[PATCH] Add aarch64 (arm64) mpi assembly

Jussi Kivilinna jussi.kivilinna at iki.fi
Sat Nov 30 22:22:43 CET 2013


* mpi/aarch64/mpi-asm-defs.h: New.
* mpi/aarch64/mpih-add1.S: New.
* mpi/aarch64/mpih-mul1.S: New.
* mpi/aarch64/mpih-mul2.S: New.
* mpi/aarch64/mpih-mul3.S: New.
* mpi/aarch64/mpih-sub1.S: New.
* mpi/config.links [host=aarch64-*-*]: Add configguration for aarch64
assembly.
* mpi/longlong.h [__aarch64__] (add_ssaaaa, sub_ddmmss, umul_ppmm)
(count_leading_zeros): New.
--

Add preliminary aarch64 assembly implementations for mpi.

Signed-off-by: Jussi Kivilinna <jussi.kivilinna at iki.fi>
---
 mpi/aarch64/mpi-asm-defs.h |    4 +
 mpi/aarch64/mpih-add1.S    |   71 ++++++++++++++++++++++++++
 mpi/aarch64/mpih-mul1.S    |   96 +++++++++++++++++++++++++++++++++++
 mpi/aarch64/mpih-mul2.S    |  108 +++++++++++++++++++++++++++++++++++++++
 mpi/aarch64/mpih-mul3.S    |  121 ++++++++++++++++++++++++++++++++++++++++++++
 mpi/aarch64/mpih-sub1.S    |   71 ++++++++++++++++++++++++++
 mpi/config.links           |    5 ++
 mpi/longlong.h             |   35 +++++++++++++
 8 files changed, 511 insertions(+)
 create mode 100644 mpi/aarch64/mpi-asm-defs.h
 create mode 100644 mpi/aarch64/mpih-add1.S
 create mode 100644 mpi/aarch64/mpih-mul1.S
 create mode 100644 mpi/aarch64/mpih-mul2.S
 create mode 100644 mpi/aarch64/mpih-mul3.S
 create mode 100644 mpi/aarch64/mpih-sub1.S

diff --git a/mpi/aarch64/mpi-asm-defs.h b/mpi/aarch64/mpi-asm-defs.h
new file mode 100644
index 0000000..6519065
--- /dev/null
+++ b/mpi/aarch64/mpi-asm-defs.h
@@ -0,0 +1,4 @@
+/* This file defines some basic constants for the MPI machinery.  We
+ * need to define the types on a per-CPU basis, so it is done with
+ * this file here.  */
+#define BYTES_PER_MPI_LIMB  (SIZEOF_UNSIGNED_LONG_LONG)
diff --git a/mpi/aarch64/mpih-add1.S b/mpi/aarch64/mpih-add1.S
new file mode 100644
index 0000000..9f7e2e6
--- /dev/null
+++ b/mpi/aarch64/mpih-add1.S
@@ -0,0 +1,71 @@
+/* ARM64 add_n -- Add two limb vectors of the same length > 0 and store
+ *                sum in a third limb vector.
+ *
+ *      Copyright © 2013 Jussi Kivilinna <jussi.kivilinna at iki.fi>
+ *
+ * This file is part of Libgcrypt.
+ *
+ * Libgcrypt is free software; you can redistribute it and/or modify
+ * it under the terms of the GNU Lesser General Public License as
+ * published by the Free Software Foundation; either version 2.1 of
+ * the License, or (at your option) any later version.
+ *
+ * Libgcrypt 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 Lesser General Public License for more details.
+ *
+ * You should have received a copy of the GNU Lesser General Public
+ * License along with this program; if not, see <http://www.gnu.org/licenses/>.
