| Message ID | 20190411114905.ivebkz234l4bflhb@kam.mff.cuni.cz |
|---|---|
| State | New |
| Headers | show |
| Series | Add parameter to limit LTO streaming parallelism | expand |
On Thu, 11 Apr 2019, Jan Hubicka wrote: > Hi, > the LTO streaming forks for every partition. With the number of > partitions incrased to 128 and relatively large memory usage (around > 5GB) needed to WPA firefox this causes kernel to spend a lot of time > probably by copying the page tables. > > This patch makes the streamer to for only lto_parallelism times > and strem num_partitions/lto_paralleism in each thread. > I have also added parameter because currently -flto=jobserv leads > to unlimited parallelism. This should be fixed by conneting to Make's > jobsever and build our own mini jobserver to distribute partitions > between worker threads, but this seems bit too involved for last minute > change in stage4. I plan to work on this and hopefully bacport it to .2 > release. > > I have tested the performance on by 32CPU 64threads box and got best > wall time with 32 partitions and therefore I set it by default. I get > > --param max-lto-streaming-parallelism=1 > Time variable usr sys wall GGC > phase stream out : 50.65 ( 30%) 20.66 ( 61%) 71.38 ( 35%) 921 kB ( 0%) > TOTAL : 170.73 33.69 204.64 7459610 kB > > --param max-lto-streaming-parallelism=4 > phase stream out : 13.79 ( 11%) 6.80 ( 35%) 20.94 ( 14%) 155 kB ( 0%) > TOTAL : 130.26 19.68 150.46 7458844 kB > > --param max-lto-streaming-parallelism=8 > phase stream out : 8.94 ( 7%) 5.21 ( 29%) 14.15 ( 10%) 83 kB ( 0%) > TOTAL : 125.28 18.09 143.54 7458773 kB > > --param max-lto-streaming-parallelism=16 > phase stream out : 4.56 ( 4%) 4.34 ( 25%) 9.46 ( 7%) 35 kB ( 0%) > TOTAL : 122.60 17.21 140.56 7458725 kB > > --param max-lto-streaming-parallelism=32 > phase stream out : 2.34 ( 2%) 5.69 ( 31%) 8.03 ( 6%) 15 kB ( 0%) > TOTAL : 118.53 18.36 137.08 7458705 kB > > --param max-lto-streaming-parallelism=64 > phase stream out : 1.63 ( 1%) 15.76 ( 55%) 17.40 ( 12%) 13 kB ( 0%) > TOTAL : 122.17 28.66 151.00 7458702 kB > > --param max-lto-streaming-parallelism=256 > phase stream out : 1.28 ( 1%) 9.24 ( 41%) 10.53 ( 8%) 13 kB ( 0%) > TOTAL : 116.78 22.56 139.53 7458702 kB > > Note that it is bit odd that 64 leads to worse results that full > parallelism but it seems to reproduce relatively well. Also the usr/sys > times for streaming are not representative since they do not account sys > time of the forked threads. I am not sure where the fork time is > accounted. > > Generally it seems that the forking performance is not at all that > bad and scales reasonably, but I still we should limit the default for > something less than 128 we do now. Definitly there are diminishing > returns after increasing from 16 or 32 and memory use goes up > noticeably. With current trunk memory use also does not seem terribly > bad (less global stream streaming makes the workers cheaper) and in all > memory traces I collected it is dominated by compilation stage during > the full rebuild. > > I did similar tests for cc1 binary. There the relative time spent in > streaming is lower so it goes from 17% to 1% (for parallelism 1 and 32 > respectively) > > Bootstrapped/regtested x86_64-linux, OK? Please document the new param in invoke.texi. Otherwise looks good to me. Btw, do we actually allocate garbage at write-out