Re: COBOL: Hoping for insight with middle-end computation time.

David Malcolm via Gcc <[email protected]>
Newsgroups gmane.comp.gcc.devel
Message-ID <[email protected]> 
On Fri, 2026-03-20 at 10:55 -0500, Robert Dubner wrote:
> 
> 
> > -----Original Message-----
> > From: David Malcolm <[email protected]>
> > Sent: Friday, March 20, 2026 10:10
> > To: Robert Dubner <[email protected]>; [email protected]
> > Subject: Re: COBOL: Hoping for insight with middle-end computation
> > time.
> > 
> > On Thu, 2026-03-19 at 18:22 -0500, Robert Dubner wrote:
> > > It happens that COBOL has the COMPUTE statement.  It takes the
> > > form
> > > of,
> > > for example, COMPUTE DDD = AAA + BBB.
> > > 
> > > We implement that by creating a temporary variable, using that as
> > > the
> > > target of an addition of AAA and BBB, and then doing an
> > > assignment to
> > > DDD.
> > > (Recall that COBOL variables can be quite complex, so we are a
> > > long
> > > way
> > > from being able to do this with an ADD_EXPR.)
> > > 
> > > We have determined that the way I've been producing GENERIC for
> > > that
> > > results in N-squared computation time somewhere in the middle
> > > end.
> > > 
> > > I have been tearing my hair out trying to figure out what's
> > > causing
> > > that
> > > N-squared behavior.  I commented away the assignment, and I got
> > > rid
> > > of the
> > > arithmetic.  All that's left is the creation of the temporary,
> > > and
> > > some IF
> > > statements that are generated to test for errors along the way of
> > > the
> > > computation.  (COBOL has a very rich error-detection and
> > > exception-generating facility.
> > > 
> > > The remaining GIMPLE for a single iteration (as shown by
> > > -fdump-tree-gimple) is shown below.  The "phase opt and generate"
> > > times
> > > for repetitions of that GIMPLE are shown here:
> > > 
> > >         phase opt	   Factor
> > > Repeats & generate
> > >   1,000       0.17
> > >   2,000       0.49      2.9
> > >   4,000       1.55      3.2
> > >   8,000       7.56      4.9
> > >  16,000      49.31      6.5
> > >  32,000     281.29      5.7
> > > 
> > > I have been struggling with this for days.  Is there an
> > > explanation
> > > for
> > > why the following GIMPLE is resulting in that N-squared behavior?
> > 
> > Hi Bob.  You cite the overall "phase opt and generate" times, but
> > no
> > data on how this is spread across the various optimization passes.
> > 
> > What's the output of -ftime-report on your workload?
> > 
> > In particular, are there any specific passes that are responsible
> > for
> > the growth in time (and thus where we can pinpoint a bug), or is
> > the
> > time evenly distributed across all of them?
> > 
> > Sorry if this is a silly question
> > Dave
> 
> The only silly thing is that in an egregious display of monumental
> arrogance 
> and ignorance, some years back I decided to take on the problem of
> code 
> generation for a new front end without having had any prior
> experience with 
> GCC internals or compiler theory, and without access to anybody who
> actually 
> knew anything.
> 
> I hope you've seen my response to Richard.  For a compilation with
> 
> phase opt and generate             :  14.95 ( 92%)   259M ( 88%)
> 
> the next big component is
> 
> thread pro- & epilogue             :  10.45 ( 64%)  4104  (  0%)
> 
> What that means is yet one more mystery to me.

64% of the wallclock time is being accounted to this timing item. 
Looking in timevar.def, I see that this timing item is:

DEFTIMEVAR (TV_THREAD_PROLOGUE_AND_EPILOGUE, "thread pro- & epilogue")

Grepping for TV_THREAD_PROLOGUE_AND_EPILOGUE, I see two passes in
function.cc that account their time to this timevar:
pass_thread_prologue_and_epilogue and
pass_late_thread_prologue_and_epilogue.

Both of these passes ultimately call the function
rest_of_handle_thread_prologue_and_epilogue.

So the slowdown is presumably somewhere inside there.

