移动GPU可重构背面剔除单元设计与实现
|
摘要
针对移动图形处理器对高能效和灵活性的需求,提出一种可重构背面剔除单元的设计方法。根据背面剔除算法的不同实现形式,采用可重构阵列结构,通过4×4个多核处理单元对乘加运算进行并行处理,其中包括对数据的拆分与准备、乘加运算、以及对不同计算方式的可重构切换,完成可重构背面剔除单元的设计。所设计的可重构背面剔除单元,能够对不可见的背面进行正确剔除,工作频率为121.18 MHz,与专用硬件实现相比,更具灵活性。
In view of the demand for high efficiency and flexibility of mobile graphics processors,a design method of a reconfigurable backface culling unit is proposed.According to the different realization forms of backface culling algorithm,using the reconfigurable array structure,the multiplication and addition operations are processed in parallel by 4 × 4 multi-core processing units,which includs disassembly and preparation of data,operations of multiplying and adding,and reconfigurable switching of different computing methods,to complete the design of the reconfigurable backface culling unit.The designed reconfigurable backface culling unit can remove the invisible back correctly,and the working frequency is 121.18 MHz,which is more flexible than the special hardware implementation.
In view of the demand for high efficiency and flexibility of mobile graphics processors,a design method of a reconfigurable backface culling unit is proposed.According to the different realization forms of backface culling algorithm,using the reconfigurable array structure,the multiplication and addition operations are processed in parallel by 4 × 4 multi-core processing units,which includs disassembly and preparation of data,operations of multiplying and adding,and reconfigurable switching of different computing methods,to complete the design of the reconfigurable backface culling unit.The designed reconfigurable backface culling unit can remove the invisible back correctly,and the working frequency is 121.18 MHz,which is more flexible than the special hardware implementation.
引文
[1]焦继业.低功耗高性能移动图形顶点处理器设计关键技术研究[D/OL].西安:西安电子科技大学.2013:1-13[2016-11-20].http://dx.chinadoi.cn/10.7666/d.D365697.
[2]杨国东.嵌入式图形处理器的研究与实现[D/OL].济南:山东大学.2010:1-3[2016-11-20].http://dx.chinadoi.cn/10.7666/d.y1791403.
[3]KIM S,KIM H E,KIM H,et al.Computing energyefficiency in the mobile GPU[C/OL]//SoC Design Conference(ISOCC),2013International.South Korea Busan:IEEE,2013:219-221[2016-11-20].https://doi.org/10.1109/ISOCC.2013.6864012.
[4]PATNEY A,TZENG S,SEITZ K A,et al.Piko:a framework for authoring programmable graphics pipelines[J/OL].ACM Transactions on Graphics,2015,34(4):147:1-13[2016-11-20].http://dx.doi.org/10.1145/2766973.
[5]KIM W Y,LEE B H,LEE K Y,et al.Design of a fully programmable shader processor for low power mobile devices[C/OL]//TENCON 2009:2009IEEE Region 10 Conference.Singapore:IEEE,2009:1-5[2016-11-22].https://doi.org/10.1109/TENCON.2009.5396077.
[6]MOREIN S L,LEFEBVRE L,GRUBER A E,et al.Graphics processing architecture employing a unified shader:EP,US8760454[P/OL].2014-06-24[2016-11-20].http://www.freepatentsonline.com/8760454.html.
[7]AKENINE-MOLLER T,HAINES E,HOFFMAN N.Real-time rendering[M].Boca Raton:CRC Press,2008:645-693.
[8]苏诚,韩俊刚,朱豪杰,等.背面剔除的一种新算法[J/OL].西安邮电大学学报,2013,18(1):82-85[2016-11-21].http://dx.chinadoi.cn/10.13682/j.issn.2095-6533.2013.01.025.
[9]张霞.计算机图形图像处理[M].北京:电子工业出版社,2010:40-43.
[10]田泽,杜斐,张骏.一种3D图形背面消隐算法及其硬件加速实现[J/OL].小型微型计算机系统,2015,36(6):1403-1408[2016-11-23].http://xwxt.sict.ac.cn/CN/Y2015/V36/I6/1403.
[11]沈绪榜,孙璐.计算模式的统一研究[J/OL].计算机学报,2014,37(7):1435-1444[2016-11-23].http://cjc.ict.ac.cn/online/onlinepaper/sxb-2014724220854.pdf.DOI:10.3724/SP.J.1016.2014.1435.
