pepperboxrevolver wrote:Sorry, you lose. TGA supports 1 bit alpha channel. It's either transparent or it's not. PNG supports shaded alpha channels.
I wonder which ugly hacks dsi_floorlogo.tga is using then
Well, so that even you understand why Nexuiz textures are in TGA and not PNG, let's just try some things out. You don't believe us when we are just telling you, but you will surely believe numbers created for the actual game data.
First of all, the total file size. PNG:19,240,811 bytes. TGA: 98,026,161 bytes. A clear difference.
Next, the size once it's compressed into a PK3 (that is, ZIP) file. PNG:19,084,742 bytes. TGA: 21,300,766 bytes. So we would be saving a total of TWO MEGABYTES if we used PNG instead of TGA.
Okay, two megabytes are more than nothing. But what about the loading time? The time that counts is the time of unzipping the PK3 plus the time of loading the images. I don't have a tool here that just loads an image, but for benchmarking I could use converting them to a raw RGB stream and throwing that stream away.
Time for unzipping the PK3 files: the PNGs take 0.93s, the TGAs take 2.26s. But now comes the really important part: time for loading all those images into gqview and generating thumbnails. This took 4.6 seconds for the TGAs and 6.4 seconds for the PNGs.
Makes a total of 7.3s for PNG and 6.9s for TGA.
Of course, you might get different results on your computer... but generally we just "found out": just don't care if it's PNG or TGA. The difference between them is neglegible. You save a slight amount of space when using PNG, but reading them in takes a slight amount of additional time. What speaks for TGA is that it's easier to read and DP has integrated routines for that which are quite simple - I would even go to the point to say that DP's TGA routines are bug free. In contrast, libpng was more than once subject to security holes, so my guess is that libpng still has critical bugs.
and the file mentioned in this thread to the Radiant developers to get official Nexuiz support?