[Gmp-commit] /var/hg/gmp: 7 new changesets
mercurial at gmplib.org
mercurial at gmplib.org
Tue Nov 22 17:14:39 CET 2011
details: /var/hg/gmp/rev/3e51a0916ae6
changeset: 14463:3e51a0916ae6
user: Torbjorn Granlund <tege at gmplib.org>
date: Tue Nov 22 14:30:39 2011 +0100
description:
Retune.
details: /var/hg/gmp/rev/deeee90516e6
changeset: 14464:deeee90516e6
user: Torbjorn Granlund <tege at gmplib.org>
date: Tue Nov 22 16:57:06 2011 +0100
description:
Add more cycle numbers.
details: /var/hg/gmp/rev/41ecdd46db4c
changeset: 14465:41ecdd46db4c
user: Torbjorn Granlund <tege at gmplib.org>
date: Tue Nov 22 16:58:46 2011 +0100
description:
Align loop for slightly better power5 performance.
details: /var/hg/gmp/rev/3f75a16a665a
changeset: 14466:3f75a16a665a
user: Torbjorn Granlund <tege at gmplib.org>
date: Tue Nov 22 17:01:22 2011 +0100
description:
Add more cycle numbers.
details: /var/hg/gmp/rev/f58c57bbc481
changeset: 14467:f58c57bbc481
user: Torbjorn Granlund <tege at gmplib.org>
date: Tue Nov 22 17:03:17 2011 +0100
description:
Add more cycle numbers.
details: /var/hg/gmp/rev/ffae33a12bbc
changeset: 14468:ffae33a12bbc
user: Torbjorn Granlund <tege at gmplib.org>
date: Tue Nov 22 17:14:32 2011 +0100
description:
Don't fail fat builds under 64-bit DOS.
details: /var/hg/gmp/rev/17be5ae5aaec
changeset: 14469:17be5ae5aaec
user: Torbjorn Granlund <tege at gmplib.org>
date: Tue Nov 22 17:14:35 2011 +0100
description:
*** empty log message ***
diffstat:
ChangeLog | 5 ++++
configure.in | 14 +++++++++--
mpn/powerpc32/aors_n.asm | 19 +++++++++------
mpn/powerpc64/mode64/aors_n.asm | 5 ++-
mpn/powerpc64/mode64/p6/gmp-mparam.h | 44 ++++++++++++++++++------------------
mpn/x86/tabselect.asm | 4 +-
mpn/x86_64/tabselect.asm | 5 ++-
7 files changed, 57 insertions(+), 39 deletions(-)
diffs (209 lines):
diff -r 90b7d0b13cf3 -r 17be5ae5aaec ChangeLog
--- a/ChangeLog Tue Nov 22 00:11:49 2011 +0100
+++ b/ChangeLog Tue Nov 22 17:14:35 2011 +0100
@@ -1,5 +1,10 @@
2011-11-22 Torbjorn Granlund <tege at gmplib.org>
+ * configure.in: Don't fail fat builds under 64-bit DOS.
+
+ * mpn/powerpc64/mode64/aors_n.asm: Align loop for slightly better
+ power5 performance.
+
* mpn/powerpc64/mode64/p6/addmul_1.asm: New file.
2011-11-21 Torbjorn Granlund <tege at gmplib.org>
diff -r 90b7d0b13cf3 -r 17be5ae5aaec configure.in
--- a/configure.in Tue Nov 22 00:11:49 2011 +0100
+++ b/configure.in Tue Nov 22 17:14:35 2011 +0100
@@ -1930,9 +1930,17 @@
if test "$abi" = 64; then
gcc_64_cflags=""
- extra_functions_64="$extra_functions_64 fat fat_entry"
- path_64="x86_64/fat x86_64"
- fat_path="x86_64 x86_64/fat x86_64/pentium4 x86_64/core2 x86_64/coreinhm x86_64/coreisbr x86_64/atom x86_64/nano"
+ case $host in
+ *-*-mingw* | *-*-cygwin)
+ path_64="" # Windows amd64 calling conventions are *different*
+ fat_path=""
+ ;;
+ *)
+ extra_functions_64="$extra_functions_64 fat fat_entry"
+ path_64="x86_64/fat x86_64"
+ fat_path="x86_64 x86_64/fat x86_64/pentium4 x86_64/core2 x86_64/coreinhm x86_64/coreisbr x86_64/atom x86_64/nano"
+ ;;
+ esac
fi
fat_functions="add_n addmul_1 copyd copyi
diff -r 90b7d0b13cf3 -r 17be5ae5aaec mpn/powerpc32/aors_n.asm
--- a/mpn/powerpc32/aors_n.asm Tue Nov 22 00:11:49 2011 +0100
+++ b/mpn/powerpc32/aors_n.asm Tue Nov 22 17:14:35 2011 +0100
@@ -19,14 +19,17 @@
include(`../config.m4')
-C cycles/limb
-C 603e: ?
