VxWorks

开发平台：

C/C++

l_entry.s：源码内容

/* l_entry.s - Motorola 68040 FP entry points (LIB) */
.data
.globl _copyright_wind_river
.long _copyright_wind_river
/*
modification history
--------------------
01j,28jan97,ms fixed SPR 7822 (bug from 01i check-in - never part of product)
01i,31may96,ms updated to mototorola version 2.3
01h,21jul93,kdl added .text (SPR #2372).
01g,18sep92,kdl restored title line clobbered in 01f checkin.
01f,18sep92,kdl changed multi-reg (fpsr/fpcr) saves to use offset USER_FPCR
instead of USER_FPSR, to reflect actual save order (SPR #1455).
01e,23aug92,jcf changed bxxx to jxx.
01d,20aug92,kdl added changes from Motorola FPSP v2.2 to correctly save fpcr.
01c,26may92,rrr the tree shuffle
01b,09jan92,kdl added modification history; general cleanup.
01a,15aug91,kdl original version, from Motorola FPSP v2.0.
*/
/*
DESCRIPTION
This file contains Motorola FPSP library interface entry points for
various floating point operations.
*/
| section 8
#include "fpsp040L.h"
| xref __l_tag
| xref __l_szero
| xref __l_sinf
| xref __l_sopr_inf
| xref __l_sone
| xref __l_spi_2
| xref __l_szr_inf
| xref __l_src_nan
| xref __l_t_operr
| xref __l_t_dz2
| xref __l_snzrinx
| xref __l_ld_pone
| xref __l_ld_pinf
| xref __l_ld_ppi2
| xref __l_ssincosz
| xref __l_ssincosi
| xref __l_ssincosnan
| xref __l_setoxm1i
|
| MONADIC.GEN 1.3 4/30/91
|
| MONADIC.GEN --- generic MONADIC template
|
| This version saves all registers that will be used by the emulation
| routines and restores all but FP0 on exit. The FPSR is
| updated to reflect the result of the operation. Return value
| is placed in FP0 for single, double and extended results.
|
| The package subroutines expect the incoming fpcr to be zeroed
| since they need extended precision to work properly. The
/* 'final' fpcr is expected in d1 so that the calculated result */
| can be properly sized and rounded. Also, if the incoming fpcr
| has enabled any exceptions, the exception will be taken on the
| final fmovem in this template.
|
| Customizations:
| 1. Remove the moveml at the entry and exit of
| each routine if your compiler treats those
| registers as scratch.
/* 2. Likewise, don't save FP0/FP1 if they are scratch */
| registers.
| 3. Delete handling of the fpsr if you only care about
| the result.
| 4. Some (most?) C compilers convert all float arguments
| to double, and provide no support at all for extended
| precision so remove the __l_facoss and __l_facosx entry points.
| 5. Move the result to d0/d1 if the compiler is that old.
|
| THIS IS UNPUBLISHED PROPRIETARY SOURCE CODE OF MOTOROLA
| The copyright notice above does not evidence any
| actual or intended publication of such source code.
| xref __l_tag
| xref __l_sacos
| xref __l_ld_ppi2
| xref __l_t_operr
| xref __l_mon_nan
| xref __l_sacosd
.text
.globl __l_facoss
__l_facoss:
link a6,#-LOCAL_SIZE
moveml d0-d1/a0-a1,a6@(USER_DA)
fmovemx fp0-fp3,a6@(USER_FP0)
fmovel fpsr,a6@(USER_FPSR)
fmovel fpcr,a6@(USER_FPCR)
fmovel fpcr,d1 /* | user's rounding mode/precision */
fmovel #0,fpcr | force rounding mode/prec to extended,rn
|
| copy, convert and tag input argument
|
fmoves a6@(8),fp0
fmovex fp0,a6@(ETEMP)
lea a6@(ETEMP),a0
bsrl __l_tag
moveb d0,a6@(STAG)
tstb d0
jne L_1012
bsrl __l_sacos | normalized (regular) number
jra L_1016
L_1012:
cmpb #0x20,d0 | zero?
jne L_1013
bsrl __l_ld_ppi2
jra L_1016
L_1013:
cmpb #0x40,d0 | infinity?
jne L_1014
bsrl __l_t_operr
jra L_1016
L_1014:
cmpb #0x60,d0 | NaN?
jne L_1015
bsrl __l_mon_nan
jra L_1016
L_1015:
bsrl __l_sacosd | assuming a denorm...
L_1016:
fmovel fpsr,d0 | update status register
orb a6@(USER_FPSR+3),d0 | add previously accrued exceptions
fmovel d0,fpsr
|
| Result is now in FP0
|
moveml a6@(USER_DA),d0-d1/a0-a1
fmovemx a6@(USER_FP1),fp1-fp3 | note: fp0 not restored
fmovel a6@(USER_FPCR),fpcr | fpcr restored
unlk a6
rts
.globl __l_facosd
__l_facosd:
link a6,#-LOCAL_SIZE
moveml d0-d1/a0-a1,a6@(USER_DA)
fmovemx fp0-fp3,a6@(USER_FP0)
fmovel fpsr,a6@(USER_FPSR)
fmovel fpcr,a6@(USER_FPCR)
fmovel fpcr,d1 /* | user's rounding mode/precision */
fmovel #0,fpcr | force rounding mode/prec to extended,rn
|
| copy, convert and tag input argument
|
fmoved a6@(8),fp0
fmovex fp0,a6@(ETEMP)
lea a6@(ETEMP),a0
bsrl __l_tag
moveb d0,a6@(STAG)
tstb d0
jne L_1017
bsrl __l_sacos | normalized (regular) number
jra L_101B
L_1017:
cmpb #0x20,d0 | zero?
jne L_1018
bsrl __l_ld_ppi2
jra L_101B
L_1018:
cmpb #0x40,d0 | infinity?
jne L_1019
bsrl __l_t_operr
jra L_101B
L_1019:
cmpb #0x60,d0 | NaN?
jne L_101A
bsrl __l_mon_nan
jra L_101B
L_101A:
bsrl __l_sacosd | assuming a denorm...
L_101B:
fmovel fpsr,d0 | update status register
orb a6@(USER_FPSR+3),d0 | add previously accrued exceptions
fmovel d0,fpsr
|
| Result is now in FP0
|
moveml a6@(USER_DA),d0-d1/a0-a1
fmovemx a6@(USER_FP1),fp1-fp3 | note: fp0 not restored
fmovel a6@(USER_FPCR),fpcr | fpcr restored
unlk a6
rts
.globl __l_facosx
__l_facosx:
link a6,#-LOCAL_SIZE
moveml d0-d1/a0-a1,a6@(USER_DA)
fmovemx fp0-fp3,a6@(USER_FP0)
fmovel fpsr,a6@(USER_FPSR)
fmovel fpcr,a6@(USER_FPCR)
fmovel fpcr,d1 /* | user's rounding mode/precision */
fmovel #0,fpcr | force rounding mode/prec to extended,rn
|
| copy, convert and tag input argument
|
fmovex a6@(8),fp0
fmovex fp0,a6@(ETEMP)
lea a6@(ETEMP),a0
bsrl __l_tag
moveb d0,a6@(STAG)
tstb d0
jne L_101C
bsrl __l_sacos | normalized (regular) number
jra L_101G
L_101C:
cmpb #0x20,d0 | zero?
jne L_101D
bsrl __l_ld_ppi2
jra L_101G
L_101D:
cmpb #0x40,d0 | infinity?
jne L_101E
bsrl __l_t_operr
jra L_101G
L_101E:
cmpb #0x60,d0 | NaN?
jne L_101F
bsrl __l_mon_nan
jra L_101G
L_101F:
bsrl __l_sacosd | assuming a denorm...
L_101G:
fmovel fpsr,d0 | update status register
orb a6@(USER_FPSR+3),d0 | add previously accrued exceptions
fmovel d0,fpsr
|
| Result is now in FP0
|
moveml a6@(USER_DA),d0-d1/a0-a1
fmovemx a6@(USER_FP1),fp1-fp3 | note: fp0 not restored
fmovel a6@(USER_FPCR),fpcr | fpcr restored
unlk a6
rts
|
| MONADIC.GEN 1.3 4/30/91
|
| MONADIC.GEN --- generic MONADIC template
|
| This version saves all registers that will be used by the emulation
| routines and restores all but FP0 on exit. The FPSR is
| updated to reflect the result of the operation. Return value
| is placed in FP0 for single, double and extended results.
|
| The package subroutines expect the incoming fpcr to be zeroed
| since they need extended precision to work properly. The
/* | 'final' fpcr is expected in d1 so that the calculated result */
| can be properly sized and rounded. Also, if the incoming fpcr
| has enabled any exceptions, the exception will be taken on the
| final fmovem in this template.
|
| Customizations:
| 1. Remove the moveml at the entry and exit of
| each routine if your compiler treats those
| registers as scratch.
/* | 2. Likewise, don't save FP0/FP1 if they are scratch */
| registers.
| 3. Delete handling of the fpsr if you only care about
| the result.
| 4. Some (most?) C compilers convert all float arguments
| to double, and provide no support at all for extended
| precision so remove the __l_fasins and __l_fasinx entry points.
| 5. Move the result to d0/d1 if the compiler is that old.
|
| THIS IS UNPUBLISHED PROPRIETARY SOURCE CODE OF MOTOROLA
| The copyright notice above does not evidence any
| actual or intended publication of such source code.
| xref __l_tag
| xref __l_sasin
| xref __l_szero
| xref __l_t_operr
| xref __l_mon_nan
| xref __l_sasind
.globl __l_fasins
__l_fasins:
link a6,#-LOCAL_SIZE
moveml d0-d1/a0-a1,a6@(USER_DA)
fmovemx fp0-fp3,a6@(USER_FP0)
fmovel fpsr,a6@(USER_FPSR)
fmovel fpcr,a6@(USER_FPCR)
fmovel fpcr,d1 /* | user's rounding mode/precision */
fmovel #0,fpcr | force rounding mode/prec to extended,rn
|
| copy, convert and tag input argument
|
fmoves a6@(8),fp0
fmovex fp0,a6@(ETEMP)
lea a6@(ETEMP),a0
bsrl __l_tag
moveb d0,a6@(STAG)
tstb d0
jne L_1022
bsrl __l_sasin | normalized (regular) number
jra L_1026
L_1022:
cmpb #0x20,d0 | zero?
jne L_1023
bsrl __l_szero
jra L_1026
L_1023:
cmpb #0x40,d0 | infinity?
jne L_1024
bsrl __l_t_operr
jra L_1026
L_1024:
cmpb #0x60,d0 | NaN?
jne L_1025
bsrl __l_mon_nan
jra L_1026
L_1025:
bsrl __l_sasind | assuming a denorm...
L_1026:
fmovel fpsr,d0 | update status register
orb a6@(USER_FPSR+3),d0 | add previously accrued exceptions
fmovel d0,fpsr
|
| Result is now in FP0
|
moveml a6@(USER_DA),d0-d1/a0-a1
fmovemx a6@(USER_FP1),fp1-fp3 | note: fp0 not restored
fmovel a6@(USER_FPCR),fpcr | fpcr restored
unlk a6
rts
.globl __l_fasind
__l_fasind:
link a6,#-LOCAL_SIZE
moveml d0-d1/a0-a1,a6@(USER_DA)
fmovemx fp0-fp3,a6@(USER_FP0)
fmovel fpsr,a6@(USER_FPSR)
fmovel fpcr,a6@(USER_FPCR)
fmovel fpcr,d1 /* | user's rounding mode/precision */
fmovel #0,fpcr | force rounding mode/prec to extended,rn
|
| copy, convert and tag input argument
|
fmoved a6@(8),fp0
fmovex fp0,a6@(ETEMP)
lea a6@(ETEMP),a0
bsrl __l_tag
moveb d0,a6@(STAG)
tstb d0
jne L_1027
bsrl __l_sasin | normalized (regular) number
jra L_102B
L_1027:
cmpb #0x20,d0 | zero?
jne L_1028
bsrl __l_szero
jra L_102B
L_1028:
cmpb #0x40,d0 | infinity?
jne L_1029
bsrl __l_t_operr
jra L_102B
L_1029:
cmpb #0x60,d0 | NaN?
jne L_102A
bsrl __l_mon_nan
jra L_102B
L_102A:
bsrl __l_sasind | assuming a denorm...
