[v3,1/8] sched-deps.cc (find_modifiable_mems): Avoid exponential behavior

Message ID 20231122111415.815147-2-maxim.kuvyrkov@linaro.org
State Committed
Commit 0c42d1782e48d8ad578ace2065cce9b3615f97c0
Headers
Series [v3,1/8] sched-deps.cc (find_modifiable_mems): Avoid exponential behavior |

Commit Message

Maxim Kuvyrkov Nov. 22, 2023, 11:14 a.m. UTC
  This patch avoids sched-deps.cc:find_inc() creating exponential number
of dependencies, which become memory and compilation time hogs.
Consider example (simplified from PR96388) ...
===
sp=sp-4 // sp_insnA
mem_insnA1[sp+A1]
...
mem_insnAN[sp+AN]
sp=sp-4 // sp_insnB
mem_insnB1[sp+B1]
...
mem_insnBM[sp+BM]
===

[For simplicity, let's assume find_inc(backwards==true)].
In this example find_modifiable_mems() will arrange for mem_insnA*
to be able to pass sp_insnA, and, while doing this, will create
dependencies between all mem_insnA*s and sp_insnB -- because sp_insnB
is a consumer of sp_insnA.  After this sp_insnB will have N new
backward dependencies.
Then find_modifiable_mems() gets to mem_insnB*s and starts to create
N new dependencies for _every_ mem_insnB*.  This gets us N*M new
dependencies.

In PR96833's testcase N and M are 10k-15k, which causes RAM usage of
30GB and compilation time of 30 minutes, with sched2 accounting for
95% of both metrics.  After this patch the RAM usage is down to 1GB
and compilation time is down to 3-4 minutes, with sched2 no longer
standing out on -ftime-report or memory usage.

gcc/ChangeLog:

	PR rtl-optimization/96388
	PR rtl-optimization/111554
	* sched-deps.cc (find_inc): Avoid exponential behavior.
---
 gcc/sched-deps.cc | 48 +++++++++++++++++++++++++++++++++++++++++++----
 1 file changed, 44 insertions(+), 4 deletions(-)
  

Comments

Maxim Kuvyrkov Jan. 15, 2024, 12:56 p.m. UTC | #1
Hi Vladimir,
Hi Jeff,

Richard and Alexander have reviewed this patch and [I assume] have no
further comments.  OK to merge?

On Wed, 22 Nov 2023 at 15:14, Maxim Kuvyrkov <maxim.kuvyrkov@linaro.org>
wrote:

