Run speed tests with newest libsbml

[luciansmith]

The version of libsbml in SVN right now, and which will be part of the next release, changes how ASTNodes are handled in such a way that it might slow things down in roadrunner. The developers are interested to find out if this is the case or not, and if so, in what contexts, so that speedups can be looked for. If we can do some speed tests with roadrunner comparing libsbml v5.17.0 (the current release) with v5.17.1 (the version in SVN right now), the developers will have a chance to fix any slowdowns we find before the next official release.

[hsauro]

I presume this would only slow down the loading time of the sbml file? At runtime we generate our own code.

Herbert

On Tue, Sep 25, 2018 at 8:17 AM Lucian Smith notifications@github.com wrote:

The version of libsbml in SVN right now, and which will be part of the next release, changes how ASTNodes are handled in such a way that it might slow things down in roadrunner. The developers are interested to find out if this is the case or not, and if so, in what contexts, so that speedups can be looked for. If we can do some speed tests with roadrunner comparing libsbml v5.17.0 (the current release) with v5.17.1 (the version in SVN right now), the developers will have a chance to fix any slowdowns we find before the next official release.

— You are receiving this because you are subscribed to this thread. Reply to this email directly, view it on GitHub https://github.com/sys-bio/tellurium/issues/408, or mute the thread https://github.com/notifications/unsubscribe-auth/ABAZDpwpuYyU1Jtxb0W3RD2vn6FGINx2ks5uekikgaJpZM4W42qu .

-- Herbert Sauro, Associate Professor University of Washington, Bioengineering 206-685-2119, www.sys-bio.org hsauro@uw.edu Books: http://books.analogmachine.org/

[luciansmith]

Andy was very adamant back in the day that slowing down the 'iterating over ASTNodes' would slow down roadrunner. I assumed from that, that the actual calculations were carried out by iterating over ASTNodes. But maybe not! Maybe it really is just a loading SBML thing?

[hsauro]

PS This is Andrew Hu, Not Andy S

On Thu, Sep 27, 2018 at 11:25 AM Herbert M Sauro hsauro@uw.edu wrote:

Andrew will know more but we get the speed by not using ASTNodes, we actually compile the rate laws to machine code

Herbert

On Wed, Sep 26, 2018 at 9:19 AM Lucian Smith notifications@github.com wrote:

Andy was very adamant back in the day that slowing down the 'iterating over ASTNodes' would slow down roadrunner. I assumed from that, that the actual calculations were carried out by iterating over ASTNodes. But maybe not! Maybe it really is just a loading SBML thing?

— You are receiving this because you commented. Reply to this email directly, view it on GitHub https://github.com/sys-bio/tellurium/issues/408#issuecomment-424775198, or mute the thread https://github.com/notifications/unsubscribe-auth/ABAZDtPuDM6dIvwi8ciAT2m0WoRqr3SYks5ue6hxgaJpZM4W42qu .

-- Herbert Sauro, Associate Professor University of Washington, Bioengineering 206-685-2119, www.sys-bio.org hsauro@uw.edu Books: http://books.analogmachine.org/

-- Herbert Sauro, Associate Professor University of Washington, Bioengineering 206-685-2119, www.sys-bio.org hsauro@uw.edu Books: http://books.analogmachine.org/

[hsauro]

Andrew will know more but we get the speed by not using ASTNodes, we actually compile the rate laws to machine code

Herbert

On Wed, Sep 26, 2018 at 9:19 AM Lucian Smith notifications@github.com wrote:

Andy was very adamant back in the day that slowing down the 'iterating over ASTNodes' would slow down roadrunner. I assumed from that, that the actual calculations were carried out by iterating over ASTNodes. But maybe not! Maybe it really is just a loading SBML thing?

— You are receiving this because you commented. Reply to this email directly, view it on GitHub https://github.com/sys-bio/tellurium/issues/408#issuecomment-424775198, or mute the thread https://github.com/notifications/unsubscribe-auth/ABAZDtPuDM6dIvwi8ciAT2m0WoRqr3SYks5ue6hxgaJpZM4W42qu .

-- Herbert Sauro, Associate Professor University of Washington, Bioengineering 206-685-2119, www.sys-bio.org hsauro@uw.edu Books: http://books.analogmachine.org/

[matthiaskoenig]

Yes, please run speed tests. Attached a model of mine, it would be great if this could be part of the speedtests. Simulations take around 6-7 seconds with current roadrunner version. timing.zip

I worked a lot with Andy to bring simulations down under 10 seconds per run (with earlier versions of roadrunner it took around 30s - 1min for a single simulation.

[hsauro]

Since we don't use ASTNodes at runtime, in theory, it shouldn't have an effect. It could affect loading times however. I believe a 5.17.1 is in the works and will coincide with support for Python 3.7.

We also have a major upgrade to the LLVM backend. It's complete but it needs to go through some more testing. One thing we need to check is whether it has an impact on runtime speed or not.

Herbert

On Fri, Oct 19, 2018 at 4:21 AM Matthias König notifications@github.com wrote:

Yes, please run speed tests. Attached a model of mine, it would be great if this could be part of the speedtests. Simulations take around 6-7 seconds with current roadrunner version. timing.zip https://github.com/sys-bio/tellurium/files/2495583/timing.zip

I worked a lot with Andy to bring simulations down under 10 seconds per run (with earlier versions of roadrunner it took around 30s - 1min for a single simulation.

— You are receiving this because you commented. Reply to this email directly, view it on GitHub https://github.com/sys-bio/tellurium/issues/408#issuecomment-431330924, or mute the thread https://github.com/notifications/unsubscribe-auth/ABAZDumP-MkQcGMRG-wnEQdUsfPF_Av6ks5umbWggaJpZM4W42qu .

-- Herbert Sauro, Associate Professor University of Washington, Bioengineering 206-685-2119, www.sys-bio.org hsauro@uw.edu Books: http://books.analogmachine.org/