diff options
| -rw-r--r-- | stdlib/gmp-impl.h | 20 | ||||
| -rw-r--r-- | stdlib/gmp.h | 11 | ||||
| -rw-r--r-- | stdlib/longlong.h | 88 | ||||
| -rw-r--r-- | sysdeps/generic/mul_n.c | 20 | ||||
| -rw-r--r-- | sysdeps/i386/i586/lshift.S | 11 | ||||
| -rw-r--r-- | sysdeps/i386/i586/rshift.S | 11 | ||||
| -rw-r--r-- | sysdeps/m68k/add_n.S | 76 | ||||
| -rw-r--r-- | sysdeps/m68k/sub_n.S | 76 | ||||
| -rw-r--r-- | sysdeps/m88k/add_n.s | 103 | ||||
| -rw-r--r-- | sysdeps/m88k/mul_1.s | 128 | ||||
| -rw-r--r-- | sysdeps/m88k/sub_n.s | 104 | ||||
| -rw-r--r-- | sysdeps/rs6000/add_n.s | 2 | ||||
| -rw-r--r-- | sysdeps/rs6000/sub_n.s | 2 | ||||
| -rw-r--r-- | sysdeps/sparc/add_n.S | 304 | ||||
| -rw-r--r-- | sysdeps/sparc/lshift.S | 94 | ||||
| -rw-r--r-- | sysdeps/sparc/rshift.S | 91 | ||||
| -rw-r--r-- | sysdeps/sparc/sub_n.S | 391 |
17 files changed, 1246 insertions, 286 deletions
diff --git a/stdlib/gmp-impl.h b/stdlib/gmp-impl.h index 0d2a8fcede..2f0956d960 100644 --- a/stdlib/gmp-impl.h +++ b/stdlib/gmp-impl.h @@ -179,24 +179,22 @@ void _mp_default_free (); strings in base 2..36. */ struct bases { - /* Number of digits in the conversion base that always fits in - an mp_limb. For example, for base 10 this is 10, since - 2**32 = 4294967296 has ten digits. */ + /* Number of digits in the conversion base that always fits in an mp_limb. + For example, for base 10 on a machine where a mp_limb has 32 bits this + is 9, since 10**9 is the largest number that fits into a mp_limb. */ int chars_per_limb; /* log(2)/log(conversion_base) */ float chars_per_bit_exactly; - /* big_base is conversion_base**chars_per_limb, i.e. the biggest - number that fits a word, built by factors of conversion_base. - Exception: For 2, 4, 8, etc, big_base is log2(base), i.e. the - number of bits used to represent each digit in the base. */ + /* base**chars_per_limb, i.e. the biggest number that fits a word, built by + factors of base. Exception: For 2, 4, 8, etc, big_base is log2(base), + i.e. the number of bits used to represent each digit in the base. */ mp_limb big_base; - /* big_base_inverted is a BITS_PER_MP_LIMB bit approximation to - 1/big_base, represented as a fixed-point number. Instead of - dividing by big_base an application can choose to multiply - by big_base_inverted. */ + /* A BITS_PER_MP_LIMB bit approximation to 1/big_base, represented as a + fixed-point number. Instead of dividing by big_base an application can + choose to multiply by big_base_inverted. */ mp_limb big_base_inverted; }; diff --git a/stdlib/gmp.h b/stdlib/gmp.h index 0b2cb29014..243779996d 100644 --- a/stdlib/gmp.h +++ b/stdlib/gmp.h @@ -23,6 +23,7 @@ the Free Software Foundation, Inc., 675 Mass Ave, Cambridge, MA 02139, USA. */ #ifndef __GNU_MP__ #define __need_size_t #include <stddef.h> +#undef __need_size_t #if defined (__STDC__) #define __gmp_const const @@ -40,7 +41,7 @@ the Free Software Foundation, Inc., 675 Mass Ave, Cambridge, MA 02139, USA. */ typedef unsigned int mp_limb; typedef int mp_limb_signed; #else -#if _LONG_LONG_LIMB +#ifdef _LONG_LONG_LIMB typedef unsigned long