diff options
Diffstat (limited to 'erts/doc/src/erl.xml')
| -rw-r--r-- | erts/doc/src/erl.xml | 168 |
1 files changed, 104 insertions, 64 deletions
diff --git a/erts/doc/src/erl.xml b/erts/doc/src/erl.xml index 39c79a29df..cfbc38f176 100644 --- a/erts/doc/src/erl.xml +++ b/erts/doc/src/erl.xml @@ -41,25 +41,11 @@ to scroll back to text which has scrolled off the screen. The <c><![CDATA[erl]]></c> program must be used, however, in pipelines or if you want to redirect standard input or output.</p> - <note><p>As of ERTS version 5.8 (OTP-R14A) the runtime system will by - default bind schedulers to logical processors using the - <c>default_bind</c> bind type if the amount of schedulers are - at least equal to the amount of logical processors configured, - binding of schedulers is supported, and a CPU topology is - available at startup. - </p><p> - If the Erlang runtime system is the only operating system - process that binds threads to logical processors, this - improves the performance of the runtime system. However, - if other operating system processes (as for example - another Erlang runtime system) also bind threads to - logical processors, there might be a performance penalty - instead. If this is the case you, are are advised to - unbind the schedulers using the - <seealso marker="#+sbt">+sbtu</seealso> command line argument, - or by invoking - <seealso marker="erlang#system_flag_scheduler_bind_type">erlang:system_flag(scheduler_bind_type, - unbound)</seealso>.</p> + <note><p>As of ERTS version 5.9 (OTP-R15B) the runtime system will by + default <em>not</em> bind schedulers to logical processors. + For more information see documentation of the + <seealso marker="#+sbt">+sbt</seealso> system flag. + </p> </note> </description> <funcs> @@ -679,41 +665,66 @@ </p> <taglist> <tag><c>u</c></tag> - <item><p>Same as - <seealso marker="erlang#system_flag_scheduler_bind_type">erlang:system_flag(scheduler_bind_type, unbound)</seealso>. - </p></item> + <item> + <p><c>unbound</c> - Schedulers will not be bound to logical + processors, i.e., the operating system decides where the + scheduler threads execute, and when to migrate them. This is + the default.</p> + </item> <tag><c>ns</c></tag> - <item><p>Same as - <seealso marker="erlang#system_flag_scheduler_bind_type">erlang:system_flag(scheduler_bind_type, no_spread)</seealso>. - </p></item> + <item> + <p><c>no_spread</c> - Schedulers with close scheduler + identifiers will be bound as close as possible in hardware.</p> + </item> <tag><c>ts</c></tag> - <item><p>Same as - <seealso marker="erlang#system_flag_scheduler_bind_type">erlang:system_flag(scheduler_bind_type, thread_spread)</seealso>. - </p></item> + <item> + <p><c>thread_spread</c> - Thread refers to hardware threads + (e.g. Intel's hyper-threads). Schedulers with low scheduler + identifiers, will be bound to the first hardware thread of + each core, then schedulers with higher scheduler identifiers + will be bound to the second hardware thread of each core, + etc.</p> + </item> <tag><c>ps</c></tag> - <item><p>Same as - <seealso marker="erlang#system_flag_scheduler_bind_type">erlang:system_flag(scheduler_bind_type, processor_spread)</seealso>. - </p></item> + <item> + <p><c>processor_spread</c> - Schedulers will be spread like + <c>thread_spread</c>, but also over physical processor chips.</p> + </item> <tag><c>s</c></tag> - <item><p>Same as - <seealso marker="erlang#system_flag_scheduler_bind_type">erlang:system_flag(scheduler_bind_type, spread)</seealso>. - </p></item> + <item> + <p><c>spread</c> - Schedulers will be spread as much as + possible.</p> + </item> <tag><c>nnts</c></tag> - <item><p>Same as - <seealso marker="erlang#system_flag_scheduler_bind_type">erlang:system_flag(scheduler_bind_type, no_node_thread_spread)</seealso>. - </p></item> + <item> + <p><c>no_node_thread_spread</c> - Like <c>thread_spread</c>, + but if multiple NUMA (Non-Uniform Memory Access) nodes exists, + schedulers will be spread over one NUMA node at a time, + i.e., all logical processors of one NUMA node will be bound + to schedulers in sequence.</p> + </item> <tag><c>nnps</c></tag> - <item><p>Same as - <seealso marker="erlang#system_flag_scheduler_bind_type">erlang:system_flag(scheduler_bind_type, no_node_processor_spread)</seealso>. - </p></item> + <item> + <p><c>no_node_processor_spread</c> - Like + <c>processor_spread</c>, but if multiple NUMA nodes exists, + schedulers will be spread over one NUMA node at a time, i.e., + all logical processors of one NUMA node will be bound to + schedulers in sequence.</p> + </item> <tag><c>tnnps</c></tag> - <item><p>Same as - <seealso marker="erlang#system_flag_scheduler_bind_type">erlang:system_flag(scheduler_bind_type, thread_no_node_processor_spread)</seealso>. - </p></item> + <item> + <p><c>thread_no_node_processor_spread</c> - A combination of + <c>thread_spread</c>, and <c>no_node_processor_spread</c>. + Schedulers will be spread over hardware threads across NUMA + nodes, but schedulers will only be spread over processors + internally in one NUMA node at a time.