iv/linux
1
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity
101246ec02
linux/include/math-emu/op-common.h

877 lines
26 KiB
C
Raw Normal View History

Linux-2.6.12-rc2 Initial git repository build. I'm not bothering with the full history, even though we have it. We can create a separate "historical" git archive of that later if we want to, and in the meantime it's about 3.2GB when imported into git - space that would just make the early git days unnecessarily complicated, when we don't have a lot of good infrastructure for it. Let it rip!
2005-04-17 02:20:36 +04:00
/* Software floating-point emulation. Common operations.
Copyright (C) 1997,1998,1999 Free Software Foundation, Inc.
This file is part of the GNU C Library.
Contributed by Richard Henderson (rth@cygnus.com),
Jakub Jelinek (jj@ultra.linux.cz),
David S. Miller (davem@redhat.com) and
Peter Maydell (pmaydell@chiark.greenend.org.uk).
The GNU C Library is free software; you can redistribute it and/or
modify it under the terms of the GNU Library General Public License as
published by the Free Software Foundation; either version 2 of the
License, or (at your option) any later version.
The GNU C Library 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
Library General Public License for more details.
You should have received a copy of the GNU Library General Public
License along with the GNU C Library; see the file COPYING.LIB. If
not, write to the Free Software Foundation, Inc.,
59 Temple Place - Suite 330, Boston, MA 02111-1307, USA. */
#ifndef __MATH_EMU_OP_COMMON_H__
#define __MATH_EMU_OP_COMMON_H__
#define _FP_DECL(wc, X) \
_FP_I_TYPE X##_c=0, X##_s=0, X##_e=0; \
_FP_FRAC_DECL_##wc(X)
/*
Fix misspellings of "truly" in comments. Some comments misspell "truly"; this fixes them. No code changes. Signed-off-by: Adam Buchbinder <adam.buchbinder@gmail.com> Signed-off-by: Jiri Kosina <jkosina@suse.cz>
2009-12-12 00:35:39 +03:00
* Finish truly unpacking a native fp value by classifying the kind
Linux-2.6.12-rc2 Initial git repository build. I'm not bothering with the full history, even though we have it. We can create a separate "historical" git archive of that later if we want to, and in the meantime it's about 3.2GB when imported into git - space that would just make the early git days unnecessarily complicated, when we don't have a lot of good infrastructure for it. Let it rip!
2005-04-17 02:20:36 +04:00
* of fp value and normalizing both the exponent and the fraction.
*/
#define _FP_UNPACK_CANONICAL(fs, wc, X) \
do { \
switch (X##_e) \
{ \
default: \
_FP_FRAC_HIGH_RAW_##fs(X) |= _FP_IMPLBIT_##fs; \
_FP_FRAC_SLL_##wc(X, _FP_WORKBITS); \
X##_e -= _FP_EXPBIAS_##fs; \
X##_c = FP_CLS_NORMAL; \
break; \
\
case 0: \
if (_FP_FRAC_ZEROP_##wc(X)) \
X##_c = FP_CLS_ZERO; \
else \
{ \
/* a denormalized number */ \
_FP_I_TYPE _shift; \
_FP_FRAC_CLZ_##wc(_shift, X); \
_shift -= _FP_FRACXBITS_##fs; \
_FP_FRAC_SLL_##wc(X, (_shift+_FP_WORKBITS)); \
X##_e -= _FP_EXPBIAS_##fs - 1 + _shift; \
X##_c = FP_CLS_NORMAL; \
FP_SET_EXCEPTION(FP_EX_DENORM); \
if (FP_DENORM_ZERO) \
{ \
FP_SET_EXCEPTION(FP_EX_INEXACT); \
X##_c = FP_CLS_ZERO; \
} \
} \
break; \
\
case _FP_EXPMAX_##fs: \
if (_FP_FRAC_ZEROP_##wc(X)) \
X##_c = FP_CLS_INF; \
else \
{ \
X##_c = FP_CLS_NAN; \
/* Check for signaling NaN */ \
if (!(_FP_FRAC_HIGH_RAW_##fs(X) & _FP_QNANBIT_##fs)) \
math-emu: Add support for reporting exact invalid exception Some architectures (like powerpc) provide status information on the exact type of invalid exception. This is pretty straight forward as we already report invalid exceptions via FP_SET_EXCEPTION. We add new flags (FP_EX_INVALID_*) the architecture code can define if it wants the exact invalid exception reported. We had to split out the INF/INF and 0/0 cases for divide to allow reporting the two invalid forms properly. Signed-off-by: Kumar Gala <galak@kernel.crashing.org> Acked-by: David S. Miller <davem@davemloft.net>
2008-06-27 18:39:00 +04:00
FP_SET_EXCEPTION(FP_EX_INVALID | FP_EX_INVALID_SNAN); \
Linux-2.6.12-rc2 Initial git repository build. I'm not bothering with the full history, even though we have it. We can create a separate "historical" git archive of that later if we want to, and in the meantime it's about 3.2GB when imported into git - space that would just make the early git days unnecessarily complicated, when we don't have a lot of good infrastructure for it. Let it rip!
2005-04-17 02:20:36 +04:00
} \
break; \
} \
} while (0)
/*
* Before packing the bits back into the native fp result, take care
* of such mundane things as rounding and overflow. Also, for some
* kinds of fp values, the original parts may not have been fully
* extracted -- but that is ok, we can regenerate them now.
