Convert normal maps to RGTC?

[STiFU 569]

Do you have a converter for RGTC @duzenko? I'd like to conduct an experiment so see how loading times are affected. I assume their is no code support needed to load RGTC normals at this point?

[duzenko 627]

3 hours ago, STiFU said:

Do you have a converter for RGTC @duzenko? I'd like to conduct an experiment so see how loading times are affected. I assume their is no code support needed to load RGTC normals at this point?

Good question that I can't answer fully ATM. D3 offers two automatic conversion options:

• via the r_writeCompress_something cvar that is supposed to let the GL driver do the conversion and then simply bit copy the result from the VRAM. You'd think that driver's on-the-fly compression is fast but image quality might suffer.
• using an external command-line tool. Supposedly this is the recommended option. We still need to decide which 3rd party compressor we want to use and how to wire it to TDM (command line params, etc)

Feel free to try either of these but keep in mind that texture loading is only a fraction of total map loading time.

I'd like to get the RGTC conversion completed but afraid to touch it before 2.08 is out

[STiFU 569]

40 minutes ago, duzenko said:

Feel free to try either of these but keep in mind that texture loading is only a fraction of total map loading time.

How can you tell? Did you profile map loading? If so, I'd be very interested to see the results!!

Quote

I'd like to get the RGTC conversion completed but afraid to touch it before 2.08 is out

Yes, this is definitely a 2.09 task.

[duzenko 627]

20 minutes ago, STiFU said:

How can you tell? Did you profile map loading? If so, I'd be very interested to see the results!!

Yes, this is definitely a 2.09 task.

Unfortunately VS2019 profiling does not work for me

But image_preload 0 will do the job - try it.

[stgatilov 800]

 

17 hours ago, STiFU said:

Do you have a converter for RGTC @duzenko? I'd like to conduct an experiment so see how loading times are affected. I assume their is no code support needed to load RGTC normals at this point?

If you have time to have fun, you can try to implement your own in something like Python. All those texture compression formats are so simple compared to PNG and JPG, that writing a basic converter is a one-day problem

13 hours ago, duzenko said:

Unfortunately VS2019 profiling does not work for me

I'm afraid CPU profiling is not the right way to conduct experiments in this case. There are many things involved which you won't be able to properly measure with profiler: disk accesses, OpenGL driver work, GPU transfers, etc. I suggest doing more complicated measurements with temporarily disabling some parts in the code. I guess most of the time should be taken by texture loading, models loading, and sound loading (also covered in zip compression unless you are on SVN).

[duzenko 627]

2 hours ago, stgatilov said:

 I guess most of the time should be taken by texture loading, models loading, and sound loading (also covered in zip compression unless you are on SVN).

.map/.proc/.cm files can take a while to load, as well as AAS parsing

Then, there are also scripts, def's, materials

[HMart 283]

On 3/26/2020 at 8:01 PM, duzenko said:

Good question that I can't answer fully ATM. D3 offers two automatic conversion options:

• via the r_writeCompress_something cvar that is supposed to let the GL driver do the conversion and then simply bit copy the result from the VRAM. You'd think that driver's on-the-fly compression is fast but image quality might suffer.
• using an external command-line tool. Supposedly this is the recommended option. We still need to decide which 3rd party compressor we want to use and how to wire it to TDM (command line params, etc)

Feel free to try either of these but keep in mind that texture loading is only a fraction of total map loading time.

I'd like to get the RGTC conversion completed but afraid to touch it before 2.08 is out

What about ATI/AMD compressionator? It was the tool that idsoftware used to compress normal maps for Doom 3 and splash damage for quake wars, you can find the old version used in Doom 3 here.

Or perhaps ODCompress from Overdose?

[duzenko 627]

48 minutes ago, HMart said:

What about ATI/AMD compressionator? It was the tool that idsoftware used to compress normal maps for Doom 3 and splash damage for quake wars, you can find the old version used in Doom 3 here.

Or perhaps ODCompress from Overdose?

Does either of them support RGTC?

[New Horizon 481]

I don't see RGTC in the Overdose menu and I doubt compressonator would support it since it's an old program.

[HMart 283]

3 hours ago, duzenko said:

Does either of them support RGTC?

Don't know a thing about texture compression but latest compressionator "Can compress to a wide range of compression formats including ASTC, ATC, ATInN, BCn, ETCn, DXTn and swizzled DXTn formats, and supports conversion of textures between 8bit fixed, 16bit fixed and 32bit float formats."

 

Taken from this link:   "ASTC provides a single-channel and two-channel version, equivalent to RGTC's formats."

  

Edited March 29 by HMart

[nbohr1more 1975]

As I recall, RGTC is just another name for Microsoft's BC4 or 5

This seems to confirm it:

RGTC1_RED = BC4U, RGTC1_SIGNED_RED = BC4S, RGTC2_RG = BC5U, RGTC2_SIGNED_RG = BC5S

 

[duzenko 627]

6 hours ago, HMart said:

Don't know a thing about texture compression but latest compressionator "Can compress to a wide range of compression formats including ASTC, ATC, ATInN, BCn, ETCn, DXTn and swizzled DXTn formats, and supports conversion of textures between 8bit fixed, 16bit fixed and 32bit float formats."

 

Taken from this link:   "ASTC provides a single-channel and two-channel version, equivalent to RGTC's formats."

  

Can't see any reference of a compression tool on that link

[HMart 283]

4 hours ago, duzenko said:

Can't see any reference of a compression tool on that link

Nor was that my intention I was just showing that I've read that ASTC is equivalent to RGTC but I don't know if is really true or not. 

Ha now I know what you meant by that, the first quote in dark gray from my reply above was from the latest AMD compressonator page, the link to it is on my reply above with the links for the tools, the other quote in white, is from another link, the "taken from this link" was referring to the white quote only.    

Edited March 29 by HMart