LuxRender 0.9-dev Windows build of 04-03-2012

[Lord Crc]

Glorious new builds!

Please check out SATtva's post for an update on what's new and fresh.

These builds are built using Boost 1.47 still, so there may still be some issues with network rendering. Please report any issues

These builds contain LuxRender, LuxConsole, LuxMerger, LuxBlend25 and Pylux for Blender 2.6x, and LuxBlend for Blender 2.49.
The NoOpenCL builds does not require any OpenCL driver or capabilities, but also does not use the GPU for the "hybrid" renderer.

LuxRender_CVS_20120304_64bit_OpenCL.exe
LuxRender_CVS_20120304_64bit_NoOpenCL.exe
LuxRender_CVS_20120304_32bit_SSE2_OpenCL.exe
LuxRender_CVS_20120304_32bit_SSE2_NoOpenCL.exe
LuxRender_CVS_20120304_32bit_SSE1_NoOpenCL.exe

Built with Visual Studio 2008 SP1 + Windows 7 SDK on Win 7 64bit, using AMD APP SDK 2.5 (formerly ATI Stream SDK).

Changelog

Code: Select all

85 minutes ago   Jens Verwiebe   LuxConsole: don´t show LUX_ERROR for --help, rather show LUX_INFO
33 hours ago   Asbjørn Heid   Added sanity checking to QueryableRegistry registration and deregistration.
2 days ago   Asbjørn Heid   Added missing changes to api files in previous commit.
2 days ago   Asbjørn Heid   Made master ignore debug and info level log entries from slaves, unless master is running in verbose mode.
2 days ago   Asbjørn Heid   Fixed issue with transmitting log from slave to master.
2 days ago   Asbjørn Heid   Added some logging in the server's ServerDisconnect handler to indicate that the server is ready for a new session.
2 days ago   Asbjørn Heid   Added ServerReconnect command to the networking protocol.
2 days ago   Asbjørn Heid   Fixed typos in SQBVH preventing compilation.
2 days ago   Asbjørn Heid   Exposed QueryableAttribute's description field in API.
3 days ago   David Bucciarelli   Fixed the total SAH cost statistics for (S)QBVH
3 days ago   David Bucciarelli   Merge
3 days ago   David Bucciarelli   If a primitive isn't a polygon, spatial-split in SQBVH will clip the bounding box
4 days ago   Guillaume Bouchard   BSDF: reduce memory usage of Lambertian BxDF
4 days ago   David Bucciarelli   Minor code cleanup in SQBVH
4 days ago   Jean-Philippe Grimaldi   New rendering hints light sampling strategy code:
5 days ago   Jens Verwiebe   OSX: added a comment to understand how flags are assembled, cmake is a bit intransparent here
5 days ago   Jens Verwiebe   OSX: cleanup for last commit, it legal to use global compileflags in "add_definition"
5 days ago   David Bucciarelli   Merge
5 days ago   David Bucciarelli   First work for reference unsplitting in SQBVH code
6 days ago   David Bucciarelli   A small clean up in SQBVH splitting code
6 days ago   David Bucciarelli   Updated SQBVH spatial-split code in order to handle AreaLightPrimitive too
6 days ago   Jens Verwiebe   OSX: better distinguish between debug/release flags, don´t use global definitions for compileflags


Enjoy

[Piita]

Glorious news! Many thanks.

[cwichura]

Did something change with the way the total section of the statistics are calculated? In prior versions, the total number would be pretty far off until a couple of cycles of retrieving samples from the slaves. But I've been running this all day with two slaves, and my total estimated samples/pixel is WAY off. I started the morning resuming a film file that has about 1200S/p. The master and slaves combined have knocked out another 800 or so since I started it. Yet the estimated total is saying ~4950S/p.

And while it's nice to try and estimate what the total is if one were to pool all the slaves together right at that moment, in truth, I'd be more interested in having a real, concrete number displayed for the actual number of samples that the master has currently...

Thanks

[J the Ninja]

Did something change with the way the total section of the statistics are calculated? In prior versions, the total number would be pretty far off until a couple of cycles of retrieving samples from the slaves. But I've been running this all day with two slaves, and my total estimated samples/pixel is WAY off. I started the morning resuming a film file that has about 1200S/p. The master and slaves combined have knocked out another 800 or so since I started it. Yet the estimated total is saying ~4950S/p.

And while it's nice to try and estimate what the total is if one were to pool all the slaves together right at that moment, in truth, I'd be more interested in having a real, concrete number displayed for the actual number of samples that the master has currently...

Thanks

It's been that bad since total was first introduced. It gets worse the more s/px the film you resumed from has:

http://www.luxrender.net/mantis/view.php?id=956

[cwichura]

Thanks for the pointer to the bugtracker entry. In the short term, I'd again suggest just showning the actual known total samples that the master has collected.

I've also for the first time had a slave tick over to 1000+ samples. On the network tab, the slave's sample count is being displayed in scientific notation ("1.19e+03 S/p"), which seems unnecessary since the field is plenty wide enough to display the sample count in decimal.

[Lord Crc]

I've also for the first time had a slave tick over to 1000+ samples. On the network tab, the slave's sample count is being displayed in scientific notation ("1.19e+03 S/p"), which seems unnecessary since the field is plenty wide enough to display the sample count in decimal.

I was pretty sure I had committed a fix for this...

[cwichura]

I am running this most current weekly on all my machines now in order to get your changes for slave reconnect. So it must not have made it in.

I would also like to make another suggestion for the network tab. Right now, it displays the cumulative number of S/p received from each slave. It'd also be nice if it had columns for current S/s (and/or C/s) rate and current S/p where these columns would be updated each time it fetches the logs from the slave. Given how broken the totals are, I used to eyeball where things are by looking at the log messages retrieved from the slaves on the master. (I generally use 1 hour as the interval between slave sample retrieval, so the every-five-minutes (with my settings) log fetch would give me a much better idea of where things are at.) But with this new build, the status info lines are now filtered out on the master, so I no-longer can see where each slave is at unless I remote into the box.

And if you want to put a bow on it, add a row at the bottom for "totals", if only for the two S/p columns (fetched and current working as of last log pull).

[cwichura]

And an extra special bow on top: Add an "update now" button next to the update interval setting. I know this can be triggered by setting the interval to something very low like 1 and then setting it back after the update cycle has started. But would be much easier to just click a button...

[jeanphi]

Hi,

The button is already present at the bottom of the imaging pane.

Jeanphi

[J the Ninja]

Hi,

The button is already present at the bottom of the imaging pane.

Jeanphi

I think he meant for network rendering?

Also, should that imaging "Apply Now" button be changed to "update now"? I mean, it applies itself automatically every time the tone mapping interval rolls around.