> This patch avoids sched-deps.cc:find_inc() creating exponential number
> of dependencies, which become memory and compilation time hogs.
> Consider example (simplified from PR96388) ...
> ===
> sp=sp-4 // sp_insnA
> mem_insnA1[sp+A1]
> ...
> mem_insnAN[sp+AN]
> sp=sp-4 // sp_insnB
> mem_insnB1[sp+B1]
> ...
> mem_insnBM[sp+BM]
> ===
>
> [For simplicity, let's assume find_inc(backwards==true)].
> In this example find_modifiable_mems() will arrange for mem_insnA*
> to be able to pass sp_insnA, and, while doing this, will create
> dependencies between all mem_insnA*s and sp_insnB -- because sp_insnB
> is a consumer of sp_insnA.  After this sp_insnB will have N new
> backward dependencies.
> Then find_modifiable_mems() gets to mem_insnB*s and starts to create
> N new dependencies for _every_ mem_insnB*.  This gets us N*M new
> dependencies.
>
> In PR96833's testcase N and M are 10k-15k, which causes RAM usage of
> 30GB and compilation time of 30 minutes, with sched2 accounting for
> 95% of both metrics.  After this patch the RAM usage is down to 1GB
> and compilation time is down to 3-4 minutes, with sched2 no longer
> standing out on -ftime-report or memory usage.
>
> gcc/ChangeLog:
>
>         PR rtl-optimization/96388
>         PR rtl-optimization/111554
>         * sched-deps.cc (find_inc): Avoid exponential behavior.
> ---
>  gcc/sched-deps.cc | 48 +++++++++++++++++++++++++++++++++++++++++++----
>  1 file changed, 44 insertions(+), 4 deletions(-)
>
> diff --git a/gcc/sched-deps.cc b/gcc/sched-deps.cc
> index c23218890f3..005fc0f567e 100644
> --- a/gcc/sched-deps.cc
> +++ b/gcc/sched-deps.cc
> @@ -4779,24 +4779,59 @@ parse_add_or_inc (struct mem_inc_info *mii,
> rtx_insn *insn, bool before_mem)
>  /* Once a suitable mem reference has been found and the corresponding data
>     in MII has been filled in, this function is called to find a suitable
>     add or inc insn involving the register we found in the memory
> -   reference.  */
> +   reference.
> +   If successful, this function will create additional dependencies
> between
> +   - mii->inc_insn's producers and mii->mem_insn as a consumer (if
> backwards)
> +   - mii->inc_insn's consumers and mii->mem_insn as a producer (if
> !backwards).
> +*/
>
>  static bool
>  find_inc (struct mem_inc_info *mii, bool backwards)
>  {
>    sd_iterator_def sd_it;
>    dep_t dep;
> +  sd_list_types_def mem_deps = backwards ? SD_LIST_HARD_BACK :
> SD_LIST_FORW;
> +  int n_mem_deps = sd_lists_size (mii->mem_insn, mem_deps);
>
> -  sd_it = sd_iterator_start (mii->mem_insn,
> -                            backwards ? SD_LIST_HARD_BACK : SD_LIST_FORW);
> +  sd_it = sd_iterator_start (mii->mem_insn, mem_deps);
>    while (sd_iterator_cond (&sd_it, &dep))
>      {
>        dep_node_t node = DEP_LINK_NODE (*sd_it.linkp);
>        rtx_insn *pro = DEP_PRO (dep);
>        rtx_insn *con = DEP_CON (dep);
> -      rtx_insn *inc_cand = backwards ? pro : con;
> +      rtx_insn *inc_cand;
> +      int n_inc_deps;
> +
>        if (DEP_NONREG (dep) || DEP_MULTIPLE (dep))
>         goto next;
> +
> +      if (backwards)
> +       {
> +         inc_cand = pro;
> +         n_inc_deps = sd_lists_size (inc_cand, SD_LIST_BACK);
> +       }
> +      else
> +       {
> +         inc_cand = con;
> +         n_inc_deps = sd_lists_size (inc_cand, SD_LIST_FORW);
> +       }
> +
> +      /* In the FOR_EACH_DEP loop below we will create additional
> n_inc_deps
> +        for mem_insn.  This by itself is not a problem, since each
> mem_insn
> +        will have only a few inc_insns associated with it.  However, if
> +        we consider that a single inc_insn may have a lot of mem_insns,
> AND,
> +        on top of that, a few other inc_insns associated with it --
> +        those _other inc_insns_ will get (n_mem_deps * number of MEM
> insns)
> +        dependencies created for them.  This may cause an exponential
> +        growth of memory usage and scheduling time.
> +        See PR96388 for details.
> +        We [heuristically] use n_inc_deps as a proxy for the number of MEM
> +        insns, and drop opportunities for breaking modifiable_mem
> dependencies
> +        when dependency lists grow beyond reasonable size.  */
> +      if (n_mem_deps * n_inc_deps
> +         >= param_max_pending_list_length * param_max_pending_list_length)
> +       goto next;
> +
>        if (parse_add_or_inc (mii, inc_cand, backwards))
>         {
>           struct dep_replacement *desc;
> @@ -4838,6 +4873,11 @@ find_inc (struct mem_inc_info *mii, bool backwards)
>           desc->insn = mii->mem_insn;
>           move_dep_link (DEP_NODE_BACK (node), INSN_HARD_BACK_DEPS (con),
>                          INSN_SPEC_BACK_DEPS (con));
> +
> +         /* Make sure that n_inc_deps above is consistent with
> dependencies
> +            we create.  */
> +         gcc_assert (mii->inc_insn == inc_cand);
> +
>           if (backwards)
>             {
>               FOR_EACH_DEP (mii->inc_insn, SD_LIST_BACK, sd_it, dep)
> --
> 2.34.1
>
>
  
Vladimir Makarov Jan. 15, 2024, 6:26 p.m. UTC | #2
On 1/15/24 07:56, Maxim Kuvyrkov wrote:
> Hi Vladimir,
> Hi Jeff,
>
> Richard and Alexander have reviewed this patch and [I assume] have no 
> further comments.  OK to merge?
>
>
I trust Richard and Alexander therefore I did not do additional review 
of the patches and have no any comment.  Richard's or Alexander's 
approval is enough for comitting the patches.
  
Jeff Law Jan. 16, 2024, 2:52 p.m. UTC | #3
On 1/15/24 05:56, Maxim Kuvyrkov wrote:
> Hi Vladimir,
> Hi Jeff,
> 
> Richard and Alexander have reviewed this patch and [I assume] have no 
> further comments.  OK to merge?
I think the question is whether or not we're too late.  I know that 
Richard S has held off on his late-combine pass and I'm holding off on 
the ext-dce work due to the fact that we're well past stage1 close.

