Merge branch 'master' into baserock/morph

2 files changed, 237 insertions, 50 deletions

diff --git a/mpn/ia64/gmp-mparam.h b/mpn/ia64/gmp-mparam.h
index 0841c82aa..77e02f518 100644
--- a/mpn/ia64/gmp-mparam.h
+++ b/mpn/ia64/gmp-mparam.h
@@ -1,6 +1,6 @@
 /* gmp-mparam.h -- Compiler/machine parameter header file.
 
-Copyright 2000, 2001, 2002, 2003, 2004, 2005, 2009, 2010 Free Software
+Copyright 2000, 2001, 2002, 2003, 2004, 2005, 2009, 2010, 2011 Free Software
 Foundation, Inc.
 
 This file is part of the GNU MP Library.
@@ -21,70 +21,94 @@ along with the GNU MP Library.  If not, see http://www.gnu.org/licenses/.  */
 #define GMP_LIMB_BITS 64
 #define BYTES_PER_MP_LIMB 8
 
-/* 1300MHz Itanium2 (babe.fsffrance.org) */
-
+/* 900MHz Itanium2 (titanic.gmplib.org) */
 
+#define MOD_1_1P_METHOD                      2
 #define MOD_1_NORM_THRESHOLD                 0  /* always */
 #define MOD_1_UNNORM_THRESHOLD               0  /* always */
 #define MOD_1N_TO_MOD_1_1_THRESHOLD          4
-#define MOD_1U_TO_MOD_1_1_THRESHOLD          8
-#define MOD_1_1_TO_MOD_1_2_THRESHOLD         0  /* never mpn_mod_1_1p */
-#define MOD_1_2_TO_MOD_1_4_THRESHOLD        21
-#define PREINV_MOD_1_TO_MOD_1_THRESHOLD     22
+#define MOD_1U_TO_MOD_1_1_THRESHOLD          5
+#define MOD_1_1_TO_MOD_1_2_THRESHOLD        26
+#define MOD_1_2_TO_MOD_1_4_THRESHOLD         0  /* never mpn_mod_1s_2p */
+#define PREINV_MOD_1_TO_MOD_1_THRESHOLD     10
 #define USE_PREINV_DIVREM_1                  1  /* native */
+#define DIV_QR_2_PI2_THRESHOLD              12
 #define DIVEXACT_1_THRESHOLD                 0  /* always (native) */
 #define BMOD_1_TO_MOD_1_THRESHOLD        MP_SIZE_T_MAX  /* never */
 
 #define MUL_TOOM22_THRESHOLD                40
-#define MUL_TOOM33_THRESHOLD               122
-#define MUL_TOOM44_THRESHOLD               212
+#define MUL_TOOM33_THRESHOLD               129
+#define MUL_TOOM44_THRESHOLD               214
 #define MUL_TOOM6H_THRESHOLD               318
 #define MUL_TOOM8H_THRESHOLD               430
 
-#define MUL_TOOM32_TO_TOOM43_THRESHOLD      93
-#define MUL_TOOM32_TO_TOOM53_THRESHOLD     146
-#define MUL_TOOM42_TO_TOOM53_THRESHOLD     129
+#define MUL_TOOM32_TO_TOOM43_THRESHOLD      97
+#define MUL_TOOM32_TO_TOOM53_THRESHOLD     145
+#define MUL_TOOM42_TO_TOOM53_THRESHOLD     126
 #define MUL_TOOM42_TO_TOOM63_THRESHOLD     151
 
 #define SQR_BASECASE_THRESHOLD              11
 #define SQR_TOOM2_THRESHOLD                 84
-#define SQR_TOOM3_THRESHOLD                125
+#define SQR_TOOM3_THRESHOLD                135
 #define SQR_TOOM4_THRESHOLD                494
-#define SQR_TOOM6_THRESHOLD                  0  /* never toom4 */
-#define SQR_TOOM8_THRESHOLD                  0  /* never toom6 */
+#define SQR_TOOM6_THRESHOLD                  0  /* always */
+#define SQR_TOOM8_THRESHOLD                  0  /* always */
+
+#define MULMID_TOOM42_THRESHOLD             98
 
