aboutsummaryrefslogtreecommitdiff
path: root/gmp-6.3.0/mpn/pa32/hppa2_0
diff options
context:
space:
mode:
Diffstat (limited to 'gmp-6.3.0/mpn/pa32/hppa2_0')
-rw-r--r--gmp-6.3.0/mpn/pa32/hppa2_0/add_n.asm107
-rw-r--r--gmp-6.3.0/mpn/pa32/hppa2_0/gmp-mparam.h167
-rw-r--r--gmp-6.3.0/mpn/pa32/hppa2_0/sqr_diagonal.asm112
-rw-r--r--gmp-6.3.0/mpn/pa32/hppa2_0/sub_n.asm107
4 files changed, 493 insertions, 0 deletions
diff --git a/gmp-6.3.0/mpn/pa32/hppa2_0/add_n.asm b/gmp-6.3.0/mpn/pa32/hppa2_0/add_n.asm
new file mode 100644
index 0000000..8d881b8
--- /dev/null
+++ b/gmp-6.3.0/mpn/pa32/hppa2_0/add_n.asm
@@ -0,0 +1,107 @@
+dnl HP-PA 2.0 32-bit mpn_add_n -- Add two limb vectors of the same length > 0
+dnl and store sum in a third limb vector.
+
+dnl Copyright 1997, 1998, 2000-2002 Free Software Foundation, Inc.
+
+dnl This file is part of the GNU MP Library.
+dnl
+dnl The GNU MP Library is free software; you can redistribute it and/or modify
+dnl it under the terms of either:
+dnl
+dnl * the GNU Lesser General Public License as published by the Free
+dnl Software Foundation; either version 3 of the License, or (at your
+dnl option) any later version.
+dnl
+dnl or
+dnl
+dnl * the GNU General Public License as published by the Free Software
+dnl Foundation; either version 2 of the License, or (at your option) any
+dnl later version.
+dnl
+dnl or both in parallel, as here.
+dnl
+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 General Public License
+dnl for more details.
+dnl
+dnl You should have received copies of the GNU General Public License and the
+dnl GNU Lesser General Public License along with the GNU MP Library. If not,
+dnl see https://www.gnu.org/licenses/.
+
+include(`../config.m4')
+
+C INPUT PARAMETERS
+C res_ptr gr26
+C s1_ptr gr25
+C s2_ptr gr24
+C size gr23
+
+C This runs at 2 cycles/limb on PA8000.
+
+ASM_START()
+PROLOGUE(mpn_add_n)
+ sub %r0,%r23,%r22
+ zdep %r22,30,3,%r28 C r28 = 2 * (-n & 7)
+ zdep %r22,29,3,%r22 C r22 = 4 * (-n & 7)
+ sub %r25,%r22,%r25 C offset s1_ptr
+ sub %r24,%r22,%r24 C offset s2_ptr
+ sub %r26,%r22,%r26 C offset res_ptr
+ blr %r28,%r0 C branch into loop
+ add %r0,%r0,%r0 C reset carry
+
+LDEF(loop)
+ ldw 0(%r25),%r20
+ ldw 0(%r24),%r31
+ addc %r20,%r31,%r20
+ stw %r20,0(%r26)
+
+LDEF(7)
+ ldw 4(%r25),%r21
+ ldw 4(%r24),%r19
+ addc %r21,%r19,%r21
+ stw %r21,4(%r26)
+
+LDEF(6)
+ ldw 8(%r25),%r20
+ ldw 8(%r24),%r31