I think the release managers ought to have the final say on this.

jeff
  
Richard Biener Jan. 17, 2024, 6:51 a.m. UTC | #4
On Tue, Jan 16, 2024 at 3:52 PM Jeff Law <jeffreyalaw@gmail.com> wrote:
>
>
>
> On 1/15/24 05:56, Maxim Kuvyrkov wrote:
> > Hi Vladimir,
> > Hi Jeff,
> >
> > Richard and Alexander have reviewed this patch and [I assume] have no
> > further comments.  OK to merge?
> I think the question is whether or not we're too late.  I know that
> Richard S has held off on his late-combine pass and I'm holding off on
> the ext-dce work due to the fact that we're well past stage1 close.
>
> I think the release managers ought to have the final say on this.

I'm fine with this now, it doesn't change code generation.

Richard.

>
> jeff
  
Maxim Kuvyrkov Jan. 17, 2024, 7:39 a.m. UTC | #5
> On Jan 17, 2024, at 10:51, Richard Biener <richard.guenther@gmail.com> wrote:
> 
> On Tue, Jan 16, 2024 at 3:52 PM Jeff Law <jeffreyalaw@gmail.com> wrote:
>> 
>> 
>> 
>> On 1/15/24 05:56, Maxim Kuvyrkov wrote:
>>> Hi Vladimir,
>>> Hi Jeff,
>>> 
>>> Richard and Alexander have reviewed this patch and [I assume] have no
>>> further comments.  OK to merge?
>> I think the question is whether or not we're too late.  I know that
>> Richard S has held off on his late-combine pass and I'm holding off on
>> the ext-dce work due to the fact that we're well past stage1 close.
>> 
>> I think the release managers ought to have the final say on this.
> 
> I'm fine with this now, it doesn't change code generation.

Thanks, Richard.

I'll merge the fix for PR96388 and PR111554 -- patch 1/8.  I'll commit cleanups and improvements to scheduler logging -- patches 2/8 - 8/8 -- when stage1 opens.

Regards,

--
Maxim Kuvyrkov
https://www.linaro.org
  
Richard Biener Jan. 17, 2024, 3:02 p.m. UTC | #6
On Wed, Jan 17, 2024 at 8:39 AM Maxim Kuvyrkov
<maxim.kuvyrkov@linaro.org> wrote:
>
> > On Jan 17, 2024, at 10:51, Richard Biener <richard.guenther@gmail.com> wrote:
> >
> > On Tue, Jan 16, 2024 at 3:52 PM Jeff Law <jeffreyalaw@gmail.com> wrote:
> >>
> >>
> >>
> >> On 1/15/24 05:56, Maxim Kuvyrkov wrote:
> >>> Hi Vladimir,
> >>> Hi Jeff,
> >>>
> >>> Richard and Alexander have reviewed this patch and [I assume] have no
> >>> further comments.  OK to merge?
> >> I think the question is whether or not we're too late.  I know that
> >> Richard S has held off on his late-combine pass and I'm holding off on
> >> the ext-dce work due to the fact that we're well past stage1 close.
> >>
> >> I think the release managers ought to have the final say on this.
> >
> > I'm fine with this now, it doesn't change code generation.
>
> Thanks, Richard.
>
> I'll merge the fix for PR96388 and PR111554 -- patch 1/8.  I'll commit cleanups and improvements to scheduler logging -- patches 2/8 - 8/8 -- when stage1 opens.

This seems to have caused a compare-debug bootstrap issue on x86_64-linux,

gcc/fortran/f95-lang.o differs

does n_mem_deps or n_inc_deps include debug insns?

Richard.

