Re: [PATCH 2/2] d3dx9_36: Implement triangle filter when create textures
2015-06-09 10:49 GMT+02:00 Sergey Isakov <isakov-sl(a)bk.ru>:
This algo just tested long ago. May be it can be rewritten to float point math but it requires additional efforts.
It does differ from MS algo because it is my algo not related to MS.
Manually observe? I did this and sent a picture. Can't imagine what should be done with tests suit.
The test I mentioned is not meant for the Wine test suite (even though it would be nice to have some kind of Wine test eventually). You would write a standalone application (or hack something in the d3dx9 tests, it might be easier) along the lines of what I described and run it against native d3dx9_36.dll, ideally on Windows. You'd save the filtered images and eyeball them to try to figure out what's the filtering algorithm used by MS. I have no idea whether it will be doable in practice or not but it seems worth a try.
OK, quick test WineHQ + d3dx9_36 builtin d3dx9_36 native my mod of d3dx9_36 Yes, not same result. On 09.06.2015, at 12:09, Matteo Bruni wrote:
2015-06-09 10:49 GMT+02:00 Sergey Isakov <isakov-sl(a)bk.ru>:
This algo just tested long ago. May be it can be rewritten to float point math but it requires additional efforts.
It does differ from MS algo because it is my algo not related to MS.
Manually observe? I did this and sent a picture. Can't imagine what should be done with tests suit.
The test I mentioned is not meant for the Wine test suite (even though it would be nice to have some kind of Wine test eventually). You would write a standalone application (or hack something in the d3dx9 tests, it might be easier) along the lines of what I described and run it against native d3dx9_36.dll, ideally on Windows. You'd save the filtered images and eyeball them to try to figure out what's the filtering algorithm used by MS. I have no idea whether it will be doable in practice or not but it seems worth a try.
participants (2)
-
Matteo Bruni -
Sergey Isakov