Non-Arithmetic Floating-Point Instructions; Non-Arithmetic Floating-Point Pseudo-Operations - Intel ITANIUM ARCHITECTURE - SOFTWARE DEVELOPERS MANUAL VOLUME 1 REV 2.3 Manual

The fneg pseudo-operation (see
and is therefore not equivalent to the IEEE negation operation. For the IEEE negation
operation, an fnma using FR 1 as the multiplicand and FR 0 as the addend must be
used.
Table 5-14
non-arithmetic pseudo-operation definitions.
Table 5-14.
Floating-point classify
Floating-point merge sign
Parallel FP merge sign
Floating-point merge negative sign
Parallel FP merge negative sign
Floating-point merge sign and exponent
Parallel FP merge sign and exponent
Floating-point mix left
Floating-point mix right
Floating-point mix left-right
Floating-point sign-extend left
Floating-point sign-extend right
Floating-point pack
Floating-point swap
Floating-point swap and negate left
Floating-point swap and negate right
Floating-point And
Floating-point And Complement
Floating-point Or
Floating-point Xor
Floating-point Select
Table 5-15.
Floating-point absolute value
Parallel FP absolute value
Floating-point negate
Parallel FP negate
Floating-point negate absolute value
Parallel FP negate absolute value
5.3.5 Floating-point Status Register (FPSR) Status Field
Instructions
Speculation of floating-point operations requires that the status flags be stored
temporarily in one of the alternate status fields (not FPSR.sf0). After a speculative
execution chain has been committed, a fchkf instruction can be used to update the
main status field flags (FPSR.sf0.flags). This operation will preserve the correctness of
the IEEE flags. The fchkf instruction does this by comparing the flags of the status field
1:100
Table
lists the non-arithmetic floating-point instructions and

Non-arithmetic Floating-point Instructions

Operation

Non-arithmetic Floating-point Pseudo-operations

Operation
fabs
fpabs
fneg
fpneg
fnegabs
fpnegabs
5-15) simply reverses the sign bit of the operand
fclass.fcrel.fctype
fmerge.s
fpmerge.s
fmerge.ns
fpmerge.ns
fmerge.se
fpmerge.se
fmix.l
fmix.r
fmix.lr
fsxt.l
fsxt.r
fpack
fswap
fswap.nl
fswap.nr
fand
fandcm
for
fxor
fselect
Mnemonic
fmerge.s, with sign from FR 0
fpmerge.s, with sign from FR 0
fmerge.ns
fpmerge.ns
fmerge.ns, with sign from FR 0
fpmerge.ns, with sign from FR 0
Volume 1, Part 1: Floating-point Programming Model
Table 5-15 lists the
Mnemonic(s)
Operation Used