| 1 |
/* Copyright (C) 2006-2025 Free Software Foundation, Inc. |
| 2 |
Contributor: Joern Rennecke <[email protected]> |
| 3 |
on behalf of Synopsys Inc. |
| 4 |
|
| 5 |
This file is part of GCC. |
| 6 |
|
| 7 |
GCC is free software; you can redistribute it and/or modify it under |
| 8 |
the terms of the GNU General Public License as published by the Free |
| 9 |
Software Foundation; either version 3, or (at your option) any later |
| 10 |
version. |
| 11 |
|
| 12 |
GCC is distributed in the hope that it will be useful, but WITHOUT ANY |
| 13 |
WARRANTY; without even the implied warranty of MERCHANTABILITY or |
| 14 |
FITNESS FOR A PARTICULAR PURPOSE. See the GNU General Public License |
| 15 |
for more details. |
| 16 |
|
| 17 |
Under Section 7 of GPL version 3, you are granted additional |
| 18 |
permissions described in the GCC Runtime Library Exception, version |
| 19 |
3.1, as published by the Free Software Foundation. |
| 20 |
|
| 21 |
You should have received a copy of the GNU General Public License and |
| 22 |
a copy of the GCC Runtime Library Exception along with this program; |
| 23 |
see the files COPYING3 and COPYING.RUNTIME respectively. If not, see |
| 24 |
<http://www.gnu.org/licenses/>. */ |
| 25 |
|
| 26 |
#include "arc-ieee-754.h" |
| 27 |
|
| 28 |
#if 0 /* DEBUG */ |
| 29 |
FUNC(__truncdfsf2) |
| 30 |
.global __truncdfsf2 |
| 31 |
.balign 4 |
| 32 |
__truncdfsf2: |
| 33 |
push_s blink |
| 34 |
push_s r0 |
| 35 |
bl.d __truncdfsf2_c |
| 36 |
push_s r1 |
| 37 |
mov_s r2,r0 |
| 38 |
pop_s r1 |
| 39 |
ld r0,[sp] |
| 40 |
bl.d __truncdfsf2_asm |
| 41 |
st r2,[sp] |
| 42 |
pop_s r1 |
| 43 |
pop_s blink |
| 44 |
cmp r0,r1 |
| 45 |
jeq_s [blink] |
| 46 |
and r12,r0,r1 |
| 47 |
bic.f 0,0x7f800000,r12 |
| 48 |
bne 0f |
| 49 |
bmsk.f 0,r0,22 |
| 50 |
bmsk.ne.f r1,r1,22 |
| 51 |
jne_s [blink] ; both NaN -> OK |
| 52 |
0: bl abort |
| 53 |
ENDFUNC(__truncdfsf2) |
| 54 |
#define __truncdfsf2 __truncdfsf2_asm |
| 55 |
#endif /* DEBUG */ |
| 56 |
|
| 57 |
.global __truncdfsf2 |
| 58 |
.balign 4 |
| 59 |
FUNC(__truncdfsf2) |
| 60 |
__truncdfsf2: |
| 61 |
lsr r2,DBL0H,20 |
| 62 |
asl_s DBL0H,DBL0H,12 |
| 63 |
sub r12,r2,0x380 |
| 64 |
bclr.f r3,r12,11 |
| 65 |
brhs r3,0xff,.Lill_exp |
| 66 |
beq_l .Ldenorm0 |
| 67 |
asl_s r12,r12,23 |
| 68 |
tst DBL0L, \ |
| 69 |
0x2fffffff /* Check if msb guard bit wants rounding up. */ |
| 70 |
lsr_s DBL0L,DBL0L,28 |
| 71 |
lsr_s DBL0H,DBL0H,8 |
| 72 |
add.ne DBL0L,DBL0L,1 |
| 73 |
add_s DBL0H,DBL0H,DBL0L |
| 74 |
lsr_s DBL0H,DBL0H |
| 75 |
btst_s r2,11 |
| 76 |
add_s r0,DBL0H,r12 |
| 77 |
j_s.d [blink] |
| 78 |
bxor.ne r0,r0,31 |
| 79 |
.balign 4 |
| 80 |
.Lill_exp: |
| 81 |
bbit1 r2,10,.Linf_nan |
| 82 |
bmsk_s r12,r12,9 |
| 83 |
rsub.f r12,r12,8+0x400-32 ; Go from 9 to 1 guard bit in MSW. */ |
| 84 |
bhs_s .Lzero |
| 85 |
lsr r3,DBL0L,21 |
| 86 |
rrc DBL0H,DBL0H ; insert leading 1 |
| 87 |
asl.f 0,DBL0L,8 ; check lower 24 guard bits |
| 88 |
add_s r3,DBL0H,r3 |
| 89 |
add.pnz r3,r3,1 ; assemble fraction with compressed guard bits. |
| 90 |
lsr r0,r3,r12 |
| 91 |
neg_s r12,r12 |
| 92 |
btst_s r0,1 |
| 93 |
asl.eq.f r3,r3,r12 |
| 94 |
add.ne r0,r0,1 |
| 95 |
btst_s r2,11 |
| 96 |
lsr_s r0,r0 |
| 97 |
j_s.d [blink] |
| 98 |
bxor.ne r0,r0,31 |
| 99 |
.Lzero: |
| 100 |
lsr_s r2,r2,11 |
| 101 |
j_s.d [blink] |
| 102 |
asl r0,r2,31 |
| 103 |
.Ldenorm0: |
| 104 |
asl_s r12,r12,20 |
| 105 |
tst DBL0L, \ |
| 106 |
0x5fffffff /* Check if msb guard bit wants rounding up. */ |
| 107 |
lsr_s DBL0L,DBL0L,29 |
| 108 |
lsr_s DBL0H,DBL0H,9 |
| 109 |
add.ne DBL0L,DBL0L,1 |
| 110 |
bset_s DBL0H,DBL0H,23 |
| 111 |
add_s DBL0H,DBL0H,DBL0L |
| 112 |
lsr_s DBL0H,DBL0H |
| 113 |
j_s.d [blink] |
| 114 |
add_l r0,DBL0H,r12 |
| 115 |
|
| 116 |
/* We would generally say that NaNs must have a non-zero high fraction part, |
| 117 |
but to allow hardware double precision floating point to interoperate |
| 118 |
with single precision software floating point, we make an exception here. |
| 119 |
The cost is to replace a tst_s DBL0H with an or.f DBL0L,DBL0L,DBL0H . |
| 120 |
As we start out unaligned, and there is an odd number of other short insns, |
| 121 |
we have a choice of letting this cost us a misalign penalty or |
| 122 |
4 more bytes (if we align the code). We choose the former here because |
| 123 |
infinity / NaN is not expected to be prevalent in time-critical code. */ |
| 124 |
.Linf_nan: |
| 125 |
or.f DBL0L,DBL0L,DBL0H |
| 126 |
mov_s r0,1 |
| 127 |
add.ne r2,r2,1 |
| 128 |
tst r2,0x7ff |
| 129 |
asl.ne r0,r0,23 |
| 130 |
btst_s r12,11 |
| 131 |
neg r0,r0 |
| 132 |
j_s.d [blink] |
| 133 |
bxor.eq r0,r0,31 |
| 134 |
ENDFUNC(__truncdfsf2) |