[binutils-gdb] [gdbsupport] Improve thread scheduling in parallel_for_each

[Tom de Vries <vries@sourceware.org>]

https://sourceware.org/git/gitweb.cgi?p=binutils-gdb.git;h=4319180c8139e579a6337eb020309c95fa1e00b3

commit 4319180c8139e579a6337eb020309c95fa1e00b3
Author: Tom de Vries <tdevries@suse.de>
Date:   Mon Jul 18 08:34:06 2022 +0200

    [gdbsupport] Improve thread scheduling in parallel_for_each
    
    When running a task using parallel_for_each, we get the following
    distribution:
    ...
    Parallel for: n_elements: 7271
    Parallel for: minimum elements per thread: 10
    Parallel for: elts_per_thread: 1817
    Parallel for: elements on worker thread 0       : 1817
    Parallel for: elements on worker thread 1       : 1817
    Parallel for: elements on worker thread 2       : 1817
    Parallel for: elements on worker thread 3       : 0
    Parallel for: elements on main thread           : 1820
    ...
    
    Note that there are 4 active threads, and scheduling elts_per_thread on each
    of those handles 4 * 1817 == 7268, leaving 3 "left over" elements.
    
    These leftovers are currently handled in the main thread.
    
    That doesn't seem to matter much for this example, but for say 10 threads and
    99 elements, you'd have 9 threads handling 9 elements and 1 thread handling 18
    elements.
    
    Instead, distribute the left over elements over the worker threads, such that
    we have:
    ...
    Parallel for: elements on worker thread 0       : 1818
    Parallel for: elements on worker thread 1       : 1818
    Parallel for: elements on worker thread 2       : 1818
    Parallel for: elements on worker thread 3       : 0
    Parallel for: elements on main thread           : 1817
    ...
    
    Tested on x86_64-linux.

Diff:
---
 gdbsupport/parallel-for.h | 8 ++++++++
 1 file changed, 8 insertions(+)

diff --git a/gdbsupport/parallel-for.h b/gdbsupport/parallel-for.h
index cfe8a6e4f09..bf40f125f0f 100644
--- a/gdbsupport/parallel-for.h
+++ b/gdbsupport/parallel-for.h
@@ -147,6 +147,8 @@ parallel_for_each (unsigned n, RandomIt first, RandomIt last,
   size_t n_threads = n_worker_threads;
   size_t n_elements = last - first;
   size_t elts_per_thread = 0;
+  size_t elts_left_over = 0;
+
   if (n_threads > 1)
     {
       /* Require that there should be at least N elements in a
@@ -155,6 +157,8 @@ parallel_for_each (unsigned n, RandomIt first, RandomIt last,
       if (n_elements / n_threads < n)
 	n_threads = std::max (n_elements / n, (size_t) 1);
       elts_per_thread = n_elements / n_threads;
+      elts_left_over = n_elements % n_threads;
+      /* n_elements == n_threads * elts_per_thread + elts_left_over. */
     }
 
   size_t count = n_threads == 0 ? 0 : n_threads - 1;
@@ -170,6 +174,10 @@ parallel_for_each (unsigned n, RandomIt first, RandomIt last,
   for (int i = 0; i < count; ++i)
     {
       RandomIt end = first + elts_per_thread;
+      if (i < elts_left_over)
+	/* Distribute the leftovers over the worker threads, to avoid having
+	   to handle all of them in a single thread.  */
+	end++;
       if (parallel_for_each_debug)
 	debug_printf (_("Parallel for: elements on worker thread %i\t: %zu\n"),
 		      i, (size_t)(end - first));