libsigrok/input/vcd.c

548 lines
14 KiB
C
Raw Normal View History

/*
2013-04-23 20:24:30 +00:00
* This file is part of the libsigrok project.
*
* Copyright (C) 2012 Petteri Aimonen <jpa@sr.mail.kapsi.fi>
*
* This program is free software: you can redistribute it and/or 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 your option) any later version.
*
* This program is distributed in the hope that it will be useful,
* but WITHOUT ANY WARRANTY; without even the implied warranty of
* MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
* GNU General Public License for more details.
*
* You should have received a copy of the GNU General Public License
* along with this program. If not, see <http://www.gnu.org/licenses/>.
*/
/* The VCD input module has the following options:
*
* numprobes: Maximum number of probes to use. The probes are
* detected in the same order as they are listed
* in the $var sections of the VCD file.
*
* skip: Allows skipping until given timestamp in the file.
* This can speed up analyzing of long captures.
*
* Value < 0: Skip until first timestamp listed in
* the file. (default)
*
* Value = 0: Do not skip, instead generate samples
* beginning from timestamp 0.
*
* Value > 0: Start at the given timestamp.
*
* downsample: Divide the samplerate by the given factor.
* This can speed up analyzing of long captures.
*
2012-11-22 21:15:27 +00:00
* compress: Compress idle periods longer than this value.
* This can speed up analyzing of long captures.
* Default 0 = don't compress.
*
* Based on Verilog standard IEEE Std 1364-2001 Version C
*
* Supported features:
* - $var with 'wire' and 'reg' types of scalar variables
* - $timescale definition for samplerate
* - multiple character variable identifiers
*
* Most important unsupported features:
* - vector variables (bit vectors etc.)
* - analog, integer and real number variables
* - $dumpvars initial value declaration
* - $scope namespaces
2013-11-26 21:11:48 +00:00
* - more than 64 probes
*/
#include <stdlib.h>
#include <glib.h>
#include <stdio.h>
2012-11-20 20:24:17 +00:00
#include <string.h>
#include "libsigrok.h"
#include "libsigrok-internal.h"
#define LOG_PREFIX "input/vcd"
#define DEFAULT_NUM_PROBES 8
2013-11-26 21:11:48 +00:00
#define CHUNKSIZE 1024
struct context {
uint64_t samplerate;
int maxprobes;
int probecount;
int downsample;
unsigned compress;
int64_t skip;
GSList *probes;
};
struct probe {
gchar *name;
gchar *identifier;
};
/* Read until specific type of character occurs in file.
* Skip input if dest is NULL.
* Modes:
* 'W' read until whitespace
* 'N' read until non-whitespace, and ungetc() the character
* '$' read until $end
*/
static gboolean read_until(FILE *file, GString *dest, char mode)
{
char prev[4] = "";
long startpos = ftell(file);
2013-11-26 21:11:48 +00:00
for(;;) {
int c = fgetc(file);
2013-11-26 21:11:48 +00:00
if (c == EOF) {
if (mode == '$')
sr_err("Unexpected EOF, read started at %ld.", startpos);
return FALSE;
}
if (mode == 'W' && g_ascii_isspace(c))
return TRUE;
2013-11-26 21:11:48 +00:00
if (mode == 'N' && !g_ascii_isspace(c)) {
ungetc(c, file);
return TRUE;
}
2013-11-26 21:11:48 +00:00
if (mode == '$') {
prev[0] = prev[1]; prev[1] = prev[2]; prev[2] = prev[3]; prev[3] = c;
2013-11-26 21:11:48 +00:00
if (prev[0] == '$' && prev[1] == 'e' && prev[2] == 'n' && prev[3] == 'd') {
if (dest != NULL)
g_string_truncate(dest, dest->len - 3);
return TRUE;
}
}
if (dest != NULL)
g_string_append_c(dest, c);
}
}
2013-11-26 21:11:48 +00:00
/*
* Reads a single VCD section from input file and parses it to structure.
