1. Update Windows Powershell scripts in mpirbench to refer to MPIR rather than GMP

2. Update MPIR version number in Windows config files to 1.1.2 3. Add an MSVC inline definition in in gmp-h.in 4. Correct locale test (as per GMP correction) 5. Add Windows x64 set/copy intrinsics to mul_fft.c (improves FFT speed score by 2%)
2009-05-09 13:26:27 +00:00 · 2009-05-09 13:26:27 +00:00 · d942415a1c
commit d942415a1c
parent 0ba06242c6
13 changed files with 65 additions and 47 deletions
--- a/bench/build.vc9/runbench.32.ps1
+++ b/bench/build.vc9/runbench.32.ps1
@ -40,19 +40,19 @@ foreach ($cat in $tests)
 	    	"      $t $parms" | Out-Host
 			$out = Invoke-Expression ".\$dir\$t $parms" 
 			foreach ($s in $out) { if($s -match "RESULT: [0-9]*") { $res = $s.Split(' ')[1] } }
-			"      GMPbench.$cat.$t.$pcm result: $res" | Out-Host
+			"      MPIRbench.$cat.$t.$pcm result: $res" | Out-Host
 			$acc = Invoke-Expression ".\$dir\gexpr -prec 10 `"$acc*$res`""
 			$n++
 		}
 		$out = Invoke-Expression ".\$dir\gexpr `"$acc^(1/$n)`""
-		"    GMPbench.$cat.$t result: $out" | Out-Host
+		"    MPIRbench.$cat.$t result: $out" | Out-Host
 		$acc1 = Invoke-Expression ".\$dir\gexpr -prec 10 `"$acc1*$acc^(1/$n)`""
 		$n1++
 	}
 	$out = Invoke-Expression ".\$dir\gexpr `"$acc1^(1/$n1)`""
-	"  GMPbench.$cat result: $out" | Out-Host
+	"  MPIRbench.$cat result: $out" | Out-Host
 	$acc2 = Invoke-Expression ".\$dir\gexpr -round -prec 10 `"$acc2*$acc1^(1/$n1)`""
 	$n2++
 }
 $out = Invoke-Expression ".\$dir\gexpr -round -prec 10 `"$acc2^(1/$n2)`""
-"GMPbench result: $out" | Out-Host
+"MPIRbench result: $out" | Out-Host
--- a/bench/build.vc9/runbench.64.ps1
+++ b/bench/build.vc9/runbench.64.ps1
@ -40,19 +40,19 @@ foreach ($cat in $tests)
 	    	"      $t $parms" | Out-Host
 			$out = Invoke-Expression ".\$dir\$t $parms" 
 			foreach ($s in $out) { if($s -match "RESULT: [0-9]*") { $res = $s.Split(' ')[1] } }
-			"      GMPbench.$cat.$t.$pcm result: $res" | Out-Host
+			"      MPIRbench.$cat.$t.$pcm result: $res" | Out-Host
 			$acc = Invoke-Expression ".\$dir\gexpr -prec 10 `"$acc*$res`""
 			$n++
 		}
 		$out = Invoke-Expression ".\$dir\gexpr `"$acc^(1/$n)`""
-		"    GMPbench.$cat.$t result: $out" | Out-Host
+		"    MPIRbench.$cat.$t result: $out" | Out-Host
 		$acc1 = Invoke-Expression ".\$dir\gexpr -prec 10 `"$acc1*$acc^(1/$n)`""
 		$n1++
 	}
 	$out = Invoke-Expression ".\$dir\gexpr `"$acc1^(1/$n1)`""
-	"  GMPbench.$cat result: $out" | Out-Host
+	"  MPIRbench.$cat result: $out" | Out-Host
 	$acc2 = Invoke-Expression ".\$dir\gexpr -round -prec 10 `"$acc2*$acc1^(1/$n1)`""
 	$n2++
 }
 $out = Invoke-Expression ".\$dir\gexpr -round -prec 10 `"$acc2^(1/$n2)`""
-"GMPbench result: $out" | Out-Host
+"MPIRbench result: $out" | Out-Host
--- a/build.vc9/config.amd64
+++ b/build.vc9/config.amd64
@ -479,7 +479,7 @@ MA 02111-1307, USA. */

