string resetw("i_reset");
|
if (async_reset)
|
if (async_reset)
|
resetw = std::string("i_areset_n");
|
resetw = std::string("i_areset_n");
|
|
|
fprintf(fp,
|
fprintf(fp,
|
"module\tqtrstage%s(i_clk, %s, i_ce, i_sync, i_data, o_data, o_sync%s);\n"
|
"module\tqtrstage%s(i_clk, %s, i_ce, i_sync, i_data, o_data, o_sync%s);\n"
|
"\tparameter IWIDTH=%d, OWIDTH=IWIDTH+1;\n"
|
"\tparameter IWIDTH=%d, OWIDTH=IWIDTH+1;\n"
|
"\t// Parameters specific to the core that should be changed when this\n"
|
"\t// Parameters specific to the core that should be changed when this\n"
|
"\t// core is built ... Note that the minimum LGSPAN is 2. Smaller\n"
|
"\t// core is built ... Note that the minimum LGSPAN is 2. Smaller\n"
|
"\t// spans must use the fftdoubles stage.\n"
|
"\t// spans must use the fftdoubles stage.\n"
|
"\tparameter\tLGWIDTH=%d, ODD=0, INVERSE=0,SHIFT=0;\n"
|
"\tparameter\tLGWIDTH=%d, ODD=0, INVERSE=0,SHIFT=0;\n"
|
"\tinput\t i_clk, %s, i_ce, i_sync;\n"
|
"\tinput\twire i_clk, %s, i_ce, i_sync;\n"
|
"\tinput\t [(2*IWIDTH-1):0] i_data;\n"
|
"\tinput\twire [(2*IWIDTH-1):0] i_data;\n"
|
"\toutput\treg [(2*OWIDTH-1):0] o_data;\n"
|
"\toutput\treg [(2*OWIDTH-1):0] o_data;\n"
|
"\toutput\treg o_sync;\n"
|
"\toutput\treg o_sync;\n"
|
"\t\n", (dbg)?"_dbg":"",
|
"\t\n", (dbg)?"_dbg":"",
|
resetw.c_str(),
|
resetw.c_str(),
|
(dbg)?", o_dbg":"", TST_QTRSTAGE_IWIDTH,
|
(dbg)?", o_dbg":"", TST_QTRSTAGE_IWIDTH,
|
TST_QTRSTAGE_LGWIDTH, resetw.c_str());
|
TST_QTRSTAGE_LGWIDTH, resetw.c_str());
|
if (dbg) { fprintf(fp, "\toutput\twire\t[33:0]\t\t\to_dbg;\n"
|
if (dbg) { fprintf(fp, "\toutput\twire\t[33:0]\t\t\to_dbg;\n"
|
"\tassign\to_dbg = { ((o_sync)&&(i_ce)), i_ce, o_data[(2*OWIDTH-1):(2*OWIDTH-16)],\n"
|
"\tassign\to_dbg = { ((o_sync)&&(i_ce)), i_ce, o_data[(2*OWIDTH-1):(2*OWIDTH-16)],\n"
|
"\t\t\t\t\to_data[(OWIDTH-1):(OWIDTH-16)] };\n"
|
"\t\t\t\t\to_data[(OWIDTH-1):(OWIDTH-16)] };\n"
|
"\n");
|
"\n");
|
}
|
}
|
fprintf(fp,
|
fprintf(fp,
|
"\treg\t wait_for_sync;\n"
|
"\treg\t wait_for_sync;\n"
|
"\treg\t[3:0] pipeline;\n"
|
"\treg\t[3:0] pipeline;\n"
|
"\n"
|
"\n"
|
"\treg\t[(IWIDTH):0] sum_r, sum_i, diff_r, diff_i;\n"
|
"\treg\t[(IWIDTH):0] sum_r, sum_i, diff_r, diff_i;\n"
|
"\n"
|
"\n"
|
"\treg\t[(2*OWIDTH-1):0]\tob_a;\n"
|
"\treg\t[(2*OWIDTH-1):0]\tob_a;\n"
|
"\twire\t[(2*OWIDTH-1):0]\tob_b;\n"
|
"\twire\t[(2*OWIDTH-1):0]\tob_b;\n"
|
"\treg\t[(OWIDTH-1):0]\t\tob_b_r, ob_b_i;\n"
|
"\treg\t[(OWIDTH-1):0]\t\tob_b_r, ob_b_i;\n"
|
"\tassign\tob_b = { ob_b_r, ob_b_i };\n"
|
"\tassign\tob_b = { ob_b_r, ob_b_i };\n"
|
"\n"
|
"\n"
|
"\treg\t[(LGWIDTH-1):0]\t\tiaddr;\n"
|
"\treg\t[(LGWIDTH-1):0]\t\tiaddr;\n"
|
"\treg\t[(2*IWIDTH-1):0]\timem;\n"
|
"\treg\t[(2*IWIDTH-1):0]\timem;\n"
|
"\n"
|
"\n"
|
"\twire\tsigned\t[(IWIDTH-1):0]\timem_r, imem_i;\n"
|
"\twire\tsigned\t[(IWIDTH-1):0]\timem_r, imem_i;\n"
|
"\tassign\timem_r = imem[(2*IWIDTH-1):(IWIDTH)];\n"
|
"\tassign\timem_r = imem[(2*IWIDTH-1):(IWIDTH)];\n"
|
"\tassign\timem_i = imem[(IWIDTH-1):0];\n"
|
"\tassign\timem_i = imem[(IWIDTH-1):0];\n"
|
"\n"
|
"\n"
|
"\twire\tsigned\t[(IWIDTH-1):0]\ti_data_r, i_data_i;\n"
|
"\twire\tsigned\t[(IWIDTH-1):0]\ti_data_r, i_data_i;\n"
|
"\tassign\ti_data_r = i_data[(2*IWIDTH-1):(IWIDTH)];\n"
|
"\tassign\ti_data_r = i_data[(2*IWIDTH-1):(IWIDTH)];\n"
|
"\tassign\ti_data_i = i_data[(IWIDTH-1):0];\n"
|
"\tassign\ti_data_i = i_data[(IWIDTH-1):0];\n"
|
"\n"
|
"\n"
|
"\treg [(2*OWIDTH-1):0] omem;\n"
|
"\treg [(2*OWIDTH-1):0] omem;\n"
|
"\n");
|
"\n");
|
fprintf(fp,
|
fprintf(fp,
|
"\twire\tsigned\t[(OWIDTH-1):0]\trnd_sum_r, rnd_sum_i, rnd_diff_r, rnd_diff_i,\n");
|
"\twire\tsigned\t[(OWIDTH-1):0]\trnd_sum_r, rnd_sum_i, rnd_diff_r, rnd_diff_i,\n");
|
fprintf(fp,
|
fprintf(fp,
|
"\t\t\t\t\tn_rnd_diff_r, n_rnd_diff_i;\n");
|
"\t\t\t\t\tn_rnd_diff_r, n_rnd_diff_i;\n");
|
fprintf(fp,
|
fprintf(fp,
|
"\t%s #(IWIDTH+1,OWIDTH,SHIFT)\tdo_rnd_sum_r(i_clk, i_ce,\n"
|
"\t%s #(IWIDTH+1,OWIDTH,SHIFT)\tdo_rnd_sum_r(i_clk, i_ce,\n"
|
"\t\t\t\tsum_r, rnd_sum_r);\n\n", rnd_string);
|
"\t\t\t\tsum_r, rnd_sum_r);\n\n", rnd_string);
|
fprintf(fp,
|
fprintf(fp,
|
"\t%s #(IWIDTH+1,OWIDTH,SHIFT)\tdo_rnd_sum_i(i_clk, i_ce,\n"
|
"\t%s #(IWIDTH+1,OWIDTH,SHIFT)\tdo_rnd_sum_i(i_clk, i_ce,\n"
|
"\t\t\t\tsum_i, rnd_sum_i);\n\n", rnd_string);
|
"\t\t\t\tsum_i, rnd_sum_i);\n\n", rnd_string);
|
fprintf(fp,
|
fprintf(fp,
|
"\t%s #(IWIDTH+1,OWIDTH,SHIFT)\tdo_rnd_diff_r(i_clk, i_ce,\n"
|
"\t%s #(IWIDTH+1,OWIDTH,SHIFT)\tdo_rnd_diff_r(i_clk, i_ce,\n"
|
"\t\t\t\tdiff_r, rnd_diff_r);\n\n", rnd_string);
|
"\t\t\t\tdiff_r, rnd_diff_r);\n\n", rnd_string);
|
fprintf(fp,
|
fprintf(fp,
|
"\t%s #(IWIDTH+1,OWIDTH,SHIFT)\tdo_rnd_diff_i(i_clk, i_ce,\n"
|
"\t%s #(IWIDTH+1,OWIDTH,SHIFT)\tdo_rnd_diff_i(i_clk, i_ce,\n"
|
"\t\t\t\tdiff_i, rnd_diff_i);\n\n", rnd_string);
|
"\t\t\t\tdiff_i, rnd_diff_i);\n\n", rnd_string);
|
fprintf(fp, "\tassign n_rnd_diff_r = - rnd_diff_r;\n"
|
fprintf(fp, "\tassign n_rnd_diff_r = - rnd_diff_r;\n"
|
"\tassign n_rnd_diff_i = - rnd_diff_i;\n");
|
"\tassign n_rnd_diff_i = - rnd_diff_i;\n");
|
/*
|
/*
|
fprintf(fp,
|
fprintf(fp,
|
"\twire [(IWIDTH-1):0] rnd;\n"
|
"\twire [(IWIDTH-1):0] rnd;\n"
|
"\tgenerate\n"
|
"\tgenerate\n"
|
"\tif ((ROUND)&&((IWIDTH+1-OWIDTH-SHIFT)>0))\n"
|
"\tif ((ROUND)&&((IWIDTH+1-OWIDTH-SHIFT)>0))\n"
|
"\t\tassign rnd = { {(IWIDTH-1){1\'b0}}, 1\'b1 };\n"
|
"\t\tassign rnd = { {(IWIDTH-1){1\'b0}}, 1\'b1 };\n"
|
"\telse\n"
|
"\telse\n"
|
"\t\tassign rnd = { {(IWIDTH){1\'b0}}};\n"
|
"\t\tassign rnd = { {(IWIDTH){1\'b0}}};\n"
|
"\tendgenerate\n"
|
"\tendgenerate\n"
|
"\n"
|
"\n"
|
*/
|
*/
|
fprintf(fp,
|
fprintf(fp,
|
"\tinitial wait_for_sync = 1\'b1;\n"
|
"\tinitial wait_for_sync = 1\'b1;\n"
|
"\tinitial iaddr = 0;\n");
|
"\tinitial iaddr = 0;\n");
|
if (async_reset)
|
if (async_reset)
|
fprintf(fp,
|
fprintf(fp,
|
"\talways @(posedge i_clk, negedge i_areset_n)\n"
|
"\talways @(posedge i_clk, negedge i_areset_n)\n"
|
"\t\tif (!i_reset)\n");
|
"\t\tif (!i_reset)\n");
|
else
|
else
|
fprintf(fp,
|
fprintf(fp,
|
"\talways @(posedge i_clk)\n"
|
"\talways @(posedge i_clk)\n"
|
"\t\tif (i_reset)\n");
|
"\t\tif (i_reset)\n");
|
fprintf(fp,
|
fprintf(fp,
|
"\t\tbegin\n"
|
"\t\tbegin\n"
|
"\t\t\twait_for_sync <= 1\'b1;\n"
|
"\t\t\twait_for_sync <= 1\'b1;\n"
|
"\t\t\tiaddr <= 0;\n"
|
"\t\t\tiaddr <= 0;\n"
|
"\t\tend else if ((i_ce)&&((!wait_for_sync)||(i_sync)))\n"
|
"\t\tend else if ((i_ce)&&((!wait_for_sync)||(i_sync)))\n"
|
"\t\tbegin\n"
|
"\t\tbegin\n"
|
"\t\t\tiaddr <= iaddr + { {(LGWIDTH-1){1\'b0}}, 1\'b1 };\n"
|
"\t\t\tiaddr <= iaddr + { {(LGWIDTH-1){1\'b0}}, 1\'b1 };\n"
|
"\t\t\twait_for_sync <= 1\'b0;\n"
|
"\t\t\twait_for_sync <= 1\'b0;\n"
|
"\t\tend\n\n"
|
"\t\tend\n\n"
|
"\talways @(posedge i_clk)\n"
|
"\talways @(posedge i_clk)\n"
|
"\t\tif (i_ce)\n"
|
"\t\tif (i_ce)\n"
|
"\t\t\timem <= i_data;\n"
|
"\t\t\timem <= i_data;\n"
|
"\n\n");
|
"\n\n");
|
fprintf(fp,
|
fprintf(fp,
|
"\t// Note that we don\'t check on wait_for_sync or i_sync here.\n"
|
"\t// Note that we don\'t check on wait_for_sync or i_sync here.\n"
|
"\t// Why not? Because iaddr will always be zero until after the\n"
|
"\t// Why not? Because iaddr will always be zero until after the\n"
|
"\t// first i_ce, so we are safe.\n"
|
"\t// first i_ce, so we are safe.\n"
|
"\tinitial pipeline = 4\'h0;\n");
|
"\tinitial pipeline = 4\'h0;\n");
|
if (async_reset)
|
if (async_reset)
|
fprintf(fp,
|
fprintf(fp,
|
"\talways\t@(posedge i_clk, negedge i_areset_n)\n"
|
"\talways\t@(posedge i_clk, negedge i_areset_n)\n"
|
"\t\tif (!i_reset)\n");
|
"\t\tif (!i_reset)\n");
|
else
|
else
|
fprintf(fp,
|
fprintf(fp,
|
"\talways\t@(posedge i_clk)\n"
|
"\talways\t@(posedge i_clk)\n"
|
"\t\tif (i_reset)\n");
|
"\t\tif (i_reset)\n");
|
|
|
fprintf(fp,
|
fprintf(fp,
|
"\t\t\tpipeline <= 4\'h0;\n"
|
"\t\t\tpipeline <= 4\'h0;\n"
|
"\t\telse if (i_ce) // is our pipeline process full? Which stages?\n"
|
"\t\telse if (i_ce) // is our pipeline process full? Which stages?\n"
|
"\t\t\tpipeline <= { pipeline[2:0], iaddr[0] };\n\n");
|
"\t\t\tpipeline <= { pipeline[2:0], iaddr[0] };\n\n");
|
fprintf(fp,
|
fprintf(fp,
|
"\t// This is the pipeline[-1] stage, pipeline[0] will be set next.\n"
|
"\t// This is the pipeline[-1] stage, pipeline[0] will be set next.\n"
|
"\talways\t@(posedge i_clk)\n"
|
"\talways\t@(posedge i_clk)\n"
|
"\t\tif ((i_ce)&&(iaddr[0]))\n"
|
"\t\tif ((i_ce)&&(iaddr[0]))\n"
|
"\t\tbegin\n"
|
"\t\tbegin\n"
|
"\t\t\tsum_r <= imem_r + i_data_r;\n"
|
"\t\t\tsum_r <= imem_r + i_data_r;\n"
|
"\t\t\tsum_i <= imem_i + i_data_i;\n"
|
"\t\t\tsum_i <= imem_i + i_data_i;\n"
|
"\t\t\tdiff_r <= imem_r - i_data_r;\n"
|
"\t\t\tdiff_r <= imem_r - i_data_r;\n"
|
"\t\t\tdiff_i <= imem_i - i_data_i;\n"
|
"\t\t\tdiff_i <= imem_i - i_data_i;\n"
|
"\t\tend\n\n");
|
"\t\tend\n\n");
|
fprintf(fp,
|
fprintf(fp,
|
"\t// pipeline[1] takes sum_x and diff_x and produces rnd_x\n\n");
|
"\t// pipeline[1] takes sum_x and diff_x and produces rnd_x\n\n");
|
fprintf(fp,
|
fprintf(fp,
|
"\t// Now for pipeline[2]. We can actually do this at all i_ce\n"
|
"\t// Now for pipeline[2]. We can actually do this at all i_ce\n"
|
"\t// clock times, since nothing will listen unless pipeline[3]\n"
|
"\t// clock times, since nothing will listen unless pipeline[3]\n"
|
"\t// on the next clock. Thus, we simplify this logic and do\n"
|
"\t// on the next clock. Thus, we simplify this logic and do\n"
|
"\t// it independent of pipeline[2].\n"
|
"\t// it independent of pipeline[2].\n"
|
"\talways\t@(posedge i_clk)\n"
|
"\talways\t@(posedge i_clk)\n"
|
"\t\tif (i_ce)\n"
|
"\t\tif (i_ce)\n"
|
"\t\tbegin\n"
|
"\t\tbegin\n"
|
"\t\t\tob_a <= { rnd_sum_r, rnd_sum_i };\n"
|
"\t\t\tob_a <= { rnd_sum_r, rnd_sum_i };\n"
|
"\t\t\t// on Even, W = e^{-j2pi 1/4 0} = 1\n"
|
"\t\t\t// on Even, W = e^{-j2pi 1/4 0} = 1\n"
|
"\t\t\tif (ODD == 0)\n"
|
"\t\t\tif (ODD == 0)\n"
|
"\t\t\tbegin\n"
|
"\t\t\tbegin\n"
|
"\t\t\t\tob_b_r <= rnd_diff_r;\n"
|
"\t\t\t\tob_b_r <= rnd_diff_r;\n"
|
"\t\t\t\tob_b_i <= rnd_diff_i;\n"
|
"\t\t\t\tob_b_i <= rnd_diff_i;\n"
|
"\t\t\tend else if (INVERSE==0) begin\n"
|
"\t\t\tend else if (INVERSE==0) begin\n"
|
"\t\t\t\t// on Odd, W = e^{-j2pi 1/4} = -j\n"
|
"\t\t\t\t// on Odd, W = e^{-j2pi 1/4} = -j\n"
|
"\t\t\t\tob_b_r <= rnd_diff_i;\n"
|
"\t\t\t\tob_b_r <= rnd_diff_i;\n"
|
"\t\t\t\tob_b_i <= n_rnd_diff_r;\n"
|
"\t\t\t\tob_b_i <= n_rnd_diff_r;\n"
|
"\t\t\tend else begin\n"
|
"\t\t\tend else begin\n"
|
"\t\t\t\t// on Odd, W = e^{j2pi 1/4} = j\n"
|
"\t\t\t\t// on Odd, W = e^{j2pi 1/4} = j\n"
|
"\t\t\t\tob_b_r <= n_rnd_diff_i;\n"
|
"\t\t\t\tob_b_r <= n_rnd_diff_i;\n"
|
"\t\t\t\tob_b_i <= rnd_diff_r;\n"
|
"\t\t\t\tob_b_i <= rnd_diff_r;\n"
|
"\t\t\tend\n"
|
"\t\t\tend\n"
|
"\t\tend\n\n");
|
"\t\tend\n\n");
|
fprintf(fp,
|
fprintf(fp,
|
"\talways\t@(posedge i_clk)\n"
|
"\talways\t@(posedge i_clk)\n"
|
"\t\tif (i_ce)\n"
|
"\t\tif (i_ce)\n"
|
"\t\tbegin // In sequence, clock = 3\n"
|
"\t\tbegin // In sequence, clock = 3\n"
|
"\t\t\tif (pipeline[3])\n"
|
"\t\t\tif (pipeline[3])\n"
|
"\t\t\tbegin\n"
|
"\t\t\tbegin\n"
|
"\t\t\t\tomem <= ob_b;\n"
|
"\t\t\t\tomem <= ob_b;\n"
|
"\t\t\t\to_data <= ob_a;\n"
|
"\t\t\t\to_data <= ob_a;\n"
|
"\t\t\tend else\n"
|
"\t\t\tend else\n"
|
"\t\t\t\to_data <= omem;\n"
|
"\t\t\t\to_data <= omem;\n"
|
"\t\tend\n\n");
|
"\t\tend\n\n");
|
|
|
fprintf(fp,
|
fprintf(fp,
|
"\t// Don\'t forget in the sync check that we are running\n"
|
"\t// Don\'t forget in the sync check that we are running\n"
|
"\t// at two clocks per sample. Thus we need to\n"
|
"\t// at two clocks per sample. Thus we need to\n"
|
"\t// produce a sync every 2^(LGWIDTH-1) clocks.\n"
|
"\t// produce a sync every 2^(LGWIDTH-1) clocks.\n"
|
"\tinitial\to_sync = 1\'b0;\n");
|
"\tinitial\to_sync = 1\'b0;\n");
|
|
|
if (async_reset)
|
if (async_reset)
|
fprintf(fp,
|
fprintf(fp,
|
"\talways\t@(posedge i_clk, negedge i_areset_n)\n"
|
"\talways\t@(posedge i_clk, negedge i_areset_n)\n"
|
"\t\tif (!i_areset_n)\n");
|
"\t\tif (!i_areset_n)\n");
|
else
|
else
|
fprintf(fp,
|
fprintf(fp,
|
"\talways\t@(posedge i_clk)\n"
|
"\talways\t@(posedge i_clk)\n"
|
"\t\tif (i_reset)\n");
|
"\t\tif (i_reset)\n");
|
fprintf(fp,
|
fprintf(fp,
|
"\t\t\to_sync <= 1\'b0;\n"
|
"\t\t\to_sync <= 1\'b0;\n"
|
"\t\telse if (i_ce)\n"
|
"\t\telse if (i_ce)\n"
|
"\t\t\to_sync <= &(~iaddr[(LGWIDTH-2):3]) && (iaddr[2:0] == 3'b101);\n");
|
"\t\t\to_sync <= &(~iaddr[(LGWIDTH-2):3]) && (iaddr[2:0] == 3'b101);\n");
|
fprintf(fp, "endmodule\n");
|
fprintf(fp, "endmodule\n");
|
}
|
}
|
|
|
void build_snglquarters(const char *fname, ROUND_T rounding, const bool async_reset=false, const bool dbg=false) {
|
void build_snglquarters(const char *fname, ROUND_T rounding, const bool async_reset=false, const bool dbg=false) {
|
FILE *fp = fopen(fname, "w");
|
FILE *fp = fopen(fname, "w");
|
if (NULL == fp) {
|
if (NULL == fp) {
|
fprintf(stderr, "Could not open \'%s\' for writing\n", fname);
|
fprintf(stderr, "Could not open \'%s\' for writing\n", fname);
|
perror("O/S Err was:");
|
perror("O/S Err was:");
|
return;
|
return;
|
}
|
}
|
const char *rnd_string;
|
const char *rnd_string;
|
if (rounding == RND_TRUNCATE)
|
if (rounding == RND_TRUNCATE)
|
rnd_string = "truncate";
|
rnd_string = "truncate";
|
else if (rounding == RND_FROMZERO)
|
else if (rounding == RND_FROMZERO)
|
rnd_string = "roundfromzero";
|
rnd_string = "roundfromzero";
|
else if (rounding == RND_HALFUP)
|
else if (rounding == RND_HALFUP)
|
rnd_string = "roundhalfup";
|
rnd_string = "roundhalfup";
|
else
|
else
|
rnd_string = "convround";
|
rnd_string = "convround";
|
|
|
|
|
fprintf(fp,
|
fprintf(fp,
|
SLASHLINE
|
SLASHLINE
|
"//\n"
|
"//\n"
|
"// Filename:\tqtrstage%s.v\n"
|
"// Filename:\tqtrstage%s.v\n"
|
"//\n"
|
"//\n"
|
"// Project:\t%s\n"
|
"// Project:\t%s\n"
|
"//\n"
|
"//\n"
|
"// Purpose: This file encapsulates the 4 point stage of a decimation in\n"
|
"// Purpose: This file encapsulates the 4 point stage of a decimation in\n"
|
"// frequency FFT. This particular implementation is optimized\n"
|
"// frequency FFT. This particular implementation is optimized\n"
|
"// so that all of the multiplies are accomplished by additions and\n"
|
"// so that all of the multiplies are accomplished by additions and\n"