*/
#define _FP_PACK_CANONICAL(fs, wc, X) \
do { \
switch (X##_c) \
{ \
case FP_CLS_NORMAL: \
X##_e += _FP_EXPBIAS_##fs; \
if (X##_e > 0) \
{ \
_FP_ROUND(wc, X); \
if (_FP_FRAC_OVERP_##wc(fs, X)) \
{ \
_FP_FRAC_CLEAR_OVERP_##wc(fs, X); \
X##_e++; \
} \
_FP_FRAC_SRL_##wc(X, _FP_WORKBITS); \
if (X##_e >= _FP_EXPMAX_##fs) \
{ \
/* overflow */ \
switch (FP_ROUNDMODE) \
{ \
case FP_RND_NEAREST: \
X##_c = FP_CLS_INF; \
break; \
case FP_RND_PINF: \
if (!X##_s) X##_c = FP_CLS_INF; \
break; \
case FP_RND_MINF: \
if (X##_s) X##_c = FP_CLS_INF; \
break; \
} \
if (X##_c == FP_CLS_INF) \
{ \
/* Overflow to infinity */ \
X##_e = _FP_EXPMAX_##fs; \
_FP_FRAC_SET_##wc(X, _FP_ZEROFRAC_##wc); \
} \
else \
{ \
/* Overflow to maximum normal */ \
X##_e = _FP_EXPMAX_##fs - 1; \
_FP_FRAC_SET_##wc(X, _FP_MAXFRAC_##wc); \
} \
FP_SET_EXCEPTION(FP_EX_OVERFLOW); \
FP_SET_EXCEPTION(FP_EX_INEXACT); \
} \
} \
else \
{ \
/* we've got a denormalized number */ \
X##_e = -X##_e + 1; \
if (X##_e <= _FP_WFRACBITS_##fs) \
{ \
_FP_FRAC_SRS_##wc(X, X##_e, _FP_WFRACBITS_##fs); \
if (_FP_FRAC_HIGH_##fs(X) \
& (_FP_OVERFLOW_##fs >> 1)) \
{ \
X##_e = 1; \
_FP_FRAC_SET_##wc(X, _FP_ZEROFRAC_##wc); \
} \
else \
{ \
math-emu: Fix signalling of underflow and inexact while packing result. I'm trying to move the powerpc math-emu code to use the include/math-emu bits. In doing so I've been using TestFloat to see how good or bad we are doing. For the most part the current math-emu code that PPC uses has a number of issues that the code in include/math-emu seems to solve (plus bugs we've had for ever that no one every realized). Anyways, I've come across a case that we are flagging underflow and inexact because we think we have a denormalized result from a double precision divide: 000.FFFFFFFFFFFFF / 3FE.FFFFFFFFFFFFE soft: 001.0000000000000 ..... syst: 001.0000000000000 ...ux What it looks like is the results out of FP_DIV_D are: D: sign: 0 mantissa: 01000000 00000000 exp: -1023 (0) The problem seems like we aren't normalizing the result and bumping the exp. Now that I'm digging into this a bit I'm thinking my issue has to do with the fix DaveM put in place from back in Aug 2007 (commit 405849610fd96b4f34cd1875c4c033228fea6c0f): [MATH-EMU]: Fix underflow exception reporting. 2) we ended up rounding back up to normal (this is the case where we set the exponent to 1 and set the fraction to zero), this should set inexact too ... Another example, "0x0.0000000000001p-1022 / 16.0", should signal both inexact and underflow. The cpu implementations and ieee1754 literature is very clear about this. This is case #2 above. Here is the distilled glibc test case from Jakub Jelinek which prompted that commit: -------------------- #include <float.h> #include <fenv.h> #include <stdio.h> volatile double d = DBL_MIN; volatile double e = 0x0.0000000000001p-1022; volatile double f = 16.0; int main (void) { printf ("%x\n", fetestexcept (FE_UNDERFLOW)); d /= f; printf ("%x\n", fetestexcept (FE_UNDERFLOW)); e /= f; printf ("%x\n", fetestexcept (FE_UNDERFLOW)); return 0; } -------------------- It looks like the case I have we are exact before rounding, but think it looks like the rounding case since it appears as if "overflow is set". 000.FFFFFFFFFFFFF / 3FE.FFFFFFFFFFFFE = 001.0000000000000 I think the following adds the check for my case and still works for the issue your commit was trying to resolve. Signed-off-by: David S. Miller <davem@davemloft.net>
2008-10-22 09:19:00 +04:00
_FP_ROUND(wc, X); \
if (_FP_FRAC_HIGH_##fs(X) \
& (_FP_OVERFLOW_##fs >> 1)) \
{ \
X##_e = 1; \
_FP_FRAC_SET_##wc(X, _FP_ZEROFRAC_##wc); \
FP_SET_EXCEPTION(FP_EX_INEXACT); \
} \
else \
{ \
X##_e = 0; \
_FP_FRAC_SRL_##wc(X, _FP_WORKBITS); \
} \
Linux-2.6.12-rc2 Initial git repository build. I'm not bothering with the full history, even though we have it. We can create a separate "historical" git archive of that later if we want to, and in the meantime it's about 3.2GB when imported into git - space that would just make the early git days unnecessarily complicated, when we don't have a lot of good infrastructure for it. Let it rip!
2005-04-17 02:20:36 +04:00
} \
[MATH-EMU]: Fix underflow exception reporting. The underflow exception cases were wrong. This is one weird area of ieee1754 handling in that the underflow behavior changes based upon whether underflow is enabled in the trap enable mask of the FPU control register. As a specific case the Sparc V9 manual gives us the following description: -------------------- If UFM = 0: Underflow occurs if a nonzero result is tiny and a loss of accuracy occurs. Tininess may be detected before or after rounding. Loss of accuracy may be either a denormalization loss or an inexact result. If UFM = 1: Underflow occurs if a nonzero result is tiny. Tininess may be detected before or after rounding. -------------------- What this amounts to in the packing case is if we go subnormal, we set underflow if any of the following are true: 1) rounding sets inexact 2) we ended up rounding back up to normal (this is the case where we set the exponent to 1 and set the fraction to zero), this should set inexact too 3) underflow is set in FPU control register trap-enable mask The initially discovered example was "DBL_MIN / 16.0" which incorrectly generated an underflow. It should not, unless underflow is set in the trap-enable mask of the FPU csr. Another example, "0x0.0000000000001p-1022 / 16.0", should signal both inexact and underflow. The cpu implementations and ieee1754 literature is very clear about this. This is case #2 above. However, if underflow is set in the trap enable mask, only underflow should be set and reported as a trap. That is handled properly by the prioritization logic in arch/sparc{,64}/math-emu/math.c:record_exception(). Based upon a report and test case from Jakub Jelinek. Signed-off-by: David S. Miller <davem@davemloft.net>
2007-08-17 09:59:49 +04:00
if ((FP_CUR_EXCEPTIONS & FP_EX_INEXACT) || \
(FP_TRAPPING_EXCEPTIONS & FP_EX_UNDERFLOW)) \
FP_SET_EXCEPTION(FP_EX_UNDERFLOW); \
Linux-2.6.12-rc2 Initial git repository build. I'm not bothering with the full history, even though we have it. We can create a separate "historical" git archive of that later if we want to, and in the meantime it's about 3.2GB when imported into git - space that would just make the early git days unnecessarily complicated, when we don't have a lot of good infrastructure for it. Let it rip!
