1.首先并不需要每个texture对应一个pipeline，可以在同一pipeline中使用不同的BindGroup，每个group对应不同的texture binding. e.g.

let sampler = device.createSampler({
    magFilter: 'linear',
    minFilter: 'linear',
});
let texture1 = device.createTexture({ ... });
let texture2 = device.createTexture({ ... });
const group1 = device.createBindGroup({
    layout: pipeline.getBindGroupLayout(0),
    entries: [
        {
            binding: 1,
            resource: sampler,
        },
        {
            binding: 2,
            resource: texture1.createView(),
        }
    ]
});
const group2 = device.createBindGroup({
    layout: pipeline.getBindGroupLayout(0),
    entries: [
        {
            binding: 1,
            resource: sampler,
        },
        {
            binding: 2,
            resource: texture2.createView(),
        }
    ]
});

然后在 rendering loop 中通过 setBindGroup 切换 group1 和 group 2 即可。

2.如果一组 textures 是同样的大小和格式，可以直接使用 texture_2d_array 去存储多个 image，这样即使一个group内，也可以在shader中通过切换array的index，来直接切换 texutre.

对于动态的 video texuture，我们可以直接使用 importExternalTexture 进行引入，随着video播放，texture 会自动更新

const video = document.createElement('video');
video.loop = true;
video.autoplay = true;
video.muted = true;
video.src = '...';
await video.play();

const group = device.createBindGroup({
    layout: pipeline.getBindGroupLayout(0),
    entries: [
        {
            binding: 1,
            resource: sampler,
        },
        {
            binding: 2,
            resource: device.importExternalTexture({
                source: video,
            })
        }
    ]
});

3.对于更新静态的 texutre，可以根据目标 image/texture 的类型进行 copy 更新，目前 webgpu 提供多个copy command，我们一般最主要会用到的：copyExternalImageToTexture ，可以更新外部image到gpu texture:

let texture = device.createTexture({ ... });
const img1 = document.createElement('img');
img1.src = '....';
await img1.decode();
const image1 = await createImageBitmap(img1);
device.queue.copyExternalImageToTexture(
      { source: image1 },
      { texture: texture },
      [image1.width, image1.height]
);
...
...js
// 更新texture的image
const img2 = document.createElement('img');
img2.src = '....';
await img2.decode();
const image2 = await createImageBitmap(img2);
device.queue.copyExternalImageToTexture(
      { source: image2 },
      { texture: texture },
      [image2.width, image2.height]
);

另外，也可以用 commandEncoder 执行 copyTextureToTexture 进行两个 gpu texutre 之间的copy 更新

// e.g. 创建两个texture，texture1 用于显示，tempTexture用于接收外部图片，图片loading 后进行 copy 更新
let texture1 = device.createTexture({ ... });
...
let tempTexture = device.createTexture({ ... });
let newImage = await loadImage() // 异步加载图片 
device.queue.copyExternalImageToTexture(
    { source: newImage },
    { texture: tempTexture },
    [newImage.width, newImage.height]
);
const commandEncoder = device.createCommandEncoder();
commandEncoder.copyTextureToTexture(
    { texture: tempTexture },
    { texture: texture1 }
);
device.queue.submit([commandEncoder.finish()]);

其他的更新texture API 使用方法，可以参阅 https://www.orillusion.com/webgpu.html#image-copies

4.对于 sampler，目前 webgpu 没有直接更新 sampler 的API，只能创建新的 sampler，并使用新的group进行渲染，e.g.

let linearSampler = device.createSampler({
    magFilter: 'linear',
    minFilter: 'linear',
});
const group = device.createBindGroup({
    layout: pipeline.getBindGroupLayout(0),
    entries: [
        {
            binding: 1,
            resource: linearSampler,
        },
        {
            binding: 2,
            resource: texture
        }
    ]
});
let nearestSampler = device.createSampler({
    magFilter: 'nearest',
    minFilter: 'nearest',
});
const newGroup = device.createBindGroup({
    layout: pipeline.getBindGroupLayout(0),
    entries: [
        {
            binding: 1,
            resource: nearestSampler,
        },
        {
            binding: 2,
            resource: texture
        }
    ]
});

group中其他的resource并不需要重新创建，可以复用

Orillusion

0

524

151

408

shuangliu