使用 Android Studio，我的代码将浮点数组渲染为传递给 GLSL 的纹理，每个纹理像素一个浮点数在 0 到 1 的范围内，就像灰度纹理一样。为此，我使用 GL_LUMINANCE 作为 glTexImage2D 的 internalFormat 和格式，使用 GL_FLOAT 作为类型。在 android 设备模拟器上运行应用程序可以正常工作（它使用我的 PC 的 GPU），但在真实设备（三星 Galaxy S7）上调用 glTexImage2D 会出现错误 1282，GL_INVALID_OPERATION。我认为这可能是两个纹理的非幂的问题，但宽度和高度肯定是二的幂。
该代码使用输出网格密度值的Jos Stam 流体模拟 C 代码（使用 NDK 编译，未移植）。
mSizeN 是流体模拟网格的宽度（与高度相同），虽然流体模拟为边界条件在其上加了 2，所以返回的数组的宽度为 mSizeN + 2；在这种情况下为 128。
坐标系设置为正投影，屏幕左上角为 0.0,0.0，右下角为 1.0,1.0。我只是绘制了一个全屏四边形，并使用 GLSL 中四边形的插值位置作为包含密度值的数组的纹理坐标。渲染它的好方法。
这是渲染器类。
public class GLFluidsimRenderer implements GLWallpaperService.Renderer {
private final String TAG = "GLFluidsimRenderer";
private FluidSolver mFluidSolver = new FluidSolver();
private float[] mProjectionMatrix = new float[16];
private final FloatBuffer mFullScreenQuadVertices;
private Context mActivityContext;
private int mProgramHandle;
private int mProjectionMatrixHandle;
private int mDensityArrayHandle;
private int mPositionHandle;
private int mGridSizeHandle;
private final int mBytesPerFloat = 4;
private final int mStrideBytes = 3 * mBytesPerFloat;
private final int mPositionOffset = 0;
private final int mPositionDataSize = 3;
private int mDensityTexId;
public static int mSizeN = 126;
public GLFluidsimRenderer(final Context activityContext) {
mActivityContext = activityContext;
final float[] fullScreenQuadVerticesData = {
0.0f, 0.0f, 0.0f,
0.0f, 1.0f, 0.0f,
1.0f, 0.0f, 0.0f,
1.0f, 1.0f, 0.0f,
};
mFullScreenQuadVertices = ByteBuffer.allocateDirect(fullScreenQuadVerticesData.length * mBytesPerFloat)
.order(ByteOrder.nativeOrder()).asFloatBuffer();
mFullScreenQuadVertices.put(fullScreenQuadVerticesData).position(0);
}
public void onTouchEvent(MotionEvent event) {
}
@Override
public void onSurfaceCreated(GL10 glUnused, EGLConfig config) {
GLES20.glDisable(GLES20.GL_DEPTH_TEST);
GLES20.glClearColor(0.5f, 0.5f, 0.5f, 0.5f);
String vertShader = AssetReader.getStringAsset(mActivityContext, "fluidVertShader");
String fragShader = AssetReader.getStringAsset(mActivityContext, "fluidFragDensityShader");
final int vertexShaderHandle = ShaderHelper.compileShader(GLES20.GL_VERTEX_SHADER, vertShader);
final int fragmentShaderHandle = ShaderHelper.compileShader(GLES20.GL_FRAGMENT_SHADER, fragShader);
mProgramHandle = ShaderHelper.createAndLinkProgram(vertexShaderHandle, fragmentShaderHandle,
new String[] {"a_Position"});
mDensityTexId = TextureHelper.loadTextureLumF(mActivityContext, null, mSizeN + 2, mSizeN + 2);
}
@Override
public void onSurfaceChanged(GL10 glUnused, int width, int height) {
mFluidSolver.init(width, height, mSizeN);
GLES20.glViewport(0, 0, width, height);
Matrix.setIdentityM(mProjectionMatrix, 0);
Matrix.orthoM(mProjectionMatrix, 0, 0.0f, 1.0f, 1.0f, 0.0f, 0.0f, 1.0f);
}
@Override
public void onDrawFrame(GL10 glUnused) {
GLES20.glClear(GLES20.GL_DEPTH_BUFFER_BIT | GLES20.GL_COLOR_BUFFER_BIT);
GLES20.glUseProgram(mProgramHandle);
mFluidSolver.step();
TextureHelper.updateTextureLumF(mFluidSolver.get_density(), mDensityTexId, mSizeN + 2, mSizeN + 2);
mProjectionMatrixHandle = GLES20.glGetUniformLocation(mProgramHandle, "u_ProjectionMatrix");
mDensityArrayHandle = GLES20.glGetUniformLocation(mProgramHandle, "u_aDensity");
mGridSizeHandle = GLES20.glGetUniformLocation(mProgramHandle, "u_GridSize");
mPositionHandle = GLES20.glGetAttribLocation(mProgramHandle, "a_Position");
double start = System.nanoTime();
drawQuad(mFullScreenQuadVertices);
double end = System.nanoTime();
}
private void drawQuad(final FloatBuffer aQuadBuffer) {
// Pass in the position information
aQuadBuffer.position(mPositionOffset);
GLES20.glVertexAttribPointer(mPositionHandle, mPositionDataSize, GLES20.GL_FLOAT, false,
mStrideBytes, aQuadBuffer);
GLES20.glEnableVertexAttribArray(mPositionHandle);
// Attach density array to texture unit 0
GLES20.glActiveTexture(GLES20.GL_TEXTURE0);
GLES20.glBindTexture(GLES20.GL_TEXTURE_2D, mDensityTexId);
GLES20.glUniform1i(mDensityArrayHandle, 0);
// Pass in the actual size of the grid.
