rygorous

        // Dasher: refine check to allow stencil only early Z/stencil
        if (((prcs->ControlDWORD & CONTROL_DWORD_INTERPOLATED_USED_FOR_Z) || !prcs->DepthState.ZEnable) &&
-            !(pCompiled->MiscState.AlphaToCoverageEnable ||
-              pCompiled->MiscState.AlphaTestEnable ||
-              OMSHADER_COMPILEDOBJ_I::PSKillsPixel(prcs->pShader)))
+            !((prcs->DepthState.ZWrite || prcs->StencilState.FrontFace.StencilEnable || prcs->StencilState.BackFace.StencilEnable) &&
+                 (pCompiled->MiscState.AlphaTestEnable ||
+                  pCompiled->MiscState.AlphaToCoverageEnable ||
+                  OMSHADER_COMPILEDOBJ_I::PSKillsPixel(prcs->pShader))))
        {

rygorous

  <redacted1>-05: -0.6%
 
Change 27056 by mikeab@mikeapc on 2010/06/22 15:40:14
 
 Fabian's clever trick to fool the compiler into not doing movzx.
 

rygorous

Weird rendering problem:

We need to render a 3D object such that the z values getting passed on to depth test/write for all pixels
are all exactly the same value (constant per batch), and we need to be able to choose that value freely.

This is what we'd like to do, but it doesn't work:

  // at the end of the VS
  out.pos.z = ourZValue * out.pos.w;

rygorous

From: Casey Muratori, 10/17/12

Well, you will notice that that loop _does not_ stat unless it's
running on a filesystem that doesn't support readdir dtypes (because
it needs to know whether or not to recurse, which you can do with the
dtype on ext*, but not on NTFS et al, since they didn't bother to
implement it or something - everything comes back as DT_UNKNOWN).  I
haven't looked at the code to find out if/where they actually do stat
as a matter of course elsewhere, but as I mentioned last night, it is
_only the stat_ that causes the catastrophic case.  In my experiments,

rygorous

///////////////////////////////////////////////////////////////////////////////
//
//  Extensions (we map to GL 2.0 function names for a uniform interface
//  across platforms)
//

#define GDRAW_GL_EXTENSION_LIST \
   /*  identifier                      import                              procname */ \
   /* GL_ARB_multitexture */ \
   GLE(ActiveTexture,                  "ActiveTextureARB",                 ACTIVETEXTUREARB) \

rygorous

First mail I ever got from Jeff:

----

From: Jeff Roberts
Subject: dude!
To: Fabian "ryg" Giesen
Date: 5/17/2009 2:42PM

Hey, man - I don't think we have ever talked before directly!

rygorous

With Tundra 1.2 I get:

----

C:\Store\Code\tundra-1.2a-win32\examples\hello-world>..\..\tundra.exe
Build script execution failed
[string "tundra.tools.msvc-vs2008"]:31: RegOpenKeyExA: The system cannot find the file specified.

stack traceback:
        [C]: in function 'assert'

rygorous

Okay, here's the different op splitting/fusing strategies for different cores, as far as I've been able to discern them:

• Pentium: Complex instructions are U-pipe only but execute directly, they don't get split.
• Atom (Bonnel/Saltwell): Certain complex instructions don't get split.
• Pentium Pro/2/3: All ops get split into, tracked as, and executed as uOps.
• Pentium 4: This never happened.
• Pentium M/Core: All ops get split into uOps. Post-split, the core can fuse two types of multi-uOp sequences into a larger fused op used for tracking:

• For stores, address generation + actual store uOps can get fused.

rygorous

#include <stdio.h>
#include <float.h>
#include <math.h>

struct Vec3f
{
    float v[3];

    Vec3f() {}
    Vec3f(float x, float y, float z)

rygorous

// This function finds a pair of approximations
// [c1 -s1]       [c2 -s2]
// [s1  c1]  and  [s2  c2]
// to given planar rotations subject to the following constraints:
// 1. c1, s1, c2, s2 are all integer
// 2. c1*c1 + s1*s1 == c2*c2 + s2*s2, i.e. the approximations have the same norm.
// This is done while simultaneously trying to minimize:
// 1. The relative errors of the approximations
// 2. The size of the integers in question (smaller is better)
// 3. The cost of computing this approximation when no integer multiplies are available.