+ *
+ */
+
+#include "sysdep.h"
+#include "asm-syntax.h"
+
+/*******************
+ *  mpi_limb_t
+ *  _gcry_mpih_add_n( mpi_ptr_t res_ptr,	x0
+ *		   mpi_ptr_t s1_ptr,		x1
+ *		   mpi_ptr_t s2_ptr,		x2
+ *		   mpi_size_t size)		x3
+ */
+
+.text
+
+.globl _gcry_mpih_add_n
+.type  _gcry_mpih_add_n,%function
+_gcry_mpih_add_n:
+	and	x5, x3, #3;
+	adds	xzr, xzr, xzr; /* clear carry flag */
+
+	cbz	x5, .Large_loop;
+
+.Loop:
+	ldr	x4, [x1], #8;
+	sub	x3, x3, #1;
+	ldr	x11, [x2], #8;
+	and	x5, x3, #3;
+	adcs	x4, x4, x11;
+	str	x4, [x0], #8;
+	cbz	x3, .Lend;
+	cbnz	x5, .Loop;
+
+.Large_loop:
+	ldp	x4, x6, [x1], #16;
+	ldp	x5, x7, [x2], #16;
+	ldp	x8, x10, [x1], #16;
+	ldp	x9, x11, [x2], #16;
+	sub	x3, x3, #4;
+	adcs	x4, x4, x5;
+	adcs	x6, x6, x7;
+	adcs	x8, x8, x9;
+	adcs	x10, x10, x11;
+	stp	x4, x6, [x0], #16;
+	stp	x8, x10, [x0], #16;
+	cbnz	x3, .Large_loop;
+
+.Lend:
+	adc	x0, xzr, xzr;
+	ret;
+.size _gcry_mpih_add_n,.-_gcry_mpih_add_n;
diff --git a/mpi/aarch64/mpih-mul1.S b/mpi/aarch64/mpih-mul1.S
new file mode 100644
index 0000000..cbb333f
--- /dev/null
+++ b/mpi/aarch64/mpih-mul1.S
@@ -0,0 +1,96 @@
+/* ARM64 mul_1 -- Multiply a limb vector with a limb and store the result in
+ *                a second limb vector.
+ *
+ *      Copyright © 2013 Jussi Kivilinna <jussi.kivilinna at iki.fi>
+ *
+ * This file is part of Libgcrypt.
+ *
+ * Libgcrypt is free software; you can redistribute it and/or modify
+ * it under the terms of the GNU Lesser General Public License as
+ * published by the Free Software Foundation; either version 2.1 of
+ * the License, or (at your option) any later version.
+ *
+ * Libgcrypt 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 Lesser General Public License for more details.
+ *
+ * You should have received a copy of the GNU Lesser General Public
+ * License along with this program; if not, see <http://www.gnu.org/licenses/>.
+ *
+ */
+
+#include "sysdep.h"
+#include "asm-syntax.h"
+
+/*******************
+ * mpi_limb_t
+ * _gcry_mpih_mul_1( mpi_ptr_t res_ptr,		x0
+ *		  mpi_ptr_t s1_ptr,		x1
+ *		  mpi_size_t s1_size,		x2
+ *		  mpi_limb_t s2_limb)		x3
+ */
+
+.text
+
+.globl _gcry_mpih_mul_1
+.type  _gcry_mpih_mul_1,%function
+_gcry_mpih_mul_1:
+	and	x5, x2, #3;
+	mov	x4, xzr;
+
+	cbz	x5, .Large_loop;
+
+.Loop:
+	ldr	x5, [x1], #8;
+	sub	x2, x2, #1;
+	mul	x9, x5, x3;
+	umulh	x10, x5, x3;
+	and	x5, x2, #3;
+	adds	x4, x4, x9;
+	str	x4, [x0], #8;
+	adc	x4, x10, xzr;
+
+	cbz	x2, .Lend;
+	cbnz	x5, .Loop;
+
+.Large_loop:
+	ldp	x5, x6, [x1];
+	sub	x2, x2, #4;
+
+	mul	x9, x5, x3;
+	ldp	x7, x8, [x1, #16];
+	umulh	x10, x5, x3;
+	add	x1, x1, #32;
+
+	adds	x4, x4, x9;
+	str	x4, [x0], #8;
+	mul	x11, x6, x3;
+	adc	x4, x10, xzr;
+
+	umulh	x12, x6, x3;
+
+	adds	x4, x4, x11;
+	str	x4, [x0], #8;
+	mul	x13, x7, x3;
+	adc	x4, x12, xzr;
+
+	umulh	x14, x7, x3;
+
+	adds	x4, x4, x13;
+	str	x4, [x0], #8;
+	mul	x15, x8, x3;
+	adc	x4, x14, xzr;
+
+	umulh	x16, x8, x3;
+
+	adds	x4, x4, x15;
+	str	x4, [x0], #8;
+	adc	x4, x16, xzr;
+
+	cbnz	x2, .Large_loop;
+
+.Lend:
+	mov	x0, x4;
+	ret;
+.size _gcry_mpih_mul_1,.-_gcry_mpih_mul_1;
diff --git a/mpi/aarch64/mpih-mul2.S b/mpi/aarch64/mpih-mul2.S
new file mode 100644
index 0000000..bfb3571
--- /dev/null
+++ b/mpi/aarch64/mpih-mul2.S
@@ -0,0 +1,108 @@
+/* ARM64 mul_2 -- Multiply a limb vector with a limb and add the result to
+ *                a second limb vector.