+ addc %r20,%r31,%r20
+ stw %r20,8(%r26)
+
+LDEF(5)
+ ldw 12(%r25),%r21
+ ldw 12(%r24),%r19
+ addc %r21,%r19,%r21
+ stw %r21,12(%r26)
+
+LDEF(4)
+ ldw 16(%r25),%r20
+ ldw 16(%r24),%r31
+ addc %r20,%r31,%r20
+ stw %r20,16(%r26)
+
+LDEF(3)
+ ldw 20(%r25),%r21
+ ldw 20(%r24),%r19
+ addc %r21,%r19,%r21
+ stw %r21,20(%r26)
+
+LDEF(2)
+ ldw 24(%r25),%r20
+ ldw 24(%r24),%r31
+ addc %r20,%r31,%r20
+ stw %r20,24(%r26)
+
+LDEF(1)
+ ldw 28(%r25),%r21
+ ldo 32(%r25),%r25
+ ldw 28(%r24),%r19
+ addc %r21,%r19,%r21
+ stw %r21,28(%r26)
+ ldo 32(%r24),%r24
+ addib,> -8,%r23,L(loop)
+ ldo 32(%r26),%r26
+
+ bv (%r2)
+ addc %r0,%r0,%r28
+EPILOGUE()
diff --git a/gmp-6.3.0/mpn/pa32/hppa2_0/gmp-mparam.h b/gmp-6.3.0/mpn/pa32/hppa2_0/gmp-mparam.h
new file mode 100644
index 0000000..6016274
--- /dev/null
+++ b/gmp-6.3.0/mpn/pa32/hppa2_0/gmp-mparam.h
@@ -0,0 +1,167 @@
+/* gmp-mparam.h -- Compiler/machine parameter header file.
+
+Copyright 1991, 1993, 1994, 1999-2002, 2009, 2010 Free Software Foundation,
+Inc.
+
+This file is part of the GNU MP Library.
+
+The GNU MP Library is free software; you can redistribute it and/or modify
+it under the terms of either:
+
+ * the GNU Lesser General Public License as published by the Free
+ Software Foundation; either version 3 of the License, or (at your
+ option) any later version.
+
+or
+
+ * the GNU General Public License as published by the Free Software
+ Foundation; either version 2 of the License, or (at your option) any
+ later version.
+
+or both in parallel, as here.
+
+The GNU MP Library is distributed in the hope that it will be useful, but
+WITHOUT ANY WARRANTY; without even the implied warranty of MERCHANTABILITY
+or FITNESS FOR A PARTICULAR PURPOSE. See the GNU General Public License
+for more details.
+
+You should have received copies of the GNU General Public License and the
+GNU Lesser General Public License along with the GNU MP Library. If not,
+see https://www.gnu.org/licenses/. */
+
+#define GMP_LIMB_BITS 32
+#define GMP_LIMB_BYTES 4
+
+/* 552 MHz PA8600 (gcc61.fsffrance.org) */
+
+#define DIVREM_1_NORM_THRESHOLD 3
+#define DIVREM_1_UNNORM_THRESHOLD 3
+#define MOD_1_NORM_THRESHOLD 3
+#define MOD_1_UNNORM_THRESHOLD 4
+#define MOD_1N_TO_MOD_1_1_THRESHOLD 11
+#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 22
+#define PREINV_MOD_1_TO_MOD_1_THRESHOLD 28
+#define USE_PREINV_DIVREM_1 1
+#define DIVEXACT_1_THRESHOLD 0 /* always */
