cachepc-linux

Fork of AMDESE/linux with modifications for CachePC side-channel attack
git clone https://git.sinitax.com/sinitax/cachepc-linux
Log | Files | Refs | README | LICENSE | sfeed.txt

stanh.S (4074B)


      1|
      2|	stanh.sa 3.1 12/10/90
      3|
      4|	The entry point sTanh computes the hyperbolic tangent of
      5|	an input argument; sTanhd does the same except for denormalized
      6|	input.
      7|
      8|	Input: Double-extended number X in location pointed to
      9|		by address register a0.
     10|
     11|	Output: The value tanh(X) returned in floating-point register Fp0.
     12|
     13|	Accuracy and Monotonicity: The returned result is within 3 ulps in
     14|		64 significant bit, i.e. within 0.5001 ulp to 53 bits if the
     15|		result is subsequently rounded to double precision. The
     16|		result is provably monotonic in double precision.
     17|
     18|	Speed: The program stanh takes approximately 270 cycles.
     19|
     20|	Algorithm:
     21|
     22|	TANH
     23|	1. If |X| >= (5/2) log2 or |X| <= 2**(-40), go to 3.
     24|
     25|	2. (2**(-40) < |X| < (5/2) log2) Calculate tanh(X) by
     26|		sgn := sign(X), y := 2|X|, z := expm1(Y), and
     27|		tanh(X) = sgn*( z/(2+z) ).
     28|		Exit.
     29|
     30|	3. (|X| <= 2**(-40) or |X| >= (5/2) log2). If |X| < 1,
     31|		go to 7.
     32|
     33|	4. (|X| >= (5/2) log2) If |X| >= 50 log2, go to 6.
     34|
     35|	5. ((5/2) log2 <= |X| < 50 log2) Calculate tanh(X) by
     36|		sgn := sign(X), y := 2|X|, z := exp(Y),
     37|		tanh(X) = sgn - [ sgn*2/(1+z) ].
     38|		Exit.
     39|
     40|	6. (|X| >= 50 log2) Tanh(X) = +-1 (round to nearest). Thus, we
     41|		calculate Tanh(X) by
     42|		sgn := sign(X), Tiny := 2**(-126),
     43|		tanh(X) := sgn - sgn*Tiny.
     44|		Exit.
     45|
     46|	7. (|X| < 2**(-40)). Tanh(X) = X.	Exit.
     47|
     48
     49|		Copyright (C) Motorola, Inc. 1990
     50|			All Rights Reserved
     51|
     52|       For details on the license for this file, please see the
     53|       file, README, in this same directory.
     54
     55|STANH	idnt	2,1 | Motorola 040 Floating Point Software Package
     56
     57	|section	8
     58
     59#include "fpsp.h"
     60
     61	.set	X,FP_SCR5
     62	.set	XDCARE,X+2
     63	.set	XFRAC,X+4
     64
     65	.set	SGN,L_SCR3
     66
     67	.set	V,FP_SCR6
     68
     69BOUNDS1:	.long 0x3FD78000,0x3FFFDDCE | ... 2^(-40), (5/2)LOG2
     70
     71	|xref	t_frcinx
     72	|xref	t_extdnrm
     73	|xref	setox
     74	|xref	setoxm1
     75
     76	.global	stanhd
     77stanhd:
     78|--TANH(X) = X FOR DENORMALIZED X
     79
     80	bra		t_extdnrm
     81
     82	.global	stanh
     83stanh:
     84	fmovex		(%a0),%fp0	| ...LOAD INPUT
     85
     86	fmovex		%fp0,X(%a6)
