Add gmp-arch and udiv_qrnnd

To enable “longlong.h” removal, the udiv_qrnnd is moved to a gmp-arch.h
file.  It allows each architecture to implement its own arch-specific
optimizations. The generic implementation now uses a static inline,
which provides better type checking than the GNU extension to cast the
asm constraint (and it works better with clang).

Most of the architecture uses the generic implementation, which is
expanded from a macro, except for alpha, x86, m68k, sh, and sparc.
I kept that alpha, which uses out-of-the-line implementations and x86,
where there is no easy way to use the div{q} instruction from C code.
For the rest, the compiler generates good enough code.

The hppa also provides arch-specific implementations, but they are not
routed in “longlong.h” and thus never used.

diff --git a/soft-fp/soft-fp.h b/soft-fp/soft-fp.h
index 1f087148318cafeb8a509c766aa82dcaccc5e2e6..4c0859fa068105c67f6b0bcf5f35e3b30d8625d2 100644 (file)
--- a/soft-fp/soft-fp.h
+++ b/soft-fp/soft-fp.h
@@ -316,22 +316,6 @@
 #include "op-8.h"
 #include "op-common.h"
 
-/* Sigh.  Silly things longlong.h needs.  */
-#define UWtype         _FP_W_TYPE
-#define W_TYPE_SIZE    _FP_W_TYPE_SIZE
-
-typedef int QItype __attribute__ ((mode (QI)));
-typedef int SItype __attribute__ ((mode (SI)));
-typedef int DItype __attribute__ ((mode (DI)));
-typedef unsigned int UQItype __attribute__ ((mode (QI)));
-typedef unsigned int USItype __attribute__ ((mode (SI)));
-typedef unsigned int UDItype __attribute__ ((mode (DI)));
-#if _FP_W_TYPE_SIZE == 32
-typedef unsigned int UHWtype __attribute__ ((mode (HI)));
-#elif _FP_W_TYPE_SIZE == 64
-typedef USItype UHWtype;
-#endif
-
 #ifndef CMPtype
 # define CMPtype       int
 #endif
@@ -341,7 +325,10 @@ typedef USItype UHWtype;
 
 #ifndef umul_ppmm
 # ifdef _LIBC
+#  include <gmp.h>
+#  include <stdlib/gmp-impl.h>
 #  include <stdlib/longlong.h>
+#  include <gmp-arch.h>
 # else
 #  include "longlong.h"
 # endif

diff --git a/stdio-common/_itoa.c b/stdio-common/_itoa.c
index 08859f0dd086b5295ceeff8edcd852419e3ca580..e9cbcfab3b66dfcff3fb5a0b5da21d742d0acd9c 100644 (file)
--- a/stdio-common/_itoa.c
+++ b/stdio-common/_itoa.c
@@ -26,6 +26,7 @@
 #include <limits.h>
 #include <stdlib/gmp-impl.h>
 #include <stdlib/longlong.h>
+#include <gmp-arch.h>
 
 #include <_itoa.h>
 
@@ -308,8 +309,8 @@ _itoa (unsigned long long int value, char *buflim, unsigned int base,
                if (big_normalization_steps == 0)
                  xh = 0;
                else
-                 xh = (mp_limb_t) (value >> 64 - big_normalization_steps);
-               xl = (mp_limb_t) (value >> 32 - big_normalization_steps);
+                 xh = (mp_limb_t) (value >> (64 - big_normalization_steps));
+               xl = (mp_limb_t) (value >> (32 - big_normalization_steps));
                udiv_qrnnd (x1hi, r, xh, xl, big_base_norm);
 
                xl = ((mp_limb_t) value) << big_normalization_steps;
@@ -320,7 +321,7 @@ _itoa (unsigned long long int value, char *buflim, unsigned int base,
                  xh = x1hi;
                else
                  xh = ((x1hi << big_normalization_steps)
-                       | (x1lo >> 32 - big_normalization_steps));
+                       | (x1lo >> (32 - big_normalization_steps)));
                xl = x1lo << big_normalization_steps;
                udiv_qrnnd (t[0], x, xh, xl, big_base_norm);
                t[1] = x >> big_normalization_steps;

diff --git a/stdio-common/_itowa.c b/stdio-common/_itowa.c
index 6443b293a489a28bf6ca692160874bfa9af12a8d..2da7489502cd082d7c43ae5dc137b1ed7c6fbc94 100644 (file)
--- a/stdio-common/_itowa.c
+++ b/stdio-common/_itowa.c
@@ -21,6 +21,7 @@
 #include <limits.h>
 #include <stdlib/gmp-impl.h>
 #include <stdlib/longlong.h>
+#include <gmp-arch.h>
 
