Intel Itanium 9150M, Manual

Page: 106 / 420

Volume 3: Instruction Reference 3:97
fpamax
fpamax — Floating-point Parallel Absolute Maximum
Format: (
qp
) fpamax.
sf f
1
=
f
2
,
f
3
F8
Description:
The paired single precision values in the significands of FR
f
2
and FR
f
3
are compared.
The operands with the larger absolute value are returned in the significand field of FR
f
1
.
If the magnitude of high (low) FR
f
3
is less than the magnitude of high (low) FR
f
2
, high
(low) FR
f
1
gets high (low) FR
f
2
. Otherwise high (low) FR
f
1
gets high (low) FR
f
3
.
If high (low) FR
f
2
or high (low) FR
f
3
is a NaN, and neither FR
f
2
or FR
f
3
is a NaTVal, high
(low) FR
f
1
gets high (low) FR
f
3
.
The exponent field of FR
f
1
is set to the biased exponent for 2.0
63
(0x1003E) and the
sign field of FR
f
1
is set to positive (0).
If either FR
f
2
or FR
f
3
is a NaTVal, FR
f
1
is set to NaTVal instead of the computed result.
This operation does not propagate NaNs the same way as other arithmetic
floating-point instructions. The Invalid Operation is signaled in the same manner as for
the
fpcmp.lt
operation.
The mnemonic values for
sf
are given in Table 2-23 on page 3:56 .
Operation: if (PR[ qp ]) {
fp_check_target_register( f
1
);
if (tmp_isrcode = fp_reg_disabled( f
1
, f
2
, f
3
, 0))
disabled_fp_register_fault(tmp_isrcode, 0);
if (fp_is_natval(FR[ f
2
]) || fp_is_natval(FR[ f
3
])) {
FR[ f
1
] = NATVAL;
} else {
fpminmax_exception_fault_check( f
2
, f
3
, sf , &tmp_fp_env);
if (fp_raise_fault(tmp_fp_env))
fp_exception_fault(fp_decode_fault(tmp_fp_env));
tmp_fr2 = tmp_right = fp_reg_read_hi( f
2
);
tmp_fr3 = tmp_left = fp_reg_read_hi( f
3
);
tmp_right.sign = FP_SIGN_POSITIVE;
tmp_left.sign = FP_SIGN_POSITIVE;
tmp_bool_res = fp_less_than(tmp_left, tmp_right);
tmp_res_hi = fp_single(tmp_bool_res ? tmp_fr2: tmp_fr3);
tmp_fr2 = tmp_right = fp_reg_read_lo( f
2
);
tmp_fr3 = tmp_left = fp_reg_read_lo( f
3
);
tmp_right.sign = FP_SIGN_POSITIVE;
tmp_left.sign = FP_SIGN_POSITIVE;
tmp_bool_res = fp_less_than(tmp_left, tmp_right);
tmp_res_lo = fp_single(tmp_bool_res ? tmp_fr2: tmp_fr3);
FR[ f
1
].significand = fp_concatenate(tmp_res_hi, tmp_res_lo);
FR[ f
1
].exponent = FP_INTEGER_EXP;
FR[ f
1
].sign = FP_SIGN_POSITIVE;
fp_update_fpsr( sf , tmp_fp_env);
}
fp_update_psr(
f
1
);
}