 #define MULMOD_BNM1_THRESHOLD               23
-#define SQRMOD_BNM1_THRESHOLD               25
+#define SQRMOD_BNM1_THRESHOLD               28
+
+#define POWM_SEC_TABLE  2,29,130,905
 
-#define MUL_FFT_MODF_THRESHOLD             444  /* k = 5 */
+#define MUL_FFT_MODF_THRESHOLD             476  /* k = 5 */
 #define MUL_FFT_TABLE3                                      \
-  { {    444, 5}, {     27, 6}, {     14, 5}, {     29, 6}, \
-    {     35, 7}, {     18, 6}, {     37, 7}, {     19, 6}, \
+  { {    476, 5}, {     27, 6}, {     14, 5}, {     29, 6}, \
+    {     33, 7}, {     17, 6}, {     37, 7}, {     19, 6}, \
     {     39, 7}, {     21, 6}, {     43, 7}, {     33, 8}, \
     {     17, 7}, {     37, 8}, {     19, 7}, {     39, 8}, \
-    {     21, 7}, {     43, 8}, {     29, 9}, {     15, 8}, \
-    {     37, 9}, {     19, 8}, {     43, 9}, {     23, 8}, \
-    {     49, 9}, {     27, 8}, {     57, 9}, {     31, 8}, \
-    {     63, 9}, {     35, 8}, {     71, 9}, {     43,10}, \
+    {     21, 7}, {     43, 8}, {     37, 9}, {     19, 8}, \
+    {     43, 9}, {     23, 8}, {     51, 9}, {     27, 8}, \
+    {     57, 9}, {     31, 8}, {     63, 9}, {     43,10}, \
     {     23, 9}, {     59,10}, {     31, 9}, {     71,10}, \
-    {     39, 9}, {     87,10}, {     47, 9}, {     99,10}, \
+    {     39, 9}, {     83,10}, {     47, 9}, {     99,10}, \
     {     55,11}, {     31,10}, {     87,11}, {     47,10}, \
     {    111,12}, {     31,11}, {     63,10}, {    143,11}, \
     {     79,10}, {    167,11}, {     95,10}, {    191,11}, \
     {    111,12}, {     63,11}, {    143,10}, {    287, 9}, \
     {    575,10}, {    303,11}, {    159,10}, {    319,12}, \
     {     95,11}, {    191,10}, {    399,11}, {    207,10}, \
-    {    431,13}, {   8192,14}, {  16384,15}, {  32768,16}, \
-    {  65536,17}, { 131072,18}, { 262144,19}, { 524288,20}, \
-    {1048576,21}, {2097152,22}, {4194304,23}, {8388608,24} }
-#define MUL_FFT_TABLE3_SIZE 76
+    {    431,13}, {     63,12}, {    127,11}, {    271,10}, \
+    {    543,11}, {    287,10}, {    575,11}, {    303,12}, \
+    {    159,11}, {    335,10}, {    671,11}, {    367,12}, \
+    {    191,11}, {    399,10}, {    799,11}, {    431,12}, \
+    {    223,11}, {    447,13}, {    127,12}, {    255,11}, \
+    {    543,12}, {    287,11}, {    607,12}, {    319,11}, \
+    {    671,12}, {    351,11}, {    703,13}, {    191,12}, \
+    {    415,11}, {    863,12}, {    447,14}, {    127,13}, \
+    {    255,12}, {    607,13}, {    319,12}, {    735,13}, \
+    {    383,12}, {    799,11}, {   1599,12}, {    863,13}, \
+    {    447,12}, {    927,11}, {   1855,14}, {    255,13}, \
+    {    511,12}, {   1055,13}, {    575,12}, {   1215,13}, \
+    {    639,12}, {   1279,13}, {    703,14}, {    383,13}, \
+    {    767,12}, {   1535,13}, {    831,12}, {   1663,13}, \
+    {    895,12}, {   1791,15}, {    255,14}, {    511,13}, \
+    {   1087,12}, {   2175,13}, {   1215,14}, {    639,13}, \
+    {   1343,12}, {   2687,13}, {   1471,14}, {    767,13}, \
+    {   1599,12}, {   3199,13}, {   1663,14}, {    895,13}, \
+    {   1855,15}, {    511,14}, {   1023,13}, {   2175,14}, \
+    {   1151,13}, {   2431,14}, {   1279,13}, {   2687,14}, \
+    {   1407,15}, {    767,14}, {   1535,13}, {   3199,14}, \
+    {   1663,13}, {   3455,14}, {   1791,16}, {  65536,17}, \
+    { 131072,18}, { 262144,19}, { 524288,20}, {1048576,21}, \
+    {2097152,22}, {4194304,23}, {8388608,24} }
+#define MUL_FFT_TABLE3_SIZE 155
 #define MUL_FFT_THRESHOLD                 5760
 