L_102B:
fmovel fpsr,d0 | update status register
orb a6@(USER_FPSR+3),d0 | add previously accrued exceptions
fmovel d0,fpsr
|
| Result is now in FP0
|
moveml a6@(USER_DA),d0-d1/a0-a1
fmovemx a6@(USER_FP1),fp1-fp3 | note: fp0 not restored
fmovel a6@(USER_FPCR),fpcr | fpcr restored
unlk a6
rts
.globl __l_fasinx
__l_fasinx:
link a6,#-LOCAL_SIZE
moveml d0-d1/a0-a1,a6@(USER_DA)
fmovemx fp0-fp3,a6@(USER_FP0)
fmovel fpsr,a6@(USER_FPSR)
fmovel fpcr,a6@(USER_FPCR)
fmovel fpcr,d1 /* | user's rounding mode/precision */
fmovel #0,fpcr | force rounding mode/prec to extended,rn
|
| copy, convert and tag input argument
|
fmovex a6@(8),fp0
fmovex fp0,a6@(ETEMP)
lea a6@(ETEMP),a0
bsrl __l_tag
moveb d0,a6@(STAG)
tstb d0
jne L_102C
bsrl __l_sasin | normalized (regular) number
jra L_102G
L_102C:
cmpb #0x20,d0 | zero?
jne L_102D
bsrl __l_szero
jra L_102G
L_102D:
cmpb #0x40,d0 | infinity?
jne L_102E
bsrl __l_t_operr
jra L_102G
L_102E:
cmpb #0x60,d0 | NaN?
jne L_102F
bsrl __l_mon_nan
jra L_102G
L_102F:
bsrl __l_sasind | assuming a denorm...
L_102G:
fmovel fpsr,d0 | update status register
orb a6@(USER_FPSR+3),d0 | add previously accrued exceptions
fmovel d0,fpsr
|
| Result is now in FP0
|
moveml a6@(USER_DA),d0-d1/a0-a1
fmovemx a6@(USER_FP1),fp1-fp3 | note: fp0 not restored
fmovel a6@(USER_FPCR),fpcr | fpcr restored
unlk a6
rts
|
| MONADIC.GEN 1.3 4/30/91
|
| MONADIC.GEN --- generic MONADIC template
|
| This version saves all registers that will be used by the emulation
| routines and restores all but FP0 on exit. The FPSR is
| updated to reflect the result of the operation. Return value
| is placed in FP0 for single, double and extended results.
|
| The package subroutines expect the incoming fpcr to be zeroed
| since they need extended precision to work properly. The
/* | 'final' fpcr is expected in d1 so that the calculated result */
| can be properly sized and rounded. Also, if the incoming fpcr
| has enabled any exceptions, the exception will be taken on the
| final fmovem in this template.
|
| Customizations:
| 1. Remove the moveml at the entry and exit of
| each routine if your compiler treats those
| registers as scratch.
/* | 2. Likewise, don't save FP0/FP1 if they are scratch */
| registers.
| 3. Delete handling of the fpsr if you only care about
| the result.
| 4. Some (most?) C compilers convert all float arguments
| to double, and provide no support at all for extended
| precision so remove the __l_fatans and __l_fatanx entry points.
| 5. Move the result to d0/d1 if the compiler is that old.
|
| THIS IS UNPUBLISHED PROPRIETARY SOURCE CODE OF MOTOROLA
| The copyright notice above does not evidence any
| actual or intended publication of such source code.
| xref __l_tag
| xref __l_satan
| xref __l_szero
| xref __l_spi_2
| xref __l_mon_nan
| xref __l_satand
.globl __l_fatans
__l_fatans:
link a6,#-LOCAL_SIZE
moveml d0-d1/a0-a1,a6@(USER_DA)
fmovemx fp0-fp3,a6@(USER_FP0)
fmovel fpsr,a6@(USER_FPSR)
fmovel fpcr,a6@(USER_FPCR)
fmovel fpcr,d1 /* | user's rounding mode/precision */
fmovel #0,fpcr | force rounding mode/prec to extended,rn
|
| copy, convert and tag input argument
|
fmoves a6@(8),fp0
fmovex fp0,a6@(ETEMP)
lea a6@(ETEMP),a0
bsrl __l_tag
moveb d0,a6@(STAG)
tstb d0
jne L_1032
bsrl __l_satan | normalized (regular) number
jra L_1036
L_1032:
cmpb #0x20,d0 | zero?
jne L_1033
bsrl __l_szero
jra L_1036
L_1033:
cmpb #0x40,d0 | infinity?
jne L_1034
bsrl __l_spi_2
jra L_1036
L_1034:
cmpb #0x60,d0 | NaN?
jne L_1035
bsrl __l_mon_nan
jra L_1036
L_1035:
bsrl __l_satand | assuming a denorm...
L_1036:
fmovel fpsr,d0 | update status register
orb a6@(USER_FPSR+3),d0 | add previously accrued exceptions
fmovel d0,fpsr
|
| Result is now in FP0
|
moveml a6@(USER_DA),d0-d1/a0-a1
fmovemx a6@(USER_FP1),fp1-fp3 | note: fp0 not restored
fmovel a6@(USER_FPCR),fpcr | fpcr restored
unlk a6
rts
.globl __l_fatand
__l_fatand:
link a6,#-LOCAL_SIZE
moveml d0-d1/a0-a1,a6@(USER_DA)
fmovemx fp0-fp3,a6@(USER_FP0)
fmovel fpsr,a6@(USER_FPSR)
fmovel fpcr,a6@(USER_FPCR)
fmovel fpcr,d1 /* | user's rounding mode/precision */
fmovel #0,fpcr | force rounding mode/prec to extended,rn
|
| copy, convert and tag input argument
|
fmoved a6@(8),fp0
fmovex fp0,a6@(ETEMP)
lea a6@(ETEMP),a0
bsrl __l_tag
moveb d0,a6@(STAG)
tstb d0
jne L_1037
bsrl __l_satan | normalized (regular) number
jra L_103B
L_1037:
cmpb #0x20,d0 | zero?
jne L_1038
bsrl __l_szero
jra L_103B
L_1038:
cmpb #0x40,d0 | infinity?
jne L_1039
bsrl __l_spi_2
jra L_103B
L_1039:
cmpb #0x60,d0 | NaN?
jne L_103A
bsrl __l_mon_nan
jra L_103B
L_103A:
bsrl __l_satand | assuming a denorm...
L_103B:
fmovel fpsr,d0 | update status register
orb a6@(USER_FPSR+3),d0 | add previously accrued exceptions
fmovel d0,fpsr
|
| Result is now in FP0
|
moveml a6@(USER_DA),d0-d1/a0-a1
fmovemx a6@(USER_FP1),fp1-fp3 | note: fp0 not restored
fmovel a6@(USER_FPCR),fpcr | fpcr restored
unlk a6
rts
.globl __l_fatanx
__l_fatanx:
link a6,#-LOCAL_SIZE
moveml d0-d1/a0-a1,a6@(USER_DA)
fmovemx fp0-fp3,a6@(USER_FP0)
fmovel fpsr,a6@(USER_FPSR)
fmovel fpcr,a6@(USER_FPCR)
fmovel fpcr,d1 /* | user's rounding mode/precision */
fmovel #0,fpcr | force rounding mode/prec to extended,rn
|
| copy, convert and tag input argument
|
fmovex a6@(8),fp0
fmovex fp0,a6@(ETEMP)
lea a6@(ETEMP),a0
bsrl __l_tag
moveb d0,a6@(STAG)
tstb d0
jne L_103C
bsrl __l_satan | normalized (regular) number
jra L_103G
L_103C:
cmpb #0x20,d0 | zero?
jne L_103D
bsrl __l_szero
jra L_103G
L_103D:
cmpb #0x40,d0 | infinity?
jne L_103E
bsrl __l_spi_2
jra L_103G
L_103E:
cmpb #0x60,d0 | NaN?
jne L_103F
bsrl __l_mon_nan
jra L_103G
L_103F:
bsrl __l_satand | assuming a denorm...
L_103G:
fmovel fpsr,d0 | update status register
orb a6@(USER_FPSR+3),d0 | add previously accrued exceptions
fmovel d0,fpsr
|
| Result is now in FP0
|
moveml a6@(USER_DA),d0-d1/a0-a1
fmovemx a6@(USER_FP1),fp1-fp3 | note: fp0 not restored
fmovel a6@(USER_FPCR),fpcr | fpcr restored
unlk a6
rts
|
| MONADIC.GEN 1.3 4/30/91
|
| MONADIC.GEN --- generic MONADIC template
|
| This version saves all registers that will be used by the emulation
| routines and restores all but FP0 on exit. The FPSR is
| updated to reflect the result of the operation. Return value
| is placed in FP0 for single, double and extended results.
|
| The package subroutines expect the incoming fpcr to be zeroed
| since they need extended precision to work properly. The
/* | 'final' fpcr is expected in d1 so that the calculated result */
| can be properly sized and rounded. Also, if the incoming fpcr
| has enabled any exceptions, the exception will be taken on the
| final fmovem in this template.
|
| Customizations:
| 1. Remove the moveml at the entry and exit of
| each routine if your compiler treats those
| registers as scratch.
/* | 2. Likewise, don't save FP0/FP1 if they are scratch */
| registers.
| 3. Delete handling of the fpsr if you only care about
| the result.
| 4. Some (most?) C compilers convert all float arguments
| to double, and provide no support at all for extended
| precision so remove the __l_fatanhs and __l_fatanhx entry points.
| 5. Move the result to d0/d1 if the compiler is that old.
|
| THIS IS UNPUBLISHED PROPRIETARY SOURCE CODE OF MOTOROLA
| The copyright notice above does not evidence any
| actual or intended publication of such source code.
| xref __l_tag
| xref __l_satanh
| xref __l_szero
| xref __l_t_operr
| xref __l_mon_nan
| xref __l_satanhd
.globl __l_fatanhs
__l_fatanhs:
link a6,#-LOCAL_SIZE
moveml d0-d1/a0-a1,a6@(USER_DA)
fmovemx fp0-fp3,a6@(USER_FP0)
fmovel fpsr,a6@(USER_FPSR)
fmovel fpcr,a6@(USER_FPCR)
fmovel fpcr,d1 /* | user's rounding mode/precision */
fmovel #0,fpcr | force rounding mode/prec to extended,rn
|
| copy, convert and tag input argument
|
fmoves a6@(8),fp0
fmovex fp0,a6@(ETEMP)
lea a6@(ETEMP),a0
bsrl __l_tag
moveb d0,a6@(STAG)
tstb d0
jne L_1042
bsrl __l_satanh | normalized (regular) number
jra L_1046
L_1042:
cmpb #0x20,d0 | zero?
jne L_1043
bsrl __l_szero
jra L_1046
L_1043:
cmpb #0x40,d0 | infinity?
jne L_1044
bsrl __l_t_operr
jra L_1046
L_1044:
cmpb #0x60,d0 | NaN?
jne L_1045
bsrl __l_mon_nan
jra L_1046
L_1045:
bsrl __l_satanhd | assuming a denorm...
L_1046:
fmovel fpsr,d0 | update status register
orb a6@(USER_FPSR+3),d0 | add previously accrued exceptions
fmovel d0,fpsr
|
| Result is now in FP0
|
moveml a6@(USER_DA),d0-d1/a0-a1
fmovemx a6@(USER_FP1),fp1-fp3 | note: fp0 not restored
fmovel a6@(USER_FPCR),fpcr | fpcr restored
unlk a6
rts
.globl __l_fatanhd
__l_fatanhd:
link a6,#-LOCAL_SIZE
moveml d0-d1/a0-a1,a6@(USER_DA)
fmovemx fp0-fp3,a6@(USER_FP0)
fmovel fpsr,a6@(USER_FPSR)
fmovel fpcr,a6@(USER_FPCR)
fmovel fpcr,d1 /* | user's rounding mode/precision */
fmovel #0,fpcr | force rounding mode/prec to extended,rn
|
| copy, convert and tag input argument
|
fmoved a6@(8),fp0
fmovex fp0,a6@(ETEMP)
lea a6@(ETEMP),a0
bsrl __l_tag
moveb d0,a6@(STAG)
tstb d0
jne L_1047
bsrl __l_satanh | normalized (regular) number
jra L_104B
L_1047:
cmpb #0x20,d0 | zero?
jne L_1048
bsrl __l_szero
jra L_104B
L_1048:
cmpb #0x40,d0 | infinity?
jne L_1049
bsrl __l_t_operr
jra L_104B
L_1049:
cmpb #0x60,d0 | NaN?
jne L_104A
bsrl __l_mon_nan
jra L_104B
L_104A:
bsrl __l_satanhd | assuming a denorm...
