zeux

--!strict

--[[
BSD Zero Clause License

Copyright (c) 2022 Arseny Kapoulkine

Permission to use, copy, modify, and/or distribute this software for any
purpose with or without fee is hereby granted.

zeux

State of Roblox graphics API across all platforms, with percentage deltas since EOY 2020. Updated December 31 2021.

Windows

API	Share
Direct3D 11+	92% (+3%)
Direct3D 10.1	5% (-2%)
Direct3D 10.0	3% (-0.5%)

zeux

Algorithms used for Cone* preprocess the mesh in some way, then split sequentially into 64-triangle clusters:
ConeBase: optimize mesh for transform cache
ConeSort: split mesh into large planar connected clusters, bin clusters into 6 buckets by cardinal axes, optimize each bucket for transform cache
ConeAcmr: optimize mesh for transform cache, split sequentially into variable length clusters that are relatively planar, sort clusters by avg normal
ConeCash: optimize mesh for transform cache, picking triangles that reduce ACMR but prioritizing those that keep current cluster planar
MaskBase: split sequentially into 64-triangle clusters, store a 64-bit conservative triangle mask for 6 frustums (cube faces)
ManyConeN: split sequentially into 64-triangle clusters, store N (up to 4) cones for each cluster and a cone id per triangle (2 bit)

Note that all Cone* solutions get significantly worse results with 128 or 256 triangle clusters; it doesn't matter much for Mask.
The biggest challenge with Cone* algorithms is t

zeux

// This file is designed to be included in D3D9-dependent code instead of d3d9.h, while adding minimal amount of junk
#pragma once
 
#include <BaseTyps.h>
#include <BaseTsd.h>
 
// stdlib.h
#ifndef _INC_STDLIB
#define _INC_STDLIB
#endif

zeux

A friend recently learned about Proebsting's law and mentioned it to me off hand. I knew about the law's existence but I never really asked myself - do I believe in it?

For people who aren't aware, Proebsting's law states:

Compiler Advances Double Computing Power Every 18 Years

Which is to say, if you upgrade your compiler every 18 years, you would expect on average your code to double in performance on the same hardware.

Let's C about this

zeux

import argparse
import json
import os
import safetensors
import safetensors.torch
import sys
import time
import torch

def fast_save_file(tensors, filename, metadata=None):

zeux

// brew install libomp
// cc -o matbench matbench.c -O3 -ffast-math -Xclang -fopenmp -I/opt/homebrew/opt/libomp/include -L/opt/homebrew/opt/libomp/lib -lomp
// ./matbench

#include <assert.h>
#include <math.h>
#include <omp.h>
#include <stdio.h>
#include <time.h>

zeux

// This code looks at precision impact of transforming a vector repeatedly by a slightly-non-unit quaternion
// Slightly-non-unit quaternions are important: they result in the process of quaternion computations naturally
// Repeated transformations are important: they may occur during simulation or complex long chains of computation

// Note that because this code runs in JS in double precision, this doesn't model floating-point roundoff.

function applyQuaternion1( q, v ) {

	const x = v.x, y = v.y, z = v.z;
	const qx = q.x, qy = q.y, qz = q.z, qw = q.w;

zeux

State of Roblox graphics API across all platforms, with percentage deltas since EOY 2018. Updated December 29 2019.

Windows

API	Share
Direct3D 11+	85% (+5%)
Direct3D 10.1	8.5% (-1.5%)
Direct3D 10.0	5.5% (-2.5%)
Direct3D 9	1% (-1%)

zeux

#version 450

// 2D Polyhedral Bounds of a Clipped, Perspective-Projected 3D Sphere. Michael Mara, Morgan McGuire. 2013
bool projectSphereView(vec3 c, float r, float znear, float P00, float P11, out vec4 aabb)
{
    if (c.z < r + znear) return false;

    vec3 cr = c * r;
    float czr2 = c.z * c.z - r * r;