Update

[daviddias]

Finding some issues while updating the deps:

• fix missing code in js-multihash
• update test to not assume that we implement all the hash functions
• Figure out why there are tests still failing

[daviddias]

@Beanow I followed #187 (comment) and started updating the deps.

I'm being unsuccessful getting unixfs-engine tests to pass with pull-block 1.2.1. With 1.2.0 they pass just fine. Did you have the chance to try running them in your machine?

//cc @dignifiedquire @pgte

[Beanow]

@diasdavid nice find. It's looking like a performance regression for a test-case that I hadn't considered in previous tests. I was able to reproduce and quantify this with the other benchmarking tools I've set up.

In chromium:

Test	Version	Min	AVG	Max
perByteSmall	v1.2.1	4ms	5ms	21ms
perByteSmall	v1.2.0	1ms	2ms	4ms
perByteMedium	v1.2.1	67ms	69ms	127ms
perByteMedium	v1.2.0	23ms	24ms	25ms

Essentially in test/test-fixed-size-chunker.js
The testing method used is

pull(
  pull.values(rawFile),
  chunker(KiB256),
  //...
)

This usage of pull.values(rawFile) feeds the chunker one byte at a time as opposed to more sizable buffers. This will cause it to do 262144 receptions of a single byte buffer and 262144 merges of those buffers per block. Which contains more logic in the later version.

I'll have a look at this scenario and you can look forward to pull-block 1.2.2 soon 🤓

Note that the tests fail due to an 80 second timeout. So the output is still correct, just very slow with these particular tests. I'll let you decide if you want to tweak parameters and release 1.2.1 or wait for a fix.

[dignifiedquire]

we should check and make sure we don’t do this anywhere in the actual code, as that would mean a large inefficency even when we optimize pull-block more

…

On 21. Oct 2017, 16:22 +0200, Robin van Boven ***@***.***>, wrote: @diasdavid nice find. It's looking like a performance regression for a test-case that I hadn't considered in previous tests. I was able to reproduce and quantify this with the other benchmarking tools I've set up. Test Version Min AVG Max perByteSmall v1.2.1 12ms 14ms 37ms perByteSmall v1.2.0 3ms 4ms 10ms perByteMedium v1.2.1 230ms 246ms 476ms perByteMedium v1.2.0 42ms 44ms 47ms Essentially in test/test-fixed-size-chunker.js The testing method used is pull( pull.values(rawFile), chunker(KiB256), //... ) This usage of pull.values(rawFile) feeds the chunker one byte at a time as opposed to more sizable buffers. This will cause it to do 262144 receptions of a single byte buffer and 262144 merges of those buffers per block. Which contains more logic in the later version. I'll have a look at this scenario and you can look forward to pull-block 1.2.2 soon 🤓 Note that the tests fail due to an 80 second timeout. So the output is still correct, just very slow with these particular tests. I'll let you decide if you want to tweak parameters and release 1.2.1 or wait for a fix. — You are receiving this because you were mentioned. Reply to this email directly, view it on GitHub, or mute the thread.

[daviddias]

I'm locking pull-block to 1.2.0 to unblock other updates. Once pull-block 1.2.2 is finished, I'll updated it right away :)