@matheusbsbs / matheus-bagliano / commits / 007062ab4e

index-pack, unpack-objects: increase input buffer from 4 KiB to 128 KiB

index-pack and unpack-objects both read pack data from stdin through a 4 KiB static buffer. In index-pack, each fill() flushes consumed bytes to the pack file via write_or_die(), capping every write(2) at 4 KiB. unpack-objects uses the same buffer pattern for reads. On FUSE-backed filesystems every write(2) is a synchronous round trip through the FUSE protocol (userspace -> kernel -> userspace -> back), so the 4 KiB buffer turns a clone into many unnecessary tiny writes with noticeable latency overhead. Increase the buffer from 4 KiB to 128 KiB. Introduce a shared DEFAULT_IO_BUFFER_SIZE constant in git-compat-util.h (next to MAX_IO_SIZE) and use it in index-pack, unpack-objects, and the hashfile layer in csum-file (which already used 128 KiB but hardcoded the value). Pack file writes to a FUSE filesystem with writeback caching disabled during HTTPS clones of git/git (~293 MB pack): 74,958 -> 4,687 (94% fewer) Wall-clock time of git clone over HTTPS onto a FUSE passthrough filesystem with writeback caching disabled, 3 runs per variant: vscode (~1.26 GB pack): 84.5s -> 75.7s avg (10% faster) git/git (~306 MB pack): 22.6s -> 20.0s avg (11% faster) Signed-off-by: Scott Bauersfeld <sbauersfeld@g.ucla.edu> Signed-off-by: Junio C Hamano <gitster@pobox.com>

Scott Bauersfeld committed Apr 28, 2026 at 14:47 UTC 007062ab4e01b8a1b1fe497610b8be4a57b76f4e

4 files changed +9 -5

builtin/index-pack.c

+1 -2

@@ -145,8 +145,7 @@ static int check_self_contained_and_connected;

145

146 static struct progress *progress;

147

148 -/* We always read in 4kB chunks. */

149 -static unsigned char input_buffer[4096];

148 +static unsigned char input_buffer[DEFAULT_IO_BUFFER_SIZE];

149 static unsigned int input_offset, input_len;

150 static off_t consumed_bytes;

151 static off_t max_input_size;

builtin/unpack-objects.c

+1 -2

@@ -23,8 +23,7 @@

23 static int dry_run, quiet, recover, has_errors, strict;

24 static const char unpack_usage[] = "git unpack-objects [-n] [-q] [-r] [--strict]";

26 -/* We always read in 4kB chunks. */

27 -static unsigned char buffer[4096];

26 +static unsigned char buffer[DEFAULT_IO_BUFFER_SIZE];

27 static unsigned int offset, len;

28 static off_t consumed_bytes;

29 static off_t max_input_size;

csum-file.c

+1 -1

@@ -178,7 +178,7 @@ struct hashfile *hashfd_ext(const struct git_hash_algo *algop,

178 f->algop = unsafe_hash_algo(algop);

179 f->algop->init_fn(&f->ctx);

180

181 - f->buffer_len = opts->buffer_len ? opts->buffer_len : 128 * 1024;

181 + f->buffer_len = opts->buffer_len ? opts->buffer_len : DEFAULT_IO_BUFFER_SIZE;

182 f->buffer = xmalloc(f->buffer_len);

183 f->check_buffer = NULL;

184

git-compat-util.h

@@ -712,6 +712,12 @@ static inline uint64_t u64_add(uint64_t a, uint64_t b)

712 # endif

713 #endif

714

715 +/*

716 + * Default buffer size for buffered I/O in index-pack, unpack-objects,

717 + * and the hashfile layer in csum-file.

718 + */

719 +#define DEFAULT_IO_BUFFER_SIZE (128 * 1024)

720 +

721 #ifdef HAVE_ALLOCA_H

722 # include <alloca.h>

723 # define xalloca(size) (alloca(size))