Stockfish v2.1 now available

[Amstaff]

Thanks Jim for you good service to us.
Gerald Grimsley

[ernest]

Here is re-compile which will show correct searched nodes.

Thanks Jim!
Now on my Intel Core2 Duo, I get indeed the same infinite analysis.

...and here, the Intel compile is 6% faster than the Msvc compile!

Intel compiler
info depth 21 seldepth 32 multipv 1 score cp 20 nodes 32817584 nps 821713 time 3
9938 pv e2e4 e7e6 g1f3 d7d5 b1c3 g8f6 e4d5 e6d5 d2d4 f8d6 f1d3 e8g8 e1g1 b8c6 c1
g5 c6b4 c3b5 c8g4 f1e1 h7h6 b5d6 d8d6
info depth 22

Msvc compiler
info depth 21 seldepth 32 multipv 1 score cp 20 nodes 32817584 nps 775884 time 4
2297 pv e2e4 e7e6 g1f3 d7d5 b1c3 g8f6 e4d5 e6d5 d2d4 f8d6 f1d3 e8g8 e1g1 b8c6 c1
g5 c6b4 c3b5 c8g4 f1e1 h7h6 b5d6 d8d6
info depth 22

[Sven]

The problem is here, in book.h file:

Code: Select all

template<int n> inline uint64_t Book::get_int() { return 256 * get_int<n-1>() + get(); }

This line recurively reads from the book file a number of chars specified by <int n>.

For this to work it is critical that the recursive function get_int<n-1>() is called _before_ reading the next character through calling get().

[...]

I would assume this bug is mine, not of MSVC because C++ standard does not seem to guarantee the order of evaluation of a term in a sum expression, so that we have to rewrite the code in a way to enforce the ordering, one quick hack (that works with MSVC) is to substitute the above line with:

Code: Select all

template<int n> inline uint64_t Book::get_int() { uint64_t v = 256 * get_int<n-1>(); return v + get(); }

so that 'v' is evaluated before get(), but I don't like it a lot and probably I will find something less tricky....

The limit of elegance is correctness

What about this? It is a lot less "tricky", comes without the ugly cast from uint64_t to the concrete type, avoids the special case get_int<1>(), shows the order of operation at runtime, keeps the elegance of the operator>> usage in Book::read_entry(), and should generate about the same code as the "tricky" one, after unrolling the loop ...

Code: Select all

template<typename T>
inline void Book::get_int(T& v) {
    v = bookFile.get();
    for (int i = 2; i <= sizeof(T); i++) {
        v = 256 * v + bookFile.get();
    }
}

template<typename T>
inline Book& Book::operator>>(T& n) { get_int(n); return *this; }

Sven

[Daniel Shawul]

Best use of template meta-programming I have encounted so far is with tensor operations. One template class is enough for all tensor dimensions and single/double precision unlike its fortran counterpars. And all are optimized at compile time by loop unrolling using that method. Another nice use is for operator overloading on two big matrices. To get the c++ equivalent with templates of the following code, some nifty tricks are used to avoid storing a 'copy' of one of the matrices and produce code as efficient as this straightforward method.

Code: Select all

for(int i = 0;i < N;i++)
for(int j = 0;j < M;j++) 
C[i][j] = A[i][j] + B[i][j]

equivalent

Code: Select all

C = A + B;

not

Code: Select all

C = A;
C += B;

[mcostalba]

What about this?

Incidentally I have already committed something very similar to your approach, here is current code:

Code: Select all

template<typename T>
void Book::get_number(T& n) {

  n = 0;

  for (size_t i = 0; i < sizeof(T); i++)
      n = (n << 8) + (T)bookFile.get();
}

Anyhow thanks for the suggestion !

[poisonedpawn]

is it just me or is anybody else seeing many losses on time with this release??

[gerold]

is it just me or is anybody else seeing many losses on time with this release??

Using win. 32 in Arena. 5/3 tc over 500 games. No time lost here.

[jpqy]

No darren..after 250 games no single time loose (yet) seen.

Stockfish 2.1 do it very well on my system..great work Stockfish Team!

JP.

[royb]

is it just me or is anybody else seeing many losses on time with this release??

Another data point ...

I've run 200 games with SF-2.1 32-bit in Arena (under Wine on Linux) with no time losses.

[poisonedpawn]

maybe it is a fritz GUI issue then. i never use arena. i had no probs with time losses at all with 2.01 only with 2.1. strange...