2005-04-17 02:20:36 +04:00
} \
else \
{ \
/* underflow to zero */ \
X##_e = 0; \
if (!_FP_FRAC_ZEROP_##wc(X)) \
{ \
_FP_FRAC_SET_##wc(X, _FP_MINFRAC_##wc); \
_FP_ROUND(wc, X); \
_FP_FRAC_LOW_##wc(X) >>= (_FP_WORKBITS); \
} \
FP_SET_EXCEPTION(FP_EX_UNDERFLOW); \
} \
} \
break; \
\
case FP_CLS_ZERO: \
X##_e = 0; \
_FP_FRAC_SET_##wc(X, _FP_ZEROFRAC_##wc); \
break; \
\
case FP_CLS_INF: \
X##_e = _FP_EXPMAX_##fs; \
_FP_FRAC_SET_##wc(X, _FP_ZEROFRAC_##wc); \
break; \
\
case FP_CLS_NAN: \
X##_e = _FP_EXPMAX_##fs; \
if (!_FP_KEEPNANFRACP) \
{ \
_FP_FRAC_SET_##wc(X, _FP_NANFRAC_##fs); \
X##_s = _FP_NANSIGN_##fs; \
} \
else \
_FP_FRAC_HIGH_RAW_##fs(X) |= _FP_QNANBIT_##fs; \
break; \
} \
} while (0)
/* This one accepts raw argument and not cooked, returns
* 1 if X is a signaling NaN.
*/
#define _FP_ISSIGNAN(fs, wc, X) \
({ \
int __ret = 0; \
if (X##_e == _FP_EXPMAX_##fs) \
{ \
if (!_FP_FRAC_ZEROP_##wc(X) \
&& !(_FP_FRAC_HIGH_RAW_##fs(X) & _FP_QNANBIT_##fs)) \
__ret = 1; \
} \
__ret; \
})
/*
* Main addition routine. The input values should be cooked.
*/
#define _FP_ADD_INTERNAL(fs, wc, R, X, Y, OP) \
do { \
switch (_FP_CLS_COMBINE(X##_c, Y##_c)) \
{ \
case _FP_CLS_COMBINE(FP_CLS_NORMAL,FP_CLS_NORMAL): \
{ \
/* shift the smaller number so that its exponent matches the larger */ \
_FP_I_TYPE diff = X##_e - Y##_e; \
\
if (diff < 0) \
{ \
diff = -diff; \
if (diff <= _FP_WFRACBITS_##fs) \
_FP_FRAC_SRS_##wc(X, diff, _FP_WFRACBITS_##fs); \
else if (!_FP_FRAC_ZEROP_##wc(X)) \
_FP_FRAC_SET_##wc(X, _FP_MINFRAC_##wc); \
R##_e = Y##_e; \
} \
else \
{ \
if (diff > 0) \
{ \
if (diff <= _FP_WFRACBITS_##fs) \
_FP_FRAC_SRS_##wc(Y, diff, _FP_WFRACBITS_##fs); \
else if (!_FP_FRAC_ZEROP_##wc(Y)) \
_FP_FRAC_SET_##wc(Y, _FP_MINFRAC_##wc); \
} \
R##_e = X##_e; \
} \
\
R##_c = FP_CLS_NORMAL; \
\
if (X##_s == Y##_s) \
{ \
R##_s = X##_s; \
_FP_FRAC_ADD_##wc(R, X, Y); \
if (_FP_FRAC_OVERP_##wc(fs, R)) \
{ \
_FP_FRAC_SRS_##wc(R, 1, _FP_WFRACBITS_##fs); \
R##_e++; \
} \
} \
else \
{ \
R##_s = X##_s; \
_FP_FRAC_SUB_##wc(R, X, Y); \
if (_FP_FRAC_ZEROP_##wc(R)) \
{ \
/* return an exact zero */ \
if (FP_ROUNDMODE == FP_RND_MINF) \
R##_s |= Y##_s; \
else \
R##_s &= Y##_s; \
R##_c = FP_CLS_ZERO; \
} \
else \
{ \
if (_FP_FRAC_NEGP_##wc(R)) \
{ \
_FP_FRAC_SUB_##wc(R, Y, X); \
R##_s = Y##_s; \
} \
\
/* renormalize after subtraction */ \
_FP_FRAC_CLZ_##wc(diff, R); \
diff -= _FP_WFRACXBITS_##fs; \
if (diff) \
{ \
R##_e -= diff; \
_FP_FRAC_SLL_##wc(R, diff); \
} \
} \
} \
break; \
} \
\
case _FP_CLS_COMBINE(FP_CLS_NAN,FP_CLS_NAN): \
_FP_CHOOSENAN(fs, wc, R, X, Y, OP); \
break; \
\
case _FP_CLS_COMBINE(FP_CLS_NORMAL,FP_CLS_ZERO): \
R##_e = X##_e; \
case _FP_CLS_COMBINE(FP_CLS_NAN,FP_CLS_NORMAL): \
case _FP_CLS_COMBINE(FP_CLS_NAN,FP_CLS_INF): \
case _FP_CLS_COMBINE(FP_CLS_NAN,FP_CLS_ZERO): \
_FP_FRAC_COPY_##wc(R, X); \
R##_s = X##_s; \
R##_c = X##_c; \
break; \
\
case _FP_CLS_COMBINE(FP_CLS_ZERO,FP_CLS_NORMAL): \
R##_e = Y##_e; \
case _FP_CLS_COMBINE(FP_CLS_NORMAL,FP_CLS_NAN): \
case _FP_CLS_COMBINE(FP_CLS_INF,FP_CLS_NAN): \
case _FP_CLS_COMBINE(FP_CLS_ZERO,FP_CLS_NAN): \
_FP_FRAC_COPY_##wc(R, Y); \
R##_s = Y##_s; \
R##_c = Y##_c; \
break; \
\
case _FP_CLS_COMBINE(FP_CLS_INF,FP_CLS_INF): \
if (X##_s != Y##_s) \
{ \
/* +INF + -INF => NAN */ \
_FP_FRAC_SET_##wc(R, _FP_NANFRAC_##fs); \
R##_s = _FP_NANSIGN_##fs; \
R##_c = FP_CLS_NAN; \
math-emu: Add support for reporting exact invalid exception Some architectures (like powerpc) provide status information on the exact type of invalid exception. This is pretty straight forward as we already report invalid exceptions via FP_SET_EXCEPTION. We add new flags (FP_EX_INVALID_*) the architecture code can define if it wants the exact invalid exception reported. We had to split out the INF/INF and 0/0 cases for divide to allow reporting the two invalid forms properly. Signed-off-by: Kumar Gala <galak@kernel.crashing.org> Acked-by: David S. Miller <davem@davemloft.net>
2008-06-27 18:39:00 +04:00
FP_SET_EXCEPTION(FP_EX_INVALID | FP_EX_INVALID_ISI); \
Linux-2.6.12-rc2 Initial git repository build. I'm not bothering with the full history, even though we have it. We can create a separate "historical" git archive of that later if we want to, and in the meantime it's about 3.2GB when imported into git - space that would just make the early git days unnecessarily complicated, when we don't have a lot of good infrastructure for it. Let it rip!
