Hacker News Reader

Comment by Robin_Message | Hacker News Reader

I can't see why my instruction decoder is failing: https://imgur.com/a/PCgkQhb

Otherwise, it's fun, although the node connections are fiddly! Maybe allow dragging from either end? Or snapping together.

Edit to add: and on the next level, the m flag doesn't exist. So, will stop there, but fun!

View on HN

You need to special case "target a", it should be 1 in case bit 15 is 0. The error reporting could definitely be improved. It should probably display the exact mismatch, something like "with input XXX connector Y was 0 but was expected to be 1". Otherwise it gets to tedious to debug.

The m issue is a bug in the spec, thanks for discovering it.

Thank you for the feedback, it is much appreciated.

As for debugging, why not just put checkboxes in the input-output lookup table(s) shown on the left. And highlight the currently input line + mark wrong outputs. I think that would be intuitive.

To further build on this; If there is an intermediate state, show its pins in the diagram like output and input. If the table becomes wide, perhaps show it on the bottom. Perhaps let users add a "debug probe" from the toolbox that becomes a column in that table.

I've looked through the code a bit, I think your assembler is actually buggy. The expected results for the tests are correct, however the encoded instructions that you're feeding to the decoder are not. For example, A=1 should be 0x89F8, but you're encoding it as 0x8FE0. This causes the tests to fail despite the logic being correct.

Thank you very much for bringing this to my attention!

Actually there was a bug both in the text and in the code. A=1 should be 0x8FE0, but this was not specified correctly. I have fixed it now (I hope).

Thanks! I think this could be improved further by changing the names of the destination and source flags (all four of them) to make things clearer.

While I'm at it, here's another bug: the last test of the program engine feeds a 5 into j, which breaks things because the counter doesn't expect it at all. Also the final level doesn't have any tests but I'm sure you know that.

I want to add that besides some issues that have already been pointed out, it's a really nicely made tutorial. I went through all of it pretty fast because I already have a good background, but I would definitely point a beginner to it as a learning resource.

Thank you for finding the bugs! It shouldn't even be possible feed a 5 to a one-bit connector.

Yes I think I shot myself in the foot by designing the connectors so the labels could only be one or two letters. There is no technical reason they couldn't have descriptive labels like "clock" rather than "c". But it would probably require to turn the design "on the side", so the connectors are left-right rather than top-bottom.

I know I miss tests for a few of the levels.

Thanks for the nice words, it is encouraging!

Reply on news.ycombinator.com