Dave




> 
> Here's the entire output of -ftime-report-details:
> 
> Time variable                                  wall           GGC
>  phase setup                        :   0.01 (  0%)   150k (  0%)
>  phase parsing                      :   1.33 (  8%)    36M ( 12%)
>  phase opt and generate             :  14.95 ( 92%)   259M ( 88%)
>  phase last asm                     :   0.02 (  0%)   377k (  0%)
>  phase finalize                     :   0.01 (  0%)     0  (  0%)
>  garbage collection                 :   0.09 (  1%)     0  (  0%)
>  callgraph construction             :   0.13 (  1%)    12M (  4%)
>  `- CFG verifier                    :   0.02 (  0%)     0  (  0%)
>  `- tree STMT verifier              :   0.16 (  1%)     0  (  0%)
>  `- symout                          :   0.01 (  0%)    10M (  4%)
>  `- tree SSA verifier               :   0.04 (  0%)     0  (  0%)
>  `- garbage collection              :   0.01 (  0%)     0  (  0%)
>  callgraph optimization             :   0.04 (  0%)     0  (  0%)
>  `- dominance computation           :   0.01 (  0%)     0  (  0%)
>  `- CFG verifier                    :   0.01 (  0%)     0  (  0%)
>  `- tree STMT verifier              :   0.21 (  1%)     0  (  0%)
>  `- tree SSA verifier               :   0.06 (  0%)     0  (  0%)
>  `- garbage collection              :   0.03 (  0%)     0  (  0%)
>  callgraph ipa passes               :   1.14 (  7%)    35M ( 12%)
>  ipa function summary               :   0.00 (  0%)  1832  (  0%)
>  `- tree SSA verifier               :   0.01 (  0%)     0  (  0%)
>  `- tree STMT verifier              :   0.02 (  0%)     0  (  0%)
>  ipa inlining heuristics            :   0.00 (  0%)     0  (  0%)
>  `- tree SSA verifier               :   0.01 (  0%)     0  (  0%)
>  `- tree STMT verifier              :   0.02 (  0%)     0  (  0%)
>  ipa comdats                        :   0.00 (  0%)     0  (  0%)
>  `- tree SSA verifier               :   0.01 (  0%)     0  (  0%)
>  `- tree STMT verifier              :   0.02 (  0%)     0  (  0%)
>  ipa free lang data                 :   0.00 (  0%)     0  (  0%)
>  `- tree STMT verifier              :   0.02 (  0%)     0  (  0%)
>  ipa free inline summary            :   0.00 (  0%)     0  (  0%)
>  `- tree SSA verifier               :   0.01 (  0%)     0  (  0%)
>  `- tree STMT verifier              :   0.04 (  0%)     0  (  0%)
>  ipa modref                         :   0.00 (  0%)     0  (  0%)
>  `- tree SSA verifier               :   0.01 (  0%)     0  (  0%)
>  `- tree STMT verifier              :   0.02 (  0%)     0  (  0%)
>  cfg construction                   :   0.01 (  0%)  1232  (  0%)
>  `- rebuild jump labels             :   0.01 (  0%)     0  (  0%)
>  `- verify RTL sharing              :   0.03 (  0%)     0  (  0%)
>  `- CFG verifier                    :   0.03 (  0%)     0  (  0%)
>  cfg cleanup                        :   0.01 (  0%)   208  (  0%)
>  `- CFG verifier                    :   0.03 (  0%)     0  (  0%)
>  CFG verifier                       :   0.40 (  2%)     0  (  0%)
>  trivially dead code                :   0.02 (  0%)     0  (  0%)
>  df scan insns                      :   0.07 (  0%)    96  (  0%)
>  `- verify RTL sharing              :   0.01 (  0%)     0  (  0%)
>  `- CFG verifier                    :   0.01 (  0%)     0  (  0%)
>  df live regs                       :   0.03 (  0%)     0  (  0%)
>  df reg dead/unused notes           :   0.03 (  0%)  2628k (  1%)
>  register information               :   0.01 (  0%)     0  (  0%)
>  alias analysis                     :   0.01 (  0%)  1024k (  0%)
>  rebuild jump labels                :   0.01 (  0%)   168  (  0%)
>  parser (global)                    :   1.33 (  8%)    36M ( 12%)
>  early inlining heuristics          :   0.00 (  0%)     0  (  0%)
>  `- tree SSA verifier               :   0.01 (  0%)     0  (  0%)
>  `- tree STMT verifier              :   0.02 (  0%)     0  (  0%)
>  inline parameters                  :   0.03 (  0%)   672  (  0%)
>  `- tree SSA verifier               :   0.02 (  0%)     0  (  0%)
>  `- tree STMT verifier              :   0.04 (  0%)     0  (  0%)
>  tree gimplify                      :   0.09 (  1%)    50M ( 17%)
>  tree eh                            :   0.01 (  0%)   584  (  0%)
>  tree CFG construction              :   0.02 (  0%)    13M (  5%)