2005-04-17 02:20:36 +04:00
break; \
} \
/* FALLTHRU */ \
\
case _FP_CLS_COMBINE(FP_CLS_INF,FP_CLS_NORMAL): \
case _FP_CLS_COMBINE(FP_CLS_INF,FP_CLS_ZERO): \
R##_s = X##_s; \
R##_c = FP_CLS_INF; \
break; \
\
case _FP_CLS_COMBINE(FP_CLS_NORMAL,FP_CLS_INF): \
case _FP_CLS_COMBINE(FP_CLS_ZERO,FP_CLS_INF): \
R##_s = Y##_s; \
R##_c = FP_CLS_INF; \
break; \
\
case _FP_CLS_COMBINE(FP_CLS_ZERO,FP_CLS_ZERO): \
/* make sure the sign is correct */ \
if (FP_ROUNDMODE == FP_RND_MINF) \
R##_s = X##_s | Y##_s; \
else \
R##_s = X##_s & Y##_s; \
R##_c = FP_CLS_ZERO; \
break; \
\
default: \
abort(); \
} \
} while (0)
#define _FP_ADD(fs, wc, R, X, Y) _FP_ADD_INTERNAL(fs, wc, R, X, Y, '+')
#define _FP_SUB(fs, wc, R, X, Y) \
do { \
if (Y##_c != FP_CLS_NAN) Y##_s ^= 1; \
_FP_ADD_INTERNAL(fs, wc, R, X, Y, '-'); \
} while (0)
/*
* Main negation routine. FIXME -- when we care about setting exception
* bits reliably, this will not do. We should examine all of the fp classes.
*/
#define _FP_NEG(fs, wc, R, X) \
do { \
_FP_FRAC_COPY_##wc(R, X); \
R##_c = X##_c; \
R##_e = X##_e; \
R##_s = 1 ^ X##_s; \
} while (0)
/*
* Main multiplication routine. The input values should be cooked.
*/
#define _FP_MUL(fs, wc, R, X, Y) \
do { \
R##_s = X##_s ^ Y##_s; \
switch (_FP_CLS_COMBINE(X##_c, Y##_c)) \
{ \
case _FP_CLS_COMBINE(FP_CLS_NORMAL,FP_CLS_NORMAL): \
R##_c = FP_CLS_NORMAL; \
R##_e = X##_e + Y##_e + 1; \
\
_FP_MUL_MEAT_##fs(R,X,Y); \
\
if (_FP_FRAC_OVERP_##wc(fs, R)) \
_FP_FRAC_SRS_##wc(R, 1, _FP_WFRACBITS_##fs); \
else \
R##_e--; \
break; \
\
case _FP_CLS_COMBINE(FP_CLS_NAN,FP_CLS_NAN): \
_FP_CHOOSENAN(fs, wc, R, X, Y, '*'); \
break; \
\
case _FP_CLS_COMBINE(FP_CLS_NAN,FP_CLS_NORMAL): \
case _FP_CLS_COMBINE(FP_CLS_NAN,FP_CLS_INF): \
case _FP_CLS_COMBINE(FP_CLS_NAN,FP_CLS_ZERO): \
R##_s = X##_s; \
\
case _FP_CLS_COMBINE(FP_CLS_INF,FP_CLS_INF): \
case _FP_CLS_COMBINE(FP_CLS_INF,FP_CLS_NORMAL): \
case _FP_CLS_COMBINE(FP_CLS_ZERO,FP_CLS_NORMAL): \
case _FP_CLS_COMBINE(FP_CLS_ZERO,FP_CLS_ZERO): \
_FP_FRAC_COPY_##wc(R, X); \
R##_c = X##_c; \
break; \
\
case _FP_CLS_COMBINE(FP_CLS_NORMAL,FP_CLS_NAN): \
case _FP_CLS_COMBINE(FP_CLS_INF,FP_CLS_NAN): \
case _FP_CLS_COMBINE(FP_CLS_ZERO,FP_CLS_NAN): \
R##_s = Y##_s; \
\
case _FP_CLS_COMBINE(FP_CLS_NORMAL,FP_CLS_INF): \
case _FP_CLS_COMBINE(FP_CLS_NORMAL,FP_CLS_ZERO): \
_FP_FRAC_COPY_##wc(R, Y); \
R##_c = Y##_c; \
break; \
\
case _FP_CLS_COMBINE(FP_CLS_INF,FP_CLS_ZERO): \
case _FP_CLS_COMBINE(FP_CLS_ZERO,FP_CLS_INF): \
R##_s = _FP_NANSIGN_##fs; \
R##_c = FP_CLS_NAN; \
_FP_FRAC_SET_##wc(R, _FP_NANFRAC_##fs); \
math-emu: Add support for reporting exact invalid exception Some architectures (like powerpc) provide status information on the exact type of invalid exception. This is pretty straight forward as we already report invalid exceptions via FP_SET_EXCEPTION. We add new flags (FP_EX_INVALID_*) the architecture code can define if it wants the exact invalid exception reported. We had to split out the INF/INF and 0/0 cases for divide to allow reporting the two invalid forms properly. Signed-off-by: Kumar Gala <galak@kernel.crashing.org> Acked-by: David S. Miller <davem@davemloft.net>
2008-06-27 18:39:00 +04:00
FP_SET_EXCEPTION(FP_EX_INVALID | FP_EX_INVALID_IMZ);\
Linux-2.6.12-rc2 Initial git repository build. I'm not bothering with the full history, even though we have it. We can create a separate "historical" git archive of that later if we want to, and in the meantime it's about 3.2GB when imported into git - space that would just make the early git days unnecessarily complicated, when we don't have a lot of good infrastructure for it. Let it rip!
2005-04-17 02:20:36 +04:00
break; \
\
default: \
abort(); \
} \
} while (0)
/*
* Main division routine. The input values should be cooked.