|
"// multiplexers only.\n"
|
"// multiplexers only.\n"
|
"//\n"
|
"//\n"
|
"// Operation:\n"
|
"// Operation:\n"
|
"// The operation of this stage is identical to the regular stages of\n"
|
"// The operation of this stage is identical to the regular stages of\n"
|
"// the FFT (see them for details), with one additional and critical\n"
|
"// the FFT (see them for details), with one additional and critical\n"
|
"// difference: this stage doesn't require any hardware multiplication.\n"
|
"// difference: this stage doesn't require any hardware multiplication.\n"
|
"// The multiplies within it may all be accomplished using additions and\n"
|
"// The multiplies within it may all be accomplished using additions and\n"
|
"// subtractions.\n"
|
"// subtractions.\n"
|
"//\n"
|
"//\n"
|
"// Let's see how this is done. Given x[n] and x[n+2], cause thats the\n"
|
"// Let's see how this is done. Given x[n] and x[n+2], cause thats the\n"
|
"// stage we are working on, with i_sync true for x[0] being input,\n"
|
"// stage we are working on, with i_sync true for x[0] being input,\n"
|
"// produce the output:\n"
|
"// produce the output:\n"
|
"//\n"
|
"//\n"
|
"// y[n ] = x[n] + x[n+2]\n"
|
"// y[n ] = x[n] + x[n+2]\n"
|
"// y[n+2] = (x[n] - x[n+2]) * e^{-j2pi n/2} (forward transform)\n"
|
"// y[n+2] = (x[n] - x[n+2]) * e^{-j2pi n/2} (forward transform)\n"
|
"// = (x[n] - x[n+2]) * -j^n\n"
|
"// = (x[n] - x[n+2]) * -j^n\n"
|
"//\n"
|
"//\n"
|
"// y[n].r = x[n].r + x[n+2].r (This is the easy part)\n"
|
"// y[n].r = x[n].r + x[n+2].r (This is the easy part)\n"
|
"// y[n].i = x[n].i + x[n+2].i\n"
|
"// y[n].i = x[n].i + x[n+2].i\n"
|
"//\n"
|
"//\n"
|
"// y[2].r = x[0].r - x[2].r\n"
|
"// y[2].r = x[0].r - x[2].r\n"
|
"// y[2].i = x[0].i - x[2].i\n"
|
"// y[2].i = x[0].i - x[2].i\n"
|
"//\n"
|
"//\n"
|
"// y[3].r = (x[1].i - x[3].i) (forward transform)\n"
|
"// y[3].r = (x[1].i - x[3].i) (forward transform)\n"
|
"// y[3].i = - (x[1].r - x[3].r)\n"
|
"// y[3].i = - (x[1].r - x[3].r)\n"
|
"//\n"
|
"//\n"
|
"// y[3].r = - (x[1].i - x[3].i) (inverse transform)\n"
|
"// y[3].r = - (x[1].i - x[3].i) (inverse transform)\n"
|
"// y[3].i = (x[1].r - x[3].r) (INVERSE = 1)\n"
|
"// y[3].i = (x[1].r - x[3].r) (INVERSE = 1)\n"
|
// "//\n"
|
// "//\n"
|
// "// When the FFT is run in the two samples per clock mode, this quarter\n"
|
// "// When the FFT is run in the two samples per clock mode, this quarter\n"
|
// "// stage will operate on either x[0] and x[2] (ODD = 0), or x[1] and\n"
|
// "// stage will operate on either x[0] and x[2] (ODD = 0), or x[1] and\n"
|
// "// x[3] (ODD = 1). In all other cases, it will operate on all four\n"
|
// "// x[3] (ODD = 1). In all other cases, it will operate on all four\n"
|
// "// values.\n"
|
// "// values.\n"
|
"//\n%s"
|
"//\n%s"
|
"//\n",
|
"//\n",
|
(dbg)?"_dbg":"", prjname, creator);
|
(dbg)?"_dbg":"", prjname, creator);
|
fprintf(fp, "%s", cpyleft);
|
fprintf(fp, "%s", cpyleft);
|
fprintf(fp, "//\n//\n`default_nettype\tnone\n//\n");
|
fprintf(fp, "//\n//\n`default_nettype\tnone\n//\n");
|
|
|
std::string resetw("i_reset");
|
std::string resetw("i_reset");
|
if (async_reset)
|
if (async_reset)
|
resetw = std::string("i_areset_n");
|
resetw = std::string("i_areset_n");
|
|
|
fprintf(fp,
|
fprintf(fp,
|
"module\tqtrstage%s(i_clk, %s, i_ce, i_sync, i_data, o_data, o_sync%s);\n"
|
"module\tqtrstage%s(i_clk, %s, i_ce, i_sync, i_data, o_data, o_sync%s);\n"
|
"\tparameter IWIDTH=%d, OWIDTH=IWIDTH+1;\n"
|
"\tparameter IWIDTH=%d, OWIDTH=IWIDTH+1;\n"
|
"\tparameter\tLGWIDTH=%d, INVERSE=0,SHIFT=0;\n"
|
"\tparameter\tLGWIDTH=%d, INVERSE=0,SHIFT=0;\n"
|
"\tinput\t i_clk, %s, i_ce, i_sync;\n"
|
"\tinput\twire i_clk, %s, i_ce, i_sync;\n"
|
"\tinput\t [(2*IWIDTH-1):0] i_data;\n"
|
"\tinput\twire [(2*IWIDTH-1):0] i_data;\n"
|
"\toutput\treg [(2*OWIDTH-1):0] o_data;\n"
|
"\toutput\treg [(2*OWIDTH-1):0] o_data;\n"
|
"\toutput\treg o_sync;\n"
|
"\toutput\treg o_sync;\n"
|
"\t\n", (dbg)?"_dbg":"", resetw.c_str(),
|
"\t\n", (dbg)?"_dbg":"", resetw.c_str(),
|
(dbg)?", o_dbg":"", TST_QTRSTAGE_IWIDTH,
|
(dbg)?", o_dbg":"", TST_QTRSTAGE_IWIDTH,
|
TST_QTRSTAGE_LGWIDTH, resetw.c_str());
|
TST_QTRSTAGE_LGWIDTH, resetw.c_str());
|
if (dbg) { fprintf(fp, "\toutput\twire\t[33:0]\t\t\to_dbg;\n"
|
if (dbg) { fprintf(fp, "\toutput\twire\t[33:0]\t\t\to_dbg;\n"
|
"\tassign\to_dbg = { ((o_sync)&&(i_ce)), i_ce, o_data[(2*OWIDTH-1):(2*OWIDTH-16)],\n"
|
"\tassign\to_dbg = { ((o_sync)&&(i_ce)), i_ce, o_data[(2*OWIDTH-1):(2*OWIDTH-16)],\n"
|
"\t\t\t\t\to_data[(OWIDTH-1):(OWIDTH-16)] };\n"
|
"\t\t\t\t\to_data[(OWIDTH-1):(OWIDTH-16)] };\n"
|
"\n");
|
"\n");
|
}
|
}
|
|
|
fprintf(fp,
|
fprintf(fp,
|
"\treg\t wait_for_sync;\n"
|
"\treg\t wait_for_sync;\n"
|
"\treg\t[2:0] pipeline;\n"
|
"\treg\t[2:0] pipeline;\n"
|
"\n"
|
"\n"
|
"\treg\tsigned [(IWIDTH):0] sum_r, sum_i, diff_r, diff_i;\n"
|
"\treg\tsigned [(IWIDTH):0] sum_r, sum_i, diff_r, diff_i;\n"
|
"\n"
|
"\n"
|
"\treg\t[(2*OWIDTH-1):0]\tob_a;\n"
|
"\treg\t[(2*OWIDTH-1):0]\tob_a;\n"
|
"\twire\t[(2*OWIDTH-1):0]\tob_b;\n"
|
"\twire\t[(2*OWIDTH-1):0]\tob_b;\n"
|
"\treg\t[(OWIDTH-1):0]\t\tob_b_r, ob_b_i;\n"
|
"\treg\t[(OWIDTH-1):0]\t\tob_b_r, ob_b_i;\n"
|
"\tassign\tob_b = { ob_b_r, ob_b_i };\n"
|
"\tassign\tob_b = { ob_b_r, ob_b_i };\n"
|
"\n"
|
"\n"
|
"\treg\t[(LGWIDTH-1):0]\t\tiaddr;\n"
|
"\treg\t[(LGWIDTH-1):0]\t\tiaddr;\n"
|
"\treg\t[(2*IWIDTH-1):0]\timem\t[0:1];\n"
|
"\treg\t[(2*IWIDTH-1):0]\timem\t[0:1];\n"
|
"\n"
|
"\n"
|
"\twire\tsigned\t[(IWIDTH-1):0]\timem_r, imem_i;\n"
|
"\twire\tsigned\t[(IWIDTH-1):0]\timem_r, imem_i;\n"
|
"\tassign\timem_r = imem[1][(2*IWIDTH-1):(IWIDTH)];\n"
|
"\tassign\timem_r = imem[1][(2*IWIDTH-1):(IWIDTH)];\n"
|
"\tassign\timem_i = imem[1][(IWIDTH-1):0];\n"
|
"\tassign\timem_i = imem[1][(IWIDTH-1):0];\n"
|
"\n"
|
"\n"
|
"\twire\tsigned\t[(IWIDTH-1):0]\ti_data_r, i_data_i;\n"
|
"\twire\tsigned\t[(IWIDTH-1):0]\ti_data_r, i_data_i;\n"
|
"\tassign\ti_data_r = i_data[(2*IWIDTH-1):(IWIDTH)];\n"
|
"\tassign\ti_data_r = i_data[(2*IWIDTH-1):(IWIDTH)];\n"
|
"\tassign\ti_data_i = i_data[(IWIDTH-1):0];\n"
|
"\tassign\ti_data_i = i_data[(IWIDTH-1):0];\n"
|
"\n"
|
"\n"
|
"\treg [(2*OWIDTH-1):0] omem [0:1];\n"
|
"\treg [(2*OWIDTH-1):0] omem [0:1];\n"
|
"\n");
|
"\n");
|
|
|
fprintf(fp, "\t//\n"
|
fprintf(fp, "\t//\n"
|
"\t// Round our output values down to OWIDTH bits\n"
|
"\t// Round our output values down to OWIDTH bits\n"
|
"\t//\n");
|
"\t//\n");
|
|
|
fprintf(fp,
|
fprintf(fp,
|
"\twire\tsigned\t[(OWIDTH-1):0]\trnd_sum_r, rnd_sum_i,\n"
|
"\twire\tsigned\t[(OWIDTH-1):0]\trnd_sum_r, rnd_sum_i,\n"
|
"\t\t\trnd_diff_r, rnd_diff_i, n_rnd_diff_r, n_rnd_diff_i;\n"
|
"\t\t\trnd_diff_r, rnd_diff_i, n_rnd_diff_r, n_rnd_diff_i;\n"
|
"\t%s #(IWIDTH+1,OWIDTH,SHIFT)\tdo_rnd_sum_r(i_clk, i_ce,\n"
|
"\t%s #(IWIDTH+1,OWIDTH,SHIFT)\tdo_rnd_sum_r(i_clk, i_ce,\n"
|
"\t\t\t\tsum_r, rnd_sum_r);\n\n", rnd_string);
|
"\t\t\t\tsum_r, rnd_sum_r);\n\n", rnd_string);
|
fprintf(fp,
|
fprintf(fp,
|
"\t%s #(IWIDTH+1,OWIDTH,SHIFT)\tdo_rnd_sum_i(i_clk, i_ce,\n"
|
"\t%s #(IWIDTH+1,OWIDTH,SHIFT)\tdo_rnd_sum_i(i_clk, i_ce,\n"
|
"\t\t\t\tsum_i, rnd_sum_i);\n\n", rnd_string);
|
"\t\t\t\tsum_i, rnd_sum_i);\n\n", rnd_string);
|
fprintf(fp,
|
fprintf(fp,
|
"\t%s #(IWIDTH+1,OWIDTH,SHIFT)\tdo_rnd_diff_r(i_clk, i_ce,\n"
|
"\t%s #(IWIDTH+1,OWIDTH,SHIFT)\tdo_rnd_diff_r(i_clk, i_ce,\n"
|
"\t\t\t\tdiff_r, rnd_diff_r);\n\n", rnd_string);
|
"\t\t\t\tdiff_r, rnd_diff_r);\n\n", rnd_string);
|
fprintf(fp,
|
fprintf(fp,
|
"\t%s #(IWIDTH+1,OWIDTH,SHIFT)\tdo_rnd_diff_i(i_clk, i_ce,\n"
|
"\t%s #(IWIDTH+1,OWIDTH,SHIFT)\tdo_rnd_diff_i(i_clk, i_ce,\n"
|
"\t\t\t\tdiff_i, rnd_diff_i);\n\n", rnd_string);
|
"\t\t\t\tdiff_i, rnd_diff_i);\n\n", rnd_string);
|
fprintf(fp, "\tassign n_rnd_diff_r = - rnd_diff_r;\n"
|
fprintf(fp, "\tassign n_rnd_diff_r = - rnd_diff_r;\n"
|
"\tassign n_rnd_diff_i = - rnd_diff_i;\n");
|
"\tassign n_rnd_diff_i = - rnd_diff_i;\n");
|
fprintf(fp,
|
fprintf(fp,
|
"\tinitial wait_for_sync = 1\'b1;\n"
|
"\tinitial wait_for_sync = 1\'b1;\n"
|
"\tinitial iaddr = 0;\n");
|
"\tinitial iaddr = 0;\n");
|
if (async_reset)
|
if (async_reset)
|
fprintf(fp,
|
fprintf(fp,
|
"\talways @(posedge i_clk, negedge i_areset_n)\n"
|
"\talways @(posedge i_clk, negedge i_areset_n)\n"
|
"\t\tif (!i_reset)\n");
|
"\t\tif (!i_reset)\n");
|
else
|
else
|
fprintf(fp,
|
fprintf(fp,
|
"\talways @(posedge i_clk)\n"
|
"\talways @(posedge i_clk)\n"
|
"\t\tif (i_reset)\n");
|
"\t\tif (i_reset)\n");
|
|
|
fprintf(fp, "\t\tbegin\n"
|
fprintf(fp, "\t\tbegin\n"
|
"\t\t\twait_for_sync <= 1\'b1;\n"
|
"\t\t\twait_for_sync <= 1\'b1;\n"
|
"\t\t\tiaddr <= 0;\n"
|
"\t\t\tiaddr <= 0;\n"
|
"\t\tend else if ((i_ce)&&((!wait_for_sync)||(i_sync)))\n"
|
"\t\tend else if ((i_ce)&&((!wait_for_sync)||(i_sync)))\n"
|
"\t\tbegin\n"
|
"\t\tbegin\n"
|
"\t\t\tiaddr <= iaddr + 1\'b1;\n"
|
"\t\t\tiaddr <= iaddr + 1\'b1;\n"
|
"\t\t\twait_for_sync <= 1\'b0;\n"
|
"\t\t\twait_for_sync <= 1\'b0;\n"
|
"\t\tend\n\n"
|
"\t\tend\n\n"
|
"\talways @(posedge i_clk)\n"
|
"\talways @(posedge i_clk)\n"
|
"\t\tif (i_ce)\n"
|
"\t\tif (i_ce)\n"
|
"\t\tbegin\n"
|
"\t\tbegin\n"
|
"\t\t\timem[0] <= i_data;\n"
|
"\t\t\timem[0] <= i_data;\n"
|
"\t\t\timem[1] <= imem[0];\n"
|
"\t\t\timem[1] <= imem[0];\n"
|
"\t\tend\n"
|
"\t\tend\n"
|
"\n\n");
|
"\n\n");
|
fprintf(fp,
|
fprintf(fp,
|
"\t// Note that we don\'t check on wait_for_sync or i_sync here.\n"
|
"\t// Note that we don\'t check on wait_for_sync or i_sync here.\n"
|
"\t// Why not? Because iaddr will always be zero until after the\n"
|
"\t// Why not? Because iaddr will always be zero until after the\n"
|
"\t// first i_ce, so we are safe.\n"
|
"\t// first i_ce, so we are safe.\n"
|
"\tinitial pipeline = 3\'h0;\n");
|
"\tinitial pipeline = 3\'h0;\n");
|
|
|
if (async_reset)
|
if (async_reset)
|
fprintf(fp,
|
fprintf(fp,
|
"\talways\t@(posedge i_clk, negedge i_areset_n)\n"
|
"\talways\t@(posedge i_clk, negedge i_areset_n)\n"
|
"\t\tif (!i_reset)\n");
|
"\t\tif (!i_reset)\n");
|
else
|
else
|
fprintf(fp,
|
fprintf(fp,
|
"\talways\t@(posedge i_clk)\n"
|
"\talways\t@(posedge i_clk)\n"
|
"\t\tif (i_reset)\n");
|
"\t\tif (i_reset)\n");
|
|
|
fprintf(fp,
|
fprintf(fp,
|
"\t\t\tpipeline <= 3\'h0;\n"
|
"\t\t\tpipeline <= 3\'h0;\n"
|
"\t\telse if (i_ce) // is our pipeline process full? Which stages?\n"
|
"\t\telse if (i_ce) // is our pipeline process full? Which stages?\n"
|
"\t\t\tpipeline <= { pipeline[1:0], iaddr[1] };\n\n");
|
"\t\t\tpipeline <= { pipeline[1:0], iaddr[1] };\n\n");
|
fprintf(fp,
|
fprintf(fp,
|
"\t// This is the pipeline[-1] stage, pipeline[0] will be set next.\n"
|
"\t// This is the pipeline[-1] stage, pipeline[0] will be set next.\n"
|
"\talways\t@(posedge i_clk)\n"
|
"\talways\t@(posedge i_clk)\n"
|
"\t\tif ((i_ce)&&(iaddr[1]))\n"
|
"\t\tif ((i_ce)&&(iaddr[1]))\n"
|
"\t\tbegin\n"
|
"\t\tbegin\n"
|
"\t\t\tsum_r <= imem_r + i_data_r;\n"
|
"\t\t\tsum_r <= imem_r + i_data_r;\n"
|
"\t\t\tsum_i <= imem_i + i_data_i;\n"
|
"\t\t\tsum_i <= imem_i + i_data_i;\n"
|
"\t\t\tdiff_r <= imem_r - i_data_r;\n"
|
"\t\t\tdiff_r <= imem_r - i_data_r;\n"
|
"\t\t\tdiff_i <= imem_i - i_data_i;\n"
|
"\t\t\tdiff_i <= imem_i - i_data_i;\n"
|
"\t\tend\n\n");
|
"\t\tend\n\n");
|
fprintf(fp,
|
fprintf(fp,
|
"\t// pipeline[1] takes sum_x and diff_x and produces rnd_x\n\n");
|
"\t// pipeline[1] takes sum_x and diff_x and produces rnd_x\n\n");
|
|
|
fprintf(fp,
|
fprintf(fp,
|
"\t// Now for pipeline[2]. We can actually do this at all i_ce\n"
|
"\t// Now for pipeline[2]. We can actually do this at all i_ce\n"
|
"\t// clock times, since nothing will listen unless pipeline[3]\n"
|
"\t// clock times, since nothing will listen unless pipeline[3]\n"
|
"\t// on the next clock. Thus, we simplify this logic and do\n"
|
"\t// on the next clock. Thus, we simplify this logic and do\n"
|
"\t// it independent of pipeline[2].\n"
|
"\t// it independent of pipeline[2].\n"
|
"\talways\t@(posedge i_clk)\n"
|
"\talways\t@(posedge i_clk)\n"
|
"\t\tif (i_ce)\n"
|
"\t\tif (i_ce)\n"
|
"\t\tbegin\n"
|
"\t\tbegin\n"
|
"\t\t\tob_a <= { rnd_sum_r, rnd_sum_i };\n"
|
"\t\t\tob_a <= { rnd_sum_r, rnd_sum_i };\n"
|
"\t\t\t// on Even, W = e^{-j2pi 1/4 0} = 1\n"
|
"\t\t\t// on Even, W = e^{-j2pi 1/4 0} = 1\n"
|
"\t\t\tif (!iaddr[0])\n"
|
"\t\t\tif (!iaddr[0])\n"
|
"\t\t\tbegin\n"
|
"\t\t\tbegin\n"
|
"\t\t\t\tob_b_r <= rnd_diff_r;\n"
|
"\t\t\t\tob_b_r <= rnd_diff_r;\n"
|
"\t\t\t\tob_b_i <= rnd_diff_i;\n"
|
"\t\t\t\tob_b_i <= rnd_diff_i;\n"
|
"\t\t\tend else if (INVERSE==0) begin\n"
|
"\t\t\tend else if (INVERSE==0) begin\n"
|
"\t\t\t\t// on Odd, W = e^{-j2pi 1/4} = -j\n"
|
"\t\t\t\t// on Odd, W = e^{-j2pi 1/4} = -j\n"
|
"\t\t\t\tob_b_r <= rnd_diff_i;\n"
|
"\t\t\t\tob_b_r <= rnd_diff_i;\n"
|
"\t\t\t\tob_b_i <= n_rnd_diff_r;\n"
|
"\t\t\t\tob_b_i <= n_rnd_diff_r;\n"
|
"\t\t\tend else begin\n"
|
"\t\t\tend else begin\n"
|
"\t\t\t\t// on Odd, W = e^{j2pi 1/4} = j\n"
|
"\t\t\t\t// on Odd, W = e^{j2pi 1/4} = j\n"
|
"\t\t\t\tob_b_r <= n_rnd_diff_i;\n"
|
"\t\t\t\tob_b_r <= n_rnd_diff_i;\n"
|
"\t\t\t\tob_b_i <= rnd_diff_r;\n"
|
"\t\t\t\tob_b_i <= rnd_diff_r;\n"
|
"\t\t\tend\n"
|
"\t\t\tend\n"
|
"\t\tend\n\n");
|
"\t\tend\n\n");
|
fprintf(fp,
|
fprintf(fp,
|
"\talways\t@(posedge i_clk)\n"
|
"\talways\t@(posedge i_clk)\n"
|
"\t\tif (i_ce)\n"
|
"\t\tif (i_ce)\n"
|
"\t\tbegin // In sequence, clock = 3\n"
|
"\t\tbegin // In sequence, clock = 3\n"
|
"\t\t\tomem[0] <= ob_b;\n"
|
"\t\t\tomem[0] <= ob_b;\n"
|
"\t\t\tomem[1] <= omem[0];\n"
|
"\t\t\tomem[1] <= omem[0];\n"
|
"\t\t\tif (pipeline[2])\n"
|
"\t\t\tif (pipeline[2])\n"
|
"\t\t\t\to_data <= ob_a;\n"
|
"\t\t\t\to_data <= ob_a;\n"
|
"\t\t\telse\n"
|
"\t\t\telse\n"
|
"\t\t\t\to_data <= omem[1];\n"
|
"\t\t\t\to_data <= omem[1];\n"
|
"\t\tend\n\n");
|
"\t\tend\n\n");
|
|
|
fprintf(fp,
|
fprintf(fp,
|
"\tinitial\to_sync = 1\'b0;\n");
|
"\tinitial\to_sync = 1\'b0;\n");
|
|
|
if (async_reset)
|
if (async_reset)
|
fprintf(fp,
|
fprintf(fp,
|
"\talways\t@(posedge i_clk, negedge i_areset_n)\n"
|
"\talways\t@(posedge i_clk, negedge i_areset_n)\n"
|
"\t\tif (!i_areset_n)\n");
|
"\t\tif (!i_areset_n)\n");
|
else
|
else
|
fprintf(fp,
|
fprintf(fp,
|
"\talways\t@(posedge i_clk)\n"
|
"\talways\t@(posedge i_clk)\n"
|