-C 604e: ? old: 3.25
-C 75x (G3): ? old: 3.5
-C 7400,7410 (G4): 3.25
-C 744x,745x (G4+): 4
-C power4/ppc970: ? old: 2.0
-C power5: ? old: 2.5
+C cycles/limb
+C 603e: ?
+C 604e: ? old: 3.25
+C 75x (G3): ? old: 3.5
+C 7400,7410 (G4): 3.25
+C 744x,745x (G4+): 4
+C POWER3/PPC630 2
+C POWER4/PPC970 2.4
+C POWER5 2.75
+C POWER6 40-140
+C POWER7 3
C INPUT PARAMETERS
define(`rp', `r3')
diff -r 90b7d0b13cf3 -r 17be5ae5aaec mpn/powerpc64/mode64/aors_n.asm
--- a/mpn/powerpc64/mode64/aors_n.asm Tue Nov 22 00:11:49 2011 +0100
+++ b/mpn/powerpc64/mode64/aors_n.asm Tue Nov 22 17:14:35 2011 +0100
@@ -1,6 +1,6 @@
dnl PowerPC-64 mpn_add_n/mpn_sub_n -- mpn addition and subtraction.
-dnl Copyright 1999, 2000, 2001, 2003, 2004, 2005, 2007 Free Software
+dnl Copyright 1999, 2000, 2001, 2003, 2004, 2005, 2007, 2011 Free Software
dnl Foundation, Inc.
dnl This file is part of the GNU MP Library.
@@ -23,7 +23,7 @@
C cycles/limb
C POWER3/PPC630 1.5
C POWER4/PPC970 2
-C POWER5 2.25
+C POWER5 2
C POWER6 2.63
C POWER7 2.25-2.87
@@ -137,6 +137,7 @@
addi r4, r4, 32
addi r5, r5, 32
+ ALIGN(16)
L(top): ADDSUBC r28, r7, r6
ld r6, 0(r4) C load s1 limb
ld r7, 0(r5) C load s2 limb
diff -r 90b7d0b13cf3 -r 17be5ae5aaec mpn/powerpc64/mode64/p6/gmp-mparam.h
--- a/mpn/powerpc64/mode64/p6/gmp-mparam.h Tue Nov 22 00:11:49 2011 +0100
+++ b/mpn/powerpc64/mode64/p6/gmp-mparam.h Tue Nov 22 17:14:35 2011 +0100
@@ -1,7 +1,7 @@
/* POWER6 gmp-mparam.h -- Compiler/machine parameter header file.
-Copyright 1991, 1993, 1994, 1999, 2000, 2001, 2002, 2003, 2009, 2010 Free
-Software Foundation, Inc.
+Copyright 1991, 1993, 1994, 1999, 2000, 2001, 2002, 2003, 2009, 2010, 2011
+Free Software Foundation, Inc.
This file is part of the GNU MP Library.
