Compression Costs And Benefits

The primary benefit of compressing sections is a reduction in object size. However, compression imposes additional costs in runtime and memory use at all stages of development.

• The compiler must produce each section in uncompressed form, allocate additional memory for the compressed version, and perform the compression.

• When reading an input object, the link-editor must read the compressed data into memory, allocate additional memory to hold the decompressed data, and perform the decompression.

• On output, the link-editor must create an uncompressed version of the resulting sections. If compression is requested, additional memory and time are used to create the compressed version.

• When a debugger reads an object with compressed debug sections, the debugger must allocate additional memory to hold the decompressed data, and perform the decompression.

Compressed sections may be beneficial in cases where disk space is scarce, or for production objects that are copied widely and debugged rarely. Compressed sections allow for smaller files, but do not increase the amount of logical information that can be represented. A common example involves 32-bit objects, which are fundamentally limited to 4 Gbytes due to the use of 32-bit file offsets and sizes within them. It is sometimes assumed that compressing debug data might allow for more debug information to be generated. However, the format of 32-bit debug data also contains 32-bit offsets, and so, is logically constrained to 4 Gbytes in uncompressed form. Ancillary objects, or 64-bit objects with their much larger size limits, can be used in such cases.