The existing check was completely wrong, confused about the definition of the (previous) float_flag_{input,output}_denormal flags, then making sure that DN, the flush-to-zero bit, was off.
Update for the introduction of float_flag_inorm_denormal and float_flag_result_denormal, taking into account that DN now sets the softfloat flush-to-zero bits. Cc: Philippe Mathieu-Daudé <f4...@amsat.org> Cc: Yoshinori Sato <ys...@users.sourceforge.jp> Signed-off-by: Richard Henderson <richard.hender...@linaro.org> --- target/rx/op_helper.c | 8 +++++--- 1 file changed, 5 insertions(+), 3 deletions(-) diff --git a/target/rx/op_helper.c b/target/rx/op_helper.c index b1772e9538..c2e4f9a5e3 100644 --- a/target/rx/op_helper.c +++ b/target/rx/op_helper.c @@ -97,9 +97,11 @@ static void update_fpsw(CPURXState *env, float32 ret, uintptr_t retaddr) if (xcpt & float_flag_inexact) { SET_FPSW(X); } - if ((xcpt & (float_flag_iflush_denormal - | float_flag_oflush_denormal)) - && !FIELD_EX32(env->fpsw, FPSW, DN)) { + /* + * If any input or output denormals, not flushed to zero, raise CE: + * unimplemented processing has been encountered. + */ + if (xcpt & (float_flag_inorm_denormal | float_flag_result_denormal)) { env->fpsw = FIELD_DP32(env->fpsw, FPSW, CE, 1); } -- 2.25.1