[Gmp-commit] /var/hg/gmp: 4 new changesets
mercurial at gmplib.org
mercurial at gmplib.org
Sun Nov 29 22:45:48 UTC 2020
details: /var/hg/gmp/rev/0d19622b27a9
changeset: 18165:0d19622b27a9
user: Torbjorn Granlund <tg at gmplib.org>
date: Sun Nov 29 23:39:12 2020 +0100
description:
Recognise CPU "applem1".
details: /var/hg/gmp/rev/7f4dd5915721
changeset: 18166:7f4dd5915721
user: Torbjorn Granlund <tg at gmplib.org>
date: Sun Nov 29 23:41:34 2020 +0100
description:
New file.
details: /var/hg/gmp/rev/fc3d6bcb38e9
changeset: 18167:fc3d6bcb38e9
user: Torbjorn Granlund <tg at gmplib.org>
date: Sun Nov 29 23:45:15 2020 +0100
description:
New file.
details: /var/hg/gmp/rev/607b394997d1
changeset: 18168:607b394997d1
user: Torbjorn Granlund <tg at gmplib.org>
date: Sun Nov 29 23:45:44 2020 +0100
description:
ChangeLog
diffstat:
ChangeLog | 9 +
config.guess | 9 +
config.sub | 2 +-
configure.ac | 12 +-
mpn/arm64/applem1/aorsmul_1.asm | 153 ++++++++++++++++++++++++++
mpn/arm64/divrem_1.asm | 231 ++++++++++++++++++++++++++++++++++++++++
6 files changed, 413 insertions(+), 3 deletions(-)
diffs (truncated from 475 to 300 lines):
diff -r 5f32dbc41afc -r 607b394997d1 ChangeLog
--- a/ChangeLog Sun Nov 29 22:31:40 2020 +0100
+++ b/ChangeLog Sun Nov 29 23:45:44 2020 +0100
@@ -1,3 +1,12 @@
+2020-11-29 Torbjörn Granlund <tg at gmplib.org>
+
+ * mpn/arm64/divrem_1.asm: New file.
+ * mpn/arm64/applem1/aorsmul_1.asm: New file.
+
+ * configure.ac, config.guess, config.sub: Recognise CPU "applem1".
+
+ * mpn/arm64: Avoid the x18 register since it is reserved on Darwin.
+
2020-11-28 Marco Bodrato <bodrato at mail.dm.unipi.it>
* tests/mpz/t-pprime_p.c (check_fermat_mersenne): Check return
diff -r 5f32dbc41afc -r 607b394997d1 config.guess
--- a/config.guess Sun Nov 29 22:31:40 2020 +0100
+++ b/config.guess Sun Nov 29 23:45:44 2020 +0100
@@ -173,6 +173,15 @@
;;
arm*-*-* | aarch64-*-*)
+ if [ $((uname -s) 2>/dev/null) = Darwin ]; then
+ case $(system_profiler SPHardwareDataType | grep Chip) in
+ *Apple\ M1)
+ exact_cpu="applem1";;
+ *Apple\ M2)
+ exact_cpu="applem2";;
+ esac
+ fi
+
cpu_code=`sed -n 's/^CPU part.*\(0x.*\)$/\1/p' /proc/cpuinfo 2>/dev/null | sort -r | head -n 1 2>/dev/null`
cpu_implementer=`sed -n 's/^CPU implementer.*\(0x.*\)$/\1/p' /proc/cpuinfo 2>/dev/null | head -n 1 2>/dev/null`
case "${cpu_implementer}_${cpu_code}" in
diff -r 5f32dbc41afc -r 607b394997d1 config.sub
--- a/config.sub Sun Nov 29 22:31:40 2020 +0100