*/
#define _FP_DIV(fs, wc, R, X, Y) \
do { \
R##_s = X##_s ^ Y##_s; \
switch (_FP_CLS_COMBINE(X##_c, Y##_c)) \
{ \
case _FP_CLS_COMBINE(FP_CLS_NORMAL,FP_CLS_NORMAL): \
R##_c = FP_CLS_NORMAL; \
R##_e = X##_e - Y##_e; \
\
_FP_DIV_MEAT_##fs(R,X,Y); \
break; \
\
case _FP_CLS_COMBINE(FP_CLS_NAN,FP_CLS_NAN): \
_FP_CHOOSENAN(fs, wc, R, X, Y, '/'); \
break; \
\
case _FP_CLS_COMBINE(FP_CLS_NAN,FP_CLS_NORMAL): \
case _FP_CLS_COMBINE(FP_CLS_NAN,FP_CLS_INF): \
case _FP_CLS_COMBINE(FP_CLS_NAN,FP_CLS_ZERO): \
R##_s = X##_s; \
_FP_FRAC_COPY_##wc(R, X); \
R##_c = X##_c; \
break; \
\
case _FP_CLS_COMBINE(FP_CLS_NORMAL,FP_CLS_NAN): \
case _FP_CLS_COMBINE(FP_CLS_INF,FP_CLS_NAN): \
case _FP_CLS_COMBINE(FP_CLS_ZERO,FP_CLS_NAN): \
R##_s = Y##_s; \
_FP_FRAC_COPY_##wc(R, Y); \
R##_c = Y##_c; \
break; \
\
case _FP_CLS_COMBINE(FP_CLS_NORMAL,FP_CLS_INF): \
case _FP_CLS_COMBINE(FP_CLS_ZERO,FP_CLS_INF): \
case _FP_CLS_COMBINE(FP_CLS_ZERO,FP_CLS_NORMAL): \
R##_c = FP_CLS_ZERO; \
break; \
\
case _FP_CLS_COMBINE(FP_CLS_NORMAL,FP_CLS_ZERO): \
FP_SET_EXCEPTION(FP_EX_DIVZERO); \
case _FP_CLS_COMBINE(FP_CLS_INF,FP_CLS_ZERO): \
case _FP_CLS_COMBINE(FP_CLS_INF,FP_CLS_NORMAL): \
R##_c = FP_CLS_INF; \
break; \
\
case _FP_CLS_COMBINE(FP_CLS_INF,FP_CLS_INF): \
math-emu: Add support for reporting exact invalid exception Some architectures (like powerpc) provide status information on the exact type of invalid exception. This is pretty straight forward as we already report invalid exceptions via FP_SET_EXCEPTION. We add new flags (FP_EX_INVALID_*) the architecture code can define if it wants the exact invalid exception reported. We had to split out the INF/INF and 0/0 cases for divide to allow reporting the two invalid forms properly. Signed-off-by: Kumar Gala <galak@kernel.crashing.org> Acked-by: David S. Miller <davem@davemloft.net>
2008-06-27 18:39:00 +04:00
R##_s = _FP_NANSIGN_##fs; \
R##_c = FP_CLS_NAN; \
_FP_FRAC_SET_##wc(R, _FP_NANFRAC_##fs); \
FP_SET_EXCEPTION(FP_EX_INVALID | FP_EX_INVALID_IDI);\
math-emu: Fix thinko in _FP_DIV In commit 48d6c64311ddb6417b901639530ccbc47bdc7635 ("math-emu: Add support for reporting exact invalid exception") code was added to set the new FP_EX_INVALID_{IDI,ZDZ} exception flag bits. However there is a missing break statement for the _FP_CLS_COMBINE(FP_CLS_INF,FP_CLS_INF) switch case, the code just falls into _FP_CLS_COMBINE(FP_CLS_ZERO,FP_CLS_ZERO) which then proceeds to overwrite all of the settings. Fix by adding the missing break. Signed-off-by: David S. Miller <davem@davemloft.net>
2008-10-23 09:09:59 +04:00
break; \
\
Linux-2.6.12-rc2 Initial git repository build. I'm not bothering with the full history, even though we have it. We can create a separate "historical" git archive of that later if we want to, and in the meantime it's about 3.2GB when imported into git - space that would just make the early git days unnecessarily complicated, when we don't have a lot of good infrastructure for it. Let it rip!
2005-04-17 02:20:36 +04:00
case _FP_CLS_COMBINE(FP_CLS_ZERO,FP_CLS_ZERO): \
R##_s = _FP_NANSIGN_##fs; \
R##_c = FP_CLS_NAN; \
_FP_FRAC_SET_##wc(R, _FP_NANFRAC_##fs); \
math-emu: Add support for reporting exact invalid exception Some architectures (like powerpc) provide status information on the exact type of invalid exception. This is pretty straight forward as we already report invalid exceptions via FP_SET_EXCEPTION. We add new flags (FP_EX_INVALID_*) the architecture code can define if it wants the exact invalid exception reported. We had to split out the INF/INF and 0/0 cases for divide to allow reporting the two invalid forms properly. Signed-off-by: Kumar Gala <galak@kernel.crashing.org> Acked-by: David S. Miller <davem@davemloft.net>
2008-06-27 18:39:00 +04:00
FP_SET_EXCEPTION(FP_EX_INVALID | FP_EX_INVALID_ZDZ);\
Linux-2.6.12-rc2 Initial git repository build. I'm not bothering with the full history, even though we have it. We can create a separate "historical" git archive of that later if we want to, and in the meantime it's about 3.2GB when imported into git - space that would just make the early git days unnecessarily complicated, when we don't have a lot of good infrastructure for it. Let it rip!
2005-04-17 02:20:36 +04:00
break; \
\
default: \
abort(); \
} \
} while (0)
/*
* Main differential comparison routine. The inputs should be raw not
* cooked. The return is -1,0,1 for normal values, 2 otherwise.
