dnl PowerPC-32/VMX and PowerPC-64/VMX mpn_popcount.
dnl Copyright 2006, 2010 Free Software Foundation, Inc.
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 modify
dnl it under the terms of either:
dnl
dnl * the GNU Lesser General Public License as published by the Free
dnl Software Foundation; either version 3 of the License, or (at your
dnl option) any later version.
dnl
dnl or
dnl
dnl * the GNU General Public License as published by the Free Software
dnl Foundation; either version 2 of the License, or (at your option) any
dnl later version.
dnl
dnl or both in parallel, as here.
dnl
dnl The GNU MP Library is distributed in the hope that it will be useful, but
dnl WITHOUT ANY WARRANTY; without even the implied warranty of MERCHANTABILITY
dnl or FITNESS FOR A PARTICULAR PURPOSE. See the GNU General Public License
dnl for more details.
dnl
dnl You should have received copies of the GNU General Public License and the
dnl GNU Lesser General Public License along with the GNU MP Library. If not,
dnl see https://www.gnu.org/licenses/.
include(`../config.m4')
C cycles/limb
C 7400,7410 (G4): ?
C 744x,745x (G4+): 1.125
C 970 (G5): 2.25
C TODO
C * Rewrite the awkward huge n outer loop code.
C * Two lvx, two vperm, and two vxor could make us a similar hamdist.
C * Compress cnsts table in 64-bit mode, only half the values are needed.
C The inner loop handles up to 2^34 bits, i.e., 2^31 64-limbs, due to overflow
C in vsum4ubs. For large operands, we work in chunks, of size LIMBS_PER_CHUNK.
define(`LIMBS_PER_CHUNK', 0x1000)
define(`LIMBS_CHUNK_THRES', 0x1001)
ASM_START()
PROLOGUE(mpn_popcount,toc)
mfspr r10, 256
oris r0, r10, 0xfffc C Set VRSAVE bit 0-13
mtspr 256, r0
ifdef(`HAVE_ABI_mode32',
` rldicl n, n, 0, 32') C zero extend n
C Load various constants into vector registers
LEAL( r11, cnsts)
li r12, 16
vspltisb cnt4, 4 C 0x0404...04 used as shift count
li r7, 160
lvx rtab, 0, r11
LIMB64(`lis r0, LIMBS_CHUNK_THRES ')
LIMB64(`cmpd cr7, n, r0 ')
LIMB32(`rlwinm r8, ap, 30,30,31 ')
LIMB64(`rlwinm r8, ap, 29,31,31 ') add n, n, r8 C compensate n for rounded down `ap'
vxor v1, v1, v1
li r8, 0 C grand total count
vxor v12, v12, v12 C zero total count
vxor v13, v13, v13 C zero total count
addic. n, n, -LIMBS_PER_VR
ble L(sum)
addic. n, n, -LIMBS_PER_VR
ble L(lsum)
C For 64-bit machines, handle huge n that would overflow vsum4ubs
LIMB64(`ble cr7, L(small) ')
LIMB64(`addis r9, n, -LIMBS_PER_CHUNK ') C remaining n
LIMB64(`lis n, LIMBS_PER_CHUNK ')
ALIGN(16)
L(small):
LIMB32(`srwi r7, n, 3 ') C loop count corresponding to n
LIMB64(`srdi r7, n, 2 ') C loop count corresponding to n
addi r7, r7, 1
mtctr r7 C copy n to count register
b L(ent)
C Handle outer loop for huge n. We inherit cr7 and r0 from above.
LIMB64(`ble cr7, L(ret)
vxor v12, v12, v12 C zero total count
vxor v13, v13, v13 C zero total count
mr n, r9
cmpd cr7, n, r0
ble cr7, L(2)
addis r9, n, -LIMBS_PER_CHUNK C remaining n
lis n, LIMBS_PER_CHUNK
L(2): srdi r7, n, 2 C loop count corresponding to n
mtctr r7 C copy n to count register
b L(top) ')
DEF_OBJECT(cnsts,16)
C Counts for vperm
.byte 0x00,0x01,0x01,0x02,0x01,0x02,0x02,0x03
.byte 0x01,0x02,0x02,0x03,0x02,0x03,0x03,0x04
C Masks for high end of number
.byte 0xff,0xff,0xff,0xff,0xff,0xff,0xff,0xff
.byte 0xff,0xff,0xff,0xff,0xff,0xff,0xff,0xff
.byte 0xff,0xff,0xff,0xff,0xff,0xff,0xff,0xff
.byte 0xff,0xff,0xff,0xff,0x00,0x00,0x00,0x00
C Masks for low end of number
.byte 0xff,0xff,0xff,0xff,0xff,0xff,0xff,0xff
.byte 0xff,0xff,0xff,0xff,0xff,0xff,0xff,0xff
Die Informationen auf dieser Webseite wurden
nach bestem Wissen sorgfältig zusammengestellt. Es wird jedoch weder Vollständigkeit, noch Richtigkeit,
noch Qualität der bereit gestellten Informationen zugesichert.
Bemerkung:
Die farbliche Syntaxdarstellung und die Messung sind noch experimentell.