Possibly refactor CACHED_CONTAINER

[Simon Richter]

This appears to be a reimplementation of std::vector, for the most part.

* Why does it exist?
* Can it be removed?

[Maciej Suminski (orsonmmz)]

What is needed here is a capability to store continuous chunks of data (of various size, as every item may have a different demand), effectively creating a one big continuous block of memory that can be easily transferred to the GPU memory.

Now when I think about the code, perhaps we could replace the raw pointer (VERTEX_CONTAINER::m_vertices) with a std::vector. CACHED_CONTAINER keeps a list of free memory chunks that can be quickly used without frequent memcpying. I am not sure if it can be done automatically with std::vector, so I would rather keep that part in place.

If you can come up with a more efficient implementation, or just an idea how to improve speed - I will eagerly apply your suggestions.

[Simon Richter (sjr)]

Hi,

On 04.08.2015 09:54, Maciej Sumiński wrote:

> If you can come up with a more efficient implementation, or just an idea
> how to improve speed - I will eagerly apply your suggestions.

Sure, just won't happen before the release, as I'm quite busy --
that's why I made myself a bug report.

The only reason I stumbled across this code was the log2() call in the
grow function.

   Simon

[Maciej Suminski (orsonmmz)]

The report is not valid anymore. Current CACHED_CONTAINER implementation manages mapped video memory, which cannot be easily replaced with std::vector.