long int mp_limb; typedef long long int mp_limb_signed; #else @@ -110,11 +111,11 @@ typedef __mpq_struct mpq_t[1]; typedef struct { - mp_size_t alloc; /* Number of *limbs* allocated and pointed - to by the D field. */ mp_size_t prec; /* Max precision, in number of `mp_limb's. Set by mpf_init and modified by - mpf_set_prec. */ + mpf_set_prec. The area pointed to + by the `d' field contains `prec' + 1 + limbs. */ mp_size_t size; /* abs(SIZE) is the number of limbs the last field points to. If SIZE is negative this is a negative @@ -127,7 +128,7 @@ typedef struct typedef __mpf_struct mpf_t[1]; /* Types for function declarations in gmp files. */ -/* ??? Should not pollute user name space ??? */ +/* ??? Should not pollute user name space with these ??? */ typedef __gmp_const __mpz_struct *mpz_srcptr; typedef __mpz_struct *mpz_ptr; typedef __gmp_const __mpf_struct *mpf_srcptr; diff --git a/stdlib/longlong.h b/stdlib/longlong.h index bbb92e3af8..e52bf32dba 100644 --- a/stdlib/longlong.h +++ b/stdlib/longlong.h @@ -139,6 +139,7 @@ the Free Software Foundation, Inc., 675 Mass Ave, Cambridge, MA 02139, USA. */ __asm__ ("clz %0,%1" \ : "=r" ((USItype)(count)) \ : "r" ((USItype)(x))) +#define COUNT_LEADING_ZEROS_0 32 #endif /* __a29k__ */ #if defined (__alpha__) && W_TYPE_SIZE == 64 @@ -298,9 +299,9 @@ extern UDItype __udiv_qrnnd (); struct {USItype __h, __l;} __i; \ } __xx; \ __asm__ ("xmpyu %1,%2,%0" \ - : "=fx" (__xx.__ll) \ - : "fx" ((USItype)(u)), \ - "fx" ((USItype)(v))); \ + : "=*f" (__xx.__ll) \ + : "*f" ((USItype)(u)), \ + "*f" ((USItype)(v))); \ (wh) = __xx.__i.__h; \ (wl) = __xx.__i.__l; \ } while (0) @@ -339,7 +340,7 @@ extern USItype __udiv_qrnnd (); sub %0,%1,%0 ; Subtract it. " : "=r" (count), "=r" (__tmp) : "1" (x)); \ } while (0) -#endif +#endif /* hppa */ #if (defined (__i370__) || defined (__mvs__)) && W_TYPE_SIZE == 32 #define umul_ppmm(xh, xl, m0, m1) \ @@ -431,7 +432,29 @@ extern USItype __udiv_qrnnd (); #endif #endif /* 80x86 */ +#if defined (__i860__) && W_TYPE_SIZE == 32 +#define rshift_rhlc(r,h,l,c) \ + __asm__ ("shr %3,r0,r0\;shrd %1,%2,%0" \ + "=r" (r) : "r" (h), "r" (l), "rn" (c)) +#endif /* i860 */ + #if defined (__i960__) && W_TYPE_SIZE == 32 +#define add_ssaaaa(sh, sl, ah, al, bh, bl) \ + __asm__ ("cmpo 1,0\;addc %5,%4,%1\;addc %3,%2,%0" \ + : "=r" ((USItype)(sh)), \ + "=&r" ((USItype)(sl)) \ + : "%dI" ((USItype)(ah)), \ + "dI" ((USItype)(bh)), \ + "%dI" ((USItype)(al)), \ + "dI" ((USItype)(bl))) +#define sub_ddmmss(sh, sl, ah, al, bh, bl) \ + __asm__ ("cmpo 0,0\;subc %5,%4,%1\;subc %3,%2,%0" \ + : "=r" ((USItype)(sh)), \ + "=&r" ((USItype)(sl)) \ + : "dI" ((USItype)(ah)), \ + "dI" ((USItype)(bh)), \ + "dI" ((USItype)(al)), \ + "dI" ((USItype)(bl))) #define umul_ppmm(w1, w0, u, v) \ ({union {UDItype __ll; \ struct {USItype __l, __h;} __i; \ @@ -448,7 +471,39 @@ extern USItype __udiv_qrnnd (); : "%dI" ((USItype)(u)), \ "dI" ((USItype)(v))); \ __w; }) -#endif /* __i960__ */ +#define udiv_qrnnd(q, r, nh, nl, d) \ + do { \ + union {UDItype __ll; \ + struct {USItype __l, __h;} __i; \ + } __nn; \ + __nn.__i.