I see with -march=native on haswell in addition to -Ofast -flto ~5% runtime regression for 481.wrf with the last known good rev. being r269093 and the first known bad one r269146.
Also seen here: https://lnt.opensuse.org/db_default/v4/SPEC/graph?plot.0=2.270.0&plot.1=7.270.0&plot.2=4.270.0&plot.3=8.270.0 I'll bisect that.
Suspicious changes include the fix for PR87609 and the new pass_remove_partial_avx_dependency. I see no other relevant changes.
(In reply to Richard Biener from comment #2) > Suspicious changes include the fix for PR87609 and the new > pass_remove_partial_avx_dependency. I see no other relevant changes. Yes, started with r269098.
Investigating, but I guess there's nothing to fix :/
So it looks like the change from r269097 to r269098 is only 2.5% (I've included the followup fix ontop of r269098): 481.wrf 11170 245 45.7 * 11170 251 44.5 S 481.wrf 11170 246 45.4 S 11170 251 44.5 * Checking r269096 reveals (BASE = r269096, PEAK = r269098 + followup): 481.wrf 11170 244 45.7 S 11170 254 44.0 S 481.wrf 11170 244 45.8 * 11170 252 44.3 * Note that fortran guarantees on parameter aliasing are stronger than those of restrict qualified pointers so using restrict isn't perfect and it might suffer from the correctness fix unnecessarily. In a similar fashion performing inlining might make PTA do more conservative choices than when looking at the fnspec guarantees. For r269096 there's the possibility of simply never recomputing restrict (gate it with !cfun->after_inlining). Overall I see 9.90% 201223 wrf_peak.amd64- wrf_peak.amd64-m64-gcc42-nn [.] solve_interface_ 8.50% 172668 wrf_base.amd64- wrf_base.amd64-m64-gcc42-nn [.] solve_interface_ which explains why we only see this with -flto (this function does nothing but call other functions...). It doesn't really explain why r269096 makes such a big difference though (I guess it must be the PRE PTA recompute triggering this). But the function is a huge mess and the profile quite flat and similar...
One difference that is clearly visible is missed vectorization of module_small_step_em.fppized.f90:1399:14: note: LOOP VECTORIZED module_small_step_em.fppized.f90:1376:14: note: LOOP VECTORIZED -module_small_step_em.fppized.f90:1376:14: note: LOOP VECTORIZED module_small_step_em.fppized.f90:1354:14: note: LOOP VECTORIZED This is the following inner loop which we vectorize before the change but not after. Before we have create runtime check for data references MEM[(float[0:D.14326] *)_891][_13070] and MEM[(float[0:D.14326] *)_891][_4424] module_small_step_em.fppized.f90:1376:14: note: created 1 versioning for alias checks. module_small_step_em.fppized.f90:1376:14: optimized: loop versioned for vectorization because of possible aliasing where afterwards module_small_step_em.fppized.f90:1376:14: missed: number of versioning for alias run-time tests exceeds 10 (--param vect-max-version-for-alias-checks) SUBROUTINE advance_w( w, rw_tend, ww, u, v, & ... ! Jammed 3 doubly nested loops over k/i into 1 for slight improvement ! in efficiency. No change in results (bit-for-bit). JM 20040514 ! (left a blank line where the other two k/i-loops were) ! DO k=2,k_end DO i=i_start, i_end w(i,k,j)=w(i,k,j)+dts*rw_tend(i,k,j) & + msft_inv(i)*cqw(i,k,j)*( & +.5*dts*g*mut_inv(i)*rdn(k)* & (c2a(i,k ,j)*rdnw(k ) & *((1.+epssm)*(rhs(i,k+1 )-rhs(i,k )) & +(1.-epssm)*(ph(i,k+1,j)-ph(i,k ,j))) & -c2a(i,k-1,j)*rdnw(k-1) & *((1.+epssm)*(rhs(i,k )-rhs(i,k-1 )) & +(1.-epssm)*(ph(i,k ,j)-ph(i,k-1,j))))) & +dts*g*msft_inv(i)*(rdn(k)* & (c2a(i,k ,j)*alt(i,k ,j)*t_2ave(i,k ,j) & -c2a(i,k-1,j)*alt(i,k-1,j)*t_2ave(i,k-1,j)) & +(rdn(k)*(c2a(i,k ,j)*alb(i,k ,j) & -c2a(i,k-1,j)*alb(i,k-1,j))*mut_inv(i)-1.) & *muave(i,j)) ENDDO ENDDO There is almost no difference in the IL after if-conversion but changes like - _14697 = MEM[(float[0:D.14440] *)_959 clique 40 base 47][_15604]; + _14697 = MEM[(float[0:D.14440] *)_959 clique 71 base 47][_15604]; _9355 = _14697 * 4.905000209808349609375e+0; - _9371 = MEM[(float[0:D.14436] *)_957 clique 40 base 28][_15604]; + _9371 = MEM[(float[0:D.14436] *)_957 clique 71 base 28][_15604]; which is introduced by unroll-and-jam by means of unrolling the outer loop once. The restrict clique remapping is strictly speaking required here. There are of course cases where remapping can be elided but they are difficult to reconstruct. Basically each clique > 1 belongs to a specific inline block and whether we need to remap it when copying a stmt depends on whether we duplicate "within" that block or we duplicate the block itself. A good "hint" would be (for example for unroll-and-jam) whether cliques mentioned in the to be copied loop are also mentioned in stmts belonging exclusively to the outer loop that is unrolled (but then they could have been moved there by invariant motion for example). I'm not sure if we want to look at the BLOCK tree. Or whether we want to build a similar thing into the loop structure, noting inline contexts (the single clique that needs not remapping when copying the loop). Note I haven't been able to verify whether the above loop is executed at all but -fno-loop-unroll-and-jam makes performance equal (and faster than ever...).
