ash: rewrite waitpid_child() to improve performance

After the previous commit there was still a slight disparity
between shell performance with and without background jobs.
Reduce this disparity by the following changes:

- Remove the pidlist array.  The PID of the exiting process can be
  obtained from its handle.

- Save the proclist array between calls.  It will always grow to
  support the required number of PIDs.

- Combine the loops to compute the required size of proclist and
  to fill it.  This saves a few bytes with no measurable impact
  on performance.

Saves 8 bytes in a 32-bit build, adds 32 in 64-bit.

(GitHub issue #434)

1 files changed, 19 insertions, 32 deletions

diff --git a/shell/ash.c b/shell/ash.c
index ba29e3ea0..340c7ce2f 100644
--- a/shell/ash.c
+++ b/shell/ash.c
@@ -4925,50 +4925,37 @@ static pid_t
 waitpid_child(int *status, int wait_flags)
 {
         struct job *jb;
-        struct procstat *ps;
         int pid_nr = 0;
-        pid_t *pidlist;
-        HANDLE *proclist;
+        static HANDLE *proclist = NULL;
+        static int pid_max = 0;
         pid_t pid = -1;
         DWORD win_status, idx;
         int i;
 
         for (jb = curjob; jb; jb = jb->prev_job) {
-                if (jb->state != JOBDONE)
-                        pid_nr += jb->nprocs;
-        }
-        if (pid_nr == 0)
-                return -1;
-
-        pidlist = ckmalloc(sizeof(*pidlist)*pid_nr);
-        proclist = ckmalloc(sizeof(*proclist)*pid_nr);
+                if (jb->state != JOBDONE) {
+                        if (pid_nr + jb->nprocs > pid_max) {
+                                pid_max = pid_nr + jb->nprocs;
+                                proclist = ckrealloc(proclist, sizeof(*proclist) * pid_max);
+                        }
 
-        pid_nr = 0;
-        for (jb = curjob; jb; jb = jb->prev_job) {
-                if (jb->state == JOBDONE)
-                        continue;
-                ps = jb->ps;
-                for (i = 0; i < jb->nprocs; ++i) {
-                        if (ps[i].ps_proc) {
-                                pidlist[pid_nr] = ps[i].ps_pid;
-                                proclist[pid_nr++] = ps[i].ps_proc;
+                        for (i = 0; i < jb->nprocs; ++i) {
+                                if (jb->ps[i].ps_proc) {
+                                        proclist[pid_nr++] = jb->ps[i].ps_proc;
+                                }
                         }
                 }
         }
 
-        if (pid_nr == 0)
-                goto done;
-
-        idx = WaitForMultipleObjects(pid_nr, proclist, FALSE,
-                                wait_flags & WNOHANG ? 0 : INFINITE);
-        if (idx < pid_nr) {
-                GetExitCodeProcess(proclist[idx], &win_status);
-                *status = exit_code_to_wait_status(win_status);
-                pid = pidlist[idx];
+        if (pid_nr) {
+                idx = WaitForMultipleObjects(pid_nr, proclist, FALSE,
+                                        wait_flags & WNOHANG ? 0 : INFINITE);
+                if (idx < pid_nr) {
+                        GetExitCodeProcess(proclist[idx], &win_status);
+                        *status = exit_code_to_wait_status(win_status);
+                        pid = GetProcessId(proclist[idx]);
+                }
         }
- done:
-        free(pidlist);
-        free(proclist);
         return pid;
 }
 #define waitpid(p, s, f) waitpid_child(s, f)