time? Thus, would using threads work as well? Thanks, Richard. > * params.def (PARAM_MAX_LTO_STREAMING_PARALLELISM): New parameter. > * lto.c (do_stream_out): rename to ... > (stream_out): ... this one; move original code to ... > (stream_out_partitions_1, stream_out_partitions): ... these new > functions. > (lto_wpa_write_files): Honnor lto_parallelism > Index: params.def > =================================================================== > --- params.def (revision 270143) > +++ params.def (working copy) > @@ -1146,6 +1146,11 @@ DEFPARAM (MAX_PARTITION_SIZE, > "Maximal size of a partition for LTO (in estimated instructions).", > 1000000, 0, INT_MAX) > > +DEFPARAM (PARAM_MAX_LTO_STREAMING_PARALLELISM, > + "max-lto-streaming-parallelism", > + "maximal number of LTO partitions streamed in parallel.", > + 32, 1, 0) > + > /* Diagnostic parameters. */ > > DEFPARAM (CXX_MAX_NAMESPACES_FOR_DIAGNOSTIC_HELP, > Index: lto/lto.c > =================================================================== > --- lto/lto.c (revision 270143) > +++ lto/lto.c (working copy) > @@ -2304,7 +2304,7 @@ static lto_file *current_lto_file; > /* Actually stream out ENCODER into TEMP_FILENAME. */ > > static void > -do_stream_out (char *temp_filename, lto_symtab_encoder_t encoder, int part) > +stream_out (char *temp_filename, lto_symtab_encoder_t encoder, int part) > { > lto_file *file = lto_obj_file_open (temp_filename, true); > if (!file) > @@ -2352,19 +2352,31 @@ wait_for_child () > } > #endif > > +static void > +stream_out_partitions_1 (char *temp_filename, int blen, int min, int max) > +{ > + /* Write all the nodes in SET. */ > + for (int p = min; p < max; p ++) > + { > + sprintf (temp_filename + blen, "%u.o", p); > + stream_out (temp_filename, ltrans_partitions[p]->encoder, p); > + ltrans_partitions[p]->encoder = NULL; > + } > +} > + > /* Stream out ENCODER into TEMP_FILENAME > Fork if that seems to help. */ > > static void > -stream_out (char *temp_filename, lto_symtab_encoder_t encoder, > - bool ARG_UNUSED (last), int part) > +stream_out_partitions (char *temp_filename, int blen, int min, int max, > + bool ARG_UNUSED (last)) > { > #ifdef HAVE_WORKING_FORK > static int nruns; > > if (lto_parallelism <= 1) > { > - do_stream_out (temp_filename, encoder, part); > + stream_out_partitions_1 (temp_filename, blen, min, max); > return; > } > > @@ -2384,12 +2396,12 @@ stream_out (char *temp_filename, lto_sym > if (!cpid) > { > setproctitle ("lto1-wpa-streaming"); > - do_stream_out (temp_filename, encoder, part); > + stream_out_partitions_1 (temp_filename, blen, min, max); > exit (0); > } > /* Fork failed; lets do the job ourseleves. */ > else if (cpid == -1) > - do_stream_out (temp_filename, encoder, part); > + stream_out_partitions_1 (temp_filename, blen, min, max); > else > nruns++; > } > @@ -2397,13 +2409,13 @@ stream_out (char *temp_filename, lto_sym > else > { > int i; > - do_stream_out (temp_filename, encoder, part); > + stream_out_partitions_1 (temp_filename, blen, min, max); > for (i = 0; i < nruns; i++) > wait_for_child (); > } > asm_nodes_output = true; > #else > - do_stream_out (temp_filename, encoder, part); > + stream_out_partitions_1 (temp_filename, blen, min, max); > #endif > } > > @@ -2445,6 +2457,13 @@ lto_wpa_write_files (void) > blen = strlen (temp_filename); > > n_sets = ltrans_partitions.length (); > + unsigned sets_per_worker = n_sets; > + if (lto_parallelism > 1) > + { > + if (lto_parallelism > (int)n_sets) > + lto_parallelism = n_sets; > + sets_per_worker = (n_sets + lto_parallelism - 1) / lto_parallelism; > + } > > for (i = 0; i < n_sets; i++) > { > @@ -2493,13 +2512,17 @@ lto_wpa_write_files (void) > } > gcc_checking_assert (lto_symtab_encoder_size (part->encoder) || !i); > > - stream_out (temp_filename, part->encoder, i == n_sets - 1, i); > - > - part->encoder = NULL; > - > temp_priority.safe_push (part->insns); > temp_filenames.safe_push (xstrdup (temp_filename)); > } > + > + for (int set = 0; set < MAX (lto_parallelism, 1); set++) > + { > + stream_out_partitions (temp_filename, blen, set * sets_per_worker, > + MIN ((set + 1) * sets_per_worker, n_sets), > + set == MAX (lto_parallelism, 1) - 1); > + } > + > ltrans_output_list_stream = fopen (ltrans_output_list, "w"); > if (ltrans_output_list_stream == NULL) > fatal_error (input_location, > @@ -3113,14 +3136,16 @@ do_whole_program_analysis (void) > > lto_parallelism = 1; > > - /* TODO: jobserver communicatoin is not supported, yet. */ > + /* TODO: jobserver communication is not supported, yet. */ > if (!strcmp (flag_wpa, "jobserver")) > - lto_parallelism = -1; > + lto_parallelism = PARAM_VALUE (PARAM_MAX_LTO_STREAMING_PARALLELISM); > else > { > lto_parallelism = atoi (flag_wpa); > if (lto_parallelism <= 0) > lto_parallelism = 0; > + if (lto_parallelism >= PARAM_VALUE (PARAM_MAX_LTO_STREAMING_PARALLELISM)) > + lto_parallelism = PARAM_VALUE (PARAM_MAX_LTO_STREAMING_PARALLELISM); > } > > timevar_start (TV_PHASE_OPT_GEN); >
> On Thu, 11 Apr 2019, Jan Hubicka wrote: > > > Hi, > > the LTO streaming forks for every partition. With the number of > > partitions incrased to 128 and relatively large memory usage (around > > 5GB) needed to WPA firefox this causes kernel to spend a lot of time > > probably by copying the page tables. > > > > This patch makes the streamer to for only lto_parallelism times > > and strem num_partitions/lto_paralleism in each thread. > > I have also added parameter because currently -flto=jobserv leads > > to unlimited parallelism. This should be fixed by conneting to Make's > > jobsever and build our own mini jobserver to distribute partitions > > between worker threads, but this seems bit too involved for last minute > > change in stage4. I plan to work on this and hopefully bacport it to .2 > > release. > > > > I have tested the performance on by 32CPU 64threads box and got best > > wall time with 32 partitions and therefore I set it by default. I get > > > > --param max-lto-streaming-parallelism=1 > > Time variable usr sys wall GGC > > phase stream out : 50.65 ( 30%) 20.66 ( 61%) 71.38 ( 35%) 921 kB ( 0%) > > TOTAL : 170.73 33.69 204.64 7459610 kB > > > > --param max-lto-streaming-parallelism=4 > > phase stream out : 13.79 ( 11%) 6.80 ( 35%) 20.94 ( 14%) 155 kB ( 0%) > > TOTAL : 130.26 19.68 150.46 7458844 kB > > > > --param max-lto-streaming-parallelism=8 > > phase stream out : 8.94 ( 7%) 5.21 ( 29%) 14.15 ( 10%) 83 kB ( 0%) > > TOTAL : 125.28 18.09 143.54 7458773 kB > > > > --param max-lto-streaming-parallelism=16 > > phase stream out : 4.56 ( 4%) 4.34 ( 25%) 9.46 ( 7%) 35 kB ( 0%) > > TOTAL : 122.60 17.21 140.56 7458725 kB > > > > --param max-lto-streaming-parallelism=32 > > phase stream out : 2.34 ( 2%) 5.69 ( 31%) 8.03 ( 6%) 15 kB ( 0%) > > TOTAL : 118.53 18.36 137.08 7458705 kB > > > > --param max-lto-streaming-parallelism=64 > > phase stream out : 1.63 ( 1%) 15.76 ( 55%) 17.40 ( 12%) 13 kB ( 0%) > > TOTAL : 122.17 28.66 151.00 7458702 kB > > > > --param max-lto-streaming-parallelism=256 > > phase stream out : 1.28 ( 1%) 9.24 ( 41%) 10.53 ( 8%) 13 kB ( 0%) > > TOTAL : 116.78 22.56 139.53 7458702 kB > > > > Note that it is bit odd that 64 leads to worse results that full > > parallelism but it seems to reproduce relatively well. Also the usr/sys > > times for streaming are not representative since they do not account sys > > time of the forked threads. I am not sure where the fork time is > > accounted. > > > > Generally it seems that the forking performance is not at all that > > bad and scales reasonably, but I still we should limit the default for > > something less than 128 we do now. Definitly there are diminishing > > returns after increasing from 16 or 32 and memory use goes up > > noticeably. With current trunk memory use also does not seem terribly > > bad (less global stream streaming makes the workers cheaper) and in all > > memory traces I collected it is dominated by compilation stage during > > the full rebuild. > > > > I did similar tests for cc1 binary. There the relative time spent in > > streaming is lower so it goes from 17% to 1% (for parallelism 1 and 32 > > respectively) > > > > Bootstrapped/regtested x86_64-linux, OK? > > Please document the new param in invoke.texi. Otherwise looks good > to me. Btw, do we actually allocate garbage at write-out time? > Thus, would using threads work as well? It is on my TODO to get this working. Last time i checked by adding abort into ggc_alloc there was some occurences but I think that can be cleanded up. I wonder how much performance hit we would get for enabling pthreads for lto1 binary and thus building libbackend with it? Honza
On Thu, 11 Apr 2019, Jan Hubicka wrote: > > On Thu, 11 Apr 2019, Jan Hubicka wrote: > > > > > Hi, > > > the LTO streaming forks for every partition. With the number of > > > partitions incrased to 128 and relatively large memory usage (around > > > 5GB) needed to WPA firefox this causes kernel to spend a lot of time > > > probably by copying the page tables. > > > > > > This patch makes the streamer to for only lto_parallelism times > > > and strem num_partitions/lto_paralleism in each thread. > > > I have also added parameter because currently -flto=jobserv leads > > > to unlimited parallelism. This should be fixed by conneting to Make's > > > jobsever and build our own mini jobserver to distribute partitions > > > between worker threads, but this seems bit too involved for last minute > > > change in stage4. I plan to work on this and hopefully bacport it to .2 > > > release. > > > > > > I have tested the performance on by 32CPU 64threads box and got best > > > wall time with 32 partitions and therefore I set it by default. I get > > > > > > --param max-lto-streaming-parallelism=1 > > > Time variable usr sys wall GGC > > > phase stream out : 50.65 ( 30%) 20.66 ( 61%) 71.38 ( 35%) 921 kB ( 0%) > > > TOTAL : 170.73 33.69 204.64 7459610 kB > > > > > > --param max-lto-streaming-parallelism=4 > > > phase stream out : 13.79 ( 11%) 6.80 ( 35%) 20.94 ( 14%) 155 kB ( 0%) > > > TOTAL : 130.26 19.68 150.46 7458844 kB > > > > > > --param max-lto-streaming-parallelism=8 > > > phase stream out : 8.94 ( 7%) 5.21 ( 29%) 14.15 ( 10%) 83 kB ( 0%) > > > TOTAL : 125.28 18.09 143.54 7458773 kB > > > > > > --param max-lto-streaming-parallelism=16 > > > phase stream out : 4.56 ( 4%) 4.34 ( 25%) 9.46 ( 7%) 35 kB ( 0%) > > > TOTAL : 122.60 17.21 140.56 7458725 kB > > > > > > --param max-lto-streaming-parallelism=32 > > > phase stream out : 2.34 ( 2%) 5.69 ( 31%) 8.03 ( 6%) 15 kB ( 0%) > > > TOTAL : 118.53 18.36 137.08 7458705 kB > > > > > > --param max-lto-streaming-parallelism=64 > > > phase stream out : 1.63 ( 1%) 15.76 ( 55%) 17.40 ( 12%) 13 kB ( 0%) > > > TOTAL : 122.17 28.66 151.00 7458702 kB > > > > > > --param max-lto-streaming-parallelism=256 > > > phase stream out : 1.28 ( 1%) 9.24 ( 41%) 10.53 ( 8%) 13 kB ( 0%) > > > TOTAL : 116.78 22.56 139.53 7458702 kB > > > > > > Note that it is bit odd that 64 leads to worse results that full > > > parallelism but it seems to reproduce relatively well. Also the usr/sys > > > times for streaming are not representative since they do not account sys > > > time of the forked threads. I am not sure where the fork time is > > > accounted. > > > > > > Generally it seems that the forking performance is not at all that > > > bad and scales reasonably, but I still we should limit the default for > > > something less than 128 we do now. Definitly there are diminishing > > > returns after increasing from 16 or 32 and memory use goes up > > > noticeably. With current trunk memory use also does not seem terribly > > > bad (less global stream streaming makes the workers cheaper) and in all > > > memory traces I collected it is dominated by compilation stage during > > > the full rebuild. > > > > > > I did similar tests for cc1 binary. There the relative time spent in > > > streaming is lower so it goes from 17% to 1% (for parallelism 1 and 32 > > > respectively) > > > > > > Bootstrapped/regtested x86_64-linux, OK? > > > > Please document the new param in invoke.texi. Otherwise looks good > > to me. Btw, do we actually allocate garbage at write-out time? > > Thus, would using threads work as well? > > It is on my TODO to get this working. Last time i checked by adding > abort into ggc_alloc there was some occurences but I think that can be > cleanded up. > > I wonder how much performance hit we would get for enabling pthreads for > lto1 binary and thus building libbackend with it? Is there any performance impact before the first thread creation? (besides eventually a few well-predicted if (threads_are_running) checks?) Richard.
Index: params.def =================================================================== --- params.def (revision 270143) +++ params.def (working copy) @@ -1146,6 +1146,11 @@ DEFPARAM (MAX_PARTITION_SIZE, "Maximal size of a partition for LTO (in estimated instructions).", 1000000, 0, INT_MAX) +DEFPARAM (PARAM_MAX_LTO_STREAMING_PARALLELISM, + "max-lto-streaming-parallelism", + "maximal number of LTO partitions streamed in parallel.", + 32, 1, 0) + /* Diagnostic parameters. */ DEFPARAM (CXX_MAX_NAMESPACES_FOR_DIAGNOSTIC_HELP, Index: lto/lto.c =================================================================== --- lto/lto.c (revision 270143) +++ lto/lto.c (working copy) @@ -2304,7 +2304,7 @@ static lto_file *current_lto_file; /* Actually stream out ENCODER into TEMP_FILENAME. */ static void -do_stream_out (char *temp_filename, lto_symtab_encoder_t encoder, int part) +stream_out (char *temp_filename, lto_symtab_encoder_t