+ *
+ *      Copyright © 2013 Jussi Kivilinna <jussi.kivilinna at iki.fi>
+ *
+ * This file is part of Libgcrypt.
+ *
+ * Libgcrypt is free software; you can redistribute it and/or modify
+ * it under the terms of the GNU Lesser General Public License as
+ * published by the Free Software Foundation; either version 2.1 of
+ * the License, or (at your option) any later version.
+ *
+ * Libgcrypt 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 Lesser General Public License for more details.
+ *
+ * You should have received a copy of the GNU Lesser General Public
+ * License along with this program; if not, see <http://www.gnu.org/licenses/>.
+ *
+ */
+
+#include "sysdep.h"
+#include "asm-syntax.h"
+
+/*******************
+ * mpi_limb_t
+ * _gcry_mpih_addmul_1( mpi_ptr_t res_ptr,	x0
+ *		     mpi_ptr_t s1_ptr,		x1
+ *		     mpi_size_t s1_size,	x2
+ *		     mpi_limb_t s2_limb)	x3
+ */
+
+.text
+
+.globl _gcry_mpih_addmul_1
+.type  _gcry_mpih_addmul_1,%function
+_gcry_mpih_addmul_1:
+	and	x5, x2, #3;
+	mov	x6, xzr;
+	mov	x7, xzr;
+
+	cbz	x5, .Large_loop;
+
+.Loop:
+	ldr	x5, [x1], #8;
+
+	mul	x12, x5, x3;
+	ldr	x4, [x0];
+	umulh	x13, x5, x3;
+	sub	x2, x2, #1;
+
+	adds	x12, x12, x4;
+	and	x5, x2, #3;
+	adc	x13, x13, x7;
+	adds	x12, x12, x6;
+	str	x12, [x0], #8;
+	adc	x6, x7, x13;
+
+	cbz	x2, .Lend;
+	cbnz	x5, .Loop;
+
+.Large_loop:
+	ldp	x5, x9, [x1], #16;
+	sub	x2, x2, #4;
+	ldp	x4, x8, [x0];
+
+	mul	x12, x5, x3;
+	umulh	x13, x5, x3;
+
+	adds	x12, x12, x4;
+	mul	x14, x9, x3;
+	adc	x13, x13, x7;
+	adds	x12, x12, x6;
+	umulh	x15, x9, x3;
+	str	x12, [x0], #8;
+	adc	x6, x7, x13;
+
+	adds	x14, x14, x8;
+	ldp	x5, x9, [x1], #16;
+	adc	x15, x15, x7;
+	adds	x14, x14, x6;
+	mul	x12, x5, x3;
+	str	x14, [x0], #8;
+	ldp	x4, x8, [x0];
+	umulh	x13, x5, x3;
+	adc	x6, x7, x15;
+
+	adds	x12, x12, x4;
+	mul	x14, x9, x3;
+	adc	x13, x13, x7;
+	adds	x12, x12, x6;
+	umulh	x15, x9, x3;
+	str	x12, [x0], #8;
+	adc	x6, x7, x13;
+
+	adds	x14, x14, x8;
+	adc	x15, x15, x7;
+	adds	x14, x14, x6;
+	str	x14, [x0], #8;
+	adc	x6, x7, x15;
+
+	cbnz	x2, .Large_loop;
+
+.Lend:
+	mov	x0, x6;
+	ret;
+.size _gcry_mpih_addmul_1,.-_gcry_mpih_addmul_1;
diff --git a/mpi/aarch64/mpih-mul3.S b/mpi/aarch64/mpih-mul3.S
new file mode 100644
index 0000000..6f12b7b
--- /dev/null
+++ b/mpi/aarch64/mpih-mul3.S
@@ -0,0 +1,121 @@
+/* ARM mul_3 -- Multiply a limb vector with a limb and subtract the result
+ *              from a second limb vector.
+ *
+ *      Copyright © 2013 Jussi Kivilinna <jussi.kivilinna at iki.fi>
+ *
+ * This file is part of Libgcrypt.
+ *
+ * Libgcrypt is free software; you can redistribute it and/or modify
+ * it under the terms of the GNU Lesser General Public License as
+ * published by the Free Software Foundation; either version 2.1 of
+ * the License, or (at your option) any later version.
+ *
+ * Libgcrypt 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 Lesser General Public License for more details.
+ *
+ * You should have received a copy of the GNU Lesser General Public
+ * License along with this program; if not, see <http://www.gnu.org/licenses/>.
