Disable POPCNT support per default

This is mainly intended to allow 64 bit compiles on any
system and avoid to crash when the binary, compiled on a
box where POPCNT is not supported, is run on a Core i7
system or similar CPU.

What could happen is that when compiled in a standard 64 bit
system, because the correct headers for the POPCNT intrinsic
are not found, the compiler creates dummy bit count functions
instead, these are never called at runtime on the machine where
Stockfish has been compiled. But if we run the same binary on a
Core i7 system, because POPCNT is detected at run time, the dummy
bitcount functions will be called giving false results that will
crash the application.

Note that would be possible to fallback on software bit count in
these cases, but this is even more subtle because POPCNT path is not
optimized so that we have an application working but at sub-optimal
speed, so better to crash, at least user is loudly warned that there
is something wrong.

If, instead, Stockfish is compiled on a Core i7 system with POPCNT
enabled, then if the PGO compile has been done properly, the same binary
will run at optimal speed _both_ on the Core i7 machine and on any other
64 bit standard machine. This is the ideal mode for binary distribution.

Finally this patch disables bsfq support under Windows, because it seems
inline assembly is not supported both by MSVC and by Intel Windows version.

Signed-off-by: Marco Costalba <mcostalba@gmail.com>

diff --git a/Readme.txt b/Readme.txt
index 0c6f5c7210bec5e62f17f3a12d40e77c78673151..1a3713985e7cec3e47746503892284cd518f4c64 100644 (file)
--- a/Readme.txt
+++ b/Readme.txt
@@ -54,6 +54,15 @@ The exception is computer with big-endian CPUs, like PowerPC
 Macintoshes. Some of the bitboard routines in the current version of\r
 Stockfish are endianness-sensitive, and won't work on a big-endian CPU.\r
 \r
 Macintoshes. Some of the bitboard routines in the current version of\r
 Stockfish are endianness-sensitive, and won't work on a big-endian CPU.\r
 \r

+Stockfish has POPCNT instruction runtime detection and support. This can\r
+give an extra speed on Core i7 or similar systems. To enable this feature\r
+(disabled by default) simply uncomment #define USE_POPCNT in bitcount.h\r
+before to compile.\r
+\r
+On 64 bit Unix-like systems the 'bsfq' assembly instruction will be used\r
+for bit counting. Detection is automatic at compile time, but in case you\r
+experience compile problems you can comment out #define USE_BSFQ line in types.h\r
+\r

 \r
 5. Terms of use\r
 ---------------\r
 \r
 5. Terms of use\r
 ---------------\r

diff --git a/src/bitcount.h b/src/bitcount.h
index 9a3b481169b667943dedbb28cf538f977e02ca5e..ef841f3a7984dc7eba803e19df2763ab302c9086 100644 (file)
--- a/src/bitcount.h
+++ b/src/bitcount.h
@@ -22,18 +22,19 @@
 #if !defined(BITCOUNT_H_INCLUDED)
 #define BITCOUNT_H_INCLUDED
 
 #if !defined(BITCOUNT_H_INCLUDED)
 #define BITCOUNT_H_INCLUDED
 

-// To disable POPCNT support uncomment NO_POPCNT define. You should do it only
-// in PGO compiling to exercise the default fallback path. Don't forget to
-// re-comment the line for the final optimized compile though ;-)
+// To enable POPCNT support uncomment USE_POPCNT define. For PGO compile on a Core i7
+// you may want to collect profile data first with USE_POPCNT disabled and then, in a
+// second profiling session, with USE_POPCNT enabled so to exercise both paths. Don't
+// forget to leave USE_POPCNT enabled for the final optimized compile though ;-)

 
 

-//#define NO_POPCNT
+//#define USE_POPCNT

 
 
 #include "types.h"
 
 // Select type of intrinsic bit count instruction to use
 
 
 
 #include "types.h"
 
 // Select type of intrinsic bit count instruction to use
 

-#if defined(_MSC_VER) && defined(IS_64BIT) && !defined(NO_POPCNT) // Microsoft compiler
+#if defined(_MSC_VER) && defined(IS_64BIT) && defined(USE_POPCNT) // Microsoft compiler

 
 #include <intrin.h>
 
 
 #include <intrin.h>
 


@@ -54,7 +55,7 @@ template<typename T> unsigned __popcnt64(T) { return 0; } // Is never called
 
 #define POPCNT_INTRINSIC(x) __popcnt64(x)
 
 
 #define POPCNT_INTRINSIC(x) __popcnt64(x)
 

-#elif defined(__INTEL_COMPILER) && defined(IS_64BIT) && !defined(NO_POPCNT) // Intel compiler
+#elif defined(__INTEL_COMPILER) && defined(IS_64BIT) && defined(USE_POPCNT) // Intel compiler

 
 #include <nmmintrin.h>
 
 
 #include <nmmintrin.h>
 


@@ -70,7 +71,7 @@ template<typename T> unsigned _mm_popcnt_u64(T) { return 0; } // Is never called
 
 #define POPCNT_INTRINSIC(x) _mm_popcnt_u64(x)
 
 
 #define POPCNT_INTRINSIC(x) _mm_popcnt_u64(x)
 

-#else // Safe fallback for unsupported compilers or when NO_POPCNT is defined
+#else // Safe fallback for unsupported compilers or when USE_POPCNT is disabled

 
 inline bool cpu_has_popcnt() { return false; }
 
 
 inline bool cpu_has_popcnt() { return false; }
 


@@ -145,12 +146,8 @@ inline int count_1s_max_15(Bitboard b) {
 
 // Global constant initialized at startup that is set to true if
 // CPU on which application runs supports POPCNT intrinsic. Unless
 
 // Global constant initialized at startup that is set to true if
 // CPU on which application runs supports POPCNT intrinsic. Unless

-// NO_POPCNT is defined.
-#if defined(NO_POPCNT)
-const bool CpuHasPOPCNT = false;
-#else
+// USE_POPCNT is not defined.

 const bool CpuHasPOPCNT = cpu_has_popcnt();
 const bool CpuHasPOPCNT = cpu_has_popcnt();

-#endif

 
 
 // Global constant used to print info about the use of 64 optimized
 
 
 // Global constant used to print info about the use of 64 optimized

diff --git a/src/types.h b/src/types.h
index af5e2455dd1434d3c58cb9312f7d5337e13dde53..70e47079112833b0a3bdf2dce775df81ae1db6d2 100644 (file)
--- a/src/types.h
+++ b/src/types.h
@@ -62,7 +62,7 @@ typedef uint64_t Bitboard;
 #define IS_64BIT
 #endif
 
 #define IS_64BIT
 #endif
 

-#if defined(IS_64BIT) && (defined(__GNUC__) || defined(__INTEL_COMPILER))
+#if defined(IS_64BIT) && !defined(_WIN64) && (defined(__GNUC__) || defined(__INTEL_COMPILER))

 #define USE_BSFQ
 #endif
 
 #define USE_BSFQ
 #endif