"\t\tif (i_reset)\n");
|
"\t\tif (i_reset)\n");
|
fprintf(fp,
|
fprintf(fp,
|
"\t\t\to_sync <= 1\'b0;\n"
|
"\t\t\to_sync <= 1\'b0;\n"
|
"\t\telse if (i_ce)\n"
|
"\t\telse if (i_ce)\n"
|
"\t\t\to_sync <= (iaddr[2:0] == 3'b101);\n\n");
|
"\t\t\to_sync <= (iaddr[2:0] == 3'b101);\n\n");
|
|
|
if (formal_property_flag) {
|
if (formal_property_flag) {
|
fprintf(fp,
|
fprintf(fp,
|
"`ifdef FORMAL\n"
|
"`ifdef FORMAL\n"
|
"\treg f_past_valid;\n"
|
"\treg f_past_valid;\n"
|
"\tinitial f_past_valid = 1'b0;\n"
|
"\tinitial f_past_valid = 1'b0;\n"
|
"\talways @(posedge i_clk)\n"
|
"\talways @(posedge i_clk)\n"
|
"\t f_past_valid = 1'b1;\n"
|
"\t f_past_valid = 1'b1;\n"
|
"\n"
|
"\n"
|
"`ifdef QTRSTAGE\n"
|
"`ifdef QTRSTAGE\n"
|
"\talways @(posedge i_clk)\n"
|
"\talways @(posedge i_clk)\n"
|
"\t assume((i_ce)||($past(i_ce))||($past(i_ce,2)));\n"
|
"\t assume((i_ce)||($past(i_ce))||($past(i_ce,2)));\n"
|
"`endif\n"
|
"`endif\n"
|
"\n"
|
"\n"
|
"\t// The below logic only works if the rounding stage does nothing\n"
|
"\t// The below logic only works if the rounding stage does nothing\n"
|
"\tinitial assert(IWIDTH+1 == OWIDTH);\n"
|
"\tinitial assert(IWIDTH+1 == OWIDTH);\n"
|
"\n"
|
"\n"
|
"\treg signed [IWIDTH-1:0] f_piped_real [0:7];\n"
|
"\treg signed [IWIDTH-1:0] f_piped_real [0:7];\n"
|
"\treg signed [IWIDTH-1:0] f_piped_imag [0:7];\n"
|
"\treg signed [IWIDTH-1:0] f_piped_imag [0:7];\n"
|
"\n"
|
"\n"
|
"\talways @(posedge i_clk)\n"
|
"\talways @(posedge i_clk)\n"
|
"\tif (i_ce)\n"
|
"\tif (i_ce)\n"
|
"\tbegin\n"
|
"\tbegin\n"
|
"\t f_piped_real[0] <= i_data[2*IWIDTH-1:IWIDTH];\n"
|
"\t f_piped_real[0] <= i_data[2*IWIDTH-1:IWIDTH];\n"
|
"\t f_piped_imag[0] <= i_data[ IWIDTH-1:0];\n"
|
"\t f_piped_imag[0] <= i_data[ IWIDTH-1:0];\n"
|
"\n"
|
"\n"
|
"\t f_piped_real[1] <= f_piped_real[0];\n"
|
"\t f_piped_real[1] <= f_piped_real[0];\n"
|
"\t f_piped_imag[1] <= f_piped_imag[0];\n"
|
"\t f_piped_imag[1] <= f_piped_imag[0];\n"
|
"\n"
|
"\n"
|
"\t f_piped_real[2] <= f_piped_real[1];\n"
|
"\t f_piped_real[2] <= f_piped_real[1];\n"
|
"\t f_piped_imag[2] <= f_piped_imag[1];\n"
|
"\t f_piped_imag[2] <= f_piped_imag[1];\n"
|
"\n"
|
"\n"
|
"\t f_piped_real[3] <= f_piped_real[2];\n"
|
"\t f_piped_real[3] <= f_piped_real[2];\n"
|
"\t f_piped_imag[3] <= f_piped_imag[2];\n"
|
"\t f_piped_imag[3] <= f_piped_imag[2];\n"
|
"\n"
|
"\n"
|
"\t f_piped_real[4] <= f_piped_real[3];\n"
|
"\t f_piped_real[4] <= f_piped_real[3];\n"
|
"\t f_piped_imag[4] <= f_piped_imag[3];\n"
|
"\t f_piped_imag[4] <= f_piped_imag[3];\n"
|
"\n"
|
"\n"
|
"\t f_piped_real[5] <= f_piped_real[4];\n"
|
"\t f_piped_real[5] <= f_piped_real[4];\n"
|
"\t f_piped_imag[5] <= f_piped_imag[4];\n"
|
"\t f_piped_imag[5] <= f_piped_imag[4];\n"
|
"\n"
|
"\n"
|
"\t f_piped_real[6] <= f_piped_real[5];\n"
|
"\t f_piped_real[6] <= f_piped_real[5];\n"
|
"\t f_piped_imag[6] <= f_piped_imag[5];\n"
|
"\t f_piped_imag[6] <= f_piped_imag[5];\n"
|
"\n"
|
"\n"
|
"\t f_piped_real[7] <= f_piped_real[6];\n"
|
"\t f_piped_real[7] <= f_piped_real[6];\n"
|
"\t f_piped_imag[7] <= f_piped_imag[6];\n"
|
"\t f_piped_imag[7] <= f_piped_imag[6];\n"
|
"\tend\n"
|
"\tend\n"
|
"\n"
|
"\n"
|
"\treg f_rsyncd;\n"
|
"\treg f_rsyncd;\n"
|
"\twire f_syncd;\n"
|
"\twire f_syncd;\n"
|
"\n"
|
"\n"
|
"\tinitial f_rsyncd = 0;\n"
|
"\tinitial f_rsyncd = 0;\n"
|
"\talways @(posedge i_clk)\n"
|
"\talways @(posedge i_clk)\n"
|
"\tif(i_reset)\n"
|
"\tif(i_reset)\n"
|
"\t f_rsyncd <= 1'b0;\n"
|
"\t f_rsyncd <= 1'b0;\n"
|
"\telse if (!f_rsyncd)\n"
|
"\telse if (!f_rsyncd)\n"
|
"\t f_rsyncd <= (o_sync);\n"
|
"\t f_rsyncd <= (o_sync);\n"
|
"\tassign f_syncd = (f_rsyncd)||(o_sync);\n"
|
"\tassign f_syncd = (f_rsyncd)||(o_sync);\n"
|
"\n"
|
"\n"
|
"\treg [1:0] f_state;\n"
|
"\treg [1:0] f_state;\n"
|
"\n"
|
"\n"
|
"\n"
|
"\n"
|
"\tinitial f_state = 0;\n"
|
"\tinitial f_state = 0;\n"
|
"\talways @(posedge i_clk)\n"
|
"\talways @(posedge i_clk)\n"
|
"\tif (i_reset)\n"
|
"\tif (i_reset)\n"
|
"\t f_state <= 0;\n"
|
"\t f_state <= 0;\n"
|
"\telse if ((i_ce)&&((!wait_for_sync)||(i_sync)))\n"
|
"\telse if ((i_ce)&&((!wait_for_sync)||(i_sync)))\n"
|
"\t f_state <= f_state + 1;\n"
|
"\t f_state <= f_state + 1;\n"
|
"\n"
|
"\n"
|
"\talways @(*)\n"
|
"\talways @(*)\n"
|
"\tif (f_state != 0)\n"
|
"\tif (f_state != 0)\n"
|
"\t assume(!i_sync);\n"
|
"\t assume(!i_sync);\n"
|
"\n"
|
"\n"
|
"\talways @(posedge i_clk)\n"
|
"\talways @(posedge i_clk)\n"
|
"\t assert(f_state[1:0] == iaddr[1:0]);\n"
|
"\t assert(f_state[1:0] == iaddr[1:0]);\n"
|
"\n"
|
"\n"
|
"\twire signed [2*IWIDTH-1:0] f_i_real, f_i_imag;\n"
|
"\twire signed [2*IWIDTH-1:0] f_i_real, f_i_imag;\n"
|
"\tassign f_i_real = i_data[2*IWIDTH-1:IWIDTH];\n"
|
"\tassign f_i_real = i_data[2*IWIDTH-1:IWIDTH];\n"
|
"\tassign f_i_imag = i_data[ IWIDTH-1:0];\n"
|
"\tassign f_i_imag = i_data[ IWIDTH-1:0];\n"
|
"\n"
|
"\n"
|
"\twire signed [OWIDTH-1:0] f_o_real, f_o_imag;\n"
|
"\twire signed [OWIDTH-1:0] f_o_real, f_o_imag;\n"
|
"\tassign f_o_real = o_data[2*OWIDTH-1:OWIDTH];\n"
|
"\tassign f_o_real = o_data[2*OWIDTH-1:OWIDTH];\n"
|
"\tassign f_o_imag = o_data[ OWIDTH-1:0];\n"
|
"\tassign f_o_imag = o_data[ OWIDTH-1:0];\n"
|
"\n"
|
"\n"
|
"\talways @(posedge i_clk)\n"
|
"\talways @(posedge i_clk)\n"
|
"\tif (f_state == 2'b11)\n"
|
"\tif (f_state == 2'b11)\n"
|
"\tbegin\n"
|
"\tbegin\n"
|
"\t assume(f_piped_real[0] != 3'sb100);\n"
|
"\t assume(f_piped_real[0] != 3'sb100);\n"
|
"\t assume(f_piped_real[2] != 3'sb100);\n"
|
"\t assume(f_piped_real[2] != 3'sb100);\n"
|
"\t assert(sum_r == f_piped_real[2] + f_piped_real[0]);\n"
|
"\t assert(sum_r == f_piped_real[2] + f_piped_real[0]);\n"
|
"\t assert(sum_i == f_piped_imag[2] + f_piped_imag[0]);\n"
|
"\t assert(sum_i == f_piped_imag[2] + f_piped_imag[0]);\n"
|
"\n"
|
"\n"
|
"\t assert(diff_r == f_piped_real[2] - f_piped_real[0]);\n"
|
"\t assert(diff_r == f_piped_real[2] - f_piped_real[0]);\n"
|
"\t assert(diff_i == f_piped_imag[2] - f_piped_imag[0]);\n"
|
"\t assert(diff_i == f_piped_imag[2] - f_piped_imag[0]);\n"
|
"\tend\n"
|
"\tend\n"
|
"\n"
|
"\n"
|
"\talways @(posedge i_clk)\n"
|
"\talways @(posedge i_clk)\n"
|
"\tif ((f_state == 2'b00)&&((f_syncd)||(iaddr >= 4)))\n"
|
"\tif ((f_state == 2'b00)&&((f_syncd)||(iaddr >= 4)))\n"
|
"\tbegin\n"
|
"\tbegin\n"
|
"\t assert(rnd_sum_r == f_piped_real[3]+f_piped_real[1]);\n"
|
"\t assert(rnd_sum_r == f_piped_real[3]+f_piped_real[1]);\n"
|
"\t assert(rnd_sum_i == f_piped_imag[3]+f_piped_imag[1]);\n"
|
"\t assert(rnd_sum_i == f_piped_imag[3]+f_piped_imag[1]);\n"
|
"\t assert(rnd_diff_r == f_piped_real[3]-f_piped_real[1]);\n"
|
"\t assert(rnd_diff_r == f_piped_real[3]-f_piped_real[1]);\n"
|
"\t assert(rnd_diff_i == f_piped_imag[3]-f_piped_imag[1]);\n"
|
"\t assert(rnd_diff_i == f_piped_imag[3]-f_piped_imag[1]);\n"
|
"\tend\n"
|
"\tend\n"
|
"\n"
|
"\n"
|
"\talways @(posedge i_clk)\n"
|
"\talways @(posedge i_clk)\n"
|
"\tif ((f_state == 2'b10)&&(f_syncd))\n"
|
"\tif ((f_state == 2'b10)&&(f_syncd))\n"
|
"\tbegin\n"
|
"\tbegin\n"
|
"\t // assert(o_sync);\n"
|
"\t // assert(o_sync);\n"
|
"\t assert(f_o_real == f_piped_real[5] + f_piped_real[3]);\n"
|
"\t assert(f_o_real == f_piped_real[5] + f_piped_real[3]);\n"
|
"\t assert(f_o_imag == f_piped_imag[5] + f_piped_imag[3]);\n"
|
"\t assert(f_o_imag == f_piped_imag[5] + f_piped_imag[3]);\n"
|
"\tend\n"
|
"\tend\n"
|
"\n"
|
"\n"
|
"\talways @(posedge i_clk)\n"
|
"\talways @(posedge i_clk)\n"
|
"\tif ((f_state == 2'b11)&&(f_syncd))\n"
|
"\tif ((f_state == 2'b11)&&(f_syncd))\n"
|
"\tbegin\n"
|
"\tbegin\n"
|
"\t assert(!o_sync);\n"
|
"\t assert(!o_sync);\n"
|
"\t assert(f_o_real == f_piped_real[5] + f_piped_real[3]);\n"
|
"\t assert(f_o_real == f_piped_real[5] + f_piped_real[3]);\n"
|
"\t assert(f_o_imag == f_piped_imag[5] + f_piped_imag[3]);\n"
|
"\t assert(f_o_imag == f_piped_imag[5] + f_piped_imag[3]);\n"
|
"\tend\n"
|
"\tend\n"
|
"\n"
|
"\n"
|
"\talways @(posedge i_clk)\n"
|
"\talways @(posedge i_clk)\n"
|
"\tif ((f_state == 2'b00)&&(f_syncd))\n"
|
"\tif ((f_state == 2'b00)&&(f_syncd))\n"
|
"\tbegin\n"
|
"\tbegin\n"
|
"\t assert(!o_sync);\n"
|
"\t assert(!o_sync);\n"
|
"\t assert(f_o_real == f_piped_real[7] - f_piped_real[5]);\n"
|
"\t assert(f_o_real == f_piped_real[7] - f_piped_real[5]);\n"
|
"\t assert(f_o_imag == f_piped_imag[7] - f_piped_imag[5]);\n"
|
"\t assert(f_o_imag == f_piped_imag[7] - f_piped_imag[5]);\n"
|
"\tend\n"
|
"\tend\n"
|
"\n"
|
"\n"
|
"\talways @(*)\n"
|
"\talways @(*)\n"
|
"\tif ((iaddr[2:0] == 0)&&(!wait_for_sync))\n"
|
"\tif ((iaddr[2:0] == 0)&&(!wait_for_sync))\n"
|
"\t assume(i_sync);\n"
|
"\t assume(i_sync);\n"
|
"\n"
|
"\n"
|
"\talways @(*)\n"
|
"\talways @(*)\n"
|
"\tif (wait_for_sync)\n"
|
"\tif (wait_for_sync)\n"
|
"\t assert((iaddr == 0)&&(f_state == 2'b00)&&(!o_sync)&&(!f_rsyncd));\n"
|
"\t assert((iaddr == 0)&&(f_state == 2'b00)&&(!o_sync)&&(!f_rsyncd));\n"
|
"\n"
|
"\n"
|
"\talways @(posedge i_clk)\n"
|
"\talways @(posedge i_clk)\n"
|
"\tif ((f_past_valid)&&($past(i_ce))&&($past(i_sync))&&(!$past(i_reset)))\n"
|
"\tif ((f_past_valid)&&($past(i_ce))&&($past(i_sync))&&(!$past(i_reset)))\n"
|
"\t assert(!wait_for_sync);\n"
|
"\t assert(!wait_for_sync);\n"
|
"\n"
|
"\n"
|
"\talways @(posedge i_clk)\n"
|
"\talways @(posedge i_clk)\n"
|
"\tif ((f_state == 2'b01)&&(f_syncd))\n"
|
"\tif ((f_state == 2'b01)&&(f_syncd))\n"
|
"\tbegin\n"
|
"\tbegin\n"
|
"\t assert(!o_sync);\n"
|
"\t assert(!o_sync);\n"
|
"\t if (INVERSE)\n"
|
"\t if (INVERSE)\n"
|
"\t begin\n"
|
"\t begin\n"
|
"\t assert(f_o_real == -f_piped_imag[7]+f_piped_imag[5]);\n"
|
"\t assert(f_o_real == -f_piped_imag[7]+f_piped_imag[5]);\n"
|
"\t assert(f_o_imag == f_piped_real[7]-f_piped_real[5]);\n"
|
"\t assert(f_o_imag == f_piped_real[7]-f_piped_real[5]);\n"
|
"\t end else begin\n"
|
"\t end else begin\n"
|
"\t assert(f_o_real == f_piped_imag[7]-f_piped_imag[5]);\n"
|
"\t assert(f_o_real == f_piped_imag[7]-f_piped_imag[5]);\n"
|
"\t assert(f_o_imag == -f_piped_real[7]+f_piped_real[5]);\n"
|
"\t assert(f_o_imag == -f_piped_real[7]+f_piped_real[5]);\n"
|
"\t end\n"
|
"\t end\n"
|
"\tend\n"
|
"\tend\n"
|
"\n"
|
"\n"
|
"`endif\n");
|
"`endif\n");
|
}
|
}
|
|
|
fprintf(fp, "endmodule\n");
|
fprintf(fp, "endmodule\n");
|
}
|
}
|
|
|
|
|
void build_sngllast(const char *fname, const bool async_reset = false) {
|
void build_sngllast(const char *fname, const bool async_reset = false) {
|
FILE *fp = fopen(fname, "w");
|
FILE *fp = fopen(fname, "w");
|
if (NULL == fp) {
|
if (NULL == fp) {
|
fprintf(stderr, "Could not open \'%s\' for writing\n", fname);
|
fprintf(stderr, "Could not open \'%s\' for writing\n", fname);
|
perror("O/S Err was:");
|
perror("O/S Err was:");
|
return;
|
return;
|
}
|
}
|
|
|
std::string resetw("i_reset");
|
std::string resetw("i_reset");
|
if (async_reset)
|
if (async_reset)
|
resetw = std::string("i_areset_n");
|
resetw = std::string("i_areset_n");
|
|
|
fprintf(fp,
|
fprintf(fp,
|
SLASHLINE
|
SLASHLINE
|
"//\n"
|
"//\n"
|
"// Filename:\tlaststage.v\n"
|
"// Filename:\tlaststage.v\n"
|
"//\n"
|
"//\n"
|
"// Project: %s\n"
|
"// Project: %s\n"
|
"//\n"
|
"//\n"
|
"// Purpose: This is part of an FPGA implementation that will process\n"
|
"// Purpose: This is part of an FPGA implementation that will process\n"
|
"// the final stage of a decimate-in-frequency FFT, running\n"
|
"// the final stage of a decimate-in-frequency FFT, running\n"
|
"// through the data at one sample per clock.\n"
|
"// through the data at one sample per clock.\n"
|
"//\n"
|
"//\n"
|
"//\n%s"
|
"//\n%s"
|
"//\n", prjname, creator);
|
"//\n", prjname, creator);
|
fprintf(fp, "%s", cpyleft);
|
fprintf(fp, "%s", cpyleft);
|
fprintf(fp, "//\n//\n`default_nettype\tnone\n//\n");
|
fprintf(fp, "//\n//\n`default_nettype\tnone\n//\n");
|
|
|
fprintf(fp,
|
fprintf(fp,
|
"module laststage(i_clk, %s, i_ce, i_sync, i_val, o_val, o_sync);\n"
|
"module laststage(i_clk, %s, i_ce, i_sync, i_val, o_val, o_sync);\n"
|
" parameter IWIDTH=16,OWIDTH=IWIDTH+1, SHIFT=0;\n"
|
" parameter IWIDTH=16,OWIDTH=IWIDTH+1, SHIFT=0;\n"
|
" input i_clk, %s, i_ce, i_sync;\n"
|
" input wire i_clk, %s, i_ce, i_sync;\n"
|
" input [(2*IWIDTH-1):0] i_val;\n"
|
" input wire [(2*IWIDTH-1):0] i_val;\n"
|
" output wire [(2*OWIDTH-1):0] o_val;\n"
|
" output wire [(2*OWIDTH-1):0] o_val;\n"
|
" output reg o_sync;\n\n",
|
" output reg o_sync;\n\n",
|
resetw.c_str(), resetw.c_str());
|
resetw.c_str(), resetw.c_str());
|
|
|
fprintf(fp,
|
fprintf(fp,
|
" reg signed [(IWIDTH-1):0] m_r, m_i;\n"
|
" reg signed [(IWIDTH-1):0] m_r, m_i;\n"
|
" wire signed [(IWIDTH-1):0] i_r, i_i;\n"
|
" wire signed [(IWIDTH-1):0] i_r, i_i;\n"
|
"\n"
|
"\n"
|
" assign i_r = i_val[(2*IWIDTH-1):(IWIDTH)]; \n"
|
" assign i_r = i_val[(2*IWIDTH-1):(IWIDTH)]; \n"
|
" assign i_i = i_val[(IWIDTH-1):0]; \n"
|
" assign i_i = i_val[(IWIDTH-1):0]; \n"
|
"\n"
|
"\n"
|
" // Don't forget that we accumulate a bit by adding two values\n"
|
" // Don't forget that we accumulate a bit by adding two values\n"
|
" // together. Therefore our intermediate value must have one more\n"
|
" // together. Therefore our intermediate value must have one more\n"
|
" // bit than the two originals.\n"
|
" // bit than the two originals.\n"
|
" reg signed [(IWIDTH):0] rnd_r, rnd_i, sto_r, sto_i;\n"
|
" reg signed [(IWIDTH):0] rnd_r, rnd_i, sto_r, sto_i;\n"
|
" reg wait_for_sync, stage;\n"
|
" reg wait_for_sync, stage;\n"
|
" reg [1:0] sync_pipe;\n"
|
" reg [1:0] sync_pipe;\n"
|
"\n"
|
"\n"
|
" initial wait_for_sync = 1'b1;\n"
|
" initial wait_for_sync = 1'b1;\n"
|
" initial stage = 1'b0;\n");
|
" initial stage = 1'b0;\n");
|
|
|
if (async_reset)
|
if (async_reset)
|
fprintf(fp, "\talways @(posedge i_clk, negedge i_areset_n)\n\t\tif (!i_areset_n)\n");
|
fprintf(fp, "\talways @(posedge i_clk, negedge i_areset_n)\n\t\tif (!i_areset_n)\n");
|
else
|
else
|
fprintf(fp, "\talways @(posedge i_clk)\n\t\tif (i_reset)\n");
|
fprintf(fp, "\talways @(posedge i_clk)\n\t\tif (i_reset)\n");
|
fprintf(fp,
|
fprintf(fp,
|
" begin\n"
|
" begin\n"
|
" wait_for_sync <= 1'b1;\n"
|
" wait_for_sync <= 1'b1;\n"
|
" stage <= 1'b0;\n"
|
" stage <= 1'b0;\n"
|
" end else if ((i_ce)&&((!wait_for_sync)||(i_sync))&&(!stage))\n"
|
" end else if ((i_ce)&&((!wait_for_sync)||(i_sync))&&(!stage))\n"
