New subject: [1/7] WineD3D: Don't single-allocate new gl shaders

19 Jan 2009

      2009/1/19 Stefan Dösinger stefan@codeweavers.com:
...

       shader->shader_array_size += max(1, old_size * 0.5);

I don't know for sure if the compiler is able to get rid of the
floating point multiplication, but wouldn't a right shift or divide by
2 work at least as well?
...

       shader->gl_shaders = HeapReAlloc(GetProcessHeap(), 0, old_array,

                                        (shader->shader_array_size + 1) * sizeof(*shader->gl_shaders));

This looks more complicated than it has to be. It's easier to just
change "old_size <= shader->num_gl_shaders" into "old_size <
shader->num_gl_shaders" and drop the +1 here. Also note that
"old_array" isn't particularly useful here, you overwrite
shader->gl_shaders regardless of whether the allocation fails or
succeeds.
The typical way to do this would be:
if (shader->num_gl_shaders >= shader->shader_array_size)
{
    UINT new_size;
    struct ps_compiled_shader *new_array;
if (shader->gl_shaders)
    {
        new_size = shader->shader_array_size << 1;
        new_array = HeapReAlloc(GetProcessHeap(), 0,
shader->gl_shaders, new_size * sizeof(*new_array));
    }
    else
    {
        new_size = 1;
        new_array = HeapAlloc(GetProcessHeap(), 0, new_size *
sizeof(*new_array));
    }
if (!new_array)
    {
        ERR("Out of memory\n");
        return 0;
    }
shader->shader_array_size = new_size;
    shader->gl_shaders = new_array;
}
This applies to 4/7 as well.

Re: [1/7] WineD3D: Don't single-allocate new gl shaders