__h = (nh); __nn.__i.__l = (nl); \ + __asm__ ("ediv %d,%n,%0" \ + : "=d" (__rq.__ll) \ + : "dI" (__nn.__ll), \ + "dI" ((USItype)(d))); \ + (r) = __rq.__i.__l; (q) = __rq.__i.__h; \ + } while (0) +#define count_leading_zeros(count, x) \ + do { \ + USItype __cbtmp; \ + __asm__ ("scanbit %1,%0" \ + : "=r" (__cbtmp) \ + : "r" ((USItype)(x))); \ + (count) = __cbtmp ^ 31; \ + } while (0) +#define COUNT_LEADING_ZEROS_0 (-32) /* sic */ +#if defined (__i960mx) /* what is the proper symbol to test??? */ +#define rshift_rhlc(r,h,l,c) \ + do { \ + union {UDItype __ll; \ + struct {USItype __l, __h;} __i; \ + } __nn; \ + __nn.__i.__h = (h); __nn.__i.__l = (l); \ + __asm__ ("shre %2,%1,%0" \ + : "=d" (r) : "dI" (__nn.__ll), "dI" (c)); \ + } +#endif /* i960mx */ +#endif /* i960 */ #if (defined (__mc68000__) || defined (__mc68020__) || defined (__NeXT__) || defined(mc68020)) && W_TYPE_SIZE == 32 #define add_ssaaaa(sh, sl, ah, al, bh, bl) \ @@ -469,7 +524,7 @@ extern USItype __udiv_qrnnd (); "d" ((USItype)(bh)), \ "1" ((USItype)(al)), \ "g" ((USItype)(bl))) -#if (defined (__mc68020__) || defined (__NeXT__) || defined(mc68020)) && W_TYPE_SIZE == 32 +#if (defined (__mc68020__) || defined (__NeXT__) || defined(mc68020)) #define umul_ppmm(w1, w0, u, v) \ __asm__ ("mulu%.l %3,%1:%0" \ : "=d" ((USItype)(w0)), \ @@ -496,8 +551,9 @@ extern USItype __udiv_qrnnd (); __asm__ ("bfffo %1{%b2:%b2},%0" \ : "=d" ((USItype)(count)) \ : "od" ((USItype)(x)), "n" (0)) +#define COUNT_LEADING_ZEROS_0 32 #else /* not mc68020 */ -#define umul_ppmmxx(xh, xl, a, b) \ +#define umul_ppmm(xh, xl, a, b) \ do { USItype __umul_tmp1, __umul_tmp2; \ __asm__ ("| Inlined umul_ppmm move%.l %5,%3 @@ -557,6 +613,7 @@ extern USItype __udiv_qrnnd (); : "r" ((USItype)(x))); \ (count) = __cbtmp ^ 31; \ } while (0) +#define COUNT_LEADING_ZEROS_0 63 /* sic */ #if defined (__m88110__) #define umul_ppmm(wh, wl, u, v) \ do { \ @@ -738,6 +795,7 @@ extern USItype __udiv_qrnnd (); __asm__ ("{cntlz|cntlzw} %0,%1" \ : "=r" ((USItype)(count)) \ : "r" ((USItype)(x))) +#define COUNT_LEADING_ZEROS_0 32 #if defined (_ARCH_PPC) #define umul_ppmm(ph, pl, m0, m1) \ do { \ @@ -887,7 +945,7 @@ extern USItype __udiv_qrnnd (); (count) += 16; \ } \ } while (0) -#endif +#endif /* RT/ROMP */ #if defined (__sh2__) && W_TYPE_SIZE == 32 #define umul_ppmm(w1, w0, u, v) \ @@ -1154,20 +1212,6 @@ extern USItype __udiv_qrnnd (); (xh) += ((((signed int) __m0 >> 15) & __m1) \ + (((signed int) __m1 >> 15) & __m0)); \ } while (0) -#define umul_ppmm_off(xh, xl, m0, m1) \ - do { \ - union {long int __ll; \ - struct {unsigned int __h, __l;} __i; \ - } __xx; \ - __asm__ ("mult %S0,%H3" \ - : "=r" (__xx.__i.__h), \ - "=r" (__xx.__i.__l) \ - : "%1" (m0), \ - "rQR" (m1)); \ - (xh) = __xx.__i.__h + ((((signed int) m0 >> 15) & m1) \ - + (((signed int) m1 >> 15) & m0)); \ - (xl) = __xx.__i.