L_104B:
fmovel fpsr,d0 | update status register
orb a6@(USER_FPSR+3),d0 | add previously accrued exceptions
fmovel d0,fpsr
|
| Result is now in FP0
|
moveml a6@(USER_DA),d0-d1/a0-a1
fmovemx a6@(USER_FP1),fp1-fp3 | note: fp0 not restored
fmovel a6@(USER_FPCR),fpcr | fpcr restored
unlk a6
rts
.globl __l_fatanhx
__l_fatanhx:
link a6,#-LOCAL_SIZE
moveml d0-d1/a0-a1,a6@(USER_DA)
fmovemx fp0-fp3,a6@(USER_FP0)
fmovel fpsr,a6@(USER_FPSR)
fmovel fpcr,a6@(USER_FPCR)
fmovel fpcr,d1 /* | user's rounding mode/precision */
fmovel #0,fpcr | force rounding mode/prec to extended,rn
|
| copy, convert and tag input argument
|
fmovex a6@(8),fp0
fmovex fp0,a6@(ETEMP)
lea a6@(ETEMP),a0
bsrl __l_tag
moveb d0,a6@(STAG)
tstb d0
jne L_104C
bsrl __l_satanh | normalized (regular) number
jra L_104G
L_104C:
cmpb #0x20,d0 | zero?
jne L_104D
bsrl __l_szero
jra L_104G
L_104D:
cmpb #0x40,d0 | infinity?
jne L_104E
bsrl __l_t_operr
jra L_104G
L_104E:
cmpb #0x60,d0 | NaN?
jne L_104F
bsrl __l_mon_nan
jra L_104G
L_104F:
bsrl __l_satanhd | assuming a denorm...
L_104G:
fmovel fpsr,d0 | update status register
orb a6@(USER_FPSR+3),d0 | add previously accrued exceptions
fmovel d0,fpsr
|
| Result is now in FP0
|
moveml a6@(USER_DA),d0-d1/a0-a1
fmovemx a6@(USER_FP1),fp1-fp3 | note: fp0 not restored
fmovel a6@(USER_FPCR),fpcr | fpcr restored
unlk a6
rts
|
| MONADIC.GEN 1.3 4/30/91
|
| MONADIC.GEN --- generic MONADIC template
|
| This version saves all registers that will be used by the emulation
| routines and restores all but FP0 on exit. The FPSR is
| updated to reflect the result of the operation. Return value
| is placed in FP0 for single, double and extended results.
|
| The package subroutines expect the incoming fpcr to be zeroed
| since they need extended precision to work properly. The
/* | 'final' fpcr is expected in d1 so that the calculated result */
| can be properly sized and rounded. Also, if the incoming fpcr
| has enabled any exceptions, the exception will be taken on the
| final fmovem in this template.
|
| Customizations:
| 1. Remove the moveml at the entry and exit of
| each routine if your compiler treats those
| registers as scratch.
/* | 2. Likewise, don't save FP0/FP1 if they are scratch */
| registers.
| 3. Delete handling of the fpsr if you only care about
| the result.
| 4. Some (most?) C compilers convert all float arguments
| to double, and provide no support at all for extended
| precision so remove the __l_fcoss and __l_fcosx entry points.
| 5. Move the result to d0/d1 if the compiler is that old.
|
| THIS IS UNPUBLISHED PROPRIETARY SOURCE CODE OF MOTOROLA
| The copyright notice above does not evidence any
| actual or intended publication of such source code.
| xref __l_tag
| xref __l_scos
| xref __l_ld_pone
| xref __l_t_operr
| xref __l_mon_nan
| xref __l_scosd
.globl __l_fcoss
__l_fcoss:
link a6,#-LOCAL_SIZE
moveml d0-d1/a0-a1,a6@(USER_DA)
fmovemx fp0-fp3,a6@(USER_FP0)
fmovel fpsr,a6@(USER_FPSR)
fmovel fpcr,a6@(USER_FPCR)
fmovel fpcr,d1 /* | user's rounding mode/precision */
fmovel #0,fpcr | force rounding mode/prec to extended,rn
|
| copy, convert and tag input argument
|
fmoves a6@(8),fp0
fmovex fp0,a6@(ETEMP)
lea a6@(ETEMP),a0
bsrl __l_tag
moveb d0,a6@(STAG)
tstb d0
jne L_1052
bsrl __l_scos | normalized (regular) number
jra L_1056
L_1052:
cmpb #0x20,d0 | zero?
jne L_1053
bsrl __l_ld_pone
jra L_1056
L_1053:
cmpb #0x40,d0 | infinity?
jne L_1054
bsrl __l_t_operr
jra L_1056
L_1054:
cmpb #0x60,d0 | NaN?
jne L_1055
bsrl __l_mon_nan
jra L_1056
L_1055:
bsrl __l_scosd | assuming a denorm...
L_1056:
fmovel fpsr,d0 | update status register
orb a6@(USER_FPSR+3),d0 | add previously accrued exceptions
fmovel d0,fpsr
|
| Result is now in FP0
|
moveml a6@(USER_DA),d0-d1/a0-a1
fmovemx a6@(USER_FP1),fp1-fp3 | note: fp0 not restored
fmovel a6@(USER_FPCR),fpcr | fpcr restored
unlk a6
rts
.globl __l_fcosd
__l_fcosd:
link a6,#-LOCAL_SIZE
moveml d0-d1/a0-a1,a6@(USER_DA)
fmovemx fp0-fp3,a6@(USER_FP0)
fmovel fpsr,a6@(USER_FPSR)
fmovel fpcr,a6@(USER_FPCR)
fmovel fpcr,d1 /* | user's rounding mode/precision */
fmovel #0,fpcr | force rounding mode/prec to extended,rn
|
| copy, convert and tag input argument
|
fmoved a6@(8),fp0
fmovex fp0,a6@(ETEMP)
lea a6@(ETEMP),a0
bsrl __l_tag
moveb d0,a6@(STAG)
tstb d0
jne L_1057
bsrl __l_scos | normalized (regular) number
jra L_105B
L_1057:
cmpb #0x20,d0 | zero?
jne L_1058
bsrl __l_ld_pone
jra L_105B
L_1058:
cmpb #0x40,d0 | infinity?
jne L_1059
bsrl __l_t_operr
jra L_105B
L_1059:
cmpb #0x60,d0 | NaN?
jne L_105A
bsrl __l_mon_nan
jra L_105B
L_105A:
bsrl __l_scosd | assuming a denorm...
L_105B:
fmovel fpsr,d0 | update status register
orb a6@(USER_FPSR+3),d0 | add previously accrued exceptions
fmovel d0,fpsr
|
| Result is now in FP0
|
moveml a6@(USER_DA),d0-d1/a0-a1
fmovemx a6@(USER_FP1),fp1-fp3 | note: fp0 not restored
fmovel a6@(USER_FPCR),fpcr | fpcr restored
unlk a6
rts
.globl __l_fcosx
__l_fcosx:
link a6,#-LOCAL_SIZE
moveml d0-d1/a0-a1,a6@(USER_DA)
fmovemx fp0-fp3,a6@(USER_FP0)
fmovel fpsr,a6@(USER_FPSR)
fmovel fpcr,a6@(USER_FPCR)
fmovel fpcr,d1 /* | user's rounding mode/precision */
fmovel #0,fpcr | force rounding mode/prec to extended,rn
|
| copy, convert and tag input argument
|
fmovex a6@(8),fp0
fmovex fp0,a6@(ETEMP)
lea a6@(ETEMP),a0
bsrl __l_tag
moveb d0,a6@(STAG)
tstb d0
jne L_105C
bsrl __l_scos | normalized (regular) number
jra L_105G
L_105C:
cmpb #0x20,d0 | zero?
jne L_105D
bsrl __l_ld_pone
jra L_105G
L_105D:
cmpb #0x40,d0 | infinity?
jne L_105E
bsrl __l_t_operr
jra L_105G
L_105E:
cmpb #0x60,d0 | NaN?
jne L_105F
bsrl __l_mon_nan
jra L_105G
L_105F:
bsrl __l_scosd | assuming a denorm...
L_105G:
fmovel fpsr,d0 | update status register
orb a6@(USER_FPSR+3),d0 | add previously accrued exceptions
fmovel d0,fpsr
|
| Result is now in FP0
|
moveml a6@(USER_DA),d0-d1/a0-a1
fmovemx a6@(USER_FP1),fp1-fp3 | note: fp0 not restored
fmovel a6@(USER_FPCR),fpcr | fpcr restored
unlk a6
rts
|
| MONADIC.GEN 1.3 4/30/91
|
| MONADIC.GEN --- generic MONADIC template
|
| This version saves all registers that will be used by the emulation
| routines and restores all but FP0 on exit. The FPSR is
| updated to reflect the result of the operation. Return value
| is placed in FP0 for single, double and extended results.
|
| The package subroutines expect the incoming fpcr to be zeroed
| since they need extended precision to work properly. The
/* | 'final' fpcr is expected in d1 so that the calculated result */
| can be properly sized and rounded. Also, if the incoming fpcr
| has enabled any exceptions, the exception will be taken on the
| final fmovem in this template.
|
| Customizations:
| 1. Remove the moveml at the entry and exit of
| each routine if your compiler treats those
| registers as scratch.
/* | 2. Likewise, don't save FP0/FP1 if they are scratch */
| registers.
| 3. Delete handling of the fpsr if you only care about
| the result.
| 4. Some (most?) C compilers convert all float arguments
| to double, and provide no support at all for extended
| precision so remove the __l_fcoshs and __l_fcoshx entry points.
| 5. Move the result to d0/d1 if the compiler is that old.
|
| THIS IS UNPUBLISHED PROPRIETARY SOURCE CODE OF MOTOROLA
| The copyright notice above does not evidence any
| actual or intended publication of such source code.
| xref __l_tag
| xref __l_scosh
| xref __l_ld_pone
| xref __l_ld_pinf
| xref __l_mon_nan
| xref __l_scoshd
.globl __l_fcoshs
__l_fcoshs:
link a6,#-LOCAL_SIZE
moveml d0-d1/a0-a1,a6@(USER_DA)
fmovemx fp0-fp3,a6@(USER_FP0)
fmovel fpsr,a6@(USER_FPSR)
fmovel fpcr,a6@(USER_FPCR)
fmovel fpcr,d1 /* | user's rounding mode/precision */
fmovel #0,fpcr | force rounding mode/prec to extended,rn
|
| copy, convert and tag input argument
|
fmoves a6@(8),fp0
fmovex fp0,a6@(ETEMP)
lea a6@(ETEMP),a0
bsrl __l_tag
moveb d0,a6@(STAG)
tstb d0
jne L_1062
bsrl __l_scosh | normalized (regular) number
jra L_1066
L_1062:
cmpb #0x20,d0 | zero?
jne L_1063
bsrl __l_ld_pone
jra L_1066
L_1063:
cmpb #0x40,d0 | infinity?
jne L_1064
bsrl __l_ld_pinf
jra L_1066
L_1064:
cmpb #0x60,d0 | NaN?
jne L_1065
bsrl __l_mon_nan
jra L_1066
L_1065:
bsrl __l_scoshd | assuming a denorm...
