nerdcruft/libcruft-util
1
0
Fork 0
Code Issues Releases Activity

maths: fix hi-lo normalisation shift size

Browse Source
This commit is contained in:
Danny Robson 2015-10-29 18:22:46 +11:00
parent 0369b16f73
commit ac86ef3bd5
1 changed files with 7 additions and 1 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

8
maths.hpp
View File

@ -446,6 +446,9 @@ typename std::enable_if<
>::type
renormalise [[gnu::pure]] (T t)
{
static_assert (sizeof (T) > sizeof (U),
"assumes right shift is sufficient");
// we have excess bits ,just shift and return
constexpr auto shift = 8 * (sizeof (T) - sizeof (U));
return t >> shift;
@ -462,6 +465,9 @@ typename std::enable_if<
>::type
renormalise [[gnu::pure]] (T t)
{
static_assert (sizeof (T) < sizeof (U),
"assumes bit creation is required to fill space");
// we need to create bits. fill the output integer with copies of ourself.
// this is approximately correct in the general case (introducing a small
// linear positive bias), but allows us to fill the output space in the
@ -473,7 +479,7 @@ renormalise [[gnu::pure]] (T t)
U out = 0;
for (size_t i = 0; i < sizeof (U) / sizeof (T); ++i)
out |= t << sizeof (U) * 8 * i;
out |= U (t) << sizeof (T) * 8 * i;
return out;
}