+#define BMOD_1_TO_MOD_1_THRESHOLD 36
+
+#define MUL_TOOM22_THRESHOLD 18
+#define MUL_TOOM33_THRESHOLD 65
+#define MUL_TOOM44_THRESHOLD 166
+#define MUL_TOOM6H_THRESHOLD 202
+#define MUL_TOOM8H_THRESHOLD 333
+
+#define MUL_TOOM32_TO_TOOM43_THRESHOLD 105
+#define MUL_TOOM32_TO_TOOM53_THRESHOLD 138
+#define MUL_TOOM42_TO_TOOM53_THRESHOLD 105
+#define MUL_TOOM42_TO_TOOM63_THRESHOLD 102
+
+#define SQR_BASECASE_THRESHOLD 7
+#define SQR_TOOM2_THRESHOLD 55
+#define SQR_TOOM3_THRESHOLD 93
+#define SQR_TOOM4_THRESHOLD 250
+#define SQR_TOOM6_THRESHOLD 306
+#define SQR_TOOM8_THRESHOLD 527
+
+#define MULMOD_BNM1_THRESHOLD 13
+#define SQRMOD_BNM1_THRESHOLD 15
+
+#define MUL_FFT_MODF_THRESHOLD 244 /* k = 5 */
+#define MUL_FFT_TABLE3 \
+ { { 244, 5}, { 8, 4}, { 17, 5}, { 13, 6}, \
+ { 7, 5}, { 17, 6}, { 9, 5}, { 19, 6}, \
+ { 13, 7}, { 7, 6}, { 17, 7}, { 9, 6}, \
+ { 19, 7}, { 11, 6}, { 24, 7}, { 13, 8}, \
+ { 7, 7}, { 19, 8}, { 11, 7}, { 25, 8}, \
+ { 15, 7}, { 33, 8}, { 23, 9}, { 15, 8}, \
+ { 39, 9}, { 23,10}, { 15, 9}, { 31, 8}, \
+ { 67, 9}, { 39, 8}, { 79, 9}, { 47,10}, \
+ { 31, 9}, { 71, 8}, { 143, 9}, { 79,10}, \
+ { 47,11}, { 31,10}, { 63, 9}, { 127, 8}, \
+ { 255, 9}, { 135, 8}, { 271, 9}, { 143,10}, \
+ { 79, 9}, { 159, 8}, { 319, 9}, { 175, 8}, \
+ { 351,10}, { 95, 9}, { 191, 8}, { 383, 9}, \
+ { 207,11}, { 63,10}, { 127, 9}, { 255, 8}, \
+ { 511, 9}, { 271,10}, { 143, 9}, { 287, 8}, \
+ { 575,10}, { 159, 9}, { 319,10}, { 175, 9}, \
+ { 351,11}, { 95,10}, { 191, 9}, { 383,10}, \
+ { 207, 9}, { 415,12}, { 63,11}, { 127,10}, \
+ { 255, 9}, { 511,10}, { 271, 9}, { 543, 8}, \
+ { 1087,10}, { 287, 9}, { 575,10}, { 303,11}, \
+ { 159,10}, { 351, 9}, { 703, 8}, { 1407,11}, \
+ { 191,10}, { 415, 9}, { 831,11}, { 223, 9}, \
+ { 895,10}, { 479,12}, { 127,11}, { 255,10}, \
+ { 543, 9}, { 1087,11}, { 287,10}, { 607, 9}, \
+ { 1215,11}, { 351,10}, { 703, 9}, { 1407,12}, \
+ { 191,11}, { 415,10}, { 831,11}, { 479,13}, \
+ { 8192,14}, { 16384,15}, { 32768,16} }
+#define MUL_FFT_TABLE3_SIZE 107
+#define MUL_FFT_THRESHOLD 2112
+
+#define SQR_FFT_MODF_THRESHOLD 240 /* k = 5 */
+#define SQR_FFT_TABLE3 \
+ { { 240, 5}, { 8, 4}, { 17, 5}, { 19, 6}, \
+ { 17, 7}, { 9, 6}, { 20, 7}, { 11, 6}, \
+ { 23, 7}, { 13, 8}, { 7, 7}, { 19, 8}, \
+ { 11, 7}, { 25, 8}, { 15, 7}, { 33, 8}, \
+ { 19, 7}, { 39, 8}, { 23, 9}, { 15, 8}, \
+ { 39, 9}, { 23,10}, { 15, 9}, { 31, 8}, \