+++ b/config.sub Sun Nov 29 23:45:44 2020 +0100
@@ -130,7 +130,7 @@
armsa1 | armxscale | arm9tdmi | arm9te | \
arm10* | arm11mpcore | armsa1 | arm1136 | arm1156 | arm1176 | \
armcortex[arm][0-9] | armcortex[arm][0-9][0-9] | \
-arm*neon | armxgene1 | armexynosm1 | armthunderx)
+arm*neon | armxgene1 | armexynosm1 | armthunderx | applem[1-9])
test_cpu="arm";;
*)
diff -r 5f32dbc41afc -r 607b394997d1 configure.ac
--- a/configure.ac Sun Nov 29 22:31:40 2020 +0100
+++ b/configure.ac Sun Nov 29 23:45:44 2020 +0100
@@ -612,7 +612,7 @@
;;
- arm*-*-* | aarch64*-*-*)
+ arm*-*-* | aarch64*-*-* | [applem[1-9]-*-*])
abilist="32"
gcc_cflags="$gcc_cflags $fomit_frame_pointer"
gcc_cflags_optlist="arch fpmode neon tune"
@@ -778,6 +778,14 @@
gcc_cflags_neon="-mfpu=neon"
gcc_cflags_tune=""
;;
+ [applem[1-9]*])
+ abilist="64"
+ path_64="arm64/applem1 arm64"
+ gcc_cflags_arch="-march=armv8-a"
+ gcc_cflags_neon="-mfpu=neon"
+ gcc_cflags_tune=""
+ CALLING_CONVENTIONS_OBJS=""
+ ;;
*)
path="arm"
gcc_cflags_arch="-march=armv4"
@@ -3696,7 +3704,7 @@
GMP_ASM_ALIGN_LOG
case $host in
- arm*-*-* | aarch64*-*-*)
+ arm*-*-* | aarch64*-*-* | [applem[1-9]-*-*])
case $ABI in
32)
GMP_INCLUDE_MPN(arm/arm-defs.m4) ;;
diff -r 5f32dbc41afc -r 607b394997d1 mpn/arm64/applem1/aorsmul_1.asm
--- /dev/null Thu Jan 01 00:00:00 1970 +0000
+++ b/mpn/arm64/applem1/aorsmul_1.asm Sun Nov 29 23:45:44 2020 +0100
@@ -0,0 +1,153 @@
+dnl ARM64 mpn_addmmul_1 and mpn_submul_1.
+
+dnl Contributed to the GNU project by Torbjörn Granlund.
+
+dnl Copyright 2020 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 cycles/limb
+C Cortex-A53
+C Cortex-A55
+C Cortex-A57
+C Cortex-A72
+C Cortex-A73
+C X-Gene
+C Apple M1 1.25
+
+changecom(blah)
+
+define(`rp', x0)
+define(`up', x1)
+define(`n', x2)
+define(`v0', x3)
+
+define(`CY',x17)
+
+ifdef(`OPERATION_addmul_1', `
+ define(`ADDSUB', adds)
+ define(`ADDSUBC', adcs)
+ define(`COND', `cc')
+ define(`func', mpn_addmul_1)')
+ifdef(`OPERATION_submul_1', `
+ define(`ADDSUB', subs)
+ define(`ADDSUBC', sbcs)
+ define(`COND', `cs')
+ define(`func', mpn_submul_1)')
+
+MULFUNC_PROLOGUE(mpn_addmul_1 mpn_submul_1)
+
+PROLOGUE(func)
+ mov CY, #0 C W0
+ lsr x16, n, #2
+ tbz n, #0, L(bx0)
+
+L(bx1): ldr x4, [up], #8
+ mul x8, x4, v0
+ umulh x4, x4, v0
+ tbz n, #1, L(b01)
+
+L(b11): ldp x5,x6, [up], #16
+ ldp x12,x13, [rp]
+ ldr x14, [rp,#16]
+ mul x9, x5, v0
+ umulh x5, x5, v0
+ mul x10, x6, v0
