From: Gerhard Sittig <redacted>
Date: Sat, 21 Dec 2019 11:31:56 +0000 (+0100)
Subject: output/wavedrom: rephrase accumulation of output data
X-Git-Url: https://sigrok.org/gitweb/?p=libsigrok.git;a=commitdiff_plain;h=38950645429a6220bdca129f0ff9b72d4ba6254b

output/wavedrom: rephrase accumulation of output data

Iterate over the received sample set first, before iterating over the
respective sample's number of channels. This avoids redundant extraction
of sampled bits (which saves only little), but also increases locality
of processed data (though string accumulation still may be expensive).

It also adds the future option of RLE compression during accumulation of
output data, which perfectly matches the WaveDrom syntax for repeated
bit patterns.
---

diff --git a/src/output/wavedrom.c b/src/output/wavedrom.c
index 12e506d2..65324bdc 100644
--- a/src/output/wavedrom.c
+++ b/src/output/wavedrom.c
@@ -73,24 +73,47 @@ static void process_logic(const struct context *ctx,
 	const struct sr_datafeed_logic *logic)
 {
 	size_t sample_count, ch, i;
-	uint8_t *sample;
+	uint8_t *sample, bit;
+	GString *accu;
 
-	sample_count = logic->length / logic->unitsize;
+	if (!ctx->channel_count)
+		return;
 
 	/*
 	 * Extract the logic bits for each channel and store them
 	 * as wavedrom letters (1/0) in each channel's text string.
+	 * This transforms the input which consists of sample sets
+	 * that span multiple channels into output stripes per logic
+	 * channel which consist of bits for that individual channel.
+	 *
+	 * TODO Reduce memory consumption during accumulation of
+	 * output data.
+	 *
+	 * Ideally we'd accumulate binary chunks, and defer conversion
+	 * to the text format. Analog data already won't get here, only
+	 * logic data does. When the per-channel transformation also
+	 * gets deferred until later, then the only overhead would be
+	 * for disabled logic channels. Which may be acceptable or even
+	 * negligable.
+	 *
+	 * An optional addition to the above mentioned accumulation of
+	 * binary data is RLE compression. Mark both the position in the
+	 * accumulated data as well as a repetition counter, instead of
+	 * repeatedly storing the same sample set. The efficiency of
+	 * this approach of course depends on the change rate of input
+	 * data. But the approach perfectly matches the WaveDrom syntax
+	 * for repeated bit patterns, and thus is easily handled in the
+	 * text rendering stage of the output module.
 	 */
-	for (ch = 0; ch < ctx->channel_count; ch++) {
-		if (ctx->channels[ch]) {
-			for (i = 0; i < sample_count; i++) {
-				sample = logic->data + i * logic->unitsize;
-
-				if (ctx->channel_outputs[ch]) {
-					g_string_append_c(ctx->channel_outputs[ch],
-						sample[ch / 8] & (1 << (ch % 8)) ? '1' : '0');
-				}
-			}
+	sample_count = logic->length / logic->unitsize;
+	for (i = 0; i < sample_count; i++) {
+		sample = logic->data + i * logic->unitsize;
+		for (ch = 0; ch < ctx->channel_count; ch++) {
+			accu = ctx->channel_outputs[ch];
+			if (!accu)
+				continue;
+			bit = sample[ch / 8] & (1 << (ch % 8));
+			g_string_append_c(accu, bit ? '1' : '0');
 		}
 	}
 }