Various changes to treatment of integers.

Details:
- Added a new cpp macro in bli_config.h, BLIS_INT_TYPE_SIZE, which can be
  assigned values of 32, 64, or some other value. The former two result in
  defining gint_t/guint_t in terms of 32- or 64-bit integers, while the latter
  causes integers to be defined in terms of a default type (e.g. long int).
- Updated bli_config.h in reference and clarksville configurations according
  to above changes.
- Updated test drivers in test and testsuite to avoid type warnings associated
  with format specifiers not matching the types of their arguments to printf()
  and scanf().
- Inserted missing #include "bli_system.h" into blis.h (which was slated for
  inclusion in d141f9e).
- Added explicit typecasting of dim_t and inc_t to macros in
  bli_blas_macro_defs.h (which are used in BLAS compatibility layer).
- Slight changes to CREDITS and INSTALL files.
- Slight tweaks to Windows build system, mostly in the form of switching to
  Windows-style CRLF newlines for certain files.

CREDITS

@@ -20,6 +20,7 @@ but many others have contributed input and feedback, including:
   Devin Matthews              (The University of Texas at Austin)
   Tze Meng Low                (The University of Texas at Austin)
   Mikhail Smelyanskiy         (Intel, Parallel Computing Lab)
+  Tyler Smith                 (The University of Texas at Austin)
   Rhys Ulerich                (The University of Texas at Austin)
   Robert van de Geijn         (The University of Texas at Austin)
   Zhang Xianyi                (Chinese Academy of Sciences)

INSTALL

@@ -10,11 +10,11 @@ located at the BLIS project's official website:
   http://code.google.com/p/blis/wiki/BuildSystem
 
 This document will always contain the most up-to-date information related
-to instantiating a BLIS library from the framework source code. If you
-have any further questions or wish to provide feedback, please contact the
-BLIS community by posting your message to the general user mailing list:
+to instantiating a BLIS library from the framework source code. If you have
+any further questions or wish to provide feedback, please contact the BLIS
+community by posting your message to the BLIS developer's mailing list:
 
-  https://groups.google.com/d/forum/blis-discuss
+  https://groups.google.com/d/forum/blis-devel
 
 Thanks for your interest in the BLIS framework!
 

config/clarksville/bli_config.h

@@ -40,6 +40,18 @@
 
 
 
+// -- INTEGER PROPERTIES -------------------------------------------------------
+
+// The bit size of the integer type used to track values such as dimensions,
+// strides, diagonal offsets. A value of 32 results in BLIS using 32-bit signed
+// integers while 64 results in 64-bit integers. Any other value results in use
+// of the C99 type "long int". Note that this ONLY affects integers used
+// internally within BLIS as well as those exposed in the native BLAS-like BLIS
+// interface.
+#define BLIS_INT_TYPE_SIZE               64
+
+
+
 // -- FLOATING-POINT PROPERTIES ------------------------------------------------
 
 #define BLIS_NUM_FP_TYPES                4
@@ -131,9 +143,13 @@
 // Enable the BLAS compatibility layer?
 #define BLIS_ENABLE_BLAS2BLIS
 
-// Enable 64-bit integers in the BLAS compatibility layer? If disabled,
-// these integers will be defined as 32-bit.
-#define BLIS_ENABLE_BLAS2BLIS_INT64
+// The bit size of the integer type used to track values such as dimensions and
+// leading dimensions (ie: column strides) within the BLAS compatibility layer.
+// A value of 32 results in the compatibility layer using 32-bit signed integers
+// while 64 results in 64-bit integers. Any other value results in use of the
+// C99 type "long int". Note that this ONLY affects integers used within the
+// BLAS compatibility layer.
+#define BLIS_BLAS2BLIS_INT_TYPE_SIZE     64
 
 // Fortran-77 name-mangling macros.
 #define PASTEF770(name)                        name ## _

config/reference/bli_config.h

@@ -40,6 +40,18 @@
 
 
 