GLES20.glUniform1i(mGridSizeHandle, mSizeN + 2);
GLES20.glUniformMatrix4fv(mProjectionMatrixHandle, 1, false, mProjectionMatrix, 0);
GLES20.glDrawArrays(GLES20.GL_TRIANGLE_STRIP, 0, 4);
}
}
这是纹理辅助函数。
public static int loadTextureLumF(final Context context, final float[] data, final int width, final int height) {
final int[] textureHandle = new int[1];
GLES20.glGenTextures(1, textureHandle, 0);
if (textureHandle[0] != 0) {
GLES20.glBindTexture(GLES20.GL_TEXTURE_2D, textureHandle[0]);
GLES20.glTexParameteri(GLES20.GL_TEXTURE_2D, GLES20.GL_TEXTURE_MIN_FILTER, GLES20.GL_NEAREST);
GLES20.glTexParameteri(GLES20.GL_TEXTURE_2D, GLES20.GL_TEXTURE_MAG_FILTER, GLES20.GL_NEAREST);
GLES20.glTexParameteri(GLES20.GL_TEXTURE_2D, GLES20.GL_TEXTURE_WRAP_S, GLES20.GL_CLAMP_TO_EDGE);
GLES20.glTexParameteri(GLES20.GL_TEXTURE_2D, GLES20.GL_TEXTURE_WRAP_T, GLES20.GL_CLAMP_TO_EDGE);
GLES20.glPixelStorei(GLES20.GL_UNPACK_ALIGNMENT, 1);
GLES20.glPixelStorei(GLES20.GL_PACK_ALIGNMENT, 1);
GLES20.glTexImage2D(GLES20.GL_TEXTURE_2D, 0, GLES20.GL_LUMINANCE,
(int) width, (int) height, 0, GLES20.GL_LUMINANCE, GLES20.GL_FLOAT,
(data != null ? FloatBuffer.wrap(data) : null));
}
if (textureHandle[0] == 0)
throw new RuntimeException("Error loading texture.");
return textureHandle[0];
}
public static void updateTextureLumF(final float[] data, final int texId, final int w, final int h) {
GLES20.glBindTexture(GLES20.GL_TEXTURE_2D, texId);
GLES20.glTexSubImage2D(GLES20.GL_TEXTURE_2D, 0, 0, 0, (int)w, (int)h, GLES20.GL_LUMINANCE, GLES20.GL_FLOAT, (data != null ? FloatBuffer.wrap(data) : null));
}
片段着色器。
precision mediump float;
uniform sampler2D u_aDensity;
uniform int u_GridSize;
varying vec4 v_Color;
varying vec4 v_Position;
void main()
{
gl_FragColor = texture2D(u_aDensity, vec2(v_Position.x, v_Position.y));
}
OpenGL ES 2 不支持 GL_FLOAT 和 GL_LUMINANCE 的组合吗？
安卓模拟器图片。
编辑：补充一下，我是否正确地说，当使用 glTexImage2D （等）传输时，每个浮点值将减少为 8 位整数分量，所以大部分浮点精度会丢失？在这种情况下，最好重新考虑模拟器的实现以输出定点。这很容易做到，Stam 甚至在他的论文中进行了描述。