CUDA Academy · 课时

将过滤器融合到一个内核中

减少启动次数与全局内存流量

将过滤器融合到一个内核中 是 CoddyKit 上的免费 CUDA Academy 课时。这是第 2 节课，共 4 节。你可以在下方免费阅读本课时的完整内容 — 然后在浏览器中使用内置代码编辑器和全天候 AI 导师进行实践。这是 CUDA Academy 学习路径的一部分，你的进度在网页和 CoddyKit 应用中同步。 CUDA Academy 课程共包含 4 节课。

本课时的部分内容尚未翻译，以英文显示。

Why Fuse At All

Running five kernels means five launches and five round-trips to global memory. Fusing filters into one kernel cuts both, often giving a big speedup. 🚀

Launch Overhead Adds Up

Every kernel launch costs a few microseconds. On a tiny image that fixed launch overhead can dwarf the real work, so fewer launches means more useful time.

Global Traffic Is the Enemy

Separate stages write a pixel to global memory then read it right back. Fusing keeps that value in a register, erasing the wasted round-trip entirely.

Load Once Per Thread

In a fused kernel each thread reads its pixel a single time. That one load then feeds every filter in sequence without touching global memory again.

float v = input[idx];

Chain Operations in Registers

Apply each filter to the value already in hand. The pixel flows through brightness, gamma, and contrast as plain math, all kept in fast registers.

v = v * brightness;
v = powf(v, gamma);
v = (v - 0.5f) * contrast + 0.5f;

Write Once At the End

After the whole chain runs, store the final result. One store replaces the many writes that separate kernels would have made.

output[idx] = v;

Pointwise Filters Fuse Cleanly

Filters that touch only one pixel are pointwise and fuse with zero fuss. Brightness, gamma, and color tweaks are the easiest wins to combine.

Neighborhood Filters Are Harder

A blur reads nearby pixels, so fusing it needs shared memory tiles, not just registers. Fuse pointwise stages freely and treat stencils with extra care.

Watch Register Pressure

A big fused kernel uses more registers per thread. Too many and occupancy drops or values spill, so fuse aggressively but keep an eye on the cost.

Verify After Fusing

Fusing reorders work, so always compare the fused output against the staged version. A quick diff confirms the math still matches before you celebrate.

One Kernel, Many Filters

The payoff is real: a single launch reads, transforms, and writes each pixel exactly once. That is the heart of a well-tuned fused image kernel.

Quick Check

You merged three pointwise filters into one kernel. What is the main performance win?

Recap

You learned to fuse filters: load once, chain pointwise math in registers, write once. It cuts launches and global traffic, but watch register pressure and verify the output. ✅

常见问题解答

「将过滤器融合到一个内核中」课时是免费的吗？

是的 — 「将过滤器融合到一个内核中」的完整文本可在网页上免费阅读。要进行交互式练习（内置代码编辑器和全天候 AI 导师）并解锁 CUDA Academy 课程的其余内容，请升级到 CoddyKit PRO。 CUDA Academy 课程共包含 4 节课。

「将过滤器融合到一个内核中」这节课中我会学到什么？

减少启动次数与全局内存流量你通过在浏览器中直接运行的动手代码来练习 CUDA Academy，全天候 AI 导师会在你学习这节课的过程中回答你的问题。

学习 CUDA Academy 需要有经验吗？

无需任何先前经验。CoddyKit 上的 CUDA Academy 课程适合初学者到高级学习者，你可以从这里开始或从头开始，按照自己的节奏学习。这是第 2 节课，共 4 节。

「将过滤器融合到一个内核中」课时需要多长时间？

大多数 CoddyKit 课程大约需要 5–10 分钟。每节课都很精短且互动，所以你能稳步进步，并在网页和应用中从离开的地方继续。

我能在这节 CUDA Academy 课中编写并运行代码吗？

能。每节 CUDA Academy 课都包含内置代码编辑器，你可以在浏览器中直接编写并运行真实代码，并获得即时 AI 反馈 — 无需本地设置。

此课程中的所有课时

← 返回 CUDA Academy