+// -- INTEGER PROPERTIES -------------------------------------------------------
+
+// The bit size of the integer type used to track values such as dimensions,
+// strides, diagonal offsets. A value of 32 results in BLIS using 32-bit signed
+// integers while 64 results in 64-bit integers. Any other value results in use
+// of the C99 type "long int". Note that this ONLY affects integers used
+// internally within BLIS as well as those exposed in the native BLAS-like BLIS
+// interface.
+#define BLIS_INT_TYPE_SIZE               32
+
+
+
 // -- FLOATING-POINT PROPERTIES ------------------------------------------------
 
 #define BLIS_NUM_FP_TYPES                4
@@ -55,7 +67,7 @@
 // -- MULTITHREADING -----------------------------------------------------------
 
 // The maximum number of BLIS threads that will run concurrently.
-#define BLIS_MAX_NUM_THREADS             24
+#define BLIS_MAX_NUM_THREADS             1
 
 
 
@@ -131,9 +143,13 @@
 // Enable the BLAS compatibility layer?
 #define BLIS_ENABLE_BLAS2BLIS
 
-// Enable 64-bit integers in the BLAS compatibility layer? If disabled,
-// these integers will be defined as 32-bit.
-//#define BLIS_ENABLE_BLAS2BLIS_INT64
+// The bit size of the integer type used to track values such as dimensions and
+// leading dimensions (ie: column strides) within the BLAS compatibility layer.
+// A value of 32 results in the compatibility layer using 32-bit signed integers
+// while 64 results in 64-bit integers. Any other value results in use of the
+// C99 type "long int". Note that this ONLY affects integers used within the
+// BLAS compatibility layer.
+#define BLIS_BLAS2BLIS_INT_TYPE_SIZE     32
 
 // Fortran-77 name-mangling macros.
 #define PASTEF770(name)                        name ## _

frame/include/bli_blas_macro_defs.h

@@ -41,8 +41,8 @@
 
 #define bli_convert_blas_dim1( n_blas, n_blis )\
 { \
-	if ( n_blas < 0 ) n_blis = 0; \
-	else              n_blis = n_blas; \
+	if ( n_blas < 0 ) n_blis = ( dim_t )0; \
+	else              n_blis = ( dim_t )n_blas; \
 }
 
 // Macro to reposition vector pointers and flip signs of increments
@@ -54,12 +54,12 @@
 	if ( incx_blas < 0 ) \
 	{ \
 		x_blis    = (x_blas) + (n-1) * (incx_blas); \
-		incx_blis = -(incx_blas); \
+		incx_blis = ( inc_t )(-incx_blas); \
 	} \
 	else \
 	{ \
 		x_blis    = (x_blas); \
-		incx_blis = (incx_blas); \
+		incx_blis = ( inc_t )(incx_blas); \
 	} \
 }
 

frame/include/bli_type_defs.h

@@ -48,48 +48,46 @@
   #include <stdint.h>
 #else
   // When stdint.h is not available, manually typedef the types we will use.
-  #if 0
-  typedef   signed long int   int64_t;
-  typedef unsigned long int  uint64_t;
+  #ifdef _WIN32
+    typedef          __int32  int32_t;
+    typedef unsigned __int32 uint32_t;
+    typedef          __int64  int64_t;
+    typedef unsigned __int64 uint64_t;
   #else
-  typedef   signed long int   int32_t;
-  typedef unsigned long int  uint32_t;
+    #error "Attempting to compile on pre-C99 system without stdint.h."
   #endif
 #endif
 
 // -- General-purpose integers --
 
-// NOTE: Here we define our general-purpose integers in terms of the C types,
-// rather than stdint types, because some systems will need larger integers
-// than others, and "long int" seems to be interpreted to the correct size
-// on most of our systems.
-//#if 0
-//typedef     int64_t  gint_t; // general signed integer
-//typedef    uint64_t guint_t; // general unsigned integer
-//#else
-//typedef     int32_t  gint_t; // general signed integer
-//typedef    uint32_t guint_t; // general unsigned integer
-//#endif
-typedef    signed long int  gint_t; // general signed integer
-typedef  unsigned long int guint_t; // general unsigned integer
+// Define integer types depending on what size integer was requested.
+#if   BLIS_INT_TYPE_SIZE == 32
+typedef           int32_t  gint_t;
+typedef          uint32_t guint_t;
+#elif BLIS_INT_TYPE_SIZE == 64
+typedef           int64_t  gint_t;
+typedef          uint64_t guint_t;
+#else
+typedef   signed long int  gint_t;
+typedef unsigned long int guint_t;
+#endif
 