>  `- tree STMT verifier              :   0.02 (  0%)     0  (  0%)
>  tree CFG cleanup                   :   0.02 (  0%)     0  (  0%)
>  `- tree SSA verifier               :   0.01 (  0%)     0  (  0%)
>  `- tree STMT verifier              :   0.02 (  0%)     0  (  0%)
>  `- dominance computation           :   0.01 (  0%)     0  (  0%)
>  `- CFG verifier                    :   0.01 (  0%)     0  (  0%)
>  tree SSA other                     :   0.00 (  0%)     0  (  0%)
>  `- tree SSA verifier               :   0.01 (  0%)     0  (  0%)
>  `- tree STMT verifier              :   0.02 (  0%)     0  (  0%)
>  tree SSA rewrite                   :   0.03 (  0%)    20M (  7%)
>  `- tree STMT verifier              :   0.02 (  0%)     0  (  0%)
>  `- tree operand scan               :   0.01 (  0%)  8470k (  3%)
>  `- tree SSA verifier               :   0.01 (  0%)     0  (  0%)
>  tree operand scan                  :   0.01 (  0%)  8470k (  3%)
>  tree SSA verifier                  :   0.36 (  2%)     0  (  0%)
>  `- dominance computation           :   0.06 (  0%)     0  (  0%)
>  tree STMT verifier                 :   0.88 (  5%)     0  (  0%)
>  tree switch lowering               :   0.00 (  0%)     0  (  0%)
>  `- tree SSA verifier               :   0.01 (  0%)     0  (  0%)
>  `- tree STMT verifier              :   0.02 (  0%)     0  (  0%)
>  callgraph verifier                 :   0.06 (  0%)     0  (  0%)
>  `- callgraph verifier              :   0.06 (  0%)     0  (  0%)
>  dominance computation              :   0.12 (  1%)     0  (  0%)
>  out of ssa                         :   0.03 (  0%)   952  (  0%)
>  expand vars                        :   0.01 (  0%)  7040k (  2%)
>  expand                             :   0.16 (  1%)    86M ( 29%)
>  `- CFG verifier                    :   0.01 (  0%)     0  (  0%)
>  `- verify RTL sharing              :   0.01 (  0%)     0  (  0%)
>  `- out of ssa                      :   0.03 (  0%)   952  (  0%)
>  `- expand vars                     :   0.01 (  0%)  7040k (  2%)
>  `- post expand cleanups            :   0.01 (  0%)    96  (  0%)
>  post expand cleanups               :   0.01 (  0%)  3712  (  0%)
>  `- rebuild jump labels             :   0.01 (  0%)   168  (  0%)
>  `- CFG verifier                    :   0.01 (  0%)     0  (  0%)
>  jump                               :   0.00 (  0%)     0  (  0%)
>  `- verify RTL sharing              :   0.03 (  0%)     0  (  0%)
>  `- trivially dead code             :   0.01 (  0%)     0  (  0%)
>  `- CFG verifier                    :   0.02 (  0%)     0  (  0%)
>  loop init                          :   0.01 (  0%)  3040  (  0%)
>  mode switching                     :   0.00 (  0%)     0  (  0%)
>  `- verify RTL sharing              :   0.01 (  0%)     0  (  0%)
>  `- CFG verifier                    :   0.01 (  0%)     0  (  0%)
>  integrated RA                      :   0.49 (  3%)    12M (  4%)
>  `- register information            :   0.01 (  0%)     0  (  0%)
>  `- verify RTL sharing              :   0.02 (  0%)     0  (  0%)
>  `- df reg dead/unused notes        :   0.03 (  0%)  2628k (  1%)
>  `- alias analysis                  :   0.01 (  0%)  1024k (  0%)
>  `- trivially dead code             :   0.01 (  0%)     0  (  0%)
>  `- CFG verifier                    :   0.01 (  0%)     0  (  0%)
>  `- df live regs                    :   0.01 (  0%)     0  (  0%)
>  LRA non-specific                   :   0.22 (  1%)   352  (  0%)
>  `- LRA hard reg assignment         :   0.01 (  0%)     0  (  0%)
>  `- LRA virtuals elimination        :   0.17 (  1%)    23M (  8%)
>  `- LRA create live ranges          :   0.01 (  0%)     0  (  0%)
>  LRA virtuals elimination           :   0.17 (  1%)    23M (  8%)
>  LRA create live ranges             :   0.01 (  0%)     0  (  0%)
>  LRA hard reg assignment            :   0.01 (  0%)     0  (  0%)
>  reload                             :   0.00 (  0%)    48  (  0%)
>  `- verify RTL sharing              :   0.02 (  0%)     0  (  0%)
>  `- integrated RA                   :   0.05 (  0%)    96  (  0%)
>  `- LRA non-specific                :   0.06 (  0%)    24  (  0%)
>  `- CFG verifier                    :   0.01 (  0%)     0  (  0%)
>  thread pro- & epilogue             :  10.45 ( 64%)  4104  (  0%)
>  `- CFG verifier                    :   0.03 (  0%)     0  (  0%)