+ { 63, 9}, { 47,10}, { 31, 9}, { 63, 8}, \
+ { 127, 9}, { 71, 8}, { 143, 9}, { 79,10}, \
+ { 47,11}, { 31,10}, { 63, 9}, { 127, 8}, \
+ { 255, 7}, { 511, 9}, { 135, 8}, { 271, 9}, \
+ { 143,10}, { 79, 9}, { 159, 8}, { 319, 9}, \
+ { 175, 8}, { 351, 7}, { 703,10}, { 95, 9}, \
+ { 191, 8}, { 383, 9}, { 207,11}, { 63,10}, \
+ { 127, 9}, { 255, 8}, { 511, 9}, { 271,10}, \
+ { 143, 9}, { 287, 8}, { 575,10}, { 159, 9}, \
+ { 319,10}, { 175, 9}, { 351, 8}, { 703,11}, \
+ { 95,10}, { 191, 9}, { 383,10}, { 207, 9}, \
+ { 415,12}, { 63,11}, { 127,10}, { 255, 9}, \
+ { 511,10}, { 271, 9}, { 543, 8}, { 1087,10}, \
+ { 287, 9}, { 575,11}, { 159,10}, { 319, 9}, \
+ { 639,10}, { 351, 9}, { 703, 8}, { 1407,11}, \
+ { 191,10}, { 415, 9}, { 831,11}, { 223, 8}, \
+ { 1791,10}, { 479, 9}, { 959,12}, { 127,11}, \
+ { 255,10}, { 543,11}, { 287,10}, { 607,11}, \
+ { 319,10}, { 639,11}, { 351,10}, { 703, 9}, \
+ { 1407,12}, { 191,11}, { 415,10}, { 831,11}, \
+ { 479,10}, { 959,13}, { 8192,14}, { 16384,15}, \
+ { 32768,16} }
+#define SQR_FFT_TABLE3_SIZE 109
+#define SQR_FFT_THRESHOLD 1600
+
+#define MULLO_BASECASE_THRESHOLD 0 /* always */
+#define MULLO_DC_THRESHOLD 116
+#define MULLO_MUL_N_THRESHOLD 3574
+
+#define DC_DIV_QR_THRESHOLD 100
+#define DC_DIVAPPR_Q_THRESHOLD 348
+#define DC_BDIV_QR_THRESHOLD 109
+#define DC_BDIV_Q_THRESHOLD 254
+
+#define INV_MULMOD_BNM1_THRESHOLD 34
+#define INV_NEWTON_THRESHOLD 276
+#define INV_APPR_THRESHOLD 276
+
+#define BINV_NEWTON_THRESHOLD 278
+#define REDC_1_TO_REDC_N_THRESHOLD 78
+
+#define MU_DIV_QR_THRESHOLD 979
+#define MU_DIVAPPR_Q_THRESHOLD 263
+#define MUPI_DIV_QR_THRESHOLD 102
+#define MU_BDIV_QR_THRESHOLD 807
+#define MU_BDIV_Q_THRESHOLD 1187
+
+#define MATRIX22_STRASSEN_THRESHOLD 11
+#define HGCD_THRESHOLD 100
+#define GCD_DC_THRESHOLD 379
+#define GCDEXT_DC_THRESHOLD 249
+#define JACOBI_BASE_METHOD 2
+
+#define GET_STR_DC_THRESHOLD 7
+#define GET_STR_PRECOMPUTE_THRESHOLD 16
+#define SET_STR_DC_THRESHOLD 270
+#define SET_STR_PRECOMPUTE_THRESHOLD 782
diff --git a/gmp-6.3.0/mpn/pa32/hppa2_0/sqr_diagonal.asm b/gmp-6.3.0/mpn/pa32/hppa2_0/sqr_diagonal.asm
new file mode 100644
index 0000000..c55112f
--- /dev/null
+++ b/gmp-6.3.0/mpn/pa32/hppa2_0/sqr_diagonal.asm
@@ -0,0 +1,112 @@
+dnl HP-PA 32-bit mpn_sqr_diagonal optimized for the PA8x00.
+
+dnl Copyright 2001, 2002 Free Software Foundation, Inc.
+
+dnl This file is part of the GNU MP Library.