 // -- Boolean type --
 
-typedef      gint_t bool_t;
+typedef  gint_t  bool_t;
 
 // -- Special-purpose integers --
 
 // This cpp guard provides a temporary hack to allow libflame
 // interoperability with BLIS.
 #ifndef _DEFINED_DIM_T
 #define _DEFINED_DIM_T
-//typedef unsigned long int dim_t;  // dimension type
-typedef     guint_t dim_t;   // dimension type
+typedef guint_t  dim_t;   // dimension type
 #endif
-typedef     guint_t inc_t;   // increment/stride type
-typedef      gint_t doff_t;  // diagonal offset type
-typedef     guint_t siz_t;   // byte size type
-typedef     guint_t info_t;  // object information bit field
+typedef guint_t  inc_t;   // increment/stride type
+typedef  gint_t  doff_t;  // diagonal offset type
+typedef guint_t  siz_t;   // byte size type
+typedef guint_t  info_t;  // object information bit field
 
 // -- Complex types --
 
@@ -143,11 +141,16 @@ typedef dcomplex atom_t;
 // Note: These types are typically only used by BLAS compatibility layer, but
 // we must define them even when the compatibility layer isn't being built
 // because they also occur in bli_slamch() and bli_dlamch().
-#ifdef BLIS_ENABLE_BLAS2BLIS_INT64
+
+// Define f77_int depending on what size of integer was requested.
+#if   BLIS_BLAS2BLIS_INT_TYPE_SIZE == 32
+typedef int32_t   f77_int;
+#elif BLIS_BLAS2BLIS_INT_TYPE_SIZE == 64
 typedef int64_t   f77_int;
 #else
-typedef int32_t   f77_int;
+typedef long int  f77_int;
 #endif
+
 typedef char      f77_char;
 typedef float     f77_float;
 typedef double    f77_double;

frame/include/blis.h

@@ -55,13 +55,7 @@ extern "C" {
 
 // -- System headers --
 
-#include <stdio.h>
-#include <stdlib.h>
-#include <math.h>
-
-// gettimeofday() needs this.
-#include <sys/time.h>
-#include <time.h>
+#include "bli_system.h"
 
 
 // -- BLIS definitions --

test/Makefile

@@ -158,9 +158,9 @@ TEST_OBJS      := $(patsubst $(TEST_SRC_PATH)/%.c, \
 CFLAGS         += -I$(BLIS_INC_PATH) -I$(TEST_SRC_PATH)
 
 LINKER         := $(CC)
-LDFLAGS        := -L/usr/lib/gcc/x86_64-redhat-linux/4.1.2 -L/usr/lib/gcc/x86_64-redhat-linux/4.1.2/../../../../lib64 -L/lib/../lib64 -L/usr/lib/../lib64 -lgfortranbegin -lgfortran -lm
-LDFLAGS        += -lpthread
-#LDFLAGS        := -L/usr/lib/gcc/i486-linux-gnu/4.4.3 -L/usr/lib/gcc/i486-linux-gnu/4.4.3/../../../../lib -L/lib/../lib -L/usr/lib/../lib -L/usr/lib/gcc/i486-linux-gnu/4.4.3/../../.. -L/usr/lib/i486-linux-gnu -lgfortranbegin -lgfortran -lm
+#LDFLAGS        := -L/usr/lib/gcc/x86_64-redhat-linux/4.1.2 -L/usr/lib/gcc/x86_64-redhat-linux/4.1.2/../../../../lib64 -L/lib/../lib64 -L/usr/lib/../lib64 -lgfortranbegin -lgfortran -lm
+#LDFLAGS        += -lpthread
+LDFLAGS        := -L/usr/lib/gcc/i486-linux-gnu/4.4.3 -L/usr/lib/gcc/i486-linux-gnu/4.4.3/../../../../lib -L/lib/../lib -L/usr/lib/../lib -L/usr/lib/gcc/i486-linux-gnu/4.4.3/../../.. -L/usr/lib/i486-linux-gnu -lgfortranbegin -lgfortran -lm
 