encoder, int part) { lto_file *file = lto_obj_file_open (temp_filename, true); if (!file) @@ -2352,19 +2352,31 @@ wait_for_child () } #endif +static void +stream_out_partitions_1 (char *temp_filename, int blen, int min, int max) +{ + /* Write all the nodes in SET. */ + for (int p = min; p < max; p ++) + { + sprintf (temp_filename + blen, "%u.o", p); + stream_out (temp_filename, ltrans_partitions[p]->encoder, p); + ltrans_partitions[p]->encoder = NULL; + } +} + /* Stream out ENCODER into TEMP_FILENAME Fork if that seems to help. */ static void -stream_out (char *temp_filename, lto_symtab_encoder_t encoder, - bool ARG_UNUSED (last), int part) +stream_out_partitions (char *temp_filename, int blen, int min, int max, + bool ARG_UNUSED (last)) { #ifdef HAVE_WORKING_FORK static int nruns; if (lto_parallelism <= 1) { - do_stream_out (temp_filename, encoder, part); + stream_out_partitions_1 (temp_filename, blen, min, max); return; } @@ -2384,12 +2396,12 @@ stream_out (char *temp_filename, lto_sym if (!cpid) { setproctitle ("lto1-wpa-streaming"); - do_stream_out (temp_filename, encoder, part); + stream_out_partitions_1 (temp_filename, blen, min, max); exit (0); } /* Fork failed; lets do the job ourseleves. */ else if (cpid == -1) - do_stream_out (temp_filename, encoder, part); + stream_out_partitions_1 (temp_filename, blen, min, max); else nruns++; } @@ -2397,13 +2409,13 @@ stream_out (char *temp_filename, lto_sym else { int i; - do_stream_out (temp_filename, encoder, part); + stream_out_partitions_1 (temp_filename, blen, min, max); for (i = 0; i < nruns; i++) wait_for_child (); } asm_nodes_output = true; #else - do_stream_out (temp_filename, encoder, part); + stream_out_partitions_1 (temp_filename, blen, min, max); #endif } @@ -2445,6 +2457,13 @@ lto_wpa_write_files (void) blen = strlen (temp_filename); n_sets = ltrans_partitions.length (); + unsigned sets_per_worker = n_sets; + if (lto_parallelism > 1) + { + if (lto_parallelism > (int)n_sets) + lto_parallelism = n_sets; + sets_per_worker = (n_sets + lto_parallelism - 1) / lto_parallelism; + } for (i = 0; i < n_sets; i++) { @@ -2493,13 +2512,17 @@ lto_wpa_write_files (void) } gcc_checking_assert (lto_symtab_encoder_size (part->encoder) || !i); - stream_out (temp_filename, part->encoder, i == n_sets - 1, i); - - part->encoder = NULL; - temp_priority.safe_push (part->insns); temp_filenames.safe_push (xstrdup (temp_filename)); } + + for (int set = 0; set < MAX (lto_parallelism, 1); set++) + { + stream_out_partitions (temp_filename, blen, set * sets_per_worker, + MIN ((set + 1) * sets_per_worker, n_sets), + set == MAX (lto_parallelism, 1) - 1); + } + ltrans_output_list_stream = fopen (ltrans_output_list, "w"); if (ltrans_output_list_stream == NULL) fatal_error (input_location, @@ -3113,14 +3136,16 @@ do_whole_program_analysis (void) lto_parallelism = 1; - /* TODO: jobserver communicatoin is not supported, yet. */ + /* TODO: jobserver communication is not supported, yet. */ if (!strcmp (flag_wpa, "jobserver")) - lto_parallelism = -1; + lto_parallelism = PARAM_VALUE (PARAM_MAX_LTO_STREAMING_PARALLELISM); else { lto_parallelism = atoi (flag_wpa); if (lto_parallelism <= 0) lto_parallelism = 0; + if (lto_parallelism >= PARAM_VALUE (PARAM_MAX_LTO_STREAMING_PARALLELISM)) + lto_parallelism = PARAM_VALUE (PARAM_MAX_LTO_STREAMING_PARALLELISM); } timevar_start (TV_PHASE_OPT_GEN);