__l; \ - } while (0) #endif /* __z8000__ */ #endif /* __GNUC__ */ diff --git a/sysdeps/generic/mul_n.c b/sysdeps/generic/mul_n.c index 7900988143..e37c5d8290 100644 --- a/sysdeps/generic/mul_n.c +++ b/sysdeps/generic/mul_n.c @@ -216,15 +216,7 @@ ____mpn_mul_n (prodp, up, vp, size, tspace) cy += __mpn_add_n (prodp + hsize, prodp + hsize, tspace, size); if (cy) - { - if (cy > 0) - __mpn_add_1 (prodp + hsize + size, prodp + hsize + size, hsize, cy); - else - { - __mpn_sub_1 (prodp + hsize + size, prodp + hsize + size, hsize, cy); - abort (); - } - } + __mpn_add_1 (prodp + hsize + size, prodp + hsize + size, hsize, cy); MPN_COPY (prodp, tspace, hsize); cy = __mpn_add_n (prodp + hsize, prodp + hsize, tspace + hsize, hsize); @@ -362,15 +354,7 @@ ____mpn_sqr_n (prodp, up, size, tspace) cy += __mpn_add_n (prodp + hsize, prodp + hsize, tspace, size); if (cy) - { - if (cy > 0) - __mpn_add_1 (prodp + hsize + size, prodp + hsize + size, hsize, cy); - else - { - __mpn_sub_1 (prodp + hsize + size, prodp + hsize + size, hsize, cy); - abort (); - } - } + __mpn_add_1 (prodp + hsize + size, prodp + hsize + size, hsize, cy); MPN_COPY (prodp, tspace, hsize); cy = __mpn_add_n (prodp + hsize, prodp + hsize, tspace + hsize, hsize); diff --git a/sysdeps/i386/i586/lshift.S b/sysdeps/i386/i586/lshift.S index b9f8131297..c41f74e17d 100644 --- a/sysdeps/i386/i586/lshift.S +++ b/sysdeps/i386/i586/lshift.S @@ -43,12 +43,15 @@ C_SYMBOL_NAME(__mpn_lshift:) movl 28(%esp),%ebp /* size */ movl 32(%esp),%ecx /* cnt */ +/* We can use faster code for shift-by-1 under certain conditions. */ cmp $1,%ecx jne Lnormal - movl %edi,%eax - subl %esi,%eax - cmpl %ebp,%eax - jnc Lspecial + leal 4(%esi),%eax + cmpl %edi,%eax + jnc Lspecial /* jump if s_ptr + 1 >= res_ptr */ + leal (%esi,%ebp,4),%eax + cmpl %eax,%edi + jnc Lspecial /* jump if res_ptr >= s_ptr + size */ Lnormal: leal -4(%edi,%ebp,4),%edi diff --git a/sysdeps/i386/i586/rshift.S b/sysdeps/i386/i586/rshift.S index 51cde8f07f..a820a79bc7 100644 --- a/sysdeps/i386/i586/rshift.S +++ b/sysdeps/i386/i586/rshift.S @@ -43,12 +43,15 @@ C_SYMBOL_NAME(__mpn_rshift:) movl 28(%esp),%ebp /* size */ movl 32(%esp),%ecx /* cnt */ +/* We can use faster code for shift-by-1 under certain conditions. */ cmp $1,%ecx jne Lnormal - movl %edi,%eax - subl %esi,%eax - cmpl %ebp,%eax - jnc Lspecial + leal 4(%edi),%eax + cmpl %esi,%eax + jnc Lspecial /* jump if res_ptr + 1 >= s_ptr */ + leal (%edi,%ebp,4),%eax + cmpl %eax,%esi + jnc Lspecial /* jump if s_ptr >= res_ptr + size */ Lnormal: movl (%esi),%edx diff --git a/sysdeps/m68k/add_n.S b/sysdeps/m68k/add_n.S new file mode 100644 index 0000000000..ea7a4458ea --- /dev/null +++ b/sysdeps/m68k/add_n.S @@ -0,0 +1,76 @@ +/* mc68020 __mpn_add_n -- Add two limb vectors of the same length > 0 and store + sum in a third limb vector. + +Copyright (C) 1992, 1994 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 Library General Public License as published by +the Free Software Foundation; either version 2 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 Library General Public +License for more details. + +You should have received a copy of the GNU Library General Public License +along with the GNU MP Library; see the file COPYING.LIB. If not, write to +the Free Software Foundation, Inc., 675 Mass Ave, Cambridge, MA 02139, USA. */ + +/* + INPUT