+ *
+ */
+
+#include "sysdep.h"
+#include "asm-syntax.h"
+
+/*******************
+ * mpi_limb_t
+ * _gcry_mpih_submul_1( mpi_ptr_t res_ptr,	x0
+ *		     mpi_ptr_t s1_ptr,		x1
+ *		     mpi_size_t s1_size,	x2
+ *		     mpi_limb_t s2_limb)	x3
+ */
+
+.text
+
+.globl _gcry_mpih_submul_1
+.type  _gcry_mpih_submul_1,%function
+_gcry_mpih_submul_1:
+	and	x5, x2, #3;
+	mov	x7, xzr;
+	cbz	x5, .Large_loop;
+
+	subs	xzr, xzr, xzr;
+
+.Loop:
+	ldr	x4, [x1], #8;
+	cinc	x7, x7, cc;
+	ldr	x5, [x0];
+	sub	x2, x2, #1;
+
+	mul	x6, x4, x3;
+	subs	x5, x5, x7;
+	umulh	x4, x4, x3;
+	and	x10, x2, #3;
+
+	cset	x7, cc;
+	subs	x5, x5, x6;
+	add	x7, x7, x4;
+	str	x5, [x0], #8;
+
+	cbz	x2, .Loop_end;
+	cbnz	x10, .Loop;
+
+	cinc	x7, x7, cc;
+
+.Large_loop:
+	ldp	x4, x8, [x1], #16;
+	sub	x2, x2, #4;
+	ldp	x5, x9, [x0];
+
+	mul	x6, x4, x3;
+	subs	x5, x5, x7;
+	umulh	x4, x4, x3;
+
+	cset	x7, cc;
+	subs	x5, x5, x6;
+	mul	x6, x8, x3;
+	add	x7, x7, x4;
+	str	x5, [x0], #8;
+	cinc	x7, x7, cc;
+
+	umulh	x8, x8, x3;
+
+	subs	x9, x9, x7;
+	cset	x7, cc;
+	subs	x9, x9, x6;
+	ldp	x4, x10, [x1], #16;
+	str	x9, [x0], #8;
+	add	x7, x7, x8;
+	ldp	x5, x9, [x0];
+	cinc	x7, x7, cc;
+
+	mul	x6, x4, x3;
+	subs	x5, x5, x7;
+	umulh	x4, x4, x3;
+
+	cset	x7, cc;
+	subs	x5, x5, x6;
+	mul	x6, x10, x3;
+	add	x7, x7, x4;
+	str	x5, [x0], #8;
+	cinc	x7, x7, cc;
+
+	umulh	x10, x10, x3;
+
+	subs	x9, x9, x7;
+	cset	x7, cc;
+	subs	x9, x9, x6;
+	add	x7, x7, x10;
+	str	x9, [x0], #8;
+	cinc	x7, x7, cc;
+
+	cbnz	x2, .Large_loop;
+
+	mov	x0, x7;
+	ret;
+
+.Loop_end:
+	cinc	x0, x7, cc;
+	ret;
+.size _gcry_mpih_submul_1,.-_gcry_mpih_submul_1;
diff --git a/mpi/aarch64/mpih-sub1.S b/mpi/aarch64/mpih-sub1.S
new file mode 100644
index 0000000..f18b1cd
--- /dev/null
+++ b/mpi/aarch64/mpih-sub1.S
@@ -0,0 +1,71 @@
+/* ARM64 sub_n -- Subtract two limb vectors of the same length > 0 and store
+ *                sum in a third limb vector.
+ *
+ *      Copyright © 2013 Jussi Kivilinna <jussi.kivilinna at iki.fi>
+ *
+ * This file is part of Libgcrypt.
+ *
+ * Libgcrypt is free software; you can redistribute it and/or modify
+ * it under the terms of the GNU Lesser General Public License as
+ * published by the Free Software Foundation; either version 2.1 of
+ * the License, or (at your option) any later version.
+ *
+ * Libgcrypt 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 Lesser General Public License for more details.
+ *
+ * You should have received a copy of the GNU Lesser General Public
+ * License along with this program; if not, see <http://www.gnu.org/licenses/>.