L_1066:
fmovel fpsr,d0 | update status register
orb a6@(USER_FPSR+3),d0 | add previously accrued exceptions
fmovel d0,fpsr
|
| Result is now in FP0
|
moveml a6@(USER_DA),d0-d1/a0-a1
fmovemx a6@(USER_FP1),fp1-fp3 | note: fp0 not restored
fmovel a6@(USER_FPCR),fpcr | fpcr restored
unlk a6
rts
.globl __l_fcoshd
__l_fcoshd:
link a6,#-LOCAL_SIZE
moveml d0-d1/a0-a1,a6@(USER_DA)
fmovemx fp0-fp3,a6@(USER_FP0)
fmovel fpsr,a6@(USER_FPSR)
fmovel fpcr,a6@(USER_FPCR)
fmovel fpcr,d1 /* | user's rounding mode/precision */
fmovel #0,fpcr | force rounding mode/prec to extended,rn
|
| copy, convert and tag input argument
|
fmoved a6@(8),fp0
fmovex fp0,a6@(ETEMP)
lea a6@(ETEMP),a0
bsrl __l_tag
moveb d0,a6@(STAG)
tstb d0
jne L_1067
bsrl __l_scosh | normalized (regular) number
jra L_106B
L_1067:
cmpb #0x20,d0 | zero?
jne L_1068
bsrl __l_ld_pone
jra L_106B
L_1068:
cmpb #0x40,d0 | infinity?
jne L_1069
bsrl __l_ld_pinf
jra L_106B
L_1069:
cmpb #0x60,d0 | NaN?
jne L_106A
bsrl __l_mon_nan
jra L_106B
L_106A:
bsrl __l_scoshd | assuming a denorm...
L_106B:
fmovel fpsr,d0 | update status register
orb a6@(USER_FPSR+3),d0 | add previously accrued exceptions
fmovel d0,fpsr
|
| Result is now in FP0
|
moveml a6@(USER_DA),d0-d1/a0-a1
fmovemx a6@(USER_FP1),fp1-fp3 | note: fp0 not restored
fmovel a6@(USER_FPCR),fpcr | fpcr restored
unlk a6
rts
.globl __l_fcoshx
__l_fcoshx:
link a6,#-LOCAL_SIZE
moveml d0-d1/a0-a1,a6@(USER_DA)
fmovemx fp0-fp3,a6@(USER_FP0)
fmovel fpsr,a6@(USER_FPSR)
fmovel fpcr,a6@(USER_FPCR)
fmovel fpcr,d1 /* | user's rounding mode/precision */
fmovel #0,fpcr | force rounding mode/prec to extended,rn
|
| copy, convert and tag input argument
|
fmovex a6@(8),fp0
fmovex fp0,a6@(ETEMP)
lea a6@(ETEMP),a0
bsrl __l_tag
moveb d0,a6@(STAG)
tstb d0
jne L_106C
bsrl __l_scosh | normalized (regular) number
jra L_106G
L_106C:
cmpb #0x20,d0 | zero?
jne L_106D
bsrl __l_ld_pone
jra L_106G
L_106D:
cmpb #0x40,d0 | infinity?
jne L_106E
bsrl __l_ld_pinf
jra L_106G
L_106E:
cmpb #0x60,d0 | NaN?
jne L_106F
bsrl __l_mon_nan
jra L_106G
L_106F:
bsrl __l_scoshd | assuming a denorm...
L_106G:
fmovel fpsr,d0 | update status register
orb a6@(USER_FPSR+3),d0 | add previously accrued exceptions
fmovel d0,fpsr
|
| Result is now in FP0
|
moveml a6@(USER_DA),d0-d1/a0-a1
fmovemx a6@(USER_FP1),fp1-fp3 | note: fp0 not restored
fmovel a6@(USER_FPCR),fpcr | fpcr restored
unlk a6
rts
|
| MONADIC.GEN 1.3 4/30/91
|
| MONADIC.GEN --- generic MONADIC template
|
| This version saves all registers that will be used by the emulation
| routines and restores all but FP0 on exit. The FPSR is
| updated to reflect the result of the operation. Return value
| is placed in FP0 for single, double and extended results.
|
| The package subroutines expect the incoming fpcr to be zeroed
| since they need extended precision to work properly. The
/* | 'final' fpcr is expected in d1 so that the calculated result */
| can be properly sized and rounded. Also, if the incoming fpcr
| has enabled any exceptions, the exception will be taken on the
| final fmovem in this template.
|
| Customizations:
| 1. Remove the moveml at the entry and exit of
| each routine if your compiler treats those
| registers as scratch.
/* | 2. Likewise, don't save FP0/FP1 if they are scratch */
| registers.
| 3. Delete handling of the fpsr if you only care about
| the result.
| 4. Some (most?) C compilers convert all float arguments
| to double, and provide no support at all for extended
| precision so remove the __l_fetoxs and __l_fetoxx entry points.
| 5. Move the result to d0/d1 if the compiler is that old.
|
| THIS IS UNPUBLISHED PROPRIETARY SOURCE CODE OF MOTOROLA
| The copyright notice above does not evidence any
| actual or intended publication of such source code.
| xref __l_tag
| xref __l_setox
| xref __l_ld_pone
| xref __l_szr_inf
| xref __l_mon_nan
| xref __l_setoxd
.globl __l_fetoxs
__l_fetoxs:
link a6,#-LOCAL_SIZE
moveml d0-d1/a0-a1,a6@(USER_DA)
fmovemx fp0-fp3,a6@(USER_FP0)
fmovel fpsr,a6@(USER_FPSR)
fmovel fpcr,a6@(USER_FPCR)
fmovel fpcr,d1 /* | user's rounding mode/precision */
fmovel #0,fpcr | force rounding mode/prec to extended,rn
|
| copy, convert and tag input argument
|
fmoves a6@(8),fp0
fmovex fp0,a6@(ETEMP)
lea a6@(ETEMP),a0
bsrl __l_tag
moveb d0,a6@(STAG)
tstb d0
jne L_1072
bsrl __l_setox | normalized (regular) number
jra L_1076
L_1072:
cmpb #0x20,d0 | zero?
jne L_1073
bsrl __l_ld_pone
jra L_1076
L_1073:
cmpb #0x40,d0 | infinity?
jne L_1074
bsrl __l_szr_inf
jra L_1076
L_1074:
cmpb #0x60,d0 | NaN?
jne L_1075
bsrl __l_mon_nan
jra L_1076
L_1075:
bsrl __l_setoxd | assuming a denorm...
L_1076:
fmovel fpsr,d0 | update status register
orb a6@(USER_FPSR+3),d0 | add previously accrued exceptions
fmovel d0,fpsr
|
| Result is now in FP0
|
moveml a6@(USER_DA),d0-d1/a0-a1
fmovemx a6@(USER_FP1),fp1-fp3 | note: fp0 not restored
fmovel a6@(USER_FPCR),fpcr | fpcr restored
unlk a6
rts
.globl __l_fetoxd
__l_fetoxd:
link a6,#-LOCAL_SIZE
moveml d0-d1/a0-a1,a6@(USER_DA)
fmovemx fp0-fp3,a6@(USER_FP0)
fmovel fpsr,a6@(USER_FPSR)
fmovel fpcr,a6@(USER_FPCR)
fmovel fpcr,d1 /* | user's rounding mode/precision */
fmovel #0,fpcr | force rounding mode/prec to extended,rn
|
| copy, convert and tag input argument
|
fmoved a6@(8),fp0
fmovex fp0,a6@(ETEMP)
lea a6@(ETEMP),a0
bsrl __l_tag
moveb d0,a6@(STAG)
tstb d0
jne L_1077
bsrl __l_setox | normalized (regular) number
jra L_107B
L_1077:
cmpb #0x20,d0 | zero?
jne L_1078
bsrl __l_ld_pone
jra L_107B
L_1078:
cmpb #0x40,d0 | infinity?
jne L_1079
bsrl __l_szr_inf
jra L_107B
L_1079:
cmpb #0x60,d0 | NaN?
jne L_107A
bsrl __l_mon_nan
jra L_107B
L_107A:
bsrl __l_setoxd | assuming a denorm...
L_107B:
fmovel fpsr,d0 | update status register
orb a6@(USER_FPSR+3),d0 | add previously accrued exceptions
fmovel d0,fpsr
|
| Result is now in FP0
|
moveml a6@(USER_DA),d0-d1/a0-a1
fmovemx a6@(USER_FP1),fp1-fp3 | note: fp0 not restored
fmovel a6@(USER_FPCR),fpcr | fpcr restored
unlk a6
rts
.globl __l_fetoxx
__l_fetoxx:
link a6,#-LOCAL_SIZE
moveml d0-d1/a0-a1,a6@(USER_DA)
fmovemx fp0-fp3,a6@(USER_FP0)
fmovel fpsr,a6@(USER_FPSR)
fmovel fpcr,a6@(USER_FPCR)
fmovel fpcr,d1 /* | user's rounding mode/precision */
fmovel #0,fpcr | force rounding mode/prec to extended,rn
|
| copy, convert and tag input argument
|
fmovex a6@(8),fp0
fmovex fp0,a6@(ETEMP)
lea a6@(ETEMP),a0
bsrl __l_tag
moveb d0,a6@(STAG)
tstb d0
jne L_107C
bsrl __l_setox | normalized (regular) number
jra L_107G
L_107C:
cmpb #0x20,d0 | zero?
jne L_107D
bsrl __l_ld_pone
jra L_107G
L_107D:
cmpb #0x40,d0 | infinity?
jne L_107E
bsrl __l_szr_inf
jra L_107G
L_107E:
cmpb #0x60,d0 | NaN?
jne L_107F
bsrl __l_mon_nan
jra L_107G
L_107F:
bsrl __l_setoxd | assuming a denorm...
L_107G:
fmovel fpsr,d0 | update status register
orb a6@(USER_FPSR+3),d0 | add previously accrued exceptions
fmovel d0,fpsr
|
| Result is now in FP0
|
moveml a6@(USER_DA),d0-d1/a0-a1
fmovemx a6@(USER_FP1),fp1-fp3 | note: fp0 not restored
fmovel a6@(USER_FPCR),fpcr | fpcr restored
unlk a6
rts
|
| MONADIC.GEN 1.3 4/30/91
|
| MONADIC.GEN --- generic MONADIC template
|
| This version saves all registers that will be used by the emulation
| routines and restores all but FP0 on exit. The FPSR is
| updated to reflect the result of the operation. Return value
| is placed in FP0 for single, double and extended results.
|
| The package subroutines expect the incoming fpcr to be zeroed
| since they need extended precision to work properly. The
/* | 'final' fpcr is expected in d1 so that the calculated result */
| can be properly sized and rounded. Also, if the incoming fpcr
| has enabled any exceptions, the exception will be taken on the
| final fmovem in this template.
|
| Customizations:
| 1. Remove the moveml at the entry and exit of
| each routine if your compiler treats those
| registers as scratch.
/* | 2. Likewise, don't save FP0/FP1 if they are scratch */
| registers.
| 3. Delete handling of the fpsr if you only care about
| the result.
| 4. Some (most?) C compilers convert all float arguments
| to double, and provide no support at all for extended
| precision so remove the __l_fetoxm1s and __l_fetoxm1x entry points.
| 5. Move the result to d0/d1 if the compiler is that old.
|
| THIS IS UNPUBLISHED PROPRIETARY SOURCE CODE OF MOTOROLA
| The copyright notice above does not evidence any
| actual or intended publication of such source code.
| xref __l_tag
| xref __l_setoxm1
| xref __l_szero
| xref __l_setoxm1i
| xref __l_mon_nan
| xref __l_setoxm1d
.globl __l_fetoxm1s
__l_fetoxm1s:
link a6,#-LOCAL_SIZE
moveml d0-d1/a0-a1,a6@(USER_DA)
fmovemx fp0-fp3,a6@(USER_FP0)
fmovel fpsr,a6@(USER_FPSR)
fmovel fpcr,a6@(USER_FPCR)
fmovel fpcr,d1 /* | user's rounding mode/precision */
fmovel #0,fpcr | force rounding mode/prec to extended,rn
|
| copy, convert and tag input argument
|
fmoves a6@(8),fp0
fmovex fp0,a6@(ETEMP)
lea a6@(ETEMP),a0
bsrl __l_tag
moveb d0,a6@(STAG)
tstb d0
jne L_1082
bsrl __l_setoxm1 | normalized (regular) number
jra L_1086
L_1082:
cmpb #0x20,d0 | zero?
jne L_1083
bsrl __l_szero
jra L_1086
L_1083:
cmpb #0x40,d0 | infinity?
jne L_1084
bsrl __l_setoxm1i
jra L_1086
L_1084:
cmpb #0x60,d0 | NaN?
jne L_1085
bsrl __l_mon_nan
jra L_1086
L_1085:
bsrl __l_setoxm1d | assuming a denorm...