 #include <_itowa.h>
 
@@ -228,8 +229,8 @@ _itowa (unsigned long long int value, wchar_t *buflim, unsigned int base,
                if (big_normalization_steps == 0)
                  xh = 0;
                else
-                 xh = (mp_limb_t) (value >> 64 - big_normalization_steps);
-               xl = (mp_limb_t) (value >> 32 - big_normalization_steps);
+                 xh = (mp_limb_t) (value >> (64 - big_normalization_steps));
+               xl = (mp_limb_t) (value >> (32 - big_normalization_steps));
                udiv_qrnnd (x1hi, r, xh, xl, big_base_norm);
 
                xl = ((mp_limb_t) value) << big_normalization_steps;
@@ -240,7 +241,7 @@ _itowa (unsigned long long int value, wchar_t *buflim, unsigned int base,
                  xh = x1hi;
                else
                  xh = ((x1hi << big_normalization_steps)
-                       | (x1lo >> 32 - big_normalization_steps));
+                       | (x1lo >> (32 - big_normalization_steps)));
                xl = x1lo << big_normalization_steps;
                udiv_qrnnd (t[0], x, xh, xl, big_base_norm);
                t[1] = x >> big_normalization_steps;

diff --git a/stdlib/Makefile b/stdlib/Makefile
index f2b34852ecdea2b2beaf963c80bce473f262ddda..a46f166aeb896952b47e8aa1c84e9c48f270b09c 100644 (file)
--- a/stdlib/Makefile
+++ b/stdlib/Makefile
@@ -486,7 +486,6 @@ mpn-routines := \
   rshift \
   sub_n \
   submul_1 \
-  udiv_qrnnd \
   # mpn-routines
 mpn-headers = \
   asm-syntax.h \

diff --git a/stdlib/addmul_1.c b/stdlib/addmul_1.c
index 52411c3d105826c42206a21f945ba137aa1cdb60..a5bf7fd37e4c2c1f198d5fc084bb14425eca125c 100644 (file)
--- a/stdlib/addmul_1.c
+++ b/stdlib/addmul_1.c
@@ -24,6 +24,7 @@ along with the GNU MP Library; see the file COPYING.LIB.  If not, see
 #include <gmp.h>
 #include "gmp-impl.h"
 #include "longlong.h"
+#include <gmp-arch.h>
 
 mp_limb_t
 mpn_addmul_1 (register mp_ptr res_ptr, register mp_srcptr s1_ptr,

diff --git a/stdlib/divmod_1.c b/stdlib/divmod_1.c
index b91ab9e5934812022205236e2969116352db9799..676145d72720d54cbf7b45ec93de6943274bcd0d 100644 (file)
--- a/stdlib/divmod_1.c
+++ b/stdlib/divmod_1.c
@@ -28,6 +28,7 @@ along with the GNU MP Library; see the file COPYING.LIB.  If not, see
 #include <stdbit.h>
 #include "gmp-impl.h"
 #include "longlong.h"
+#include <gmp-arch.h>
 
 #ifndef UMUL_TIME
 #define UMUL_TIME 1

diff --git a/stdlib/mod_1.c b/stdlib/mod_1.c
index 74c1f6a5214b4f9522c56c6fd12ad053e2acf0c5..be3ae3d648f9c2b65a59ff3c89105bbd77c96216 100644 (file)
--- a/stdlib/mod_1.c
+++ b/stdlib/mod_1.c
@@ -25,6 +25,7 @@ along with the GNU MP Library; see the file COPYING.LIB.  If not, see
 #include <stdbit.h>
 #include "gmp-impl.h"
 #include "longlong.h"
+#include <gmp-arch.h>
 
 #ifndef UMUL_TIME
 #define UMUL_TIME 1

diff --git a/stdlib/mul_1.c b/stdlib/mul_1.c
index 225b4917d98ced03f97e129a0f8723bcebe31df6..0a04a8eb1e3cf27f1b7bf39c3f952eb11080f0c7 100644 (file)
--- a/stdlib/mul_1.c
+++ b/stdlib/mul_1.c
@@ -22,6 +22,7 @@ along with the GNU MP Library; see the file COPYING.LIB.  If not, see
 #include <gmp.h>
 #include "gmp-impl.h"
 #include "longlong.h"
+#include <gmp-arch.h>
 
 mp_limb_t
 mpn_mul_1 (register mp_ptr res_ptr, register mp_srcptr s1_ptr,

diff --git a/stdlib/strtod_l.c b/stdlib/strtod_l.c
index 5814475e4059e5559abd77db9435093d1c076cb8..7618adb28547c0a452b0fd7c1f1eae13b2e2439f 100644 (file)
--- a/stdlib/strtod_l.c
+++ b/stdlib/strtod_l.c
@@ -83,6 +83,7 @@ extern double ____strtod_l_internal (const char *, char **, int, locale_t);
 #include <gmp.h>
 #include "gmp-impl.h"
 #include "longlong.h"
+#include "gmp-arch.h"
 #include "fpioconst.h"
 
