getting a giant bitmap into a container

[440gtx]

I would like to represent used/free clusters on a disk drive in a
standard container. Pardon platform specifics, but the cluster data
comes from a POD header followed by the cluster bitmap as defined by
FSCTL_GET_VOLUME_BITMAP. One choice is to create an empty vector
<bool> and manually copy the bits. Since there can be billions of
clusters, this is very inefficient in terms of needing double the
memory and then the time to copy billions of bits:

vector <bool> bitmap;
VOLUME_BITMAP_BUFFER *out = (VOLUME_BITMAP_BUFFER *)malloc(out_len);

…call ioctl…

bitmap.resize(out->BitmapSize.QuadPart);

for (vector <bool>::size_type i = 0; i < bitmap.size(); ++i)
bitmap = (out->Buffer[i/8] >> (i % 8)) & 1;

Are there more appropriate ways?

 

[Öö Tiib]

I would like to represent used/free clusters on a disk drive in a
standard container. Pardon platform specifics, but the cluster data
comes from a POD header followed by the cluster bitmap as defined by
FSCTL_GET_VOLUME_BITMAP. One choice is to create an empty vector
<bool> and manually copy the bits. Since there can be billions of
clusters, this is very inefficient in terms of needing double the
memory and then the time to copy billions of bits:

You need double memory each time when you copy something to something
else so if this is concern then do not copy and do not convert. When
unsure then measure.

vector <bool> bitmap;
VOLUME_BITMAP_BUFFER *out   = (VOLUME_BITMAP_BUFFER *)malloc(out_len);

…call ioctl…

bitmap.resize(out->BitmapSize.QuadPart);

for (vector <bool>::size_type i = 0; i < bitmap.size(); ++i)
    bitmap = (out->Buffer[i/8] >> (i % 8)) & 1;

Are there more appropriate ways?

boost::dynamic_bitset or vector of std::bitset are perhaps better
containers for bits than std::vector<bool>.

 

[440gtx]

boost::dynamic_bitset or vector of std::bitset are perhaps

better containers for bits than std::vector<bool>.

Better in what way? Do they solve the performance and memory usage
issue of needing to copy a billion bits one by one from one memory
region to another?

 

[Öö Tiib]

Better in what way? Do they solve the performance and memory usage
issue of needing to copy a billion bits one by one from one memory
region to another?

Better since bitset constructor takes at least value (unsigned long)
in, so your copying code does not look like ">(>%)8&":

unsigned char original;
std::bitset<8> copied( original );