blackfin architecture
This adds support for the Analog Devices Blackfin processor architecture, and
currently supports the BF533, BF532, BF531, BF537, BF536, BF534, and BF561
(Dual Core) devices, with a variety of development platforms including those
avaliable from Analog Devices (BF533-EZKit, BF533-STAMP, BF537-STAMP,
BF561-EZKIT), and Bluetechnix! Tinyboards.
The Blackfin architecture was jointly developed by Intel and Analog Devices
Inc. (ADI) as the Micro Signal Architecture (MSA) core and introduced it in
December of 2000. Since then ADI has put this core into its Blackfin
processor family of devices. The Blackfin core has the advantages of a clean,
orthogonal,RISC-like microprocessor instruction set. It combines a dual-MAC
(Multiply/Accumulate), state-of-the-art signal processing engine and
single-instruction, multiple-data (SIMD) multimedia capabilities into a single
instruction-set architecture.
The Blackfin architecture, including the instruction set, is described by the
ADSP-BF53x/BF56x Blackfin Processor Programming Reference
http://blackfin.uclinux.org/gf/download/frsrelease/29/2549/Blackfin_PRM.pdf
The Blackfin processor is already supported by major releases of gcc, and
there are binary and source rpms/tarballs for many architectures at:
http://blackfin.uclinux.org/gf/project/toolchain/frs There is complete
documentation, including "getting started" guides available at:
http://docs.blackfin.uclinux.org/ which provides links to the sources and
patches you will need in order to set up a cross-compiling environment for
bfin-linux-uclibc
This patch, as well as the other patches (toolchain, distribution,
uClibc) are actively supported by Analog Devices Inc, at:
http://blackfin.uclinux.org/
We have tested this on LTP, and our test plan (including pass/fails) can
be found at:
http://docs.blackfin.uclinux.org/doku.php?id=testing_the_linux_kernel
[m.kozlowski@tuxland.pl: balance parenthesis in blackfin header files]
Signed-off-by: Bryan Wu <bryan.wu@analog.com>
Signed-off-by: Mariusz Kozlowski <m.kozlowski@tuxland.pl>
Signed-off-by: Aubrey Li <aubrey.li@analog.com>
Signed-off-by: Jie Zhang <jie.zhang@analog.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2007-05-06 15:50:22 -06:00
|
|
|
/*
|
2009-09-24 08:11:24 -06:00
|
|
|
* Copyright 2004-2009 Analog Devices Inc.
|
blackfin architecture
This adds support for the Analog Devices Blackfin processor architecture, and
currently supports the BF533, BF532, BF531, BF537, BF536, BF534, and BF561
(Dual Core) devices, with a variety of development platforms including those
avaliable from Analog Devices (BF533-EZKit, BF533-STAMP, BF537-STAMP,
BF561-EZKIT), and Bluetechnix! Tinyboards.
The Blackfin architecture was jointly developed by Intel and Analog Devices
Inc. (ADI) as the Micro Signal Architecture (MSA) core and introduced it in
December of 2000. Since then ADI has put this core into its Blackfin
processor family of devices. The Blackfin core has the advantages of a clean,
orthogonal,RISC-like microprocessor instruction set. It combines a dual-MAC
(Multiply/Accumulate), state-of-the-art signal processing engine and
single-instruction, multiple-data (SIMD) multimedia capabilities into a single
instruction-set architecture.
