直接使用开源版本 无需License

H265ENC V2.0 Feature Improvement h265enc v2.0 a. Update the I/P prediction algorithms, and the hardware pipeline to improve the coding efficiency b. Fix the bugs in the previous version c. Check the RTL V2.0 by nLint, and fix the errors and warnings d. Meet the 400MHz synthesized by Design Complier，and fix critical path，warning e. IP Core V2.0 has been validated by the FPGA PreI – Intra mode decision a. Support CTU-level Rate control，based on the CABAC bit number to change the QP adaptively PosI – Intra partition decision a. Prediction for the partition decision is based on the original pixels b. The RD cost for the partition decision is based on the SATD and simplified rate estimation IME – Integer motion estimation a. Configurable IME parameters by the ime_cfg.dat, such as the search range, search center , and search shape b. Extend the search range to ±64 c. By the H-V SRAMs, the search direction of novel IME can be the slope d. By the Bit Truncation, the pixels bit depth is truncated into 4 bit from 8bit with only less than 1dB PSNR loss, but with great improvement on the area and power FME – Fractional motion estimation a. Support the SKIP and MERGE decision by the RD cost similar to the FME decision b. Fix the bugs in the interpolation REC – Reconstruction loop a． Support Intra CTU in Inter Frame by the costs from PosI and FME to decide I/P of current CTU b． Support the SKIP CU, where the prediction samples will be the reconstruction samples c． Rearrange the DCT RTL and fix the bugs DBSAO – Deblocking filter & sample adaptive offset a. Update the DB filter order and remove the transpose memory to reduce memory usage b. Support the SAO EO mode CABAC – Entropy coding a. Update the RTL and fix the bugs b. Support SKIP、MERGE、Intra CTU in Inter Frame and so on c. Remove the ROM in the previous version

@nicu_233 是的 264 的更新。 265 更新我们稍晚也会放出来

@shixiaogang http://openasic.org/topic/32/2017-1-13-ic-caffe线下研讨会ppt文档

introduction和user guide两个文档确实不能下载

这个具体要看我们发表得论文

@duoerpu 好的，你可以联系我， fanyibo@fudan.edu.cn

@ruibo 这个不是我们开发的代码，我也搞不清楚，你可以问问原作者

C代码目前没开源

@asuperbird 编码器延迟理论上很低的，大概5w个时钟周期左右。通过仿真能看到精确的延迟。编码系统延迟还包括帧缓存延迟，bitstream的缓存延迟，通讯延迟等。

@Scofield 我们正在做这方面的工作

@shen_lee x_total y_total 超过你要的边界， 还需要padding出边界到64的倍数

@eesiukei 目前还没有。