|
" begin\n"
|
" begin\n"
|
" wait_for_sync <= 1'b0;\n"
|
" wait_for_sync <= 1'b0;\n"
|
" //\n"
|
" //\n"
|
" stage <= 1'b1;\n"
|
" stage <= 1'b1;\n"
|
" //\n"
|
" //\n"
|
" end else if (i_ce)\n"
|
" end else if (i_ce)\n"
|
" stage <= 1'b0;\n\n");
|
" stage <= 1'b0;\n\n");
|
|
|
fprintf(fp, "\tinitial\tsync_pipe = 0;\n");
|
fprintf(fp, "\tinitial\tsync_pipe = 0;\n");
|
if (async_reset)
|
if (async_reset)
|
fprintf(fp,
|
fprintf(fp,
|
"\talways @(posedge i_clk, negedge i_areset_n)\n"
|
"\talways @(posedge i_clk, negedge i_areset_n)\n"
|
"\tif (!i_areset_n)\n");
|
"\tif (!i_areset_n)\n");
|
else
|
else
|
fprintf(fp,
|
fprintf(fp,
|
"\talways @(posedge i_clk)\n"
|
"\talways @(posedge i_clk)\n"
|
"\tif (i_reset)\n");
|
"\tif (i_reset)\n");
|
|
|
fprintf(fp,
|
fprintf(fp,
|
"\t\tsync_pipe <= 0;\n"
|
"\t\tsync_pipe <= 0;\n"
|
"\telse if (i_ce)\n"
|
"\telse if (i_ce)\n"
|
"\t\tsync_pipe <= { sync_pipe[0], i_sync };\n\n");
|
"\t\tsync_pipe <= { sync_pipe[0], i_sync };\n\n");
|
|
|
fprintf(fp, "\tinitial\to_sync = 1\'b0;\n");
|
fprintf(fp, "\tinitial\to_sync = 1\'b0;\n");
|
if (async_reset)
|
if (async_reset)
|
fprintf(fp,
|
fprintf(fp,
|
"\talways @(posedge i_clk, negedge i_areset_n)\n"
|
"\talways @(posedge i_clk, negedge i_areset_n)\n"
|
"\tif (!i_areset_n)\n");
|
"\tif (!i_areset_n)\n");
|
else
|
else
|
fprintf(fp,
|
fprintf(fp,
|
"\talways @(posedge i_clk)\n"
|
"\talways @(posedge i_clk)\n"
|
"\tif (i_reset)\n");
|
"\tif (i_reset)\n");
|
|
|
fprintf(fp,
|
fprintf(fp,
|
"\t\to_sync <= 1\'b0;\n"
|
"\t\to_sync <= 1\'b0;\n"
|
"\telse if (i_ce)\n"
|
"\telse if (i_ce)\n"
|
"\t\to_sync <= sync_pipe[1];\n\n");
|
"\t\to_sync <= sync_pipe[1];\n\n");
|
|
|
fprintf(fp,
|
fprintf(fp,
|
" always @(posedge i_clk)\n"
|
" always @(posedge i_clk)\n"
|
" if (i_ce)\n"
|
" if (i_ce)\n"
|
" begin\n"
|
" begin\n"
|
" if (!stage)\n"
|
" if (!stage)\n"
|
" begin\n"
|
" begin\n"
|
" // Clock 1\n"
|
" // Clock 1\n"
|
" m_r <= i_r;\n"
|
" m_r <= i_r;\n"
|
" m_i <= i_i;\n"
|
" m_i <= i_i;\n"
|
" // Clock 3\n"
|
" // Clock 3\n"
|
" rnd_r <= sto_r;\n"
|
" rnd_r <= sto_r;\n"
|
" rnd_i <= sto_i;\n"
|
" rnd_i <= sto_i;\n"
|
" //\n"
|
" //\n"
|
" end else begin\n"
|
" end else begin\n"
|
" // Clock 2\n"
|
" // Clock 2\n"
|
" rnd_r <= m_r + i_r;\n"
|
" rnd_r <= m_r + i_r;\n"
|
" rnd_i <= m_i + i_i;\n"
|
" rnd_i <= m_i + i_i;\n"
|
" //\n"
|
" //\n"
|
" sto_r <= m_r - i_r;\n"
|
" sto_r <= m_r - i_r;\n"
|
" sto_i <= m_i - i_i;\n"
|
" sto_i <= m_i - i_i;\n"
|
" //\n"
|
" //\n"
|
" end\n"
|
" end\n"
|
" end\n"
|
" end\n"
|
"\n"
|
"\n"
|
" // Now that we have our results, let's round them and report them\n"
|
" // Now that we have our results, let's round them and report them\n"
|
" wire signed [(OWIDTH-1):0] o_r, o_i;\n"
|
" wire signed [(OWIDTH-1):0] o_r, o_i;\n"
|
"\n"
|
"\n"
|
" convround #(IWIDTH+1,OWIDTH,SHIFT) do_rnd_r(i_clk, i_ce, rnd_r, o_r);\n"
|
" convround #(IWIDTH+1,OWIDTH,SHIFT) do_rnd_r(i_clk, i_ce, rnd_r, o_r);\n"
|
" convround #(IWIDTH+1,OWIDTH,SHIFT) do_rnd_i(i_clk, i_ce, rnd_i, o_i);\n"
|
" convround #(IWIDTH+1,OWIDTH,SHIFT) do_rnd_i(i_clk, i_ce, rnd_i, o_i);\n"
|
"\n"
|
"\n"
|
" assign o_val = { o_r, o_i };\n"
|
" assign o_val = { o_r, o_i };\n"
|
"\n");
|
"\n");
|
|
|
|
|
if (formal_property_flag) {
|
if (formal_property_flag) {
|
fprintf(fp,
|
fprintf(fp,
|
"`ifdef FORMAL\n"
|
"`ifdef FORMAL\n"
|
"\treg f_past_valid;\n"
|
"\treg f_past_valid;\n"
|
"\tinitial f_past_valid = 1'b0;\n"
|
"\tinitial f_past_valid = 1'b0;\n"
|
"\talways @(posedge i_clk)\n"
|
"\talways @(posedge i_clk)\n"
|
"\t f_past_valid <= 1'b1;\n"
|
"\t f_past_valid <= 1'b1;\n"
|
"\n"
|
"\n"
|
"`ifdef LASTSTAGE\n"
|
"`ifdef LASTSTAGE\n"
|
"\talways @(posedge i_clk)\n"
|
"\talways @(posedge i_clk)\n"
|
"\t assume((i_ce)||($past(i_ce))||($past(i_ce,2)));\n"
|
"\t assume((i_ce)||($past(i_ce))||($past(i_ce,2)));\n"
|
"`endif\n"
|
"`endif\n"
|
"\n"
|
"\n"
|
"\tinitial assert(IWIDTH+1 == OWIDTH);\n"
|
"\tinitial assert(IWIDTH+1 == OWIDTH);\n"
|
"\n"
|
"\n"
|
"\treg signed [IWIDTH-1:0] f_piped_real [0:3];\n"
|
"\treg signed [IWIDTH-1:0] f_piped_real [0:3];\n"
|
"\treg signed [IWIDTH-1:0] f_piped_imag [0:3];\n"
|
"\treg signed [IWIDTH-1:0] f_piped_imag [0:3];\n"
|
"\talways @(posedge i_clk)\n"
|
"\talways @(posedge i_clk)\n"
|
"\tif (i_ce)\n"
|
"\tif (i_ce)\n"
|
"\tbegin\n"
|
"\tbegin\n"
|
"\t f_piped_real[0] <= i_val[2*IWIDTH-1:IWIDTH];\n"
|
"\t f_piped_real[0] <= i_val[2*IWIDTH-1:IWIDTH];\n"
|
"\t f_piped_imag[0] <= i_val[ IWIDTH-1:0];\n"
|
"\t f_piped_imag[0] <= i_val[ IWIDTH-1:0];\n"
|
"\n"
|
"\n"
|
"\t f_piped_real[1] <= f_piped_real[0];\n"
|
"\t f_piped_real[1] <= f_piped_real[0];\n"
|
"\t f_piped_imag[1] <= f_piped_imag[0];\n"
|
"\t f_piped_imag[1] <= f_piped_imag[0];\n"
|
"\n"
|
"\n"
|
"\t f_piped_real[2] <= f_piped_real[1];\n"
|
"\t f_piped_real[2] <= f_piped_real[1];\n"
|
"\t f_piped_imag[2] <= f_piped_imag[1];\n"
|
"\t f_piped_imag[2] <= f_piped_imag[1];\n"
|
"\n"
|
"\n"
|
"\t f_piped_real[3] <= f_piped_real[2];\n"
|
"\t f_piped_real[3] <= f_piped_real[2];\n"
|
"\t f_piped_imag[3] <= f_piped_imag[2];\n"
|
"\t f_piped_imag[3] <= f_piped_imag[2];\n"
|
"\tend\n"
|
"\tend\n"
|
"\n"
|
"\n"
|
"\twire f_syncd;\n"
|
"\twire f_syncd;\n"
|
"\treg f_rsyncd;\n"
|
"\treg f_rsyncd;\n"
|
"\n"
|
"\n"
|
"\tinitial f_rsyncd = 0;\n"
|
"\tinitial f_rsyncd = 0;\n"
|
"\talways @(posedge i_clk)\n"
|
"\talways @(posedge i_clk)\n"
|
"\tif (i_reset)\n"
|
"\tif (i_reset)\n"
|
"\t f_rsyncd <= 1'b0;\n"
|
"\t f_rsyncd <= 1'b0;\n"
|
"\telse if (!f_rsyncd)\n"
|
"\telse if (!f_rsyncd)\n"
|
"\t f_rsyncd <= o_sync;\n"
|
"\t f_rsyncd <= o_sync;\n"
|
"\tassign f_syncd = (f_rsyncd)||(o_sync);\n"
|
"\tassign f_syncd = (f_rsyncd)||(o_sync);\n"
|
"\n"
|
"\n"
|
"\treg f_state;\n"
|
"\treg f_state;\n"
|
"\tinitial f_state = 0;\n"
|
"\tinitial f_state = 0;\n"
|
"\talways @(posedge i_clk)\n"
|
"\talways @(posedge i_clk)\n"
|
"\tif (i_reset)\n"
|
"\tif (i_reset)\n"
|
"\t f_state <= 0;\n"
|
"\t f_state <= 0;\n"
|
"\telse if ((i_ce)&&((!wait_for_sync)||(i_sync)))\n"
|
"\telse if ((i_ce)&&((!wait_for_sync)||(i_sync)))\n"
|
"\t f_state <= f_state + 1;\n"
|
"\t f_state <= f_state + 1;\n"
|
"\n"
|
"\n"
|
"\talways @(*)\n"
|
"\talways @(*)\n"
|
"\tif (f_state != 0)\n"
|
"\tif (f_state != 0)\n"
|
"\t assume(!i_sync);\n"
|
"\t assume(!i_sync);\n"
|
"\n"
|
"\n"
|
"\talways @(*)\n"
|
"\talways @(*)\n"
|
"\t assert(stage == f_state[0]);\n"
|
"\t assert(stage == f_state[0]);\n"
|
"\n"
|
"\n"
|
"\talways @(posedge i_clk)\n"
|
"\talways @(posedge i_clk)\n"
|
"\tif ((f_state == 1'b1)&&(f_syncd))\n"
|
"\tif ((f_state == 1'b1)&&(f_syncd))\n"
|
"\tbegin\n"
|
"\tbegin\n"
|
"\t assert(o_r == f_piped_real[2] + f_piped_real[1]);\n"
|
"\t assert(o_r == f_piped_real[2] + f_piped_real[1]);\n"
|
"\t assert(o_i == f_piped_imag[2] + f_piped_imag[1]);\n"
|
"\t assert(o_i == f_piped_imag[2] + f_piped_imag[1]);\n"
|
"\tend\n"
|
"\tend\n"
|
"\n"
|
"\n"
|
"\talways @(posedge i_clk)\n"
|
"\talways @(posedge i_clk)\n"
|
"\tif ((f_state == 1'b0)&&(f_syncd))\n"
|
"\tif ((f_state == 1'b0)&&(f_syncd))\n"
|
"\tbegin\n"
|
"\tbegin\n"
|
"\t assert(!o_sync);\n"
|
"\t assert(!o_sync);\n"
|
"\t assert(o_r == f_piped_real[3] - f_piped_real[2]);\n"
|
"\t assert(o_r == f_piped_real[3] - f_piped_real[2]);\n"
|
"\t assert(o_i == f_piped_imag[3] - f_piped_imag[2]);\n"
|
"\t assert(o_i == f_piped_imag[3] - f_piped_imag[2]);\n"
|
"\tend\n"
|
"\tend\n"
|
"\n"
|
"\n"
|
"\talways @(*)\n"
|
"\talways @(*)\n"
|
"\tif (wait_for_sync)\n"
|
"\tif (wait_for_sync)\n"
|
"\tbegin\n"
|
"\tbegin\n"
|
"\t assert(!f_rsyncd);\n"
|
"\t assert(!f_rsyncd);\n"
|
"\t assert(!o_sync);\n"
|
"\t assert(!o_sync);\n"
|
"\t assert(f_state == 0);\n"
|
"\t assert(f_state == 0);\n"
|
"\tend\n\n");
|
"\tend\n\n");
|
}
|
}
|
|
|
fprintf(fp,
|
fprintf(fp,
|
"`endif // FORMAL\n"
|
"`endif // FORMAL\n"
|
"endmodule\n");
|
"endmodule\n");
|
|
|
fclose(fp);
|
fclose(fp);
|
}
|
}
|
|
|
void usage(void) {
|
void usage(void) {
|
fprintf(stderr,
|
fprintf(stderr,
|
"USAGE:\tfftgen [-f <size>] [-d dir] [-c cbits] [-n nbits] [-m mxbits] [-s]\n"
|
"USAGE:\tfftgen [-f <size>] [-d dir] [-c cbits] [-n nbits] [-m mxbits] [-s]\n"
|
// "\tfftgen -i\n"
|
// "\tfftgen -i\n"
|
"\t-1\tBuild a normal FFT, running at one clock per complex sample, or\n"
|
"\t-1\tBuild a normal FFT, running at one clock per complex sample, or\n"
|
"\t\t(for a real FFT) at one clock per two real input samples.\n"
|
"\t\t(for a real FFT) at one clock per two real input samples.\n"
|
"\t-a <hdrname> Create a header of information describing the built-in\n"
|
"\t-a <hdrname> Create a header of information describing the built-in\n"
|
"\t\tparameters, useful for module-level testing with Verilator\n"
|
"\t\tparameters, useful for module-level testing with Verilator\n"
|
"\t-c <cbits>\tCauses all internal complex coefficients to be\n"
|
"\t-c <cbits>\tCauses all internal complex coefficients to be\n"
|
"\t\tlonger than the corresponding data bits, to help avoid\n"
|
"\t\tlonger than the corresponding data bits, to help avoid\n"
|
"\t\tcoefficient truncation errors. The default is %d bits longer\n"
|
"\t\tcoefficient truncation errors. The default is %d bits longer\n"
|
"\t\tthan the data bits.\n"
|
"\t\tthan the data bits.\n"
|
"\t-d <dir> Places all of the generated verilog files into <dir>.\n"
|
"\t-d <dir> Places all of the generated verilog files into <dir>.\n"
|
"\t\tThe default is a subdirectory of the current directory\n"
|
"\t\tThe default is a subdirectory of the current directory\n"
|
"\t\tnamed %s.\n"
|
"\t\tnamed %s.\n"
|
"\t-f <size> Sets the size of the FFT as the number of complex\n"
|
"\t-f <size> Sets the size of the FFT as the number of complex\n"
|
"\t\tsamples input to the transform. (No default value, this is\n"
|
"\t\tsamples input to the transform. (No default value, this is\n"
|
"\t\ta required parameter.)\n"
|
"\t\ta required parameter.)\n"
|
"\t-i\tAn inverse FFT, meaning that the coefficients are\n"
|
"\t-i\tAn inverse FFT, meaning that the coefficients are\n"
|
"\t\tgiven by e^{ j 2 pi k/N n }. The default is a forward FFT, with\n"
|
"\t\tgiven by e^{ j 2 pi k/N n }. The default is a forward FFT, with\n"
|
"\t\tcoefficients given by e^{ -j 2 pi k/N n }.\n"
|
"\t\tcoefficients given by e^{ -j 2 pi k/N n }.\n"
|
"\t-k #\tSets # clocks per sample, used to minimize multiplies. Also\n"
|
"\t-k #\tSets # clocks per sample, used to minimize multiplies. Also\n"
|
"\t\tsets one sample in per i_ce clock (opt -1)\n"
|
"\t\tsets one sample in per i_ce clock (opt -1)\n"
|
"\t-m <mxbits>\tSets the maximum bit width that the FFT should ever\n"
|
"\t-m <mxbits>\tSets the maximum bit width that the FFT should ever\n"
|
"\t\tproduce. Internal values greater than this value will be\n"
|
"\t\tproduce. Internal values greater than this value will be\n"
|
"\t\ttruncated to this value. (The default value grows the input\n"
|
"\t\ttruncated to this value. (The default value grows the input\n"
|
"\t\tsize by one bit for every two FFT stages.)\n"
|
"\t\tsize by one bit for every two FFT stages.)\n"
|
"\t-n <nbits>\tSets the bitwidth for values coming into the (i)FFT.\n"
|
"\t-n <nbits>\tSets the bitwidth for values coming into the (i)FFT.\n"
|
"\t\tThe default is %d bits input for each component of the two\n"
|
"\t\tThe default is %d bits input for each component of the two\n"
|
"\t\tcomplex values into the FFT.\n"
|
"\t\tcomplex values into the FFT.\n"
|
"\t-p <nmpy> Sets the number of hardware multiplies (DSPs) to use, versus\n"
|
"\t-p <nmpy> Sets the number of hardware multiplies (DSPs) to use, versus\n"
|
"\t\tshift-add emulation. The default is not to use any hardware\n"
|
"\t\tshift-add emulation. The default is not to use any hardware\n"
|
"\t\tmultipliers.\n"
|
"\t\tmultipliers.\n"
|
"\t-r\tBuild a real-FFT at four input points per sample, rather than a\n"
|
"\t-r\tBuild a real-FFT at four input points per sample, rather than a\n"
|
"\t\tcomplex FFT. (Default is a Complex FFT.)\n"
|
"\t\tcomplex FFT. (Default is a Complex FFT.)\n"
|
|
"\t\tThis option is a place-holder. The real-FFT has not (yet) been\n"
|
|
"\t\timplemented.\n"
|
"\t-s\tSkip the final bit reversal stage. This is useful in\n"
|
"\t-s\tSkip the final bit reversal stage. This is useful in\n"
|
"\t\talgorithms that need to apply a filter without needing to do\n"
|
"\t\talgorithms that need to apply a filter without needing to do\n"
|
"\t\tbin shifting, as these algorithms can, with this option, just\n"
|
"\t\tbin shifting, as these algorithms can, with this option, just\n"
|
"\t\tmultiply by a bit reversed correlation sequence and then\n"
|
"\t\tmultiply by a bit reversed correlation sequence and then\n"
|
"\t\tinverse FFT the (still bit reversed) result. (You would need\n"
|
"\t\tinverse FFT the (still bit reversed) result. (You would need\n"
|
"\t\ta decimation in time inverse to do this, which this program does\n"
|
"\t\ta decimation in time inverse to do this, which this program does\n"
|
"\t\tnot yet provide.)\n"
|
"\t\tnot yet provide.)\n"
|
"\t-S\tInclude the final bit reversal stage (default).\n"
|
"\t-S\tInclude the final bit reversal stage (default).\n"
|
"\t-x <xtrabits>\tUse this many extra bits internally, before any final\n"
|
"\t-x <xtrabits>\tUse this many extra bits internally, before any final\n"
|
"\t\trounding or truncation of the answer to the final number of\n"
|
"\t\trounding or truncation of the answer to the final number of\n"
|
"\t\tbits. The default is to use %d extra bits internally.\n",
|
"\t\tbits. The default is to use %d extra bits internally.\n",
|
/*
|
/*
|
"\t-0\tA forward FFT (default), meaning that the coefficients are\n"
|
"\t-0\tA forward FFT (default), meaning that the coefficients are\n"
|
"\t\tgiven by e^{-j 2 pi k/N n }.\n"
|
"\t\tgiven by e^{-j 2 pi k/N n }.\n"
|
"\t-1\tAn inverse FFT, meaning that the coefficients are\n"
|
"\t-1\tAn inverse FFT, meaning that the coefficients are\n"
|
"\t\tgiven by e^{ j 2 pi k/N n }.\n",
|
"\t\tgiven by e^{ j 2 pi k/N n }.\n",
|
*/
|
*/
|
DEF_XTRACBITS, DEF_COREDIR, DEF_NBITSIN, DEF_XTRAPBITS);
|
DEF_XTRACBITS, DEF_COREDIR, DEF_NBITSIN, DEF_XTRAPBITS);
|
}
|
}
|
|
|
// Features still needed:
|
// Features still needed:
|
// Interactivity.