The Blackfin architecture, including the instruction set, is described by the
ADSP-BF53x/BF56x Blackfin Processor Programming Reference
http://blackfin.uclinux.org/gf/download/frsrelease/29/2549/Blackfin_PRM.pdf
The Blackfin processor is already supported by major releases of gcc, and
there are binary and source rpms/tarballs for many architectures at:
http://blackfin.uclinux.org/gf/project/toolchain/frs There is complete
documentation, including "getting started" guides available at:
http://docs.blackfin.uclinux.org/ which provides links to the sources and
patches you will need in order to set up a cross-compiling environment for
bfin-linux-uclibc
This patch, as well as the other patches (toolchain, distribution,
uClibc) are actively supported by Analog Devices Inc, at:
http://blackfin.uclinux.org/
We have tested this on LTP, and our test plan (including pass/fails) can
be found at:
http://docs.blackfin.uclinux.org/doku.php?id=testing_the_linux_kernel
[m.kozlowski@tuxland.pl: balance parenthesis in blackfin header files]
Signed-off-by: Bryan Wu <bryan.wu@analog.com>
Signed-off-by: Mariusz Kozlowski <m.kozlowski@tuxland.pl>
Signed-off-by: Aubrey Li <aubrey.li@analog.com>
Signed-off-by: Jie Zhang <jie.zhang@analog.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2007-05-06 15:50:22 -06:00
|
|
|
*
|
2012-05-17 00:45:27 -06:00
|
|
|
* Licensed under the Clear BSD license or the GPL-2 (or later)
|
2009-09-24 08:11:24 -06:00
|
|
|
*
|
|
|
|
* 16 / 32 bit signed division.
|
blackfin architecture
This adds support for the Analog Devices Blackfin processor architecture, and
currently supports the BF533, BF532, BF531, BF537, BF536, BF534, and BF561
(Dual Core) devices, with a variety of development platforms including those
avaliable from Analog Devices (BF533-EZKit, BF533-STAMP, BF537-STAMP,
BF561-EZKIT), and Bluetechnix! Tinyboards.
The Blackfin architecture was jointly developed by Intel and Analog Devices
Inc. (ADI) as the Micro Signal Architecture (MSA) core and introduced it in
December of 2000. Since then ADI has put this core into its Blackfin
processor family of devices. The Blackfin core has the advantages of a clean,
orthogonal,RISC-like microprocessor instruction set. It combines a dual-MAC
(Multiply/Accumulate), state-of-the-art signal processing engine and
single-instruction, multiple-data (SIMD) multimedia capabilities into a single
instruction-set architecture.
The Blackfin architecture, including the instruction set, is described by the
ADSP-BF53x/BF56x Blackfin Processor Programming Reference
http://blackfin.uclinux.org/gf/download/frsrelease/29/2549/Blackfin_PRM.pdf
The Blackfin processor is already supported by major releases of gcc, and
there are binary and source rpms/tarballs for many architectures at:
http://blackfin.uclinux.org/gf/project/toolchain/frs There is complete
documentation, including "getting started" guides available at:
http://docs.blackfin.uclinux.org/ which provides links to the sources and
patches you will need in order to set up a cross-compiling environment for
bfin-linux-uclibc
This patch, as well as the other patches (toolchain, distribution,
uClibc) are actively supported by Analog Devices Inc, at:
http://blackfin.uclinux.org/
We have tested this on LTP, and our test plan (including pass/fails) can
be found at:
http://docs.blackfin.uclinux.org/doku.php?id=testing_the_linux_kernel
[m.kozlowski@tuxland.pl: balance parenthesis in blackfin header files]
Signed-off-by: Bryan Wu <bryan.wu@analog.com>
Signed-off-by: Mariusz Kozlowski <m.kozlowski@tuxland.pl>
Signed-off-by: Aubrey Li <aubrey.li@analog.com>
Signed-off-by: Jie Zhang <jie.zhang@analog.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2007-05-06 15:50:22 -06:00
|
|
|
* Special cases :
|
|
|
|
* 1) If(numerator == 0)
|
|
|
|
* return 0
|
|
|
|
* 2) If(denominator ==0)
|
|
|
|
* return positive max = 0x7fffffff
|
|
|
|
* 3) If(numerator == denominator)
|
|
|
|
* return 1
|
|
|
|
* 4) If(denominator ==1)
|
|
|
|
* return numerator
|
|
|
|
* 5) If(denominator == -1)
|
|
|
|
* return -numerator
|
|
|
|
*
|
|
|
|
* Operand : R0 - Numerator (i)
|
|
|
|
* R1 - Denominator (i)
|
|
|
|
* R0 - Quotient (o)
|
|
|
|
* Registers Used : R2-R7,P0-P2
|
|
|
|
*
|
|
|
|
*/
|
|
|
|
|
|
|
|
.global ___divsi3;
|
2007-06-11 01:31:30 -06:00
|
|
|
.type ___divsi3, STT_FUNC;
|
blackfin architecture
This adds support for the Analog Devices Blackfin processor architecture, and
currently supports the BF533, BF532, BF531, BF537, BF536, BF534, and BF561
(Dual Core) devices, with a variety of development platforms including those
avaliable from Analog Devices (BF533-EZKit, BF533-STAMP, BF537-STAMP,
BF561-EZKIT), and Bluetechnix! Tinyboards.