     87	movel		(%a0),%d0
     88	movew		4(%a0),%d0
     89	movel		%d0,X(%a6)
     90	andl		#0x7FFFFFFF,%d0
     91	cmp2l		BOUNDS1(%pc),%d0	| ...2**(-40) < |X| < (5/2)LOG2 ?
     92	bcss		TANHBORS
     93
     94|--THIS IS THE USUAL CASE
     95|--Y = 2|X|, Z = EXPM1(Y), TANH(X) = SIGN(X) * Z / (Z+2).
     96
     97	movel		X(%a6),%d0
     98	movel		%d0,SGN(%a6)
     99	andl		#0x7FFF0000,%d0
    100	addl		#0x00010000,%d0	| ...EXPONENT OF 2|X|
    101	movel		%d0,X(%a6)
    102	andl		#0x80000000,SGN(%a6)
    103	fmovex		X(%a6),%fp0		| ...FP0 IS Y = 2|X|
    104
    105	movel		%d1,-(%a7)
    106	clrl		%d1
    107	fmovemx	%fp0-%fp0,(%a0)
    108	bsr		setoxm1		| ...FP0 IS Z = EXPM1(Y)
    109	movel		(%a7)+,%d1
    110
    111	fmovex		%fp0,%fp1
    112	fadds		#0x40000000,%fp1	| ...Z+2
    113	movel		SGN(%a6),%d0
    114	fmovex		%fp1,V(%a6)
    115	eorl		%d0,V(%a6)
    116
    117	fmovel		%d1,%FPCR		|restore users exceptions
    118	fdivx		V(%a6),%fp0
    119	bra		t_frcinx
    120
    121TANHBORS:
    122	cmpl		#0x3FFF8000,%d0
    123	blt		TANHSM
    124
    125	cmpl		#0x40048AA1,%d0
    126	bgt		TANHHUGE
    127
    128|-- (5/2) LOG2 < |X| < 50 LOG2,
    129|--TANH(X) = 1 - (2/[EXP(2X)+1]). LET Y = 2|X|, SGN = SIGN(X),
    130|--TANH(X) = SGN -	SGN*2/[EXP(Y)+1].
    131
    132	movel		X(%a6),%d0
    133	movel		%d0,SGN(%a6)
    134	andl		#0x7FFF0000,%d0
    135	addl		#0x00010000,%d0	| ...EXPO OF 2|X|
    136	movel		%d0,X(%a6)		| ...Y = 2|X|
    137	andl		#0x80000000,SGN(%a6)
    138	movel		SGN(%a6),%d0
    139	fmovex		X(%a6),%fp0		| ...Y = 2|X|
    140
    141	movel		%d1,-(%a7)
    142	clrl		%d1
    143	fmovemx	%fp0-%fp0,(%a0)
    144	bsr		setox		| ...FP0 IS EXP(Y)
    145	movel		(%a7)+,%d1
    146	movel		SGN(%a6),%d0
    147	fadds		#0x3F800000,%fp0	| ...EXP(Y)+1
    148
    149	eorl		#0xC0000000,%d0	| ...-SIGN(X)*2
    150	fmoves		%d0,%fp1		| ...-SIGN(X)*2 IN SGL FMT
    151	fdivx		%fp0,%fp1		| ...-SIGN(X)2 / [EXP(Y)+1 ]
    152
    153	movel		SGN(%a6),%d0
    154	orl		#0x3F800000,%d0	| ...SGN
    155	fmoves		%d0,%fp0		| ...SGN IN SGL FMT
    156
    157	fmovel		%d1,%FPCR		|restore users exceptions
    158	faddx		%fp1,%fp0
    159
    160	bra		t_frcinx
    161
    162TANHSM:
    163	movew		#0x0000,XDCARE(%a6)
    164
    165	fmovel		%d1,%FPCR		|restore users exceptions
    166	fmovex		X(%a6),%fp0		|last inst - possible exception set
    167
    168	bra		t_frcinx
    169
    170TANHHUGE:
    171|---RETURN SGN(X) - SGN(X)EPS
    172	movel		X(%a6),%d0
    173	andl		#0x80000000,%d0
    174	orl		#0x3F800000,%d0
    175	fmoves		%d0,%fp0
    176	andl		#0x80000000,%d0
    177	eorl		#0x80800000,%d0	| ...-SIGN(X)*EPS
    178
    179	fmovel		%d1,%FPCR		|restore users exceptions
    180	fadds		%d0,%fp0
    181
    182	bra		t_frcinx
    183
    184	|end