|
// Interactivity.
|
int main(int argc, char **argv) {
|
int main(int argc, char **argv) {
|
int fftsize = -1, lgsize = -1;
|
int fftsize = -1, lgsize = -1;
|
int nbitsin = DEF_NBITSIN, xtracbits = DEF_XTRACBITS,
|
int nbitsin = DEF_NBITSIN, xtracbits = DEF_XTRACBITS,
|
nummpy=DEF_NMPY, nmpypstage=6, mpy_stages;
|
nummpy=DEF_NMPY, nmpypstage=6, mpy_stages;
|
int nbitsout, maxbitsout = -1, xtrapbits=DEF_XTRAPBITS, ckpce = 0;
|
int nbitsout, maxbitsout = -1, xtrapbits=DEF_XTRAPBITS, ckpce = 0;
|
const char *EMPTYSTR = "";
|
const char *EMPTYSTR = "";
|
bool bitreverse = true, inverse=false,
|
bool bitreverse = true, inverse=false,
|
verbose_flag = false,
|
verbose_flag = false,
|
single_clock = false,
|
single_clock = true,
|
real_fft = false,
|
real_fft = false,
|
async_reset = false;
|
async_reset = false;
|
FILE *vmain;
|
FILE *vmain;
|
std::string coredir = DEF_COREDIR, cmdline = "", hdrname = "";
|
std::string coredir = DEF_COREDIR, cmdline = "", hdrname = "";
|
ROUND_T rounding = RND_CONVERGENT;
|
ROUND_T rounding = RND_CONVERGENT;
|
// ROUND_T rounding = RND_HALFUP;
|
// ROUND_T rounding = RND_HALFUP;
|
|
|
bool dbg = false;
|
bool dbg = false;
|
int dbgstage = 128;
|
int dbgstage = 128;
|
|
|
if (argc <= 1)
|
if (argc <= 1)
|
usage();
|
usage();
|
|
|
// Copy the original command line before we mess with it
|
// Copy the original command line before we mess with it
|
cmdline = argv[0];
|
cmdline = argv[0];
|
for(int argn=1; argn<argc; argn++) {
|
for(int argn=1; argn<argc; argn++) {
|
cmdline += " ";
|
cmdline += " ";
|
cmdline += argv[argn];
|
cmdline += argv[argn];
|
}
|
}
|
|
|
{ int c;
|
{ int c;
|
while((c = getopt(argc, argv, "12Aa:c:d:D:f:hik:m:n:p:rsSx:v")) != -1) {
|
while((c = getopt(argc, argv, "12Aa:c:d:D:f:hik:m:n:p:rsSx:v")) != -1) {
|
switch(c) {
|
switch(c) {
|
case '1': single_clock = true; break;
|
case '1': single_clock = true; break;
|
case '2': single_clock = false; break;
|
case '2': single_clock = false; break;
|
case 'A': async_reset = true; break;
|
case 'A': async_reset = true; break;
|
case 'a': hdrname = strdup(optarg); break;
|
case 'a': hdrname = strdup(optarg); break;
|
case 'c': xtracbits = atoi(optarg); break;
|
case 'c': xtracbits = atoi(optarg); break;
|
case 'd': coredir = std::string(optarg); break;
|
case 'd': coredir = std::string(optarg); break;
|
case 'D': dbgstage = atoi(optarg); break;
|
case 'D': dbgstage = atoi(optarg); break;
|
case 'f': fftsize = atoi(optarg);
|
case 'f': fftsize = atoi(optarg);
|
{ int sln = strlen(optarg);
|
{ int sln = strlen(optarg);
|
if (!isdigit(optarg[sln-1])){
|
if (!isdigit(optarg[sln-1])){
|
switch(optarg[sln-1]) {
|
switch(optarg[sln-1]) {
|
case 'k': case 'K':
|
case 'k': case 'K':
|
fftsize <<= 10;
|
fftsize <<= 10;
|
break;
|
break;
|
case 'm': case 'M':
|
case 'm': case 'M':
|
fftsize <<= 20;
|
fftsize <<= 20;
|
break;
|
break;
|
case 'g': case 'G':
|
case 'g': case 'G':
|
fftsize <<= 30;
|
fftsize <<= 30;
|
break;
|
break;
|
default:
|
default:
|
printf("ERR: Unknown FFT size, %s!\n", optarg);
|
printf("ERR: Unknown FFT size, %s!\n", optarg);
|
exit(EXIT_FAILURE);
|
exit(EXIT_FAILURE);
|
}
|
}
|
}} break;
|
}} break;
|
case 'h': usage(); exit(EXIT_SUCCESS); break;
|
case 'h': usage(); exit(EXIT_SUCCESS); break;
|
case 'i': inverse = true; break;
|
case 'i': inverse = true; break;
|
case 'k': ckpce = atoi(optarg);
|
case 'k': ckpce = atoi(optarg);
|
single_clock = true;
|
single_clock = true;
|
break;
|
break;
|
case 'm': maxbitsout = atoi(optarg); break;
|
case 'm': maxbitsout = atoi(optarg); break;
|
case 'n': nbitsin = atoi(optarg); break;
|
case 'n': nbitsin = atoi(optarg); break;
|
case 'p': nummpy = atoi(optarg); break;
|
case 'p': nummpy = atoi(optarg); break;
|
case 'r': real_fft = true; break;
|
case 'r': real_fft = true; break;
|
case 'S': bitreverse = true; break;
|
case 'S': bitreverse = true; break;
|
case 's': bitreverse = false; break;
|
case 's': bitreverse = false; break;
|
case 'x': xtrapbits = atoi(optarg); break;
|
case 'x': xtrapbits = atoi(optarg); break;
|
case 'v': verbose_flag = true; break;
|
case 'v': verbose_flag = true; break;
|
// case 'z': variable_size = true; break;
|
// case 'z': variable_size = true; break;
|
default:
|
default:
|
printf("Unknown argument, -%c\n", c);
|
printf("Unknown argument, -%c\n", c);
|
usage();
|
usage();
|
exit(EXIT_FAILURE);
|
exit(EXIT_FAILURE);
|
}
|
}
|
}}
|
}}
|
|
|
if (verbose_flag) {
|
if (verbose_flag) {
|
if (inverse)
|
if (inverse)
|
printf("Building a %d point inverse FFT module, with %s outputs\n",
|
printf("Building a %d point inverse FFT module, with %s outputs\n",
|
fftsize,
|
fftsize,
|
(real_fft)?"real ":"complex");
|
(real_fft)?"real ":"complex");
|
else
|
else
|
printf("Building a %d point %sforward FFT module\n",
|
printf("Building a %d point %sforward FFT module\n",
|
fftsize,
|
fftsize,
|
(real_fft)?"real ":"");
|
(real_fft)?"real ":"");
|
if (!single_clock)
|
if (!single_clock)
|
printf(" that accepts two inputs per clock\n");
|
printf(" that accepts two inputs per clock\n");
|
if (async_reset)
|
if (async_reset)
|
printf(" using a negative logic ASYNC reset\n");
|
printf(" using a negative logic ASYNC reset\n");
|
|
|
printf("The core will be placed into the %s/ directory\n", coredir.c_str());
|
printf("The core will be placed into the %s/ directory\n", coredir.c_str());
|
|
|
if (hdrname[0])
|
if (hdrname[0])
|
printf("A C header file, %s, will be written capturing these\n"
|
printf("A C header file, %s, will be written capturing these\n"
|
"options for a Verilator testbench\n",
|
"options for a Verilator testbench\n",
|
hdrname.c_str());
|
hdrname.c_str());
|
// nummpy
|
// nummpy
|
// xtrapbits
|
// xtrapbits
|
}
|
}
|
|
|
if (real_fft) {
|
if (real_fft) {
|
printf("The real FFT option is not implemented yet, but still on\nmy to do list. Please try again later.\n");
|
printf("The real FFT option is not implemented yet, but still on\nmy to do list. Please try again later.\n");
|
exit(EXIT_FAILURE);
|
exit(EXIT_FAILURE);
|
}
|
}
|
|
|
if (ckpce < 1)
|
if (ckpce < 1)
|
ckpce = 1;
|
ckpce = 1;
|
if (!bitreverse) {
|
if (!bitreverse) {
|
printf("WARNING: While I can skip the bit reverse stage, the code to do\n");
|
printf("WARNING: While I can skip the bit reverse stage, the code to do\n");
|
printf("an inverse FFT on a bit--reversed input has not yet been\n");
|
printf("an inverse FFT on a bit--reversed input has not yet been\n");
|
printf("built.\n");
|
printf("built.\n");
|
}
|
}
|
|
|
if ((lgsize < 0)&&(fftsize > 1)) {
|
if ((lgsize < 0)&&(fftsize > 1)) {
|
for(lgsize=1; (1<<lgsize) < fftsize; lgsize++)
|
for(lgsize=1; (1<<lgsize) < fftsize; lgsize++)
|
;
|
;
|
}
|
}
|
|
|
if ((fftsize <= 0)||(nbitsin < 1)||(nbitsin>48)) {
|
if ((fftsize <= 0)||(nbitsin < 1)||(nbitsin>48)) {
|
printf("INVALID PARAMETERS!!!!\n");
|
printf("INVALID PARAMETERS!!!!\n");
|
exit(EXIT_FAILURE);
|
exit(EXIT_FAILURE);
|
}
|
}
|
|
|
|
|
if (nextlg(fftsize) != fftsize) {
|
if (nextlg(fftsize) != fftsize) {
|
fprintf(stderr, "ERR: FFTSize (%d) *must* be a power of two\n",
|
fprintf(stderr, "ERR: FFTSize (%d) *must* be a power of two\n",
|
fftsize);
|
fftsize);
|
exit(EXIT_FAILURE);
|
exit(EXIT_FAILURE);
|
} else if (fftsize < 2) {
|
} else if (fftsize < 2) {
|
fprintf(stderr, "ERR: Minimum FFTSize is 2, not %d\n",
|
fprintf(stderr, "ERR: Minimum FFTSize is 2, not %d\n",
|
fftsize);
|
fftsize);
|
if (fftsize == 1) {
|
if (fftsize == 1) {
|
fprintf(stderr, "You do realize that a 1 point FFT makes very little sense\n");
|
fprintf(stderr, "You do realize that a 1 point FFT makes very little sense\n");
|
fprintf(stderr, "in an FFT operation that handles two samples per clock?\n");
|
fprintf(stderr, "in an FFT operation that handles two samples per clock?\n");
|
fprintf(stderr, "If you really need to do an FFT of this size, the output\n");
|
fprintf(stderr, "If you really need to do an FFT of this size, the output\n");
|
fprintf(stderr, "can be connected straight to the input.\n");
|
fprintf(stderr, "can be connected straight to the input.\n");
|
} else {
|
} else {
|
fprintf(stderr, "Indeed, a size of %d doesn\'t make much sense to me at all.\n", fftsize);
|
fprintf(stderr, "Indeed, a size of %d doesn\'t make much sense to me at all.\n", fftsize);
|
fprintf(stderr, "Is such an operation even defined?\n");
|
fprintf(stderr, "Is such an operation even defined?\n");
|
}
|
}
|
exit(EXIT_FAILURE);
|
exit(EXIT_FAILURE);
|
}
|
}
|
|
|
// Calculate how many output bits we'll have, and what the log
|
// Calculate how many output bits we'll have, and what the log
|
// based two size of our FFT is.
|
// based two size of our FFT is.
|
{
|
{
|
int tmp_size = fftsize;
|
int tmp_size = fftsize;
|
|
|
// The first stage always accumulates one bit, regardless
|
// The first stage always accumulates one bit, regardless
|
// of whether you need to or not.
|
// of whether you need to or not.
|
nbitsout = nbitsin + 1;
|
nbitsout = nbitsin + 1;
|
tmp_size >>= 1;
|
tmp_size >>= 1;
|
|
|
while(tmp_size > 4) {
|
while(tmp_size > 4) {
|
nbitsout += 1;
|
nbitsout += 1;
|
tmp_size >>= 2;
|
tmp_size >>= 2;
|
}
|
}
|
|
|
if (tmp_size > 1)
|
if (tmp_size > 1)
|
nbitsout ++;
|
nbitsout ++;
|
|
|
if (fftsize <= 2)
|
if (fftsize <= 2)
|
bitreverse = false;
|
bitreverse = false;
|
} if ((maxbitsout > 0)&&(nbitsout > maxbitsout))
|
} if ((maxbitsout > 0)&&(nbitsout > maxbitsout))
|
nbitsout = maxbitsout;
|
nbitsout = maxbitsout;
|
|
|
if (verbose_flag) {
|
if (verbose_flag) {
|
printf("Output samples will be %d bits wide\n", nbitsout);
|
printf("Output samples will be %d bits wide\n", nbitsout);
|
printf("This %sFFT will take %d-bit samples in, and produce %d samples out\n", (inverse)?"i":"", nbitsin, nbitsout);
|
printf("This %sFFT will take %d-bit samples in, and produce %d samples out\n", (inverse)?"i":"", nbitsin, nbitsout);
|
if (maxbitsout > 0)
|
if (maxbitsout > 0)
|
printf(" Internally, it will allow items to accumulate to %d bits\n", maxbitsout);
|
printf(" Internally, it will allow items to accumulate to %d bits\n", maxbitsout);
|
printf(" Twiddle-factors of %d bits will be used\n",
|
printf(" Twiddle-factors of %d bits will be used\n",
|
nbitsin+xtracbits);
|
nbitsin+xtracbits);
|
if (!bitreverse)
|
if (!bitreverse)
|
printf(" The output will be left in bit-reversed order\n");
|
printf(" The output will be left in bit-reversed order\n");
|
}
|
}
|
|
|
// Figure out how many multiply stages to use, and how many to skip
|
// Figure out how many multiply stages to use, and how many to skip
|
if (!single_clock) {
|
if (!single_clock) {
|
nmpypstage = 6;
|
nmpypstage = 6;
|
} else if (ckpce <= 1) {
|
} else if (ckpce <= 1) {
|
nmpypstage = 3;
|
nmpypstage = 3;
|
} else if (ckpce == 2) {
|
} else if (ckpce == 2) {
|
nmpypstage = 2;
|
nmpypstage = 2;
|
} else
|
} else
|
nmpypstage = 1;
|
nmpypstage = 1;
|
|
|
mpy_stages = nummpy / nmpypstage;
|
mpy_stages = nummpy / nmpypstage;
|
if (mpy_stages > lgval(fftsize)-2)
|
if (mpy_stages > lgval(fftsize)-2)
|
mpy_stages = lgval(fftsize)-2;
|
mpy_stages = lgval(fftsize)-2;
|
|
|
{
|
{
|
struct stat sbuf;
|
struct stat sbuf;
|
if (lstat(coredir.c_str(), &sbuf)==0) {
|
if (lstat(coredir.c_str(), &sbuf)==0) {
|
if (!S_ISDIR(sbuf.st_mode)) {
|
if (!S_ISDIR(sbuf.st_mode)) {
|
fprintf(stderr, "\'%s\' already exists, and is not a directory!\n", coredir.c_str());
|
fprintf(stderr, "\'%s\' already exists, and is not a directory!\n", coredir.c_str());
|
fprintf(stderr, "I will stop now, lest I overwrite something you care about.\n");
|
fprintf(stderr, "I will stop now, lest I overwrite something you care about.\n");
|
fprintf(stderr, "To try again, please remove this file.\n");
|
fprintf(stderr, "To try again, please remove this file.\n");
|
exit(EXIT_FAILURE);
|
exit(EXIT_FAILURE);
|
}
|
}
|
} else
|
} else
|
mkdir(coredir.c_str(), 0755);
|
mkdir(coredir.c_str(), 0755);
|
if (access(coredir.c_str(), X_OK|W_OK) != 0) {
|
if (access(coredir.c_str(), X_OK|W_OK) != 0) {
|
fprintf(stderr, "I have no access to the directory \'%s\'.\n", coredir.c_str());
|
fprintf(stderr, "I have no access to the directory \'%s\'.\n", coredir.c_str());
|
exit(EXIT_FAILURE);
|
exit(EXIT_FAILURE);
|
}
|
}
|
}
|
}
|
|
|
if (hdrname.length() > 0) {
|
if (hdrname.length() > 0) {
|
FILE *hdr = fopen(hdrname.c_str(), "w");
|
FILE *hdr = fopen(hdrname.c_str(), "w");
|
if (hdr == NULL) {
|
if (hdr == NULL) {
|
fprintf(stderr, "ERROR: Cannot open %s to create header file\n", hdrname.c_str());
|
fprintf(stderr, "ERROR: Cannot open %s to create header file\n", hdrname.c_str());
|
perror("O/S Err:");
|
perror("O/S Err:");
|
exit(EXIT_FAILURE);
|
exit(EXIT_FAILURE);
|
}
|
}
|
|
|
fprintf(hdr,
|
fprintf(hdr,
|
SLASHLINE
|
SLASHLINE
|
"//\n"
|
"//\n"
|
"// Filename:\t%s\n"
|
"// Filename:\t%s\n"
|
"//\n"
|
"//\n"
|
"// Project:\t%s\n"
|
"// Project:\t%s\n"
|
"//\n"
|
"//\n"
|
"// Purpose: This simple header file captures the internal constants\n"
|
"// Purpose: This simple header file captures the internal constants\n"
|
"// within the FFT that were used to build it, for the purpose\n"
|
"// within the FFT that were used to build it, for the purpose\n"
|
"// of making C++ integration (and test bench testing) simpler. That is,\n"
|
"// of making C++ integration (and test bench testing) simpler. That is,\n"
|
"// should the FFT change size, this will note that size change and thus\n"
|
"// should the FFT change size, this will note that size change and thus\n"
|
"// any test bench or other C++ program dependent upon either the size of\n"
|
"// any test bench or other C++ program dependent upon either the size of\n"
|
"// the FFT, the number of bits in or out of it, etc., can pick up the\n"
|
"// the FFT, the number of bits in or out of it, etc., can pick up the\n"
|
"// changes in the defines found within this file.\n"
|
"// changes in the defines found within this file.\n"
|
"//\n",
|
"//\n",
|
hdrname.c_str(), prjname);
|
hdrname.c_str(), prjname);
|
fprintf(hdr, "%s", creator);
|
fprintf(hdr, "%s", creator);
|
fprintf(hdr, "//\n");
|
fprintf(hdr, "//\n");
|
fprintf(hdr, "%s", cpyleft);
|
fprintf(hdr, "%s", cpyleft);
|
fprintf(hdr, "//\n"
|
fprintf(hdr, "//\n"
|
"//\n"
|
"//\n"
|
"#ifndef %sFFTHDR_H\n"
|
"#ifndef %sFFTHDR_H\n"
|
"#define %sFFTHDR_H\n"
|
"#define %sFFTHDR_H\n"
|
"\n"
|
"\n"
|
"#define\t%sFFT_IWIDTH\t%d\n"
|
"#define\t%sFFT_IWIDTH\t%d\n"
|
"#define\t%sFFT_OWIDTH\t%d\n"
|
"#define\t%sFFT_OWIDTH\t%d\n"
|
"#define\t%sFFT_LGWIDTH\t%d\n"
|
"#define\t%sFFT_LGWIDTH\t%d\n"
|
"#define\t%sFFT_SIZE\t(1<<%sFFT_LGWIDTH)\n\n",
|
"#define\t%sFFT_SIZE\t(1<<%sFFT_LGWIDTH)\n\n",
|
(inverse)?"I":"", (inverse)?"I":"",
|
(inverse)?"I":"", (inverse)?"I":"",
|
(inverse)?"I":"", nbitsin,
|
(inverse)?"I":"", nbitsin,
|
(inverse)?"I":"", nbitsout,
|
(inverse)?"I":"", nbitsout,
|
(inverse)?"I":"", lgsize,
|
(inverse)?"I":"", lgsize,
|
(inverse)?"I":"", (inverse)?"I":"");
|
(inverse)?"I":"", (inverse)?"I":"");
|
if (ckpce > 0)
|
if (ckpce > 0)
|
fprintf(hdr, "#define\t%sFFT_CKPCE\t%d\t// Clocks per CE\n",
|
fprintf(hdr, "#define\t%sFFT_CKPCE\t%d\t// Clocks per CE\n",
