lemire

// with jemalloc

$ strace ./lemirebenchmark
execve("./lemirebenchmark", ["./lemirebenchmark"], [/* 28 vars */]) = 0
brk(NULL)                               = 0x79d0000
openat(AT_FDCWD, "/usr/local/lib/libjemalloc.so.2", O_RDONLY|O_CLOEXEC) = 3
read(3, "\177ELF\2\1\1\0\0\0\0\0\0\0\0\0\3\0\267\0\1\0\0\0\340b\0\0\0\0\0\0"..., 832) = 832
fstat(3, {st_mode=S_IFREG|0755, st_size=4887824, ...}) = 0
mmap(NULL, 1113224, PROT_READ|PROT_EXEC, MAP_PRIVATE|MAP_DENYWRITE, 3, 0) = 0xffffaa020000
mmap(0xffffaa090000, 131072, PROT_READ|PROT_WRITE, MAP_PRIVATE|MAP_FIXED|MAP_DENYWRITE, 3, 0x60000) = 0xffffaa090000

lemire

// with jemalloc

$ strace ./lemirebenchmark
execve("./lemirebenchmark", ["./lemirebenchmark"], [/* 28 vars */]) = 0
brk(NULL)                               = 0x79d0000
openat(AT_FDCWD, "/usr/local/lib/libjemalloc.so.2", O_RDONLY|O_CLOEXEC) = 3
read(3, "\177ELF\2\1\1\0\0\0\0\0\0\0\0\0\3\0\267\0\1\0\0\0\340b\0\0\0\0\0\0"..., 832) = 832
fstat(3, {st_mode=S_IFREG|0755, st_size=4887824, ...}) = 0
mmap(NULL, 1113224, PROT_READ|PROT_EXEC, MAP_PRIVATE|MAP_DENYWRITE, 3, 0) = 0xffffaa020000
mmap(0xffffaa090000, 131072, PROT_READ|PROT_WRITE, MAP_PRIVATE|MAP_FIXED|MAP_DENYWRITE, 3, 0x60000) = 0xffffaa090000

lemire

#include <string.h>
#include <stdint.h>

static inline
uint32_t _zigzag_encode_32 (int32_t val) {
	return (val + val) ^ (val >> 31);
}

void zigzag_encode(const int32_t * in, uint32_t * out, size_t N) {
    for(size_t i = 0; i < N; i++)

lemire

#ifndef SIMDJSON_ALLOCATOR_H
#define SIMDJSON_ALLOCATOR_H

#include "common_defs.h"
#include <cstddef>
#include <cstdint>
#include <limits>

template <class T> class padded_allocator {
public:

lemire

// gcc -o t2 t2.c  -mavx2
#include <stdio.h>
#include <string.h>
#include <stdbool.h>
#include <x86intrin.h>

 __attribute__((no_sanitize("memory")))
__attribute__ ((noinline))
bool hash_zero_byte(char *va) {
  __m256i v = _mm256_loadu_si256((__m256i *) va);

lemire

def f(x):
  r1 = (x + 212) % 256
  r2 = x | 32
  table = [44, 125, 0, 0, 192, 0, 0, 0, 0, 0, 0, 0, 0, 0, 58, 123 ]
  if ((r1&0x80)==0x80):
    r3 = 0
  else:
    r3 = table[r1& 0xF]
  r4 = r2 == r3
  return r4

lemire

// the following works fine under clang:
#pragma clang attribute push(__attribute__((target("sse4.2,pclmul"))), apply_to=function)
void f(){}
#pragma clang attribute pop


////////////
/// The following DOES NOT WORK: 'error: expected unqualified-id'
//////////
namespace joe {

lemire

// this is more or less a port of travis_avx2_w to plain C

namespace {
// this is __m256i for people who don't have avx
struct poor__m256i {
    uint32_t val[8];
};

#ifndef really_inline

lemire

package Main

import (
	"math"
	"reflect"
	"testing"

	"github.com/RoaringBitmap/roaring"
	"github.com/willf/bitset"
)

lemire

// prefetching can be useful for large documents
// where page walking becomes a significant overhead
#ifdef _MSC_VER
// annoyingly, Visual Studio does not appear to have a portable
// prefetch
#ifdef IS_ARM64
inline void prefetch(const void * p) {
  __prefetch(p);
}