轻量封装WebGPU渲染系统示例<33>- 单精度浮点纹理(源码)

在WebGPU中创建纹理使用纹理很方便,只是js中只有Float32Array而默认不支持Float16Array,所以略微费点事。不过网上的大神多的是,摇摇小手就能获得解决方案。

废话多了容易挨胖揍,看代码。

js中float16单精度float数值转换:

javascript 复制代码
// thanks: https://esdiscuss.org/topic/float16array
const toFloat16 = (function() {
	var floatView = new Float32Array(1);
	var int32View = new Int32Array(floatView.buffer);
	/* This method is faster than the OpenEXR implementation (very often
	 * used, eg. in Ogre), with the additional benefit of rounding, inspired
	 * by James Tursa?s half-precision code. */
	return function toHalf(val: number) {
		floatView[0] = val;
		var x = int32View[0];

		var bits = (x >> 16) & 0x8000; /* Get the sign */
		var m = (x >> 12) & 0x07ff; /* Keep one extra bit for rounding */
		var e = (x >> 23) & 0xff; /* Using int is faster here */

		/* If zero, or denormal, or exponent underflows too much for a denormal
		 * half, return signed zero. */
		if (e < 103) {
			return bits;
		}

		/* If NaN, return NaN. If Inf or exponent overflow, return Inf. */
		if (e > 142) {
			bits |= 0x7c00;
			/* If exponent was 0xff and one mantissa bit was set, it means NaN,
			 * not Inf, so make sure we set one mantissa bit too. */
			bits |= (e == 255 ? 0 : 1) && x & 0x007fffff;
			return bits;
		}

		/* If exponent underflows but not too much, return a denormal */
		if (e < 113) {
			m |= 0x0800;
			/* Extra rounding may overflow and set mantissa to 0 and exponent
			 * to 1, which is OK. */
			bits |= (m >> (114 - e)) + ((m >> (113 - e)) & 1);
			return bits;
		}

		bits |= ((e - 112) << 10) | (m >> 1);
		/* Extra rounding. An overflow will set mantissa to 0 and increment
		 * the exponent, which is OK. */
		bits += m & 1;
		return bits;
	};
})();

基于上述数值转换,实现单精度浮点数纹理(rgba16float格式纹理):

javascript 复制代码
private createFloat16Texture(width: number, height: number): GPUTexture {

		let data = new Uint16Array(width * height * 4);
		let scale = 1.0;
		let k = 0;
		for (let i = 0; i < height; ++i) {
			for (let j = 0; j < width; ++j) {
				k = (width * i + j) * 4;
				data[k] = toFloat16(scale * (j/width));
				data[k+1] = toFloat16(scale * (0.5 + 0.5 * Math.sin(10.0 * (1.0 - j/width))));
				data[k+2] = toFloat16(scale * (1.0 - (i * j)/(width * height)));
				data[k+3] = toFloat16(scale * 1.0);
			}
		}

		const texture = device.createTexture({
			size: { width, height },
			format: "rgba16float",
			usage: GPUTextureUsage.TEXTURE_BINDING | GPUTextureUsage.COPY_DST
		});
        // 下面的8表示一个像素需要4个float16也就是8个字节的内存来读写
		device.queue.writeTexture({ texture }, data, {bytesPerRow: width * 8, rowsPerImage: height}, { width, height });
		return texture;
	}

上述代码呈现的纹理效果:

此引擎系统当前示例RGBA8纹理和Float16纹理运行效果:

此示例基于此渲染系统实现,当前示例TypeScript源码如下:

javascript 复制代码
export class DataTextureTest {
	private mRscene = new RendererScene();

	initialize(): void {

		this.initScene();
	}
	private applyRGBAFloat16Tex(): void {
		let rc = this.mRscene;

		let width = 256;
		let height = 256;

		let dataFs32 = new Float32Array(width * height * 4);
		let scale = 10.0;
		let k = 0;
		for (let i = 0; i < height; ++i) {
			for (let j = 0; j < width; ++j) {
				k = (width * i + j) * 4;
				dataFs32[k] = scale * (j / width);
				dataFs32[k + 1] = scale * (0.5 + 0.5 * Math.sin(10.0 * (1.0 - j / width)));
				dataFs32[k + 2] = scale * (1.0 - (i * j) / (width * height));
				dataFs32[k + 3] = scale * 1.0;
			}
		}
		const tex = {
			diffuse: { uuid: "rtt0", dataTexture: { data: dataFs32, width, height }, format: "rgba16float", generateMipmaps: true }
		};

		let entity = new FixScreenPlaneEntity({ extent: [-0.8, -0.8, 0.8, 0.8], textures: [tex] });
		entity.color = [0.1, 0.1, 0.1, 0.1];
		rc.addEntity(entity);
	}

	private applyRGBA8Tex(): void {
		let rc = this.mRscene;

		let width = 256;
		let height = 256;

		let dataU8 = new Uint8Array(width * height * 4);
		let k = 0;
		for (let i = 0; i < height; ++i) {
			for (let j = 0; j < width; ++j) {
				k = (width * i + j) * 4;
				dataU8[k] = ((j / width) * 255) | 0;
				dataU8[k + 1] = ((0.5 + 0.5 * Math.sin(10.0 * (1.0 - j / width))) * 255) | 0;
				dataU8[k + 2] = ((1.0 - (i * j) / (width * height)) * 255) | 0;
				dataU8[k + 3] = 255;
			}
		}

		let tex = {
			diffuse: { uuid: "rtt1", dataTexture: { data: dataU8, width, height }, format: "rgba8unorm", generateMipmaps: true }
		};

		let entity = new FixScreenPlaneEntity({ extent: [0.0, 0.0, 0.8, 0.8], textures: [tex] });
		rc.addEntity(entity);
	}
	private initScene(): void {
		this.applyRGBAFloat16Tex();
		this.applyRGBA8Tex();
	}

	run(): void {
		this.mRscene.run();
	}
}

注: 这里的纹理应用基于系统的对应功能的封装实现。

相关推荐
zhongqu_3dnest18 分钟前
3D可视化:开启多维洞察新时代
3d·3d建模·空间计算·3d可视化·三维空间·沉浸式体验
试着13 小时前
【数据标注师】3D标注
3d·数据标注师·3d标注
工业3D_大熊10 天前
3D模式格式转换工具HOOPS Exchange如何将3D PDF转换为STEP格式?
3d·pdf·3d格式转换·3d模型格式转换·cad格式转换·cad数据格式转换·3d模型可视化
广州华锐视点10 天前
浅议 3D 展示技术为线上车展新体验带来的助力
3d
大霸王龙11 天前
AR眼镜与3D建模社区建设
3d·ar
杀生丸学AI12 天前
【物理重建】SPLART:基于3D高斯泼溅的铰链估计与部件级重建
3d·aigc·三维重建·视觉大模型·世界模型·空间智能·动态重建
Love__Tay12 天前
【Python小练习】3D散点图
开发语言·python·3d
海伯森技术12 天前
海伯森3D闪测传感器,工业检测领域的高精度利器
3d
:mnong12 天前
三维空间数据格式全景图:OSGB | 3Dtiles | I3S | S3M
3d