> Regards,
>
> --
> Maxim Kuvyrkov
> https://www.linaro.org
>
  
Maxim Kuvyrkov Jan. 17, 2024, 3:05 p.m. UTC | #7
> On Jan 17, 2024, at 19:02, Richard Biener <richard.guenther@gmail.com> wrote:
> 
> On Wed, Jan 17, 2024 at 8:39 AM Maxim Kuvyrkov
> <maxim.kuvyrkov@linaro.org> wrote:
>> 
>>> On Jan 17, 2024, at 10:51, Richard Biener <richard.guenther@gmail.com> wrote:
>>> 
>>> On Tue, Jan 16, 2024 at 3:52 PM Jeff Law <jeffreyalaw@gmail.com> wrote:
>>>> 
>>>> 
>>>> 
>>>> On 1/15/24 05:56, Maxim Kuvyrkov wrote:
>>>>> Hi Vladimir,
>>>>> Hi Jeff,
>>>>> 
>>>>> Richard and Alexander have reviewed this patch and [I assume] have no
>>>>> further comments.  OK to merge?
>>>> I think the question is whether or not we're too late.  I know that
>>>> Richard S has held off on his late-combine pass and I'm holding off on
>>>> the ext-dce work due to the fact that we're well past stage1 close.
>>>> 
>>>> I think the release managers ought to have the final say on this.
>>> 
>>> I'm fine with this now, it doesn't change code generation.
>> 
>> Thanks, Richard.
>> 
>> I'll merge the fix for PR96388 and PR111554 -- patch 1/8.  I'll commit cleanups and improvements to scheduler logging -- patches 2/8 - 8/8 -- when stage1 opens.
> 
> This seems to have caused a compare-debug bootstrap issue on x86_64-linux,
> 
> gcc/fortran/f95-lang.o differs
> 
> does n_mem_deps or n_inc_deps include debug insns?

Thanks, investigating.

--
Maxim Kuvyrkov
https://www.linaro.org
  
Maxim Kuvyrkov Jan. 17, 2024, 3:44 p.m. UTC | #8
> On Jan 17, 2024, at 19:05, Maxim Kuvyrkov <maxim.kuvyrkov@linaro.org> wrote:
> 
>> On Jan 17, 2024, at 19:02, Richard Biener <richard.guenther@gmail.com> wrote:
>> 
>> On Wed, Jan 17, 2024 at 8:39 AM Maxim Kuvyrkov
>> <maxim.kuvyrkov@linaro.org> wrote:
>>> 
>>>> On Jan 17, 2024, at 10:51, Richard Biener <richard.guenther@gmail.com> wrote:
>>>> 
>>>> On Tue, Jan 16, 2024 at 3:52 PM Jeff Law <jeffreyalaw@gmail.com> wrote:
>>>>> 
>>>>> 
>>>>> 
>>>>> On 1/15/24 05:56, Maxim Kuvyrkov wrote:
>>>>>> Hi Vladimir,
>>>>>> Hi Jeff,
>>>>>> 
>>>>>> Richard and Alexander have reviewed this patch and [I assume] have no
>>>>>> further comments.  OK to merge?
>>>>> I think the question is whether or not we're too late.  I know that
>>>>> Richard S has held off on his late-combine pass and I'm holding off on
>>>>> the ext-dce work due to the fact that we're well past stage1 close.
>>>>> 
>>>>> I think the release managers ought to have the final say on this.
>>>> 
>>>> I'm fine with this now, it doesn't change code generation.
>>> 
>>> Thanks, Richard.
>>> 
>>> I'll merge the fix for PR96388 and PR111554 -- patch 1/8.  I'll commit cleanups and improvements to scheduler logging -- patches 2/8 - 8/8 -- when stage1 opens.
>> 
>> This seems to have caused a compare-debug bootstrap issue on x86_64-linux,
>> 
>> gcc/fortran/f95-lang.o differs
>> 
>> does n_mem_deps or n_inc_deps include debug insns?
> 
> Thanks, investigating.

Hi Richard,

Yes, both n_mem_deps or n_inc_deps include debug insns, I posted a patch for this in https://gcc.gnu.org/pipermail/gcc-patches/2024-January/643267.html .  Testing it now.

If you prefer, I can revert the fix for PR96388 and PR111554.

Kind regards,

--
Maxim Kuvyrkov
https://www.linaro.org
  
H.J. Lu Jan. 17, 2024, 6:54 p.m. UTC | #9
On Wed, Jan 17, 2024 at 7:02 AM Richard Biener
<richard.guenther@gmail.com> wrote:
>
> On Wed, Jan 17, 2024 at 8:39 AM Maxim Kuvyrkov
> <maxim.kuvyrkov@linaro.org> wrote:
> >
> > > On Jan 17, 2024, at 10:51, Richard Biener <richard.guenther@gmail.com> wrote:
> > >
> > > On Tue, Jan 16, 2024 at 3:52 PM Jeff Law <jeffreyalaw@gmail.com> wrote:
> > >>
> > >>
> > >>
> > >> On 1/15/24 05:56, Maxim Kuvyrkov wrote:
> > >>> Hi Vladimir,
> > >>> Hi Jeff,
> > >>>
> > >>> Richard and Alexander have reviewed this patch and [I assume] have no
> > >>> further comments.  OK to merge?
> > >> I think the question is whether or not we're too late.  I know that
> > >> Richard S has held off on his late-combine pass and I'm holding off on
> > >> the ext-dce work due to the fact that we're well past stage1 close.
> > >>
> > >> I think the release managers ought to have the final say on this.
> > >
> > > I'm fine with this now, it doesn't change code generation.
> >
> > Thanks, Richard.
> >
> > I'll merge the fix for PR96388 and PR111554 -- patch 1/8.  I'll commit cleanups and improvements to scheduler logging -- patches 2/8 - 8/8 -- when stage1 opens.
>
> This seems to have caused a compare-debug bootstrap issue on x86_64-linux,
>
> gcc/fortran/f95-lang.o differs
>
> does n_mem_deps or n_inc_deps include debug insns?
>
> Richard.