 #include <assert.h>

diff --git a/stdlib/submul_1.c b/stdlib/submul_1.c
index 3eb4d015f558a27d4470aa446f8577c2d87a3802..0005c2e04e6e2ee6bfc10b6bbe47460ae82be788 100644 (file)
--- a/stdlib/submul_1.c
+++ b/stdlib/submul_1.c
@@ -24,6 +24,7 @@ along with the GNU MP Library; see the file COPYING.LIB.  If not, see
 #include <gmp.h>
 #include "gmp-impl.h"
 #include "longlong.h"
+#include <gmp-arch.h>
 
 mp_limb_t
 mpn_submul_1 (register mp_ptr res_ptr, register mp_srcptr s1_ptr,

diff --git a/stdlib/udiv_qrnnd.c b/stdlib/udiv_qrnnd.c
deleted file mode 100644 (file)
index d32796c..0000000
--- a/stdlib/udiv_qrnnd.c
+++ /dev/null
@@ -1,10 +0,0 @@
-/* For some machines GNU MP needs to define an auxiliary function:
-
-   udiv_qrnnd (quotient, remainder, high_numerator, low_numerator, denominator)
-
-   Divides a two-word unsigned integer, composed by the integers
-   HIGH_NUMERATOR and LOW_NUMERATOR, by DENOMINATOR and places the quotient
-   in QUOTIENT and the remainder in REMAINDER.  HIGH_NUMERATOR must be less
-   than DENOMINATOR for correct operation.  If, in addition, the most
-   significant bit of DENOMINATOR must be 1, then the pre-processor symbol
-   UDIV_NEEDS_NORMALIZATION is defined to 1.  */

diff --git a/sysdeps/alpha/Makefile b/sysdeps/alpha/Makefile
index 39dea0d72cb4210a08d6b1cefdf2bb2c7c2d5785..cae60186c82080cc79f01bc73d3a4fc476ea399b 100644 (file)
--- a/sysdeps/alpha/Makefile
+++ b/sysdeps/alpha/Makefile
@@ -32,6 +32,10 @@ ifeq ($(subdir),string)
 sysdep_routines += stxcpy stxncpy
 endif
 
+ifeq ($(subdir),stdlib)
+sysdep_routines += udiv_qrnnd
+endif
+
 ifeq ($(subdir),elf)
 # The ld.so startup code cannot use literals until it self-relocates.
 CFLAGS-rtld.c = -mbuild-constants

diff --git a/sysdeps/alpha/gmp-arch.h b/sysdeps/alpha/gmp-arch.h
new file mode 100644 (file)
index 0000000..c6e4c8f
--- /dev/null
+++ b/sysdeps/alpha/gmp-arch.h
@@ -0,0 +1,2 @@
+#define __EXTERNAL_QRNND 1
+#include <sysdeps/generic/gmp-arch.h>