@@ -53,7 +53,7 @@
#define SQR_TOOM6_THRESHOLD 274
#define SQR_TOOM8_THRESHOLD 410
-#define MULMID_TOOM42_THRESHOLD 24
+#define MULMID_TOOM42_THRESHOLD 36
#define MULMOD_BNM1_THRESHOLD 14
#define SQRMOD_BNM1_THRESHOLD 14
@@ -111,36 +111,36 @@
#define SQR_FFT_THRESHOLD 2368
#define MULLO_BASECASE_THRESHOLD 5
-#define MULLO_DC_THRESHOLD 28
-#define MULLO_MUL_N_THRESHOLD 3084
+#define MULLO_DC_THRESHOLD 61
+#define MULLO_MUL_N_THRESHOLD 5558
-#define DC_DIV_QR_THRESHOLD 23
+#define DC_DIV_QR_THRESHOLD 29
#define DC_DIVAPPR_Q_THRESHOLD 112
-#define DC_BDIV_QR_THRESHOLD 29
-#define DC_BDIV_Q_THRESHOLD 79
+#define DC_BDIV_QR_THRESHOLD 70
+#define DC_BDIV_Q_THRESHOLD 168
-#define INV_MULMOD_BNM1_THRESHOLD 51
+#define INV_MULMOD_BNM1_THRESHOLD 61
#define INV_NEWTON_THRESHOLD 93
#define INV_APPR_THRESHOLD 91
-#define BINV_NEWTON_THRESHOLD 132
-#define REDC_1_TO_REDC_N_THRESHOLD 39
+#define BINV_NEWTON_THRESHOLD 222
+#define REDC_1_TO_REDC_N_THRESHOLD 63
-#define MU_DIV_QR_THRESHOLD 855
+#define MU_DIV_QR_THRESHOLD 807
#define MU_DIVAPPR_Q_THRESHOLD 807
-#define MUPI_DIV_QR_THRESHOLD 23
-#define MU_BDIV_QR_THRESHOLD 807
-#define MU_BDIV_Q_THRESHOLD 872
+#define MUPI_DIV_QR_THRESHOLD 27
+#define MU_BDIV_QR_THRESHOLD 872
+#define MU_BDIV_Q_THRESHOLD 1078
#define MATRIX22_STRASSEN_THRESHOLD 13
-#define HGCD_THRESHOLD 69
-#define HGCD_APPR_THRESHOLD 50
+#define HGCD_THRESHOLD 94
+#define HGCD_APPR_THRESHOLD 55
#define HGCD_REDUCE_THRESHOLD 2121
-#define GCD_DC_THRESHOLD 268
-#define GCDEXT_DC_THRESHOLD 209
+#define GCD_DC_THRESHOLD 253
+#define GCDEXT_DC_THRESHOLD 217
#define JACOBI_BASE_METHOD 4
-#define GET_STR_DC_THRESHOLD 17
-#define GET_STR_PRECOMPUTE_THRESHOLD 27
+#define GET_STR_DC_THRESHOLD 16
+#define GET_STR_PRECOMPUTE_THRESHOLD 29
#define SET_STR_DC_THRESHOLD 532
-#define SET_STR_PRECOMPUTE_THRESHOLD 1648
+#define SET_STR_PRECOMPUTE_THRESHOLD 1561
diff -r 90b7d0b13cf3 -r 17be5ae5aaec mpn/x86/tabselect.asm
--- a/mpn/x86/tabselect.asm Tue Nov 22 00:11:49 2011 +0100
+++ b/mpn/x86/tabselect.asm Tue Nov 22 17:14:35 2011 +0100
@@ -27,12 +27,12 @@
C P6 model 13 (Dothan) ?
C P4 model 0 (Willamette) ?
C P4 model 1 (?) ?
-C P4 model 2 (Northwood) ?
+C P4 model 2 (Northwood) 4.5
C P4 model 3 (Prescott) ?
C P4 model 4 (Nocona) ?
C Intel Atom ?
C AMD K6 ?
-C AMD K7 ?
+C AMD K7 3.4
C AMD K8 ?
C AMD K10 ?
diff -r 90b7d0b13cf3 -r 17be5ae5aaec mpn/x86_64/tabselect.asm
--- a/mpn/x86_64/tabselect.asm Tue Nov 22 00:11:49 2011 +0100
+++ b/mpn/x86_64/tabselect.asm Tue Nov 22 17:14:35 2011 +0100
@@ -23,11 +23,12 @@
C cycles/limb
C AMD K8,K9 2.5
C AMD K10 2.5
+C AMD bobcat 3.5
C Intel P4 4
-C Intel core2 2.3
+C Intel core2 2.33
C Intel NHM 2.5
C Intel SBR 2.2
-C Intel atom ?
+C Intel atom 5
C VIA nano 3.5
C NOTES
More information about the gmp-commit
mailing list