I plumbed through the build and tested it. The new patchset is: 1. https://github.com/AOSC-Dev/ntfs-3g/commit/418f794de68da0784c0df57bda7b1a0cd8eb4f00.patch 2. https://github.com/AOSC-Dev/ntfs-3g/commit/5fb9fe6177d412ab6069f6b5d6f57cabf1ac04c8.patch
For some reason the internal FUSE is faster now that the write size is back on the same level. I did a very shobby test with a usual file on a HARD drive and mounted as an NTFS partition. dd bs=4M if=/dev/random runs at ~47MB/s external vs ~60 MB/s internal, while the CPU usage is ~23% ext vs ~12% int. I don't really understand the speedup, since even the most aggressive change around bufsize yields equivalent behavior as libfuse v3 (barring the splice magic). It could be our simpler logic or just static linking (on modern AMD64, really?). There is an additional, very small patch against some GCC warnings: https://github.com/AOSC-Dev/ntfs-3g/commit/b73b8da7a88affc751d4c58a9d80191a9118e424.patch -- Mingye Wang (Artoria2e5) _______________________________________________ ntfs-3g-devel mailing list ntfs-3g-devel@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/ntfs-3g-devel
