ssvb

/+dub.sdl:+/
// Siarhei Siamashka - Programming Challange from Erann Gat:
//   http://www.flownet.com/ron/papers/lisp-java/
// This work is marked with CC0 1.0. To view a copy of this license, visit
//   http://creativecommons.org/publicdomain/zero/1.0
@safe:
import std.stdio, std.file, std.conv, std.ascii, std.algorithm, std.string;
import core.bitop, std.bigint;

// Can be changed to 'long' for more speed, but will overflow in extreme cases

ssvb

/+dub.sdl:+/
// Siarhei Siamashka - Programming Challange from Erann Gat:
//   http://www.flownet.com/ron/papers/lisp-java/
// This work is marked with CC0 1.0. To view a copy of this license, visit
//   http://creativecommons.org/publicdomain/zero/1.0
@safe:
import std.stdio, std.file, std.conv, std.ascii, std.algorithm, std.string;
import core.bitop;

immutable letter2digit = (){

ssvb

/+dub.sdl:+/
// Siarhei Siamashka - Programming Challange from Erann Gat:
//   http://www.flownet.com/ron/papers/lisp-java/
// This work is marked with CC0 1.0. To view a copy of this license, visit
//   http://creativecommons.org/publicdomain/zero/1.0
@safe:
import std.stdio, std.file, std.conv, std.ascii, std.algorithm, std.string;

immutable letter2digit = (){
  char[256] tmp;

ssvb

import std.stdio, std.range, std.algorithm, std.compiler, core.bitop;

// Optimizing compilers should do tail call optimization here, so
// 0xBADF00D magic constant won't be found in stack if optimizations
// are enabled. Is there a cleaner way to detect optimizations?
bool detect_opt() {
  int[100] filler;
  bool detect_tail_call_opt(int depth, int magic) {
    if (depth > 100) {
      int x; foreach (i ; 20 .. 80) if (*(&x + i) == magic) return false;

ssvb

#!/usr/bin/env ruby

$data = {}
$func_name = nil

# binutils confuses the sizes of the first and the second arguments
# this function contains a fixup workaround
def fixup_vshll(l)
  if l =~ /(.*vshll.u16\s+)d(\d+),\s*q(\d+)(.*)/
    reg1 = $2.to_i

ssvb

require "benchmark"

# Access two 64-bit random seed values
struct Crystal::Hasher
  def self.get_seed
    @@seed
  end
end

# Implementation of FunnyHash from

ssvb

The real starting rating is 1400 (but displayed as 0). Your rating delta after your first contest was -110 and your real rating changed to 1290 (displayed as 390). Your rating delta after your second contest was -43 and your real rating changed to 1247 (displayed as 697).

The following Python code can be used to convert your display rating to your real rating and the other way around:

def get_display_rating(real_rating, number_of_contests):
  if number_of_contests >= 6:
    return real_rating
  display_delta = [1400, 900, 550, 300, 150, 50]
 return real_rating - display_delta[number_of_contests]

ssvb

Deactivated jsimd_h2v1_downsample_dspr2 and jsimd_h2v2_downsample_dspr2

Running tests...
Test project /tmp/libjpeg-turbo.bigendian/build
        Start   1: tjunittest-static
  1/151 Test   #1: tjunittest-static .................................***Failed   73.19 sec
        Start   2: tjunittest-static-alloc
  2/151 Test   #2: tjunittest-static-alloc ...........................***Failed   72.38 sec
        Start   3: tjunittest-static-yuv

ssvb

Little endian MIPS74K @480MHz without FPU, GCC 7.2.0

Without DSP r2:

# JSIMD_FORCENONE=1 ./tjbench-static test-image.ppm 75 -fastdct

Using fastest DCT/IDCT algorithm

>>>>>  BGR (Top-down) <--> JPEG GRAY Q75  <<<<<

ssvb

<< MIPSELo32 >> OpenWrt ~/libjpeg-turbo/build # JSIMD_FORCENONE=1 ./tjbench-static test-image.ppm 95

>>>>>  BGR (Top-down) <--> JPEG GRAY Q95  <<<<<

Image size: 2592 x 1944
Compress      --> Frame rate:         0.448023 fps
                  Output image size:  1225591 bytes
                  Compression ratio:  12.334085:1
                  Throughput:         2.257520 Megapixels/sec
                  Output bit stream:  4.392744 Megabits/sec