BT is painfully slow



  • Im running monte carlo simulations using BT, mainly fine tuning some parameters that are involved in the algo.

    I had an algo that was taking about 1 minute per 1 mo. of historical data. I implemented some scikit learn modules that made it slower.. and is now taking 3 minutes.

    System:

    • Processor Intel(R) Core(TM) i7-4700MQ CPU @ 2.40GHz, 2401 Mhz, 4 Core(s), 8 Logical Processor(s)
    • Installed Physical Memory (RAM) 12.0 GB
    • Hyper-V - VM Monitor Mode Extensions Yes
    • Hyper-V - Second Level Address Translation Extensions Yes
    • Hyper-V - Virtualization Enabled in Firmware Yes
    • Hyper-V - Data Execution Protection Yes
    • Windows 8, running on bare metal (not virtual machine)
    • Python 3.6 on cygwin

    I tried running on a larger VPS with 8+ vCPU cores, but for some reason, I can't get BT to use all the cores. htop shows only 1 in usage even with maxcpus=8 declaration.

    1. What are some "low hanging" optimizations? I know pypy is supposed to be magical, but pypy support of python 3.5 isn't released for windows

    2. What are the obvious optimizations you can see? is running on python2 recommended?

    3. Is there any performance enhancements planned for BT? Mainly, integration with cython, numba, Nuitka, or any of the more recent dev. in python performance?


  • administrators

    @Taewoo-Kim said in BT is painfully slow:

    BT is painfully slow

    Using it is free. Switching to something else no so painful is also fee.

    1. What are some "low hanging" optimizations? I know pypy is supposed to be magical, but pypy support of python 3.5 isn't released for windows

    Windows 8, running on bare metal (not virtual machine)

    Use Linux

    Python 3.6 on cygwin

    Use the Windows version of Python and not one that's running atop a posix-like simulation layer

    @Taewoo-Kim said in BT is painfully slow:

    2 What are the obvious optimizations you can see? is running on python2 recommended?

    Python 2 vs 3 shouldn't make such a difference. Python is slow per se.

    3 Is there any performance enhancements planned for BT? Mainly, integration with cython, numba, Nuitka, or any of the more recent dev. in python performance?

    See Community - EoDevelopment 1.x



  • deleted: <spoke too soon>


Log in to reply
 

Looks like your connection to Backtrader Community was lost, please wait while we try to reconnect.