+dnl
+dnl The GNU MP Library is free software; you can redistribute it and/or modify
+dnl it under the terms of either:
+dnl
+dnl * the GNU Lesser General Public License as published by the Free
+dnl Software Foundation; either version 3 of the License, or (at your
+dnl option) any later version.
+dnl
+dnl or
+dnl
+dnl * the GNU General Public License as published by the Free Software
+dnl Foundation; either version 2 of the License, or (at your option) any
+dnl later version.
+dnl
+dnl or both in parallel, as here.
+dnl
+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 General Public License
+dnl for more details.
+dnl
+dnl You should have received copies of the GNU General Public License and the
+dnl GNU Lesser General Public License along with the GNU MP Library. If not,
+dnl see https://www.gnu.org/licenses/.
+
+include(`../config.m4')
+
+C This code runs at 6 cycles/limb on the PA7100 and 2 cycles/limb on PA8x00.
+C The 2-way unrolling is actually not helping the PA7100.
+
+C INPUT PARAMETERS
+define(`rp',`%r26')
+define(`up',`%r25')
+define(`n',`%r24')
+
+ASM_START()
+PROLOGUE(mpn_sqr_diagonal)
+
+ fldws,ma 4(up),%fr4r
+ addib,= -1,n,L(end1)
+ ldo 4(rp),rp
+
+ fldws,ma 4(up),%fr6r
+ addib,= -1,n,L(end2)
+ xmpyu %fr4r,%fr4r,%fr5
+
+ fldws,ma 4(up),%fr4r
+ addib,= -1,n,L(end3)
+ xmpyu %fr6r,%fr6r,%fr7
+
+
+LDEF(loop)
+ fldws,ma 4(up),%fr6r
+ fstws %fr5r,-4(rp)
+ fstws,ma %fr5l,8(rp)
+ addib,= -1,n,L(exite)
+ xmpyu %fr4r,%fr4r,%fr5
+ fldws,ma 4(up),%fr4r
+ fstws %fr7r,-4(rp)
+ fstws,ma %fr7l,8(rp)
+ addib,<> -1,n,L(loop)
+ xmpyu %fr6r,%fr6r,%fr7
+
+LDEF(exito)
+ fstws %fr5r,-4(rp)
+ fstws %fr5l,0(rp)
+ xmpyu %fr4r,%fr4r,%fr5
+ fstws %fr7r,4(rp)
+ fstws %fr7l,8(rp)
+ fstws,mb %fr5r,12(rp)
+ bv 0(%r2)
+ fstws %fr5l,4(rp)
+
+LDEF(exite)
+ fstws %fr7r,-4(rp)
+ fstws %fr7l,0(rp)
+ xmpyu %fr6r,%fr6r,%fr7
+ fstws %fr5r,4(rp)
+ fstws %fr5l,8(rp)
+ fstws,mb %fr7r,12(rp)
+ bv 0(%r2)
+ fstws %fr7l,4(rp)
+
+LDEF(end1)
+ xmpyu %fr4r,%fr4r,%fr5
+ fstws %fr5r,-4(rp)
+ bv 0(%r2)
+ fstws,ma %fr5l,8(rp)
+
+LDEF(end2)
+ xmpyu %fr6r,%fr6r,%fr7
+ fstws %fr5r,-4(rp)
+ fstws %fr5l,0(rp)
+ fstws %fr7r,4(rp)
+ bv 0(%r2)
+ fstws %fr7l,8(rp)
+
+LDEF(end3)
+ fstws %fr5r,-4(rp)
+ fstws %fr5l,0(rp)
+ xmpyu %fr4r,%fr4r,%fr5
+ fstws %fr7r,4(rp)
+ fstws %fr7l,8(rp)
+ fstws,mb %fr5r,12(rp)
+ bv 0(%r2)
+ fstws %fr5l,4(rp)
+EPILOGUE(mpn_sqr_diagonal)
diff --git a/gmp-6.3.0/mpn/pa32/hppa2_0/sub_n.asm b/gmp-6.3.0/mpn/pa32/hppa2_0/sub_n.asm
new file mode 100644
index 0000000..47b3163
--- /dev/null
+++ b/gmp-6.3.0/mpn/pa32/hppa2_0/sub_n.asm
@@ -0,0 +1,107 @@
+dnl HP-PA 2.0 32-bit mpn_sub_n -- Subtract two limb vectors of the same
+dnl length > 0 and store difference in a third limb vector.
