{ $class-description "Symbols of this type represent floating-point exceptions. They are used to get and set the floating-point unit's exception flags (using " { $link fp-exception-flags } " and " { $link set-fp-exception-flags } ") and to control processor traps (using " { $link with-fp-traps } "). The following symbols are defined:"
{ $list
{ { $link +fp-invalid-operation+ } " indicates that an invalid floating-point operation occurred, such as taking the square root of a negative number or dividing zero by zero." }
{ { $link +fp-overflow+ } " indicates that a floating-point operation gave a result larger than the maximum representable value of the type used to perform the calculation." }
{ { $link +fp-underflow+ } " indicates that a floating-point operation gave a result smaller than the minimum representable normalized value of the type used to perform the calculation." }
{ { $link +fp-zero-divide+ } " indicates that a floating-point division by zero was attempted." }
{ { $link +fp-inexact+ } " indicates that a floating-point operation gave an inexact result that needed to be rounded." }
} } ;
HELP:+fp-invalid-operation+
{ $class-description "This symbol represents a invalid operation " { $link fp-exception } "." } ;
HELP:+fp-overflow+
{ $class-description "This symbol represents an overflow " { $link fp-exception } "." } ;
HELP:+fp-underflow+
{ $class-description "This symbol represents an underflow " { $link fp-exception } "." } ;
HELP:+fp-zero-divide+
{ $class-description "This symbol represents a division-by-zero " { $link fp-exception } "." } ;
HELP:+fp-inexact+
{ $class-description "This symbol represents an inexact result " { $link fp-exception } "." } ;
HELP:fp-rounding-mode
{ $class-description "Symbols of this type represent floating-point rounding modes. They are passed to the " { $link with-rounding-mode } " word to control how inexact values are calculated when exact results cannot fit in a floating-point type. The following symbols are defined:"
{ $list
{ { $link +round-nearest+ } " rounds the exact result to the nearest representable value, using the even value when the result is halfway between its two nearest values." }
{ { $link +round-zero+ } " rounds the exact result toward zero, that is, down for positive values, and up for negative values." }
{ { $link +round-down+ } " always rounds the exact result down." }
{ { $link +round-up+ } " always rounds the exact result up." }
} } ;
HELP:+round-nearest+
{ $class-description "This symbol represents the round-to-nearest " { $link fp-rounding-mode } "." } ;
HELP:+round-zero+
{ $class-description "This symbol represents the round-toward-zero " { $link fp-rounding-mode } "." } ;
HELP:+round-down+
{ $class-description "This symbol represents the round-down " { $link fp-rounding-mode } "." } ;
HELP:+round-up+
{ $class-description "This symbol represents the round-up " { $link fp-rounding-mode } "." } ;
HELP:fp-denormal-mode
{ $class-description "Symbols of this type represent floating-point denormal modes. They are passed to the " { $link with-denormal-mode } " word to control whether denormals are generated as outputs of floating-point operations and how they are treated when given as inputs."
{ $list
{ { $link +denormal-keep+ } " causes denormal results to be generated and accepted as inputs as required by IEEE 754." }
{ { $link +denormal-flush+ } " causes denormal results to be flushed to zero and be treated as zero when given as inputs. This mode may allow floating point operations to give results that are not compliant with the IEEE 754 standard." }
} } ;
HELP:+denormal-keep+
{ $class-description "This symbol represents the IEEE 754 compliant keep-denormals " { $link fp-denormal-mode } "." } ;
HELP:+denormal-flush+
{ $class-description "This symbol represents the non-IEEE-754-compliant flush-denormals-to-zero " { $link fp-denormal-mode } "." } ;
{ $description "Clears the floating-point exception flags and then calls " { $snippet "quot" } ", returning the set of floating-point exceptions raised during its execution and placing them on the datastack on " { $snippet "quot" } "'s completion." } ;
{ $description "Sets the floating-point denormal mode to " { $snippet "mode" } " for the dynamic extent of " { $snippet "quot" } ", restoring the denormal mode to its original value on " { $snippet "quot" } "'s completion." } ;
{ $description "Sets the floating-point rounding mode to " { $snippet "mode" } " for the dynamic extent of " { $snippet "quot" } ", restoring the rounding mode to its original value on " { $snippet "quot" } "'s completion." } ;
{ $description "Replaces the floating-point exception mask to enable processor traps to be raised for the set of exception conditions specified in " { $snippet "exceptions" } " for the dynamic extent of " { $snippet "quot" } ", restoring the original exception mask on " { $snippet "quot" } "'s completion." } ;
HELP:without-fp-traps
{ $values { "quot" quotation } }
{ $description "Disables all floating-pointer processor traps for the dynamic extent of " { $snippet "quot" } ", restoring the original exception mask on " { $snippet "quot" } "'s completion." } ;
"On x86, the above words only modify the SSE unit's state (in particular, the MXCSR register); the x87 unit is unaffected. On PowerPC, the above words only modify the scalar FPU's state (in FPSCR); the AltiVec unit is unaffected." } ;
