* @todo create a sample example and unit tests
* @todo set limit on buffer capacity, frequency
* @todo support device/trigger number as input
- * @todo support for trigger frequency
* @todo support specific channels as input
- * @todo support multiple buffers
- * @todo support blocksize for multiple samples per buffer
* @todo support buffer timestamps based on IIO channel timestamp
+ * @todo support one-shot mode
*
*
* This is the plugin to capture data from sensors
#define DEFAULT_MERGE_CHANNELS TRUE
/**
- * blocksize for buffer
- */
-#define BLOCKSIZE 1
-
-/**
* @brief IIO devices/triggers
*/
#define DEVICE "device"
* @note they should have matching type and shape to form 1 other/tensors
* @note extra dimension should be available for other/tensors
* @note order of merge is stable
+ * @note merging into 1 tensor only supported using innermost dimension.
* @param[in/out] info Tensor info to be merged
* @param[in] size Info array size
+ * @param[in] dir Innermost/outermost/innermost-outer (0/1/2) available dimension
* @returns >=0 number of valid entries in the info after merge
* -1 failed due to missing extra dimension/mismatch shape/type
- * @todo merge tensors (in stable order) into buckets based on type
*/
static gint
-gst_tensor_src_merge_tensor_by_type (GstTensorInfo * info, guint size)
+gst_tensor_src_merge_tensor_by_type (GstTensorInfo * info, guint size,
+ guint dir)
{
gint info_idx, base_idx, dim_idx;
- gboolean mismatch = FALSE, dim_avail = TRUE;
+ gboolean mismatch = FALSE, dim_avail = FALSE;
+ gint merge_dim = -1;
/** base error control check */
g_return_val_if_fail (size > 0, 0);
base_idx = 0;
- /** verify extra dimension */
- if (info[base_idx].dimension[NNS_TENSOR_RANK_LIMIT - 1] != 1) {
- dim_avail = FALSE;
+ /** verify extra dimension (innermost to outermost) */
+ for (dim_idx = 0; dim_idx < NNS_TENSOR_RANK_LIMIT; dim_idx++) {
+ if (info[base_idx].dimension[dim_idx] == 1) {
+ dim_avail = TRUE;
+ }
}
/** verify that all the types and shapes match */
}
}
- /** Now merge into 1 tensor using the outermost dimension*/
- for (dim_idx = NNS_TENSOR_RANK_LIMIT - 1; dim_idx > 0; dim_idx--) {
- if (info[base_idx].dimension[dim_idx] != 1) {
- dim_idx += 1;
- break;
+ /**
+ * If there are multiple available dimensions to merge along, we use dir
+ * to choose which the dimension to merge. If there is just 1 dimension,
+ * dir variable has no effect
+ */
+ if (dir == 0) {
+ for (dim_idx = 0; dim_idx < NNS_TENSOR_RANK_LIMIT; dim_idx++) {
+ if (info[base_idx].dimension[dim_idx] == 1) {
+ merge_dim = dim_idx;
+ break;
+ }
+ }
+ } else if (dir == 1) {
+ for (dim_idx = NNS_TENSOR_RANK_LIMIT - 1; dim_idx >= 0; dim_idx--) {
+ if (info[base_idx].dimension[dim_idx] == 1) {
+ merge_dim = dim_idx;
+ break;
+ }
}
+ } else if (dir == 2) {
+ for (dim_idx = NNS_TENSOR_RANK_LIMIT - 1; dim_idx >= 0; dim_idx--) {
+ if (info[base_idx].dimension[dim_idx] != 1) {
+ merge_dim = dim_idx + 1;
+ break;
+ }
+ /** No outer dimension available to merge */
+ if (merge_dim > NNS_TENSOR_RANK_LIMIT) {
+ return size;
+ }
+ }
+ } else {
+ return -1;
}
- info[0].dimension[dim_idx] = size;
+ /** Now merge into 1 tensor using the selected dimension*/
+ info[0].dimension[merge_dim] = size;
return 1;
}
for (dim_idx = 0; dim_idx < NNS_TENSOR_RANK_LIMIT; dim_idx++) {
info[info_idx].dimension[dim_idx] = 1;
}
+ info[info_idx].dimension[1] = tensor_src_iio->buffer_capacity;
