Re: [PATCH v4 2/8] grep: skip pthreads overhead when using one thread

From: Ævar Arnfjörð Bjarmason <avarab@gmail.com>
Date: Jun 1, 2017, 21:36 UTC
Message-ID: <CACBZZX7hffa3iGndzyJMKYAwDqjjYO6XacWLrHnSo29xYSKAsQ@mail.gmail.com>
In-Reply-To: <CAGZ79kb6O7hraY4caY8tdFn1d0Fi+LRr9cHk2UuXf79LbnPdhw@mail.gmail.com>

On Thu, Jun 1, 2017 at 11:20 PM, Stefan Beller <sbeller@google.com> wrote:

Show 39 quoted lines

> On Thu, Jun 1, 2017 at 11:20 AM, Ævar Arnfjörð Bjarmason
> <avarab@gmail.com> wrote:
>
>> +       if (num_threads == 1)
>> +               num_threads = 0;
>
> I would think that it is easier to maintain the code when keep the 1
> hard coded, and apply the following diff instead. If we encounter
> a 0 later on, it is not clear what the original user input was.
> (Did the user ask for 0 as a proxy for GREP_NUM_THREADS_DEFAULT ?
> do they care about the number of threads?)
> It is less complexity in the decision logic here.
>
> --8<-- (white space broken)
> diff --git a/builtin/grep.c b/builtin/grep.c
> index c6c26e9b9e..6ad9b3da20 100644
> --- a/builtin/grep.c
> +++ b/builtin/grep.c
> @@ -1231,7 +1231,7 @@ int cmd_grep(int argc, const char **argv, const
> char *prefix)
>  #endif
>
>  #ifndef NO_PTHREADS
> -       if (num_threads) {
> +       if (num_threads > 1) {
>                 if (!(opt.name_only || opt.unmatch_name_only || opt.count)
>                     && (opt.pre_context || opt.post_context ||
>                         opt.file_break || opt.funcbody))
> @@ -1295,7 +1295,7 @@ int cmd_grep(int argc, const char **argv, const
> char *prefix)
>                 hit = grep_objects(&opt, &pathspec, &list);
>         }
>
> -       if (num_threads)
> +       if (num_threads > 1)
>                 hit |= wait_all();
>         if (hit && show_in_pager)
>                 run_pager(&opt, prefix);
> --8<--

If I've understood you correctly (what I applied on top of this based on the above at the end of the mail) this segfaults because now we won't compile the pattern.

I agree that this logic is a bit tricky, but it must be considered with the "!num_threads" logic in preceding "grep: don't redundantly compile throwaway patterns under threading" patch.

I.e. we already have num_threads being assigned to 0 under NO_PTHREADS, doing the same for the PTHREADS codepath seemed like the simplest solution to me.

diff --git a/builtin/grep.c b/builtin/grep.c
index a0a3922f92..6d16df2526 100644
--- a/builtin/grep.c
+++ b/builtin/grep.c
@@ -1238,25 +1238,23 @@ int cmd_grep(int argc, const char **argv,
const char *prefix)
                num_threads = GREP_NUM_THREADS_DEFAULT;
        else if (num_threads < 0)
                die(_("invalid number of threads specified (%d)"), num_threads);
-       if (num_threads == 1)
-               num_threads = 0;
 #else
        if (num_threads)
                warning(_("no threads support, ignoring --threads"));
-       num_threads = 0;
+       num_threads = 0; /* If we have no threads... */
 #endif

-       if (!num_threads)
+       if (num_threads) /* Or if we've decided not to use them... */
                /*
                 * The compiled patterns on the main path are only
                 * used when not using threading. Otherwise
                 * start_threads() below calls compile_grep_patterns()
                 * for each thread.
                 */
-               compile_grep_patterns(&opt);
+               compile_grep_patterns(&opt); /* We'll compile the
pattern here */
 #ifndef NO_PTHREADS
-       if (num_threads) {
+       if (num_threads > 1) {
                if (!(opt.name_only || opt.unmatch_name_only || opt.count)
                    && (opt.pre_context || opt.post_context ||
                        opt.file_break || opt.funcbody))
@@ -1320,7 +1318,7 @@ int cmd_grep(int argc, const char **argv, const
char *prefix)
                hit = grep_objects(&opt, &pathspec, &list);
        }

-       if (num_threads)
+       if (num_threads > 1)
                hit |= wait_all();
        if (hit && show_in_pager)
                run_pager(&opt, prefix);