- UID
- 2
- 积分
- 8682
- 帖子
- 2905
- 主题
- 199
- 论坛币
- 11740
- 威望
- 16
- EP值
- 2349
- MP值
- 15
- 阅读权限
- 200
- 注册时间
- 2011-8-3
- 在线时间
- 2597 小时
- 最后登录
- 2024-8-28
|
I spent several days on updating the libtcas library, for I had got some good ideas to improve it. Two main conceptual guidelines are that using multi-threading and caching to improve the overall performance of rendering. Resulted in two sets of high level APIs, the hla_file_cache and hla_frame_cache, these APIs are more pertinent to the performance issues than hla_mt and hla_mt_mm APIs.
I implemented them without too much difficulties, since the logic was very clear. And the techniques are also some kind of mature. With the two newly developed API sets, the performance of the Avisynth interface has been improved dramatically, however, though haven't figured out yet, the performance of the DirectShow interface doesn't improve, but even decreased. So I kept it using the old way of rendering.
I believe now is the time to create more TCAS FXs, for both experimental purpose and showing what TCAS can do.
|
|