The Blackfin architecture was jointly developed by Intel and Analog Devices
Inc. (ADI) as the Micro Signal Architecture (MSA) core and introduced it in
December of 2000. Since then ADI has put this core into its Blackfin
processor family of devices. The Blackfin core has the advantages of a clean,
orthogonal,RISC-like microprocessor instruction set. It combines a dual-MAC
(Multiply/Accumulate), state-of-the-art signal processing engine and
single-instruction, multiple-data (SIMD) multimedia capabilities into a single
instruction-set architecture.
The Blackfin architecture, including the instruction set, is described by the
ADSP-BF53x/BF56x Blackfin Processor Programming Reference
http://blackfin.uclinux.org/gf/download/frsrelease/29/2549/Blackfin_PRM.pdf
The Blackfin processor is already supported by major releases of gcc, and
there are binary and source rpms/tarballs for many architectures at:
http://blackfin.uclinux.org/gf/project/toolchain/frs There is complete
documentation, including "getting started" guides available at:
http://docs.blackfin.uclinux.org/ which provides links to the sources and
patches you will need in order to set up a cross-compiling environment for
bfin-linux-uclibc
This patch, as well as the other patches (toolchain, distribution,
uClibc) are actively supported by Analog Devices Inc, at:
http://blackfin.uclinux.org/
We have tested this on LTP, and our test plan (including pass/fails) can
be found at:
http://docs.blackfin.uclinux.org/doku.php?id=testing_the_linux_kernel
[m.kozlowski@tuxland.pl: balance parenthesis in blackfin header files]
Signed-off-by: Bryan Wu <bryan.wu@analog.com>
Signed-off-by: Mariusz Kozlowski <m.kozlowski@tuxland.pl>
Signed-off-by: Aubrey Li <aubrey.li@analog.com>
Signed-off-by: Jie Zhang <jie.zhang@analog.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2007-05-06 15:50:22 -06:00
|
|
|
|
|
|
|
#ifdef CONFIG_ARITHMETIC_OPS_L1
|
|
|
|
.section .l1.text
|
|
|
|
#else
|
|
|
|
.text
|
|
|
|
#endif
|
|
|
|
|
|
|
|
.align 2;
|
|
|
|
___divsi3 :
|
|
|
|
|
|
|
|
|
|
|
|
R3 = R0 ^ R1;
|
|
|
|
R0 = ABS R0;
|
|
|
|
|
|
|
|
CC = V;
|
|
|
|
|
|
|
|
r3 = rot r3 by -1;
|
|
|
|
r1 = abs r1; /* now both positive, r3.30 means "negate result",
|
|
|
|
** r3.31 means overflow, add one to result
|
|
|
|
*/
|
|
|
|
cc = r0 < r1;
|
|
|
|
if cc jump .Lret_zero;
|
|
|
|
r2 = r1 >> 15;
|
|
|
|
cc = r2;
|
|
|
|
if cc jump .Lidents;
|
|
|
|
