Fixes and improvements in the floating point exception
handling in the hppa system and user emulation.
-----BEGIN PGP SIGNATURE-----
iHUEABYKAB0WIQS86RI+GtKfB8BJu973ErUQojoPXwUCaC7b/QAKCRD3ErUQojoP
X8IPAQCZGjG3+iwJJuJkL6dSu7WKWgidX16zIxeRYQl+FXTpkAD9FBQ/nDP+EAFN
5hQTeiixyj3xIsZHPeHwfzhWiBUwQAo=
=Q3zi
-----END PGP SIGNATURE-----
Merge tag 'hppa-fpe-fixup-pull-request' of https://github.com/hdeller/qemu-hppa into staging
HPPA floating point exception handling fixes
Fixes and improvements in the floating point exception
handling in the hppa system and user emulation.
# -----BEGIN PGP SIGNATURE-----
#
# iHUEABYKAB0WIQS86RI+GtKfB8BJu973ErUQojoPXwUCaC7b/QAKCRD3ErUQojoP
# X8IPAQCZGjG3+iwJJuJkL6dSu7WKWgidX16zIxeRYQl+FXTpkAD9FBQ/nDP+EAFN
# 5hQTeiixyj3xIsZHPeHwfzhWiBUwQAo=
# =Q3zi
# -----END PGP SIGNATURE-----
# gpg: Signature made Thu 22 May 2025 04:10:37 EDT
# gpg: using EDDSA key BCE9123E1AD29F07C049BBDEF712B510A23A0F5F
# gpg: Good signature from "Helge Deller <deller@gmx.de>" [unknown]
# gpg: aka "Helge Deller <deller@kernel.org>" [unknown]
# gpg: WARNING: This key is not certified with a trusted signature!
# gpg: There is no indication that the signature belongs to the owner.
# Primary key fingerprint: 4544 8228 2CD9 10DB EF3D 25F8 3E5F 3D04 A7A2 4603
# Subkey fingerprint: BCE9 123E 1AD2 9F07 C049 BBDE F712 B510 A23A 0F5F
* tag 'hppa-fpe-fixup-pull-request' of https://github.com/hdeller/qemu-hppa:
target/hppa: Fix FPE exceptions
linux-user/hppa: Send proper si_code on SIGFPE exception
target/hppa: Copy instruction code into fr1 on FPU assist fault
Signed-off-by: Stefan Hajnoczi <stefanha@redhat.com>
Implement FP exception register #1 (lower 32-bits of 64-bit fr[0]).
A proper implementation is necessary to allow the Linux kernel in
system mode and the qemu linux-user to send proper si_code values
on SIGFPE signal.
Always set the T-bit on taken exception, and merge over- and underflow
in system mode to just set overflow bit to mimic the behaviour I tested
on a physical machine.
The test program below can be used to verify correct behaviour. Note
that behaviour on SIGFPE may vary on different platforms. The program
should always detect the correct signal, but it may or may not be able
to sucessfully continue afterwards.
#define _GNU_SOURCE
#include <signal.h>
#include <stdio.h>
#include <fenv.h>
#include <float.h>
static void fpe_func(int sig, siginfo_t *i, void *v) {
sigset_t set;
sigemptyset(&set);
sigaddset(&set, SIGFPE);
sigprocmask(SIG_UNBLOCK, &set, NULL);
printf("GOT signal %d with si_code %ld\n", sig, i->si_code);
}
int main(int argc, char *argv[]) {
struct sigaction action = {
.sa_sigaction = fpe_func,
.sa_flags = SA_RESTART|SA_SIGINFO };
sigaction(SIGFPE, &action, 0);
feenableexcept(FE_OVERFLOW | FE_UNDERFLOW);
double x = DBL_MIN;
return printf("%lf\n", argc > 1
? 1.7976931348623158E308*1.7976931348623158E308
: x / 10);
}
Signed-off-by: Helge Deller <deller@gmx.de>
"exec/exec-all.h" is now fully empty, let's remove it.
Mechanical change running:
$ sed -i '/exec\/exec-all.h/d' $(git grep -wl exec/exec-all.h)
Signed-off-by: Philippe Mathieu-Daudé <philmd@linaro.org>
Reviewed-by: Mark Cave-Ayland <mark.caveayland@nutanix.com>
Signed-off-by: Richard Henderson <richard.henderson@linaro.org>
Message-ID: <20250424202412.91612-14-philmd@linaro.org>
Currently we have a compile-time shortcut where we return a hardcode
value from snan_bit_is_one() on everything except MIPS, because we
know that's the only target that needs to change
status->no_signaling_nans at runtime.
