# Copyright (c) 1985 Regents of the University of California. # All rights reserved. # # %sccs.include.redist.sh% # # @(#)tan.s 5.4 (Berkeley) 10/09/90 # .data .align 2 _sccsid: .asciz "@(#)tan.s 1.1 (Berkeley) 8/21/85; 5.4 (ucb.elefunt) 10/09/90" # This is the implementation of Peter Tang's double precision # tangent for the VAX using Bob Corbett's argument reduction. # # Notes: # under 1,024,000 random arguments testing on [0,2*pi] # tan() observed maximum error = 2.15 ulps # # double tan(arg) # double arg; # method: true range reduction to [-pi/4,pi/4], P. Tang & B. Corbett # S. McDonald, April 4, 1985 # .globl _tan .text .align 1 _tan: .word 0xffc # save r2-r11 movq 4(ap),r0 bicw3 $0x807f,r0,r2 beql 1f # if x is zero or reserved operand then return x # # Save the PSL's IV & FU bits on the stack. # movpsl r2 bicw3 $0xff9f,r2,-(sp) # # Clear the IV & FU bits. # bicpsw $0x0060 jsb libm$argred # # At this point, # r0 contains the quadrant number, 0, 1, 2, or 3; # r2/r1 contains the reduced argument as a D-format number; # r3 contains a F-format extension to the reduced argument; # # Save r3/r0 so that we can call cosine after calling sine. # movq r2,-(sp) movq r0,-(sp) # # Call sine. r4 = 0 implies sine. # movl $0,r4 jsb libm$sincos # # Save sin(x) in r11/r10 . # movd r0,r10 # # Call cosine. r4 = 1 implies cosine. # movq (sp)+,r0 movq (sp)+,r2 movl $1,r4 jsb libm$sincos divd3 r0,r10,r0 bispsw (sp)+ 1: ret