+ umulh x6, x6, v0
+ ADDSUB x8, x12, x8
+ ADDSUBC x4, x13, x4
+ ADDSUBC x5, x14, x5
+ csinc x6, x6, x6, COND
+ ADDSUB x8, x8, CY
+ ADDSUBC x4, x4, x9
+ ADDSUBC x5, x5, x10
+ csinc CY, x6, x6, COND
+ stp x8, x4, [rp], #16
+ str x5, [rp], #8
+ cbnz x16, L(top)
+ mov x0, CY
+ ret
+
+L(b01): ldr x12, [rp]
+ ADDSUB x8, x12, x8
+ csinc x4, x4, x4, COND
+ ADDSUB x8, x8, CY
+ csinc CY, x4, x5, COND
+ str x8, [rp], #8
+ cbnz x16, L(top)
+ mov x0, CY
+ ret
+
+L(bx0): ldp x4,x5, [up], #16
+ tbz n, #1, L(top)+4
+
+L(b10): ldp x12,x13, [rp]
+ mul x8, x4, v0
+ umulh x4, x4, v0
+ mul x9, x5, v0
+ umulh x5, x5, v0
+ ADDSUB x8, x12, x8
+ ADDSUBC x4, x13, x4
+ csinc x5, x5, x5, COND
+ ADDSUB x8, x8, CY
+ ADDSUBC x4, x4, x9
+ csinc CY, x5, x5, COND
+ stp x8, x4, [rp], #16
+ cbz x16, L(done)
+
+L(top): ldp x4,x5, [up], #16 C W0 W1
+ ldp x6,x7, [up], #16 C W2 W3
+ ldp x12,x13, [rp] C W0 W1
+ ldp x14,x15, [rp,#16] C W2 W3
+ mul x8, x4, v0 C W0
+ umulh x4, x4, v0 C W1
+ mul x9, x5, v0 C W1
+ umulh x5, x5, v0 C W2
+ mul x10, x6, v0 C W2
+ umulh x6, x6, v0 C W3
+ mul x11, x7, v0 C W3
+ umulh x7, x7, v0 C W4
+ ADDSUB x8, x12, x8 C W0
+ ADDSUBC x4, x13, x4 C W1
+ ADDSUBC x5, x14, x5 C W2
+ ADDSUBC x6, x15, x6 C W3
+ csinc x7, x7, x7, COND C W4
+ ADDSUB x8, x8, CY C W0 carry-in
+ ADDSUBC x4, x4, x9 C W1
+ ADDSUBC x5, x5, x10 C W2
+ ADDSUBC x6, x6, x11 C W2
+ csinc CY, x7, x7, COND C W3 carry-out
+ stp x8, x4, [rp], #16
+ stp x5, x6, [rp], #16
+ sub x16, x16, #1
+ cbnz x16, L(top)
+
+L(done):mov x0, CY
+ ret
+EPILOGUE()
diff -r 5f32dbc41afc -r 607b394997d1 mpn/arm64/divrem_1.asm
--- /dev/null Thu Jan 01 00:00:00 1970 +0000
+++ b/mpn/arm64/divrem_1.asm Sun Nov 29 23:45:44 2020 +0100
@@ -0,0 +1,231 @@
+dnl ARM64 mpn_divrem_1 and mpn_preinv_divrem_1.
+
+dnl Contributed to the GNU project by Torbjörn Granlund.
+
+dnl Copyright 2020 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')
+
+dnl TODO
+dnl * Handle the most significant quotient limb for the unnormalised case
+dml specially, just like in the C code. (It is very often 0.)
+
+define(`qp_arg', x0)
+define(`fn_arg', x1)
+define(`np_arg', x2)
+define(`n_arg', x3)
+define(`d_arg', x4)
+define(`dinv_arg', x5)
+define(`cnt_arg', x6)
+
+define(`qp', x19)
+define(`np', x20)
+define(`n', x21)
+define(`d', x22)
+define(`fn', x24)
+define(`dinv', x0)
+define(`cnt', x23)
+define(`tnc', x8)
+
+dnl mp_limb_t
More information about the gmp-commit
mailing list