Package: libgoogle-perftools4 Version: 2.2.1-0.2 Severity: important Dear Maintainer,
With the update from libgoogle-perftools4 2.1 to 2.2.1 I saw a serious performance regression in a benchmark which is sensitive to memory allocation performance. Runtime of the benchmark increased from less than 5 seconds to over 13 seconds (with glibc's allocator finishes the benchmark in 7s). `perf report -s dso` show almost 2/3 of the cycles spent in libtcmalloc.so.4.2.2 for the benchmark. Compiling the debian source package on the host results in a libtcmalloc which restores the performance and perf shows only 24% of the cycles for the benchmark spent in libtcmalloc. libtcmalloc-minimal4 has the same problem. -- System Information: Debian Release: jessie/sid APT prefers testing APT policy: (500, 'testing') Architecture: armhf (armv7l) Kernel: Linux 3.8.13.28 (SMP w/4 CPU cores; PREEMPT) Locale: LANG=C, LC_CTYPE=en_US.utf8 (charmap=UTF-8) Shell: /bin/sh linked to /bin/dash Versions of packages libgoogle-perftools4 depends on: ii libc6 2.19-11 ii libgcc1 1:4.9.1-16 ii libstdc++6 4.9.1-16 ii libtcmalloc-minimal4 2.2.1-0.2 libgoogle-perftools4 recommends no packages. libgoogle-perftools4 suggests no packages. -- no debconf information -- To UNSUBSCRIBE, email to [email protected] with a subject of "unsubscribe". Trouble? Contact [email protected]
