| Message ID | 20221005121551.27957-1-victor.colombo@eldorado.org.br (mailing list archive) |
|---|---|
| State | New, archived |
| Headers | show |
| Series | target/ppc: Fix xvcmp* clearing FI bit | expand |
On 10/5/22 05:15, Víctor Colombo wrote: > Vector instructions in general are not supposed to change the FI bit. > However, xvcmp* instructions are calling gen_helper_float_check_status, > which is leading to a cleared FI flag where it should be kept > unchanged. > As helper_float_check_status only affects inexact, overflow and > underflow, and the xvcmp* instructions don't change these flags, this > issue can be fixed by removing the call to helper_float_check_status. > By doing this, the FI bit in FPSCR will be preserved as expected. > > Fixes: 00084a25adf ("target/ppc: introduce separate VSX_CMP macro for xvcmp* instructions") > Signed-off-by: Víctor Colombo<victor.colombo@eldorado.org.br> > --- > target/ppc/translate/vsx-impl.c.inc | 1 - > 1 file changed, 1 deletion(-) Reviewed-by: Richard Henderson <richard.henderson@linaro.org> r~
Queued in gitlab.com/danielhb/qemu/tree/ppc-next. Thanks, Daniel On 10/5/22 09:15, Víctor Colombo wrote: > Vector instructions in general are not supposed to change the FI bit. > However, xvcmp* instructions are calling gen_helper_float_check_status, > which is leading to a cleared FI flag where it should be kept > unchanged. > As helper_float_check_status only affects inexact, overflow and > underflow, and the xvcmp* instructions don't change these flags, this > issue can be fixed by removing the call to helper_float_check_status. > By doing this, the FI bit in FPSCR will be preserved as expected. > > Fixes: 00084a25adf ("target/ppc: introduce separate VSX_CMP macro for xvcmp* instructions") > Signed-off-by: Víctor Colombo <victor.colombo@eldorado.org.br> > --- > target/ppc/translate/vsx-impl.c.inc | 1 - > 1 file changed, 1 deletion(-) > > diff --git a/target/ppc/translate/vsx-impl.c.inc b/target/ppc/translate/vsx-impl.c.inc > index 7acdbceec4..e6e5c45ffd 100644 > --- a/target/ppc/translate/vsx-impl.c.inc > +++ b/target/ppc/translate/vsx-impl.c.inc > @@ -810,7 +810,6 @@ static void gen_##name(DisasContext *ctx) \ > gen_helper_##name(ignored, cpu_env, xt, xa, xb); \ > tcg_temp_free_i32(ignored); \ > } \ > - gen_helper_float_check_status(cpu_env); \ > tcg_temp_free_ptr(xt); \ > tcg_temp_free_ptr(xa); \ > tcg_temp_free_ptr(xb); \
diff --git a/target/ppc/translate/vsx-impl.c.inc b/target/ppc/translate/vsx-impl.c.inc index 7acdbceec4..e6e5c45ffd 100644 --- a/target/ppc/translate/vsx-impl.c.inc +++ b/target/ppc/translate/vsx-impl.c.inc @@ -810,7 +810,6 @@ static void gen_##name(DisasContext *ctx) \ gen_helper_##name(ignored, cpu_env, xt, xa, xb); \ tcg_temp_free_i32(ignored); \ } \ - gen_helper_float_check_status(cpu_env); \ tcg_temp_free_ptr(xt); \ tcg_temp_free_ptr(xa); \ tcg_temp_free_ptr(xb); \
Vector instructions in general are not supposed to change the FI bit. However, xvcmp* instructions are calling gen_helper_float_check_status, which is leading to a cleared FI flag where it should be kept unchanged. As helper_float_check_status only affects inexact, overflow and underflow, and the xvcmp* instructions don't change these flags, this issue can be fixed by removing the call to helper_float_check_status. By doing this, the FI bit in FPSCR will be preserved as expected. Fixes: 00084a25adf ("target/ppc: introduce separate VSX_CMP macro for xvcmp* instructions") Signed-off-by: Víctor Colombo <victor.colombo@eldorado.org.br> --- target/ppc/translate/vsx-impl.c.inc | 1 - 1 file changed, 1 deletion(-)