diff --git a/sysdeps/generic/gmp-arch.h b/sysdeps/generic/gmp-arch.h
new file mode 100644 (file)
index 0000000..623b0ae
--- /dev/null
+++ b/sysdeps/generic/gmp-arch.h
@@ -0,0 +1,113 @@
+/* Multiprecision generic functions.
+   Copyright (C) 2025 Free Software Foundation, Inc.
+   This file is part of the GNU C Library.
+
+   The GNU C Library 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.
+
+   The GNU C 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
+   Lesser General Public License for more details.
+
+   You should have received a copy of the GNU Lesser General Public
+   License along with the GNU C Library; if not, see
+   <https://www.gnu.org/licenses/>.  */
+
+#ifndef __GMP_ARCH_H
+#define __GMP_ARCH_H
+
+#include <gmp.h>
+
+#define LL_B ((mp_limb_t) 1 << (BITS_PER_MP_LIMB / 2))
+
+static __always_inline mp_limb_t
+ll_lowpart (mp_limb_t t)
+{
+  return t & (LL_B - 1);
+}
+
+static __always_inline mp_limb_t
+ll_highpart (mp_limb_t t)
+{
+  return t >> (BITS_PER_MP_LIMB / 2);
+}
+
+/* udiv_qrnnd(quotient, remainder, high_numerator, low_numerator,
+   denominator) divides a UDWtype, composed by the UWtype integers
+   HIGH_NUMERATOR and LOW_NUMERATOR, by DENOMINATOR and places the quotient
+   in QUOTIENT and the remainder in REMAINDER.  HIGH_NUMERATOR must be less
+   than DENOMINATOR for correct operation.  If, in addition, the most
+   significant bit of DENOMINATOR must be 1, then the pre-processor symbol
+   UDIV_NEEDS_NORMALIZATION is defined to 1.  */
+#ifndef __EXTERNAL_QRNND
+static __always_inline void
+udiv_qrnnd_generic (mp_limb_t *q, mp_limb_t *r, mp_limb_t n1, mp_limb_t n0,
+                   mp_limb_t d)
+{
+  mp_limb_t d1 = ll_highpart (d),
+            d0 = ll_lowpart (d),
+            q1, q0;
+  mp_limb_t r1, r0, m;
+
+  r1 = n1 % d1;
+  q1 = n1 / d1;
+  m = q1 * d0;
+  r1 = r1 * LL_B | ll_highpart (n0);
+  if (r1 < m)
+    {
+      q1--;
+      r1 += d;
+      if (r1 >= d)
+        if (r1 < m)
+          {
+            q1--;
+            r1 += d;
+          }
+    }
+  r1 -= m;
+
+  r0 = r1 % d1;
+  q0 = r1 / d1;
+  m = q0 * d0;
+  r0 = r0 * LL_B | ll_lowpart (n0);
+  if (r0 < m)
+    {
+      q0--;
+      r0 += d;
+      if (r0 >= d)
+        if (r0 < m)
+          {
+            q0--;
+            r0 += d;
+          }
+    }
+  r0 -= m;
+
+  *q = q1 * LL_B | q0;
+  *r = r0;
+}
+# undef UDIV_NEEDS_NORMALIZATION
+# define UDIV_NEEDS_NORMALIZATION 1
+#else
+extern mp_limb_t __udiv_qrnnd (mp_limb_t *, mp_limb_t, mp_limb_t, mp_limb_t)
+     attribute_hidden;
+
+static __always_inline void
+udiv_qrnnd_generic (mp_limb_t *q, mp_limb_t *r, mp_limb_t n1, mp_limb_t n0,
+                   mp_limb_t d)
+{
+  *q = __udiv_qrnnd (r, n1, n0, d);
+}
+# undef UDIV_NEEDS_NORMALIZATION
+# define UDIV_NEEDS_NORMALIZATION 0
+#endif
+
+#undef udiv_qrnnd
+#define udiv_qrnnd(__q, __r, __n1, __n0, __d) \
+  udiv_qrnnd_generic (&__q, &__r, __n1, __n0, __d)
+
+
+#endif /* __GMP_ARHC_H */

diff --git a/sysdeps/hppa/hppa1.1/udiv_qrnnd.S b/sysdeps/hppa/hppa1.1/udiv_qrnnd.S
deleted file mode 100644 (file)
index 2f8101b..0000000
--- a/sysdeps/hppa/hppa1.1/udiv_qrnnd.S
+++ /dev/null
@@ -1,77 +0,0 @@
-;! HP-PA  __udiv_qrnnd division support, used from longlong.h.
-;! This version runs fast on PA 7000 and later.
-
-;! Copyright (C) 1993-2025 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 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.
-
-;! 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 Lesser General Public
-;! License for more details.
-
-;! You should have received a copy of the GNU Lesser General Public License
-;! along with the GNU MP Library.  If not, see
-;! <https://www.gnu.org/licenses/>.
-
-
-;! INPUT PARAMETERS
-;! rem_ptr     gr26
-;! n1          gr25
-;! n0          gr24
-;! d           gr23
-
-       .text
-L$0000:
-       .word           0x43f00000
-       .word           0x0
-       .export         __udiv_qrnnd
-__udiv_qrnnd:
-       .proc
-       .callinfo       frame=64,no_calls
-       .entry
-       ldo             64(%r30),%r30
-
-       stws            %r25,-16(%r30)  ;! n_hi
-       stws            %r24,-12(%r30)  ;! n_lo
-       b,l             L$0,%r1
-       ldo             L$0000-L$0(%r1),%r1
-L$0:
-       fldds           -16(%r30),%fr5
-       stws            %r23,-12(%r30)
-       comib,<=        0,%r25,L$1
-       fcnvxf,dbl,dbl  %fr5,%fr5
-       fldds           0(%r1),%fr4
-       fadd,dbl        %fr4,%fr5,%fr5
-L$1:
-       fcpy,sgl        %fr0,%fr6L
-       fldws           -12(%r30),%fr6R
-       fcnvxf,dbl,dbl  %fr6,%fr4
-
-       fdiv,dbl        %fr5,%fr4,%fr5
-
-       fcnvfx,dbl,dbl  %fr5,%fr4
-       fstws           %fr4R,-16(%r30)
-       xmpyu           %fr4R,%fr6R,%fr6
-       ldws            -16(%r30),%r28
-       fstds           %fr6,-16(%r30)
-       ldws            -12(%r30),%r21
-       ldws            -16(%r30),%r20
-       sub             %r24,%r21,%r22
-       subb            %r25,%r20,%r1
-       comib,=         0,%r1,L$2
-       ldo             -64(%r30),%r30
-
-       add             %r22,%r23,%r22
-       ldo             -1(%r28),%r28
-L$2:
-       bv              0(%r2)
-       stws            %r22,0(%r26)
-
-       .exit
-       .procend

