Table B–2: IEEE Floating-Point Trap Handling (Continued)

 

 

 

OS

User

 

Hardware1

PAL-

Completion

Signal

Alpha Instructions

Code

Handler

Handler

SQRTx INPUT Exceptions

 

 

 

 

 

 

 

 

 

Negative nonzero operand

Trap

Trap

Supply QNan

[Invalid Op]

+/–0

Supply +/–0

+ Denormal operand

Trap

Trap

Supply SQRT

[Denormal Op2]

– Denormal operand

Trap

Trap

Supply QNaN

[Denormal Op/

 

 

 

 

Invalid Op]

+ Infinity operand

Trap

Trap

Supply +Inf

– Infinity operand

Trap

Trap

Supply QNaN

[Invalid Op]

QNaN operand

Trap

Trap

Supply QNaN

SNaN operand

Trap

Trap

Supply QNaN

[Invalid Op]

 

 

 

 

SQRTx OUTPUT Exceptions

 

 

 

 

 

 

 

 

Exponent overflow

Not possible

 

 

 

Exponent underflow

Not possible

 

 

 

Inexact and disabled

Inexact and enabled

Supply SQRT

Trap

[Inexact]

 

 

 

 

 

1This column describes the minimum necessary hardware support.

2[Denormal Op] signals have priority over all other signals.

3[Overflow] and [Underflow] signals have priority over [Inexact] signals.

4An implementation could choose instead to trap to PALcode and have the PALcode supply a zero result on all underflows.

5An implementation could choose instead to trap to PALcode on extreme values and have the PALcode supply a truncated result on all overflows.

Other IEEE operations (software subroutines or sequences of instructions) are listed here for completeness:

Remainder

Round float to integer-valued float Convert binary to/from decimal

Compare, other combinations than the four above

IEEE Floating-Point Conformance B–11

Page 301
Image 301
Compaq ECQD2KCTE manual SQRTx Output Exceptions