 
 
@@ -274,5 +274,5 @@ test_%_blis.x: test_%_blis.o $(BLIS_LIB)
 clean: cleanx
 
 cleanx:
-	- $(RM_F) *.x
+	- $(RM_F) *.o *.x
 

test/test_gemm.c

@@ -290,8 +290,11 @@ int main( int argc, char** argv )
 #else
 		printf( "data_gemm_%s", BLAS );
 #endif
-		printf( "( %2ld, 1:5 ) = [ %4lu %4lu %4lu  %10.3e  %6.3f ];\n",
-		        (p - p_begin + 1)/p_inc + 1, m, k, n, dtime_save, gflops );
+		printf( "( %2lu, 1:5 ) = [ %4lu %4lu %4lu  %10.3e  %6.3f ];\n",
+		        ( unsigned long )(p - p_begin + 1)/p_inc + 1,
+		        ( unsigned long )m,
+		        ( unsigned long )k,
+		        ( unsigned long )n, dtime_save, gflops );
 
 #if 0
 		bli_obj_release_pack( &a_pack );

test/test_gemv.c

@@ -197,8 +197,10 @@ int main( int argc, char** argv )
 #else
 		printf( "data_gemv_%s", BLAS );
 #endif
-		printf( "( %2ld, 1:4 ) = [ %4lu %4lu  %10.3e  %6.3f ];\n",
-		        (p - p_begin + 1)/p_inc + 1, m, n, dtime_save, gflops );
+		printf( "( %2lu, 1:4 ) = [ %4lu %4lu  %10.3e  %6.3f ];\n",
+		        ( unsigned long )(p - p_begin + 1)/p_inc + 1,
+		        ( unsigned long )m,
+		        ( unsigned long )n, dtime_save, gflops );
 
 		bli_obj_free( &alpha );
 		bli_obj_free( &beta );

test/test_ger.c

@@ -166,8 +166,10 @@ int main( int argc, char** argv )
 #else
 		printf( "data_ger_%s", BLAS );
 #endif
-		printf( "( %2ld, 1:4 ) = [ %4lu %4lu  %10.3e  %6.3f ];\n",
-		        (p - p_begin + 1)/p_inc + 1, m, n, dtime_save, gflops );
+		printf( "( %2lu, 1:4 ) = [ %4lu %4lu  %10.3e  %6.3f ];\n",
+		        ( unsigned long )(p - p_begin + 1)/p_inc + 1,
+		        ( unsigned long )m,
+		        ( unsigned long )n, dtime_save, gflops );
 
 		bli_obj_free( &alpha );
 

test/test_hemm.c

@@ -299,8 +299,10 @@ int main( int argc, char** argv )
 #else
 		printf( "data_hemm_%s", BLAS );
 #endif
-		printf( "( %2ld, 1:4 ) = [ %4lu %4lu  %10.3e  %6.3f ];\n",
-		        (p - p_begin + 1)/p_inc + 1, m, n, dtime_save, gflops );
+		printf( "( %2lu, 1:4 ) = [ %4lu %4lu  %10.3e  %6.3f ];\n",
+		        ( unsigned long )(p - p_begin + 1)/p_inc + 1,
+		        ( unsigned long )m,
+		        ( unsigned long )n, dtime_save, gflops );
 
 #if 0
 		bli_blksz_obj_free( mr );

test/test_hemv.c

@@ -183,8 +183,9 @@ int main( int argc, char** argv )
 #else
 		printf( "data_hemv_%s", BLAS );
 #endif
-		printf( "( %2ld, 1:3 ) = [ %4lu  %10.3e  %6.3f ];\n",
-		        (p - p_begin + 1)/p_inc + 1, m, dtime_save, gflops );
+		printf( "( %2lu, 1:3 ) = [ %4lu  %10.3e  %6.3f ];\n",
+		        ( unsigned long )(p - p_begin + 1)/p_inc + 1,
+		        ( unsigned long )m, dtime_save, gflops );
 