|
(inverse)?"I":"", ckpce);
|
(inverse)?"I":"", ckpce);
|
else
|
else
|
fprintf(hdr, "// Two samples per i_ce\n");
|
fprintf(hdr, "// Two samples per i_ce\n");
|
if (!bitreverse)
|
if (!bitreverse)
|
fprintf(hdr, "#define\t%sFFT_SKIPS_BIT_REVERSE\n",
|
fprintf(hdr, "#define\t%sFFT_SKIPS_BIT_REVERSE\n",
|
(inverse)?"I":"");
|
(inverse)?"I":"");
|
if (real_fft)
|
if (real_fft)
|
fprintf(hdr, "#define\tRL%sFFT\n\n", (inverse)?"I":"");
|
fprintf(hdr, "#define\tRL%sFFT\n\n", (inverse)?"I":"");
|
if (!single_clock)
|
if (!single_clock)
|
fprintf(hdr, "#define\tDBLCLK%sFFT\n\n", (inverse)?"I":"");
|
fprintf(hdr, "#define\tDBLCLK%sFFT\n\n", (inverse)?"I":"");
|
else
|
else
|
fprintf(hdr, "// #define\tDBLCLK%sFFT // this FFT takes one input sample per clock\n\n", (inverse)?"I":"");
|
fprintf(hdr, "// #define\tDBLCLK%sFFT // this FFT takes one input sample per clock\n\n", (inverse)?"I":"");
|
if (USE_OLD_MULTIPLY)
|
if (USE_OLD_MULTIPLY)
|
fprintf(hdr, "#define\tUSE_OLD_MULTIPLY\n\n");
|
fprintf(hdr, "#define\tUSE_OLD_MULTIPLY\n\n");
|
|
|
fprintf(hdr, "// Parameters for testing the longbimpy\n");
|
fprintf(hdr, "// Parameters for testing the longbimpy\n");
|
fprintf(hdr, "#define\tTST_LONGBIMPY_AW\t%d\n", TST_LONGBIMPY_AW);
|
fprintf(hdr, "#define\tTST_LONGBIMPY_AW\t%d\n", TST_LONGBIMPY_AW);
|
#ifdef TST_LONGBIMPY_BW
|
#ifdef TST_LONGBIMPY_BW
|
fprintf(hdr, "#define\tTST_LONGBIMPY_BW\t%d\n\n", TST_LONGBIMPY_BW);
|
fprintf(hdr, "#define\tTST_LONGBIMPY_BW\t%d\n\n", TST_LONGBIMPY_BW);
|
#else
|
#else
|
fprintf(hdr, "#define\tTST_LONGBIMPY_BW\tTST_LONGBIMPY_AW\n\n");
|
fprintf(hdr, "#define\tTST_LONGBIMPY_BW\tTST_LONGBIMPY_AW\n\n");
|
#endif
|
#endif
|
|
|
fprintf(hdr, "// Parameters for testing the shift add multiply\n");
|
fprintf(hdr, "// Parameters for testing the shift add multiply\n");
|
fprintf(hdr, "#define\tTST_SHIFTADDMPY_AW\t%d\n", TST_SHIFTADDMPY_AW);
|
fprintf(hdr, "#define\tTST_SHIFTADDMPY_AW\t%d\n", TST_SHIFTADDMPY_AW);
|
#ifdef TST_SHIFTADDMPY_BW
|
#ifdef TST_SHIFTADDMPY_BW
|
fprintf(hdr, "#define\tTST_SHIFTADDMPY_BW\t%d\n\n", TST_SHIFTADDMPY_BW);
|
fprintf(hdr, "#define\tTST_SHIFTADDMPY_BW\t%d\n\n", TST_SHIFTADDMPY_BW);
|
#else
|
#else
|
fprintf(hdr, "#define\tTST_SHIFTADDMPY_BW\tTST_SHIFTADDMPY_AW\n\n");
|
fprintf(hdr, "#define\tTST_SHIFTADDMPY_BW\tTST_SHIFTADDMPY_AW\n\n");
|
#endif
|
#endif
|
|
|
#define TST_SHIFTADDMPY_AW 16
|
#define TST_SHIFTADDMPY_AW 16
|
#define TST_SHIFTADDMPY_BW 20 // Leave undefined to match AW
|
#define TST_SHIFTADDMPY_BW 20 // Leave undefined to match AW
|
fprintf(hdr, "// Parameters for testing the butterfly\n");
|
fprintf(hdr, "// Parameters for testing the butterfly\n");
|
fprintf(hdr, "#define\tTST_BUTTERFLY_IWIDTH\t%d\n", TST_BUTTERFLY_IWIDTH);
|
fprintf(hdr, "#define\tTST_BUTTERFLY_IWIDTH\t%d\n", TST_BUTTERFLY_IWIDTH);
|
fprintf(hdr, "#define\tTST_BUTTERFLY_CWIDTH\t%d\n", TST_BUTTERFLY_CWIDTH);
|
fprintf(hdr, "#define\tTST_BUTTERFLY_CWIDTH\t%d\n", TST_BUTTERFLY_CWIDTH);
|
fprintf(hdr, "#define\tTST_BUTTERFLY_OWIDTH\t%d\n", TST_BUTTERFLY_OWIDTH);
|
fprintf(hdr, "#define\tTST_BUTTERFLY_OWIDTH\t%d\n", TST_BUTTERFLY_OWIDTH);
|
fprintf(hdr, "#define\tTST_BUTTERFLY_MPYDELAY\t%d\n\n",
|
fprintf(hdr, "#define\tTST_BUTTERFLY_MPYDELAY\t%d\n\n",
|
bflydelay(TST_BUTTERFLY_IWIDTH,
|
bflydelay(TST_BUTTERFLY_IWIDTH,
|
TST_BUTTERFLY_CWIDTH-TST_BUTTERFLY_IWIDTH));
|
TST_BUTTERFLY_CWIDTH-TST_BUTTERFLY_IWIDTH));
|
|
|
fprintf(hdr, "// Parameters for testing the quarter stage\n");
|
fprintf(hdr, "// Parameters for testing the quarter stage\n");
|
fprintf(hdr, "#define\tTST_QTRSTAGE_IWIDTH\t%d\n", TST_QTRSTAGE_IWIDTH);
|
fprintf(hdr, "#define\tTST_QTRSTAGE_IWIDTH\t%d\n", TST_QTRSTAGE_IWIDTH);
|
fprintf(hdr, "#define\tTST_QTRSTAGE_LGWIDTH\t%d\n\n", TST_QTRSTAGE_LGWIDTH);
|
fprintf(hdr, "#define\tTST_QTRSTAGE_LGWIDTH\t%d\n\n", TST_QTRSTAGE_LGWIDTH);
|
|
|
fprintf(hdr, "// Parameters for testing the double stage\n");
|
fprintf(hdr, "// Parameters for testing the double stage\n");
|
fprintf(hdr, "#define\tTST_DBLSTAGE_IWIDTH\t%d\n", TST_DBLSTAGE_IWIDTH);
|
fprintf(hdr, "#define\tTST_DBLSTAGE_IWIDTH\t%d\n", TST_DBLSTAGE_IWIDTH);
|
fprintf(hdr, "#define\tTST_DBLSTAGE_SHIFT\t%d\n\n", TST_DBLSTAGE_SHIFT);
|
fprintf(hdr, "#define\tTST_DBLSTAGE_SHIFT\t%d\n\n", TST_DBLSTAGE_SHIFT);
|
|
|
fprintf(hdr, "// Parameters for testing the bit reversal stage\n");
|
fprintf(hdr, "// Parameters for testing the bit reversal stage\n");
|
fprintf(hdr, "#define\tTST_DBLREVERSE_LGSIZE\t%d\n\n", TST_DBLREVERSE_LGSIZE);
|
fprintf(hdr, "#define\tTST_DBLREVERSE_LGSIZE\t%d\n\n", TST_DBLREVERSE_LGSIZE);
|
fprintf(hdr, "\n" "#endif\n\n");
|
fprintf(hdr, "\n" "#endif\n\n");
|
fclose(hdr);
|
fclose(hdr);
|
}
|
}
|
|
|
{
|
{
|
std::string fname_string;
|
std::string fname_string;
|
|
|
fname_string = coredir;
|
fname_string = coredir;
|
fname_string += "/";
|
fname_string += "/";
|
if (inverse) fname_string += "i";
|
if (inverse) fname_string += "i";
|
fname_string += "fftmain.v";
|
fname_string += "fftmain.v";
|
|
|
vmain = fopen(fname_string.c_str(), "w");
|
vmain = fopen(fname_string.c_str(), "w");
|
if (NULL == vmain) {
|
if (NULL == vmain) {
|
fprintf(stderr, "Could not open \'%s\' for writing\n", fname_string.c_str());
|
fprintf(stderr, "Could not open \'%s\' for writing\n", fname_string.c_str());
|
perror("Err from O/S:");
|
perror("Err from O/S:");
|
exit(EXIT_FAILURE);
|
exit(EXIT_FAILURE);
|
}
|
}
|
|
|
if (verbose_flag)
|
if (verbose_flag)
|
printf("Opened %s\n", fname_string.c_str());
|
printf("Opened %s\n", fname_string.c_str());
|
}
|
}
|
|
|
fprintf(vmain,
|
fprintf(vmain,
|
SLASHLINE
|
SLASHLINE
|
"//\n"
|
"//\n"
|
"// Filename:\t%sfftmain.v\n"
|
"// Filename:\t%sfftmain.v\n"
|
"//\n"
|
"//\n"
|
"// Project: %s\n"
|
"// Project: %s\n"
|
"//\n"
|
"//\n"
|
"// Purpose: This is the main module in the General Purpose FPGA FFT\n"
|
"// Purpose: This is the main module in the General Purpose FPGA FFT\n"
|
"// implementation. As such, all other modules are subordinate\n"
|
"// implementation. As such, all other modules are subordinate\n"
|
"// to this one. This module accomplish a fixed size Complex FFT on\n"
|
"// to this one. This module accomplish a fixed size Complex FFT on\n"
|
"// %d data points.\n",
|
"// %d data points.\n",
|
(inverse)?"i":"",prjname, fftsize);
|
(inverse)?"i":"",prjname, fftsize);
|
if (single_clock) {
|
if (single_clock) {
|
fprintf(vmain,
|
fprintf(vmain,
|
"// The FFT is fully pipelined, and accepts as inputs one complex two\'s\n"
|
"// The FFT is fully pipelined, and accepts as inputs one complex two\'s\n"
|
"// complement sample per clock.\n");
|
"// complement sample per clock.\n");
|
} else {
|
} else {
|
fprintf(vmain,
|
fprintf(vmain,
|
"// The FFT is fully pipelined, and accepts as inputs two complex two\'s\n"
|
"// The FFT is fully pipelined, and accepts as inputs two complex two\'s\n"
|
"// complement samples per clock.\n");
|
"// complement samples per clock.\n");
|
}
|
}
|
|
|
fprintf(vmain,
|
fprintf(vmain,
|
"//\n"
|
"//\n"
|
"// Parameters:\n"
|
"// Parameters:\n"
|
"// i_clk\tThe clock. All operations are synchronous with this clock.\n"
|
"// i_clk\tThe clock. All operations are synchronous with this clock.\n"
|
"// i_%sreset%s\tSynchronous reset, active high. Setting this line will\n"
|
"// i_%sreset%s\tSynchronous reset, active high. Setting this line will\n"
|
"// \t\tforce the reset of all of the internals to this routine.\n"
|
"// \t\tforce the reset of all of the internals to this routine.\n"
|
"// \t\tFurther, following a reset, the o_sync line will go\n"
|
"// \t\tFurther, following a reset, the o_sync line will go\n"
|
"// \t\thigh the same time the first output sample is valid.\n",
|
"// \t\thigh the same time the first output sample is valid.\n",
|
(async_reset)?"a":"", (async_reset)?"_n":"");
|
(async_reset)?"a":"", (async_reset)?"_n":"");
|
if (single_clock) {
|
if (single_clock) {
|
fprintf(vmain,
|
fprintf(vmain,
|
"// i_ce\tA clock enable line. If this line is set, this module\n"
|
"// i_ce\tA clock enable line. If this line is set, this module\n"
|
"// \t\twill accept one complex input value, and produce\n"
|
"// \t\twill accept one complex input value, and produce\n"
|
"// \t\tone (possibly empty) complex output value.\n"
|
"// \t\tone (possibly empty) complex output value.\n"
|
"// i_sample\tThe complex input sample. This value is split\n"
|
"// i_sample\tThe complex input sample. This value is split\n"
|
"// \t\tinto two two\'s complement numbers, %d bits each, with\n"
|
"// \t\tinto two two\'s complement numbers, %d bits each, with\n"
|
"// \t\tthe real portion in the high order bits, and the\n"
|
"// \t\tthe real portion in the high order bits, and the\n"
|
"// \t\timaginary portion taking the bottom %d bits.\n"
|
"// \t\timaginary portion taking the bottom %d bits.\n"
|
"// o_result\tThe output result, of the same format as i_sample,\n"
|
"// o_result\tThe output result, of the same format as i_sample,\n"
|
"// \t\tonly having %d bits for each of the real and imaginary\n"
|
"// \t\tonly having %d bits for each of the real and imaginary\n"
|
"// \t\tcomponents, leading to %d bits total.\n"
|
"// \t\tcomponents, leading to %d bits total.\n"
|
"// o_sync\tA one bit output indicating the first sample of the FFT frame.\n"
|
"// o_sync\tA one bit output indicating the first sample of the FFT frame.\n"
|
"// \t\tIt also indicates the first valid sample out of the FFT\n"
|
"// \t\tIt also indicates the first valid sample out of the FFT\n"
|
"// \t\ton the first frame.\n", nbitsin, nbitsin, nbitsout, nbitsout*2);
|
"// \t\ton the first frame.\n", nbitsin, nbitsin, nbitsout, nbitsout*2);
|
} else {
|
} else {
|
fprintf(vmain,
|
fprintf(vmain,
|
"// i_ce\tA clock enable line. If this line is set, this module\n"
|
"// i_ce\tA clock enable line. If this line is set, this module\n"
|
"// \t\twill accept two complex values as inputs, and produce\n"
|
"// \t\twill accept two complex values as inputs, and produce\n"
|
"// \t\ttwo (possibly empty) complex values as outputs.\n"
|
"// \t\ttwo (possibly empty) complex values as outputs.\n"
|
"// i_left\tThe first of two complex input samples. This value is split\n"
|
"// i_left\tThe first of two complex input samples. This value is split\n"
|
"// \t\tinto two two\'s complement numbers, %d bits each, with\n"
|
"// \t\tinto two two\'s complement numbers, %d bits each, with\n"
|
"// \t\tthe real portion in the high order bits, and the\n"
|
"// \t\tthe real portion in the high order bits, and the\n"
|
"// \t\timaginary portion taking the bottom %d bits.\n"
|
"// \t\timaginary portion taking the bottom %d bits.\n"
|
"// i_right\tThis is the same thing as i_left, only this is the second of\n"
|
"// i_right\tThis is the same thing as i_left, only this is the second of\n"
|
"// \t\ttwo such samples. Hence, i_left would contain input\n"
|
"// \t\ttwo such samples. Hence, i_left would contain input\n"
|
"// \t\tsample zero, i_right would contain sample one. On the\n"
|
"// \t\tsample zero, i_right would contain sample one. On the\n"
|
"// \t\tnext clock i_left would contain input sample two,\n"
|
"// \t\tnext clock i_left would contain input sample two,\n"
|
"// \t\ti_right number three and so forth.\n"
|
"// \t\ti_right number three and so forth.\n"
|
"// o_left\tThe first of two output samples, of the same format as i_left,\n"
|
"// o_left\tThe first of two output samples, of the same format as i_left,\n"
|
"// \t\tonly having %d bits for each of the real and imaginary\n"
|
"// \t\tonly having %d bits for each of the real and imaginary\n"
|
"// \t\tcomponents, leading to %d bits total.\n"
|
"// \t\tcomponents, leading to %d bits total.\n"
|
"// o_right\tThe second of two output samples produced each clock. This has\n"
|
"// o_right\tThe second of two output samples produced each clock. This has\n"
|
"// \t\tthe same format as o_left.\n"
|
"// \t\tthe same format as o_left.\n"
|
"// o_sync\tA one bit output indicating the first valid sample produced by\n"
|
"// o_sync\tA one bit output indicating the first valid sample produced by\n"
|
"// \t\tthis FFT following a reset. Ever after, this will\n"
|
"// \t\tthis FFT following a reset. Ever after, this will\n"
|
"// \t\tindicate the first sample of an FFT frame.\n",
|
"// \t\tindicate the first sample of an FFT frame.\n",
|
nbitsin, nbitsin, nbitsout, nbitsout*2);
|
nbitsin, nbitsin, nbitsout, nbitsout*2);
|
}
|
}
|
|
|
fprintf(vmain,
|
fprintf(vmain,
|
"//\n"
|
"//\n"
|
"// Arguments:\tThis file was computer generated using the following command\n"
|
"// Arguments:\tThis file was computer generated using the following command\n"
|
"//\t\tline:\n"
|
"//\t\tline:\n"
|
"//\n");
|
"//\n");
|
fprintf(vmain, "//\t\t%% %s\n", cmdline.c_str());
|
fprintf(vmain, "//\t\t%% %s\n", cmdline.c_str());
|
fprintf(vmain, "//\n");
|
fprintf(vmain, "//\n");
|
|
fprintf(vmain, "//\tThis core will use hardware accelerated multiplies (DSPs)\n"
|
|
"//\tfor %d of the %d stages\n", mpy_stages, lgval(fftsize));
|
|
fprintf(vmain, "//\n");
|
fprintf(vmain, "%s", creator);
|
fprintf(vmain, "%s", creator);
|
fprintf(vmain, "//\n");
|
fprintf(vmain, "//\n");
|
fprintf(vmain, "%s", cpyleft);
|
fprintf(vmain, "%s", cpyleft);
|
fprintf(vmain, "//\n//\n`default_nettype\tnone\n//\n");
|
fprintf(vmain, "//\n//\n`default_nettype\tnone\n//\n");
|
|
|
|
|
std::string resetw("i_reset");
|
std::string resetw("i_reset");
|
if (async_reset)
|
if (async_reset)
|
resetw = "i_areset_n";
|
resetw = "i_areset_n";
|
|
|
fprintf(vmain, "//\n");
|
fprintf(vmain, "//\n");
|
fprintf(vmain, "//\n");
|
fprintf(vmain, "//\n");
|
fprintf(vmain, "module %sfftmain(i_clk, %s, i_ce,\n",
|
fprintf(vmain, "module %sfftmain(i_clk, %s, i_ce,\n",
|
(inverse)?"i":"", resetw.c_str());
|
(inverse)?"i":"", resetw.c_str());
|
if (single_clock) {
|
if (single_clock) {
|
fprintf(vmain, "\t\ti_sample, o_result, o_sync%s);\n",
|
fprintf(vmain, "\t\ti_sample, o_result, o_sync%s);\n",
|
(dbg)?", o_dbg":"");
|
(dbg)?", o_dbg":"");
|
} else {
|
} else {
|
fprintf(vmain, "\t\ti_left, i_right,\n");
|
fprintf(vmain, "\t\ti_left, i_right,\n");
|
fprintf(vmain, "\t\to_left, o_right, o_sync%s);\n",
|
fprintf(vmain, "\t\to_left, o_right, o_sync%s);\n",
|
(dbg)?", o_dbg":"");
|
(dbg)?", o_dbg":"");
|
}
|
}
|
fprintf(vmain, "\tparameter\tIWIDTH=%d, OWIDTH=%d, LGWIDTH=%d;\n\t//\n", nbitsin, nbitsout, lgsize);
|
fprintf(vmain,
|
|
"\t// The bit-width of the input, IWIDTH, output, OWIDTH, and the log\n"
|
|
"\t// of the FFT size. These are localparams, rather than parameters,\n"
|
|
"\t// because once the core has been generated, they can no longer be\n"
|
|
"\t// changed. (These values can be adjusted by running the core\n"
|
|
"\t// generator again.) The reason is simply that these values have\n"
|
|
"\t// been hardwired into the core at several places.\n");
|
|
fprintf(vmain, "\tlocalparam\tIWIDTH=%d, OWIDTH=%d, LGWIDTH=%d;\n\t//\n", nbitsin, nbitsout, lgsize);
|
assert(lgsize > 0);
|
assert(lgsize > 0);
|
fprintf(vmain, "\tinput\t\t\t\t\ti_clk, %s, i_ce;\n\t//\n",
