splitget

a downloader. in c. uses libcurl and pthreads.

stuff it does

• parallel chunks: handles multiple connections. thread pool thing. workers just grab stuff.
• granular resume: if it dies, it saves what's left in a .meta file. use -r to continue.
• adaptive chunking: it changes chunk size based on speed. fast net means bigger chunks. simple.
• direct i/o: writes to final file with pwrite. no .part files. no merge step.
• range detection: checks if ranges work. if not, it uses 1 thread.
• c: it's small. doesn't use much ram.

build

libcurl and pthreads.

make

it makes splitget.

use

./splitget [opts] <url>

opts

• -o <file>: output file. defaults to url name.
• -n <num> : threads. 4 is default.
• -r : resume. needs the .meta file.
• -q : quiet. no bars.
• -h : help.

example

./splitget -r -n 8 https://example.com/file.iso -o test.iso

logic

• work-stealing queue for the chunks.
• pwrite for the offsets.
• linked list for the holes.

license

mit. copyright 2026 bymehul. check the license file.