target-ppc: fix sNaN propagation
The current FPU code returns 0.0 if one of the operand is a signaling NaN and the VXSNAN exception is disabled. fload_invalid_op_excp() doesn't return a qNaN in case of a VXSNAN exception as the operand should be propagated instead of a new qNaN to be generated. Fix that by calling fload_invalid_op_excp() only for the exception generation (if enabled), and use the softfloat code to correctly compute the result. Reviewed-by: Nathan Froyd <froydnj@codesourcery.com> Signed-off-by: Aurelien Jarno <aurelien@aurel32.net>
This commit is contained in:
parent
b2bf03a90c
commit
96912e3970
@ -975,15 +975,16 @@ uint64_t helper_fadd (uint64_t arg1, uint64_t arg2)
|
||||
farg1.ll = arg1;
|
||||
farg2.ll = arg2;
|
||||
|
||||
if (unlikely(float64_is_signaling_nan(farg1.d) ||
|
||||
float64_is_signaling_nan(farg2.d))) {
|
||||
/* sNaN addition */
|
||||
farg1.ll = fload_invalid_op_excp(POWERPC_EXCP_FP_VXSNAN);
|
||||
} else if (unlikely(float64_is_infinity(farg1.d) && float64_is_infinity(farg2.d) &&
|
||||
if (unlikely(float64_is_infinity(farg1.d) && float64_is_infinity(farg2.d) &&
|
||||
float64_is_neg(farg1.d) != float64_is_neg(farg2.d))) {
|
||||
/* Magnitude subtraction of infinities */
|
||||
farg1.ll = fload_invalid_op_excp(POWERPC_EXCP_FP_VXISI);
|
||||
} else {
|
||||
if (unlikely(float64_is_signaling_nan(farg1.d) ||
|
||||
float64_is_signaling_nan(farg2.d))) {
|
||||
/* sNaN addition */
|
||||
fload_invalid_op_excp(POWERPC_EXCP_FP_VXSNAN);
|
||||
}
|
||||
farg1.d = float64_add(farg1.d, farg2.d, &env->fp_status);
|
||||
}
|
||||
|
||||
@ -998,15 +999,16 @@ uint64_t helper_fsub (uint64_t arg1, uint64_t arg2)
|
||||
farg1.ll = arg1;
|
||||
farg2.ll = arg2;
|
||||
|
||||
if (unlikely(float64_is_signaling_nan(farg1.d) ||
|
||||
float64_is_signaling_nan(farg2.d))) {
|
||||
/* sNaN subtraction */
|
||||
farg1.ll = fload_invalid_op_excp(POWERPC_EXCP_FP_VXSNAN);
|
||||
} else if (unlikely(float64_is_infinity(farg1.d) && float64_is_infinity(farg2.d) &&
|
||||
if (unlikely(float64_is_infinity(farg1.d) && float64_is_infinity(farg2.d) &&
|
||||
float64_is_neg(farg1.d) == float64_is_neg(farg2.d))) {
|
||||
/* Magnitude subtraction of infinities */
|
||||
farg1.ll = fload_invalid_op_excp(POWERPC_EXCP_FP_VXISI);
|
||||
} else {
|
||||
if (unlikely(float64_is_signaling_nan(farg1.d) ||
|
||||
float64_is_signaling_nan(farg2.d))) {
|
||||
/* sNaN subtraction */
|
||||
fload_invalid_op_excp(POWERPC_EXCP_FP_VXSNAN);
|
||||
}
|
||||
farg1.d = float64_sub(farg1.d, farg2.d, &env->fp_status);
|
||||
}
|
||||
|
||||
@ -1021,15 +1023,16 @@ uint64_t helper_fmul (uint64_t arg1, uint64_t arg2)
|
||||
farg1.ll = arg1;
|
||||
farg2.ll = arg2;
|
||||
|
||||
if (unlikely(float64_is_signaling_nan(farg1.d) ||
|
||||
float64_is_signaling_nan(farg2.d))) {
|
||||
/* sNaN multiplication */
|
||||
farg1.ll = fload_invalid_op_excp(POWERPC_EXCP_FP_VXSNAN);
|
||||
} else if (unlikely((float64_is_infinity(farg1.d) && float64_is_zero(farg2.d)) ||
|
||||
if (unlikely((float64_is_infinity(farg1.d) && float64_is_zero(farg2.d)) ||
|
||||
