Retro68/gcc/libgcc/config/bfin/lib1funcs.S
Wolfgang Thaller aaf905ce07 add gcc 4.70
2012-03-28 01:13:14 +02:00

212 lines
5.0 KiB
ArmAsm

/* libgcc functions for Blackfin.
Copyright (C) 2005, 2009 Free Software Foundation, Inc.
Contributed by Analog Devices.
This file is part of GCC.
GCC is free software; you can redistribute it and/or modify
it under the terms of the GNU General Public License as published by
the Free Software Foundation; either version 3, or (at your option)
any later version.
GCC is distributed in the hope that it will be useful,
but WITHOUT ANY WARRANTY; without even the implied warranty of
MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
GNU General Public License for more details.
Under Section 7 of GPL version 3, you are granted additional
permissions described in the GCC Runtime Library Exception, version
3.1, as published by the Free Software Foundation.
You should have received a copy of the GNU General Public License and
a copy of the GCC Runtime Library Exception along with this program;
see the files COPYING3 and COPYING.RUNTIME respectively. If not, see
<http://www.gnu.org/licenses/>. */
#ifdef L_divsi3
.text
.align 2
.global ___divsi3;
.type ___divsi3, STT_FUNC;
___divsi3:
[--SP]= RETS;
[--SP] = R7;
R2 = -R0;
CC = R0 < 0;
IF CC R0 = R2;
R7 = CC;
R2 = -R1;
CC = R1 < 0;
IF CC R1 = R2;
R2 = CC;
R7 = R7 ^ R2;
CALL ___udivsi3;
CC = R7;
R1 = -R0;
IF CC R0 = R1;
R7 = [SP++];
RETS = [SP++];
RTS;
#endif
#ifdef L_modsi3
.align 2
.global ___modsi3;
.type ___modsi3, STT_FUNC;
___modsi3:
[--SP] = RETS;
[--SP] = R0;
[--SP] = R1;
CALL ___divsi3;
R2 = [SP++];
R1 = [SP++];
R2 *= R0;
R0 = R1 - R2;
RETS = [SP++];
RTS;
#endif
#ifdef L_udivsi3
.align 2
.global ___udivsi3;
.type ___udivsi3, STT_FUNC;
___udivsi3:
P0 = 32;
LSETUP (0f, 1f) LC0 = P0;
/* upper half of dividend */
R3 = 0;
0:
/* The first time round in the loop we shift in garbage, but since we
perform 33 shifts, it doesn't matter. */
R0 = ROT R0 BY 1;
R3 = ROT R3 BY 1;
R2 = R3 - R1;
CC = R3 < R1 (IU);
1:
/* Last instruction of the loop. */
IF ! CC R3 = R2;
/* Shift in the last bit. */
R0 = ROT R0 BY 1;
/* R0 is the result, R3 contains the remainder. */
R0 = ~ R0;
RTS;
#endif
#ifdef L_umodsi3
.align 2
.global ___umodsi3;
.type ___umodsi3, STT_FUNC;
___umodsi3:
[--SP] = RETS;
CALL ___udivsi3;
R0 = R3;
RETS = [SP++];
RTS;
#endif
#ifdef L_umulsi3_highpart
.align 2
.global ___umulsi3_highpart;
.type ___umulsi3_highpart, STT_FUNC;
___umulsi3_highpart:
A1 = R1.L * R0.L (FU);
A1 = A1 >> 16;
A0 = R1.H * R0.H, A1 += R1.L * R0.H (FU);
A1 += R0.L * R1.H (FU);
A1 = A1 >> 16;
A0 += A1;
R0 = A0 (FU);
RTS;
#endif
#ifdef L_smulsi3_highpart
.align 2
.global ___smulsi3_highpart;
.type ___smulsi3_highpart, STT_FUNC;
___smulsi3_highpart:
A1 = R1.L * R0.L (FU);
A1 = A1 >> 16;
A0 = R0.H * R1.H, A1 += R0.H * R1.L (IS,M);
A1 += R1.H * R0.L (IS,M);
A1 = A1 >>> 16;
R0 = (A0 += A1);
RTS;
#endif
#ifdef L_muldi3
.align 2
.global ___muldi3;
.type ___muldi3, STT_FUNC;
/*
R1:R0 * R3:R2
= R1.h:R1.l:R0.h:R0.l * R3.h:R3.l:R2.h:R2.l
[X] = (R1.h * R3.h) * 2^96
[X] + (R1.h * R3.l + R1.l * R3.h) * 2^80
[X] + (R1.h * R2.h + R1.l * R3.l + R3.h * R0.h) * 2^64
[T1] + (R1.h * R2.l + R3.h * R0.l + R1.l * R2.h + R3.l * R0.h) * 2^48
[T2] + (R1.l * R2.l + R3.l * R0.l + R0.h * R2.h) * 2^32
[T3] + (R0.l * R2.h + R2.l * R0.h) * 2^16
[T4] + (R0.l * R2.l)
We can discard the first three lines marked "X" since we produce
only a 64 bit result. So, we need ten 16-bit multiplies.
Individual mul-acc results:
[E1] = R1.h * R2.l + R3.h * R0.l + R1.l * R2.h + R3.l * R0.h
[E2] = R1.l * R2.l + R3.l * R0.l + R0.h * R2.h
[E3] = R0.l * R2.h + R2.l * R0.h
[E4] = R0.l * R2.l
We also need to add high parts from lower-level results to higher ones:
E[n]c = E[n] + (E[n+1]c >> 16), where E4c := E4
One interesting property is that all parts of the result that depend
on the sign of the multiplication are discarded. Those would be the
multiplications involving R1.h and R3.h, but only the top 16 bit of
the 32 bit result depend on the sign, and since R1.h and R3.h only
occur in E1, the top half of these results is cut off.
So, we can just use FU mode for all of the 16-bit multiplies, and
ignore questions of when to use mixed mode. */
___muldi3:
/* [SP] technically is part of the caller's frame, but we can
use it as scratch space. */
A0 = R2.H * R1.L, A1 = R2.L * R1.H (FU) || R3 = [SP + 12]; /* E1 */
A0 += R3.H * R0.L, A1 += R3.L * R0.H (FU) || [SP] = R4; /* E1 */
A0 += A1; /* E1 */
R4 = A0.w;
A0 = R0.l * R3.l (FU); /* E2 */
A0 += R2.l * R1.l (FU); /* E2 */
A1 = R2.L * R0.L (FU); /* E4 */
R3 = A1.w;
A1 = A1 >> 16; /* E3c */
A0 += R2.H * R0.H, A1 += R2.L * R0.H (FU); /* E2, E3c */
A1 += R0.L * R2.H (FU); /* E3c */
R0 = A1.w;
A1 = A1 >> 16; /* E2c */
A0 += A1; /* E2c */
R1 = A0.w;
/* low(result) = low(E3c):low(E4) */
R0 = PACK (R0.l, R3.l);
/* high(result) = E2c + (E1 << 16) */
R1.h = R1.h + R4.l (NS) || R4 = [SP];
RTS;
.size ___muldi3, .-___muldi3
#endif