Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Performance regression #1378

Open
jsanz opened this issue Jun 28, 2019 · 4 comments
Open

Performance regression #1378

jsanz opened this issue Jun 28, 2019 · 4 comments
Assignees
Labels
type: performance Performance problem to measure & solve

Comments

@jsanz
Copy link
Contributor

jsanz commented Jun 28, 2019

Context

I'm experiencing performance problems on a generic VL example.

Steps to Reproduce

This happens only in certain set ups but it's definitely not a corner case, I've seen this happening both in our computers but also in final clients.

  1. Navigate to this example from our developer center.
  2. Without changing anything from the UI, zoom in using the scroll wheel
  3. Check your box RAM consumption

Current Result

This example performance is degraded greatly, to the point of blocking the browser and requiring a full restart. This is not happening in newer computers.

Browser and version

Happening at least on this set-up:

  • Intel(R) Core(TM) i5-5200U CPU @ 2.20GHz
  • 8GB RAM
  • Chrome: 75.0.3770.90 (Build oficial) (64 bits)

Additional info

Please add any information of interest here below

@jsanz jsanz added the type: performance Performance problem to measure & solve label Jun 28, 2019
@jsanz
Copy link
Contributor Author

jsanz commented Jun 28, 2019

I will share my test on an old Windows 10 computer. I tried the example and it worked fine, the visualization was responsive and quick. Still, the browser RAM consumption spiked around 800MB. Sharing below a few screenshots.

screenshots

Sorry for the Spanish labels

System details

00-system png

Baseline

System freshly started and just with the browser

01-base png

Example load

Loading the VL example from the dev center, without any interaction.

03-carga-pagina png

After scrolling

Zooming in and out and panning for a few seconds

04-scrolling png

@jsanz
Copy link
Contributor Author

jsanz commented Jun 28, 2019

On the other hand Nacho Brodin (Prodevelop) reported having issues with this set up

  • Intel(R) Core(TM) i5-5200U CPU @ 2.20GHz
  • 8GB RAM
  • Chrome: 75.0.3770.90 (Build oficial) (64 bits)

@joseahr
Copy link

joseahr commented Jun 28, 2019

I tested the example in my laptop:

Captura5

My GPU is NVIDIA Geforce GTX 1650 4GB

In my test:

  • Opening the browser, the RAM consumption is about 200 MB.
  • Opening the carto vl example, the RAM consumption is up to 500 MB.
  • After zooming for few seconds, the RAM consumption increases up to 2GB and about 50% of the GPU while zooming.
  • Changing the number of buckets, opacity or palette does not seem to be a problem.
Screenshots

No page opened

Captura4

Opened the example page

Captura

After a few seconds of zooming

Captura2

After a few minutes of interacting with the map

Captura3

GPU consumption while zooming

carto vl

@VictorVelarde
Copy link
Contributor

I have tried in my machine with https://carto.com/developers/carto-vl/examples/maps/advanced/classification-variables.html

  • MacBook Pro 2016
  • 2 GHz Intel Core i5
  • 16 GB 1867 MHz LPDDR3
  • Intel Iris Graphics 540 1536 MB

and I get these figures (with Chrome > More tools > Task manager > Memory use):

  • Cold start

Total: 284.4 MB (GPU process: 101 MB | Browser: 165 MB | Tab: 18.4 MB)

  • After opening example:

Total: 494 MB (GPU process: 198 MB | Browser: 167 MB | Tab: 129 MB)

  • After zooming constantly for 1 minute:

Total: 1218 MB (GPU process: 645 MB | Browser: 179 MB | Tab: 394 MB)

  • Another 5 min session with fast interaction:

Total: 1273 MB (GPU process: 704 | Browser: 178 MB | Tab: 391 MB)

So despite the memory grows, it looks pretty stable from a certain point.

While we keep analyzing this, it would be great to gather more info from your side:

  1. Can you verify in a problematic machine if this is a regression from... (some tests ahead to discard several cases; please go to https://vlmemory-fxeovxmova.now.sh and then use each different file)...

    • Mapbox v1.0.0? --> please try with '01.html' (it uses same VL 1.3.1 but Mapbox v0.54.0)
    • VL v1.3.1? --> please try with '02.html' (it uses VL v1.3.0)
    • VL v1.3? --> please try with '03.html' (it uses VL v1.2.6).
    • If at this point you didn't find any difference, then this is probably not a regression, but a different situation, and we have to keep researching.
  2. Can you check if your situation is related to Interactivity?

    • Please try with '04.html' (same latest libs, but without Interactivity).
  3. Can you reproduce this problem with a points layer (not a polygon one): for example, https://carto.com/developers/carto-vl/examples/maps/advanced/taxi-pickups.html

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Labels
type: performance Performance problem to measure & solve
Projects
None yet
Development

No branches or pull requests

3 participants