diff --git a/sysdeps/hppa/udiv_qrnnd.S b/sysdeps/hppa/udiv_qrnnd.S
deleted file mode 100644 (file)
index 317fcda..0000000
--- a/sysdeps/hppa/udiv_qrnnd.S
+++ /dev/null
@@ -1,285 +0,0 @@
-;! HP-PA  __udiv_qrnnd division support, used from longlong.h.
-;! This version runs fast on pre-PA7000 CPUs.
-
-;! Copyright (C) 1993-2025 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 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.
-
-;! 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 Lesser General Public
-;! License for more details.
-
-;! You should have received a copy of the GNU Lesser General Public License
-;! along with the GNU MP Library.  If not, see
-;! <https://www.gnu.org/licenses/>.
-
-
-;! INPUT PARAMETERS
-;! rem_ptr     gr26
-;! n1          gr25
-;! n0          gr24
-;! d           gr23
-
-;! The code size is a bit excessive.  We could merge the last two ds;addc
-;! sequences by simply moving the "bb,< Odd" instruction down.  The only
-;! trouble is the FFFFFFFF code that would need some hacking.
-
-       .text
-       .export         __udiv_qrnnd
-__udiv_qrnnd:
-       .proc
-       .callinfo       frame=0,no_calls
-       .entry
-
-       comb,<          %r23,%r0,L$largedivisor
-        sub            %r0,%r23,%r1            ;! clear cy as side-effect
-       ds              %r0,%r1,%r0
-       addc            %r24,%r24,%r24
-       ds              %r25,%r23,%r25
-       addc            %r24,%r24,%r24
-       ds              %r25,%r23,%r25
-       addc            %r24,%r24,%r24
-       ds              %r25,%r23,%r25
-       addc            %r24,%r24,%r24
-       ds              %r25,%r23,%r25
-       addc            %r24,%r24,%r24
-       ds              %r25,%r23,%r25
-       addc            %r24,%r24,%r24
-       ds              %r25,%r23,%r25
-       addc            %r24,%r24,%r24
-       ds              %r25,%r23,%r25
-       addc            %r24,%r24,%r24
-       ds              %r25,%r23,%r25
-       addc            %r24,%r24,%r24
-       ds              %r25,%r23,%r25
-       addc            %r24,%r24,%r24
-       ds              %r25,%r23,%r25
-       addc            %r24,%r24,%r24
-       ds              %r25,%r23,%r25
-       addc            %r24,%r24,%r24
-       ds              %r25,%r23,%r25
-       addc            %r24,%r24,%r24
-       ds              %r25,%r23,%r25
-       addc            %r24,%r24,%r24
-       ds              %r25,%r23,%r25
-       addc            %r24,%r24,%r24
-       ds              %r25,%r23,%r25
-       addc            %r24,%r24,%r24
-       ds              %r25,%r23,%r25
-       addc            %r24,%r24,%r24
-       ds              %r25,%r23,%r25
-       addc            %r24,%r24,%r24
-       ds              %r25,%r23,%r25
-       addc            %r24,%r24,%r24
-       ds              %r25,%r23,%r25
-       addc            %r24,%r24,%r24
-       ds              %r25,%r23,%r25
-       addc            %r24,%r24,%r24
-       ds              %r25,%r23,%r25
-       addc            %r24,%r24,%r24
-       ds              %r25,%r23,%r25
-       addc            %r24,%r24,%r24
-       ds              %r25,%r23,%r25
-       addc            %r24,%r24,%r24
-       ds              %r25,%r23,%r25
-       addc            %r24,%r24,%r24
-       ds              %r25,%r23,%r25
-       addc            %r24,%r24,%r24
-       ds              %r25,%r23,%r25
-       addc            %r24,%r24,%r24
-       ds              %r25,%r23,%r25
-       addc            %r24,%r24,%r24
-       ds              %r25,%r23,%r25
-       addc            %r24,%r24,%r24
-       ds              %r25,%r23,%r25
-       addc            %r24,%r24,%r24
-       ds              %r25,%r23,%r25
-       addc            %r24,%r24,%r24
-       ds              %r25,%r23,%r25
-       addc            %r24,%r24,%r28
-       ds              %r25,%r23,%r25
-       comclr,>=       %r25,%r0,%r0
-       addl            %r25,%r23,%r25
-       stws            %r25,0(%r26)
-       bv              0(%r2)
-        addc           %r28,%r28,%r28
-
-L$largedivisor:
-       extru           %r24,31,1,%r20          ;! r20 = n0 & 1
-       bb,<            %r23,31,L$odd
-        extru          %r23,30,31,%r22         ;! r22 = d >> 1
-       shd             %r25,%r24,1,%r24        ;! r24 = new n0
-       extru           %r25,30,31,%r25         ;! r25 = new n1
-       sub             %r0,%r22,%r21
-       ds              %r0,%r21,%r0
-       addc            %r24,%r24,%r24
-       ds              %r25,%r22,%r25
-       addc            %r24,%r24,%r24
-       ds              %r25,%r22,%r25
-       addc            %r24,%r24,%r24
-       ds              %r25,%r22,%r25
-       addc            %r24,%r24,%r24
-       ds              %r25,%r22,%r25
-       addc            %r24,%r24,%r24
-       ds              %r25,%r22,%r25
-       addc            %r24,%r24,%r24
-       ds              %r25,%r22,%r25
-       addc            %r24,%r24,%r24
-       ds              %r25,%r22,%r25
-       addc            %r24,%r24,%r24
-       ds              %r25,%r22,%r25
-       addc            %r24,%r24,%r24
-       ds              %r25,%r22,%r25
-       addc            %r24,%r24,%r24
-       ds              %r25,%r22,%r25
-       addc            %r24,%r24,%r24
-       ds              %r25,%r22,%r25
-       addc            %r24,%r24,%r24
-       ds              %r25,%r22,%r25
-       addc            %r24,%r24,%r24