+ *
+ */
+
+#include "sysdep.h"
+#include "asm-syntax.h"
+
+/*******************
+ *  mpi_limb_t
+ *  _gcry_mpih_sub_n( mpi_ptr_t res_ptr,	x0
+ *		   mpi_ptr_t s1_ptr,		x1
+ *		   mpi_ptr_t s2_ptr,		x2
+ *		   mpi_size_t size)		x3
+ */
+
+.text
+
+.globl _gcry_mpih_sub_n
+.type  _gcry_mpih_sub_n,%function
+_gcry_mpih_sub_n:
+	and	x5, x3, #3;
+	subs	xzr, xzr, xzr; /* prepare carry flag for sub */
+
+	cbz	x5, .Large_loop;
+
+.Loop:
+	ldr	x4, [x1], #8;
+	sub	x3, x3, #1;
+	ldr	x11, [x2], #8;
+	and	x5, x3, #3;
+	sbcs	x4, x4, x11;
+	str	x4, [x0], #8;
+	cbz	x3, .Lend;
+	cbnz	x5, .Loop;
+
+.Large_loop:
+	ldp	x4, x6, [x1], #16;
+	ldp	x5, x7, [x2], #16;
+	ldp	x8, x10, [x1], #16;
+	ldp	x9, x11, [x2], #16;
+	sub	x3, x3, #4;
+	sbcs	x4, x4, x5;
+	sbcs	x6, x6, x7;
+	sbcs	x8, x8, x9;
+	sbcs	x10, x10, x11;
+	stp	x4, x6, [x0], #16;
+	stp	x8, x10, [x0], #16;
+	cbnz	x3, .Large_loop;
+
+.Lend:
+	cset	x0, cc;
+	ret;
+.size _gcry_mpih_sub_n,.-_gcry_mpih_sub_n;
diff --git a/mpi/config.links b/mpi/config.links
index 90d1077..a79b03b 100644
--- a/mpi/config.links
+++ b/mpi/config.links
@@ -136,6 +136,11 @@ case "${host}" in
 	mpi_extra_modules="udiv-qrnnd"
         mpi_cpu_arch="alpha"
 	;;
+    aarch64-*-*)
+	echo '/* configured for aarch64 */' >>./mpi/asm-syntax.h
+	path="aarch64"
+	mpi_cpu_arch="aarch64"
+	;;
     arm*-*-*)
 	if test "$gcry_cv_gcc_arm_platform_as_ok" = "yes" ; then
 	  echo '/* configured for arm */' >>./mpi/asm-syntax.h
diff --git a/mpi/longlong.h b/mpi/longlong.h
index 8c8260e..aad644e 100644
--- a/mpi/longlong.h
+++ b/mpi/longlong.h
@@ -268,6 +268,41 @@ extern UDItype __udiv_qrnnd ();
 #endif /* __arm__ */
 
 /***************************************
+ **********  ARM64 / Aarch64  **********
+ ***************************************/
+#if defined(__aarch64__) && W_TYPE_SIZE == 64
+#define add_ssaaaa(sh, sl, ah, al, bh, bl) \
+  __asm__ ("adds %1, %4, %5\n"                                          \
+	   "adc  %0, %2, %3"                                            \
+	   : "=r" ((sh)),                                               \
+	     "=&r" ((sl))                                               \
+	   : "r" ((UDItype)(ah)),                                       \
+	     "r" ((UDItype)(bh)),                                       \
+	     "r" ((UDItype)(al)),                                       \
+	     "r" ((UDItype)(bl)) __CLOBBER_CC)
+#define sub_ddmmss(sh, sl, ah, al, bh, bl) \
+  __asm__ ("subs %1, %4, %5\n"                                          \
+	   "sbc  %0, %2, %3"                                            \
+	   : "=r" ((sh)),                                               \
+	     "=&r" ((sl))                                               \
+	   : "r" ((UDItype)(ah)),                                       \
+	     "r" ((UDItype)(bh)),                                       \
+	     "r" ((UDItype)(al)),                                       \
+	     "r" ((UDItype)(bl)) __CLOBBER_CC)
+#define umul_ppmm(xh, xl, a, b)                                         \
+  __asm__ ("mul   %1, %2, %3\n"                                         \
+	   "umulh %0, %2, %3"                                           \
+		   : "=&r" ((xh)),                                      \
+		     "=r" ((xl))                                        \
+		   : "r" ((UDItype)(a)),                                \
+		     "r" ((UDItype)(b)))
+#define count_leading_zeros(count, x) \
+  __asm__ ("clz %0, %1"                                                 \
+		   : "=r" ((count))                                     \
+		   : "r" ((UDItype)(x)))
+#endif /* __aarch64__ */
+
+/***************************************
  **************  CLIPPER  **************
  ***************************************/
 #if defined (__clipper__) && W_TYPE_SIZE == 32




More information about the Gcrypt-devel mailing list