Return to lshift.asm CVS log

Up to [local] / OpenXM_contrib / gmp / mpn / x86 / k6 / mmx

Annotation of OpenXM_contrib/gmp/mpn/x86/k6/mmx/lshift.asm, Revision 1.1.1.1

1.1       maekawa     1: dnl  AMD K6 mpn_lshift -- mpn left shift.
                      2: dnl
                      3: dnl  K6: 3.0 cycles/limb
                      4:
                      5:
                      6: dnl  Copyright (C) 1999, 2000 Free Software Foundation, Inc.
                      7: dnl
                      8: dnl  This file is part of the GNU MP Library.
                      9: dnl
                     10: dnl  The GNU MP Library is free software; you can redistribute it and/or
                     11: dnl  modify it under the terms of the GNU Lesser General Public License as
                     12: dnl  published by the Free Software Foundation; either version 2.1 of the
                     13: dnl  License, or (at your option) any later version.
                     14: dnl
                     15: dnl  The GNU MP Library is distributed in the hope that it will be useful,
                     16: dnl  but WITHOUT ANY WARRANTY; without even the implied warranty of
                     17: dnl  MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
                     18: dnl  Lesser General Public License for more details.
                     19: dnl
                     20: dnl  You should have received a copy of the GNU Lesser General Public
                     21: dnl  License along with the GNU MP Library; see the file COPYING.LIB.  If
                     22: dnl  not, write to the Free Software Foundation, Inc., 59 Temple Place -
                     23: dnl  Suite 330, Boston, MA 02111-1307, USA.
                     24:
                     25:
                     26: include(`../config.m4')
                     27:
                     28:
                     29: C mp_limb_t mpn_lshift (mp_ptr dst, mp_srcptr src, mp_size_t size,
                     30: C                       unsigned shift);
                     31: C
                     32: C The loop runs at 3 cycles/limb, limited by decoding and by having 3 mmx
                     33: C instructions.  This is despite every second fetch being unaligned.
                     34:
                     35:
                     36: defframe(PARAM_SHIFT,16)
                     37: defframe(PARAM_SIZE, 12)
                     38: defframe(PARAM_SRC,  8)
                     39: defframe(PARAM_DST,  4)
                     40:
                     41:        .text
                     42:        ALIGN(32)
                     43:
                     44: PROLOGUE(mpn_lshift)
                     45: deflit(`FRAME',0)
                     46:
                     47:        C The 1 limb case can be done without the push %ebx, but it's then
                     48:        C still the same speed.  The push is left as a free helping hand for
                     49:        C the two_or_more code.
                     50:
                     51:        movl    PARAM_SIZE, %eax
                     52:        pushl   %ebx                    FRAME_pushl()
                     53:
                     54:        movl    PARAM_SRC, %ebx
                     55:        decl    %eax
                     56:
                     57:        movl    PARAM_SHIFT, %ecx
                     58:        jnz     L(two_or_more)
                     59:
                     60:        movl    (%ebx), %edx            C src limb
                     61:        movl    PARAM_DST, %ebx
                     62:
                     63:        shldl(  %cl, %edx, %eax)        C return value
                     64:
                     65:        shll    %cl, %edx
                     66:
                     67:        movl    %edx, (%ebx)            C dst limb
                     68:        popl    %ebx
                     69:
                     70:        ret
                     71:
                     72:
                     73:        ALIGN(16)       C avoid offset 0x1f
                     74:        nop             C avoid bad cache line crossing
                     75: L(two_or_more):
                     76:        C eax   size-1
                     77:        C ebx   src
                     78:        C ecx   shift
                     79:        C edx
                     80:
                     81:        movl    (%ebx,%eax,4), %edx     C src high limb
                     82:        negl    %ecx
                     83:
                     84:        movd    PARAM_SHIFT, %mm6
                     85:        addl    $32, %ecx               C 32-shift
                     86:
                     87:        shrl    %cl, %edx
                     88:
                     89:        movd    %ecx, %mm7
                     90:        movl    PARAM_DST, %ecx
                     91:
                     92: L(top):
                     93:        C eax   counter, size-1 to 1
                     94:        C ebx   src
                     95:        C ecx   dst
                     96:        C edx   retval
                     97:        C
                     98:        C mm0   scratch
                     99:        C mm6   shift
                    100:        C mm7   32-shift
                    101:
                    102:        movq    -4(%ebx,%eax,4), %mm0
                    103:        decl    %eax
                    104:
                    105:        psrlq   %mm7, %mm0
                    106:
                    107:        movd    %mm0, 4(%ecx,%eax,4)
                    108:        jnz     L(top)
                    109:
                    110:
                    111:        movd    (%ebx), %mm0
                    112:        popl    %ebx
                    113:
                    114:        psllq   %mm6, %mm0
                    115:        movl    %edx, %eax
                    116:
                    117:        movd    %mm0, (%ecx)
                    118:
                    119:        emms
                    120:        ret
                    121:
                    122: EPILOGUE()