 /* Version number of package */
 #define GMP_VERSION	    "4.2.1"
-#define VERSION			"1.0.0"
+#define VERSION			"1.1.2"

 /* ./configure --enable-assert option, to enable some ASSERT()s */
 #undef WANT_ASSERT
--- a/build.vc9/config.core2
+++ b/build.vc9/config.core2
@ -471,7 +471,7 @@ MA 02111-1307, USA. */

 /* Version number of package */
 #define GMP_VERSION	    "4.2.1"
-#define VERSION			"1.0.0"
+#define VERSION			"1.1.2"

 /* ./configure --enable-assert option, to enable some ASSERT()s */
 #undef WANT_ASSERT
--- a/build.vc9/config.gc
+++ b/build.vc9/config.gc
@ -468,7 +468,7 @@ MA 02111-1307, USA. */

 /* Version number of package */
 #define GMP_VERSION	    "4.2.1"
-#define VERSION			"1.0.0"
+#define VERSION			"1.1.2"

 /* ./configure --enable-assert option, to enable some ASSERT()s */
 #undef WANT_ASSERT
--- a/build.vc9/config.p0
+++ b/build.vc9/config.p0
@ -493,7 +493,7 @@ MA 02111-1307, USA. */

 /* Version number of package */
 #define GMP_VERSION	    "4.2.1"
-#define VERSION			"1.0.0"
+#define VERSION			"1.1.2"

 /* ./configure --enable-assert option, to enable some ASSERT()s */
 #undef WANT_ASSERT
--- a/build.vc9/config.p3
+++ b/build.vc9/config.p3
@ -499,7 +499,7 @@ MA 02111-1307, USA. */

 /* Version number of package */
 #define GMP_VERSION	    "4.2.1"
-#define VERSION			"1.0.0"
+#define VERSION			"1.1.2"

 /* ./configure --enable-assert option, to enable some ASSERT()s */
 #undef WANT_ASSERT
--- a/build.vc9/config.p4
+++ b/build.vc9/config.p4
@ -503,7 +503,7 @@ MA 02111-1307, USA. */

 /* Version number of package */
 #define GMP_VERSION	    "4.2.1"
-#define VERSION			"1.0.0"
+#define VERSION			"1.1.2"

 /* ./configure --enable-assert option, to enable some ASSERT()s */
 #undef WANT_ASSERT
--- a/gmp-h.in
+++ b/gmp-h.in
@ -442,6 +442,10 @@ typedef __mpq_struct *mpq_ptr;
 #endif
 #endif

+#ifdef _MSC_VER
+#define __GMP_EXTERN_INLINE __inline
+#endif
+
 /* DEC C (eg. version 5.9) supports "static __inline foo()", even in -std1
   strict ANSI mode.  Inlining is done even when not optimizing (ie. -O0
   mode, which is the default), but an unnecessary local copy of foo is
--- a/mpn/generic/mul_fft.c
+++ b/mpn/generic/mul_fft.c
@ -116,6 +116,10 @@ MPN_FFT_ZERO (void *dst, long int n)
    __asm__ __volatile__ ("rep stosl" : "+c" (n), "+D" (dst) : "a" (0) : 
 			  "memory");
 }
+#elif defined( _MSC_VER ) && defined( _M_X64 )
+#include <intrin.h>
+#pragma intrinsic(__stosq)
+#define MPN_FFT_ZERO(d, l) __stosq(d, 0, l)
 #else
  /* Fall back to GMP's MPN_ZERO() macro */
  #define MPN_FFT_ZERO(dst, n) MPN_ZERO(dst,n)
@ -140,6 +144,10 @@ MPN_FFT_STORE (void *dst, long int n, long int d)
    __asm__ __volatile__ ("rep stosl" : "+c" (n), "+D" (dst) : "a" (d) :
 			  "memory");
 }
+#elif defined( _MSC_VER ) && defined( _M_X64 )
+#include <intrin.h>
+#pragma intrinsic(__stosq)
+#define MPN_FFT_STORE(d, l, v) __stosq(d, v, l)
 #else
 void static inline
 MPN_FFT_STORE (mp_limb_t *dst, mp_size_t n, mp_limb_t d)
@ -168,6 +176,10 @@ MPN_FFT_COPY (void *dst, const void *src, long int n)
    __asm__ __volatile__ ("rep movsl" : "+c" (n), "+S" (src), "+D" (dst) :
 			  "memory");
 }
+#elif defined( _MSC_VER ) && defined( _M_X64 )
+#include <intrin.h>
+#pragma intrinsic(__movsq)
+#define MPN_FFT_COPY(d, s, l) __movsq(d, s, l)
 #else
  /* Fall back to GMP's MPN_COPY() macro */
  #define MPN_FFT_COPY(dst, src, n) MPN_COPY(dst,src,n)
--- a/mpn/x86_64w/amd64/gmp-mparam.h
+++ b/mpn/x86_64w/amd64/gmp-mparam.h
@ -1,24 +1,24 @@
-/* Generated by tuneup.c, 2009-05-06, system compiler */
+/* Generated by tuneup.c, 2009-05-08, system compiler */

