An extflonum is like a flonum, but with 80-bit precision and
not a number in the sense of `number?': only operations such as
`extfl+' work on extflonums, and only on platforms where extflonums
can be implemented by hardware without interefering with flonums
(i.e., on platforms where SSE instructions are used for
double-precision floats).
[Patch provided by Michael Filonenko and revised by Matthew.]
The compiler tracks information about bindings that are known to
hold extflonums, but the JIT does not yet exploit this information
to unbox them (except as intermediate results).
Note: With this refactoring, `math/utils' no longer depends on `rackunit'.
* (flexp2 x) computes (flexpt 2.0 x) but in about 1/3 the time for integer
`x' using a lookup table. Written for exact argument reduction in `fllog2'
after discovering that (flexpt 2.0 x) was the main performance bottleneck.
* (fllog2 x) computes (/ (fllog x) (fllog 2.0)) with near perfect accuracy.
Invented an algorithm to compute it with at least 8 extra bits before
final rounding; quite pleased with the result. Needed `fllog2' to ensure
(fllogb 2.0 x) would be exact when `x' is a power of two.
* (fllogb b x) computes (/ (fllog x) (fllog b)) with better accuracy, and
also handles limit values in a way that's consistent with the mathematical
limits. When those are ambiguous, it's consistent with `flexpt', which
follows IEEE 754 and C99. Otherwise returns +nan.0. See docs for details.
* `bflogb' is currently just for testing `fllogb'.
* Refactored FPU testing and documented it. So far, the only documented way
to do it is by calling `test-floating-point', which runs a comprehensive
deterministic+randomized suite of tests and returns a list representing
failed tests. I'll document individual tests after I document flonum
expansions and result/error functions like `fl+/error'.
* Added `fllog2' and `fllogb' to the flonum tests.
The code used eq? to check to see if a given language was in
an list of allowed-to-create-executables languages. But the
language object was passing thru TR and so eq? didn't hold and
thus the check was buggy.
They hypothesis behind this change is that the undo is
undoing an edit that came before, not the temporary edit
that appears in insert-close-paren. So, when there are
no edits happening (ie when fixup? is #f) then don't
undo. This is only a hypothesis because we were not able
to find a small code sequence, outside of DrRacket to
cause the bad behavior, so possibly there is really more
going on here.
Thanks to Nadeem Abdul Hamid for finding this fix.
Closes PR 13454
This move moves a library from the plt distribution into the
mr distribution, which is a direction we do not generally
like to go, but the pict library is more core than slideshow
and the dependency that triggered this change is that the
2htdp/image library now interoperates with picts (hopefully
2htdp/universe will also work with them at some point).
So this teases out the pict part of slideshow (mostly,
but not completely the texpict collection) and moves only
that part to the mr distribution.
The code to draw a border has to determine the widget's position
within the drawing window. Gtk 3.0 has a function to provide that
transformation, but Gtk 2.0 doesn't seem to have one, and it seems
that the transformation implemented in `racket/gui' wasn't right.
Closes PR 13453
The code added to handle expressions mixed with definitions was
not introduced in the right way; fix it to be like macro-introduced
code.
Closes PR 13452
