Timeline for Ara matrices calculated on the GPU or on the CPU? Would built-in matrix functions be faster than my custom ones?
Current License: CC BY-SA 3.0
9 events
| when toggle format | what | by | license | comment | |
|---|---|---|---|---|---|
| Oct 29, 2011 at 22:16 | comment | added | Kasper | Functions like glLoadIdentity, glMultMatrix, glRotate, glScale etc... are deprecated. You have to write a shader and upload your own matrices using uniform variables. | |
| Oct 28, 2011 at 13:38 | vote | accept | Yevhen | ||
| Oct 28, 2011 at 9:19 | comment | added | e-MEE | what do you mean by "matrix manipulations in OpenGL are deprecated"? | |
| Oct 28, 2011 at 6:49 | comment | added | 5ound | It is worth noting that many GL fixed function matrix implementations shortcut multiply operations when possible. i.e. glIdentity doesn't necessarily set any values, it can flip a bit to flag itself as identity and the next matrix operation will replace the matrix instead of multiplying. I've studied the mesa source code and saw all sorts of shortcuts. Using someone else's optimized library is usually better than writing your own. | |
| Oct 27, 2011 at 17:04 | comment | added | Kylotan | 'the total amount of floating point operations you'll do involving matrices will be less than 1000' - that's sadly not true in most complex games. You might do more than that per character. | |
| Oct 27, 2011 at 14:41 | comment | added | Kasper | If he's talking about OpenGL then he can only transform his geometry on the GPU and the question is moot anyway. | |
| Oct 27, 2011 at 11:23 | comment | added | Valmond | ...the final multiplication of position by MVP... ... and of tangents and bitangents... and skeletal animations and ... Well, all matrix multiplications DOES NOT occur on the CPU. | |
| Oct 27, 2011 at 11:00 | comment | added | Maximus Minimus | +1 and just to add - the final multiplication of position by MVP is what's carried out on the GPU. | |
| Oct 27, 2011 at 10:36 | history | answered | Kasper | CC BY-SA 3.0 |