Merged #2418 into master.
--
Reply to this email directly or view it on GitHub:
https://github.com/rpm-software-management/rpm/pull/2418#event-8886106505
You are receiving this because you are subscribed to this thread.
Message ID:
___
Rpm-maint
Okay, since there are no further comments...
--
Reply to this email directly or view it on GitHub:
https://github.com/rpm-software-management/rpm/pull/2418#issuecomment-1490100013
You are receiving this because you are subscribed to this thread.
Message ID:
OK, I can confirm the new push fixes the `proc` calculation, thanks!
--
Reply to this email directly or view it on GitHub:
https://github.com/rpm-software-management/rpm/pull/2418#issuecomment-1485189177
You are receiving this because you are subscribed to this thread.
Message ID:
@dmnks commented on this pull request.
> +long int pagesize = sysconf(_SC_PAGESIZE);
+long int pages = sysconf(_SC_PHYS_PAGES);
+
+if (pagesize < 0)
+ pagesize = 4096;
+if (pages > 0)
+ mem = pages * pagesize;
+
+return mem;
+}
+
+static unsigned long
@pmatilai commented on this pull request.
> +long int pagesize = sysconf(_SC_PAGESIZE);
+long int pages = sysconf(_SC_PHYS_PAGES);
+
+if (pagesize < 0)
+ pagesize = 4096;
+if (pages > 0)
+ mem = pages * pagesize;
+
+return mem;
+}
+
+static unsigned long
@pmatilai commented on this pull request.
> + * Conservative estimates for thread use on 32bit systems where address
+ * space is an issue: 2GB for bare metal, 3GB for a 32bit process
+ * on a 64bit system.
+ */
+if (thread) {
+ unsigned long vmem = mem;
+#if
@pmatilai commented on this pull request.
> @@ -1174,6 +1177,89 @@ static void doShescape(MacroBuf mb, rpmMacroEntry me,
> ARGV_t argv, size_t *parse
mbAppend(mb, '\'');
}
+static unsigned long getmem_total(void)
+{
+unsigned long mem = 0;
+long int pagesize =
@pmatilai pushed 3 commits.
c013e36fa47006ef8598a74450e709911a2a0b9d Add optional total/proc/thread
arguments to %{getncpus} macro
758e92b4b4d94792387c8999d2a258cd18225020 Make rpmio parallelism memory aware
73dad5c636524d1dec6f22df7de2a6a669ebbc70 Make build parallelism memory and
@pmatilai commented on this pull request.
> + * Conservative estimates for thread use on 32bit systems where address
+ * space is an issue: 2GB for bare metal, 3GB for a 32bit process
+ * on a 64bit system.
+ */
+if (thread) {
+ unsigned long vmem = mem;
+#if
Other than the above, the overall changeset looks sane to me.
--
Reply to this email directly or view it on GitHub:
https://github.com/rpm-software-management/rpm/pull/2418#issuecomment-1479932861
You are receiving this because you are subscribed to this thread.
Message ID:
@dmnks commented on this pull request.
> +long int pagesize = sysconf(_SC_PAGESIZE);
+long int pages = sysconf(_SC_PHYS_PAGES);
+
+if (pagesize < 0)
+ pagesize = 4096;
+if (pages > 0)
+ mem = pages * pagesize;
+
+return mem;
+}
+
+static unsigned long
@dmnks commented on this pull request.
> @@ -1174,6 +1177,89 @@ static void doShescape(MacroBuf mb, rpmMacroEntry me,
> ARGV_t argv, size_t *parse
mbAppend(mb, '\'');
}
+static unsigned long getmem_total(void)
+{
+unsigned long mem = 0;
+long int pagesize =
@dmnks commented on this pull request.
> @@ -1174,6 +1177,89 @@ static void doShescape(MacroBuf mb, rpmMacroEntry me,
> ARGV_t argv, size_t *parse
mbAppend(mb, '\'');
}
+static unsigned long getmem_total(void)
+{
+unsigned long mem = 0;
+long int pagesize =
@dmnks commented on this pull request.
> @@ -1174,6 +1177,89 @@ static void doShescape(MacroBuf mb, rpmMacroEntry me,
> ARGV_t argv, size_t *parse
mbAppend(mb, '\'');
}
+static unsigned long getmem_total(void)
+{
+unsigned long mem = 0;
+long int pagesize =
@dmnks commented on this pull request.
> + * Conservative estimates for thread use on 32bit systems where address
+ * space is an issue: 2GB for bare metal, 3GB for a 32bit process
+ * on a 64bit system.
+ */
+if (thread) {
+ unsigned long vmem = mem;
+#if __WORDSIZE
@dmnks commented on this pull request.
> + * Conservative estimates for thread use on 32bit systems where address
+ * space is an issue: 2GB for bare metal, 3GB for a 32bit process
+ * on a 64bit system.
+ */
+if (thread) {
+ unsigned long vmem = mem;
+#if __WORDSIZE
That's even better.
--
Reply to this email directly or view it on GitHub:
https://github.com/rpm-software-management/rpm/pull/2418#issuecomment-1477683506
You are receiving this because you are subscribed to this thread.
Message ID: ___
Rpm-maint
Well, the point of this PR is to make %limit_build and the like redundant
entirely.
--
Reply to this email directly or view it on GitHub:
https://github.com/rpm-software-management/rpm/pull/2418#issuecomment-1477342247
You are receiving this because you are subscribed to this thread.
Message
`%getmem` could be useful for making `%limit_build` work better in Fedora, so I
think it'd be worth having.
--
Reply to this email directly or view it on GitHub:
https://github.com/rpm-software-management/rpm/pull/2418#issuecomment-1476902969
You are receiving this because you are subscribed to
In my original version this also added a %getmem macro with similar arguments.
Left it out to keep things simple and minimal, but would be trivial to add back
if people think that's useful.
--
Reply to this email directly or view it on GitHub:
@pmatilai pushed 1 commit.
e4d0e62d129569a714ae190b2ffd2fcc1296834d Make build parallelism memory and
address-space aware (#804)
--
View it on GitHub:
@pmatilai pushed 2 commits.
c859ebe35c54e58f689bc3d2097ff0acbe8172fa Make rpmio parallelism memory aware
3353d6a4234dd8e2ead28fc7cffdae9cc68e24b1 Make build parallelism memory and
address-space aware (#804)
--
View it on GitHub:
@pmatilai pushed 3 commits.
28b7237b30ad02ad2babee4d8abb13c96ae360fa Add optional total/proc/thread
arguments to %{getncpus} macro
b09677daadfa7025eb2ff89560c0a886e043302c Make rpmio parallelism memory aware
753763a1f971c0e041f934c442337a012eb7df5b Make build parallelism memory and
See commits for details, but short summary: add optional proc/thread argument
to %{getncpus} macro which consider the available memory and address space,
based on newly added tunables for tasksize.
The goal here is to avoid build failures due to gross overallocation on
constrained systems,
24 matches
Mail list logo