L_1086:
fmovel fpsr,d0 | update status register
orb a6@(USER_FPSR+3),d0 | add previously accrued exceptions
fmovel d0,fpsr
|
| Result is now in FP0
|
moveml a6@(USER_DA),d0-d1/a0-a1
fmovemx a6@(USER_FP1),fp1-fp3 | note: fp0 not restored
fmovel a6@(USER_FPCR),fpcr | fpcr restored
unlk a6
rts
.globl __l_fetoxm1d
__l_fetoxm1d:
link a6,#-LOCAL_SIZE
moveml d0-d1/a0-a1,a6@(USER_DA)
fmovemx fp0-fp3,a6@(USER_FP0)
fmovel fpsr,a6@(USER_FPSR)
fmovel fpcr,a6@(USER_FPCR)
fmovel fpcr,d1 /* | user's rounding mode/precision */
fmovel #0,fpcr | force rounding mode/prec to extended,rn
|
| copy, convert and tag input argument
|
fmoved a6@(8),fp0
fmovex fp0,a6@(ETEMP)
lea a6@(ETEMP),a0
bsrl __l_tag
moveb d0,a6@(STAG)
tstb d0
jne L_1087
bsrl __l_setoxm1 | normalized (regular) number
jra L_108B
L_1087:
cmpb #0x20,d0 | zero?
jne L_1088
bsrl __l_szero
jra L_108B
L_1088:
cmpb #0x40,d0 | infinity?
jne L_1089
bsrl __l_setoxm1i
jra L_108B
L_1089:
cmpb #0x60,d0 | NaN?
jne L_108A
bsrl __l_mon_nan
jra L_108B
L_108A:
bsrl __l_setoxm1d | assuming a denorm...
L_108B:
fmovel fpsr,d0 | update status register
orb a6@(USER_FPSR+3),d0 | add previously accrued exceptions
fmovel d0,fpsr
|
| Result is now in FP0
|
moveml a6@(USER_DA),d0-d1/a0-a1
fmovemx a6@(USER_FP1),fp1-fp3 | note: fp0 not restored
fmovel a6@(USER_FPCR),fpcr | fpcr restored
unlk a6
rts
.globl __l_fetoxm1x
__l_fetoxm1x:
link a6,#-LOCAL_SIZE
moveml d0-d1/a0-a1,a6@(USER_DA)
fmovemx fp0-fp3,a6@(USER_FP0)
fmovel fpsr,a6@(USER_FPSR)
fmovel fpcr,a6@(USER_FPCR)
fmovel fpcr,d1 /* | user's rounding mode/precision */
fmovel #0,fpcr | force rounding mode/prec to extended,rn
|
| copy, convert and tag input argument
|
fmovex a6@(8),fp0
fmovex fp0,a6@(ETEMP)
lea a6@(ETEMP),a0
bsrl __l_tag
moveb d0,a6@(STAG)
tstb d0
jne L_108C
bsrl __l_setoxm1 | normalized (regular) number
jra L_108G
L_108C:
cmpb #0x20,d0 | zero?
jne L_108D
bsrl __l_szero
jra L_108G
L_108D:
cmpb #0x40,d0 | infinity?
jne L_108E
bsrl __l_setoxm1i
jra L_108G
L_108E:
cmpb #0x60,d0 | NaN?
jne L_108F
bsrl __l_mon_nan
jra L_108G
L_108F:
bsrl __l_setoxm1d | assuming a denorm...
L_108G:
fmovel fpsr,d0 | update status register
orb a6@(USER_FPSR+3),d0 | add previously accrued exceptions
fmovel d0,fpsr
|
| Result is now in FP0
|
moveml a6@(USER_DA),d0-d1/a0-a1
fmovemx a6@(USER_FP1),fp1-fp3 | note: fp0 not restored
fmovel a6@(USER_FPCR),fpcr | fpcr restored
unlk a6
rts
|
| MONADIC.GEN 1.3 4/30/91
|
| MONADIC.GEN --- generic MONADIC template
|
| This version saves all registers that will be used by the emulation
| routines and restores all but FP0 on exit. The FPSR is
| updated to reflect the result of the operation. Return value
| is placed in FP0 for single, double and extended results.
|
| The package subroutines expect the incoming fpcr to be zeroed
| since they need extended precision to work properly. The
/* | 'final' fpcr is expected in d1 so that the calculated result */
| can be properly sized and rounded. Also, if the incoming fpcr
| has enabled any exceptions, the exception will be taken on the
| final fmovem in this template.
|
| Customizations:
| 1. Remove the moveml at the entry and exit of
| each routine if your compiler treats those
| registers as scratch.
/* | 2. Likewise, don't save FP0/FP1 if they are scratch */
| registers.
| 3. Delete handling of the fpsr if you only care about
| the result.
| 4. Some (most?) C compilers convert all float arguments
| to double, and provide no support at all for extended
| precision so remove the __l_fgetexps and __l_fgetexpx entry points.
| 5. Move the result to d0/d1 if the compiler is that old.
|
| THIS IS UNPUBLISHED PROPRIETARY SOURCE CODE OF MOTOROLA
| The copyright notice above does not evidence any
| actual or intended publication of such source code.
| xref __l_tag
| xref __l_sgetexp
| xref __l_szero
| xref __l_t_operr
| xref __l_mon_nan
| xref __l_sgetexpd
.globl __l_fgetexps
__l_fgetexps:
link a6,#-LOCAL_SIZE
moveml d0-d1/a0-a1,a6@(USER_DA)
fmovemx fp0-fp3,a6@(USER_FP0)
fmovel fpsr,a6@(USER_FPSR)
fmovel fpcr,a6@(USER_FPCR)
fmovel fpcr,d1 /* | user's rounding mode/precision */
fmovel #0,fpcr | force rounding mode/prec to extended,rn
|
| copy, convert and tag input argument
|
fmoves a6@(8),fp0
fmovex fp0,a6@(ETEMP)
lea a6@(ETEMP),a0
bsrl __l_tag
moveb d0,a6@(STAG)
tstb d0
jne L_1092
bsrl __l_sgetexp | normalized (regular) number
jra L_1096
L_1092:
cmpb #0x20,d0 | zero?
jne L_1093
bsrl __l_szero
jra L_1096
L_1093:
cmpb #0x40,d0 | infinity?
jne L_1094
bsrl __l_t_operr
jra L_1096
L_1094:
cmpb #0x60,d0 | NaN?
jne L_1095
bsrl __l_mon_nan
jra L_1096
L_1095:
bsrl __l_sgetexpd | assuming a denorm...
L_1096:
fmovel fpsr,d0 | update status register
orb a6@(USER_FPSR+3),d0 | add previously accrued exceptions
fmovel d0,fpsr
|
| Result is now in FP0
|
moveml a6@(USER_DA),d0-d1/a0-a1
fmovemx a6@(USER_FP1),fp1-fp3 | note: fp0 not restored
fmovel a6@(USER_FPCR),fpcr | fpcr restored
unlk a6
rts
.globl __l_fgetexpd
__l_fgetexpd:
link a6,#-LOCAL_SIZE
moveml d0-d1/a0-a1,a6@(USER_DA)
fmovemx fp0-fp3,a6@(USER_FP0)
fmovel fpsr,a6@(USER_FPSR)
fmovel fpcr,a6@(USER_FPCR)
fmovel fpcr,d1 /* | user's rounding mode/precision */
fmovel #0,fpcr | force rounding mode/prec to extended,rn
|
| copy, convert and tag input argument
|
fmoved a6@(8),fp0
fmovex fp0,a6@(ETEMP)
lea a6@(ETEMP),a0
bsrl __l_tag
moveb d0,a6@(STAG)
tstb d0
jne L_1097
bsrl __l_sgetexp | normalized (regular) number
jra L_109B
L_1097:
cmpb #0x20,d0 | zero?
jne L_1098
bsrl __l_szero
jra L_109B
L_1098:
cmpb #0x40,d0 | infinity?
jne L_1099
bsrl __l_t_operr
jra L_109B
L_1099:
cmpb #0x60,d0 | NaN?
jne L_109A
bsrl __l_mon_nan
jra L_109B
L_109A:
bsrl __l_sgetexpd | assuming a denorm...
L_109B:
fmovel fpsr,d0 | update status register
orb a6@(USER_FPSR+3),d0 | add previously accrued exceptions
fmovel d0,fpsr
|
| Result is now in FP0
|
moveml a6@(USER_DA),d0-d1/a0-a1
fmovemx a6@(USER_FP1),fp1-fp3 | note: fp0 not restored
fmovel a6@(USER_FPCR),fpcr | fpcr restored
unlk a6
rts
.globl __l_fgetexpx
__l_fgetexpx:
link a6,#-LOCAL_SIZE
moveml d0-d1/a0-a1,a6@(USER_DA)
fmovemx fp0-fp3,a6@(USER_FP0)
fmovel fpsr,a6@(USER_FPSR)
fmovel fpcr,a6@(USER_FPCR)
fmovel fpcr,d1 /* | user's rounding mode/precision */
fmovel #0,fpcr | force rounding mode/prec to extended,rn
|
| copy, convert and tag input argument
|
fmovex a6@(8),fp0
fmovex fp0,a6@(ETEMP)
lea a6@(ETEMP),a0
bsrl __l_tag
moveb d0,a6@(STAG)
tstb d0
jne L_109C
bsrl __l_sgetexp | normalized (regular) number
jra L_109G
L_109C:
cmpb #0x20,d0 | zero?
jne L_109D
bsrl __l_szero
jra L_109G
L_109D:
cmpb #0x40,d0 | infinity?
jne L_109E
bsrl __l_t_operr
jra L_109G
L_109E:
cmpb #0x60,d0 | NaN?
jne L_109F
bsrl __l_mon_nan
jra L_109G
L_109F:
bsrl __l_sgetexpd | assuming a denorm...
L_109G:
fmovel fpsr,d0 | update status register
orb a6@(USER_FPSR+3),d0 | add previously accrued exceptions
fmovel d0,fpsr
|
| Result is now in FP0
|
moveml a6@(USER_DA),d0-d1/a0-a1
fmovemx a6@(USER_FP1),fp1-fp3 | note: fp0 not restored
fmovel a6@(USER_FPCR),fpcr | fpcr restored
unlk a6
rts
|
| MONADIC.GEN 1.3 4/30/91
|
| MONADIC.GEN --- generic MONADIC template
|
| This version saves all registers that will be used by the emulation
| routines and restores all but FP0 on exit. The FPSR is
| updated to reflect the result of the operation. Return value
| is placed in FP0 for single, double and extended results.
|
| The package subroutines expect the incoming fpcr to be zeroed
| since they need extended precision to work properly. The
/* | 'final' fpcr is expected in d1 so that the calculated result */
| can be properly sized and rounded. Also, if the incoming fpcr
| has enabled any exceptions, the exception will be taken on the
| final fmovem in this template.
|
| Customizations:
| 1. Remove the moveml at the entry and exit of
| each routine if your compiler treats those
| registers as scratch.
/* | 2. Likewise, don't save FP0/FP1 if they are scratch */
| registers.
| 3. Delete handling of the fpsr if you only care about
| the result.
| 4. Some (most?) C compilers convert all float arguments
| to double, and provide no support at all for extended
| precision so remove the __l_fsins and __l_fsinx entry points.
| 5. Move the result to d0/d1 if the compiler is that old.
|
| THIS IS UNPUBLISHED PROPRIETARY SOURCE CODE OF MOTOROLA
| The copyright notice above does not evidence any
| actual or intended publication of such source code.
| xref __l_tag
| xref __l_ssin
| xref __l_szero
| xref __l_t_operr
| xref __l_mon_nan
| xref __l_ssind
.globl __l_fsins
__l_fsins:
link a6,#-LOCAL_SIZE
moveml d0-d1/a0-a1,a6@(USER_DA)
fmovemx fp0-fp3,a6@(USER_FP0)
fmovel fpsr,a6@(USER_FPSR)
fmovel fpcr,a6@(USER_FPCR)
fmovel fpcr,d1 /* | user's rounding mode/precision */
fmovel #0,fpcr | force rounding mode/prec to extended,rn
|
| copy, convert and tag input argument
|
fmoves a6@(8),fp0
fmovex fp0,a6@(ETEMP)
lea a6@(ETEMP),a0
bsrl __l_tag
moveb d0,a6@(STAG)
tstb d0
jne L_1102
bsrl __l_ssin | normalized (regular) number
jra L_1106
L_1102:
cmpb #0x20,d0 | zero?
jne L_1103
bsrl __l_szero
jra L_1106
L_1103:
cmpb #0x40,d0 | infinity?
jne L_1104
bsrl __l_t_operr
jra L_1106
L_1104:
cmpb #0x60,d0 | NaN?
jne L_1105
bsrl __l_mon_nan
jra L_1106
L_1105:
bsrl __l_ssind | assuming a denorm...
