vurtun

// Selection algorithm
// Partial sorting

// http://www.unicode.org/Public/UCD/latest/ucd/CaseFolding.txt
switch(rune) {
case 0x0041: return 0x0061; // LATIN CAPITAL LETTER A
case 0x0042: return 0x0062; // LATIN CAPITAL LETTER B
case 0x0043: return 0x0063; // LATIN CAPITAL LETTER C
case 0x0044: return 0x0064; // LATIN CAPITAL LETTER D
case 0x0045: return 0x0065; // LATIN CAPITAL LETTER E

vurtun

#include <stdio.h>
#include <stdlib.h>
#include <string.h>
#include <math.h>
#include <time.h>

#define szof(a) ((int)sizeof(a))
#define cntof(a) ((int)(sizeof(a) / sizeof((a)[0])))

static void

vurtun

#include <stdio.h>
#include <limits.h>

#define byte4_dup(c)        (((c)<<24u)|((c)<<16u)|((c)<<8u)|(c))
#define byte4_cmp_lt(x,n)   (((x)-~0UL/255u*(n))&~(x)&~0UL/255u*128u) // x>=0; 0<=n<=128 
#define byte4_cmp_gt(x,n)   (((x)+~0UL/255*(127-(n))|(x))&~0UL/255*128) // x>=0; 0<=n<=127 

#ifdef _MSC_VER
static int bit_ffs32(unsigned int u) {_BitScanForward(&u, u); return (int)(u);}
static int bit_fls32(unsigned int u) {_BitScanBackward(&u, u); return 31-(int)(u);}

vurtun

/*  This algorithm uses as a predicate equality of the first and the last characters from the substring. These two characters are populated in two registers
 *  F and L respectively. Then in each iteration two chunks of strings are loaded. The first chunk (A) is read from offset i (where i is the current offset)
 *  and the second chunk (B) is read from offset i + k - 1, where k is sub string's length. Then we compute a vector expression F == A and B == L.
 *  This step yields a byte vector (or a bit mask), where "true" values denote position of potential substring occurrences.
 *  Finally, just at these positions an exact comparisons of sub strings are performed.
 *
 *  Example: Let's assume 8-byte registers. We're searching for word "cat", thus:
 *
 *    F = [ c | c | c | c | c | c | c | c ]
 *    L = [ t | t | t | t | t | t | t | t ]

vurtun

#define xglue(x, y) x##y
#define glue(x, y) xglue(x, y)
#define uniqid(name) glue(name, __LINE__)

#ifdef _MSC_VER
#define swap(a,b) do { decltype((a) + 0) _t = (a); (a) = (b); (b) = _t; } while(0)
#else
#define swap(a,b) do {__typeof__((a) + 0) _t = (a); (a) = (b); (b) = _t; } while(0)
#endif

vurtun

/**
 * Advent of Code 2022
 * Day 2: Rock Paper Scissors
 * https://adventofcode.com/2022/day/2
 *
 * http://codercorner.com/Modulo3.htm
 * (Index+1)%3 = (1 << Index1) & 3
 */
#include <stdlib.h>
#include <stdio.h>

vurtun

static void
qdiag(float *restrict qres, const float *restrict A) {
  /* Diagonalizer: http://melax.github.io/diag.html
    'A' must be a symmetric matrix.
    Returns quaternion q such that corresponding matrix Q 
    can be used to Diagonalize 'A'
    Diagonal matrix D = Q * A * Transpose(Q);  and  A = QT*D*Q
    The rows of q are the eigenvectors D's diagonal is the eigenvalues
    As per 'row' convention if float3x3 Q = q.getmatrix(); then v*Q = q*v*conj(q)
  */

vurtun

/* 
Enable FMA support:
    clang -O2 -Wall -mfma
*/
static inline float
rsqrta(float x) {
  /* Exact bits: 13.71 */
  /* δ+max = 7.459289×10^−5 */
  /* δ−max = −7.450387×10^−5 */
  union { float f; int i; } v = {x};

vurtun

#include <stdio.h>
#include <stdlib.h>
#include <limits.h>
#include <float.h>
#include <string.h>
#include <assert.h>
#include <stdint.h>
#include <stdarg.h>
#include <math.h>

vurtun

static inline int
bin_search(const void *vals, int cnt, int siz, void *val,
        int(*cmp_less)(const void *a, const void *b)) {
  int half, nleft = cnt;
  const unsigned char *base = vals;
  while ((half = (nleft >> 1)) > 0) {
    const unsigned char *mid = base + half * siz;
    base = cmp_less(mid, val) ? mid : base;
    nleft -= half;
  }