什么让glBufferSubData缓冲大缓冲区？(What makes glBufferSubData slow on big buffer?)

什么让glBufferSubData缓冲大缓冲区？(What makes glBufferSubData slow on big buffer?)

我注意到随着缓冲区大小的增加，在VBO中更新少量顶点的速度非常慢。 例如，当我更新时，例如512 * 512缓冲区中的2个垂直速度要慢得多，然而当我更新整个256 * 256缓冲区时，尽管更新的数据量完全相反。 这是glBufferSubData的正常行为吗？

这是我分配缓冲区的方式：

glBufferData(GL_ARRAY_BUFFER, sizeof(uiMapView::vertex)*(size + 1)*(size + 1), m_verts, GL_DYNAMIC_DRAW);

这是我更新它的方式：

glBufferSubData(GL_ARRAY_BUFFER, sizeof(uiMapView::vertex)*iMinId, sizeof(uiMapView::vertex)*(iMaxId-iMinId), (&m_verts[iMinId]));

I've noticed that updating a small amount of vertices in VBO goes very slow as buffer size grows. For example, when I update, say, 2 verts in a 512*512 buffer is much slower then when I update the whole 256*256 buffer despite the amounts of data updated are quite opposite. Is this a normal behavior for glBufferSubData?

This is how I allocate the buffer:

glBufferData(GL_ARRAY_BUFFER, sizeof(uiMapView::vertex)*(size + 1)*(size + 1), m_verts, GL_DYNAMIC_DRAW);

This is how I update it:

glBufferSubData(GL_ARRAY_BUFFER, sizeof(uiMapView::vertex)*iMinId, sizeof(uiMapView::vertex)*(iMaxId-iMinId), (&m_verts[iMinId]));

最满意答案

那么，我对我的解决方案做了更多的研究，发现VBO的更新并不是真正的瓶颈。 减少我的程序的原因是对数据进行的计算量，这显然是最大的影响。

所以，实际上，正确的anser是： 不，这不是glBufferSubData的正常行为 。 缓冲区大小不会对glBufferSubData性能产生如此巨大的影响。

Well, I made some more research on my solution and found out that VBO updating wasn't really a bottleneck. The reason of slowing my program down was an amount of calculations made with the data, wich obviously made the biggest impact.

So, actually, the right anser is: no, this is not a normal behaviour for glBufferSubData. The buffer size doesn't make such a great impact on glBufferSubData performance.