-#define SQR_FFT_MODF_THRESHOLD             440  /* k = 5 */
+#define SQR_FFT_MODF_THRESHOLD             436  /* k = 5 */
 #define SQR_FFT_TABLE3                                      \
-  { {    440, 5}, {     14, 4}, {     29, 5}, {     29, 6}, \
-    {     15, 5}, {     31, 6}, {     35, 7}, {     18, 6}, \
-    {     37, 7}, {     33, 8}, {     17, 7}, {     37, 8}, \
+  { {    436, 5}, {     14, 4}, {     29, 5}, {     31, 6}, \
+    {     35, 7}, {     18, 6}, {     37, 7}, {     37, 8}, \
     {     19, 7}, {     40, 8}, {     37, 9}, {     19, 8}, \
     {     43, 9}, {     23, 8}, {     49, 9}, {     27, 8}, \
     {     57, 9}, {     43,10}, {     23, 9}, {     55,10}, \
@@ -93,45 +117,69 @@ along with the GNU MP Library.  If not, see http://www.gnu.org/licenses/.  */
     {     87,11}, {     47,10}, {    111,12}, {     31,11}, \
     {     63,10}, {    135,11}, {     79,10}, {    167,11}, \
     {     95,10}, {    191,11}, {    111,12}, {     63,11}, \
-    {    127,10}, {    255,11}, {    143,10}, {    303,11}, \
-    {    159,10}, {    319,12}, {     95,11}, {    191,10}, \
-    {    399,11}, {    207,10}, {    431,13}, {   8192,14}, \
-    {  16384,15}, {  32768,16}, {  65536,17}, { 131072,18}, \
-    { 262144,19}, { 524288,20}, {1048576,21}, {2097152,22}, \
-    {4194304,23}, {8388608,24} }
-#define SQR_FFT_TABLE3_SIZE 66
+    {    127,10}, {    255,11}, {    143,10}, {    287, 9}, \
+    {    575,10}, {    303,11}, {    159,10}, {    319,12}, \
+    {     95,11}, {    191,10}, {    399,11}, {    207,10}, \
+    {    431,13}, {     63,12}, {    127,11}, {    271,10}, \
+    {    543,11}, {    303,12}, {    159,11}, {    335,10}, \
+    {    671,11}, {    367,10}, {    735,12}, {    191,11}, \
+    {    399,10}, {    799,11}, {    431,12}, {    223,11}, \
+    {    463,13}, {    127,12}, {    255,11}, {    543,12}, \
+    {    287,11}, {    607,12}, {    319,11}, {    671,12}, \
+    {    351,11}, {    735,13}, {    191,12}, {    383,11}, \
+    {    799,12}, {    415,11}, {    863,12}, {    447,11}, \
+    {    895,14}, {    127,13}, {    255,12}, {    543,11}, \
+    {   1087,12}, {    607,13}, {    319,12}, {    735,13}, \
+    {    383,12}, {    863,13}, {    447,12}, {    959,14}, \
+    {    255,13}, {    511,12}, {   1087,13}, {    575,12}, \
+    {   1183,13}, {    639,12}, {   1279,13}, {    703,12}, \
+    {   1407,14}, {    383,13}, {    767,12}, {   1535,13}, \
+    {    831,12}, {   1663,13}, {    895,12}, {   1791,13}, \
+    {    959,15}, {    255,14}, {    511,13}, {   1087,12}, \
+    {   2175,13}, {   1215,14}, {    639,13}, {   1343,12}, \
+    {   2687,13}, {   1471,14}, {    767,13}, {   1663,14}, \
+    {    895,13}, {   1919,15}, {    511,14}, {   1023,13}, \
+    {   2175,14}, {   1151,13}, {   2431,14}, {   1279,13}, \
+    {   2687,14}, {   1407,15}, {    767,14}, {   1535,13}, \
+    {   3199,14}, {   1663,13}, {   3455,14}, {   1791,13}, \
+    {   8192,14}, {  16384,15}, {  32768,16}, {  65536,17}, \
+    { 131072,18}, { 262144,19}, { 524288,20}, {1048576,21}, \
+    {2097152,22}, {4194304,23}, {8388608,24} }
+#define SQR_FFT_TABLE3_SIZE 151
 #define SQR_FFT_THRESHOLD                 4032
 
