Re: [PATCH 0/17] MR541: RFC: Shader cache for vkd3d

[Giovanni Mascellani (＠giomasce)]

Giovanni Mascellani (@giomasce) commented about libs/vkd3d/cache.c:
> - return VKD3D_ERROR_NOT_IMPLEMENTED;
> - struct vkd3d_shader_cache *object;
> - size_t size;
> -
> - TRACE("%s, %p, %p.\n", debugstr_a(name), desc, cache);
> -
> - if (!name || !desc)
> - {
> -
>    WARN("No name or description, returning VKD3D_ERROR_INVALID_ARGUMENT.\n");
> -
>    return E_INVALIDARG;
> - }
> -
> - /* FIXME: This isn't thread safe and cache_mutex_initialized might overflow. Do we have a
> -
> * something like DllMain or a platform-independent InitializeOnce? */
> - if (InterlockedIncrement(&cache_mutex_initialized) == 1)
> -
>    vkd3d_mutex_init(&cache_list_mutex);

Both the platforms we support have static mutex initializers, we can just define them in the header. We can also add support for `InitializeOnce()` and `pthread_once()`, of course, but the static initializer seems better to me.

Also notice that we have platform-independent functions `vkd3d_atomic_increment()` and similar.