-#define MUL_KARATSUBA_THRESHOLD          26
-#define MUL_TOOM3_THRESHOLD             168
-#define MUL_TOOM4_THRESHOLD             842
-#define MUL_TOOM7_THRESHOLD             842
+#define MUL_KARATSUBA_THRESHOLD          28
+#define MUL_TOOM3_THRESHOLD             129
+#define MUL_TOOM4_THRESHOLD             746
+#define MUL_TOOM7_THRESHOLD             746

 #define SQR_BASECASE_THRESHOLD            0  /* always (native) */
 #define SQR_KARATSUBA_THRESHOLD          34
 #define SQR_TOOM3_THRESHOLD             218

-#define MULLOW_BASECASE_THRESHOLD         7
-#define MULLOW_DC_THRESHOLD              99
-#define MULLOW_MUL_N_THRESHOLD          210
+#define MULLOW_BASECASE_THRESHOLD        38
+#define MULLOW_DC_THRESHOLD              87
+#define MULLOW_MUL_N_THRESHOLD          177

 #define DIV_SB_PREINV_THRESHOLD           0  /* always */
-#define DIV_DC_THRESHOLD                108
-#define POWM_THRESHOLD                  170
+#define DIV_DC_THRESHOLD                106
+#define POWM_THRESHOLD                  190

-#define GCD_ACCEL_THRESHOLD               3
-#define GCDEXT_THRESHOLD                224
+#define GCD_ACCEL_THRESHOLD              14
+#define GCDEXT_THRESHOLD                208
 #define JACOBI_BASE_METHOD                2

 #define DIVREM_1_NORM_THRESHOLD       MP_SIZE_T_MAX  /* never */
@ -35,15 +35,15 @@
 #define GET_STR_PRECOMPUTE_THRESHOLD     11
 #define SET_STR_THRESHOLD              8540

-#define MUL_FFT_TABLE  { 464, 1056, 3392, 3840, 7168, 28672, 0 }
+#define MUL_FFT_TABLE  { 432, 1056, 3264, 3840, 7168, 36864, 0 }
 #define MUL_FFT_MODF_THRESHOLD         1488
 #define MUL_FFT_THRESHOLD             14336

-#define SQR_FFT_TABLE  { 560, 1056, 3392, 3840, 7168, 36864, 0 }
+#define SQR_FFT_TABLE  { 464, 1056, 3264, 3840, 7168, 36864, 0 }
 #define SQR_FFT_MODF_THRESHOLD         1488
-#define SQR_FFT_THRESHOLD             13120
+#define SQR_FFT_THRESHOLD              8916

-/* Tuneup completed successfully, took 1381 seconds */
+/* Tuneup completed successfully, took 1073 seconds */

