Go to most recent revision | Details | Last modification | View Log | RSS feed
| Rev | Author | Line No. | Line |
|---|---|---|---|
| 4358 | Serge | 1 | |
| 2 | |||
| 3 | |||
| 4 | |||
| 5 | General Public License, see the LICENSE file for details. |
||
| 6 | |||
| 7 | |||
| 8 | |||
| 9 | started. For more detaile information consult the web page. |
||
| 10 | |||
| 11 | |||
| 12 | |||
| 13 | |||
| 14 | |||
| 15 | |||
| 16 | Contributors: |
||
| 17 | Emil Briggs (briggs@bucky.physics.ncsu.edu) |
||
| 18 | David Bucciarelli (tech.hmw@plus.it) |
||
| 19 | Andreas Schiffler (schiffler@zkm.de) |
||
| 20 | |||
| 21 | |||
| 22 | |||
| 23 | |||
| 24 | Mesa 3.0. |
||
| 25 | An SMP capable machine running Linux 2.x |
||
| 26 | libpthread installed on your machine. |
||
| 27 | |||
| 28 | |||
| 29 | |||
| 30 | MITS stands for Mesa Internal Threading System. By adding |
||
| 31 | internal threading to Mesa it should be possible to improve |
||
| 32 | performance of OpenGL applications on SMP machines. |
||
| 33 | |||
| 34 | |||
| 35 | |||
| 36 | No. The threading is internal to Mesa and transparent to |
||
| 37 | applications. |
||
| 38 | |||
| 39 | |||
| 40 | |||
| 41 | No. This implementation splits the processing of the vertex buffer |
||
| 42 | over two threads. There is a certain amount of overhead involved |
||
| 43 | with the thread synchronization and if there is not enough work |
||
| 44 | to be done the extra overhead outweighs any speedup from using |
||
| 45 | dual processors. You will not for example see any speedup when |
||
| 46 | running Quake because it uses GL_POLYGON and there is only one |
||
| 47 | polygon for each vertex buffer processed. Test results on a |
||
| 48 | dual 200 Mhz. Pentium Pro system show that one needs around |
||
| 49 | 100-200 vertices in the vertex buffer before any there is any |
||
| 50 | appreciable benefit from the threading. |
||
| 51 | |||
| 52 | |||
| 53 | |||
| 54 | Yes. You can try to vary the size of the vertex buffer which is |
||
| 55 | define in VB_MAX located in the file src/vb.h from your top level |
||
| 56 | Mesa distribution. The number needs to be a multiple of 12 and |
||
| 57 | the optimum value will probably depend on the capabilities of |
||
| 58 | your machine and the particular application you are running. |
||
| 59 | |||
| 60 | |||
| 61 | |||
| 62 | performance with the MITS? |
||
| 63 | Yes. Try to use as many vertices between each Begin/End pair |
||
| 64 | as possbile. This will reduce the thread synchronization |
||
| 65 | overhead. |
||
| 66 | |||
| 67 | |||
| 68 | |||
| 69 | On some benchmarks performance gains of up to 30% have been |
||
| 70 | observerd. Others may see no gain at all and in a few rare |
||
| 71 | cases even some degradation. |
||
| 72 | |||
| 73 | |||
| 74 | |||
| 75 | Lots of testing and benchmarking. |
||
| 76 | A portable implementation that works within the Mesa thread API. |
||
| 77 | Threading of additional areas of Mesa to improve performance |
||
| 78 | even more. |
||
| 79 | |||
| 80 | |||
| 81 | |||
| 82 | |||
| 83 | |||
| 84 | |||
| 85 | from source. |
||
| 86 | 2. Place the tarball MITS.tar.gz in your top level Mesa directory. |
||
| 87 | 3. Unzip it and untar it. It will replace the following files in |
||
| 88 | your Mesa source tree so back them up if you want to save them. |
||
| 89 | |||
| 90 | |||
| 91 | |||
| 92 | Make-config |
||
| 93 | Makefile |
||
| 94 | mklib.glide |
||
| 95 | src/vbxform.c |
||
| 96 | src/vb.h |
||
| 97 | |||
| 98 | |||
| 99 | |||
| 100 | |||
| 101 |