java - matrix.multiply 返回 Nan

Question

我知道 NaN 的意思不是数字，但我不明白我是如何得到这个错误的。我正在做的是引用来 self 的形状类的我的形状对象的顶点值，并在我的渲染器类中使用它们。当我在 logcat 上打印出顶点的值时，我看到它返回前 5 个顶点，然后所有其他顶点都是 NaN。如果您查看我的 logcat，则表明在调用 matrix.multiplyMM() 之前，顶点 (vertex1) 打印良好。但是在调用之后，顶点(shapeverts)开始变成 NaN。

for (int i = 0; i < Shape.vertices.length; i += Shape.COORDS_PER_VERTEX) {//only checking one vertex
        Log.i("vertex1", String.valueOf(Shape.vertices[i] + ", " + Shape.vertices[i+1] + ", " + Shape.vertices[i+2] + ", " + Shape.vertices[i+3]));

        Matrix.multiplyMM(shapeVerts, 0, mvp, 0, Shape.vertices, 0);//vertices multiplied by model view projection matrix

        Log.i("shapevertBf", String.valueOf(shapeVerts[i] + ", " + shapeVerts[i+1] + ", " + shapeVerts[i+2] + ", " + shapeVerts[i+3]));

        shapeVerts[i] = shapeVerts[i] / shapeVerts[i + 3];    //clip.x divided by clip.w
        shapeVerts[i + 1] = shapeVerts[i + 1] / shapeVerts[i + 3];//clip.y divided by clip.w
        shapeVerts[i + 2] = shapeVerts[i + 2] / shapeVerts[i + 3];//clip.y divided by clip.w

        Log.i("shapevertAf", String.valueOf(shapeVerts[i] + ", " + shapeVerts[i+1] + ", " + shapeVerts[i+2] + ", " + shapeVerts[i+3]));

日志猫

  06-10 13:59:27.983  13121-13189/com.example.james.rollingsphere I/vertex1﹕ 0.0, 0.0, 9.0, 1.0
06-10 13:59:27.993  13121-13189/com.example.james.rollingsphere I/shapevertBf﹕ 0.0, 0.0, 19.5, 12.0
06-10 13:59:27.993  13121-13189/com.example.james.rollingsphere I/shapevertAf﹕ 0.0, 0.0, 1.625, 12.0
06-10 13:59:27.993  13121-13189/com.example.james.rollingsphere I/vertex1﹕ 0.0, 0.0, 9.0, 1.0
06-10 13:59:27.993  13121-13189/com.example.james.rollingsphere I/shapevertBf﹕ 0.0, 0.0, 19.5, 12.0
06-10 13:59:27.993  13121-13189/com.example.james.rollingsphere I/shapevertAf﹕ 0.0, 0.0, 1.625, 12.0
06-10 13:59:27.993  13121-13189/com.example.james.rollingsphere I/vertex1﹕ 0.0, 0.0, 9.0, 1.0
06-10 13:59:27.993  13121-13189/com.example.james.rollingsphere I/shapevertBf﹕ 0.0, 0.0, 19.5, 12.0
06-10 13:59:27.993  13121-13189/com.example.james.rollingsphere I/shapevertAf﹕ 0.0, 0.0, 1.625, 12.0
06-10 13:59:27.993  13121-13189/com.example.james.rollingsphere I/vertex1﹕ 0.0, 0.0, 9.0, 1.0
06-10 13:59:27.993  13121-13189/com.example.james.rollingsphere I/shapevertBf﹕ 0.0, 0.0, 19.5, 12.0
06-10 13:59:27.993  13121-13189/com.example.james.rollingsphere I/shapevertAf﹕ 0.0, 0.0, 1.625, 12.0
06-10 13:59:28.003  13121-13189/com.example.james.rollingsphere I/vertex1﹕ 0.0, 0.0, 9.0, 1.0
06-10 13:59:28.003  13121-13189/com.example.james.rollingsphere I/shapevertBf﹕ 0.0, 0.0, 0.0, 0.0
06-10 13:59:28.003  13121-13189/com.example.james.rollingsphere I/shapevertAf﹕ NaN, NaN, NaN, 0.0
06-10 13:59:28.003  13121-13189/com.example.james.rollingsphere I/vertex1﹕ 5.290067, 0.0, 7.281153, 1.0
06-10 13:59:28.003  13121-13189/com.example.james.rollingsphere I/shapevertBf﹕ 0.0, 0.0, 0.0, 0.0
06-10 13:59:28.003  13121-13189/com.example.james.rollingsphere I/shapevertAf﹕ NaN, NaN, NaN, 0.0
06-10 13:59:28.003  13121-13189/com.example.james.rollingsphere I/vertex1﹕ 1.634721, 5.031153, 7.281153, 1.0
06-10 13:59:28.003  13121-13189/com.example.james.rollingsphere I/shapevertBf﹕ 0.0, 0.0, 0.0, 0.0
06-10 13:59:28.003  13121-13189/com.example.james.rollingsphere I/shapevertAf﹕ NaN, NaN, NaN, 0.0
06-10 13:59:28.003  13121-13189/com.example.james.rollingsphere I/vertex1﹕ -4.279755, 3.109423, 7.281153, 1.0
06-10 13:59:28.013  13121-13189/com.example.james.rollingsphere I/shapevertBf﹕ 0.0, 0.0, 0.0, 0.0
06-10 13:59:28.013  13121-13189/com.example.james.rollingsphere I/shapevertAf﹕ NaN, NaN, NaN, 0.0

编辑我更新了代码和 logcat。有趣的是，w 分量在 Matrix.multiply 之后从 12 开始，但随后变为 0。

Answer 1

问题是在这种情况下使用 Matrix.multiplyMM 没有意义。 multiplyMM 用于将两个矩阵相乘以获得矩阵乘积，而您想要做的是将一个矩阵(模型 View 投影矩阵)乘以一个 vector (您在 Shape.js 中未转换的顶点)。 vertices) 得到一个 vector (你在 shapeVerts 中转换的顶点)。

您只获得 5 个结果的原因是您将顶点数组视为一个 4x4 矩阵，而实际上它是一个顶点数组，每个顶点数组包含 3 个元素。一个 4x4 矩阵有 16 个元素，占用的空间与 5 1/3 顶点相同，因此您可以在到达未初始化数据并开始获取零(导致 NaN)等之前读出 5 个(无效)值。

要解决此问题，请使用 Matrix.multiplyMV将您的顶点坐标乘以模型 View 投影矩阵，将偏移量传递到正确的顶点，如下所示:

Matrix.multiplyMV(shapeVerts, i, mvp, 0, Shape.vertices, i);