Fixed.
Author: rguenth Date: Fri Mar 8 10:20:12 2019 New Revision: 269484 URL: https://gcc.gnu.org/viewcvs?rev=269484&root=gcc&view=rev Log: 2019-03-08 Richard Biener <rguenther@suse.de> PR middle-end/89578 * cfgloop.h (struct loop): Add owned_clique field. * cfgloopmanip.c (copy_loop_info): Copy it. * tree-cfg.c (gimple_duplicate_bb): Do not remap owned_clique cliques. * tree-inline.c (copy_loops): Remap owned_clique. * lto-streamer-in.c (input_cfg): Stream owned_clique. * lto-streamer-out.c (output_cfg): Likewise. Modified: trunk/gcc/ChangeLog trunk/gcc/cfgloop.h trunk/gcc/cfgloopmanip.c trunk/gcc/lto-streamer-in.c trunk/gcc/lto-streamer-out.c trunk/gcc/tree-cfg.c trunk/gcc/tree-inline.c
Author: rguenth Date: Fri Aug 30 11:11:01 2019 New Revision: 275069 URL: https://gcc.gnu.org/viewcvs?rev=275069&root=gcc&view=rev Log: 2019-08-30 Richard Biener <rguenther@suse.de> * lto-streamer.h (LTO_minor_version): Bump. Backport from mainline 2019-05-06 Richard Biener <rguenther@suse.de> PR tree-optimization/90328 * tree-data-ref.h (dr_may_alias_p): Pass in the actual loop nest. * tree-data-ref.c (dr_may_alias_p): Check whether the clique is valid in the loop nest before using it. (initialize_data_dependence_relation): Adjust. * graphite-scop-detection.c (build_alias_set): Pass the SCOP enclosing loop as loop-nest to dr_may_alias_p. * gcc.dg/torture/pr90328.c: New testcase. 2019-03-08 Richard Biener <rguenther@suse.de> PR middle-end/89578 * cfgloop.h (struct loop): Add owned_clique field. * cfgloopmanip.c (copy_loop_info): Copy it. * tree-cfg.c (gimple_duplicate_bb): Do not remap owned_clique cliques. * tree-inline.c (copy_loops): Remap owned_clique. * lto-streamer-in.c (input_cfg): Stream owned_clique. * lto-streamer-out.c (output_cfg): Likewise. 2019-02-22 Richard Biener <rguenther@suse.de> PR tree-optimization/87609 * tree-cfg.c (gimple_duplicate_bb): Only remap inlined cliques. 2019-02-22 Richard Biener <rguenther@suse.de> PR middle-end/87609 * cfghooks.h (dependence_hash): New typedef. (struct copy_bb_data): New type. (cfg_hooks::duplicate_block): Adjust to take a copy_bb_data argument. (duplicate_block): Likewise. * cfghooks.c (duplicate_block): Pass down copy_bb_data. (copy_bbs): Create and pass down copy_bb_data. * cfgrtl.c (cfg_layout_duplicate_bb): Adjust. (rtl_duplicate_bb): Likewise. * tree-cfg.c (gimple_duplicate_bb): If the copy_bb_data arg is not NULL remap dependence info. * gcc.dg/torture/restrict-7.c: New testcase. 2019-02-22 Richard Biener <rguenther@suse.de> PR tree-optimization/87609 * tree-core.h (tree_base): Document special clique values. * tree-inline.c (remap_dependence_clique): Do not use the special clique value of one. (maybe_set_dependence_info): Use clique one. (clear_dependence_clique): New callback. (compute_dependence_clique): Clear clique one from all refs before assigning it (again). Added: branches/gcc-8-branch/gcc/testsuite/gcc.dg/torture/pr90328.c branches/gcc-8-branch/gcc/testsuite/gcc.dg/torture/restrict-7.c Modified: branches/gcc-8-branch/gcc/ChangeLog branches/gcc-8-branch/gcc/cfghooks.c branches/gcc-8-branch/gcc/cfghooks.h branches/gcc-8-branch/gcc/cfgloop.h branches/gcc-8-branch/gcc/cfgloopmanip.c branches/gcc-8-branch/gcc/cfgrtl.c branches/gcc-8-branch/gcc/graphite-scop-detection.c branches/gcc-8-branch/gcc/lto-streamer-in.c branches/gcc-8-branch/gcc/lto-streamer-out.c branches/gcc-8-branch/gcc/lto-streamer.h branches/gcc-8-branch/gcc/testsuite/ChangeLog branches/gcc-8-branch/gcc/tree-cfg.c branches/gcc-8-branch/gcc/tree-core.h branches/gcc-8-branch/gcc/tree-data-ref.c branches/gcc-8-branch/gcc/tree-data-ref.h branches/gcc-8-branch/gcc/tree-inline.c branches/gcc-8-branch/gcc/tree-ssa-structalias.c