L_1106:
fmovel fpsr,d0 | update status register
orb a6@(USER_FPSR+3),d0 | add previously accrued exceptions
fmovel d0,fpsr
|
| Result is now in FP0
|
moveml a6@(USER_DA),d0-d1/a0-a1
fmovemx a6@(USER_FP1),fp1-fp3 | note: fp0 not restored
fmovel a6@(USER_FPCR),fpcr | fpcr restored
unlk a6
rts
.globl __l_fsind
__l_fsind:
link a6,#-LOCAL_SIZE
moveml d0-d1/a0-a1,a6@(USER_DA)
fmovemx fp0-fp3,a6@(USER_FP0)
fmovel fpsr,a6@(USER_FPSR)
fmovel fpcr,a6@(USER_FPCR)
fmovel fpcr,d1 /* | user's rounding mode/precision */
fmovel #0,fpcr | force rounding mode/prec to extended,rn
|
| copy, convert and tag input argument
|
fmoved a6@(8),fp0
fmovex fp0,a6@(ETEMP)
lea a6@(ETEMP),a0
bsrl __l_tag
moveb d0,a6@(STAG)
tstb d0
jne L_1107
bsrl __l_ssin | normalized (regular) number
jra L_110B
L_1107:
cmpb #0x20,d0 | zero?
jne L_1108
bsrl __l_szero
jra L_110B
L_1108:
cmpb #0x40,d0 | infinity?
jne L_1109
bsrl __l_t_operr
jra L_110B
L_1109:
cmpb #0x60,d0 | NaN?
jne L_110A
bsrl __l_mon_nan
jra L_110B
L_110A:
bsrl __l_ssind | assuming a denorm...
L_110B:
fmovel fpsr,d0 | update status register
orb a6@(USER_FPSR+3),d0 | add previously accrued exceptions
fmovel d0,fpsr
|
| Result is now in FP0
|
moveml a6@(USER_DA),d0-d1/a0-a1
fmovemx a6@(USER_FP1),fp1-fp3 | note: fp0 not restored
fmovel a6@(USER_FPCR),fpcr | fpcr restored
unlk a6
rts
.globl __l_fsinx
__l_fsinx:
link a6,#-LOCAL_SIZE
moveml d0-d1/a0-a1,a6@(USER_DA)
fmovemx fp0-fp3,a6@(USER_FP0)
fmovel fpsr,a6@(USER_FPSR)
fmovel fpcr,a6@(USER_FPCR)
fmovel fpcr,d1 /* | user's rounding mode/precision */
fmovel #0,fpcr | force rounding mode/prec to extended,rn
|
| copy, convert and tag input argument
|
fmovex a6@(8),fp0
fmovex fp0,a6@(ETEMP)
lea a6@(ETEMP),a0
bsrl __l_tag
moveb d0,a6@(STAG)
tstb d0
jne L_110C
bsrl __l_ssin | normalized (regular) number
jra L_110G
L_110C:
cmpb #0x20,d0 | zero?
jne L_110D
bsrl __l_szero
jra L_110G
L_110D:
cmpb #0x40,d0 | infinity?
jne L_110E
bsrl __l_t_operr
jra L_110G
L_110E:
cmpb #0x60,d0 | NaN?
jne L_110F
bsrl __l_mon_nan
jra L_110G
L_110F:
bsrl __l_ssind | assuming a denorm...
L_110G:
fmovel fpsr,d0 | update status register
orb a6@(USER_FPSR+3),d0 | add previously accrued exceptions
fmovel d0,fpsr
|
| Result is now in FP0
|
moveml a6@(USER_DA),d0-d1/a0-a1
fmovemx a6@(USER_FP1),fp1-fp3 | note: fp0 not restored
fmovel a6@(USER_FPCR),fpcr | fpcr restored
unlk a6
rts
|
| MONADIC.GEN 1.3 4/30/91
|
| MONADIC.GEN --- generic MONADIC template
|
| This version saves all registers that will be used by the emulation
| routines and restores all but FP0 on exit. The FPSR is
| updated to reflect the result of the operation. Return value
| is placed in FP0 for single, double and extended results.
|
| The package subroutines expect the incoming fpcr to be zeroed
| since they need extended precision to work properly. The
/* | 'final' fpcr is expected in d1 so that the calculated result */
| can be properly sized and rounded. Also, if the incoming fpcr
| has enabled any exceptions, the exception will be taken on the
| final fmovem in this template.
|
| Customizations:
| 1. Remove the moveml at the entry and exit of
| each routine if your compiler treats those
| registers as scratch.
/* | 2. Likewise, don't save FP0/FP1 if they are scratch */
| registers.
| 3. Delete handling of the fpsr if you only care about
| the result.
| 4. Some (most?) C compilers convert all float arguments
| to double, and provide no support at all for extended
| precision so remove the __l_fsinhs and __l_fsinhx entry points.
| 5. Move the result to d0/d1 if the compiler is that old.
|
| THIS IS UNPUBLISHED PROPRIETARY SOURCE CODE OF MOTOROLA
| The copyright notice above does not evidence any
| actual or intended publication of such source code.
| xref __l_tag
| xref __l_ssinh
| xref __l_szero
| xref __l_sinf
| xref __l_mon_nan
| xref __l_ssinhd
.globl __l_fsinhs
__l_fsinhs:
link a6,#-LOCAL_SIZE
moveml d0-d1/a0-a1,a6@(USER_DA)
fmovemx fp0-fp3,a6@(USER_FP0)
fmovel fpsr,a6@(USER_FPSR)
fmovel fpcr,a6@(USER_FPCR)
fmovel fpcr,d1 /* | user's rounding mode/precision */
fmovel #0,fpcr | force rounding mode/prec to extended,rn
|
| copy, convert and tag input argument
|
fmoves a6@(8),fp0
fmovex fp0,a6@(ETEMP)
lea a6@(ETEMP),a0
bsrl __l_tag
moveb d0,a6@(STAG)
tstb d0
jne L_1112
bsrl __l_ssinh | normalized (regular) number
jra L_1116
L_1112:
cmpb #0x20,d0 | zero?
jne L_1113
bsrl __l_szero
jra L_1116
L_1113:
cmpb #0x40,d0 | infinity?
jne L_1114
bsrl __l_sinf
jra L_1116
L_1114:
cmpb #0x60,d0 | NaN?
jne L_1115
bsrl __l_mon_nan
jra L_1116
L_1115:
bsrl __l_ssinhd | assuming a denorm...
L_1116:
fmovel fpsr,d0 | update status register
orb a6@(USER_FPSR+3),d0 | add previously accrued exceptions
fmovel d0,fpsr
|
| Result is now in FP0
|
moveml a6@(USER_DA),d0-d1/a0-a1
fmovemx a6@(USER_FP1),fp1-fp3 | note: fp0 not restored
fmovel a6@(USER_FPCR),fpcr | fpcr restored
unlk a6
rts
.globl __l_fsinhd
__l_fsinhd:
link a6,#-LOCAL_SIZE
moveml d0-d1/a0-a1,a6@(USER_DA)
fmovemx fp0-fp3,a6@(USER_FP0)
fmovel fpsr,a6@(USER_FPSR)
fmovel fpcr,a6@(USER_FPCR)
fmovel fpcr,d1 /* | user's rounding mode/precision */
fmovel #0,fpcr | force rounding mode/prec to extended,rn
|
| copy, convert and tag input argument
|
fmoved a6@(8),fp0
fmovex fp0,a6@(ETEMP)
lea a6@(ETEMP),a0
bsrl __l_tag
moveb d0,a6@(STAG)
tstb d0
jne L_1117
bsrl __l_ssinh | normalized (regular) number
jra L_111B
L_1117:
cmpb #0x20,d0 | zero?
jne L_1118
bsrl __l_szero
jra L_111B
L_1118:
cmpb #0x40,d0 | infinity?
jne L_1119
bsrl __l_sinf
jra L_111B
L_1119:
cmpb #0x60,d0 | NaN?
jne L_111A
bsrl __l_mon_nan
jra L_111B
L_111A:
bsrl __l_ssinhd | assuming a denorm...
L_111B:
fmovel fpsr,d0 | update status register
orb a6@(USER_FPSR+3),d0 | add previously accrued exceptions
fmovel d0,fpsr
|
| Result is now in FP0
|
moveml a6@(USER_DA),d0-d1/a0-a1
fmovemx a6@(USER_FP1),fp1-fp3 | note: fp0 not restored
fmovel a6@(USER_FPCR),fpcr | fpcr restored
unlk a6
rts
.globl __l_fsinhx
__l_fsinhx:
link a6,#-LOCAL_SIZE
moveml d0-d1/a0-a1,a6@(USER_DA)
fmovemx fp0-fp3,a6@(USER_FP0)
fmovel fpsr,a6@(USER_FPSR)
fmovel fpcr,a6@(USER_FPCR)
fmovel fpcr,d1 /* | user's rounding mode/precision */
fmovel #0,fpcr | force rounding mode/prec to extended,rn
|
| copy, convert and tag input argument
|
fmovex a6@(8),fp0
fmovex fp0,a6@(ETEMP)
lea a6@(ETEMP),a0
bsrl __l_tag
moveb d0,a6@(STAG)
tstb d0
jne L_111C
bsrl __l_ssinh | normalized (regular) number
jra L_111G
L_111C:
cmpb #0x20,d0 | zero?
jne L_111D
bsrl __l_szero
jra L_111G
L_111D:
cmpb #0x40,d0 | infinity?
jne L_111E
bsrl __l_sinf
jra L_111G
L_111E:
cmpb #0x60,d0 | NaN?
jne L_111F
bsrl __l_mon_nan
jra L_111G
L_111F:
bsrl __l_ssinhd | assuming a denorm...
L_111G:
fmovel fpsr,d0 | update status register
orb a6@(USER_FPSR+3),d0 | add previously accrued exceptions
fmovel d0,fpsr
|
| Result is now in FP0
|
moveml a6@(USER_DA),d0-d1/a0-a1
fmovemx a6@(USER_FP1),fp1-fp3 | note: fp0 not restored
fmovel a6@(USER_FPCR),fpcr | fpcr restored
unlk a6
rts
|
| MONADIC.GEN 1.3 4/30/91
|
| MONADIC.GEN --- generic MONADIC template
|
| This version saves all registers that will be used by the emulation
| routines and restores all but FP0 on exit. The FPSR is
| updated to reflect the result of the operation. Return value
| is placed in FP0 for single, double and extended results.
|
| The package subroutines expect the incoming fpcr to be zeroed
| since they need extended precision to work properly. The
/* | 'final' fpcr is expected in d1 so that the calculated result */
| can be properly sized and rounded. Also, if the incoming fpcr
| has enabled any exceptions, the exception will be taken on the
| final fmovem in this template.
|
| Customizations:
| 1. Remove the moveml at the entry and exit of
| each routine if your compiler treats those
| registers as scratch.
/* | 2. Likewise, don't save FP0/FP1 if they are scratch */
| registers.
| 3. Delete handling of the fpsr if you only care about
| the result.
| 4. Some (most?) C compilers convert all float arguments
| to double, and provide no support at all for extended
| precision so remove the __l_ftans and __l_ftanx entry points.
| 5. Move the result to d0/d1 if the compiler is that old.
|
| THIS IS UNPUBLISHED PROPRIETARY SOURCE CODE OF MOTOROLA
| The copyright notice above does not evidence any
| actual or intended publication of such source code.
| xref __l_tag
| xref __l_stan
| xref __l_szero
| xref __l_t_operr
| xref __l_mon_nan
| xref __l_stand
.globl __l_ftans
__l_ftans:
link a6,#-LOCAL_SIZE
moveml d0-d1/a0-a1,a6@(USER_DA)
fmovemx fp0-fp3,a6@(USER_FP0)
fmovel fpsr,a6@(USER_FPSR)
fmovel fpcr,a6@(USER_FPCR)
fmovel fpcr,d1 /* | user's rounding mode/precision */
fmovel #0,fpcr | force rounding mode/prec to extended,rn
|
| copy, convert and tag input argument
|
fmoves a6@(8),fp0
fmovex fp0,a6@(ETEMP)
lea a6@(ETEMP),a0
bsrl __l_tag
moveb d0,a6@(STAG)
tstb d0
jne L_1122
bsrl __l_stan | normalized (regular) number
jra L_1126
L_1122:
cmpb #0x20,d0 | zero?
jne L_1123
bsrl __l_szero
jra L_1126
L_1123:
cmpb #0x40,d0 | infinity?
jne L_1124
bsrl __l_t_operr
jra L_1126
L_1124:
cmpb #0x60,d0 | NaN?
jne L_1125
bsrl __l_mon_nan
jra L_1126
L_1125:
bsrl __l_stand | assuming a denorm...
