nixos/nix
1
1
Fork 0
mirror of https://github.com/NixOS/nix.git synced 2025-11-17 07:52:43 +01:00
Code Activity

Run the daemon worker on the same CPU as the client

Browse source 
On a system with multiple CPUs, running Nix operations through the
daemon is significantly slower than "direct" mode:

$ NIX_REMOTE= nix-instantiate '<nixos>' -A system
real    0m0.974s
user    0m0.875s
sys     0m0.088s

$ NIX_REMOTE=daemon nix-instantiate '<nixos>' -A system
real    0m2.118s
user    0m1.463s
sys     0m0.218s

The main reason seems to be that the client and the worker get moved
to a different CPU after every call to the worker.  This patch adds a
hack to lock them to the same CPU.  With this, the overhead of going
through the daemon is very small:

$ NIX_REMOTE=daemon nix-instantiate '<nixos>' -A system
real    0m1.074s
user    0m0.809s
sys     0m0.098s
This commit is contained in:
Eelco Dolstra 2013-08-07 11:51:55 +00:00
parent 263d668222
commit a583a2bc59
9 changed files with 92 additions and 4 deletions
Download patch file Download diff file Expand all files Collapse all files

13
src/libstore/remote-store.cc
View file

@ -3,6 +3,7 @@
#include "remote-store.hh"
#include "worker-protocol.hh"
#include "archive.hh"
#include "affinity.hh"
#include "globals.hh"
#include <sys/types.h>
@ -15,7 +16,6 @@
#include <unistd.h>
#include <cstring>
namespace nix {
@ -71,8 +71,19 @@ void RemoteStore::openConnection(bool reserveSpace)
if (GET_PROTOCOL_MAJOR(daemonVersion) != GET_PROTOCOL_MAJOR(PROTOCOL_VERSION))
throw Error("Nix daemon protocol version not supported");
writeInt(PROTOCOL_VERSION, to);
if (GET_PROTOCOL_MINOR(daemonVersion) >= 14) {
int cpu = lockToCurrentCPU();
if (cpu != -1) {
writeInt(1, to);
writeInt(cpu, to);
} else
writeInt(0, to);
}
if (GET_PROTOCOL_MINOR(daemonVersion) >= 11)
writeInt(reserveSpace, to);
processStderr();
}
catch (Error & e) {