diff options
Diffstat (limited to 'rts/gmp/mpn/x86/pentium/diveby3.asm')
| -rw-r--r-- | rts/gmp/mpn/x86/pentium/diveby3.asm | 183 |
1 files changed, 183 insertions, 0 deletions
diff --git a/rts/gmp/mpn/x86/pentium/diveby3.asm b/rts/gmp/mpn/x86/pentium/diveby3.asm new file mode 100644 index 0000000000..dbac81642f --- /dev/null +++ b/rts/gmp/mpn/x86/pentium/diveby3.asm @@ -0,0 +1,183 @@ +dnl Intel P5 mpn_divexact_by3 -- mpn division by 3, expecting no remainder. +dnl +dnl P5: 15.0 cycles/limb + + +dnl Copyright (C) 2000 Free Software Foundation, Inc. +dnl +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 +dnl modify it under the terms of the GNU Lesser General Public License as +dnl published by the Free Software Foundation; either version 2.1 of the +dnl License, or (at your option) any later version. +dnl +dnl The GNU MP Library is distributed in the hope that it will be useful, +dnl but WITHOUT ANY WARRANTY; without even the implied warranty of +dnl MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU +dnl Lesser General Public License for more details. +dnl +dnl You should have received a copy of the GNU Lesser General Public +dnl License along with the GNU MP Library; see the file COPYING.LIB. If +dnl not, write to the Free Software Foundation, Inc., 59 Temple Place - +dnl Suite 330, Boston, MA 02111-1307, USA. + + +include(`../config.m4') + + +C mp_limb_t mpn_divexact_by3c (mp_ptr dst, mp_srcptr src, mp_size_t size, +C mp_limb_t carry); + +defframe(PARAM_CARRY,16) +defframe(PARAM_SIZE, 12) +defframe(PARAM_SRC, 8) +defframe(PARAM_DST, 4) + +dnl multiplicative inverse of 3, modulo 2^32 +deflit(INVERSE_3, 0xAAAAAAAB) + +dnl ceil(b/3), ceil(b*2/3) and floor(b*2/3) where b=2^32 +deflit(ONE_THIRD_CEIL, 0x55555556) +deflit(TWO_THIRDS_CEIL, 0xAAAAAAAB) +deflit(TWO_THIRDS_FLOOR, 0xAAAAAAAA) + + .text + ALIGN(8) + +PROLOGUE(mpn_divexact_by3c) +deflit(`FRAME',0) + + movl PARAM_SRC, %ecx + movl PARAM_SIZE, %edx + + decl %edx + jnz L(two_or_more) + + movl (%ecx), %edx + movl PARAM_CARRY, %eax C risk of cache bank clash here + + movl PARAM_DST, %ecx + subl %eax, %edx + + sbbl %eax, %eax C 0 or -1 + + imull $INVERSE_3, %edx, %edx + + negl %eax C 0 or 1 + cmpl $ONE_THIRD_CEIL, %edx + + sbbl $-1, %eax C +1 if edx>=ceil(b/3) + cmpl $TWO_THIRDS_CEIL, %edx + + sbbl $-1, %eax C +1 if edx>=ceil(b*2/3) + movl %edx, (%ecx) + + ret + + +L(two_or_more): + C eax + C ebx + C ecx src + C edx size-1 + C esi + C edi + C ebp + + pushl %ebx FRAME_pushl() + pushl %esi FRAME_pushl() + + pushl %edi FRAME_pushl() + pushl %ebp FRAME_pushl() + + movl PARAM_DST, %edi + movl PARAM_CARRY, %esi + + movl (%ecx), %eax C src low limb + xorl %ebx, %ebx + + sub %esi, %eax + movl $TWO_THIRDS_FLOOR, %esi + + leal (%ecx,%edx,4), %ecx C &src[size-1] + leal (%edi,%edx,4), %edi C &dst[size-1] + + adcl $0, %ebx C carry, 0 or 1 + negl %edx C -(size-1) + + +C The loop needs a source limb ready at the top, which leads to one limb +C handled separately at the end, and the special case above for size==1. +C There doesn't seem to be any scheduling that would keep the speed but move +C the source load and carry subtract up to the top. +C +C The destination cache line prefetching adds 1 cycle to the loop but is +C considered worthwhile. The slowdown is a factor of 1.07, but will prevent +C repeated write-throughs if the destination isn't in L1. A version using +C an outer loop to prefetch only every 8 limbs (a cache line) proved to be +C no faster, due to unavoidable branch mispreditions in the inner loop. +C +C setc is 2 cycles on P54, so an adcl is used instead. If the movl $0,%ebx +C could be avoided then the src limb fetch could pair up and save a cycle. +C This would probably mean going to a two limb loop with the carry limb +C alternately positive or negative, since an sbbl %ebx,%ebx will leave a +C value which is in the opposite sense to the preceding sbbl/adcl %ebx,%eax. +C +C A register is used for TWO_THIRDS_FLOOR because a cmp can't be done as +C "cmpl %edx, $n" with the immediate as the second operand. +C +C The "4" source displacement is in the loop rather than the setup because +C this gets L(top) aligned to 8 bytes at no cost. + + ALIGN(8) +L(top): + C eax source limb, carry subtracted + C ebx carry (0 or 1) + C ecx &src[size-1] + C edx counter, limbs, negative + C esi TWO_THIRDS_FLOOR + C edi &dst[size-1] + C ebp scratch (result limb) + + imull $INVERSE_3, %eax, %ebp + + cmpl $ONE_THIRD_CEIL, %ebp + movl (%edi,%edx,4), %eax C dst cache line prefetch + + sbbl $-1, %ebx C +1 if ebp>=ceil(b/3) + cmpl %ebp, %esi + + movl 4(%ecx,%edx,4), %eax C next src limb + + sbbl %ebx, %eax C and further -1 if ebp>=ceil(b*2/3) + movl $0, %ebx + + adcl $0, %ebx C new carry + movl %ebp, (%edi,%edx,4) + + incl %edx + jnz L(top) + + + + imull $INVERSE_3, %eax, %edx + + cmpl $ONE_THIRD_CEIL, %edx + movl %edx, (%edi) + + sbbl $-1, %ebx C +1 if edx>=ceil(b/3) + cmpl $TWO_THIRDS_CEIL, %edx + + sbbl $-1, %ebx C +1 if edx>=ceil(b*2/3) + popl %ebp + + movl %ebx, %eax + popl %edi + + popl %esi + popl %ebx + + ret + +EPILOGUE() |