-       ds              %r25,%r22,%r25
-       addc            %r24,%r24,%r24
-       ds              %r25,%r22,%r25
-       addc            %r24,%r24,%r24
-       ds              %r25,%r22,%r25
-       addc            %r24,%r24,%r24
-       ds              %r25,%r22,%r25
-       addc            %r24,%r24,%r24
-       ds              %r25,%r22,%r25
-       addc            %r24,%r24,%r24
-       ds              %r25,%r22,%r25
-       addc            %r24,%r24,%r24
-       ds              %r25,%r22,%r25
-       addc            %r24,%r24,%r24
-       ds              %r25,%r22,%r25
-       addc            %r24,%r24,%r24
-       ds              %r25,%r22,%r25
-       addc            %r24,%r24,%r24
-       ds              %r25,%r22,%r25
-       addc            %r24,%r24,%r24
-       ds              %r25,%r22,%r25
-       addc            %r24,%r24,%r24
-       ds              %r25,%r22,%r25
-       addc            %r24,%r24,%r24
-       ds              %r25,%r22,%r25
-       addc            %r24,%r24,%r24
-       ds              %r25,%r22,%r25
-       addc            %r24,%r24,%r24
-       ds              %r25,%r22,%r25
-       addc            %r24,%r24,%r24
-       ds              %r25,%r22,%r25
-       addc            %r24,%r24,%r24
-       ds              %r25,%r22,%r25
-       addc            %r24,%r24,%r24
-       ds              %r25,%r22,%r25
-       addc            %r24,%r24,%r24
-       ds              %r25,%r22,%r25
-       addc            %r24,%r24,%r24
-       ds              %r25,%r22,%r25
-       comclr,>=       %r25,%r0,%r0
-       addl            %r25,%r22,%r25
-       sh1addl         %r25,%r20,%r25
-       stws            %r25,0(%r26)
-       bv              0(%r2)
-        addc           %r24,%r24,%r28
-
-L$odd: addib,sv,n      1,%r22,L$FF..           ;! r22 = (d / 2 + 1)
-       shd             %r25,%r24,1,%r24        ;! r24 = new n0
-       extru           %r25,30,31,%r25         ;! r25 = new n1
-       sub             %r0,%r22,%r21
-       ds              %r0,%r21,%r0
-       addc            %r24,%r24,%r24
-       ds              %r25,%r22,%r25
-       addc            %r24,%r24,%r24
-       ds              %r25,%r22,%r25
-       addc            %r24,%r24,%r24
-       ds              %r25,%r22,%r25
-       addc            %r24,%r24,%r24
-       ds              %r25,%r22,%r25
-       addc            %r24,%r24,%r24
-       ds              %r25,%r22,%r25
-       addc            %r24,%r24,%r24
-       ds              %r25,%r22,%r25
-       addc            %r24,%r24,%r24
-       ds              %r25,%r22,%r25
-       addc            %r24,%r24,%r24
-       ds              %r25,%r22,%r25
-       addc            %r24,%r24,%r24
-       ds              %r25,%r22,%r25
-       addc            %r24,%r24,%r24
-       ds              %r25,%r22,%r25
-       addc            %r24,%r24,%r24
-       ds              %r25,%r22,%r25
-       addc            %r24,%r24,%r24
-       ds              %r25,%r22,%r25
-       addc            %r24,%r24,%r24
-       ds              %r25,%r22,%r25
-       addc            %r24,%r24,%r24
-       ds              %r25,%r22,%r25
-       addc            %r24,%r24,%r24
-       ds              %r25,%r22,%r25
-       addc            %r24,%r24,%r24
-       ds              %r25,%r22,%r25
-       addc            %r24,%r24,%r24
-       ds              %r25,%r22,%r25
-       addc            %r24,%r24,%r24
-       ds              %r25,%r22,%r25
-       addc            %r24,%r24,%r24
-       ds              %r25,%r22,%r25
-       addc            %r24,%r24,%r24
-       ds              %r25,%r22,%r25
-       addc            %r24,%r24,%r24
-       ds              %r25,%r22,%r25
-       addc            %r24,%r24,%r24
-       ds              %r25,%r22,%r25
-       addc            %r24,%r24,%r24
-       ds              %r25,%r22,%r25
-       addc            %r24,%r24,%r24
-       ds              %r25,%r22,%r25
-       addc            %r24,%r24,%r24
-       ds              %r25,%r22,%r25
-       addc            %r24,%r24,%r24
-       ds              %r25,%r22,%r25
-       addc            %r24,%r24,%r24
-       ds              %r25,%r22,%r25
-       addc            %r24,%r24,%r24
-       ds              %r25,%r22,%r25
-       addc            %r24,%r24,%r24
-       ds              %r25,%r22,%r25
-       addc            %r24,%r24,%r24
-       ds              %r25,%r22,%r25
-       addc            %r24,%r24,%r24
-       ds              %r25,%r22,%r25
-       addc            %r24,%r24,%r24
-       ds              %r25,%r22,%r25
-       addc            %r24,%r24,%r28
-       comclr,>=       %r25,%r0,%r0
-       addl            %r25,%r22,%r25
-       sh1addl         %r25,%r20,%r25
-;! We have computed (n1,,n0) / (d + 1), q' = r28, r' = r25
-       add,nuv         %r28,%r25,%r25
-       addl            %r25,%r1,%r25
-       addc            %r0,%r28,%r28
-       sub,<<          %r25,%r23,%r0
-       addl            %r25,%r1,%r25
-       stws            %r25,0(%r26)
-       bv              0(%r2)
-        addc           %r0,%r28,%r28
-
-;! This is just a special case of the code above.
-;! We come here when d == 0xFFFFFFFF
-L$FF..:        add,uv          %r25,%r24,%r24
-       sub,<<          %r24,%r23,%r0
-       ldo             1(%r24),%r24
-       stws            %r24,0(%r26)
-       bv              0(%r2)
-        addc           %r0,%r25,%r28
-
-       .exit
-       .procend

