Hello, please help me test this glibc patch. It could bring a significant speedup to certain workloads, namely those that copy lots of memory in chunks.
I wrote a micro-benchmark for this: teythoon@darnassus:~/build/time-memcpy$ bash driver.bash size iterations gcc_memcpy memcpy vm_copy 4096 16777216 6.09 6.07 15.67 8192 8388608 7.02 6.29 7.87 12288 5592405 5.19 5.29 5.22 16384 4194304 6.36 5.71 3.56 32768 2097152 7.03 7.03 2.03 65536 1048576 7.29 7.33 0.87 The results are more consistent (but similar) on a quieter system. This is copying a buffer of SIZE, ITERATIONS times, a total of 2^36 bytes. gcc_memcpy is a self-written memcpy compiled with -O3 and -fno-builtin-memcpy, memcpy is from glibc, vm_copy is using vm_copy. You can find it here http://darnassus.sceen.net/gitweb/teythoon/time-memcpy.git/ Thanks, Justus