>  `- verify RTL sharing              :   0.02 (  0%)     0  (  0%)
>  `- df live regs                    :   0.01 (  0%)     0  (  0%)
>  machine dep reorg                  :   0.00 (  0%)     0  (  0%)
>  `- verify RTL sharing              :   0.02 (  0%)     0  (  0%)
>  shorten branches                   :   0.05 (  0%)     0  (  0%)
>  `- verify RTL sharing              :   0.02 (  0%)     0  (  0%)
>  reg stack                          :   0.00 (  0%)     0  (  0%)
>  `- verify RTL sharing              :   0.04 (  0%)     0  (  0%)
>  `- CFG verifier                    :   0.02 (  0%)     0  (  0%)
>  final                              :   0.10 (  1%)  3887k (  1%)
>  `- verify RTL sharing              :   0.04 (  0%)     0  (  0%)
>  `- symout                          :   0.01 (  0%)  7192k (  2%)
>  symout                             :   0.04 (  0%)    18M (  6%)
>  access analysis                    :   0.06 (  0%)    24  (  0%)
>  `- tree SSA verifier               :   0.02 (  0%)     0  (  0%)
>  `- tree STMT verifier              :   0.04 (  0%)     0  (  0%)
>  `- dominance computation           :   0.01 (  0%)     0  (  0%)
>  early local passes                 :   0.00 (  0%)     0  (  0%)
>  `- tree STMT verifier              :   0.02 (  0%)     0  (  0%)
>  rest of compilation                :   0.10 (  1%)  1754k (  1%)
>  `- dominance computation           :   0.01 (  0%)     0  (  0%)
>  `- verify RTL sharing              :   0.24 (  1%)     0  (  0%)
>  `- garbage collection              :   0.04 (  0%)     0  (  0%)
>  `- tree STMT verifier              :   0.13 (  1%)     0  (  0%)
>  `- tree SSA verifier               :   0.06 (  0%)     0  (  0%)
>  `- CFG verifier                    :   0.13 (  1%)     0  (  0%)
>  unaccounted post reload            :   0.00 (  0%)     0  (  0%)
>  `- verify RTL sharing              :   0.02 (  0%)     0  (  0%)
>  `- CFG verifier                    :   0.01 (  0%)     0  (  0%)
>  unaccounted late compilation       :   0.00 (  0%)     0  (  0%)
>  `- verify RTL sharing              :   0.02 (  0%)     0  (  0%)
>  `- CFG verifier                    :   0.01 (  0%)     0  (  0%)
>  verify RTL sharing                 :   0.54 (  3%)     0  (  0%)
>  TOTAL                              :  16.32          296M
> 
> 
> > 
> > 
> > > 
> > > I agree that the conditionals are a bit convoluted, and Jim and I
> > > are
> > > working to do COMPUTE in a much improved way.  But I still feel a
> > > need to
> > > understand what's going on!
> > > 
> > > Thanks so much for any insight into what might be happening here.
> > > 
> > >       _intermediate__stack501_503.0.0.data =
> > > &_stack501_data_517.0;
> > >       _intermediate__stack501_503.0.0.capacity = 16;
> > >       _intermediate__stack501_503.0.0.allocated = 16;
> > >       _intermediate__stack501_503.0.0.offset = 0;
> > >       _intermediate__stack501_503.0.0.name = &"_stack501"[0];
> > >       _intermediate__stack501_503.0.0.picture = &""[0];
> > >       _intermediate__stack501_503.0.0.initial = 0B;
> > >       _intermediate__stack501_503.0.0.parent = 0B;
> > >       _intermediate__stack501_503.0.0.occurs_lower = 0;
> > >       _intermediate__stack501_503.0.0.occurs_upper = 0;
> > >       _intermediate__stack501_503.0.0.attr = 4160;
> > >       _intermediate__stack501_503.0.0.type = 6;
> > >       _intermediate__stack501_503.0.0.level = 0;
> > >       _intermediate__stack501_503.0.0.digits = 37;
> > >       _intermediate__stack501_503.0.0.rdigits = 0;
> > >       _intermediate__stack501_503.0.0.encoding = 1;
> > >       _intermediate__stack501_503.0.0.alphabet = 0;
> > >       D.2812 = 0;
> > >       D.2813 = 0;
> > >       _1013 = D.2812 & 18;
> > >       if (_1013 != 0) goto <D.9353>; else goto <D.9354>;
> > >       <D.9353>:
> > >       goto <D.9355>;
> > >       <D.9354>:
> > >       D.2814 = 0;
> > >       ..pa_erf.1519_1014 = ..pa_erf;
> > >       D.2813 = D.2813 | ..pa_erf.1519_1014;
> > >       if (D.2813 != 0) goto <D.9357>; else goto <D.9358>;
> > >       <D.9357>:
> > >       goto <D.9359>;
> > >       <D.9358>:
> > >       <D.9359>:
> > >       <D.9355>:
> > > 
> > > 
>
lmpx.com only provides a reader for public news (NNTP) servers. It is not affiliated with the servers or forums shown here and is not responsible for the content of articles, which is written by their respective authors.