L_1126:
fmovel fpsr,d0 | update status register
orb a6@(USER_FPSR+3),d0 | add previously accrued exceptions
fmovel d0,fpsr
|
| Result is now in FP0
|
moveml a6@(USER_DA),d0-d1/a0-a1
fmovemx a6@(USER_FP1),fp1-fp3 | note: fp0 not restored
fmovel a6@(USER_FPCR),fpcr | fpcr restored
unlk a6
rts
.globl __l_ftand
__l_ftand:
link a6,#-LOCAL_SIZE
moveml d0-d1/a0-a1,a6@(USER_DA)
fmovemx fp0-fp3,a6@(USER_FP0)
fmovel fpsr,a6@(USER_FPSR)
fmovel fpcr,a6@(USER_FPCR)
fmovel fpcr,d1 /* | user's rounding mode/precision */
fmovel #0,fpcr | force rounding mode/prec to extended,rn
|
| copy, convert and tag input argument
|
fmoved a6@(8),fp0
fmovex fp0,a6@(ETEMP)
lea a6@(ETEMP),a0
bsrl __l_tag
moveb d0,a6@(STAG)
tstb d0
jne L_1127
bsrl __l_stan | normalized (regular) number
jra L_112B
L_1127:
cmpb #0x20,d0 | zero?
jne L_1128
bsrl __l_szero
jra L_112B
L_1128:
cmpb #0x40,d0 | infinity?
jne L_1129
bsrl __l_t_operr
jra L_112B
L_1129:
cmpb #0x60,d0 | NaN?
jne L_112A
bsrl __l_mon_nan
jra L_112B
L_112A:
bsrl __l_stand | assuming a denorm...
L_112B:
fmovel fpsr,d0 | update status register
orb a6@(USER_FPSR+3),d0 | add previously accrued exceptions
fmovel d0,fpsr
|
| Result is now in FP0
|
moveml a6@(USER_DA),d0-d1/a0-a1
fmovemx a6@(USER_FP1),fp1-fp3 | note: fp0 not restored
fmovel a6@(USER_FPCR),fpcr | fpcr restored
unlk a6
rts
.globl __l_ftanx
__l_ftanx:
link a6,#-LOCAL_SIZE
moveml d0-d1/a0-a1,a6@(USER_DA)
fmovemx fp0-fp3,a6@(USER_FP0)
fmovel fpsr,a6@(USER_FPSR)
fmovel fpcr,a6@(USER_FPCR)
fmovel fpcr,d1 /* | user's rounding mode/precision */
fmovel #0,fpcr | force rounding mode/prec to extended,rn
|
| copy, convert and tag input argument
|
fmovex a6@(8),fp0
fmovex fp0,a6@(ETEMP)
lea a6@(ETEMP),a0
bsrl __l_tag
moveb d0,a6@(STAG)
tstb d0
jne L_112C
bsrl __l_stan | normalized (regular) number
jra L_112G
L_112C:
cmpb #0x20,d0 | zero?
jne L_112D
bsrl __l_szero
jra L_112G
L_112D:
cmpb #0x40,d0 | infinity?
jne L_112E
bsrl __l_t_operr
jra L_112G
L_112E:
cmpb #0x60,d0 | NaN?
jne L_112F
bsrl __l_mon_nan
jra L_112G
L_112F:
bsrl __l_stand | assuming a denorm...
L_112G:
fmovel fpsr,d0 | update status register
orb a6@(USER_FPSR+3),d0 | add previously accrued exceptions
fmovel d0,fpsr
|
| Result is now in FP0
|
moveml a6@(USER_DA),d0-d1/a0-a1
fmovemx a6@(USER_FP1),fp1-fp3 | note: fp0 not restored
fmovel a6@(USER_FPCR),fpcr | fpcr restored
unlk a6
rts
|
| MONADIC.GEN 1.3 4/30/91
|
| MONADIC.GEN --- generic MONADIC template
|
| This version saves all registers that will be used by the emulation
| routines and restores all but FP0 on exit. The FPSR is
| updated to reflect the result of the operation. Return value
| is placed in FP0 for single, double and extended results.
|
| The package subroutines expect the incoming fpcr to be zeroed
| since they need extended precision to work properly. The
/* | 'final' fpcr is expected in d1 so that the calculated result */
| can be properly sized and rounded. Also, if the incoming fpcr
| has enabled any exceptions, the exception will be taken on the
| final fmovem in this template.
|
| Customizations:
| 1. Remove the moveml at the entry and exit of
| each routine if your compiler treats those
| registers as scratch.
/* | 2. Likewise, don't save FP0/FP1 if they are scratch */
| registers.
| 3. Delete handling of the fpsr if you only care about
| the result.
| 4. Some (most?) C compilers convert all float arguments
| to double, and provide no support at all for extended
| precision so remove the __l_ftanhs and __l_ftanhx entry points.
| 5. Move the result to d0/d1 if the compiler is that old.
|
| THIS IS UNPUBLISHED PROPRIETARY SOURCE CODE OF MOTOROLA
| The copyright notice above does not evidence any
| actual or intended publication of such source code.
| xref __l_tag
| xref __l_stanh
| xref __l_szero
| xref __l_sone
| xref __l_mon_nan
| xref __l_stanhd
.globl __l_ftanhs
__l_ftanhs:
link a6,#-LOCAL_SIZE
moveml d0-d1/a0-a1,a6@(USER_DA)
fmovemx fp0-fp3,a6@(USER_FP0)
fmovel fpsr,a6@(USER_FPSR)
fmovel fpcr,a6@(USER_FPCR)
fmovel fpcr,d1 /* | user's rounding mode/precision */
fmovel #0,fpcr | force rounding mode/prec to extended,rn
|
| copy, convert and tag input argument
|
fmoves a6@(8),fp0
fmovex fp0,a6@(ETEMP)
lea a6@(ETEMP),a0
bsrl __l_tag
moveb d0,a6@(STAG)
tstb d0
jne L_1132
bsrl __l_stanh | normalized (regular) number
jra L_1136
L_1132:
cmpb #0x20,d0 | zero?
jne L_1133
bsrl __l_szero
jra L_1136
L_1133:
cmpb #0x40,d0 | infinity?
jne L_1134
bsrl __l_sone
jra L_1136
L_1134:
cmpb #0x60,d0 | NaN?
jne L_1135
bsrl __l_mon_nan
jra L_1136
L_1135:
bsrl __l_stanhd | assuming a denorm...
L_1136:
fmovel fpsr,d0 | update status register
orb a6@(USER_FPSR+3),d0 | add previously accrued exceptions
fmovel d0,fpsr
|
| Result is now in FP0
|
moveml a6@(USER_DA),d0-d1/a0-a1
fmovemx a6@(USER_FP1),fp1-fp3 | note: fp0 not restored
fmovel a6@(USER_FPCR),fpcr | fpcr restored
unlk a6
rts
.globl __l_ftanhd
__l_ftanhd:
link a6,#-LOCAL_SIZE
moveml d0-d1/a0-a1,a6@(USER_DA)
fmovemx fp0-fp3,a6@(USER_FP0)
fmovel fpsr,a6@(USER_FPSR)
fmovel fpcr,a6@(USER_FPCR)
fmovel fpcr,d1 /* | user's rounding mode/precision */
fmovel #0,fpcr | force rounding mode/prec to extended,rn
|
| copy, convert and tag input argument
|
fmoved a6@(8),fp0
fmovex fp0,a6@(ETEMP)
lea a6@(ETEMP),a0
bsrl __l_tag
moveb d0,a6@(STAG)
tstb d0
jne L_1137
bsrl __l_stanh | normalized (regular) number
jra L_113B
L_1137:
cmpb #0x20,d0 | zero?
jne L_1138
bsrl __l_szero
jra L_113B
L_1138:
cmpb #0x40,d0 | infinity?
jne L_1139
bsrl __l_sone
jra L_113B
L_1139:
cmpb #0x60,d0 | NaN?
jne L_113A
bsrl __l_mon_nan
jra L_113B
L_113A:
bsrl __l_stanhd | assuming a denorm...
L_113B:
fmovel fpsr,d0 | update status register
orb a6@(USER_FPSR+3),d0 | add previously accrued exceptions
fmovel d0,fpsr
|
| Result is now in FP0
|
moveml a6@(USER_DA),d0-d1/a0-a1
fmovemx a6@(USER_FP1),fp1-fp3 | note: fp0 not restored
fmovel a6@(USER_FPCR),fpcr | fpcr restored
unlk a6
rts
.globl __l_ftanhx
__l_ftanhx:
link a6,#-LOCAL_SIZE
moveml d0-d1/a0-a1,a6@(USER_DA)
fmovemx fp0-fp3,a6@(USER_FP0)
fmovel fpsr,a6@(USER_FPSR)
fmovel fpcr,a6@(USER_FPCR)
fmovel fpcr,d1 /* | user's rounding mode/precision */
fmovel #0,fpcr | force rounding mode/prec to extended,rn
|
| copy, convert and tag input argument
|
fmovex a6@(8),fp0
fmovex fp0,a6@(ETEMP)
lea a6@(ETEMP),a0
bsrl __l_tag
moveb d0,a6@(STAG)
tstb d0
jne L_113C
bsrl __l_stanh | normalized (regular) number
jra L_113G
L_113C:
cmpb #0x20,d0 | zero?
jne L_113D
bsrl __l_szero
jra L_113G
L_113D:
cmpb #0x40,d0 | infinity?
jne L_113E
bsrl __l_sone
jra L_113G
L_113E:
cmpb #0x60,d0 | NaN?
jne L_113F
bsrl __l_mon_nan
jra L_113G
L_113F:
bsrl __l_stanhd | assuming a denorm...
L_113G:
fmovel fpsr,d0 | update status register
orb a6@(USER_FPSR+3),d0 | add previously accrued exceptions
fmovel d0,fpsr
|
| Result is now in FP0
|
moveml a6@(USER_DA),d0-d1/a0-a1
fmovemx a6@(USER_FP1),fp1-fp3 | note: fp0 not restored
fmovel a6@(USER_FPCR),fpcr | fpcr restored
unlk a6
rts
|
| MONADIC.GEN 1.3 4/30/91
|
| MONADIC.GEN --- generic MONADIC template
|
| This version saves all registers that will be used by the emulation
| routines and restores all but FP0 on exit. The FPSR is
| updated to reflect the result of the operation. Return value
| is placed in FP0 for single, double and extended results.
|
| The package subroutines expect the incoming fpcr to be zeroed
| since they need extended precision to work properly. The
/* | 'final' fpcr is expected in d1 so that the calculated result */
| can be properly sized and rounded. Also, if the incoming fpcr
| has enabled any exceptions, the exception will be taken on the
| final fmovem in this template.
|
| Customizations:
| 1. Remove the moveml at the entry and exit of
| each routine if your compiler treats those
| registers as scratch.