|
fprintf(vmain, "\tinput\twire\t\t\t\ti_clk, %s, i_ce;\n\t//\n",
|
resetw.c_str());
|
resetw.c_str());
|
if (single_clock) {
|
if (single_clock) {
|
fprintf(vmain, "\tinput\t\t[(2*IWIDTH-1):0]\ti_sample;\n");
|
fprintf(vmain, "\tinput\twire\t[(2*IWIDTH-1):0]\ti_sample;\n");
|
fprintf(vmain, "\toutput\treg\t[(2*OWIDTH-1):0]\to_result;\n");
|
fprintf(vmain, "\toutput\treg\t[(2*OWIDTH-1):0]\to_result;\n");
|
} else {
|
} else {
|
fprintf(vmain, "\tinput\t\t[(2*IWIDTH-1):0]\ti_left, i_right;\n");
|
fprintf(vmain, "\tinput\twire\t[(2*IWIDTH-1):0]\ti_left, i_right;\n");
|
fprintf(vmain, "\toutput\treg\t[(2*OWIDTH-1):0]\to_left, o_right;\n");
|
fprintf(vmain, "\toutput\treg\t[(2*OWIDTH-1):0]\to_left, o_right;\n");
|
}
|
}
|
fprintf(vmain, "\toutput\treg\t\t\t\to_sync;\n");
|
fprintf(vmain, "\toutput\treg\t\t\t\to_sync;\n");
|
if (dbg)
|
if (dbg)
|
fprintf(vmain, "\toutput\twire\t[33:0]\t\to_dbg;\n");
|
fprintf(vmain, "\toutput\twire\t[33:0]\t\to_dbg;\n");
|
fprintf(vmain, "\n\n");
|
fprintf(vmain, "\n\n");
|
|
|
fprintf(vmain, "\t// Outputs of the FFT, ready for bit reversal.\n");
|
fprintf(vmain, "\t// Outputs of the FFT, ready for bit reversal.\n");
|
if (single_clock)
|
if (single_clock)
|
fprintf(vmain, "\twire\t[(2*OWIDTH-1):0]\tbr_sample;\n");
|
fprintf(vmain, "\twire\t[(2*OWIDTH-1):0]\tbr_sample;\n");
|
else
|
else
|
fprintf(vmain, "\twire\t[(2*OWIDTH-1):0]\tbr_left, br_right;\n");
|
fprintf(vmain, "\twire\t[(2*OWIDTH-1):0]\tbr_left, br_right;\n");
|
int tmp_size = fftsize, lgtmp = lgsize;
|
int tmp_size = fftsize, lgtmp = lgsize;
|
if (fftsize == 2) {
|
if (fftsize == 2) {
|
if (bitreverse) {
|
if (bitreverse) {
|
fprintf(vmain, "\treg\tbr_start;\n");
|
fprintf(vmain, "\treg\tbr_start;\n");
|
fprintf(vmain, "\tinitial br_start = 1\'b0;\n");
|
fprintf(vmain, "\tinitial br_start = 1\'b0;\n");
|
if (async_reset) {
|
if (async_reset) {
|
fprintf(vmain, "\talways @(posedge i_clk, negedge i_arese_n)\n");
|
fprintf(vmain, "\talways @(posedge i_clk, negedge i_arese_n)\n");
|
fprintf(vmain, "\t\tif (!i_areset_n)\n");
|
fprintf(vmain, "\t\tif (!i_areset_n)\n");
|
} else {
|
} else {
|
fprintf(vmain, "\talways @(posedge i_clk)\n");
|
fprintf(vmain, "\talways @(posedge i_clk)\n");
|
fprintf(vmain, "\t\tif (i_reset)\n");
|
fprintf(vmain, "\t\tif (i_reset)\n");
|
}
|
}
|
fprintf(vmain, "\t\t\tbr_start <= 1\'b0;\n");
|
fprintf(vmain, "\t\t\tbr_start <= 1\'b0;\n");
|
fprintf(vmain, "\t\telse if (i_ce)\n");
|
fprintf(vmain, "\t\telse if (i_ce)\n");
|
fprintf(vmain, "\t\t\tbr_start <= 1\'b1;\n");
|
fprintf(vmain, "\t\t\tbr_start <= 1\'b1;\n");
|
}
|
}
|
fprintf(vmain, "\n\n");
|
fprintf(vmain, "\n\n");
|
fprintf(vmain, "\tlaststage\t#(IWIDTH)\tstage_2(i_clk, %s, i_ce,\n", resetw.c_str());
|
fprintf(vmain, "\tlaststage\t#(IWIDTH)\tstage_2(i_clk, %s, i_ce,\n", resetw.c_str());
|
fprintf(vmain, "\t\t\t(%s%s), i_left, i_right, br_left, br_right);\n",
|
fprintf(vmain, "\t\t\t(%s%s), i_left, i_right, br_left, br_right);\n",
|
(async_reset)?"":"!", resetw.c_str());
|
(async_reset)?"":"!", resetw.c_str());
|
fprintf(vmain, "\n\n");
|
fprintf(vmain, "\n\n");
|
} else {
|
} else {
|
int nbits = nbitsin, dropbit=0;
|
int nbits = nbitsin, dropbit=0;
|
int obits = nbits+1+xtrapbits;
|
int obits = nbits+1+xtrapbits;
|
std::string cmem;
|
std::string cmem;
|
FILE *cmemfp;
|
FILE *cmemfp;
|
|
|
if ((maxbitsout > 0)&&(obits > maxbitsout))
|
if ((maxbitsout > 0)&&(obits > maxbitsout))
|
obits = maxbitsout;
|
obits = maxbitsout;
|
|
|
// Always do a first stage
|
// Always do a first stage
|
{
|
{
|
bool mpystage;
|
bool mpystage;
|
|
|
// Last two stages are always non-multiply stages
|
// Last two stages are always non-multiply stages
|
// since the multiplies can be done by adds
|
// since the multiplies can be done by adds
|
mpystage = ((lgtmp-2) <= mpy_stages);
|
mpystage = ((lgtmp-2) <= mpy_stages);
|
|
|
|
fprintf(vmain, "\n\n");
|
if (mpystage)
|
if (mpystage)
|
fprintf(vmain, "\t// A hardware optimized FFT stage\n");
|
fprintf(vmain, "\t// A hardware optimized FFT stage\n");
|
fprintf(vmain, "\n\n");
|
|
fprintf(vmain, "\twire\t\tw_s%d;\n", fftsize);
|
fprintf(vmain, "\twire\t\tw_s%d;\n", fftsize);
|
if (single_clock) {
|
if (single_clock) {
|
fprintf(vmain, "\twire\t[%d:0]\tw_d%d;\n", 2*(obits+xtrapbits)-1, fftsize);
|
fprintf(vmain, "\twire\t[%d:0]\tw_d%d;\n", 2*(obits+xtrapbits)-1, fftsize);
|
cmem = gen_coeff_fname(EMPTYSTR, fftsize, 1, 0, inverse);
|
cmem = gen_coeff_fname(coredir.c_str(), fftsize, 1, 0, inverse);
|
cmemfp = gen_coeff_open(cmem.c_str());
|
cmemfp = gen_coeff_open(cmem.c_str());
|
gen_coeffs(cmemfp, fftsize, nbitsin+xtracbits, 1, 0, inverse);
|
gen_coeffs(cmemfp, fftsize, nbitsin+xtracbits, 1, 0, inverse);
|
fprintf(vmain, "\tfftstage%s\t#(IWIDTH,IWIDTH+%d,%d,%d,%d,0,\n\t\t\t%d, %d, \"%s\")\n\t\tstage_%d(i_clk, %s, i_ce,\n",
|
cmem = gen_coeff_fname(EMPTYSTR, fftsize, 1, 0, inverse);
|
|
fprintf(vmain, "\tfftstage%s\t#(IWIDTH,IWIDTH+%d,%d,%d,0,\n\t\t\t%d, %d, \"%s\")\n\t\tstage_%d(i_clk, %s, i_ce,\n",
|
((dbg)&&(dbgstage == fftsize))?"_dbg":"",
|
((dbg)&&(dbgstage == fftsize))?"_dbg":"",
|
xtracbits, obits+xtrapbits,
|
xtracbits, obits+xtrapbits,
|
lgsize, lgtmp-1,
|
lgtmp-1, (mpystage)?1:0,
|
(mpystage)?1:0,
|
|
ckpce, cmem.c_str(),
|
ckpce, cmem.c_str(),
|
fftsize, resetw.c_str());
|
fftsize, resetw.c_str());
|
fprintf(vmain, "\t\t\t(%s%s), i_sample, w_d%d, w_s%d%s);\n",
|
fprintf(vmain, "\t\t\t(%s%s), i_sample, w_d%d, w_s%d%s);\n",
|
(async_reset)?"":"!", resetw.c_str(),
|
(async_reset)?"":"!", resetw.c_str(),
|
fftsize, fftsize,
|
fftsize, fftsize,
|
((dbg)&&(dbgstage == fftsize))
|
((dbg)&&(dbgstage == fftsize))
|
? ", o_dbg":"");
|
? ", o_dbg":"");
|
} else {
|
} else {
|
fprintf(vmain, "\t// verilator lint_off UNUSED\n\twire\t\tw_os%d;\n\t// verilator lint_on UNUSED\n", fftsize);
|
fprintf(vmain, "\t// verilator lint_off UNUSED\n\twire\t\tw_os%d;\n\t// verilator lint_on UNUSED\n", fftsize);
|
fprintf(vmain, "\twire\t[%d:0]\tw_e%d, w_o%d;\n", 2*(obits+xtrapbits)-1, fftsize, fftsize);
|
fprintf(vmain, "\twire\t[%d:0]\tw_e%d, w_o%d;\n", 2*(obits+xtrapbits)-1, fftsize, fftsize);
|
cmem = gen_coeff_fname(EMPTYSTR, fftsize, 2, 0, inverse);
|
cmem = gen_coeff_fname(coredir.c_str(), fftsize, 2, 0, inverse);
|
cmemfp = gen_coeff_open(cmem.c_str());
|
cmemfp = gen_coeff_open(cmem.c_str());
|
gen_coeffs(cmemfp, fftsize, nbitsin+xtracbits, 2, 0, inverse);
|
gen_coeffs(cmemfp, fftsize, nbitsin+xtracbits, 2, 0, inverse);
|
fprintf(vmain, "\tfftstage%s\t#(IWIDTH,IWIDTH+%d,%d,%d,%d,0,\n\t\t\t%d, %d, \"%s\")\n\t\tstage_e%d(i_clk, %s, i_ce,\n",
|
cmem = gen_coeff_fname(EMPTYSTR, fftsize, 2, 0, inverse);
|
|
fprintf(vmain, "\tfftstage%s\t#(IWIDTH,IWIDTH+%d,%d,%d,0,\n\t\t\t%d, %d, \"%s\")\n\t\tstage_e%d(i_clk, %s, i_ce,\n",
|
((dbg)&&(dbgstage == fftsize))?"_dbg":"",
|
((dbg)&&(dbgstage == fftsize))?"_dbg":"",
|
xtracbits, obits+xtrapbits,
|
xtracbits, obits+xtrapbits,
|
lgsize, lgtmp-2,
|
lgtmp-2, (mpystage)?1:0,
|
(mpystage)?1:0,
|
|
ckpce, cmem.c_str(),
|
ckpce, cmem.c_str(),
|
fftsize, resetw.c_str());
|
fftsize, resetw.c_str());
|
fprintf(vmain, "\t\t\t(%s%s), i_left, w_e%d, w_s%d%s);\n",
|
fprintf(vmain, "\t\t\t(%s%s), i_left, w_e%d, w_s%d%s);\n",
|
(async_reset)?"":"!", resetw.c_str(),
|
(async_reset)?"":"!", resetw.c_str(),
|
fftsize, fftsize,
|
fftsize, fftsize,
|
((dbg)&&(dbgstage == fftsize))?", o_dbg":"");
|
((dbg)&&(dbgstage == fftsize))?", o_dbg":"");
|
cmem = gen_coeff_fname(EMPTYSTR, fftsize, 2, 1, inverse);
|
cmem = gen_coeff_fname(coredir.c_str(), fftsize, 2, 1, inverse);
|
cmemfp = gen_coeff_open(cmem.c_str());
|
cmemfp = gen_coeff_open(cmem.c_str());
|
gen_coeffs(cmemfp, fftsize, nbitsin+xtracbits, 2, 1, inverse);
|
gen_coeffs(cmemfp, fftsize, nbitsin+xtracbits, 2, 1, inverse);
|
fprintf(vmain, "\tfftstage\t#(IWIDTH,IWIDTH+%d,%d,%d,%d,0,\n\t\t\t%d, %d, \"%s\")\n\t\tstage_o%d(i_clk, %s, i_ce,\n",
|
cmem = gen_coeff_fname(EMPTYSTR, fftsize, 2, 1, inverse);
|
|
fprintf(vmain, "\tfftstage\t#(IWIDTH,IWIDTH+%d,%d,%d,0,\n\t\t\t%d, %d, \"%s\")\n\t\tstage_o%d(i_clk, %s, i_ce,\n",
|
xtracbits, obits+xtrapbits,
|
xtracbits, obits+xtrapbits,
|
lgsize, lgtmp-2,
|
lgtmp-2, (mpystage)?1:0,
|
(mpystage)?1:0,
|
|
ckpce, cmem.c_str(),
|
ckpce, cmem.c_str(),
|
fftsize, resetw.c_str());
|
fftsize, resetw.c_str());
|
fprintf(vmain, "\t\t\t(%s%s), i_right, w_o%d, w_os%d);\n",
|
fprintf(vmain, "\t\t\t(%s%s), i_right, w_o%d, w_os%d);\n",
|
(async_reset)?"":"!",resetw.c_str(),
|
(async_reset)?"":"!",resetw.c_str(),
|
fftsize, fftsize);
|
fftsize, fftsize);
|
}
|
}
|
|
|
std::string fname;
|
std::string fname;
|
|
|
fname = coredir + "/";
|
fname = coredir + "/";
|
if (inverse)
|
if (inverse)
|
fname += "i";
|
fname += "i";
|
fname += "fftstage";
|
fname += "fftstage";
|
if (dbg) {
|
if (dbg) {
|
std::string dbgname(fname);
|
std::string dbgname(fname);
|
dbgname += "_dbg";
|
dbgname += "_dbg";
|
dbgname += ".v";
|
dbgname += ".v";
|
if (single_clock)
|
if (single_clock)
|
build_stage(fname.c_str(), fftsize, 1, 0, nbits, xtracbits, ckpce, async_reset, true);
|
build_stage(fname.c_str(), fftsize, 1, 0, nbits, xtracbits, ckpce, async_reset, true);
|
else
|
else
|
build_stage(fname.c_str(), fftsize/2, 2, 1, nbits, xtracbits, ckpce, async_reset, true);
|
build_stage(fname.c_str(), fftsize, 2, 1, nbits, xtracbits, ckpce, async_reset, true);
|
}
|
}
|
|
|
fname += ".v";
|
fname += ".v";
|
if (single_clock) {
|
if (single_clock) {
|
build_stage(fname.c_str(), fftsize, 1, 0,
|
build_stage(fname.c_str(), fftsize, 1, 0,
|
nbits, xtracbits, ckpce, async_reset,
|
nbits, xtracbits, ckpce, async_reset,
|
false);
|
false);
|
} else {
|
} else {
|
// All stages use the same Verilog, so we only
|
// All stages use the same Verilog, so we only
|
// need to build one
|
// need to build one
|
build_stage(fname.c_str(), fftsize/2, 2, 1,
|
build_stage(fname.c_str(), fftsize, 2, 1,
|
nbits, xtracbits, ckpce, async_reset, false);
|
nbits, xtracbits, ckpce, async_reset, false);
|
}
|
}
|
}
|
}
|
|
|
nbits = obits; // New number of input bits
|
nbits = obits; // New number of input bits
|
tmp_size >>= 1; lgtmp--;
|
tmp_size >>= 1; lgtmp--;
|
dropbit = 0;
|
dropbit = 0;
|
fprintf(vmain, "\n\n");
|
fprintf(vmain, "\n\n");
|
while(tmp_size >= 8) {
|
while(tmp_size >= 8) {
|
obits = nbits+((dropbit)?0:1);
|
obits = nbits+((dropbit)?0:1);
|
|
|
if ((maxbitsout > 0)&&(obits > maxbitsout))
|
if ((maxbitsout > 0)&&(obits > maxbitsout))
|
obits = maxbitsout;
|
obits = maxbitsout;
|
|
|
{
|
{
|
bool mpystage;
|
bool mpystage;
|
|
|
mpystage = ((lgtmp-2) <= mpy_stages);
|
mpystage = ((lgtmp-2) <= mpy_stages);
|
|
|
if (mpystage)
|
if (mpystage)
|
fprintf(vmain, "\t// A hardware optimized FFT stage\n");
|
fprintf(vmain, "\t// A hardware optimized FFT stage\n");
|
fprintf(vmain, "\twire\t\tw_s%d;\n",
|
fprintf(vmain, "\twire\t\tw_s%d;\n",
|
tmp_size);
|
tmp_size);
|
if (single_clock) {
|
if (single_clock) {
|
fprintf(vmain,"\twire\t[%d:0]\tw_d%d;\n",
|
fprintf(vmain,"\twire\t[%d:0]\tw_d%d;\n",
|
2*(obits+xtrapbits)-1,
|
2*(obits+xtrapbits)-1,
|
tmp_size);
|
tmp_size);
|
cmem = gen_coeff_fname(EMPTYSTR, tmp_size, 1, 0, inverse);
|
cmem = gen_coeff_fname(coredir.c_str(), tmp_size, 1, 0, inverse);
|
cmemfp = gen_coeff_open(cmem.c_str());
|
cmemfp = gen_coeff_open(cmem.c_str());
|
gen_coeffs(cmemfp, tmp_size,
|
gen_coeffs(cmemfp, tmp_size,
|
nbits+xtracbits+xtrapbits, 1, 0, inverse);
|
nbits+xtracbits+xtrapbits, 1, 0, inverse);
|
fprintf(vmain, "\tfftstage%s\t#(%d,%d,%d,%d,%d,%d,\n\t\t\t%d, %d, \"%s\")\n\t\tstage_%d(i_clk, %s, i_ce,\n",
|
cmem = gen_coeff_fname(EMPTYSTR, tmp_size, 1, 0, inverse);
|
|
fprintf(vmain, "\tfftstage%s\t#(%d,%d,%d,%d,%d,\n\t\t\t%d, %d, \"%s\")\n\t\tstage_%d(i_clk, %s, i_ce,\n",
|
((dbg)&&(dbgstage==tmp_size))?"_dbg":"",
|
((dbg)&&(dbgstage==tmp_size))?"_dbg":"",
|
nbits+xtrapbits,
|
nbits+xtrapbits,
|
nbits+xtracbits+xtrapbits,
|
nbits+xtracbits+xtrapbits,
|
obits+xtrapbits,
|
obits+xtrapbits,
|
lgsize, lgtmp-1,
|
lgtmp-1, (dropbit)?0:0, (mpystage)?1:0,
|
(dropbit)?0:0, (mpystage)?1:0,
|
|
ckpce,
|
ckpce,
|
cmem.c_str(), tmp_size,
|
cmem.c_str(), tmp_size,
|
resetw.c_str());
|
resetw.c_str());
|
fprintf(vmain, "\t\t\tw_s%d, w_d%d, w_d%d, w_s%d%s);\n",
|
fprintf(vmain, "\t\t\tw_s%d, w_d%d, w_d%d, w_s%d%s);\n",
|
tmp_size<<1, tmp_size<<1,
|
tmp_size<<1, tmp_size<<1,
|
tmp_size, tmp_size,
|
tmp_size, tmp_size,
|
((dbg)&&(dbgstage == tmp_size))
|
((dbg)&&(dbgstage == tmp_size))
|
?", o_dbg":"");
|
?", o_dbg":"");
|
} else {
|
} else {
|
fprintf(vmain, "\t// verilator lint_off UNUSED\n\twire\t\tw_os%d;\n\t// verilator lint_on UNUSED\n",
|
fprintf(vmain, "\t// verilator lint_off UNUSED\n\twire\t\tw_os%d;\n\t// verilator lint_on UNUSED\n",
|
tmp_size);
|
tmp_size);
|
fprintf(vmain,"\twire\t[%d:0]\tw_e%d, w_o%d;\n",
|
fprintf(vmain,"\twire\t[%d:0]\tw_e%d, w_o%d;\n",
|
2*(obits+xtrapbits)-1,
|
2*(obits+xtrapbits)-1,
|
tmp_size, tmp_size);
|
tmp_size, tmp_size);
|
cmem = gen_coeff_fname(EMPTYSTR, tmp_size, 2, 0, inverse);
|
cmem = gen_coeff_fname(coredir.c_str(), tmp_size, 2, 0, inverse);
|
cmemfp = gen_coeff_open(cmem.c_str());
|
cmemfp = gen_coeff_open(cmem.c_str());
|
gen_coeffs(cmemfp, tmp_size,
|
gen_coeffs(cmemfp, tmp_size,
|
nbits+xtracbits+xtrapbits, 2, 0, inverse);
|
nbits+xtracbits+xtrapbits, 2, 0, inverse);
|
fprintf(vmain, "\tfftstage%s\t#(%d,%d,%d,%d,%d,%d,\n\t\t\t%d, %d, \"%s\")\n\t\tstage_e%d(i_clk, %s, i_ce,\n",
|
cmem = gen_coeff_fname(EMPTYSTR, tmp_size, 2, 0, inverse);
|
|
fprintf(vmain, "\tfftstage%s\t#(%d,%d,%d,%d,%d,\n\t\t\t%d, %d, \"%s\")\n\t\tstage_e%d(i_clk, %s, i_ce,\n",
|
((dbg)&&(dbgstage==tmp_size))?"_dbg":"",
|
((dbg)&&(dbgstage==tmp_size))?"_dbg":"",
|
nbits+xtrapbits,
|
nbits+xtrapbits,
|
nbits+xtracbits+xtrapbits,
|
nbits+xtracbits+xtrapbits,
|
obits+xtrapbits,
|
obits+xtrapbits,
|
lgsize, lgtmp-2,
|
lgtmp-2, (dropbit)?0:0, (mpystage)?1:0,
|
(dropbit)?0:0, (mpystage)?1:0,
|
|
ckpce,
|
ckpce,
|
cmem.c_str(), tmp_size,
|
cmem.c_str(), tmp_size,
|
resetw.c_str());
|
resetw.c_str());
|
fprintf(vmain, "\t\t\tw_s%d, w_e%d, w_e%d, w_s%d%s);\n",
|
fprintf(vmain, "\t\t\tw_s%d, w_e%d, w_e%d, w_s%d%s);\n",
|
tmp_size<<1, tmp_size<<1,
|
tmp_size<<1, tmp_size<<1,
|
tmp_size, tmp_size,
|
tmp_size, tmp_size,
|
((dbg)&&(dbgstage == tmp_size))
|
((dbg)&&(dbgstage == tmp_size))
|
?", o_dbg":"");
|
?", o_dbg":"");
|
cmem = gen_coeff_fname(EMPTYSTR,
|
cmem = gen_coeff_fname(coredir.c_str(),
|
tmp_size, 2, 1, inverse);
|
tmp_size, 2, 1, inverse);
|
cmemfp = gen_coeff_open(cmem.c_str());
|
cmemfp = gen_coeff_open(cmem.c_str());
|
gen_coeffs(cmemfp, tmp_size,
|
gen_coeffs(cmemfp, tmp_size,
|
nbits+xtracbits+xtrapbits,
|
nbits+xtracbits+xtrapbits,
|
2, 1, inverse);
|
2, 1, inverse);
|
fprintf(vmain, "\tfftstage\t#(%d,%d,%d,%d,%d,%d,\n\t\t\t%d, %d, \"%s\")\n\t\tstage_o%d(i_clk, %s, i_ce,\n",
|
cmem = gen_coeff_fname(EMPTYSTR,
|
|
tmp_size, 2, 1, inverse);
|
|
fprintf(vmain, "\tfftstage\t#(%d,%d,%d,%d,%d,\n\t\t\t%d, %d, \"%s\")\n\t\tstage_o%d(i_clk, %s, i_ce,\n",
|
nbits+xtrapbits,
|
nbits+xtrapbits,
|
nbits+xtracbits+xtrapbits,
|
nbits+xtracbits+xtrapbits,
|