+
+dnl Copyright 1997, 1998, 2000-2002 Free Software Foundation, Inc.
+
+dnl This file is part of the GNU MP Library.
+dnl
+dnl The GNU MP Library is free software; you can redistribute it and/or modify
+dnl it under the terms of either:
+dnl
+dnl * the GNU Lesser General Public License as published by the Free
+dnl Software Foundation; either version 3 of the License, or (at your
+dnl option) any later version.
+dnl
+dnl or
+dnl
+dnl * the GNU General Public License as published by the Free Software
+dnl Foundation; either version 2 of the License, or (at your option) any
+dnl later version.
+dnl
+dnl or both in parallel, as here.
+dnl
+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 General Public License
+dnl for more details.
+dnl
+dnl You should have received copies of the GNU General Public License and the
+dnl GNU Lesser General Public License along with the GNU MP Library. If not,
+dnl see https://www.gnu.org/licenses/.
+
+include(`../config.m4')
+
+C INPUT PARAMETERS
+C res_ptr gr26
+C s1_ptr gr25
+C s2_ptr gr24
+C size gr23
+
+C This runs at 2 cycles/limb on PA8000.
+
+ASM_START()
+PROLOGUE(mpn_sub_n)
+ sub %r0,%r23,%r22
+ zdep %r22,30,3,%r28 C r28 = 2 * (-n & 7)
+ zdep %r22,29,3,%r22 C r22 = 4 * (-n & 7)
+ sub %r25,%r22,%r25 C offset s1_ptr
+ sub %r24,%r22,%r24 C offset s2_ptr
+ blr %r28,%r0 C branch into loop
+ sub %r26,%r22,%r26 C offset res_ptr and set carry
+
+LDEF(loop)
+ ldw 0(%r25),%r20
+ ldw 0(%r24),%r31
+ subb %r20,%r31,%r20
+ stw %r20,0(%r26)
+
+LDEF(7)
+ ldw 4(%r25),%r21
+ ldw 4(%r24),%r19
+ subb %r21,%r19,%r21
+ stw %r21,4(%r26)
+
+LDEF(6)
+ ldw 8(%r25),%r20
+ ldw 8(%r24),%r31
+ subb %r20,%r31,%r20
+ stw %r20,8(%r26)
+
+LDEF(5)
+ ldw 12(%r25),%r21
+ ldw 12(%r24),%r19
+ subb %r21,%r19,%r21
+ stw %r21,12(%r26)
+
+LDEF(4)
+ ldw 16(%r25),%r20
+ ldw 16(%r24),%r31
+ subb %r20,%r31,%r20
+ stw %r20,16(%r26)
+
+LDEF(3)
+ ldw 20(%r25),%r21
+ ldw 20(%r24),%r19
+ subb %r21,%r19,%r21
+ stw %r21,20(%r26)
+
+LDEF(2)
+ ldw 24(%r25),%r20
+ ldw 24(%r24),%r31
+ subb %r20,%r31,%r20
+ stw %r20,24(%r26)
+
+LDEF(1)
+ ldw 28(%r25),%r21
+ ldo 32(%r25),%r25
+ ldw 28(%r24),%r19
+ subb %r21,%r19,%r21
+ stw %r21,28(%r26)
+ ldo 32(%r24),%r24
+ addib,> -8,%r23,L(loop)
+ ldo 32(%r26),%r26
+
+ addc %r0,%r0,%r28
+ bv (%r2)
+ subi 1,%r28,%r28
+EPILOGUE()
generated by cgit v1.2.3 (git 2.25.1) at 2024-09-19 22:34:27 +0000