An Intelligent Memory Architecture for 3D Graphics

3D 图形智能内存架构

• 批准号: 14380132
• 负责人: KOBAYASHI Hiroaki
• 金额: $ 5.25万
• 依托单位: Tohoku University
• 依托单位国家: 日本
• 项目类别: Grant-in-Aid for Scientific Research (B)
• 财政年份: 2002
• 资助国家: 日本
• 起止时间: 2002 至 2004
• 项目状态: 已结题
• 来源: https://kaken.nii.ac.jp/en/grant/KAKENHI-PROJECT-14380132/
• 关键词: Memory System; Graphics Hardware; Photo-Realistic Image Synthesis; Power-Aware Computing; Media Processor; Real-Time Ray Tracing; Data Compression; Vector Quantization; グラフィックスプロセッサ; 大域照明モデル; システムLSI; レンダリングパイプライン

We have the following achievements(1)High-performance graphics algorithm and its hardwareWe analyzed parallelism and locality of reference in a graphics algorithm based on the global illumination model, and designed a novel rendering pipeline architecture for this algorithm. In addition, we designed and developed a prototype hardware based on the architecture. Through the performance evaluation of the hardware, we showed its effectiveness for realizing interactive ray-tracing. Moreover, we designed a new high-performance algorithm for generating walkthrough animations.(2)Power-efficient memory mechanismFor design of the intelligent memory architecture for mobile devices, a low-power mechanism for on-chip memory system was designed. In this mechanism, memory modules are activated and inactivated based on their activity during the program execution. We clarified the relationship between activated memory modules and sustained performance, and showed the effectiveness of power-aware computing for on-chip cache memory.(3)Data compression algorithms for graphics hardware.We applied vector quantization to volume data set to achieve efficient data compression, and designed a visualization algorithm that can directly visualize the compressed volume data. We also designed a novel data compression algorithm using data clustering for graphics hardware

(1)高性能图形算法及其硬件分析了一种基于全局光照模型的图形算法的并行性和引用局部性，并为该算法设计了一种新颖的渲染流水线结构。此外，我们还设计并开发了一个基于该体系结构的硬件原型。通过对硬件的性能评估，验证了其对实现交互式光线跟踪的有效性。此外，我们还设计了一种新的高性能漫游动画生成算法。(2)低功耗存储机制为移动设备智能存储体系结构的设计，设计了一种适用于片上存储系统的低功耗机制。在该机制中，存储器模块根据它们在程序执行期间的活动被激活和停用。针对图形硬件的数据压缩算法，将矢量量化应用于体数据集，实现了高效的数据压缩，并设计了一种可视化算法，可以直接对压缩后的体数据进行可视化。并针对图形硬件设计了一种新颖的基于数据聚类的数据压缩算法

项目成果

視覚的画質劣化を抑制するベクトル量子化手法

抑制视觉图像质量恶化的矢量量化方法

• 发表时间: 2002
• 期刊: 情報科学技術フォーラム(FIT) Information Technology Letters Vol.1
• 作者: S.Li;H.Ogawa;三浦 健
• 通讯作者: 三浦 健

Could Wave Pipeline overcome Commodity Pipelines?

波浪管道能否战胜商品管道？

• 发表时间: 2004
• 期刊: Journal of Information Vol7, no5
• 作者: X.Jiang;P.H.Ho;H.Shen;S.Horiguchi;渥美 紀寿ほか;Ryusuke Egawa
• 通讯作者: Ryusuke Egawa

Design Variation and Performance Trade-Offs by Changing Processing Rates of Functional Elements in a Reconfigurable Embedded System

通过改变可重构嵌入式系统中功能元件的处理速率来进行设计变化和性能权衡

• 发表时间: 2003
• 期刊: Proceedings of COOL Chips VI
• 作者: K.Degawa;T.Aoki;T.Higuchi;Seong-Yong Ahn
• 通讯作者: Seong-Yong Ahn

Outlook for low-power and high-performance processor design : How can we survive in an era of billion-transistor chips?

低功耗高性能处理器设计展望：我们如何在十亿晶体管芯片时代生存？

• 发表时间: 2004
• 期刊: Proceedings of COOL Chips VII
• 作者: Hiroaki Kobayashi(Ed.)

A Comparison Study Of Vector Quantization Codebook Design Algorithms Based On The Equidistortion Principle

基于等失真原理的矢量量化码本设计算法比较研究

• 发表时间: 2003
• 期刊: Proceedings of the 21st IASTED International Conference on Applied Informatics
• 作者: Inoue;Kazuko;土井章男;竹内 啓;Hiroyuki Takizawa
• 通讯作者: Hiroyuki Takizawa