(float64_is_zero(farg1.d) && float64_is_infinity(farg2.d)))) {
|
||||
/* Multiplication of zero by infinity */
|
||||
farg1.ll = fload_invalid_op_excp(POWERPC_EXCP_FP_VXIMZ);
|
||||
} else {
|
||||
if (unlikely(float64_is_signaling_nan(farg1.d) ||
|
||||
float64_is_signaling_nan(farg2.d))) {
|
||||
/* sNaN multiplication */
|
||||
fload_invalid_op_excp(POWERPC_EXCP_FP_VXSNAN);
|
||||
}
|
||||
farg1.d = float64_mul(farg1.d, farg2.d, &env->fp_status);
|
||||
}
|
||||
|
||||
@ -1044,17 +1047,18 @@ uint64_t helper_fdiv (uint64_t arg1, uint64_t arg2)
|
||||
farg1.ll = arg1;
|
||||
farg2.ll = arg2;
|
||||
|
||||
if (unlikely(float64_is_signaling_nan(farg1.d) ||
|
||||
float64_is_signaling_nan(farg2.d))) {
|
||||
/* sNaN division */
|
||||
farg1.ll = fload_invalid_op_excp(POWERPC_EXCP_FP_VXSNAN);
|
||||
} else if (unlikely(float64_is_infinity(farg1.d) && float64_is_infinity(farg2.d))) {
|
||||
if (unlikely(float64_is_infinity(farg1.d) && float64_is_infinity(farg2.d))) {
|
||||
/* Division of infinity by infinity */
|
||||
farg1.ll = fload_invalid_op_excp(POWERPC_EXCP_FP_VXIDI);
|
||||
} else if (unlikely(float64_is_zero(farg1.d) && float64_is_zero(farg2.d))) {
|
||||
/* Division of zero by zero */
|
||||
farg1.ll = fload_invalid_op_excp(POWERPC_EXCP_FP_VXZDZ);
|
||||
} else {
|
||||
if (unlikely(float64_is_signaling_nan(farg1.d) ||
|
||||
float64_is_signaling_nan(farg2.d))) {
|
||||
/* sNaN division */
|
||||
fload_invalid_op_excp(POWERPC_EXCP_FP_VXSNAN);
|
||||
}
|
||||
farg1.d = float64_div(farg1.d, farg2.d, &env->fp_status);
|
||||
}
|
||||
|
||||
@ -1232,16 +1236,17 @@ uint64_t helper_fmadd (uint64_t arg1, uint64_t arg2, uint64_t arg3)
|
||||
farg2.ll = arg2;
|
||||
farg3.ll = arg3;
|
||||
|
||||
if (unlikely(float64_is_signaling_nan(farg1.d) ||
|
||||
float64_is_signaling_nan(farg2.d) ||
|
||||
float64_is_signaling_nan(farg3.d))) {
|
||||
/* sNaN operation */
|
||||
farg1.ll = fload_invalid_op_excp(POWERPC_EXCP_FP_VXSNAN);
|
||||
} else if (unlikely((float64_is_infinity(farg1.d) && float64_is_zero(farg2.d)) ||
|
||||
if (unlikely((float64_is_infinity(farg1.d) && float64_is_zero(farg2.d)) ||
|
||||
(float64_is_zero(farg1.d) && float64_is_infinity(farg2.d)))) {
|
||||
/* Multiplication of zero by infinity */
|
||||
farg1.ll = fload_invalid_op_excp(POWERPC_EXCP_FP_VXIMZ);
|
||||
} else {
|
||||
if (unlikely(float64_is_signaling_nan(farg1.d) ||
|
||||
float64_is_signaling_nan(farg2.d) ||
|
||||
float64_is_signaling_nan(farg3.d))) {
|
||||
/* sNaN operation */
|
||||
fload_invalid_op_excp(POWERPC_EXCP_FP_VXSNAN);
|
||||
}
|
||||
#ifdef FLOAT128
|
||||
/* This is the way the PowerPC specification defines it */
|
||||
float128 ft0_128, ft1_128;
|
||||
@ -1276,16 +1281,17 @@ uint64_t helper_fmsub (uint64_t arg1, uint64_t arg2, uint64_t arg3)
|
||||
farg2.ll = arg2;
|
||||
farg3.ll = arg3;
|
||||
|
||||
if (unlikely(float64_is_signaling_nan(farg1.d) ||
|
||||
float64_is_signaling_nan(farg2.d) ||
|
||||
float64_is_signaling_nan(farg3.d))) {
|
||||
/* sNaN operation */
|
||||
farg1.ll = fload_invalid_op_excp(POWERPC_EXCP_FP_VXSNAN);
|
||||
