CUDA 纹理内存中支持双精度类型
计算能力1.3的GTX 280中可以为双精度型变量声明2D纹理吗 仅针对 float 给出声明
Can 2D textures be declared for double type variables in GTX 280 having compute capability 1.3
The declaration is given only for float
如果你对这篇内容有疑问,欢迎到本站社区发帖提问 参与讨论,获取更多帮助,或者扫码二维码加入 Web 技术交流群。
绑定邮箱获取回复消息
由于您还没有绑定你的真实邮箱,如果其他用户或者作者回复了您的评论,将不能在第一时间通知您!
发布评论
评论(2)
据我所知,你不能。但是,您可以使用 int2 类型,如第 23 点所示:http://forums。 nvidia.com/index.php?showtopic=84440
编辑:更正语法
For as far as I'm aware, you can't. However, you can use the int2 type as shown in point 23 here: http://forums.nvidia.com/index.php?showtopic=84440
edit: Corrected grammar
你不能用双打进行纹理插值,硬件不支持它。您可以做的是将双精度数存储为 int2,然后使用 tex2dfetch 收集相关点,转换为双精度数,然后在软件中进行实际插值。
You can't do texture interpolation with doubles, the hardware doesn't support it. What you can do is store the doubles as int2 and then use the tex2dfetch to collect the relevant points, convert to double, then do the actual interpolation in software.