info_idx += 1;
}
g_assert_cmpint (info_idx, ==, tensor_src_iio->num_channels_enabled);
if (tensor_src_iio->merge_channels_data) {
tensor_info_merged_size =
gst_tensor_src_merge_tensor_by_type (info,
- tensor_src_iio->num_channels_enabled);
+ tensor_src_iio->num_channels_enabled, 0);
}
/** verify the merging of the array */
for (info_idx = 0; info_idx < tensor_info_merged_size; info_idx++) {
gst_tensor_info_copy (&config->info.info[info_idx], &info[info_idx]);
}
+
+ /**
+ * buffer_capacity number of data samples are captured at once, packed
+ * together and sent downstream
+ */
config->rate_n = tensor_src_iio->sampling_frequency;
- config->rate_d = BLOCKSIZE;
+ config->rate_d = tensor_src_iio->buffer_capacity;
config->info.num_tensors = tensor_info_merged_size;
tensor_src_iio->tensors_config = config;
gchar *filename = NULL;
gint num_channels_enabled;
+ /** no support one shot mode for now */
+ if (!g_strcmp0 (self->mode, MODE_ONE_SHOT)) {
+ GST_ERROR_OBJECT (self, "One-shot mode not yet supported.");
+ goto error_return;
+ }
+
/** Find the device */
id = gst_tensor_src_iio_get_id_by_name (IIO_BASE_DIR, self->device.name,
DEVICE_PREFIX);
self->device.base_dir = g_build_filename (IIO_BASE_DIR, dirname, NULL);
g_free (dirname);
- /**
- * @todo: support scale/offset in one-shot mode for
- * shared/non-shared channels
- */
- /** no more configuration for one shot mode */
- if (!g_strcmp0 (self->mode, MODE_ONE_SHOT)) {
- goto safe_return;
- }
/** register the trigger */
if (self->trigger.name != NULL) {
/** verify if trigger is supported by our device */
}
g_free (filename);
}
+
/** setup the frequency (only verifying the frequency now) */
/** @todo verify setting up frequency */
sampling_frequency =
self->buffer_data_fp->events = POLLIN;
g_free (filename);
-safe_return:
self->configured = TRUE;
/** bytes every buffer will be fixed */
gst_base_src_set_dynamic_size (src, FALSE);
{
GstTensorSrcIIO *self;
gint status, bytes_to_read;
- guint idx_x, idx_y;
+ guint idx;
gchar *raw_data_base, *raw_data;
gfloat *map_data_float;
GstMemory *mem;
}
} else {
/** sleep for a device tick */
- g_usleep (MAX (1, (guint64) 1000000 / self->sampling_frequency));
+ g_usleep (MAX (1,
+ (guint64) 1000000 / (self->sampling_frequency /
+ self->buffer_capacity)));
}
/** read the data from file */
/**
* current assumption is that the all data is float and merged to form
* a 1 dimension data. 2nd dimension comes from buffer capacity.
- * @todo introduce 3rd dimension from blocksize
*/
- for (idx_x = 0; idx_x < BLOCKSIZE; idx_x++) {
- for (idx_y = 0; idx_y < self->buffer_capacity; idx_y++) {
- map_data_float =
- ((gfloat *) map.data) +
- idx_x * self->buffer_capacity * self->num_channels_enabled +
- idx_y * self->num_channels_enabled;
- if (!gst_tensor_src_iio_process_scanned_data (self->channels, raw_data,
- map_data_float)) {
- GST_ERROR_OBJECT (self, "Error while processing scanned data.");
- goto error_data_free;
- }
- raw_data += self->scan_size;
+ for (idx = 0; idx < self->buffer_capacity; idx++) {
+ map_data_float = ((gfloat *) map.data) + idx * self->num_channels_enabled;
+ if (!gst_tensor_src_iio_process_scanned_data (self->channels, raw_data,
+ map_data_float)) {
+ GST_ERROR_OBJECT (self, "Error while processing scanned data.");
+ goto error_data_free;
}
+ raw_data += self->scan_size;
}
/** wrap up the buffer */
projects / platform / upstream / nnstreamer.git / commitdiff