} else if (unlikely((float64_is_infinity(farg1.d) && float64_is_zero(farg2.d)) ||
|
||||
if (unlikely((float64_is_infinity(farg1.d) && float64_is_zero(farg2.d)) ||
|
||||
(float64_is_zero(farg1.d) && float64_is_infinity(farg2.d)))) {
|
||||
/* Multiplication of zero by infinity */
|
||||
farg1.ll = fload_invalid_op_excp(POWERPC_EXCP_FP_VXIMZ);
|
||||
} else {
|
||||
if (unlikely(float64_is_signaling_nan(farg1.d) ||
|
||||
float64_is_signaling_nan(farg2.d) ||
|
||||
float64_is_signaling_nan(farg3.d))) {
|
||||
/* sNaN operation */
|
||||
fload_invalid_op_excp(POWERPC_EXCP_FP_VXSNAN);
|
||||
}
|
||||
#ifdef FLOAT128
|
||||
/* This is the way the PowerPC specification defines it */
|
||||
float128 ft0_128, ft1_128;
|
||||
@ -1319,16 +1325,17 @@ uint64_t helper_fnmadd (uint64_t arg1, uint64_t arg2, uint64_t arg3)
|
||||
farg2.ll = arg2;
|
||||
farg3.ll = arg3;
|
||||
|
||||
if (unlikely(float64_is_signaling_nan(farg1.d) ||
|
||||
float64_is_signaling_nan(farg2.d) ||
|
||||
float64_is_signaling_nan(farg3.d))) {
|
||||
/* sNaN operation */
|
||||
farg1.ll = fload_invalid_op_excp(POWERPC_EXCP_FP_VXSNAN);
|
||||
} else if (unlikely((float64_is_infinity(farg1.d) && float64_is_zero(farg2.d)) ||
|
||||
if (unlikely((float64_is_infinity(farg1.d) && float64_is_zero(farg2.d)) ||
|
||||
(float64_is_zero(farg1.d) && float64_is_infinity(farg2.d)))) {
|
||||
/* Multiplication of zero by infinity */
|
||||
farg1.ll = fload_invalid_op_excp(POWERPC_EXCP_FP_VXIMZ);
|
||||
} else {
|
||||
if (unlikely(float64_is_signaling_nan(farg1.d) ||
|
||||
float64_is_signaling_nan(farg2.d) ||
|
||||
float64_is_signaling_nan(farg3.d))) {
|
||||
/* sNaN operation */
|
||||
fload_invalid_op_excp(POWERPC_EXCP_FP_VXSNAN);
|
||||
}
|
||||
#ifdef FLOAT128
|
||||
/* This is the way the PowerPC specification defines it */
|
||||
float128 ft0_128, ft1_128;
|
||||
@ -1364,16 +1371,17 @@ uint64_t helper_fnmsub (uint64_t arg1, uint64_t arg2, uint64_t arg3)
|
||||
farg2.ll = arg2;
|
||||
farg3.ll = arg3;
|
||||
|
||||
if (unlikely(float64_is_signaling_nan(farg1.d) ||
|
||||
float64_is_signaling_nan(farg2.d) ||
|
||||
float64_is_signaling_nan(farg3.d))) {
|
||||
/* sNaN operation */
|
||||
farg1.ll = fload_invalid_op_excp(POWERPC_EXCP_FP_VXSNAN);
|
||||
} else if (unlikely((float64_is_infinity(farg1.d) && float64_is_zero(farg2.d)) ||
|
||||
if (unlikely((float64_is_infinity(farg1.d) && float64_is_zero(farg2.d)) ||
|
||||
(float64_is_zero(farg1.d) && float64_is_infinity(farg2.d)))) {
|
||||
/* Multiplication of zero by infinity */
|
||||
farg1.ll = fload_invalid_op_excp(POWERPC_EXCP_FP_VXIMZ);
|
||||
} else {
|
||||
if (unlikely(float64_is_signaling_nan(farg1.d) ||
|
||||
float64_is_signaling_nan(farg2.d) ||
|
||||
float64_is_signaling_nan(farg3.d))) {
|
||||
/* sNaN operation */
|
||||
fload_invalid_op_excp(POWERPC_EXCP_FP_VXSNAN);
|
||||
}
|
||||
#ifdef FLOAT128
|
||||
/* This is the way the PowerPC specification defines it */
|
||||
float128 ft0_128, ft1_128;
|
||||
@ -1409,11 +1417,11 @@ uint64_t helper_frsp (uint64_t arg)
|
||||
|
||||
if (unlikely(float64_is_signaling_nan(farg.d))) {
|
||||
/* sNaN square root */
|
||||