Remove the ifdef, so we always look at the status flag. This means
we must update the two targets (HPPA and SH4) that were previously
hardcoded to return true so that they set the status flag correctly.
This has no behavioural change, but will be necessary if we want to
build softfloat once for all targets.
Signed-off-by: Peter Maydell <peter.maydell@linaro.org>
Reviewed-by: Philippe Mathieu-Daudé <philmd@linaro.org>
Reviewed-by: Richard Henderson <richard.henderson@linaro.org>
Message-id: 20250224111524.1101196-11-peter.maydell@linaro.org
Message-id: 20250217125055.160887-9-peter.maydell@linaro.org
Currently we handle flushing of output denormals in uncanon_normal
always before we deal with rounding. This works for architectures
that detect tininess before rounding, but is usually not the right
place when the architecture detects tininess after rounding. For
example, for x86 the SDM states that the MXCSR FTZ control bit causes
outputs to be flushed to zero "when it detects a floating-point
underflow condition". This means that we mustn't flush to zero if
the input is such that after rounding it is no longer tiny.
At least one of our guest architectures does underflow detection
after rounding but flushing of denormals before rounding (MIPS MSA);
this means we need to have a config knob for this that is separate
from our existing tininess_before_rounding setting.
Add an ftz_detection flag. For consistency with
tininess_before_rounding, we make it default to "detect ftz after
rounding"; this means that we need to explicitly set the flag to
"detect ftz before rounding" on every existing architecture that sets
flush_to_zero, so that this commit has no behaviour change.
(This means more code change here but for the long term a less
confusing API.)
For several architectures the current behaviour is either
definitely or possibly wrong; annotate those with TODO comments.
These architectures are definitely wrong (and should detect
ftz after rounding):
* x86
* Alpha
For these architectures the spec is unclear:
* MIPS (for non-MSA)
* RX
* SH4
PA-RISC makes ftz detection IMPDEF, but we aren't setting the
"tininess before rounding" setting that we ought to.
Signed-off-by: Peter Maydell <peter.maydell@linaro.org>
Reviewed-by: Richard Henderson <richard.henderson@linaro.org>
Set the default NaN pattern explicitly, and remove the ifdef from
parts64_default_nan().
Signed-off-by: Peter Maydell <peter.maydell@linaro.org>
Reviewed-by: Richard Henderson <richard.henderson@linaro.org>
Message-id: 20241202131347.498124-39-peter.maydell@linaro.org
Set the Float3NaNPropRule explicitly for HPPA, and remove the
ifdef from pickNaNMulAdd().
HPPA is the only target that was using the default branch of the
ifdef ladder (other targets either do not use muladd or set
default_nan_mode), so we can remove the ifdef fallback entirely now
(allowing the "rule not set" case to fall into the default of the
switch statement and assert).
We add a TODO note that the HPPA rule is probably wrong; this is
not a behavioural change for this refactoring.
Signed-off-by: Peter Maydell <peter.maydell@linaro.org>
Reviewed-by: Richard Henderson <richard.henderson@linaro.org>
Message-id: 20241202131347.498124-26-peter.maydell@linaro.org
Set the FloatInfZeroNaNRule explicitly for the HPPA target,
so we can remove the ifdef from pickNaNMulAdd().
As this is the last target to be converted to explicitly setting
the rule, we can remove the fallback code in pickNaNMulAdd()
entirely.
Signed-off-by: Peter Maydell <peter.maydell@linaro.org>
Reviewed-by: Richard Henderson <richard.henderson@linaro.org>
Message-id: 20241202131347.498124-14-peter.maydell@linaro.org
Set the 2-NaN propagation rule explicitly in env->fp_status.
Really we only need to do this at CPU reset (after reset has zeroed
out most of the CPU state struct, which typically includes fp_status
fields). However target/hppa does not currently implement CPU reset
at all, so leave a TODO comment to note that this could be moved if
we ever do implement reset.
Signed-off-by: Peter Maydell <peter.maydell@linaro.org>
Reviewed-by: Richard Henderson <richard.henderson@linaro.org>
Message-id: 20241025141254.2141506-7-peter.maydell@linaro.org
Define all of the context dependent field definitions.
Use FIELD_EX32 and FIELD_DP32 with named fields instead
of extract32 and deposit32 with raw constants.
Reviewed-by: Helge Deller <deller@gmx.de>
Signed-off-by: Richard Henderson <richard.henderson@linaro.org>
Signed-off-by: Philippe Mathieu-Daudé <philmd@linaro.org>
Reviewed-by: Richard Henderson <richard.henderson@linaro.org>
Message-Id: <20221217173219.8715-2-philmd@linaro.org>
