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

scosh.S (2995B)


      1|
      2|	scosh.sa 3.1 12/10/90
      3|
      4|	The entry point sCosh computes the hyperbolic cosine of
      5|	an input argument; sCoshd 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 cosh(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 sCOSH takes approximately 250 cycles.
     19|
     20|	Algorithm:
     21|
     22|	COSH
     23|	1. If |X| > 16380 log2, go to 3.
     24|
     25|	2. (|X| <= 16380 log2) Cosh(X) is obtained by the formulae
     26|		y = |X|, z = exp(Y), and
     27|		cosh(X) = (1/2)*( z + 1/z ).
     28|		Exit.
     29|
     30|	3. (|X| > 16380 log2). If |X| > 16480 log2, go to 5.
     31|
     32|	4. (16380 log2 < |X| <= 16480 log2)
     33|		cosh(X) = sign(X) * exp(|X|)/2.
     34|		However, invoking exp(|X|) may cause premature overflow.
     35|		Thus, we calculate sinh(X) as follows:
     36|		Y	:= |X|
     37|		Fact	:=	2**(16380)
     38|		Y'	:= Y - 16381 log2
     39|		cosh(X) := Fact * exp(Y').
     40|		Exit.
     41|
     42|	5. (|X| > 16480 log2) sinh(X) must overflow. Return
     43|		Huge*Huge to generate overflow and an infinity with
     44|		the appropriate sign. Huge is the largest finite number in
     45|		extended format. Exit.
     46|
     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|SCOSH	idnt	2,1 | Motorola 040 Floating Point Software Package
     56
     57	|section	8
     58
     59	|xref	t_ovfl
     60	|xref	t_frcinx
     61	|xref	setox
     62
     63T1:	.long 0x40C62D38,0xD3D64634 | ... 16381 LOG2 LEAD
     64T2:	.long 0x3D6F90AE,0xB1E75CC7 | ... 16381 LOG2 TRAIL
     65
     66TWO16380: .long 0x7FFB0000,0x80000000,0x00000000,0x00000000
     67
     68	.global	scoshd
     69scoshd:
     70|--COSH(X) = 1 FOR DENORMALIZED X
     71
     72	fmoves		#0x3F800000,%fp0
     73
     74	fmovel		%d1,%FPCR
     75	fadds		#0x00800000,%fp0
     76	bra		t_frcinx
     77
     78	.global	scosh
     79scosh:
     80	fmovex		(%a0),%fp0	| ...LOAD INPUT
     81
     82	movel		(%a0),%d0
     83	movew		4(%a0),%d0
     84	andil		#0x7FFFFFFF,%d0
     85	cmpil		#0x400CB167,%d0
     86	bgts		COSHBIG
     87
     88|--THIS IS THE USUAL CASE, |X| < 16380 LOG2
     89|--COSH(X) = (1/2) * ( EXP(X) + 1/EXP(X) )
     90
     91	fabsx		%fp0		| ...|X|
     92
     93	movel		%d1,-(%sp)
     94	clrl		%d1
     95	fmovemx	%fp0-%fp0,(%a0)	|pass parameter to setox
     96	bsr		setox		| ...FP0 IS EXP(|X|)
     97	fmuls		#0x3F000000,%fp0	| ...(1/2)EXP(|X|)
     98	movel		(%sp)+,%d1
     99
    100	fmoves		#0x3E800000,%fp1	| ...(1/4)
    101	fdivx		%fp0,%fp1		| ...1/(2 EXP(|X|))
    102
    103	fmovel		%d1,%FPCR
    104	faddx		%fp1,%fp0
    105
    106	bra		t_frcinx
    107
    108COSHBIG:
    109	cmpil		#0x400CB2B3,%d0
    110	bgts		COSHHUGE
    111
    112	fabsx		%fp0
    113	fsubd		T1(%pc),%fp0		| ...(|X|-16381LOG2_LEAD)
    114	fsubd		T2(%pc),%fp0		| ...|X| - 16381 LOG2, ACCURATE
    115
    116	movel		%d1,-(%sp)
    117	clrl		%d1
    118	fmovemx	%fp0-%fp0,(%a0)
    119	bsr		setox
    120	fmovel		(%sp)+,%fpcr
    121
    122	fmulx		TWO16380(%pc),%fp0
    123	bra		t_frcinx
    124
    125COSHHUGE:
    126	fmovel		#0,%fpsr		|clr N bit if set by source
    127	bclrb		#7,(%a0)		|always return positive value
    128	fmovemx	(%a0),%fp0-%fp0
    129	bra		t_ovfl
    130
    131	|end