farg.ll = fload_invalid_op_excp(POWERPC_EXCP_FP_VXSNAN);
|
||||
} else {
|
||||
fload_invalid_op_excp(POWERPC_EXCP_FP_VXSNAN);
|
||||
}
|
||||
f32 = float64_to_float32(farg.d, &env->fp_status);
|
||||
farg.d = float32_to_float64(f32, &env->fp_status);
|
||||
}
|
||||
|
||||
return farg.ll;
|
||||
}
|
||||
|
||||
@ -1423,13 +1431,14 @@ uint64_t helper_fsqrt (uint64_t arg)
|
||||
CPU_DoubleU farg;
|
||||
farg.ll = arg;
|
||||
|
||||
if (unlikely(float64_is_signaling_nan(farg.d))) {
|
||||
/* sNaN square root */
|
||||
farg.ll = fload_invalid_op_excp(POWERPC_EXCP_FP_VXSNAN);
|
||||
} else if (unlikely(float64_is_neg(farg.d) && !float64_is_zero(farg.d))) {
|
||||
if (unlikely(float64_is_neg(farg.d) && !float64_is_zero(farg.d))) {
|
||||
/* Square root of a negative nonzero number */
|
||||
farg.ll = fload_invalid_op_excp(POWERPC_EXCP_FP_VXSQRT);
|
||||
} else {
|
||||
if (unlikely(float64_is_signaling_nan(farg.d))) {
|
||||
/* sNaN square root */
|
||||
fload_invalid_op_excp(POWERPC_EXCP_FP_VXSNAN);
|
||||
}
|
||||
farg.d = float64_sqrt(farg.d, &env->fp_status);
|
||||
}
|
||||
return farg.ll;
|
||||
@ -1443,10 +1452,9 @@ uint64_t helper_fre (uint64_t arg)
|
||||
|
||||
if (unlikely(float64_is_signaling_nan(farg.d))) {
|
||||
/* sNaN reciprocal */
|
||||
farg.ll = fload_invalid_op_excp(POWERPC_EXCP_FP_VXSNAN);
|
||||
} else {
|
||||
farg.d = float64_div(float64_one, farg.d, &env->fp_status);
|
||||
fload_invalid_op_excp(POWERPC_EXCP_FP_VXSNAN);
|
||||
}
|
||||
farg.d = float64_div(float64_one, farg.d, &env->fp_status);
|
||||
return farg.d;
|
||||
}
|
||||
|
||||
@ -1459,12 +1467,12 @@ uint64_t helper_fres (uint64_t arg)
|
||||
|
||||
if (unlikely(float64_is_signaling_nan(farg.d))) {
|
||||
/* sNaN reciprocal */
|
||||
farg.ll = fload_invalid_op_excp(POWERPC_EXCP_FP_VXSNAN);
|
||||
} else {
|
||||
fload_invalid_op_excp(POWERPC_EXCP_FP_VXSNAN);
|
||||
}
|
||||
farg.d = float64_div(float64_one, farg.d, &env->fp_status);
|
||||
f32 = float64_to_float32(farg.d, &env->fp_status);
|
||||
farg.d = float32_to_float64(f32, &env->fp_status);
|
||||
}
|
||||
|
||||
return farg.ll;
|
||||
}
|
||||
|
||||
@ -1475,13 +1483,14 @@ uint64_t helper_frsqrte (uint64_t arg)
|
||||
float32 f32;
|
||||
farg.ll = arg;
|
||||
|
||||
if (unlikely(float64_is_signaling_nan(farg.d))) {
|
||||
/* sNaN reciprocal square root */
|
||||
farg.ll = fload_invalid_op_excp(POWERPC_EXCP_FP_VXSNAN);
|
||||
} else if (unlikely(float64_is_neg(farg.d) && !float64_is_zero(farg.d))) {
|
||||
if (unlikely(float64_is_neg(farg.d) && !float64_is_zero(farg.d))) {
|
||||
/* Reciprocal square root of a negative nonzero number */
|
||||
farg.ll = fload_invalid_op_excp(POWERPC_EXCP_FP_VXSQRT);
|
||||
} else {
|
||||
if (unlikely(float64_is_signaling_nan(farg.d))) {
|
||||
/* sNaN reciprocal square root */
|
||||
fload_invalid_op_excp(POWERPC_EXCP_FP_VXSNAN);
|
||||
}
|
||||
farg.d = float64_sqrt(farg.d, &env->fp_status);
|
||||
farg.d = float64_div(float64_one, farg.d, &env->fp_status);
|
||||
f32 = float64_to_float32(farg.d, &env->fp_status);
|
||||
|
Loading…
Reference in New Issue
Block a user