1dnl  Intel P5 mpn_popcount -- mpn bit population count.
2
3dnl  Copyright 2001, 2002 Free Software Foundation, Inc.
4dnl
5dnl  This file is part of the GNU MP Library.
6dnl
7dnl  The GNU MP Library is free software; you can redistribute it and/or
8dnl  modify it under the terms of the GNU Lesser General Public License as
9dnl  published by the Free Software Foundation; either version 3 of the
10dnl  License, or (at your option) any later version.
11dnl
12dnl  The GNU MP Library is distributed in the hope that it will be useful,
13dnl  but WITHOUT ANY WARRANTY; without even the implied warranty of
14dnl  MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
15dnl  Lesser General Public License for more details.
16dnl
17dnl  You should have received a copy of the GNU Lesser General Public License
18dnl  along with the GNU MP Library.  If not, see http://www.gnu.org/licenses/.
19
20include(`../config.m4')
21
22
23C P5: 8.0 cycles/limb
24
25
26C unsigned long mpn_popcount (mp_srcptr src, mp_size_t size);
27C
28C An arithmetic approach has been found to be slower than the table lookup,
29C due to needing too many instructions.
30
31C The slightly strange quoting here helps the renaming done by tune/many.pl.
32deflit(TABLE_NAME,
33m4_assert_defined(`GSYM_PREFIX')
34GSYM_PREFIX`'mpn_popcount``'_table')
35
36	RODATA
37	ALIGN(8)
38	GLOBL	TABLE_NAME
39TABLE_NAME:
40forloop(i,0,255,
41`	.byte	m4_popcount(i)
42')
43
44defframe(PARAM_SIZE,8)
45defframe(PARAM_SRC, 4)
46
47	TEXT
48	ALIGN(8)
49
50PROLOGUE(mpn_popcount)
51deflit(`FRAME',0)
52
53	movl	PARAM_SIZE, %ecx
54	pushl	%esi	FRAME_pushl()
55
56ifdef(`PIC',`
57	pushl	%ebx	FRAME_pushl()
58	pushl	%ebp	FRAME_pushl()
59
60	call	L(here)
61L(here):
62	popl	%ebp
63	shll	%ecx		C size in byte pairs
64
65	addl	$_GLOBAL_OFFSET_TABLE_+[.-L(here)], %ebp
66	movl	PARAM_SRC, %esi
67
68	xorl	%eax, %eax	C total
69	xorl	%ebx, %ebx	C byte
70
71	movl	TABLE_NAME@GOT(%ebp), %ebp
72	xorl	%edx, %edx	C byte
73define(TABLE,`(%ebp,$1)')
74',`
75dnl non-PIC
76	shll	%ecx		C size in byte pairs
77	movl	PARAM_SRC, %esi
78
79	pushl	%ebx	FRAME_pushl()
80	xorl	%eax, %eax	C total
81
82	xorl	%ebx, %ebx	C byte
83	xorl	%edx, %edx	C byte
84
85define(TABLE,`TABLE_NAME`'($1)')
86')
87
88
89	ALIGN(8)	C necessary on P55 for claimed speed
90L(top):
91	C eax	total
92	C ebx	byte
93	C ecx	counter, 2*size to 2
94	C edx	byte
95	C esi	src
96	C edi
97	C ebp	[PIC] table
98
99	addl	%ebx, %eax
100	movb	-1(%esi,%ecx,2), %bl
101
102	addl	%edx, %eax
103	movb	-2(%esi,%ecx,2), %dl
104
105	movb	TABLE(%ebx), %bl
106	decl	%ecx
107
108	movb	TABLE(%edx), %dl
109	jnz	L(top)
110
111
112ifdef(`PIC',`
113	popl	%ebp
114')
115	addl	%ebx, %eax
116	popl	%ebx
117
118	addl	%edx, %eax
119	popl	%esi
120
121	ret
122
123EPILOGUE()
124