FWIW, I opened:

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=113456
  

Patch

diff --git a/gcc/sched-deps.cc b/gcc/sched-deps.cc
index c23218890f3..005fc0f567e 100644
--- a/gcc/sched-deps.cc
+++ b/gcc/sched-deps.cc
@@ -4779,24 +4779,59 @@  parse_add_or_inc (struct mem_inc_info *mii, rtx_insn *insn, bool before_mem)
 /* Once a suitable mem reference has been found and the corresponding data
    in MII has been filled in, this function is called to find a suitable
    add or inc insn involving the register we found in the memory
-   reference.  */
+   reference.
+   If successful, this function will create additional dependencies between
+   - mii->inc_insn's producers and mii->mem_insn as a consumer (if backwards)
+   - mii->inc_insn's consumers and mii->mem_insn as a producer (if !backwards).
+*/
 
 static bool
 find_inc (struct mem_inc_info *mii, bool backwards)
 {
   sd_iterator_def sd_it;
   dep_t dep;
+  sd_list_types_def mem_deps = backwards ? SD_LIST_HARD_BACK : SD_LIST_FORW;
+  int n_mem_deps = sd_lists_size (mii->mem_insn, mem_deps);
 
-  sd_it = sd_iterator_start (mii->mem_insn,
-			     backwards ? SD_LIST_HARD_BACK : SD_LIST_FORW);
+  sd_it = sd_iterator_start (mii->mem_insn, mem_deps);
   while (sd_iterator_cond (&sd_it, &dep))
     {
       dep_node_t node = DEP_LINK_NODE (*sd_it.linkp);
       rtx_insn *pro = DEP_PRO (dep);
       rtx_insn *con = DEP_CON (dep);
-      rtx_insn *inc_cand = backwards ? pro : con;
+      rtx_insn *inc_cand;
+      int n_inc_deps;
+
       if (DEP_NONREG (dep) || DEP_MULTIPLE (dep))
 	goto next;
+
+      if (backwards)
+	{
+	  inc_cand = pro;
+	  n_inc_deps = sd_lists_size (inc_cand, SD_LIST_BACK);
+	}
+      else
+	{
+	  inc_cand = con;
+	  n_inc_deps = sd_lists_size (inc_cand, SD_LIST_FORW);
+	}
+
+      /* In the FOR_EACH_DEP loop below we will create additional n_inc_deps
+	 for mem_insn.  This by itself is not a problem, since each mem_insn
+	 will have only a few inc_insns associated with it.  However, if
+	 we consider that a single inc_insn may have a lot of mem_insns, AND,
+	 on top of that, a few other inc_insns associated with it --
+	 those _other inc_insns_ will get (n_mem_deps * number of MEM insns)
+	 dependencies created for them.  This may cause an exponential
+	 growth of memory usage and scheduling time.
+	 See PR96388 for details.
+	 We [heuristically] use n_inc_deps as a proxy for the number of MEM
+	 insns, and drop opportunities for breaking modifiable_mem dependencies
+	 when dependency lists grow beyond reasonable size.  */
+      if (n_mem_deps * n_inc_deps
+	  >= param_max_pending_list_length * param_max_pending_list_length)
+	goto next;
+
       if (parse_add_or_inc (mii, inc_cand, backwards))
 	{
 	  struct dep_replacement *desc;
@@ -4838,6 +4873,11 @@  find_inc (struct mem_inc_info *mii, bool backwards)
 	  desc->insn = mii->mem_insn;
 	  move_dep_link (DEP_NODE_BACK (node), INSN_HARD_BACK_DEPS (con),
 			 INSN_SPEC_BACK_DEPS (con));
+
+	  /* Make sure that n_inc_deps above is consistent with dependencies
+	     we create.  */
+	  gcc_assert (mii->inc_insn == inc_cand);
+
 	  if (backwards)
 	    {
 	      FOR_EACH_DEP (mii->inc_insn, SD_LIST_BACK, sd_it, dep)