* e.g. $timescale 1ps $end => "timescale" "1ps"
*/
static gboolean parse_section(FILE *file, gchar **name, gchar **contents)
{
gboolean status;
GString *sname, *scontents;
/* Skip any initial white-space */
if (!read_until(file, NULL, 'N')) return FALSE;
/* Section tag should start with $. */
2013-11-26 21:11:48 +00:00
if (fgetc(file) != '$') {
sr_err("Expected $ at beginning of section.");
return FALSE;
}
/* Read the section tag */
sname = g_string_sized_new(32);
status = read_until(file, sname, 'W');
/* Skip whitespace before content */
status = status && read_until(file, NULL, 'N');
/* Read the content */
scontents = g_string_sized_new(128);
status = status && read_until(file, scontents, '$');
g_strchomp(scontents->str);
/* Release strings if status is FALSE, return them if status is TRUE */
*name = g_string_free(sname, !status);
*contents = g_string_free(scontents, !status);
return status;
}
static void free_probe(void *data)
{
struct probe *probe = data;
g_free(probe->name);
g_free(probe->identifier);
g_free(probe);
}
2012-11-20 20:24:17 +00:00
static void release_context(struct context *ctx)
{
g_slist_free_full(ctx->probes, free_probe);
2012-11-20 20:24:17 +00:00
g_free(ctx);
}
/* Remove empty parts from an array returned by g_strsplit. */
static void remove_empty_parts(gchar **parts)
{
gchar **src = parts;
gchar **dest = parts;
2013-11-26 21:11:48 +00:00
while (*src != NULL) {
if (**src != '\0')
*dest++ = *src;
src++;
}
*dest = NULL;
}
2013-11-26 21:11:48 +00:00
/*
* Parse VCD header to get values for context structure.
* The context structure should be zeroed before calling this.
*/
static gboolean parse_header(FILE *file, struct context *ctx)
{
uint64_t p, q;
gchar *name = NULL, *contents = NULL;
gboolean status = FALSE;
struct probe *probe;
2013-11-26 21:11:48 +00:00
while (parse_section(file, &name, &contents)) {
sr_dbg("Section '%s', contents '%s'.", name, contents);
2013-11-26 21:11:48 +00:00
if (g_strcmp0(name, "enddefinitions") == 0) {
status = TRUE;
break;
2013-11-26 21:11:48 +00:00
} else if (g_strcmp0(name, "timescale") == 0) {
/*
* The standard allows for values 1, 10 or 100
* and units s, ms, us, ns, ps and fs.
* */
if (sr_parse_period(contents, &p, &q) == SR_OK) {
ctx->samplerate = q / p;
2013-11-26 21:11:48 +00:00
if (q % p != 0) {
/* Does not happen unless time value is non-standard */
sr_warn("Inexact rounding of samplerate, %" PRIu64 " / %" PRIu64 " to %" PRIu64 " Hz.",
q, p, ctx->samplerate);
}
sr_dbg("Samplerate: %" PRIu64, ctx->samplerate);
2013-11-26 21:11:48 +00:00
} else {
sr_err("Parsing timescale failed.");
}
2013-11-26 21:11:48 +00:00
} else if (g_strcmp0(name, "var") == 0) {
/* Format: $var type size identifier reference $end */
gchar **parts = g_strsplit_set(contents, " \r\n\t", 0);
remove_empty_parts(parts);
if (g_strv_length(parts) != 4)
sr_warn("$var section should have 4 items");
else if (g_strcmp0(parts[0], "reg") != 0 && g_strcmp0(parts[0], "wire") != 0)
sr_info("Unsupported signal type: '%s'", parts[0]);
else if (strtol(parts[1], NULL, 10) != 1)
sr_info("Unsupported signal size: '%s'", parts[1]);
else if (ctx->probecount >= ctx->maxprobes)
sr_warn("Skipping '%s' because only %d probes requested.", parts[3], ctx->maxprobes);
2013-11-26 21:11:48 +00:00
else {
sr_info("Probe %d is '%s' identified by '%s'.", ctx->probecount, parts[3], parts[2]);
probe = g_malloc(sizeof(struct probe));
probe->identifier = g_strdup(parts[2]);
probe->name = g_strdup(parts[3]);
ctx->probes = g_slist_append(ctx->probes, probe);
ctx->probecount++;
}
g_strfreev(parts);
}
g_free(name); name = NULL;
g_free(contents); contents = NULL;
}
g_free(name);
g_free(contents);
return status;
}
static int format_match(const char *filename)
{
FILE *file;
gchar *name = NULL, *contents = NULL;
gboolean status;
file = fopen(filename, "r");
if (file == NULL)
return FALSE;
2013-11-26 21:11:48 +00:00
/*
* If we can parse the first section correctly,
* then it is assumed to be a VCD file.
