05.02.2013 Views

ARM Architecture Reference Manual ARMv7-A and ARMv7-R edition

ARM Architecture Reference Manual ARMv7-A and ARMv7-R edition

ARM Architecture Reference Manual ARMv7-A and ARMv7-R edition

SHOW MORE
SHOW LESS

Create successful ePaper yourself

Turn your PDF publications into a flip-book with our unique Google optimized e-Paper software.

Application Level Programmers’ Model<br />

The criteria for the Underflow exception to occur are different in Flush-to-zero mode. For<br />

details, see Flush-to-zero on page A2-39.<br />

IXC Inexact. The flag is set to 1 if the result of an operation is not equivalent to the value that<br />

would be produced if the operation were performed with unbounded precision <strong>and</strong> exponent<br />

range.<br />

The criteria for the Inexact exception to occur are different in Flush-to-zero mode. For<br />

details, see Flush-to-zero on page A2-39.<br />

IDC Input Denormal. The flag is set to 1 if a denormalized input oper<strong>and</strong> is replaced in the<br />

computation by a zero, as described in Flush-to-zero on page A2-39.<br />

With the Advanced SIMD extension <strong>and</strong> the VFPv3 extension these are non-trapping exceptions <strong>and</strong> the<br />

data-processing instructions do not generate any trapped exceptions.<br />

With the VFPv2 <strong>and</strong> VFPv3U extensions:<br />

These exceptions can be trapped, by setting trap enable flags in the FPSCR, see VFPv3U on<br />

page A2-31. Trapped floating-point exceptions are delivered to user code in an IMPLEMENTATION<br />

DEFINED fashion.<br />

The definitions of the floating-point exceptions change as follows:<br />

— if the Underflow exception is trapped, it occurs if the absolute value of the result of an<br />

operation, produced before rounding, is less than the minimum positive normalized number<br />

for the destination precision, regardless of whether the rounded result is inexact<br />

— higher priority trapped exceptions can prevent lower priority exceptions from occurring, as<br />

described in Combinations of exceptions on page A2-44.<br />

Table A2-7 shows the default results of the floating-point exceptions:<br />

Table A2-7 Floating-point exception default results<br />

Exception type Default result for positive sign Default result for negative sign<br />

IOC, Invalid Operation Quiet NaN Quiet NaN<br />

DZC, Division by Zero +∞ (plus infinity) –∞ (minus infinity)<br />

OFC, Overflow RN, RP:<br />

RM, RZ:<br />

+∞ (plus infinity)<br />

+MaxNorm<br />

RN, RM:<br />

RP, RZ:<br />

UFC, Underflow Normal rounded result Normal rounded result<br />

IXC, Inexact Normal rounded result Normal rounded result<br />

IDC, Input Denormal Normal rounded result Normal rounded result<br />

–∞ (minus infinity)<br />

–MaxNorm<br />

<strong>ARM</strong> DDI 0406B Copyright © 1996-1998, 2000, 2004-2008 <strong>ARM</strong> Limited. All rights reserved. A2-43

Hooray! Your file is uploaded and ready to be published.

Saved successfully!

Ooh no, something went wrong!