 		bli_obj_free( &alpha );
 		bli_obj_free( &beta );

test/test_her.c

@@ -173,8 +173,9 @@ int main( int argc, char** argv )
 #else
 		printf( "data_her_%s", BLAS );
 #endif
-		printf( "( %2ld, 1:3 ) = [ %4lu  %10.3e  %6.3f ];\n",
-		        (p - p_begin + 1)/p_inc + 1, m, dtime_save, gflops );
+		printf( "( %2lu, 1:3 ) = [ %4lu  %10.3e  %6.3f ];\n",
+		        ( unsigned long )(p - p_begin + 1)/p_inc + 1,
+		        ( unsigned long )m, dtime_save, gflops );
 
 		bli_obj_free( &alpha );
 

test/test_her2.c

@@ -178,8 +178,9 @@ int main( int argc, char** argv )
 #else
 		printf( "data_her2_%s", BLAS );
 #endif
-		printf( "( %2ld, 1:3 ) = [ %4lu  %10.3e  %6.3f ];\n",
-		        (p - p_begin + 1)/p_inc + 1, m, dtime_save, gflops );
+		printf( "( %2lu, 1:3 ) = [ %4lu  %10.3e  %6.3f ];\n",
+		        ( unsigned long )(p - p_begin + 1)/p_inc + 1,
+		        ( unsigned long )m, dtime_save, gflops );
 
 		bli_obj_free( &alpha );
 

test/test_her2k.c

@@ -116,7 +116,7 @@ int main( int argc, char** argv )
 		bli_obj_create( dt_beta,  1, 1, 0, 0, &beta );
 
 		bli_obj_create( dt_a, m, k, 0, 0, &a );
-		bli_obj_create( dt_a, m, k, 0, 0, &b );
+		bli_obj_create( dt_b, m, k, 0, 0, &b );
 		bli_obj_create( dt_c, m, m, 0, 0, &c );
 		bli_obj_create( dt_c, m, m, 0, 0, &c_save );
 
@@ -286,8 +286,10 @@ int main( int argc, char** argv )
 #else
 		printf( "data_her2k_%s", BLAS );
 #endif
-		printf( "( %2ld, 1:4 ) = [ %4lu %4lu  %10.3e  %6.3f ];\n",
-		        (p - p_begin + 1)/p_inc + 1, m, k, dtime_save, gflops );
+		printf( "( %2lu, 1:4 ) = [ %4lu %4lu  %10.3e  %6.3f ];\n",
+		        ( unsigned long )(p - p_begin + 1)/p_inc + 1,
+		        ( unsigned long )m,
+		        ( unsigned long )k, dtime_save, gflops );
 
 #if 0
 		bli_blksz_obj_free( mr );

test/test_herk.c

@@ -49,7 +49,7 @@ int main( int argc, char** argv )
 	dim_t p;
 	dim_t p_begin, p_end, p_inc;
 	int   m_input, k_input;
-	num_t dt_a, dt_b, dt_c;
+	num_t dt_a, dt_c;
 	num_t dt_alpha, dt_beta;
 	int   r, n_repeats;
 
@@ -98,7 +98,6 @@ int main( int argc, char** argv )
 #endif
 
 	dt_a = BLIS_DOUBLE;
-	dt_b = BLIS_DOUBLE;
 	dt_c = BLIS_DOUBLE;
 	dt_alpha = BLIS_DOUBLE;
 	dt_beta = BLIS_DOUBLE;
@@ -278,8 +277,10 @@ int main( int argc, char** argv )
 #else
 		printf( "data_herk_%s", BLAS );
 #endif
-		printf( "( %2ld, 1:4 ) = [ %4lu %4lu  %10.3e  %6.3f ];\n",
-		        (p - p_begin + 1)/p_inc + 1, m, k, dtime_save, gflops );
+		printf( "( %2lu, 1:4 ) = [ %4lu %4lu  %10.3e  %6.3f ];\n",
+		        ( unsigned long )(p - p_begin + 1)/p_inc + 1,
+		        ( unsigned long )m,
+		        ( unsigned long )k, dtime_save, gflops );
 
 #if 0
 		bli_blksz_obj_free( mr );