Clearing everything on a new map?

[Mexicouger]

I want to clear all models/textures/sprites/gfx on Start of a new map. How and where would I go about doing this?

I am up against a wall here. I just added in tga texture support into my engine, but when I goto reload a map, and there is an external texture, the game locks up.

I refuse to believe that this is an out-of-ram problem, because I can access other parts of the menu that seemed to crash when i was low on ram. Not only that, but the tga textures have Swizzle and DXT compression, so there isn't any reason for lockups.

I think I spent a bit too much time trying to get this all fixed and working. I didn't want to resort to asking for help, but heck, I'm drained of solutions here.

So all in all, How would I flush out external textures on the start of a new map. I did see Bakers tutorial on it, but It doesn't suffice for my case, and I fail to find where I would put that function.

Help would be greatly appreciated

[Baker]

There "where" goes like this ...

cl_main.c

Code: Select all

/*
================
Host_ClearMemory

This clears all the memory used by both the client and server, but does
not reinitialize anything.
================
*/
void Host_ClearMemory (void)
{
	Con_DPrintf ("Clearing memory\n");
	D_FlushCaches ();
	Mod_ClearAll ();
	if (host_hunklevel)
		Hunk_FreeToLowMark (host_hunklevel);

	cls.signon = 0;
	memset (&sv, 0, sizeof(sv));
	memset (&cl, 0, sizeof(cl));
}

Kurok unloads textures in ...

Code: Select all

/*
===================
Mod_ClearAll
===================
*/
void Mod_ClearAll (void)
{
	int		i;
	int		texture_index;
	model_t	*mod;

	for (i=0 , mod=mod_known ; i<mod_numknown ; i++, mod++)
	{
		if (mod->type != mod_alias)
			mod->needload = qtrue;
	}

	// maybe we should check if it is new map or not
	// (so we dont unload textures unnecessary)
	while (mapTextureNameList.size() > 0) {
		texture_index = mapTextureNameList.front();
		mapTextureNameList.pop_front();
		GL_UnloadTexture(texture_index);
	}

	solidskytexture	= -1;
	alphaskytexture	= -1;

	//purge old sky textures
	for (i=0; i<6; i++)
	{
	    if (skyimage[i] && skyimage[i] != solidskytexture)
            GL_UnloadTexture(skyimage[i]);
		skyimage[i] = NULL;
	}

    //purge old lightmaps
	for (i=0; i<MAX_LIGHTMAPS; i++)
	{
	    if (lightmap_index[i] && lightmap_index[i] != 0)
            GL_UnloadTexture(lightmap_index[i]);
		lightmap_index[i] = NULL;
	}
}

Which if you look at Host_ClearMemory, calls Mod_ClearAll.

Mod_ClearAll occurs at the start of any map. But if look through CL_ParseServerInfo in cl_parse.c ... it keeps models cached.

Enhanced GLQuake modified Mod_ClearAll to accept a -freemdl command line parameter which would also clear the models out ...

Code: Select all

/*
===================
Mod_ClearAll
===================
*/
void Mod_ClearAll (void)
{
	int		i;
	model_t		*mod;
	static qboolean NoFree, Done;

	if (!Done)
	{
		// Some 3dfx miniGLs don't support glDeleteTextures (i.e. do nothing)
		NoFree = COM_CheckParm ("-nofreetex");
		
		FreeMdl = COM_CheckParm ("-freemdl"); // Free also mdls
		Done = true;
	}

	for (i=0 , mod=mod_known ; i<mod_numknown ; i++, mod++)
	{
		if (mod->type != mod_alias || FreeMdl)
			mod->needload = true;
		
		if (mod->type == mod_alias)
		{
			if (FreeMdl && mod->cache.data)
				Cache_Free(&mod->cache); // Release memory buffer

			mod->loadtimes = 0; // Clear inbetween maps
		}

		Mod_Free (mod);
	}

	if (!NoFree)
		GL_FreeTextures ();
}

FitzQuake has a modification that deletes the textures when the model is freed ...

Code: Select all

/*
==============
Cache_Free

Frees the memory and removes it from the LRU list
==============
*/
void Cache_Free (cache_user_t *c, qboolean freetextures) //johnfitz -- added second argument
{
	cache_system_t	*cs;

	if (!c->data)
		Sys_Error ("Cache_Free: not allocated");

	cs = ((cache_system_t *)c->data) - 1;

	cs->prev->next = cs->next;
	cs->next->prev = cs->prev;
	cs->next = cs->prev = NULL;

	c->data = NULL;

	Cache_UnlinkLRU (cs);

	//johnfitz -- if a model becomes uncached, free the gltextures.  This only works
	//becuase the cache_user_t is the last component of the model_t struct.  Should
	//fail harmlessly if *c is actually part of an sfx_t struct.  I FEEL DIRTY
	if (freetextures)
		TexMgr_FreeTexturesForOwner ((model_t *)(c + 1) - 1);
}

[Mexicouger]

Thanks Baker. I have browsed through Mod_ClearAll and modified it a bit, but I didn't have that much of luck. I'll see what I can do with this new info.

Thanks

[Baker]

Thanks Baker. I have browsed through Mod_ClearAll and modified it a bit, but I didn't have that much of luck. I'll see what I can do with this new info.

Thanks

It isn't going to be an easy task, I hate to say. None of the PSP engines have a texture manager. You will have to invest probably too much time to mark the textures and figure out how to only unload model textures.

[ceriux]

why not just give all of your mapping textures prefix's and if the texture has a prefix dont unload it that way only model textures are unloaded?

[Mexicouger]

Well the only Reason I am bound and determined, is because DQuake managed to do it, so I know it is possible.

Thats my only Push right now, and well, the fact that the game will look better in the end.

[Downsider]

Use an std::list? Most of the platform specific stuff is C++ anyway.

[Mexicouger]

On the DQuake Thread, I noticed that Downsider mentioned that Some libraries were needed to Compile DQuake. Do you think these libraries could be affecting the texture loading?

I don't get any linker errors without them, but I am thinking that maybe they are affecting the game.
Just some thoughts, because I can't think/find any other problems with the code

Edit: I am trying out different vram and valloc libraries, and I am definitely getting different results. With one of them, I am getting a constant screen flashing, and my tga textures are masked with purple

EDIT AGAIN: I finally got my Tgas working! I been working on them for a constant 4 days now, and I gotta say that Breadsticks post about DQuake Helped me more than anything. He listed the libraries that were required, and he mentioned a vram library. I did a search, found it, and applied it. First time I converted the functions and all, and I got some horrendous effects. I restarted(For the 7th time) and went through the code with a fine comb. To my success, I got tgas working 100%!

I think this will be one of my biggest accomplishments for a while. Oh how much I learned from doing this. I am now aware of where things are, and what alot more does.

Knowledge is great, just can't get enough of it!