*/
status = parse_section(file, &name, &contents);
status = status && (*name != '\0');
g_free(name);
g_free(contents);
fclose(file);
return status;
}
static int init(struct sr_input *in, const char *filename)
{
struct sr_probe *probe;
int num_probes, i;
char name[SR_MAX_PROBENAME_LEN + 1];
char *param;
struct context *ctx;
(void)filename;
if (!(ctx = g_try_malloc0(sizeof(*ctx)))) {
sr_err("Input format context malloc failed.");
return SR_ERR_MALLOC;
}
num_probes = DEFAULT_NUM_PROBES;
ctx->samplerate = 0;
ctx->downsample = 1;
ctx->skip = -1;
if (in->param) {
param = g_hash_table_lookup(in->param, "numprobes");
if (param) {
num_probes = strtoul(param, NULL, 10);
2013-11-26 21:11:48 +00:00
if (num_probes < 1) {
2012-11-20 20:24:17 +00:00
release_context(ctx);
return SR_ERR;
} else if (num_probes > 64) {
sr_err("No more than 64 probes supported.");
return SR_ERR;
2012-11-20 20:24:17 +00:00
}
}
param = g_hash_table_lookup(in->param, "downsample");
if (param) {
ctx->downsample = strtoul(param, NULL, 10);
if (ctx->downsample < 1)
ctx->downsample = 1;
}
2012-11-22 21:15:27 +00:00
param = g_hash_table_lookup(in->param, "compress");
2013-11-26 21:11:48 +00:00
if (param)
2012-11-22 21:15:27 +00:00
ctx->compress = strtoul(param, NULL, 10);
param = g_hash_table_lookup(in->param, "skip");
2013-11-26 21:11:48 +00:00
if (param)
ctx->skip = strtoul(param, NULL, 10) / ctx->downsample;
}
/* Maximum number of probes to parse from the VCD */
ctx->maxprobes = num_probes;
/* Create a virtual device. */
in->sdi = sr_dev_inst_new(0, SR_ST_ACTIVE, NULL, NULL, NULL);
in->internal = ctx;
for (i = 0; i < num_probes; i++) {
snprintf(name, SR_MAX_PROBENAME_LEN, "%d", i);
2013-11-26 21:11:48 +00:00
if (!(probe = sr_probe_new(i, SR_PROBE_LOGIC, TRUE, name))) {
2012-11-20 20:24:17 +00:00
release_context(ctx);
return SR_ERR;
2012-11-20 20:24:17 +00:00
}
in->sdi->probes = g_slist_append(in->sdi->probes, probe);
}
return SR_OK;
}
2012-11-20 20:24:17 +00:00
/* Send N samples of the given value. */
static void send_samples(const struct sr_dev_inst *sdi, uint64_t sample, uint64_t count)
{
struct sr_datafeed_packet packet;
struct sr_datafeed_logic logic;
uint64_t buffer[CHUNKSIZE];
uint64_t i;
unsigned chunksize = CHUNKSIZE;
2012-11-20 20:24:17 +00:00
if (count < chunksize)
chunksize = count;
for (i = 0; i < chunksize; i++)
buffer[i] = sample;
packet.type = SR_DF_LOGIC;
packet.payload = &logic;
logic.unitsize = sizeof(uint64_t);
logic.data = buffer;
2013-11-26 21:11:48 +00:00
while (count) {
2012-11-20 20:24:17 +00:00
if (count < chunksize)
chunksize = count;
logic.length = sizeof(uint64_t) * chunksize;
sr_session_send(sdi, &packet);
count -= chunksize;
}
}
/* Parse the data section of VCD */
static void parse_contents(FILE *file, const struct sr_dev_inst *sdi, struct context *ctx)
{
GString *token = g_string_sized_new(32);
uint64_t prev_timestamp = 0;
uint64_t prev_values = 0;
/* Read one space-delimited token at a time. */
2013-11-26 21:11:48 +00:00
while (read_until(file, NULL, 'N') && read_until(file, token, 'W')) {
if (token->str[0] == '#' && g_ascii_isdigit(token->str[1])) {
2012-11-20 20:24:17 +00:00
/* Numeric value beginning with # is a new timestamp value */
uint64_t timestamp;
timestamp = strtoull(token->str + 1, NULL, 10);
if (ctx->downsample > 1)
timestamp /= ctx->downsample;
2013-11-26 21:11:48 +00:00
/*
* Skip < 0 => skip until first timestamp.
* Skip = 0 => don't skip
* Skip > 0 => skip until timestamp >= skip.
