rygorous

0x00 0 cost=10
0x01 BCAX(A, 0xf5, 0x0c) cost=30
0x02 BIC(0x0a, B) cost=20
0x03 EOR(A, 0xf3) cost=20
0x04 BIC(B, 0xfa) cost=20
0x05 EOR(A, 0xf5) cost=20
0x06 BCAX(0x0a, B, A) cost=20
0x07 EOR(A, 0xf7) cost=30
0x08 BCAX(B, B, 0x0a) cost=20
0x09 EOR3(A, 0x0a, 0xf3) cost=30

rygorous

import functools
from typing import Optional

class BddContext:
    def __init__(self):
        # 0/1 leaf nodes, it's convenient to just refer to them as 0 and 1
        # a node is a (var id, lo_node, hi_node) tuple
        self.nodes = []
        self.vars = []
        self.node_dict = {}

rygorous

// The FFT alg used here was designed to be very FMA-friendly, but because we can't assume FMAs are present on
// all target HW and want consistent results everywhere, we're using FMA-less algorithms for this application.

// Notation throughout this file:
//
// Let z = a + bi. Then conj(z) = a - bi.
//
// We can swap the real and imaginary parts of z to yield s(z) = b + ai ("swap").
// Now because
//

rygorous

static size_t const kMaxN = 2048; // This is the largest straight FFT we support

struct complexf
{
    float re;
    float im;

    complexf() {}
    complexf(float r) : re(r), im(0.0f) {}
    complexf(float r, float i) : re(r), im(i) {}

rygorous

#include <immintrin.h>
#include <stdio.h>
#include <stdint.h>
#include <assert.h>

typedef uint32_t uint32;
typedef __m128i Vec128_U32;

// "Insert" two 0 bits after each of the 11 low bits of x
static uint32 Part1By2(uint32 x)

rygorous

//===================================================================
// constraint validation for preserve extremes mode

// Cold part of are_endpoitns_permitted: the actual constraint validation
static RADNOINLINE bool are_endpoints_permitted_cold(const BC7BlockState & st, const bc7rd_blockinfo& info)
{
	const BC7Flags flags = info.flags;

	if (st.mode <= 3)
	{

rygorous

{
	BC7Flags flags = in_flags;

	// Preserve extremes mode.
	//
	// This mode preserves values of 0 and 255 in the alpha channel exactly. In general,
	// we can do this in any one channel, but restricting this to alpha makes the interface
	// simpler, is consistent with what we do for BC3, and doesn't seem like a signfiicant
	// limitation for the user.
	//

rygorous

	uint EvenBitMask = 0x55555555u;
	uint HighIndexBit = PackedIndices >> 1;

	float NumWeight1 = float(countbits(EvenBitMask & ~HighIndexBit & PackedIndices));
	float NumWeight2 = float(countbits(EvenBitMask & HighIndexBit & ~PackedIndices));
	float NumWeight3 = float(countbits(EvenBitMask & HighIndexBit & PackedIndices));

// with NV LOP3: 1 shift, 3 LOP3, 3 pop count, 3 int->float = 10 insns total for 16 pixels
// without: 1 shift, 2 NOT, 5 AND, 3 pop count, 3 int->float = 14 insns total for 16 pixels

rygorous

//===================================================
// Oodle2 DDS Tool 
// (C) Copyright 1994-2022 Epic Games Tools LLC
//===================================================

#ifndef RR_DDS_INCLUDED
#define RR_DDS_INCLUDED

#include <stdint.h>
#include <stddef.h>

rygorous

use bit_set::BitSet;
use std::mem;

// Vanilla RFC 4648
const ALPHABET: &str = "ABCDEFGHIJKLMNOPQRSTUVWXYZabcdefghijklmnopqrstuvwxyz0123456789+/";
// URL-safe RFC 4648
//const ALPHABET: &str = "ABCDEFGHIJKLMNOPQRSTUVWXYZabcdefghijklmnopqrstuvwxyz0123456789-_";
const COUNT: usize = 1usize << 24; // 3 bytes worth suffices for this test

fn lookup(index: u32) -> u32 {