[12]蒋林,王杏军,刘镇弢,等.基于SystemC的可重构阵列处理器模型[J/OL].西安邮电大学学报,2016,21(3):73-78[2016-11-23].http://dx.chinadoi.cn/10.13682/j.issn.2095-6533.2016.03.012.
[13]MUNSHI A,GINSBURG D,SHREINER D.OpenGL ES 2.0 Programming Guide[M].Boston:AddisonWesley,2008:115-126.
[14]SHREINER D.OpenGl编程指南(原书第7版)[M].李军,徐波,译.北京:机械工业出版社,2010:31-41.
[15]邓军勇,李涛,蒋林,等.面向OpenGL的图形加速器设计与实现[J/OL].西安电子科技大学报(自然科学版),2015,42(6):124-130[2016-11-20].http://dx.chinadoi.cn/10.3969/j.issn.1001-2400.2015.06.022.
[2]杨国东.嵌入式图形处理器的研究与实现[D/OL].济南:山东大学.2010:1-3[2016-11-20].http://dx.chinadoi.cn/10.7666/d.y1791403.
[3]KIM S,KIM H E,KIM H,et al.Computing energyefficiency in the mobile GPU[C/OL]//SoC Design Conference(ISOCC),2013International.South Korea Busan:IEEE,2013:219-221[2016-11-20].https://doi.org/10.1109/ISOCC.2013.6864012.
[4]PATNEY A,TZENG S,SEITZ K A,et al.Piko:a framework for authoring programmable graphics pipelines[J/OL].ACM Transactions on Graphics,2015,34(4):147:1-13[2016-11-20].http://dx.doi.org/10.1145/2766973.
[5]KIM W Y,LEE B H,LEE K Y,et al.Design of a fully programmable shader processor for low power mobile devices[C/OL]//TENCON 2009:2009IEEE Region 10 Conference.Singapore:IEEE,2009:1-5[2016-11-22].https://doi.org/10.1109/TENCON.2009.5396077.
[6]MOREIN S L,LEFEBVRE L,GRUBER A E,et al.Graphics processing architecture employing a unified shader:EP,US8760454[P/OL].2014-06-24[2016-11-20].http://www.freepatentsonline.com/8760454.html.
[7]AKENINE-MOLLER T,HAINES E,HOFFMAN N.Real-time rendering[M].Boca Raton:CRC Press,2008:645-693.
[8]苏诚,韩俊刚,朱豪杰,等.背面剔除的一种新算法[J/OL].西安邮电大学学报,2013,18(1):82-85[2016-11-21].http://dx.chinadoi.cn/10.13682/j.issn.2095-6533.2013.01.025.
[9]张霞.计算机图形图像处理[M].北京:电子工业出版社,2010:40-43.
[10]田泽,杜斐,张骏.一种3D图形背面消隐算法及其硬件加速实现[J/OL].小型微型计算机系统,2015,36(6):1403-1408[2016-11-23].http://xwxt.sict.ac.cn/CN/Y2015/V36/I6/1403.
[11]沈绪榜,孙璐.计算模式的统一研究[J/OL].计算机学报,2014,37(7):1435-1444[2016-11-23].http://cjc.ict.ac.cn/online/onlinepaper/sxb-2014724220854.pdf.DOI:10.3724/SP.J.1016.2014.1435.
[12]蒋林,王杏军,刘镇弢,等.基于SystemC的可重构阵列处理器模型[J/OL].西安邮电大学学报,2016,21(3):73-78[2016-11-23].http://dx.chinadoi.cn/10.13682/j.issn.2095-6533.2016.03.012.
[13]MUNSHI A,GINSBURG D,SHREINER D.OpenGL ES 2.0 Programming Guide[M].Boston:AddisonWesley,2008:115-126.
[14]SHREINER D.OpenGl编程指南(原书第7版)[M].李军,徐波,译.北京:机械工业出版社,2010:31-41.
[15]邓军勇,李涛,蒋林,等.面向OpenGL的图形加速器设计与实现[J/OL].西安电子科技大学报(自然科学版),2015,42(6):124-130[2016-11-20].http://dx.chinadoi.cn/10.3969/j.issn.1001-2400.2015.06.022.