Hi Francois:
The more texture sheets per single object, the more texture file I/O
and mapped texture material draw calls occur.
[
EDIT]
I think 2 texture draws per "side" for a large hangar might not be too resource-intensive, if its complexity is otherwise average.
Ex: 44 meters per long side at 50 pixels applied per meter = 2,200 pixels; that could be 2 texture spans per side on a 1024 wide sheet
[
END_EDIT]
Although IIUC one could theoretically even use 2048 or 4096 square textures for FS objects rather than 1024 x 1024, 512 x 512, 256 x 256 (or even the rarely seen 128 x 128) pixel texture file, with texture files over 1024 x1024 in size, there is a rapidly diminishing return in performance on less powerful FS systems.
On a practical basis, there is also a point of diminishing returns as to just how "well-defined" or "higher resolution" ones' textures
could be before one's object fails to integrate well with the surrounding scenery when compared to nearby objects and ground textures.
Most of us might remember some of the early special effects used in movies which had superimposed flames that were unusually "sharp", water bodies / waves that looked "too wet", or buildings or rock formations / mountains that stood out too much due to overly well-defined edges, shadows, or specular shine highlights from the earlier methods used in production.
So, IMHO, one's objects must maintain uniformity relative to the remainder of the scenery in FS for acceptable "realism" at run-time.
[
EDIT]
For example, the "door side" and "roof" of the metal quonset hangar at FSAddon's Plum Island are both about 20 meters wide / long and the mapped *.DDS texture for each is just under 1024 pixels wide (about 50 pixels applied per linear meter) ... and IMHO that object renders rather nicely into the overall scene.
The "front door wing" of the FBO building at LAGO's Emma Field 2004 is around 7 meters wide, and with adjoining left (north) and right (south) walls, the
total main building long axis is 25 meters.
Brick wall portions of that (west) side of the building appear to use 3 mapped textures at 30%, 15% and 40% (85% total span) of a 512 pixel wide *.BMP (about 17 pixels applied per linear meter); and yet, IMHO, that object
also renders rather nicely into the overall scene, even though the images also appear inherently (and
purposely ?) ...not as 'sharp' as those used at Plum Island.
IMHO, while creating a realistic perception of FS visual "distance" that takes into account a "perceived" lesser resolution of textures at normal in-sim viewing distances (ex: without a drop in MDL LOD), one might expect for example, that interposed atmospheric 'haze' with dust and pollen from Fritz's 'constant gardening' (with various mowers and weed-whackers !
) could result in the Emma Field FBO looking slightly less sharp when navigating about on the grounds nearby!
PS: I'd also convert any "lossy" JPG texture source files into a "non-lossy"
ex: *.BMP, *.TIF, *.PNG, *.TGA, file
before doing even (1) editing operation of preliminary compositing, sharpening, adjusting of color balance/brightness/contrast/gamma etc. prior to conversion to the final material file format, thus avoiding a cumulative loss of image detail which otherwise occurs incrementally each time the file is 'saved' back in a 'lossy JPG' format.
[
END_EDIT]
Hope these considerations might help !
GaryGB