*/
#define _FP_CMP(fs, wc, ret, X, Y, un) \
do { \
/* NANs are unordered */ \
if ((X##_e == _FP_EXPMAX_##fs && !_FP_FRAC_ZEROP_##wc(X)) \
|| (Y##_e == _FP_EXPMAX_##fs && !_FP_FRAC_ZEROP_##wc(Y))) \
{ \
ret = un; \
} \
else \
{ \
int __is_zero_x; \
int __is_zero_y; \
\
__is_zero_x = (!X##_e && _FP_FRAC_ZEROP_##wc(X)) ? 1 : 0; \
__is_zero_y = (!Y##_e && _FP_FRAC_ZEROP_##wc(Y)) ? 1 : 0; \
\
if (__is_zero_x && __is_zero_y) \
ret = 0; \
else if (__is_zero_x) \
ret = Y##_s ? 1 : -1; \
else if (__is_zero_y) \
ret = X##_s ? -1 : 1; \
else if (X##_s != Y##_s) \
ret = X##_s ? -1 : 1; \
else if (X##_e > Y##_e) \
ret = X##_s ? -1 : 1; \
else if (X##_e < Y##_e) \
ret = X##_s ? 1 : -1; \
else if (_FP_FRAC_GT_##wc(X, Y)) \
ret = X##_s ? -1 : 1; \
else if (_FP_FRAC_GT_##wc(Y, X)) \
ret = X##_s ? 1 : -1; \
else \
ret = 0; \
} \
} while (0)
/* Simplification for strict equality. */
#define _FP_CMP_EQ(fs, wc, ret, X, Y) \
do { \
/* NANs are unordered */ \
if ((X##_e == _FP_EXPMAX_##fs && !_FP_FRAC_ZEROP_##wc(X)) \
|| (Y##_e == _FP_EXPMAX_##fs && !_FP_FRAC_ZEROP_##wc(Y))) \
{ \
ret = 1; \
} \
else \
{ \
ret = !(X##_e == Y##_e \
&& _FP_FRAC_EQ_##wc(X, Y) \
&& (X##_s == Y##_s || !X##_e && _FP_FRAC_ZEROP_##wc(X))); \
} \
} while (0)
/*
* Main square root routine. The input value should be cooked.
*/
#define _FP_SQRT(fs, wc, R, X) \
do { \
_FP_FRAC_DECL_##wc(T); _FP_FRAC_DECL_##wc(S); \
_FP_W_TYPE q; \
switch (X##_c) \
{ \
case FP_CLS_NAN: \
_FP_FRAC_COPY_##wc(R, X); \
R##_s = X##_s; \
R##_c = FP_CLS_NAN; \
break; \
case FP_CLS_INF: \
if (X##_s) \
{ \
R##_s = _FP_NANSIGN_##fs; \
R##_c = FP_CLS_NAN; /* NAN */ \
_FP_FRAC_SET_##wc(R, _FP_NANFRAC_##fs); \
FP_SET_EXCEPTION(FP_EX_INVALID); \
} \
else \
{ \
R##_s = 0; \
R##_c = FP_CLS_INF; /* sqrt(+inf) = +inf */ \
} \
break; \
case FP_CLS_ZERO: \
R##_s = X##_s; \
R##_c = FP_CLS_ZERO; /* sqrt(+-0) = +-0 */ \
break; \
case FP_CLS_NORMAL: \
R##_s = 0; \
if (X##_s) \
{ \
R##_c = FP_CLS_NAN; /* sNAN */ \
R##_s = _FP_NANSIGN_##fs; \
_FP_FRAC_SET_##wc(R, _FP_NANFRAC_##fs); \
FP_SET_EXCEPTION(FP_EX_INVALID); \
break; \
} \
R##_c = FP_CLS_NORMAL; \
if (X##_e & 1) \
_FP_FRAC_SLL_##wc(X, 1); \
R##_e = X##_e >> 1; \
_FP_FRAC_SET_##wc(S, _FP_ZEROFRAC_##wc); \
_FP_FRAC_SET_##wc(R, _FP_ZEROFRAC_##wc); \
q = _FP_OVERFLOW_##fs >> 1; \
_FP_SQRT_MEAT_##wc(R, S, T, X, q); \
} \
} while (0)
/*
* Convert from FP to integer
*/
/* RSIGNED can have following values:
* 0: the number is required to be 0..(2^rsize)-1, if not, NV is set plus
* the result is either 0 or (2^rsize)-1 depending on the sign in such case.
* 1: the number is required to be -(2^(rsize-1))..(2^(rsize-1))-1, if not, NV is
* set plus the result is either -(2^(rsize-1)) or (2^(rsize-1))-1 depending
* on the sign in such case.
* 2: the number is required to be -(2^(rsize-1))..(2^(rsize-1))-1, if not, NV is
* set plus the result is truncated to fit into destination.
* -1: the number is required to be -(2^(rsize-1))..(2^rsize)-1, if not, NV is
* set plus the result is either -(2^(rsize-1)) or (2^(rsize-1))-1 depending
* on the sign in such case.
*/
#define _FP_TO_INT(fs, wc, r, X, rsize, rsigned) \
do { \
switch (X##_c) \
{ \
case FP_CLS_NORMAL: \
if (X##_e < 0) \
{ \
FP_SET_EXCEPTION(FP_EX_INEXACT); \
case FP_CLS_ZERO: \
r = 0; \
} \
else if (X##_e >= rsize - (rsigned > 0 || X##_s) \
|| (!rsigned && X##_s)) \
{ /* overflow */ \
case FP_CLS_NAN: \
case FP_CLS_INF: \
if (rsigned == 2) \
{ \
if (X##_c != FP_CLS_NORMAL \
|| X##_e >= rsize - 1 + _FP_WFRACBITS_##fs) \
r = 0; \
else \
{ \
_FP_FRAC_SLL_##wc(X, (X##_e - _FP_WFRACBITS_##fs + 1)); \
_FP_FRAC_ASSEMBLE_##wc(r, X, rsize); \
} \
} \
else if (rsigned) \
{ \
r = 1; \
r <<= rsize - 1; \
r -= 1 - X##_s; \
} \
else \
{ \
r = 0; \
math-emu: fix floating-point to integer unsigned saturation The math-emu macros _FP_TO_INT and _FP_TO_INT_ROUND are supposed to saturate their results for out-of-range arguments, except in the case rsigned == 2 (when instead the low bits of the result are taken). However, in the case rsigned == 0 (converting to unsigned integers), they mistakenly produce 0 for positive results and the maximum unsigned integer for negative results, the opposite of correct unsigned saturation. This patch fixes the logic. Signed-off-by: Joseph Myers <joseph@codesourcery.com> Signed-off-by: Scott Wood <scottwood@freescale.com>
2013-11-04 20:53:50 +04:00
if (!X##_s) \
Linux-2.6.12-rc2 Initial git repository build. I'm not bothering with the full history, even though we have it. We can create a separate "historical" git archive of that later if we want to, and in the meantime it's about 3.2GB when imported into git - space that would just make the early git days unnecessarily complicated, when we don't have a lot of good infrastructure for it. Let it rip!