</p> + </item> <tag><c>db</c></tag> - <item><p>Same as - <seealso marker="erlang#system_flag_scheduler_bind_type">erlang:system_flag(scheduler_bind_type, default_bind)</seealso>. - </p></item> + <item> + <p><c>default_bind</c> - Binds schedulers the default way. + Currently the default is <c>thread_no_node_processor_spread</c> + (which might change in the future).</p> + </item> </taglist> <p>Binding of schedulers is currently only supported on newer Linux, Solaris, FreeBSD, and Windows systems.</p> @@ -725,24 +736,47 @@ that the <c>+sct</c> flag may have to be passed before the <c>+sbt</c> flag on the command line (in case no CPU topology has been automatically detected).</p> - <p>The runtime system will by default bind schedulers to logical - processors using the <c>default_bind</c> bind type if the amount - of schedulers are at least equal to the amount of logical - processors configured, binding of schedulers is supported, - and a CPU topology is available at startup. + <p>The runtime system will by default <em>not</em> bind schedulers + to logical processors. </p> - <p><em>NOTE:</em> If the Erlang runtime system is the only operating - system process that binds threads to logical processors, this - improves the performance of the runtime system. However, if other - operating system processes (as for example another Erlang runtime - system) also bind threads to logical processors, there might be a - performance penalty instead. If this is the case you, are advised - to unbind the schedulers using the <c>+sbtu</c> command line - argument, or by invoking - <seealso marker="erlang#system_flag_scheduler_bind_type">erlang:system_flag(scheduler_bind_type, - unbound)</seealso>.</p> - <p>For more information, see - <seealso marker="erlang#system_flag_scheduler_bind_type">erlang:system_flag(scheduler_bind_type, SchedulerBindType)</seealso>. + <p><em>NOTE:</em> If the Erlang runtime system is the only operating system + process that binds threads to logical processors, this + improves the performance of the runtime system. However, + if other operating system processes (as for example + another Erlang runtime system) also bind threads to + logical processors, there might be a performance penalty + instead. In some cases this performance penalty might be + severe. If this is the case, you are advised to not + bind the schedulers.</p> + <p>How schedulers are bound matters. For example, in + situations when there are fewer running processes than + schedulers online, the runtime system tries to migrate + processes to schedulers with low scheduler identifiers. + The more the schedulers are spread over the hardware, + the more resources will be available to the runtime + system in such situations. + </p> + <p> + <em>NOTE:</em> If a scheduler fails to bind, this + will often be silently ignored. This since it isn't always + possible to verify valid logical processor identifiers. If + an error is reported, it will be reported to the + <c>error_logger</c>. If you want to verify that the + schedulers actually have bound as requested, call + <seealso marker="erlang#system_info_scheduler_bindings">erlang:system_info(scheduler_bindings)</seealso>. + </p> + </item> + <tag><marker id="+scl"><c>+scl true|false</c></marker></tag> + <item> + <p>Enable or disable scheduler compaction of load. By default + scheduler compaction of load is enabled. When enabled, load + balancing will strive for a load distribution which causes + as many scheduler threads as possible to be fully loaded (i.e., + not run out of work). This is accomplished by migrating load + (e.g. runnable processes) into a smaller set of schedulers + when schedulers frequently run out of work. When disabled, + the frequency with which schedulers run out of work will + not be taken into account by the load balancing logic. </p> </item> <tag><marker id="+sct"><c>+sct CpuTopology</c></marker></tag> @@ -760,6 +794,12 @@ <item><c><![CDATA[<IdDefs> = <LogicalIds><ThreadIds><CoreIds><ProcessorIds><NodeIds> | <LogicalIds><ThreadIds><CoreIds><NodeIds><ProcessorIds>]]></c></item> <item><c><![CDATA[CpuTopology = <IdDefs>:<IdDefs> | <IdDefs>]]></c></item> </list> + <p>Set a user defined CPU topology. The user defined + CPU topology will override any automatically detected + CPU topology. The CPU topology is used when + <seealso marker="#+sbt">binding schedulers to logical + processors</seealso>. + </p> <p>Upper-case letters signify real identifiers and lower-case letters signify fake identifiers only used for description of the topology. Identifiers passed as real identifiers may @@ -859,7 +899,7 @@ how the real CPU topology looks like is likely to decrease the performance of the runtime system.</p> <p>For more information, see - <seealso marker="erlang#system_flag_cpu_topology">erlang:system_flag(cpu_topology, CpuTopology)</seealso>.</p> + <seealso marker="erlang#system_info_cpu_topology">erlang:system_info(cpu_topology)</seealso>.</p> </item> <tag><marker id="+swt"><c>+swt very_low|low|medium|high|very_high</c></marker></tag> <item> |