obits+xtrapbits,
|
obits+xtrapbits,
|
lgsize, lgtmp-2,
|
lgtmp-2, (dropbit)?0:0, (mpystage)?1:0,
|
(dropbit)?0:0, (mpystage)?1:0,
|
|
ckpce, cmem.c_str(), tmp_size,
|
ckpce, cmem.c_str(), tmp_size,
|
resetw.c_str());
|
resetw.c_str());
|
fprintf(vmain, "\t\t\tw_s%d, w_o%d, w_o%d, w_os%d);\n",
|
fprintf(vmain, "\t\t\tw_s%d, w_o%d, w_o%d, w_os%d);\n",
|
tmp_size<<1, tmp_size<<1,
|
tmp_size<<1, tmp_size<<1,
|
tmp_size, tmp_size);
|
tmp_size, tmp_size);
|
}
|
}
|
fprintf(vmain, "\n");
|
fprintf(vmain, "\n");
|
}
|
}
|
|
|
|
|
dropbit ^= 1;
|
dropbit ^= 1;
|
nbits = obits;
|
nbits = obits;
|
tmp_size >>= 1; lgtmp--;
|
tmp_size >>= 1; lgtmp--;
|
}
|
}
|
|
|
if (tmp_size == 4) {
|
if (tmp_size == 4) {
|
obits = nbits+((dropbit)?0:1);
|
obits = nbits+((dropbit)?0:1);
|
|
|
if ((maxbitsout > 0)&&(obits > maxbitsout))
|
if ((maxbitsout > 0)&&(obits > maxbitsout))
|
obits = maxbitsout;
|
obits = maxbitsout;
|
|
|
fprintf(vmain, "\twire\t\tw_s4;\n");
|
fprintf(vmain, "\twire\t\tw_s4;\n");
|
if (single_clock) {
|
if (single_clock) {
|
fprintf(vmain, "\twire\t[%d:0]\tw_d4;\n",
|
fprintf(vmain, "\twire\t[%d:0]\tw_d4;\n",
|
2*(obits+xtrapbits)-1);
|
2*(obits+xtrapbits)-1);
|
fprintf(vmain, "\tqtrstage%s\t#(%d,%d,%d,%d,%d)\tstage_4(i_clk, %s, i_ce,\n",
|
fprintf(vmain, "\tqtrstage%s\t#(%d,%d,%d,%d,%d)\tstage_4(i_clk, %s, i_ce,\n",
|
((dbg)&&(dbgstage==4))?"_dbg":"",
|
((dbg)&&(dbgstage==4))?"_dbg":"",
|
nbits+xtrapbits, obits+xtrapbits, lgsize,
|
nbits+xtrapbits, obits+xtrapbits, lgsize,
|
(inverse)?1:0, (dropbit)?0:0,
|
(inverse)?1:0, (dropbit)?0:0,
|
resetw.c_str());
|
resetw.c_str());
|
fprintf(vmain, "\t\t\t\t\t\tw_s8, w_d8, w_d4, w_s4%s);\n",
|
fprintf(vmain, "\t\t\t\t\t\tw_s8, w_d8, w_d4, w_s4%s);\n",
|
((dbg)&&(dbgstage==4))?", o_dbg":"");
|
((dbg)&&(dbgstage==4))?", o_dbg":"");
|
} else {
|
} else {
|
fprintf(vmain, "\t// verilator lint_off UNUSED\n\twire\t\tw_os4;\n\t// verilator lint_on UNUSED\n");
|
fprintf(vmain, "\t// verilator lint_off UNUSED\n\twire\t\tw_os4;\n\t// verilator lint_on UNUSED\n");
|
fprintf(vmain, "\twire\t[%d:0]\tw_e4, w_o4;\n", 2*(obits+xtrapbits)-1);
|
fprintf(vmain, "\twire\t[%d:0]\tw_e4, w_o4;\n", 2*(obits+xtrapbits)-1);
|
fprintf(vmain, "\tqtrstage%s\t#(%d,%d,%d,0,%d,%d)\tstage_e4(i_clk, %s, i_ce,\n",
|
fprintf(vmain, "\tqtrstage%s\t#(%d,%d,%d,0,%d,%d)\tstage_e4(i_clk, %s, i_ce,\n",
|
((dbg)&&(dbgstage==4))?"_dbg":"",
|
((dbg)&&(dbgstage==4))?"_dbg":"",
|
nbits+xtrapbits, obits+xtrapbits, lgsize,
|
nbits+xtrapbits, obits+xtrapbits, lgsize,
|
(inverse)?1:0, (dropbit)?0:0,
|
(inverse)?1:0, (dropbit)?0:0,
|
resetw.c_str());
|
resetw.c_str());
|
fprintf(vmain, "\t\t\t\t\t\tw_s8, w_e8, w_e4, w_s4%s);\n",
|
fprintf(vmain, "\t\t\t\t\t\tw_s8, w_e8, w_e4, w_s4%s);\n",
|
((dbg)&&(dbgstage==4))?", o_dbg":"");
|
((dbg)&&(dbgstage==4))?", o_dbg":"");
|
fprintf(vmain, "\tqtrstage\t#(%d,%d,%d,1,%d,%d)\tstage_o4(i_clk, %s, i_ce,\n",
|
fprintf(vmain, "\tqtrstage\t#(%d,%d,%d,1,%d,%d)\tstage_o4(i_clk, %s, i_ce,\n",
|
nbits+xtrapbits, obits+xtrapbits, lgsize, (inverse)?1:0, (dropbit)?0:0,
|
nbits+xtrapbits, obits+xtrapbits, lgsize, (inverse)?1:0, (dropbit)?0:0,
|
resetw.c_str());
|
resetw.c_str());
|
fprintf(vmain, "\t\t\t\t\t\tw_s8, w_o8, w_o4, w_os4);\n");
|
fprintf(vmain, "\t\t\t\t\t\tw_s8, w_o8, w_o4, w_os4);\n");
|
}
|
}
|
dropbit ^= 1;
|
dropbit ^= 1;
|
nbits = obits;
|
nbits = obits;
|
tmp_size >>= 1; lgtmp--;
|
tmp_size >>= 1; lgtmp--;
|
}
|
}
|
|
|
{
|
{
|
obits = nbits+((dropbit)?0:1);
|
obits = nbits+((dropbit)?0:1);
|
if (obits > nbitsout)
|
if (obits > nbitsout)
|
obits = nbitsout;
|
obits = nbitsout;
|
if ((maxbitsout>0)&&(obits > maxbitsout))
|
if ((maxbitsout>0)&&(obits > maxbitsout))
|
obits = maxbitsout;
|
obits = maxbitsout;
|
fprintf(vmain, "\twire\t\tw_s2;\n");
|
fprintf(vmain, "\twire\t\tw_s2;\n");
|
if (single_clock) {
|
if (single_clock) {
|
fprintf(vmain, "\twire\t[%d:0]\tw_d2;\n",
|
fprintf(vmain, "\twire\t[%d:0]\tw_d2;\n",
|
2*obits-1);
|
2*obits-1);
|
} else {
|
} else {
|
fprintf(vmain, "\twire\t[%d:0]\tw_e2, w_o2;\n",
|
fprintf(vmain, "\twire\t[%d:0]\tw_e2, w_o2;\n",
|
2*obits-1);
|
2*obits-1);
|
}
|
}
|
if ((nbits+xtrapbits+1 == obits)&&(!dropbit))
|
/*
|
printf("WARNING: SCALING OFF BY A FACTOR OF TWO--should\'ve dropped a bit in the last stage.\n");
|
if ((nbits+xtrapbits+1 == obits)&&(!dropbit))
|
|
printf("Warning: Less than optimal scaling\n");
|
|
*/
|
|
|
if (single_clock) {
|
if (single_clock) {
|
fprintf(vmain, "\tlaststage\t#(%d,%d,%d)\tstage_2(i_clk, %s, i_ce,\n",
|
fprintf(vmain, "\tlaststage\t#(%d,%d,%d)\tstage_2(i_clk, %s, i_ce,\n",
|
nbits+xtrapbits, obits,(dropbit)?0:1,
|
nbits+xtrapbits, obits,(dropbit)?0:1,
|
resetw.c_str());
|
resetw.c_str());
|
fprintf(vmain, "\t\t\t\t\tw_s4, w_d4, w_d2, w_s2);\n");
|
fprintf(vmain, "\t\t\t\t\tw_s4, w_d4, w_d2, w_s2);\n");
|
} else {
|
} else {
|
fprintf(vmain, "\tlaststage\t#(%d,%d,%d)\tstage_2(i_clk, %s, i_ce,\n",
|
fprintf(vmain, "\tlaststage\t#(%d,%d,%d)\tstage_2(i_clk, %s, i_ce,\n",
|
nbits+xtrapbits, obits,(dropbit)?0:1,
|
nbits+xtrapbits, obits,(dropbit)?0:1,
|
resetw.c_str());
|
resetw.c_str());
|
fprintf(vmain, "\t\t\t\t\tw_s4, w_e4, w_o4, w_e2, w_o2, w_s2);\n");
|
fprintf(vmain, "\t\t\t\t\tw_s4, w_e4, w_o4, w_e2, w_o2, w_s2);\n");
|
}
|
}
|
|
|
fprintf(vmain, "\n\n");
|
fprintf(vmain, "\n\n");
|
nbits = obits;
|
nbits = obits;
|
}
|
}
|
|
|
fprintf(vmain, "\t// Prepare for a (potential) bit-reverse stage.\n");
|
fprintf(vmain, "\t// Prepare for a (potential) bit-reverse stage.\n");
|
if (single_clock)
|
if (single_clock)
|
fprintf(vmain, "\tassign\tbr_sample= w_d2;\n");
|
fprintf(vmain, "\tassign\tbr_sample= w_d2;\n");
|
else {
|
else {
|
fprintf(vmain, "\tassign\tbr_left = w_e2;\n");
|
fprintf(vmain, "\tassign\tbr_left = w_e2;\n");
|
fprintf(vmain, "\tassign\tbr_right = w_o2;\n");
|
fprintf(vmain, "\tassign\tbr_right = w_o2;\n");
|
}
|
}
|
fprintf(vmain, "\n");
|
fprintf(vmain, "\n");
|
if (bitreverse) {
|
if (bitreverse) {
|
fprintf(vmain, "\twire\tbr_start;\n");
|
fprintf(vmain, "\twire\tbr_start;\n");
|
fprintf(vmain, "\treg\tr_br_started;\n");
|
fprintf(vmain, "\treg\tr_br_started;\n");
|
fprintf(vmain, "\tinitial\tr_br_started = 1\'b0;\n");
|
fprintf(vmain, "\tinitial\tr_br_started = 1\'b0;\n");
|
if (async_reset) {
|
if (async_reset) {
|
fprintf(vmain, "\talways @(posedge i_clk, negedge i_areset_n)\n");
|
fprintf(vmain, "\talways @(posedge i_clk, negedge i_areset_n)\n");
|
fprintf(vmain, "\t\tif (!i_areset_n)\n");
|
fprintf(vmain, "\t\tif (!i_areset_n)\n");
|
} else {
|
} else {
|
fprintf(vmain, "\talways @(posedge i_clk)\n");
|
fprintf(vmain, "\talways @(posedge i_clk)\n");
|
fprintf(vmain, "\t\tif (i_reset)\n");
|
fprintf(vmain, "\t\tif (i_reset)\n");
|
}
|
}
|
fprintf(vmain, "\t\t\tr_br_started <= 1\'b0;\n");
|
fprintf(vmain, "\t\t\tr_br_started <= 1\'b0;\n");
|
fprintf(vmain, "\t\telse if (i_ce)\n");
|
fprintf(vmain, "\t\telse if (i_ce)\n");
|
fprintf(vmain, "\t\t\tr_br_started <= r_br_started || w_s2;\n");
|
fprintf(vmain, "\t\t\tr_br_started <= r_br_started || w_s2;\n");
|
fprintf(vmain, "\tassign\tbr_start = r_br_started || w_s2;\n");
|
fprintf(vmain, "\tassign\tbr_start = r_br_started || w_s2;\n");
|
}
|
}
|
}
|
}
|
|
|
|
|
fprintf(vmain, "\n");
|
fprintf(vmain, "\n");
|
fprintf(vmain, "\t// Now for the bit-reversal stage.\n");
|
fprintf(vmain, "\t// Now for the bit-reversal stage.\n");
|
fprintf(vmain, "\twire\tbr_sync;\n");
|
fprintf(vmain, "\twire\tbr_sync;\n");
|
if (bitreverse) {
|
if (bitreverse) {
|
if (single_clock) {
|
if (single_clock) {
|
fprintf(vmain, "\twire\t[(2*OWIDTH-1):0]\tbr_o_result;\n");
|
fprintf(vmain, "\twire\t[(2*OWIDTH-1):0]\tbr_o_result;\n");
|
fprintf(vmain, "\tbitreverse\t#(%d,%d)\n\t\trevstage(i_clk, %s,\n", lgsize, nbitsout, resetw.c_str());
|
fprintf(vmain, "\tbitreverse\t#(%d,%d)\n\t\trevstage(i_clk, %s,\n", lgsize, nbitsout, resetw.c_str());
|
fprintf(vmain, "\t\t\t(i_ce & br_start), br_sample,\n");
|
fprintf(vmain, "\t\t\t(i_ce & br_start), br_sample,\n");
|
fprintf(vmain, "\t\t\tbr_o_result, br_sync);\n");
|
fprintf(vmain, "\t\t\tbr_o_result, br_sync);\n");
|
} else {
|
} else {
|
fprintf(vmain, "\twire\t[(2*OWIDTH-1):0]\tbr_o_left, br_o_right;\n");
|
fprintf(vmain, "\twire\t[(2*OWIDTH-1):0]\tbr_o_left, br_o_right;\n");
|
fprintf(vmain, "\tbitreverse\t#(%d,%d)\n\t\trevstage(i_clk, %s,\n", lgsize, nbitsout, resetw.c_str());
|
fprintf(vmain, "\tbitreverse\t#(%d,%d)\n\t\trevstage(i_clk, %s,\n", lgsize, nbitsout, resetw.c_str());
|
fprintf(vmain, "\t\t\t(i_ce & br_start), br_left, br_right,\n");
|
fprintf(vmain, "\t\t\t(i_ce & br_start), br_left, br_right,\n");
|
fprintf(vmain, "\t\t\tbr_o_left, br_o_right, br_sync);\n");
|
fprintf(vmain, "\t\t\tbr_o_left, br_o_right, br_sync);\n");
|
}
|
}
|
} else if (single_clock) {
|
} else if (single_clock) {
|
fprintf(vmain, "\tassign\tbr_o_result = br_result;\n");
|
fprintf(vmain, "\tassign\tbr_o_result = br_result;\n");
|
fprintf(vmain, "\tassign\tbr_sync = w_s2;\n");
|
fprintf(vmain, "\tassign\tbr_sync = w_s2;\n");
|
} else {
|
} else {
|
fprintf(vmain, "\tassign\tbr_o_left = br_left;\n");
|
fprintf(vmain, "\tassign\tbr_o_left = br_left;\n");
|
fprintf(vmain, "\tassign\tbr_o_right = br_right;\n");
|
fprintf(vmain, "\tassign\tbr_o_right = br_right;\n");
|
fprintf(vmain, "\tassign\tbr_sync = w_s2;\n");
|
fprintf(vmain, "\tassign\tbr_sync = w_s2;\n");
|
}
|
}
|
|
|
fprintf(vmain,
|
fprintf(vmain,
|
"\n\n"
|
"\n\n"
|
"\t// Last clock: Register our outputs, we\'re done.\n"
|
"\t// Last clock: Register our outputs, we\'re done.\n"
|
"\tinitial\to_sync = 1\'b0;\n");
|
"\tinitial\to_sync = 1\'b0;\n");
|
if (async_reset)
|
if (async_reset)
|
fprintf(vmain,
|
fprintf(vmain,
|
"\talways @(posedge i_clk, negedge i_areset_n)\n\t\tif (!i_areset_n)\n");
|
"\talways @(posedge i_clk, negedge i_areset_n)\n\t\tif (!i_areset_n)\n");
|
else {
|
else {
|
fprintf(vmain,
|
fprintf(vmain,
|
"\talways @(posedge i_clk)\n\t\tif (i_reset)\n");
|
"\talways @(posedge i_clk)\n\t\tif (i_reset)\n");
|
}
|
}
|
|
|
fprintf(vmain,
|
fprintf(vmain,
|
"\t\t\to_sync <= 1\'b0;\n"
|
"\t\t\to_sync <= 1\'b0;\n"
|
"\t\telse if (i_ce)\n"
|
"\t\telse if (i_ce)\n"
|
"\t\t\to_sync <= br_sync;\n"
|
"\t\t\to_sync <= br_sync;\n"
|
"\n"
|
"\n"
|
"\talways @(posedge i_clk)\n"
|
"\talways @(posedge i_clk)\n"
|
"\t\tif (i_ce)\n");
|
"\t\tif (i_ce)\n");
|
if (single_clock) {
|
if (single_clock) {
|
fprintf(vmain, "\t\t\to_result <= br_o_result;\n");
|
fprintf(vmain, "\t\t\to_result <= br_o_result;\n");
|
} else {
|
} else {
|
fprintf(vmain,
|
fprintf(vmain,
|
"\t\tbegin\n"
|
"\t\tbegin\n"
|
"\t\t\to_left <= br_o_left;\n"
|
"\t\t\to_left <= br_o_left;\n"
|
"\t\t\to_right <= br_o_right;\n"
|
"\t\t\to_right <= br_o_right;\n"
|
"\t\tend\n");
|
"\t\tend\n");
|
}
|
}
|
|
|
fprintf(vmain,
|
fprintf(vmain,
|
"\n\n"
|
"\n\n"
|
"endmodule\n");
|
"endmodule\n");
|
fclose(vmain);
|
fclose(vmain);
|
|
|
|
|
{
|
{
|
std::string fname;
|
std::string fname;
|
|
|
fname = coredir + "/butterfly.v";
|
fname = coredir + "/butterfly.v";
|
build_butterfly(fname.c_str(), xtracbits, rounding,
|
build_butterfly(fname.c_str(), xtracbits, rounding,
|
ckpce, async_reset);
|
ckpce, async_reset);
|
|
|
fname = coredir + "/hwbfly.v";
|
fname = coredir + "/hwbfly.v";
|
build_hwbfly(fname.c_str(), xtracbits, rounding,
|
build_hwbfly(fname.c_str(), xtracbits, rounding,
|
ckpce, async_reset);
|
ckpce, async_reset);
|
|
|
{
|
{
|
// To make debugging easier, we build both of these
|
// To make debugging easier, we build both of these
|
fname = coredir + "/shiftaddmpy.v";
|
fname = coredir + "/shiftaddmpy.v";
|
build_multiply(fname.c_str());
|
build_multiply(fname.c_str());
|
|
|
fname = coredir + "/longbimpy.v";
|
fname = coredir + "/longbimpy.v";
|
build_longbimpy(fname.c_str());
|
build_longbimpy(fname.c_str());
|
fname = coredir + "/bimpy.v";
|
fname = coredir + "/bimpy.v";
|
build_bimpy(fname.c_str());
|
build_bimpy(fname.c_str());
|
}
|
}
|
|
|
if ((dbg)&&(dbgstage == 4)) {
|
if ((dbg)&&(dbgstage == 4)) {
|
fname = coredir + "/qtrstage_dbg.v";
|
fname = coredir + "/qtrstage_dbg.v";
|
if (single_clock)
|
if (single_clock)
|
build_snglquarters(fname.c_str(), rounding,
|
build_snglquarters(fname.c_str(), rounding,
|
async_reset, true);
|
async_reset, true);
|
else
|
else
|
build_dblquarters(fname.c_str(), rounding,
|
build_dblquarters(fname.c_str(), rounding,
|
async_reset, true);
|
async_reset, true);
|
}
|
}
|
fname = coredir + "/qtrstage.v";
|
fname = coredir + "/qtrstage.v";
|
if (single_clock)
|
if (single_clock)
|
build_snglquarters(fname.c_str(), rounding,
|
build_snglquarters(fname.c_str(), rounding,
|
async_reset, false);
|
async_reset, false);
|
else
|
else
|
build_dblquarters(fname.c_str(), rounding,
|
build_dblquarters(fname.c_str(), rounding,
|
async_reset, false);
|
async_reset, false);
|
|
|
|
|
if (single_clock) {
|
if (single_clock) {
|
fname = coredir + "/laststage.v";
|
fname = coredir + "/laststage.v";
|
build_sngllast(fname.c_str(), async_reset);
|
build_sngllast(fname.c_str(), async_reset);
|
} else {
|
} else {
|
if ((dbg)&&(dbgstage == 2))
|
if ((dbg)&&(dbgstage == 2))
|
fname = coredir + "/laststage_dbg.v";
|
fname = coredir + "/laststage_dbg.v";
|
else
|
else
|
fname = coredir + "/laststage.v";
|
fname = coredir + "/laststage.v";
|
build_dblstage(fname.c_str(), rounding,
|
build_dblstage(fname.c_str(), rounding,
|
async_reset, (dbg)&&(dbgstage==2));
|
async_reset, (dbg)&&(dbgstage==2));
|
}
|
}
|
|
|
if (bitreverse) {
|
if (bitreverse) {
|
fname = coredir + "/bitreverse.v";
|
fname = coredir + "/bitreverse.v";
|
if (single_clock)
|
if (single_clock)
|
build_snglbrev(fname.c_str(), async_reset);
|
build_snglbrev(fname.c_str(), async_reset);
|
else
|
else
|
build_dblreverse(fname.c_str(), async_reset);
|
build_dblreverse(fname.c_str(), async_reset);
|
}
|
}
|
|
|
const char *rnd_string = "";
|
const char *rnd_string = "";
|
switch(rounding) {
|
switch(rounding) {
|
case RND_TRUNCATE: rnd_string = "/truncate.v"; break;
|
case RND_TRUNCATE: rnd_string = "/truncate.v"; break;
|
case RND_FROMZERO: rnd_string = "/roundfromzero.v"; break;
|
case RND_FROMZERO: rnd_string = "/roundfromzero.v"; break;
|
case RND_HALFUP: rnd_string = "/roundhalfup.v"; break;
|
case RND_HALFUP: rnd_string = "/roundhalfup.v"; break;
|
default:
|
default:
|
rnd_string = "/convround.v"; break;
|
rnd_string = "/convround.v"; break;
|
} fname = coredir + rnd_string;
|
} fname = coredir + rnd_string;
|
switch(rounding) {
|
switch(rounding) {
|
case RND_TRUNCATE: build_truncator(fname.c_str()); break;
|
case RND_TRUNCATE: build_truncator(fname.c_str()); break;
|
case RND_FROMZERO: build_roundfromzero(fname.c_str()); break;
|
case RND_FROMZERO: build_roundfromzero(fname.c_str()); break;
|
case RND_HALFUP: build_roundhalfup(fname.c_str()); break;
|
case RND_HALFUP: build_roundhalfup(fname.c_str()); break;
|
default:
|
default:
|
build_convround(fname.c_str()); break;
|
build_convround(fname.c_str()); break;
|
}
|
}
|
|
|
}
|
}
|
|
|
if (verbose_flag)
|
if (verbose_flag)
|
printf("All done -- success\n");
|
printf("All done -- success\n");
|
}
|
}
|
|
|