2005-04-17 02:20:36 +04:00
r = ~r; \
} \
FP_SET_EXCEPTION(FP_EX_INVALID); \
} \
else \
{ \
if (_FP_W_TYPE_SIZE*wc < rsize) \
{ \
_FP_FRAC_ASSEMBLE_##wc(r, X, rsize); \
r <<= X##_e - _FP_WFRACBITS_##fs; \
} \
else \
{ \
if (X##_e >= _FP_WFRACBITS_##fs) \
_FP_FRAC_SLL_##wc(X, (X##_e - _FP_WFRACBITS_##fs + 1)); \
else if (X##_e < _FP_WFRACBITS_##fs - 1) \
{ \
_FP_FRAC_SRS_##wc(X, (_FP_WFRACBITS_##fs - X##_e - 2), \
_FP_WFRACBITS_##fs); \
if (_FP_FRAC_LOW_##wc(X) & 1) \
FP_SET_EXCEPTION(FP_EX_INEXACT); \
_FP_FRAC_SRL_##wc(X, 1); \
} \
_FP_FRAC_ASSEMBLE_##wc(r, X, rsize); \
} \
if (rsigned && X##_s) \
r = -r; \
} \
break; \
} \
} while (0)
#define _FP_TO_INT_ROUND(fs, wc, r, X, rsize, rsigned) \
do { \
r = 0; \
switch (X##_c) \
{ \
case FP_CLS_NORMAL: \
if (X##_e >= _FP_FRACBITS_##fs - 1) \
{ \
if (X##_e < rsize - 1 + _FP_WFRACBITS_##fs) \
{ \
if (X##_e >= _FP_WFRACBITS_##fs - 1) \
{ \
_FP_FRAC_ASSEMBLE_##wc(r, X, rsize); \
r <<= X##_e - _FP_WFRACBITS_##fs + 1; \
} \
else \
{ \
_FP_FRAC_SRL_##wc(X, _FP_WORKBITS - X##_e \
+ _FP_FRACBITS_##fs - 1); \
_FP_FRAC_ASSEMBLE_##wc(r, X, rsize); \
} \
} \
} \
else \
{ \
math-emu: fix floating-point to integer overflow detection On overflow, the math-emu macro _FP_TO_INT_ROUND tries to saturate its result (subject to the value of rsigned specifying the desired overflow semantics). However, if the rounding step has the effect of increasing the exponent so as to cause overflow (if the rounded result is 1 larger than the largest positive value with the given number of bits, allowing for signedness), the overflow does not get detected, meaning that for unsigned results 0 is produced instead of the maximum unsigned integer with the give number of bits, without an exception being raised for overflow, and that for signed results the minimum (negative) value is produced instead of the maximum (positive) value, again without an exception. This patch makes the code check for rounding increasing the exponent and adjusts the exponent value as needed for the overflow check. Signed-off-by: Joseph Myers <joseph@codesourcery.com> Signed-off-by: Scott Wood <scottwood@freescale.com>
2013-11-04 20:54:15 +04:00
int _lz0, _lz1; \
Linux-2.6.12-rc2 Initial git repository build. I'm not bothering with the full history, even though we have it. We can create a separate "historical" git archive of that later if we want to, and in the meantime it's about 3.2GB when imported into git - space that would just make the early git days unnecessarily complicated, when we don't have a lot of good infrastructure for it. Let it rip!
2005-04-17 02:20:36 +04:00
if (X##_e <= -_FP_WORKBITS - 1) \
_FP_FRAC_SET_##wc(X, _FP_MINFRAC_##wc); \
else \
_FP_FRAC_SRS_##wc(X, _FP_FRACBITS_##fs - 1 - X##_e, \
_FP_WFRACBITS_##fs); \
math-emu: fix floating-point to integer overflow detection On overflow, the math-emu macro _FP_TO_INT_ROUND tries to saturate its result (subject to the value of rsigned specifying the desired overflow semantics). However, if the rounding step has the effect of increasing the exponent so as to cause overflow (if the rounded result is 1 larger than the largest positive value with the given number of bits, allowing for signedness), the overflow does not get detected, meaning that for unsigned results 0 is produced instead of the maximum unsigned integer with the give number of bits, without an exception being raised for overflow, and that for signed results the minimum (negative) value is produced instead of the maximum (positive) value, again without an exception. This patch makes the code check for rounding increasing the exponent and adjusts the exponent value as needed for the overflow check. Signed-off-by: Joseph Myers <joseph@codesourcery.com> Signed-off-by: Scott Wood <scottwood@freescale.com>
2013-11-04 20:54:15 +04:00
_FP_FRAC_CLZ_##wc(_lz0, X); \
Linux-2.6.12-rc2 Initial git repository build. I'm not bothering with the full history, even though we have it. We can create a separate "historical" git archive of that later if we want to, and in the meantime it's about 3.2GB when imported into git - space that would just make the early git days unnecessarily complicated, when we don't have a lot of good infrastructure for it. Let it rip!
2005-04-17 02:20:36 +04:00
_FP_ROUND(wc, X); \
math-emu: fix floating-point to integer overflow detection On overflow, the math-emu macro _FP_TO_INT_ROUND tries to saturate its result (subject to the value of rsigned specifying the desired overflow semantics). However, if the rounding step has the effect of increasing the exponent so as to cause overflow (if the rounded result is 1 larger than the largest positive value with the given number of bits, allowing for signedness), the overflow does not get detected, meaning that for unsigned results 0 is produced instead of the maximum unsigned integer with the give number of bits, without an exception being raised for overflow, and that for signed results the minimum (negative) value is produced instead of the maximum (positive) value, again without an exception. This patch makes the code check for rounding increasing the exponent and adjusts the exponent value as needed for the overflow check. Signed-off-by: Joseph Myers <joseph@codesourcery.com> Signed-off-by: Scott Wood <scottwood@freescale.com>
2013-11-04 20:54:15 +04:00
_FP_FRAC_CLZ_##wc(_lz1, X); \
if (_lz1 < _lz0) \
X##_e++; /* For overflow detection. */ \
Linux-2.6.12-rc2 Initial git repository build. I'm not bothering with the full history, even though we have it. We can create a separate "historical" git archive of that later if we want to, and in the meantime it's about 3.2GB when imported into git - space that would just make the early git days unnecessarily complicated, when we don't have a lot of good infrastructure for it. Let it rip!