r2 = r1 << 16;
|
|
|
|
cc = r2 <= r0;
|
|
|
|
if cc jump .Lidents;
|
|
|
|
|
|
|
|
DIVS(R0, R1);
|
|
|
|
DIVQ(R0, R1);
|
|
|
|
DIVQ(R0, R1);
|
|
|
|
DIVQ(R0, R1);
|
|
|
|
DIVQ(R0, R1);
|
|
|
|
DIVQ(R0, R1);
|
|
|
|
DIVQ(R0, R1);
|
|
|
|
DIVQ(R0, R1);
|
|
|
|
DIVQ(R0, R1);
|
|
|
|
DIVQ(R0, R1);
|
|
|
|
DIVQ(R0, R1);
|
|
|
|
DIVQ(R0, R1);
|
|
|
|
DIVQ(R0, R1);
|
|
|
|
DIVQ(R0, R1);
|
|
|
|
DIVQ(R0, R1);
|
|
|
|
DIVQ(R0, R1);
|
|
|
|
DIVQ(R0, R1);
|
|
|
|
|
|
|
|
R0 = R0.L (Z);
|
|
|
|
r1 = r3 >> 31; /* add overflow issue back in */
|
|
|
|
r0 = r0 + r1;
|
|
|
|
r1 = -r0;
|
|
|
|
cc = bittst(r3, 30);
|
|
|
|
if cc r0 = r1;
|
|
|
|
RTS;
|
|
|
|
|
|
|
|
/* Can't use the primitives. Test common identities.
|
|
|
|
** If the identity is true, return the value in R2.
|
|
|
|
*/
|
|
|
|
|
|
|
|
.Lidents:
|
|
|
|
CC = R1 == 0; /* check for divide by zero */
|
|
|
|
IF CC JUMP .Lident_return;
|
|
|
|
|
|
|
|
CC = R0 == 0; /* check for division of zero */
|
|
|
|
IF CC JUMP .Lzero_return;
|
|
|
|
|
|
|
|
CC = R0 == R1; /* check for identical operands */
|
|
|
|
IF CC JUMP .Lident_return;
|
|
|
|
|
|
|
|
CC = R1 == 1; /* check for divide by 1 */
|
|
|
|
IF CC JUMP .Lident_return;
|
|
|
|
|
|
|
|
R2.L = ONES R1;
|
|
|
|
R2 = R2.L (Z);
|
|
|
|
CC = R2 == 1;
|
|
|
|
IF CC JUMP .Lpower_of_two;
|
|
|
|
|
|
|
|
/* Identities haven't helped either.
|
|
|
|
** Perform the full division process.
|
|
|
|
*/
|
|
|
|
|
|
|
|
P1 = 31; /* Set loop counter */
|
|
|
|
|
|
|
|
[--SP] = (R7:5); /* Push registers R5-R7 */
|
|
|
|
R2 = -R1;
|
|
|
|
[--SP] = R2;
|
|
|
|
R2 = R0 << 1; /* R2 lsw of dividend */
|
|
|
|
R6 = R0 ^ R1; /* Get sign */
|
|
|
|
R5 = R6 >> 31; /* Shift sign to LSB */
|
|
|
|
|
|
|
|
R0 = 0 ; /* Clear msw partial remainder */
|
|
|
|
R2 = R2 | R5; /* Shift quotient bit */
|
|
|
|
R6 = R0 ^ R1; /* Get new quotient bit */
|
|
|
|
|
|
|
|
LSETUP(.Llst,.Llend) LC0 = P1; /* Setup loop */
|
|
|
|
.Llst: R7 = R2 >> 31; /* record copy of carry from R2 */
|
|
|
|
R2 = R2 << 1; /* Shift 64 bit dividend up by 1 bit */
|
|
|
|
R0 = R0 << 1 || R5 = [SP];
|
|
|
|
R0 = R0 | R7; /* and add carry */
|
|
|
|
CC = R6 < 0; /* Check quotient(AQ) */
|
|
|
|
/* we might be subtracting divisor (AQ==0) */
|
|
|
|
IF CC R5 = R1; /* or we might be adding divisor (AQ==1)*/
|
|
|
|
R0 = R0 + R5; /* do add or subtract, as indicated by AQ */
|
|
|
|
R6 = R0 ^ R1; /* Generate next quotient bit */
|
|
|
|
R5 = R6 >> 31;