 #define MUL_FFT_TABLE2 {{1, 2}, {215, 3}, {220, 2}, {404, 3}, {413, 2}, {443, 3}, {453, 2}, {507, 3}, {519, 2}, {695, 3}, {711, 2}, {727, 3}, {777, 2}, {831, 3}, {850, 2}, {1015, 3}, {1038, 2}, {1185, 3}, {1211, 2}, {1266, 3}, {1352, 2}, {1382, 3}, {1476, 4}, {1509, 2}, {1543, 3}, {1577, 2}, {1612, 3}, {1648, 2}, {2340, 3}, {2392, 2}, {2499, 3}, {2610, 2}, {2668, 3}, {2787, 2}, {3248, 3}, {3393, 2}, {3468, 3}, {3544, 2}, {3622, 3}, {3702, 2}, {3784, 3}, {3867, 4}, {3952, 3}, {4128, 2}, {4914, 3}, {5022, 4}, {5132, 3}, {5245, 2}, {5360, 3}, {5478, 4}, {5598, 2}, {5721, 3}, {6107, 2}, {6378, 3}, {6661, 2}, {6807, 3}, {6957, 2}, {7110, 3}, {7426, 2}, {7926, 3}, {8100, 4}, {8460, 2}, {9030, 3}, {9228, 4}, {9431, 2}, {9850, 3}, {10287, 2}, {10744, 3}, {10980, 4}, {11221, 3}, {11467, 4}, {11976, 3}, {13061, 4}, {13640, 3}, {13939, 4}, {14245, 3}, {14557, 4}, {14876, 5}, {15202, 4}, {16224, 3}, {17316, 4}, {18084, 2}, {18480, 3}, {19299, 2}, {19722, 3}, {20154, 2}, {20596, 3}, {21508, 4}, {21979, 3}, {22461, 2}, {23970, 3}, {24495, 4}, {25032, 5}, {25581, 6}, {26142, 7}, {26715, 8}, {27300, 9}, {28509, 7}, {29134, 8}, {31091, 9}, {33179, 10}, {33906, 9}, {37787, 10}, {39461, 9}, {42113, 10}, {43036, 9}, {43979, 10}, {45928, 11}, {46934, 10}, {47962, 9}, {51184, 10}, {52305, 9}, {54622, 10}, {55819, 8}, {57042, 9}, {59569, 10}, {60874, 11}, {62207, 12}, {63570, 10}, {64963, 9}, {67840, 10}, {72397, 9}, {75604, 10}, {77260, 9}, {78952, 10}, {86099, 11}, {87985, 10}, {91881, 11}, {93893, 10}, {95949, 11}, {98051, 12}, {MP_SIZE_T_MAX,0}}

--- a/mpn/x86_64w/core2/gmp-mparam.h
+++ b/mpn/x86_64w/core2/gmp-mparam.h
@ -1,24 +1,24 @@
-/* Generated by tuneup.c, 2009-05-06, system compiler */
+/* Generated by tuneup.c, 2009-05-08, system compiler */

 #define MUL_KARATSUBA_THRESHOLD          16
-#define MUL_TOOM3_THRESHOLD             113
-#define MUL_TOOM4_THRESHOLD             502
-#define MUL_TOOM7_THRESHOLD             502
+#define MUL_TOOM3_THRESHOLD             114
+#define MUL_TOOM4_THRESHOLD             199
+#define MUL_TOOM7_THRESHOLD             390

 #define SQR_BASECASE_THRESHOLD            0  /* always (native) */
 #define SQR_KARATSUBA_THRESHOLD          26
-#define SQR_TOOM3_THRESHOLD             177
+#define SQR_TOOM3_THRESHOLD             176

-#define MULLOW_BASECASE_THRESHOLD         6
-#define MULLOW_DC_THRESHOLD              79
-#define MULLOW_MUL_N_THRESHOLD          378
+#define MULLOW_BASECASE_THRESHOLD         7
+#define MULLOW_DC_THRESHOLD              71
+#define MULLOW_MUL_N_THRESHOLD          339

 #define DIV_SB_PREINV_THRESHOLD           0  /* always */
-#define DIV_DC_THRESHOLD                 93
-#define POWM_THRESHOLD                   79
+#define DIV_DC_THRESHOLD                 88
+#define POWM_THRESHOLD                   89