diff --git a/sysdeps/ieee754/dbl-64/dbl2mpn.c b/sysdeps/ieee754/dbl-64/dbl2mpn.c
index d69973419b7ba098a00cfefe76948261b9a61f4d..cacf5da7d80449726f28eb856466084c8bbc018b 100644 (file)
--- a/sysdeps/ieee754/dbl-64/dbl2mpn.c
+++ b/sysdeps/ieee754/dbl-64/dbl2mpn.c
@@ -18,6 +18,7 @@
 #include "gmp.h"
 #include "gmp-impl.h"
 #include "longlong.h"
+#include <gmp-arch.h>
 #include <ieee754.h>
 #include <float.h>
 #include <stdlib.h>

diff --git a/sysdeps/ieee754/ldbl-128/ldbl2mpn.c b/sysdeps/ieee754/ldbl-128/ldbl2mpn.c
index d3f3476d569db92202045a6caaee2146764ba775..fd75db1271c1006ae46adbd2fbde9246a40ddfc5 100644 (file)
--- a/sysdeps/ieee754/ldbl-128/ldbl2mpn.c
+++ b/sysdeps/ieee754/ldbl-128/ldbl2mpn.c
@@ -18,6 +18,7 @@
 #include "gmp.h"
 #include "gmp-impl.h"
 #include "longlong.h"
+#include <gmp-arch.h>
 #include <ieee754.h>
 #include <float.h>
 #include <math.h>