|
|
|
|
/* Assume AQ==1, shift in zero */
|
|
|
|
BITTGL(R5,0); /* tweak AQ to be what we want to shift in */
|
|
|
|
.Llend: R2 = R2 + R5; /* and then set shifted-in value to
|
|
|
|
** tweaked AQ.
|
|
|
|
*/
|
|
|
|
r1 = r3 >> 31;
|
|
|
|
r2 = r2 + r1;
|
|
|
|
cc = bittst(r3,30);
|
|
|
|
r0 = -r2;
|
|
|
|
if !cc r0 = r2;
|
|
|
|
SP += 4;
|
|
|
|
(R7:5)= [SP++]; /* Pop registers R6-R7 */
|
|
|
|
RTS;
|
|
|
|
|
|
|
|
.Lident_return:
|
|
|
|
CC = R1 == 0; /* check for divide by zero => 0x7fffffff */
|
|
|
|
R2 = -1 (X);
|
|
|
|
R2 >>= 1;
|
|
|
|
IF CC JUMP .Ltrue_ident_return;
|
|
|
|
|
|
|
|
CC = R0 == R1; /* check for identical operands => 1 */
|
|
|
|
R2 = 1 (Z);
|
|
|
|
IF CC JUMP .Ltrue_ident_return;
|
|
|
|
|
|
|
|
R2 = R0; /* assume divide by 1 => numerator */
|
|
|
|
/*FALLTHRU*/
|
|
|
|
|
|
|
|
.Ltrue_ident_return:
|
|
|
|
R0 = R2; /* Return an identity value */
|
|
|
|
R2 = -R2;
|
|
|
|
CC = bittst(R3,30);
|
|
|
|
IF CC R0 = R2;
|
|
|
|
.Lzero_return:
|
|
|
|
RTS; /* ...including zero */
|
|
|
|
|
|
|
|
.Lpower_of_two:
|
|
|
|
/* Y has a single bit set, which means it's a power of two.
|
|
|
|
** That means we can perform the division just by shifting
|
|
|
|
** X to the right the appropriate number of bits
|
|
|
|
*/
|
|
|
|
|
|
|
|
/* signbits returns the number of sign bits, minus one.
|
|
|
|
** 1=>30, 2=>29, ..., 0x40000000=>0. Which means we need
|
|
|
|
** to shift right n-signbits spaces. It also means 0x80000000
|
|
|
|
** is a special case, because that *also* gives a signbits of 0
|
|
|
|
*/
|
|
|
|
|
|
|
|
R2 = R0 >> 31;
|
|
|
|
CC = R1 < 0;
|
|
|
|
IF CC JUMP .Ltrue_ident_return;
|
|
|
|
|
|
|
|
R1.l = SIGNBITS R1;
|
|
|
|
R1 = R1.L (Z);
|
|
|
|
R1 += -30;
|
|
|
|
R0 = LSHIFT R0 by R1.L;
|
|
|
|
r1 = r3 >> 31;
|
|
|
|
r0 = r0 + r1;
|
|
|
|
R2 = -R0; // negate result if necessary
|
|
|
|
CC = bittst(R3,30);
|
|
|
|
IF CC R0 = R2;
|
|
|
|
RTS;
|
|
|
|
|
|
|
|
.Lret_zero:
|
|
|
|
R0 = 0;
|
|
|
|
RTS;
|
2007-06-11 01:31:30 -06:00
|
|
|
|
|
|
|
.size ___divsi3, .-___divsi3
|