-#define GCD_ACCEL_THRESHOLD              16
-#define GCDEXT_THRESHOLD                155
+#define GCD_ACCEL_THRESHOLD               3
+#define GCDEXT_THRESHOLD                173
 #define JACOBI_BASE_METHOD                1

 #define DIVREM_1_NORM_THRESHOLD       MP_SIZE_T_MAX  /* never */
@ -37,13 +37,13 @@

 #define MUL_FFT_TABLE  { 432, 864, 2368, 2816, 7168, 20480, 0 }
 #define MUL_FFT_MODF_THRESHOLD         1120
-#define MUL_FFT_THRESHOLD              5760
+#define MUL_FFT_THRESHOLD              6016

-#define SQR_FFT_TABLE  { 432, 928, 2496, 2816, 7168, 36864, 0 }
+#define SQR_FFT_TABLE  { 432, 928, 2496, 2816, 7168, 20480, 49152, 0 }
 #define SQR_FFT_MODF_THRESHOLD         1120
 #define SQR_FFT_THRESHOLD              5408

-/* Tuneup completed successfully, took 202 seconds */
+/* Tuneup completed successfully, took 194 seconds */

 #define MUL_FFT_TABLE2 {{1, 2}, {248, 3}, {254, 2}, {299, 3}, {306, 2}, {360, 3}, {368, 2}, {760, 3}, {777, 2}, {831, 3}, {869, 2}, {971, 3}, {993, 2}, {1015, 3}, {1038, 2}, {1085, 3}, {1109, 4}, {1159, 2}, {1211, 3}, {1238, 2}, {1266, 3}, {1294, 4}, {1323, 2}, {1444, 3}, {1476, 2}, {1799, 3}, {1839, 2}, {2098, 3}, {2144, 2}, {2392, 3}, {2445, 2}, {2668, 3}, {2727, 2}, {2787, 3}, {2849, 2}, {2912, 3}, {3042, 2}, {3109, 3}, {3248, 4}, {3320, 5}, {3468, 6}, {3544, 5}, {3622, 3}, {3784, 2}, {4504, 3}, {4603, 2}, {4808, 3}, {4914, 4}, {5022, 3}, {5245, 2}, {5478, 3}, {5721, 2}, {5976, 3}, {6107, 2}, {6378, 3}, {6518, 2}, {7110, 3}, {7756, 4}, {7926, 3}, {8100, 2}, {9431, 3}, {10066, 4}, {10287, 5}, {10513, 6}, {10744, 7}, {11221, 8}, {11467, 7}, {11719, 8}, {12507, 9}, {12781, 8}, {13347, 9}, {13939, 10}, {14876, 8}, {15202, 9}, {15876, 8}, {16224, 9}, {16944, 8}, {17316, 9}, {18480, 8}, {18885, 9}, {19722, 10}, {20596, 9}, {21047, 10}, {21508, 9}, {21979, 10}, {23970, 9}, {25581, 10}, {26715, 8}, {27300, 9}, {29134, 10}, {29772, 11}, {31091, 9}, {33179, 10}, {36977, 11}, {37787, 9}, {38615, 10}, {41210, 11}, {42113, 10}, {43036, 11}, {43979, 10}, {46934, 11}, {47962, 10}, {49013, 11}, {50087, 10}, {52305, 11}, {54622, 12}, {55819, 11}, {57042, 12}, {58292, 11}, {62207, 12}, {63570, 10}, {69326, 11}, {70845, 10}, {72397, 11}, {75604, 10}, {77260, 11}, {87985, 10}, {89912, 11}, {MP_SIZE_T_MAX,0}}

--- a/tests/cxx/t-locale.cc
+++ b/tests/cxx/t-locale.cc
@ -29,8 +29,10 @@ MA 02110-1301, USA. */

 using namespace std;

-
-char point_string[2];
+extern "C"
+{
+    char point_string[2];
+};

 #if HAVE_STD__LOCALE
 // Like std::numpunct, but with decimal_point coming from point_string[].