/* | 2. Likewise, don't save FP0/FP1 if they are scratch */
| registers.
| 3. Delete handling of the fpsr if you only care about
| the result.
| 4. Some (most?) C compilers convert all float arguments
| to double, and provide no support at all for extended
| precision so remove the __l_ftentoxs and __l_ftentoxx entry points.
| 5. Move the result to d0/d1 if the compiler is that old.
|
| THIS IS UNPUBLISHED PROPRIETARY SOURCE CODE OF MOTOROLA
| The copyright notice above does not evidence any
| actual or intended publication of such source code.
| xref __l_tag
| xref __l_stentox
| xref __l_ld_pone
| xref __l_szr_inf
| xref __l_mon_nan
| xref __l_stentoxd
.globl __l_ftentoxs
__l_ftentoxs:
link a6,#-LOCAL_SIZE
moveml d0-d1/a0-a1,a6@(USER_DA)
fmovemx fp0-fp3,a6@(USER_FP0)
fmovel fpsr,a6@(USER_FPSR)
fmovel fpcr,a6@(USER_FPCR)
fmovel fpcr,d1 /* | user's rounding mode/precision */
fmovel #0,fpcr | force rounding mode/prec to extended,rn
|
| copy, convert and tag input argument
|
fmoves a6@(8),fp0
fmovex fp0,a6@(ETEMP)
lea a6@(ETEMP),a0
bsrl __l_tag
moveb d0,a6@(STAG)
tstb d0
jne L_1142
bsrl __l_stentox | normalized (regular) number
jra L_1146
L_1142:
cmpb #0x20,d0 | zero?
jne L_1143
bsrl __l_ld_pone
jra L_1146
L_1143:
cmpb #0x40,d0 | infinity?
jne L_1144
bsrl __l_szr_inf
jra L_1146
L_1144:
cmpb #0x60,d0 | NaN?
jne L_1145
bsrl __l_mon_nan
jra L_1146
L_1145:
bsrl __l_stentoxd | assuming a denorm...
L_1146:
fmovel fpsr,d0 | update status register
orb a6@(USER_FPSR+3),d0 | add previously accrued exceptions
fmovel d0,fpsr
|
| Result is now in FP0
|
moveml a6@(USER_DA),d0-d1/a0-a1
fmovemx a6@(USER_FP1),fp1-fp3 | note: fp0 not restored
fmovel a6@(USER_FPCR),fpcr | fpcr restored
unlk a6
rts
.globl __l_ftentoxd
__l_ftentoxd:
link a6,#-LOCAL_SIZE
moveml d0-d1/a0-a1,a6@(USER_DA)
fmovemx fp0-fp3,a6@(USER_FP0)
fmovel fpsr,a6@(USER_FPSR)
fmovel fpcr,a6@(USER_FPCR)
fmovel fpcr,d1 /* | user's rounding mode/precision */
fmovel #0,fpcr | force rounding mode/prec to extended,rn
|
| copy, convert and tag input argument
|
fmoved a6@(8),fp0
fmovex fp0,a6@(ETEMP)
lea a6@(ETEMP),a0
bsrl __l_tag
moveb d0,a6@(STAG)
tstb d0
jne L_1147
bsrl __l_stentox | normalized (regular) number
jra L_114B
L_1147:
cmpb #0x20,d0 | zero?
jne L_1148
bsrl __l_ld_pone
jra L_114B
L_1148:
cmpb #0x40,d0 | infinity?
jne L_1149
bsrl __l_szr_inf
jra L_114B
L_1149:
cmpb #0x60,d0 | NaN?
jne L_114A
bsrl __l_mon_nan
jra L_114B
L_114A:
bsrl __l_stentoxd | assuming a denorm...
L_114B:
fmovel fpsr,d0 | update status register
orb a6@(USER_FPSR+3),d0 | add previously accrued exceptions
fmovel d0,fpsr
|
| Result is now in FP0
|
moveml a6@(USER_DA),d0-d1/a0-a1
fmovemx a6@(USER_FP1),fp1-fp3 | note: fp0 not restored
fmovel a6@(USER_FPCR),fpcr | fpcr restored
unlk a6
rts
.globl __l_ftentoxx
__l_ftentoxx:
link a6,#-LOCAL_SIZE
moveml d0-d1/a0-a1,a6@(USER_DA)
fmovemx fp0-fp3,a6@(USER_FP0)
fmovel fpsr,a6@(USER_FPSR)
fmovel fpcr,a6@(USER_FPCR)
fmovel fpcr,d1 /* | user's rounding mode/precision */
fmovel #0,fpcr | force rounding mode/prec to extended,rn
|
| copy, convert and tag input argument
|
fmovex a6@(8),fp0
fmovex fp0,a6@(ETEMP)
lea a6@(ETEMP),a0
bsrl __l_tag
moveb d0,a6@(STAG)
tstb d0
jne L_114C
bsrl __l_stentox | normalized (regular) number
jra L_114G
L_114C:
cmpb #0x20,d0 | zero?
jne L_114D
bsrl __l_ld_pone
jra L_114G
L_114D:
cmpb #0x40,d0 | infinity?
jne L_114E
bsrl __l_szr_inf
jra L_114G
L_114E:
cmpb #0x60,d0 | NaN?
jne L_114F
bsrl __l_mon_nan
jra L_114G
L_114F:
bsrl __l_stentoxd | assuming a denorm...
L_114G:
fmovel fpsr,d0 | update status register
orb a6@(USER_FPSR+3),d0 | add previously accrued exceptions
fmovel d0,fpsr
|
| Result is now in FP0
|
moveml a6@(USER_DA),d0-d1/a0-a1
fmovemx a6@(USER_FP1),fp1-fp3 | note: fp0 not restored
fmovel a6@(USER_FPCR),fpcr | fpcr restored
unlk a6
rts
|
| MONADIC.GEN 1.3 4/30/91
|
| MONADIC.GEN --- generic MONADIC template
|
| This version saves all registers that will be used by the emulation
| routines and restores all but FP0 on exit. The FPSR is
| updated to reflect the result of the operation. Return value
| is placed in FP0 for single, double and extended results.
|
| The package subroutines expect the incoming fpcr to be zeroed
| since they need extended precision to work properly. The
/* | 'final' fpcr is expected in d1 so that the calculated result */
| can be properly sized and rounded. Also, if the incoming fpcr
| has enabled any exceptions, the exception will be taken on the
| final fmovem in this template.
|
| Customizations:
| 1. Remove the moveml at the entry and exit of
| each routine if your compiler treats those
| registers as scratch.
/* | 2. Likewise, don't save FP0/FP1 if they are scratch */
| registers.
| 3. Delete handling of the fpsr if you only care about
| the result.
| 4. Some (most?) C compilers convert all float arguments
| to double, and provide no support at all for extended
| precision so remove the __l_ftwotoxs and __l_ftwotoxx entry points.
| 5. Move the result to d0/d1 if the compiler is that old.
|
| THIS IS UNPUBLISHED PROPRIETARY SOURCE CODE OF MOTOROLA
| The copyright notice above does not evidence any
| actual or intended publication of such source code.
| xref __l_tag
| xref __l_stwotox
| xref __l_ld_pone
| xref __l_szr_inf
| xref __l_mon_nan
| xref __l_stwotoxd
.globl __l_ftwotoxs
__l_ftwotoxs:
link a6,#-LOCAL_SIZE
moveml d0-d1/a0-a1,a6@(USER_DA)
fmovemx fp0-fp3,a6@(USER_FP0)
fmovel fpsr,a6@(USER_FPSR)
fmovel fpcr,a6@(USER_FPCR)
fmovel fpcr,d1 /* | user's rounding mode/precision */
fmovel #0,fpcr | force rounding mode/prec to extended,rn
|
| copy, convert and tag input argument
|
fmoves a6@(8),fp0
fmovex fp0,a6@(ETEMP)
lea a6@(ETEMP),a0
bsrl __l_tag
moveb d0,a6@(STAG)
tstb d0
jne L_1152
bsrl __l_stwotox | normalized (regular) number
jra L_1156
L_1152:
cmpb #0x20,d0 | zero?
jne L_1153
bsrl __l_ld_pone
jra L_1156
L_1153:
cmpb #0x40,d0 | infinity?
jne L_1154
bsrl __l_szr_inf
jra L_1156
L_1154:
cmpb #0x60,d0 | NaN?
jne L_1155
bsrl __l_mon_nan
jra L_1156
L_1155:
bsrl __l_stwotoxd | assuming a denorm...
L_1156:
fmovel fpsr,d0 | update status register
orb a6@(USER_FPSR+3),d0 | add previously accrued exceptions
fmovel d0,fpsr
|
| Result is now in FP0
|
moveml a6@(USER_DA),d0-d1/a0-a1
fmovemx a6@(USER_FP1),fp1-fp3 | note: fp0 not restored
fmovel a6@(USER_FPCR),fpcr | fpcr restored
unlk a6
rts
.globl __l_ftwotoxd
__l_ftwotoxd:
link a6,#-LOCAL_SIZE
moveml d0-d1/a0-a1,a6@(USER_DA)
fmovemx fp0-fp3,a6@(USER_FP0)
fmovel fpsr,a6@(USER_FPSR)
fmovel fpcr,a6@(USER_FPCR)
fmovel fpcr,d1 /* | user's rounding mode/precision */
fmovel #0,fpcr | force rounding mode/prec to extended,rn
|
| copy, convert and tag input argument
|
fmoved a6@(8),fp0
fmovex fp0,a6@(ETEMP)
lea a6@(ETEMP),a0
bsrl __l_tag
moveb d0,a6@(STAG)
tstb d0
jne L_1157
bsrl __l_stwotox | normalized (regular) number
jra L_115B
L_1157:
cmpb #0x20,d0 | zero?
jne L_1158
bsrl __l_ld_pone
jra L_115B
L_1158:
cmpb #0x40,d0 | infinity?
jne L_1159
bsrl __l_szr_inf
jra L_115B
L_1159:
cmpb #0x60,d0 | NaN?
jne L_115A
bsrl __l_mon_nan
jra L_115B
L_115A:
bsrl __l_stwotoxd | assuming a denorm...
L_115B:
fmovel fpsr,d0 | update status register
orb a6@(USER_FPSR+3),d0 | add previously accrued exceptions
fmovel d0,fpsr
|
| Result is now in FP0
|
moveml a6@(USER_DA),d0-d1/a0-a1
fmovemx a6@(USER_FP1),fp1-fp3 | note: fp0 not restored
fmovel a6@(USER_FPCR),fpcr | fpcr restored
unlk a6
rts
.globl __l_ftwotoxx
__l_ftwotoxx:
link a6,#-LOCAL_SIZE
moveml d0-d1/a0-a1,a6@(USER_DA)
fmovemx fp0-fp3,a6@(USER_FP0)
fmovel fpsr,a6@(USER_FPSR)
fmovel fpcr,a6@(USER_FPCR)
fmovel fpcr,d1 /* | user's rounding mode/precision */
fmovel #0,fpcr | force rounding mode/prec to extended,rn
|
| copy, convert and tag input argument
|
fmovex a6@(8),fp0
fmovex fp0,a6@(ETEMP)
lea a6@(ETEMP),a0
bsrl __l_tag
moveb d0,a6@(STAG)
tstb d0
jne L_115C
bsrl __l_stwotox | normalized (regular) number
jra L_115G
L_115C:
cmpb #0x20,d0 | zero?
jne L_115D
bsrl __l_ld_pone
jra L_115G
L_115D:
cmpb #0x40,d0 | infinity?
jne L_115E
bsrl __l_szr_inf
jra L_115G
L_115E:
cmpb #0x60,d0 | NaN?
jne L_115F
bsrl __l_mon_nan
jra L_115G
L_115F:
bsrl __l_stwotoxd | assuming a denorm...
L_115G:
fmovel fpsr,d0 | update status register
orb a6@(USER_FPSR+3),d0 | add previously accrued exceptions
fmovel d0,fpsr
|
| Result is now in FP0
|
moveml a6@(USER_DA),d0-d1/a0-a1
fmovemx a6@(USER_FP1),fp1-fp3 | note: fp0 not restored
fmovel a6@(USER_FPCR),fpcr | fpcr restored
unlk a6
rts
|
| MONADIC.GEN 1.3 4/30/91
|
| MONADIC.GEN --- generic MONADIC template
|
| This version saves all registers that will be used by the emulation
| routines and restores all but FP0 on exit. The FPSR is
| updated to reflect the result of the operation. Return value
| is placed in FP0 for single, double and extended results.
|
| The package subroutines expect the incoming fpcr to be zeroed
| since they need extended precision to work properly. The
/* | 'final' fpcr is expected in d1 so that the calculated result */
| can be properly sized and rounded. Also, if the incoming fpcr
| has enabled any exceptions, the exception will be taken on the
| final fmovem in this template.
|
| Customizations:
| 1. Remove the moveml at the entry and exit of
| each routine if your compiler treats those
| registers as scratch.
/* | 2. Likewise, don't save FP0/FP1 if they are scratch */