There was a major issue with the parallel search in 5.2E where the best move was not being copied at the root. This is one of those cases where you think you have created an elegant and compact code, when it fact, it was incorrect and short in logic.
I have fixed this in 5.2F.
Also, you should now be able to set the number of threads by specifying the "threads" option in the command-line. It would be something like:
MP-Thinker.exe hashsize=8 threads=4
Lastly, I have disabled the splitting in interior nodes. For now, splitting is at root nodes only. My tests show very little gain with splitting of interior nodes. So, it is off for now until I have improved it.
If you notice that the node counts don't scale, don't worry. The current code simply is not accurate in adding up the node counts from child threads. There is no node count conspiracy here (until someone finds my NodeCount div 10 code).
CThinker wrote:There was a major issue with the parallel search in 5.2E where the best move was not being copied at the root. This is one of those cases where you think you have created an elegant and compact code, when it fact, it was incorrect and short in logic.
I have fixed this in 5.2F.
Also, you should now be able to set the number of threads by specifying the "threads" option in the command-line. It would be something like:
MP-Thinker.exe hashsize=8 threads=4
Lastly, I have disabled the splitting in interior nodes. For now, splitting is at root nodes only. My tests show very little gain with splitting of interior nodes. So, it is off for now until I have improved it.
If you notice that the node counts don't scale, don't worry. The current code simply is not accurate in adding up the node counts from child threads. There is no node count conspiracy here (until someone finds my NodeCount div 10 code).
Thanks Lance for the quick fix and the great chess engine of yours
_No one can hit as hard as life.But it ain’t about how hard you can hit.It’s about how hard you can get hit and keep moving forward.How much you can take and keep moving forward….
Looked at werner post for adapter.eng hashtable settings. You need to put a file called Thinker52F.init or whatever u like to call it in Thinker52F folder.
C:\Program Files\ChessBase\Engines\Thinker52f
adapter.eng renamed to Thinker52F.eng for example in C:\Program Files\ChessBase\Engines
CThinker wrote:There was a major issue with the parallel search in 5.2E where the best move was not being copied at the root. This is one of those cases where you think you have created an elegant and compact code, when it fact, it was incorrect and short in logic.
I have fixed this in 5.2F.
Also, you should now be able to set the number of threads by specifying the "threads" option in the command-line. It would be something like:
MP-Thinker.exe hashsize=8 threads=4
Lastly, I have disabled the splitting in interior nodes. For now, splitting is at root nodes only. My tests show very little gain with splitting of interior nodes. So, it is off for now until I have improved it.
If you notice that the node counts don't scale, don't worry. The current code simply is not accurate in adding up the node counts from child threads. There is no node count conspiracy here (until someone finds my NodeCount div 10 code).
Unfortunately, i am thinking that when you set the number of threads, 2 is as low as it will allow you to go. I dont think it can be set to "1 Thread" Am i right or wrong?
CThinker wrote:There was a major issue with the parallel search in 5.2E where the best move was not being copied at the root. This is one of those cases where you think you have created an elegant and compact code, when it fact, it was incorrect and short in logic.
I have fixed this in 5.2F.
Also, you should now be able to set the number of threads by specifying the "threads" option in the command-line. It would be something like:
MP-Thinker.exe hashsize=8 threads=4
Lastly, I have disabled the splitting in interior nodes. For now, splitting is at root nodes only. My tests show very little gain with splitting of interior nodes. So, it is off for now until I have improved it.
If you notice that the node counts don't scale, don't worry. The current code simply is not accurate in adding up the node counts from child threads. There is no node count conspiracy here (until someone finds my NodeCount div 10 code).
Unfortunately, i am thinking that when you set the number of threads, 2 is as low as it will allow you to go. I dont think it can be set to "1 Thread" Am i right or wrong?
Best,
George, if you want to test it as a single core engine, you dont have to use any commands. Just play with it like you used to.
CThinker wrote:There was a major issue with the parallel search in 5.2E where the best move was not being copied at the root. This is one of those cases where you think you have created an elegant and compact code, when it fact, it was incorrect and short in logic.
I have fixed this in 5.2F.
Also, you should now be able to set the number of threads by specifying the "threads" option in the command-line. It would be something like:
MP-Thinker.exe hashsize=8 threads=4
Lastly, I have disabled the splitting in interior nodes. For now, splitting is at root nodes only. My tests show very little gain with splitting of interior nodes. So, it is off for now until I have improved it.
If you notice that the node counts don't scale, don't worry. The current code simply is not accurate in adding up the node counts from child threads. There is no node count conspiracy here (until someone finds my NodeCount div 10 code).
Unfortunately, i am thinking that when you set the number of threads, 2 is as low as it will allow you to go. I dont think it can be set to "1 Thread" Am i right or wrong?
Best,
George, if you want to test it as a single core engine, you dont have to use any commands. Just play with it like you used to.
Hmm- i will try that- but remember i am converting it to UCI with the adapters. I wonder if that has any bearing on it. At any rate, i will try.
Thanks, Tony.
geots wrote:Tried that ,Tony- the damn thing keeps using 2 threads. Any other ideas?
Best,
Thinker 5.2f is a bugfix for MP only.
Thinker 5.2e was fine on 1CPU.
I'm picking that Thinker 5.2f is an MP version only and that there are no other changes.
CThinker wrote:There was a major issue with the parallel search in 5.2E where the best move was not being copied at the root. This is one of those cases where you think you have created an elegant and compact code, when it fact, it was incorrect and short in logic.
I have fixed this in 5.2F.
Also, you should now be able to set the number of threads by specifying the "threads" option in the command-line. It would be something like:
MP-Thinker.exe hashsize=8 threads=4
Lastly, I have disabled the splitting in interior nodes. For now, splitting is at root nodes only. My tests show very little gain with splitting of interior nodes. So, it is off for now until I have improved it.
If you notice that the node counts don't scale, don't worry. The current code simply is not accurate in adding up the node counts from child threads. There is no node count conspiracy here (until someone finds my NodeCount div 10 code).
Unfortunately, i am thinking that when you set the number of threads, 2 is as low as it will allow you to go. I dont think it can be set to "1 Thread" Am i right or wrong?
Best,
5.2E and 5.2F are essentially identical except that 5.2F has parallel search.
If you need the single thread engine, please use 5.2E.