 #define MULLO_BASECASE_THRESHOLD            29
 #define MULLO_DC_THRESHOLD                  57
 #define MULLO_MUL_N_THRESHOLD            11278
 
-#define DC_DIV_QR_THRESHOLD                 59
+#define DC_DIV_QR_THRESHOLD                 64
 #define DC_DIVAPPR_Q_THRESHOLD             222
 #define DC_BDIV_QR_THRESHOLD                95
 #define DC_BDIV_Q_THRESHOLD                264
 
-#define INV_MULMOD_BNM1_THRESHOLD           82
-#define INV_NEWTON_THRESHOLD                11
-#define INV_APPR_THRESHOLD                  18
+#define INV_MULMOD_BNM1_THRESHOLD           86
+#define INV_NEWTON_THRESHOLD               139
+#define INV_APPR_THRESHOLD                 147
 
 #define BINV_NEWTON_THRESHOLD              252
-#define REDC_1_TO_REDC_2_THRESHOLD           0
+#define REDC_1_TO_REDC_2_THRESHOLD           0  /* always */
 #define REDC_2_TO_REDC_N_THRESHOLD         147
 
 #define MU_DIV_QR_THRESHOLD               1142
-#define MU_DIVAPPR_Q_THRESHOLD             998
+#define MU_DIVAPPR_Q_THRESHOLD            1142
 #define MUPI_DIV_QR_THRESHOLD                0  /* always */
-#define MU_BDIV_QR_THRESHOLD              1187
+#define MU_BDIV_QR_THRESHOLD              1210
 #define MU_BDIV_Q_THRESHOLD               1470
 
 #define MATRIX22_STRASSEN_THRESHOLD         23
 #define HGCD_THRESHOLD                     117
-#define GCD_DC_THRESHOLD                   469
+#define HGCD_APPR_THRESHOLD                111
+#define HGCD_REDUCE_THRESHOLD             3014
+#define GCD_DC_THRESHOLD                   555
 #define GCDEXT_DC_THRESHOLD                368
 #define JACOBI_BASE_METHOD                   4
 