PARAMETERS + res_ptr (sp + 4) + s1_ptr (sp + 8) + s2_ptr (sp + 16) + size (sp + 12) +*/ + +#include "asm-syntax.h" + + TEXT + ALIGN + GLOBL ___mpn_add_n + +LAB(___mpn_add_n) +/* Save used registers on the stack. */ + INSN2(move,l ,MEM_PREDEC(sp),d2) + INSN2(move,l ,MEM_PREDEC(sp),a2) + +/* Copy the arguments to registers. Better use movem? */ + INSN2(move,l ,a2,MEM_DISP(sp,12)) + INSN2(move,l ,a0,MEM_DISP(sp,16)) + INSN2(move,l ,a1,MEM_DISP(sp,20)) + INSN2(move,l ,d2,MEM_DISP(sp,24)) + + INSN2(eor,w ,d2,#1) + INSN2(lsr,l ,d2,#1) + bcc L1 + INSN2(subq,l ,d2,#1) /* clears cy as side effect */ + +LAB(Loop) + INSN2(move,l ,d0,MEM_POSTINC(a0)) + INSN2(move,l ,d1,MEM_POSTINC(a1)) + INSN2(addx,l ,d0,d1) + INSN2(move,l ,MEM_POSTINC(a2),d0) +LAB(L1) INSN2(move,l ,d0,MEM_POSTINC(a0)) + INSN2(move,l ,d1,MEM_POSTINC(a1)) + INSN2(addx,l ,d0,d1) + INSN2(move,l ,MEM_POSTINC(a2),d0) + + dbf d2,Loop /* loop until 16 lsb of %4 == -1 */ + INSN2(subx,l ,d0,d0) /* d0 <= -cy; save cy as 0 or -1 in d0 */ + INSN2(sub,l ,d2,#0x10000) + bcs L2 + INSN2(add,l ,d0,d0) /* restore cy */ + bra Loop + +LAB(L2) + INSN1(neg,l ,d0) + +/* Restore used registers from stack frame. */ + INSN2(move,l ,a2,MEM_POSTINC(sp)) + INSN2(move,l ,d2,MEM_POSTINC(sp)) + + rts diff --git a/sysdeps/m68k/sub_n.S b/sysdeps/m68k/sub_n.S new file mode 100644 index 0000000000..19f0ec1568 --- /dev/null +++ b/sysdeps/m68k/sub_n.S @@ -0,0 +1,76 @@ +/* mc68020 __mpn_sub_n -- Subtract two limb vectors of the same length > 0 and + store difference in a third limb vector. + +Copyright (C) 1992, 1994 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 Library General Public License as published by +the Free Software Foundation; either version 2 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 Library General Public +License for more details. + +You should have received a copy of the GNU Library General Public License +along with the GNU MP Library; see the file COPYING.LIB. If not, write to +the Free Software Foundation, Inc., 675 Mass Ave, Cambridge, MA 02139, USA. */ + +/* + INPUT PARAMETERS + res_ptr (sp + 4) + s1_ptr (sp + 8) + s2_ptr (sp + 16) + size (sp + 12) +*/ + +#include "asm-syntax.h" + + TEXT + ALIGN + GLOBL ___mpn_sub_n + +LAB(___mpn_sub_n) +/* Save used registers on the stack. */ + INSN2(move,l ,MEM_PREDEC(sp),d2) + INSN2(move,l ,MEM_PREDEC(sp),a2) + +/* Copy the arguments to registers. Better use movem? */ + INSN2(move,l ,a2,MEM_DISP(sp,12)) + INSN2(move,l ,a0,MEM_DISP(sp,16)) + INSN2(move,l ,a1,MEM_DISP(sp,20)) + INSN2(move,l ,d2,MEM_DISP(sp,24)) + + INSN2(eor,w ,d2,#1) + INSN2(lsr,l ,d2,#1) + bcc L1 + INSN2(subq,l ,d2,#1) /* clears cy as side effect */ + +LAB(Loop) + INSN2(move,l ,d0,MEM_POSTINC(a0)) + INSN2(move,l ,d1,MEM_POSTINC(a1)) + INSN2(subx,l ,d0,d1) + INSN2(move,l ,MEM_POSTINC(a2),d0) +LAB(L1) INSN2(move,l ,d0,MEM_POSTINC(a0)) + INSN2(move,l ,d1,MEM_POSTINC(a1)) + INSN2(subx,l ,d0,d1) + INSN2(move,l ,MEM_POSTINC(a2),d0) + + dbf d2,Loop /* loop until 16 lsb of %4 == -1 */ + INSN2(subx,l ,d0,d0) /* d0 <= -cy; save cy as 0 or -1 in d0 */ + INSN2(sub,l ,d2,#0x10000) + bcs L2 + INSN2(add,l ,d0,d0) /* restore cy */ + bra Loop + +LAB(L2) + INSN1(neg,l ,d0) + +/* Restore used registers from stack frame. */ + INSN2(move,l ,a2,MEM_POSTINC(sp)) + INSN2(move,l ,d2,MEM_POSTINC(sp)) + + rts diff --git a/sysdeps/m88k/add_n.s b/sysdeps/m88k/add_n.s new file mode 100644 index 0000000000..7e4ccccb90 --- /dev/null +++ b/sysdeps/m88k/add_n.s @@ -0,0 +1,103 @@ +; mc88100 __mpn_add -- Add two limb vectors of the same length > 0 and store +; sum in a third limb vector. + +; Copyright (C) 1992, 1994 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 Library General Public License as published by +; the Free Software Foundation; either version 2 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 Library General Public +; License for more details. + +; You should have received a copy of the GNU Library General Public License +; along with the GNU MP Library; see the file COPYING.LIB. If not, write to +; the Free Software Foundation, Inc., 675 Mass Ave, Cambridge, MA 02139, USA. + + +; INPUT PARAMETERS +; res_ptr r2 +; s1_ptr r3 +; s2_ptr r4 +; size r5 + +; This code has been optimized to run one instruction per clock, avoiding +; load stalls and writeback contention. As a result, the instruction +; order is not always natural. + +; The speed is about 4.6 clocks/limb + 18 clocks/limb-vector on an 88100, +; but on the 88110, it seems to run much slower, 6.6 clocks/limb. + + text + align 16 + global ___mpn_add_n +___mpn_add_n: + ld r6,r3,0 ; read first limb from s1_ptr + extu r10,r5,3 + ld r7,r4,0 ; read first limb from s2_ptr + + subu.co r5,r0,r5 ; (clear carry as side effect) + mak r5,r5,3<4> + bcnd eq0,r5,Lzero + + or r12,r0,lo16(Lbase) + or.u r12,r12,hi16(Lbase) + addu r12,r12,r5 ; r12 is address for entering in loop + + extu r5,r5,2 ; divide by 4 + subu r2,r2,r5 ; adjust res_ptr + subu r3,r3,r5 ; adjust s1_ptr + subu r4,r4,r5 ; adjust s2_ptr + + or r8,r6,r0 + + jmp.n r12 + or r9,r7,r0 + +Loop: addu r3,r3,32 + st r8,r2,28 + addu r4,r4,32 + ld r6,r3,0 + addu r2,r2,32 + ld r7,r4,0 +Lzero: subu r10,r10,1 ; add 0 + 8r limbs (adj loop cnt) +Lbase: ld r8,r3,4 + addu.cio r6,r6,r7 + ld r9,r4,4 + st r6,r2,0 + ld r6,r3,8 ; add 7 + 8r limbs + addu.cio r8,r8,r9 + ld r7,r4,8 + st r8,r2,4 + ld r8,r3,12 ; add 6 + 8r limbs + addu.cio r6,r6,r7 + ld r9,r4,12 + st r6,r2,8 + ld r6,r3,16 ; add 5 + 8r limbs + addu.cio r8,r8,r9 + ld r7,r4,16 + st r8,r2,12 + ld r8,r3,20 ; add 4 + 8r limbs + addu.cio r6,r6,r7 + ld r9,r4,20 + st r6,r2,16 + ld r6,r3,24 ; add 3 + 8r limbs + addu.cio r8,r8,r9 + ld r7,r4,24 + st r8,r2,20 + ld r8,r3,28 ; add 2 + 8r limbs + addu.cio r6,r6,r7 + ld r9,r4,28 + st r6,r2,24 + bcnd.n ne0,r10,Loop ; add 1 + 8r limbs + addu.cio r8,r8,r9 + + st r8,r2,28 ; store most significant limb + + jmp.n r1 + addu.ci r2,r0,r0 ; return carry-out from most sign. limb diff --git a/sysdeps/m88k/mul_1.s b/sysdeps/m88k/mul_1.s new file mode 100644 index 0000000000..35c238d570 --- /dev/null +++ b/sysdeps/m88k/mul_1.s @@ -0,0 +1,128 @@ +; mc88100 __mpn_mul_1 -- Multiply a limb vector with a single limb and +; store the prod |