2005-04-17 02:20:36 +04:00
_FP_FRAC_SRL_##wc(X, _FP_WORKBITS); \
_FP_FRAC_ASSEMBLE_##wc(r, X, rsize); \
} \
if (rsigned && X##_s) \
r = -r; \
if (X##_e >= rsize - (rsigned > 0 || X##_s) \
|| (!rsigned && X##_s)) \
{ /* overflow */ \
case FP_CLS_NAN: \
case FP_CLS_INF: \
if (!rsigned) \
{ \
r = 0; \
math-emu: fix floating-point to integer unsigned saturation The math-emu macros _FP_TO_INT and _FP_TO_INT_ROUND are supposed to saturate their results for out-of-range arguments, except in the case rsigned == 2 (when instead the low bits of the result are taken). However, in the case rsigned == 0 (converting to unsigned integers), they mistakenly produce 0 for positive results and the maximum unsigned integer for negative results, the opposite of correct unsigned saturation. This patch fixes the logic. Signed-off-by: Joseph Myers <joseph@codesourcery.com> Signed-off-by: Scott Wood <scottwood@freescale.com>
2013-11-04 20:53:50 +04:00
if (!X##_s) \
Linux-2.6.12-rc2 Initial git repository build. I'm not bothering with the full history, even though we have it. We can create a separate "historical" git archive of that later if we want to, and in the meantime it's about 3.2GB when imported into git - space that would just make the early git days unnecessarily complicated, when we don't have a lot of good infrastructure for it. Let it rip!
2005-04-17 02:20:36 +04:00
r = ~r; \
} \
else if (rsigned != 2) \
{ \
r = 1; \
r <<= rsize - 1; \
r -= 1 - X##_s; \
} \
FP_SET_EXCEPTION(FP_EX_INVALID); \
} \
break; \
case FP_CLS_ZERO: \
break; \
} \
} while (0)
#define _FP_FROM_INT(fs, wc, X, r, rsize, rtype) \
do { \
if (r) \
{ \
unsigned rtype ur_; \
X##_c = FP_CLS_NORMAL; \
\
if ((X##_s = (r < 0))) \
ur_ = (unsigned rtype) -r; \
else \
ur_ = (unsigned rtype) r; \
if (rsize <= _FP_W_TYPE_SIZE) \
__FP_CLZ(X##_e, ur_); \
else \
__FP_CLZ_2(X##_e, (_FP_W_TYPE)(ur_ >> _FP_W_TYPE_SIZE), \
(_FP_W_TYPE)ur_); \
if (rsize < _FP_W_TYPE_SIZE) \
X##_e -= (_FP_W_TYPE_SIZE - rsize); \
X##_e = rsize - X##_e - 1; \
\
math-emu: correct test for downshifting fraction in _FP_FROM_INT() The kernel's math-emu code contains a macro _FP_FROM_INT() which is used to convert an integer to a raw normalized floating-point value. It does this basically in three steps: 1. Compute the exponent from the number of leading zero bits. 2. Downshift large fractions to put the MSB in the right position for normalized fractions. 3. Upshift small fractions to put the MSB in the right position. There is an boundary error in step 2, causing a fraction with its MSB exactly one bit above the normalized MSB position to not be downshifted. This results in a non-normalized raw float, which when packed becomes a massively inaccurate representation for that input. The impact of this depends on a number of arch-specific factors, but it is known to have broken emulation of FXTOD instructions on UltraSPARC III, which was originally reported as GCC bug 44631 <http://gcc.gnu.org/bugzilla/show_bug.cgi?id=44631>. Any arch which uses math-emu to emulate conversions from integers to same-size floats may be affected. The fix is simple: the exponent comparison used to determine if the fraction should be downshifted must be "<=" not "<". I'm sending a kernel module to test this as a reply to this message. There are also SPARC user-space test cases in the GCC bug entry. Signed-off-by: Mikael Pettersson <mikpe@it.uu.se> Signed-off-by: David S. Miller <davem@davemloft.net>
2010-07-21 05:45:14 +04:00
if (_FP_FRACBITS_##fs < rsize && _FP_WFRACBITS_##fs <= X##_e) \
Linux-2.6.12-rc2 Initial git repository build. I'm not bothering with the full history, even though we have it. We can create a separate "historical" git archive of that later if we want to, and in the meantime it's about 3.2GB when imported into git - space that would just make the early git days unnecessarily complicated, when we don't have a lot of good infrastructure for it. Let it rip!
2005-04-17 02:20:36 +04:00
__FP_FRAC_SRS_1(ur_, (X##_e - _FP_WFRACBITS_##fs + 1), rsize);\
_FP_FRAC_DISASSEMBLE_##wc(X, ur_, rsize); \
if ((_FP_WFRACBITS_##fs - X##_e - 1) > 0) \
_FP_FRAC_SLL_##wc(X, (_FP_WFRACBITS_##fs - X##_e - 1)); \
} \
else \
{ \
X##_c = FP_CLS_ZERO, X##_s = 0; \
} \
} while (0)
#define FP_CONV(dfs,sfs,dwc,swc,D,S) \
do { \
_FP_FRAC_CONV_##dwc##_##swc(dfs, sfs, D, S); \
D##_e = S##_e; \
D##_c = S##_c; \
D##_s = S##_s; \
} while (0)
/*
* Helper primitives.
*/
/* Count leading zeros in a word. */
#ifndef __FP_CLZ
#if _FP_W_TYPE_SIZE < 64
/* this is just to shut the compiler up about shifts > word length -- PMM 02/1998 */
#define __FP_CLZ(r, x) \
do { \
_FP_W_TYPE _t = (x); \
r = _FP_W_TYPE_SIZE - 1; \
if (_t > 0xffff) r -= 16; \
if (_t > 0xffff) _t >>= 16; \
if (_t > 0xff) r -= 8; \
if (_t > 0xff) _t >>= 8; \
if (_t & 0xf0) r -= 4; \
if (_t & 0xf0) _t >>= 4; \
if (_t & 0xc) r -= 2; \
if (_t & 0xc) _t >>= 2; \
if (_t & 0x2) r -= 1; \
} while (0)
#else /* not _FP_W_TYPE_SIZE < 64 */
#define __FP_CLZ(r, x) \
do { \
_FP_W_TYPE _t = (x); \
r = _FP_W_TYPE_SIZE - 1; \
if (_t > 0xffffffff) r -= 32; \
if (_t > 0xffffffff) _t >>= 32; \
if (_t > 0xffff) r -= 16; \
if (_t > 0xffff) _t >>= 16; \
if (_t > 0xff) r -= 8; \
if (_t > 0xff) _t >>= 8; \
if (_t & 0xf0) r -= 4; \
if (_t & 0xf0) _t >>= 4; \
if (_t & 0xc) r -= 2; \
if (_t & 0xc) _t >>= 2; \
if (_t & 0x2) r -= 1; \
} while (0)
#endif /* not _FP_W_TYPE_SIZE < 64 */
#endif /* ndef __FP_CLZ */
#define _FP_DIV_HELP_imm(q, r, n, d) \
do { \
q = n / d, r = n % d; \
} while (0)
#endif /* __MATH_EMU_OP_COMMON_H__ */
Reference in New Issue Copy Permalink