diff --git a/sysdeps/ieee754/ldbl-96/ldbl2mpn.c b/sysdeps/ieee754/ldbl-96/ldbl2mpn.c
index 982f4273803f1a517ecd1824e828661c83ffa96d..6c7aadbca2e4e014a68ef488207cbec4d464f60d 100644 (file)
--- a/sysdeps/ieee754/ldbl-96/ldbl2mpn.c
+++ b/sysdeps/ieee754/ldbl-96/ldbl2mpn.c
@@ -18,6 +18,7 @@
 #include "gmp.h"
 #include "gmp-impl.h"
 #include "longlong.h"
+#include <gmp-arch.h>
 #include <ieee754.h>
 #include <float.h>
 #include <math.h>

diff --git a/sysdeps/wordsize-32/divdi3.c b/sysdeps/wordsize-32/divdi3.c
index 387022ab14116a06afc13f9b22c75c7db459326d..360094134f11ed44ef379745351e38a3e4d9d82c 100644 (file)
--- a/sysdeps/wordsize-32/divdi3.c
+++ b/sysdeps/wordsize-32/divdi3.c
@@ -25,20 +25,16 @@
 #error This is for 32-bit targets only
 #endif
 
-typedef unsigned int UQItype   __attribute__ ((mode (QI)));
-typedef          int SItype    __attribute__ ((mode (SI)));
-typedef unsigned int USItype   __attribute__ ((mode (SI)));
-typedef          int DItype    __attribute__ ((mode (DI)));
-typedef unsigned int UDItype   __attribute__ ((mode (DI)));
 #define Wtype SItype
 #define HWtype SItype
 #define DWtype DItype
-#define UWtype USItype
 #define UHWtype USItype
 #define UDWtype UDItype
-#define W_TYPE_SIZE 32
 
+#include <gmp.h>
+#include <stdlib/gmp-impl.h>
 #include <stdlib/longlong.h>
+#include <gmp-arch.h>
 
 #if __BYTE_ORDER == __BIG_ENDIAN
 struct DWstruct { Wtype high, low;};

diff --git a/sysdeps/x86/gmp-arch.h b/sysdeps/x86/gmp-arch.h
new file mode 100644 (file)
index 0000000..5c8b883
--- /dev/null
+++ b/sysdeps/x86/gmp-arch.h
@@ -0,0 +1,50 @@
+/* Multiprecision generic functions.
+   Copyright (C) 2025 Free Software Foundation, Inc.
+   This file is part of the GNU C Library.
+
+   The GNU C Library 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.
+
+   The GNU C 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
+   Lesser General Public License for more details.
+
+   You should have received a copy of the GNU Lesser General Public
+   License along with the GNU C Library; if not, see
+   <https://www.gnu.org/licenses/>.  */
+
+#ifndef __GMP_ARCH_H
+#define __GMP_ARCH_H
+
+#include <gmp.h>
+
+static __always_inline void
+udiv_qrnnd_x86 (mp_limb_t *q, mp_limb_t *r, mp_limb_t n1, mp_limb_t n0,
+               mp_limb_t d)
+{
+#ifdef __x86_64__
+  asm ("div{q} %4"
+       : "=a" (*q),
+         "=d" (*r)
+       : "0" (n0),
+        "1" (n1),
+        "rm" (d));
+#else
+  asm ("div{l} %4"
+       : "=a" (*q),
+         "=d" (*r)
+       : "0" (n0),
+        "1" (n1),
+        "rm" (d));
+#endif
+}
+#undef UDIV_NEEDS_NORMALIZATION
+#define UDIV_NEEDS_NORMALIZATION 0
+#undef udiv_qrnnd
+#define udiv_qrnnd(__q, __r, __n1, __n0, __d) \
+  udiv_qrnnd_x86 (&__q, &__r, __n1, __n0, __d)
+
+#endif

diff --git a/sysdeps/x86/ldbl2mpn.c b/sysdeps/x86/ldbl2mpn.c
index c41fb8da147a70a514be008ca99085647845efc3..593c472e8abb23fe322ae9795c81292c2d62df24 100644 (file)
--- a/sysdeps/x86/ldbl2mpn.c
+++ b/sysdeps/x86/ldbl2mpn.c
@@ -18,6 +18,7 @@
 #include "gmp.h"
 #include "gmp-impl.h"
 #include "longlong.h"
+#include <gmp-arch.h>
 #include <ieee754.h>
 #include <float.h>
 #include <stdlib.h>