*/
2013-11-26 21:11:48 +00:00
if (ctx->skip < 0) {
ctx->skip = timestamp;
prev_timestamp = timestamp;
2013-11-26 21:11:48 +00:00
} else if (ctx->skip > 0 && timestamp < (uint64_t)ctx->skip) {
prev_timestamp = ctx->skip;
}
2013-11-26 21:11:48 +00:00
else if (timestamp == prev_timestamp) {
/* Ignore repeated timestamps (e.g. sigrok outputs these) */
2012-11-20 20:24:17 +00:00
}
2013-11-26 21:11:48 +00:00
else {
2012-11-22 21:15:27 +00:00
if (ctx->compress != 0 && timestamp - prev_timestamp > ctx->compress)
{
/* Compress long idle periods */
prev_timestamp = timestamp - ctx->compress;
}
2012-11-20 20:24:17 +00:00
sr_dbg("New timestamp: %" PRIu64, timestamp);
/* Generate samples from prev_timestamp up to timestamp - 1. */
send_samples(sdi, prev_values, timestamp - prev_timestamp);
prev_timestamp = timestamp;
2012-11-20 20:24:17 +00:00
}
2013-11-26 21:11:48 +00:00
} else if (token->str[0] == '$' && token->len > 1) {
2012-11-20 20:24:17 +00:00
/* This is probably a $dumpvars, $comment or similar.
* $dump* contain useful data, but other tags will be skipped until $end. */
2013-11-26 21:11:48 +00:00
if (g_strcmp0(token->str, "$dumpvars") == 0
|| g_strcmp0(token->str, "$dumpon") == 0
|| g_strcmp0(token->str, "$dumpoff") == 0
|| g_strcmp0(token->str, "$end") == 0) {
/* Ignore, parse contents as normally. */
2013-11-26 21:11:48 +00:00
} else {
/* Skip until $end */
read_until(file, NULL, '$');
}
}
2013-11-26 21:11:48 +00:00
else if (strchr("bBrR", token->str[0]) != NULL) {
/* A vector value. Skip it and also the following identifier. */
read_until(file, NULL, 'N');
read_until(file, NULL, 'W');
2013-11-26 21:11:48 +00:00
} else if (strchr("01xXzZ", token->str[0]) != NULL) {
2012-11-20 20:24:17 +00:00
/* A new 1-bit sample value */
int i, bit;
GSList *l;
struct probe *probe;
2012-11-20 20:24:17 +00:00
bit = (token->str[0] == '1');
g_string_erase(token, 0, 1);
2013-11-26 21:11:48 +00:00
if (token->len == 0) {
2012-11-20 20:24:17 +00:00
/* There was a space between value and identifier.
* Read in the rest.
*/
read_until(file, NULL, 'N');
read_until(file, token, 'W');
}
2013-11-26 21:11:48 +00:00
for (i = 0, l = ctx->probes; i < ctx->probecount && l; i++, l = l->next) {
probe = l->data;
2013-11-26 21:11:48 +00:00
if (g_strcmp0(token->str, probe->identifier) == 0) {
2012-11-20 20:24:17 +00:00
sr_dbg("Probe %d new value %d.", i, bit);
/* Found our probe */
if (bit)
prev_values |= (1 << i);
2012-11-20 20:24:17 +00:00
else
prev_values &= ~(1 << i);
2012-11-20 20:24:17 +00:00
break;
}
}
if (i == ctx->probecount)
sr_dbg("Did not find probe for identifier '%s'.", token->str);
2013-11-26 21:11:48 +00:00
} else {
sr_warn("Skipping unknown token '%s'.", token->str);
}
2012-11-20 20:24:17 +00:00
g_string_truncate(token, 0);
}
g_string_free(token, TRUE);
}
static int loadfile(struct sr_input *in, const char *filename)
{
struct sr_datafeed_packet packet;
struct sr_datafeed_meta meta;
struct sr_config *src;
FILE *file;
struct context *ctx;
uint64_t samplerate;
ctx = in->internal;
if ((file = fopen(filename, "r")) == NULL)
return SR_ERR;
2013-11-26 21:11:48 +00:00
if (!parse_header(file, ctx)) {
sr_err("VCD parsing failed");
fclose(file);
return SR_ERR;
}
/* Send header packet to the session bus. */
std_session_send_df_header(in->sdi, LOG_PREFIX);
/* Send metadata about the SR_DF_LOGIC packets to come. */
packet.type = SR_DF_META;
packet.payload = &meta;
samplerate = ctx->samplerate / ctx->downsample;
src = sr_config_new(SR_CONF_SAMPLERATE, g_variant_new_uint64(samplerate));
meta.config = g_slist_append(NULL, src);
sr_session_send(in->sdi, &packet);
sr_config_free(src);
2012-11-20 20:24:17 +00:00
/* Parse the contents of the VCD file */
parse_contents(file, in->sdi, ctx);
/* Send end packet to the session bus. */
packet.type = SR_DF_END;
sr_session_send(in->sdi, &packet);
2012-11-20 20:24:17 +00:00
fclose(file);
release_context(ctx);
in->internal = NULL;
return SR_OK;
}
SR_PRIV struct sr_input_format input_vcd = {
.id = "vcd",
.description = "Value Change Dump",
.format_match = format_match,
.init = init,
.loadfile = loadfile,
};