pkrisz99 / renegade Goto Github PK

Hardware=Intel(R) Core(TM) i5-4570 3.20GHz 8GB RAM
OS=Windows 10 Pro x64
GUI=Arena 3.5.1

Renegade configuration:
[Renegade 1.1.0 x64(BMI2)]
Hash=16
Threads=1

Sandbox Info:
SBIE2101 Object name not found: Unnamed object, error OpenProcess (C0000022) access=001FFFFF initialized=1

It's sad that it's not possible to compile the newest version on Android and Linux devices.

when compiling the source of v1.1.0 (dev 24 / PR 27) on Linux (Xubuntu 22.04), using the makefile, i get this error:
NOTE: in the makefile i changed clang++ to clang++-16

Magics.cpp:145:5: error: return type 'unsigned long long' must match
previous return type 'uint64_t' (aka 'unsigned long')
when lambda expression has unspecified explicit return type
                                return 0ull;

i found a solution to meet the message and solve the error by changing return 0ull into return 0ul.

concerning this 'value construct' in C++ : at https://stackoverflow.com/questions/7036056/what-do-0ll-or-0x0ul-mean i read :

each L or l represents a long and each U or u represents an unsigned

Hey, would you consider adding a license file? 😀

As of 7eb14de this was mitigated, invalid PVs are truncated. It still outputs an incorrect PV line if but the moves are legal.

It's unclear how often this happens, when the engine was connected through lichess-bot it seemed to happen in about 10% of the moves, but these are the illegal PVs only. These issues are more prevalent, the longer a PV is, the smallest depth I've seen this issue occur was at depth=4.

hi, i just compiled your newest main source and i succeeded, but i had to adjust the Makefile for Linux. That file contains a warning, stating it's experimental .. why ? You should take some time learning how to create a proper Makefile for all OS, otherwise people can't use the engine, or they must download the asset (but there's none for Apple).

In 'Search.cpp' i read :

This is the main search code of Renegade. If you're reading this, you're probably interested in what this engine does, and I'm happy for that, feel free to try some ideas from here!

However, you may have confidence in this code, but be aware, that I don't. Here are some caveats:

- this engine does not reduce losing captures at all
- LMR is quite conservative
- the ordering score can come either directly from history, or from captures/killers etc.
- quiescence search is very basic
- there aren't any form of staged movegen implemented
- elo gain estimates are very outdated
- some stuff are just plain cursed

The current CCRL list shows your engine has rating 3552, which is impressive .. i wonder how it can improve when you focus on the above shortcomings !