开普勒是否在图形上模拟了双精度浮点？
[英]Is double precision floating point emulated on Kepler for graphics?
问题描述
Comparing the SMXes on the graphics and compute versions of Kepler, I noticed that there are no units for double precision in the block diagram for the graphics version. 
比较开普勒的图形和计算版本上的SMX，我注意到图形版本的框图中没有双精度单位。 The blocks are in yellow on the compute version. 计算版本上的块为黄色。 Is double precision emulated on the graphics chip? 图形芯片上是否模拟了双精度？ It seems possible, since throughput is only 1/24 of single precision. 这似乎是可行的，因为吞吐量仅为单精度的1/24。
1 个解决方案
解决方案1
4 已采纳 2013-02-02 04:25:24
NVIDIA has two version of Kepler SM. NVIDIA有两个版本的Kepler SM。 SM 3.0 targets graphics and non-double precision workloads. SM 3.0针对图形和非双精度工作负载。 The SM only has sufficient data paths for 8 64-bit operations per cycle (x2 if doing DFMA). SM每个周期仅具有足够的数据路径用于8个64位操作（如果执行DFMA，则为x2）。 SM 3.5 (K20) targets high performance compute workloads. SM 3.5（K20）针对高性能计算工作负载。 SM 3.5 supports 64 64-bit operations per cycle. SM 3.5每个周期支持64个64位操作。 Both SMs implement double precision operations in hardware. 两种SM均在硬件中实现双精度运算。
声明:本站的技术帖子网页，遵循CC BY-SA 4.0协议，如果您需要转载，请注明本站网址或者原文地址。任何问题请咨询:yoyou2525@163.com.