xref: /csrg-svn/lib/libm/vax/cabs.s (revision 34927)
1# Copyright (c) 1985 Regents of the University of California.
2# All rights reserved.
3#
4# Redistribution and use in source and binary forms are permitted
5# provided that the above copyright notice and this paragraph are
6# duplicated in all such forms and that any documentation,
7# advertising materials, and other materials related to such
8# distribution and use acknowledge that the software was developed
9# by the University of California, Berkeley.  The name of the
10# University may not be used to endorse or promote products derived
11# from this software without specific prior written permission.
12# THIS SOFTWARE IS PROVIDED ``AS IS'' AND WITHOUT ANY EXPRESS OR
13# IMPLIED WARRANTIES, INCLUDING, WITHOUT LIMITATION, THE IMPLIED
14# WARRANTIES OF MERCHANTIBILITY AND FITNESS FOR A PARTICULAR PURPOSE.
15#
16# All recipients should regard themselves as participants in an ongoing
17# research project and hence should feel obligated to report their
18# experiences (good or bad) with these elementary function codes, using
19# the sendbug(8) program, to the authors.
20#
21#	@(#)cabs.s	5.3 (Berkeley) 06/30/88
22#
23	.data
24	.align	2
25_sccsid:
26.asciz	"@(#)cabs.s	1.2 (Berkeley) 8/21/85; 5.3 (ucb.elefunt) 06/30/88"
27
28# double precision complex absolute value
29# CABS by W. Kahan, 9/7/80.
30# Revised for reserved operands by E. LeBlanc, 8/18/82
31# argument for complex absolute value by reference, *4(ap)
32# argument for cabs and hypot (C fcns) by value, 4(ap)
33# output is in r0:r1 (error less than 0.86 ulps)
34
35	.text
36	.align	1
37	.globl  _cabs
38	.globl  _hypot
39	.globl	_z_abs
40	.globl	libm$cdabs_r6
41	.globl	libm$dsqrt_r5
42
43#	entry for c functions cabs and hypot
44_cabs:
45_hypot:
46	.word	0x807c		# save r2-r6, enable floating overflow
47	movq	4(ap),r0	# r0:1 = x
48	movq	12(ap),r2	# r2:3 = y
49	jmp	cabs2
50#	entry for Fortran use, call by:   d = abs(z)
51_z_abs:
52	.word	0x807c		# save r2-r6, enable floating overflow
53	movl	4(ap),r2	# indirect addressing is necessary here
54	movq	(r2)+,r0	# r0:1 = x
55	movq	(r2),r2		# r2:3 = y
56
57cabs2:
58	bicw3	$0x7f,r0,r4	# r4 has signed biased exp of x
59	cmpw	$0x8000,r4
60	jeql	return		# x is a reserved operand, so return it
61	bicw3	$0x7f,r2,r5	# r5 has signed biased exp of y
62	cmpw	$0x8000,r5
63	jneq	cont		# y isn't a reserved operand
64	movq	r2,r0		# return y if it's reserved
65	ret
66
67cont:
68	bsbb	regs_set	# r0:1 = dsqrt(x^2+y^2)/2^r6
69	addw2	r6,r0		# unscaled cdabs in r0:1
70	jvc	return		# unless it overflows
71	subw2	$0x80,r0	# halve r0 to get meaningful overflow
72	addd2	r0,r0		# overflow; r0 is half of true abs value
73return:
74	ret
75
76libm$cdabs_r6:			# ENTRY POINT for cdsqrt
77				# calculates a scaled (factor in r6)
78				# complex absolute value
79
80	movq	(r4)+,r0	# r0:r1 = x via indirect addressing
81	movq	(r4),r2		# r2:r3 = y via indirect addressing
82
83	bicw3	$0x7f,r0,r5	# r5 has signed biased exp of x
84	cmpw	$0x8000,r5
85	jeql	cdreserved	# x is a reserved operand
86	bicw3	$0x7f,r2,r5	# r5 has signed biased exp of y
87	cmpw	$0x8000,r5
88	jneq	regs_set	# y isn't a reserved operand either?
89
90cdreserved:
91	movl	*4(ap),r4	# r4 -> (u,v), if x or y is reserved
92	movq	r0,(r4)+	# copy u and v as is and return
93	movq	r2,(r4)		# (again addressing is indirect)
94	ret
95
96regs_set:
97	bicw2	$0x8000,r0	# r0:r1 = dabs(x)
98	bicw2	$0x8000,r2	# r2:r3 = dabs(y)
99	cmpw	r0,r2
100	jgeq	ordered
101	movq	r0,r4
102	movq	r2,r0
103	movq	r4,r2		# force y's exp <= x's exp
104ordered:
105	bicw3	$0x7f,r0,r6	# r6 = exponent(x) + bias(129)
106	jeql	retsb		# if x = y = 0 then cdabs(x,y) = 0
107	subw2	$0x4780,r6	# r6 = exponent(x) - 14
108	subw2	r6,r0		# 2^14 <= scaled x < 2^15
109	bitw	$0xff80,r2
110	jeql	retsb		# if y = 0 return dabs(x)
111	subw2	r6,r2
112	cmpw	$0x3780,r2	# if scaled y < 2^-18
113	jgtr	retsb		#   return dabs(x)
114	emodd	r0,$0,r0,r4,r0	# r4 + r0:1 = scaled x^2
115	emodd	r2,$0,r2,r5,r2	# r5 + r2:3 = scaled y^2
116	addd2	r2,r0
117	addl2	r5,r4
118	cvtld	r4,r2
119	addd2	r2,r0		# r0:1 = scaled x^2 + y^2
120	jmp	libm$dsqrt_r5	# r0:1 = dsqrt(x^2+y^2)/2^r6
121retsb:
122	rsb			# error < 0.86 ulp
123