fpu: allow flushing of output denormals to be after rounding

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>

2 files changed, 29 insertions, 0 deletions

diff --git a/include/fpu/softfloat-helpers.h b/include/fpu/softfloat-helpers.h
index 4cb30a4822..8983c2748e 100644
--- a/include/fpu/softfloat-helpers.h
+++ b/include/fpu/softfloat-helpers.h
@@ -109,6 +109,12 @@ static inline void set_flush_inputs_to_zero(bool val, float_status *status)
     status->flush_inputs_to_zero = val;
 }
 
+static inline void set_float_ftz_detection(FloatFTZDetection d,
+                                           float_status *status)
+{
+    status->ftz_detection = d;
+}
+
 static inline void set_default_nan_mode(bool val, float_status *status)
 {
     status->default_nan_mode = val;
@@ -183,4 +189,9 @@ static inline bool get_default_nan_mode(const float_status *status)
     return status->default_nan_mode;
 }
 
+static inline FloatFTZDetection get_float_ftz_detection(const float_status *status)
+{
+    return status->ftz_detection;
+}
+
 #endif /* SOFTFLOAT_HELPERS_H */
diff --git a/include/fpu/softfloat-types.h b/include/fpu/softfloat-types.h
index bba1c397bb..53d5eb8521 100644
--- a/include/fpu/softfloat-types.h
+++ b/include/fpu/softfloat-types.h
@@ -305,6 +305,22 @@ typedef enum __attribute__((__packed__)) {
 } FloatInfZeroNaNRule;
 
 /*
+ * When flush_to_zero is set, should we detect denormal results to
+ * be flushed before or after rounding? For most architectures this
+ * should be set to match the tininess_before_rounding setting,
+ * but a few architectures, e.g. MIPS MSA, detect FTZ before
+ * rounding but tininess after rounding.
+ *
+ * This enum is arranged so that the default if the target doesn't
+ * configure it matches the default for tininess_before_rounding
+ * (i.e. "after rounding").
+ */
+typedef enum __attribute__((__packed__)) {
+    float_ftz_after_rounding = 0,
+    float_ftz_before_rounding = 1,
+} FloatFTZDetection;
+
+/*
  * Floating Point Status. Individual architectures may maintain
  * several versions of float_status for different functions. The
  * correct status for the operation is then passed by reference to
@@ -321,6 +337,8 @@ typedef struct float_status {
     bool tininess_before_rounding;
     /* should denormalised results go to zero and set output_denormal_flushed? */
     bool flush_to_zero;
+    /* do we detect and flush denormal results before or after rounding? */
+    FloatFTZDetection ftz_detection;
     /* should denormalised inputs go to zero and set input_denormal_flushed? */
     bool flush_inputs_to_zero;
     bool default_nan_mode;