Re: D/L-ing textures in draw space

New Message Reply Date view Thread view Subject view Author view

Grambo (graham++at++beast)
Mon, 2 Jan 1995 13:12:10 -0800

Kent, what I usually do is try and borrow some debugged performer code from sone
example somewhere, in this case from perfly. You are right about having to
buffer up the textures in App and Draw space, hence just use shared memory.
 Here is the code fragment:

void DownloadTex(long n, pfTexture **texList, long drawFlag)
{
    long foo, ncomp, xs, ys, i;

    fprintf(stderr, "Downloading textures");

    if(!drawFlag)
    {
        for (i = 0; i < n; i++)
        {
            fprintf(stderr, ".");
            pfApplyTex(texList[i]);
        }
    }
    else
    {
        float xc, yc;
        static pfMatrix ident = {{1., 0., 0., 0.},
                                 {0., 1., 0., 0.},
                                 {0., 0., 1., 0.},
                                 {0., 0., 0., 1.}};

        pfPushState();
        pfEnable(PFEN_TEXTURE);
        pfDisable(PFEN_FOG);

        getsize(&xs, &ys);
        ortho2(-.5, xs-.5, -.5, ys-.5);
        viewport(0, xs-1, 0, ys-1);
        xc = xs / 2.;
        yc = ys / 2.;

        pushmatrix();
        loadmatrix(ident);

        for (i = 0; i < n; i++)
        {
            long tx, ty;
            pfVec2 v;
            static pfVec2 t[] = {{0., 0.,}, {1., 0.,}, {1., 1.,}, {0., 1.,}};

            fprintf(stderr, ".");

            pfGetTexImage(texList[i], (ulong**)&foo, &ncomp, &tx, &ty, &foo);
            pfApplyTex(texList[i]);

            cpack(0);
            clear();
            cpack(0xffffffff);

            bgnpolygon();
            t2f(t[0]);
            PFSET_VEC2(v, xc - tx/2., yc - ty/2.);
            v3f(v);
            t2f(t[1]);
            PFSET_VEC2(v, xc + tx/2., yc - ty/2.);
            v3f(v);
            t2f(t[2]);
            PFSET_VEC2(v, xc + tx/2., yc + ty/2.);
            v3f(v);
            t2f(t[3]);
            PFSET_VEC2(v, xc - tx/2., yc + ty/2.);
            v3f(v);
            endpolygon();

            swapbuffers();
        }

        popmatrix();

        pfPopState();
    }
    fprintf(stderr, "\n");
    fprintf(stderr, "Textures successfully down-loaded\n");
}

On Dec 31, 7:42am, Kent Watsen wrote:
> Subject: D/L-ing textures in draw space
>
> Hello,
>
> I've just recently parallelized my DIS simulation
> application , it took less then a day, but I still
> have one nasty issue to address...
>
> In the unparallelized version, I would load all geometry
> from disk at load time. This geometry included the
> terrain and an instance of each moving model likely to
> be needed. Also at this time I would d/l the textures
> associated with the geometry into hardware texture memory.
>
> When parallelized, the above code is considered to be in the
> app process space, but the d/l-ing of textures must be in
> draw process space.
>
> I have considered several approaches to this, but they all
> seem kludgy (having to use a first-time static variable to
> indicate the need to do the work). I figure that there must
> be some support or, at least, a commonly accepted approach
> to do this work efficiently.
>
> Anybody?
>
> Kent Watsen
> DCS Corporation
> Simulation Branch
> 703.683.8430 x369
>
>
>-- End of excerpt from Kent Watsen

-- 


"It is often easier to fight for one's principles than to
live up to them." , be good..


----------------------------------------------------------------------
Graham (Grambo) Beasley		Silicon Graphics, Inc.
MTS (Simulator Guy)		(415) 390-5420		graham++at++sgi.com
----------------------------------------------------------------------

New Message Reply Date view Thread view Subject view Author view

This archive was generated by hypermail 2.0b2 on Mon Aug 10 1998 - 17:50:48 PDT

This message has been cleansed for anti-spam protection. Replace '++at++' in any mail addresses with the '@' symbol.