motorhead/qemu
1
0
Fork 0
mirror of https://github.com/Motorhead1991/qemu.git synced 2025-12-24 00:18:36 -07:00
Code Issues Projects Releases Packages Wiki Activity Actions

softfloat: Allow 2-operand NaN propagation rule to be set at runtime

Browse source 
IEEE 758 does not define a fixed rule for which NaN to pick as the
result if both operands of a 2-operand operation are NaNs.  As a
result different architectures have ended up with different rules for
propagating NaNs.

QEMU currently hardcodes the NaN propagation logic into the binary
because pickNaN() has an ifdef ladder for different targets.  We want
to make the propagation rule instead be selectable at runtime,
because:
 * this will let us have multiple targets in one QEMU binary
 * the Arm FEAT_AFP architectural feature includes letting
   the guest select a NaN propagation rule at runtime
 * x86 specifies different propagation rules for x87 FPU ops
   and for SSE ops, and specifying the rule in the float_status
   would let us emulate this, instead of wrongly using the
   x87 rules everywhere

In this commit we add an enum for the propagation rule, the field in
float_status, and the corresponding getters and setters.  We change
pickNaN to honour this, but because all targets still leave this
field at its default 0 value, the fallback logic will pick the rule
type with the old ifdef ladder.

It's valid not to set a propagation rule if default_nan_mode is
enabled, because in that case there's no need to pick a NaN; all the
callers of pickNaN() catch this case and skip calling it.  So we can
already assert that we don't get into the "no rule defined" codepath
for our four targets which always set default_nan_mode: Hexagon,
RiscV, SH4 and Tricore, and for the one target which does not have FP
at all: avr.  These targets will not need to be updated to call
set_float_2nan_prop_rule().

Signed-off-by: Peter Maydell <peter.maydell@linaro.org>
Reviewed-by: Richard Henderson <richard.henderson@linaro.org>
Message-id: 20241025141254.2141506-2-peter.maydell@linaro.org
This commit is contained in:
Peter Maydell 2024-11-05 10:09:52 +00:00
parent 11b8920ed2
commit 8a132968b2
3 changed files with 185 additions and 97 deletions
Download patch file Download diff file Expand all files Collapse all files

11
include/fpu/softfloat-helpers.h
View file

@ -75,6 +75,12 @@ static inline void set_floatx80_rounding_precision(FloatX80RoundPrec val,
status->floatx80_rounding_precision = val;
}
static inline void set_float_2nan_prop_rule(Float2NaNPropRule rule,
float_status *status)
{
status->float_2nan_prop_rule = rule;
}
static inline void set_flush_to_zero(bool val, float_status *status)
{
status->flush_to_zero = val;
@ -126,6 +132,11 @@ get_floatx80_rounding_precision(float_status *status)
return status->floatx80_rounding_precision;
}
static inline Float2NaNPropRule get_float_2nan_prop_rule(float_status *status)
{
return status->float_2nan_prop_rule;
}
static inline bool get_flush_to_zero(float_status *status)
{
return status->flush_to_zero;