 #define GET_STR_DC_THRESHOLD                13
-#define GET_STR_PRECOMPUTE_THRESHOLD        21
-#define SET_STR_DC_THRESHOLD              1204
-#define SET_STR_PRECOMPUTE_THRESHOLD      3266
+#define GET_STR_PRECOMPUTE_THRESHOLD        22
+#define SET_STR_DC_THRESHOLD              1474
+#define SET_STR_PRECOMPUTE_THRESHOLD      3168
diff --git a/mpn/ia64/tabselect.asm b/mpn/ia64/tabselect.asm
new file mode 100644
index 000000000..cc5b49b04
--- /dev/null
+++ b/mpn/ia64/tabselect.asm
@@ -0,0 +1,139 @@
+dnl  IA-64 mpn_tabselect.
+
+dnl  Copyright 2011 Free Software Foundation, Inc.
+
+dnl  This file is part of the GNU MP Library.
+
+dnl  The GNU MP Library is free software; you can redistribute it and/or modify
+dnl  it under the terms of the GNU Lesser General Public License as published
+dnl  by the Free Software Foundation; either version 3 of the License, or (at
+dnl  your option) any later version.
+
+dnl  The GNU MP Library is distributed in the hope that it will be useful, but
+dnl  WITHOUT ANY WARRANTY; without even the implied warranty of MERCHANTABILITY
+dnl  or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU Lesser General Public
+dnl  License for more details.
+
+dnl  You should have received a copy of the GNU Lesser General Public License
+dnl  along with the GNU MP Library.  If not, see http://www.gnu.org/licenses/.
+
+include(`../config.m4')
+
+C           cycles/limb
+C Itanium:       ?
+C Itanium 2:     2.5
+
+C NOTES
+C  * Using software pipelining could trivially yield 2 c/l without unrolling,
+C    or 1+epsilon with unrolling.  (This code was modelled after the powerpc64
+C    code, for simplicity.)
+
+C mpn_tabselect (mp_limb_t *rp, mp_limb_t *tp, mp_size_t n, mp_size_t nents, mp_size_t which)
+define(`rp',     `r32')
+define(`tp',     `r33')
+define(`n',      `r34')
+define(`nents',  `r35')
+define(`which',  `r36')
+
+define(`mask',   `r8')
+
+define(`rp1',     `r32')
+define(`tp1',     `r33')
+define(`rp2',     `r14')
+define(`tp2',     `r15')
+
+ASM_START()
+	TEXT
+	ALIGN(16)
+PROLOGUE(mpn_tabselect)
+	.prologue
+	.save	ar.lc, r2
+	.body
+ifdef(`HAVE_ABI_32',`
+.mmi;	addp4	rp = 0, rp		C			M I
+	addp4	tp = 0, tp		C			M I
+	zxt4	n = n			C			I
+.mii;	nop	0
+	zxt4	nents = nents		C			I
+	zxt4	which = which		C			I
+	;;
+')
+.mmi;	add	rp2 = 8, rp1
+	add	tp2 = 8, tp1
+	add	r6 = -2, n
+	;;
+.mmi;	cmp.eq	p10, p0 = 1, n
+	and	r9 = 1, n		C set cr0 for use in inner loop
+	shr.u	r6 = r6, 1		C inner loop count
+	;;
+.mmi;	cmp.eq	p8, p0 = 0, r9
+	sub	which = nents, which
+	shl	n = n, 3
+	;;
+
+L(outer):
+.mmi	cmp.eq	p6, p7 = which, nents	C are we at the selected table entry?
+	nop	0
+	mov	ar.lc = r6		C			I0
+	;;
+.mmb;
+  (p6)	mov	mask = -1
+  (p7)	mov	mask = 0
+  (p8)	br.dptk	L(top)			C branch to loop entry if n even
+	;;
+
+.mmi;	ld8	r16 = [tp1], 8
+	add	tp2 = 8, tp2
+	nop	0
+	;;
+.mmi;	ld8	r18 = [rp1]
+	and	r16 = r16, mask
+	nop	0
+	;;
+.mmi;	andcm	r18 = r18, mask
+	;;
+	or	r16 = r16, r18
+	nop	0
+	;;
+.mmb;	st8	[rp1] = r16, 8
+	add	rp2 = 8, rp2
+  (p10)	br.dpnt	L(end)
+
+	ALIGN(32)
+L(top):
+.mmi;	ld8	r16 = [tp1], 16
+	ld8	r17 = [tp2], 16
+	nop	0
+	;;
+.mmi;	ld8	r18 = [rp1]
+	and	r16 = r16, mask
+	nop	0
+.mmi;	ld8	r19 = [rp2]
+	and	r17 = r17, mask
+	nop	0
+	;;
+.mmi;	andcm	r18 = r18, mask
+	andcm	r19 = r19, mask
+	nop	0
+	;;
+.mmi;	or	r16 = r16, r18
+	or	r17 = r17, r19
+	nop	0
+	;;
+.mmb;	st8	[rp1] = r16, 16
+	st8	[rp2] = r17, 16
+	br.cloop.dptk	L(top)
+	;;
+L(end):
+.mmi;	sub	rp1 = rp1, n		C move rp back to beginning
+	sub	rp2 = rp2, n		C move rp back to beginning
+	cmp.ne	p9, p0 = 1, nents
+.mmb;	add	nents = -1, nents
+	nop	0
+  (p9)	br.dptk	L(outer)
+	;;
+
+.mib;	nop	0
+	nop	0
+	br.ret.sptk.many b0
+EPILOGUE()