Author: rguenth Date: Thu Sep 5 12:11:52 2019 New Revision: 275405 URL: https://gcc.gnu.org/viewcvs?rev=275405&root=gcc&view=rev Log: 2019-09-05 Richard Biener <rguenther@suse.de> * lto-streamer.h (LTO_minor_version): Bump. Backport from mainline 2019-05-06 Richard Biener <rguenther@suse.de> PR tree-optimization/90328 * tree-data-ref.h (dr_may_alias_p): Pass in the actual loop nest. * tree-data-ref.c (dr_may_alias_p): Check whether the clique is valid in the loop nest before using it. (initialize_data_dependence_relation): Adjust. * graphite-scop-detection.c (build_alias_set): Pass the SCOP enclosing loop as loop-nest to dr_may_alias_p. * gcc.dg/torture/pr90328.c: New testcase. 2019-03-08 Richard Biener <rguenther@suse.de> PR middle-end/89578 * cfgloop.h (struct loop): Add owned_clique field. * cfgloopmanip.c (copy_loop_info): Copy it. * tree-cfg.c (gimple_duplicate_bb): Do not remap owned_clique cliques. * tree-inline.c (copy_loops): Remap owned_clique. * lto-streamer-in.c (input_cfg): Stream owned_clique. * lto-streamer-out.c (output_cfg): Likewise. 2019-02-22 Richard Biener <rguenther@suse.de> PR tree-optimization/87609 * tree-cfg.c (gimple_duplicate_bb): Only remap inlined cliques. 2019-02-22 Richard Biener <rguenther@suse.de> PR middle-end/87609 * cfghooks.h (dependence_hash): New typedef. (struct copy_bb_data): New type. (cfg_hooks::duplicate_block): Adjust to take a copy_bb_data argument. (duplicate_block): Likewise. * cfghooks.c (duplicate_block): Pass down copy_bb_data. (copy_bbs): Create and pass down copy_bb_data. * cfgrtl.c (cfg_layout_duplicate_bb): Adjust. (rtl_duplicate_bb): Likewise. * tree-cfg.c (gimple_duplicate_bb): If the copy_bb_data arg is not NULL remap dependence info. * gcc.dg/torture/restrict-7.c: New testcase. 2019-02-22 Richard Biener <rguenther@suse.de> PR tree-optimization/87609 * tree-core.h (tree_base): Document special clique values. * tree-inline.c (remap_dependence_clique): Do not use the special clique value of one. (maybe_set_dependence_info): Use clique one. (clear_dependence_clique): New callback. (compute_dependence_clique): Clear clique one from all refs before assigning it (again). Added: branches/gcc-7-branch/gcc/testsuite/gcc.dg/torture/pr90328.c branches/gcc-7-branch/gcc/testsuite/gcc.dg/torture/restrict-7.c Modified: branches/gcc-7-branch/gcc/ChangeLog branches/gcc-7-branch/gcc/cfghooks.c branches/gcc-7-branch/gcc/cfghooks.h branches/gcc-7-branch/gcc/cfgloop.h branches/gcc-7-branch/gcc/cfgloopmanip.c branches/gcc-7-branch/gcc/cfgrtl.c branches/gcc-7-branch/gcc/graphite-scop-detection.c branches/gcc-7-branch/gcc/lto-streamer-in.c branches/gcc-7-branch/gcc/lto-streamer-out.c branches/gcc-7-branch/gcc/lto-streamer.h branches/gcc-7-branch/gcc/testsuite/ChangeLog branches/gcc-7-branch/gcc/tree-cfg.c branches/gcc-7-branch/gcc/tree-core.h branches/gcc-7-branch/gcc/tree-data-ref.c branches/gcc-7-branch/gcc/tree-data-ref.h branches/gcc-7-branch/gcc/tree-inline.c branches/gcc-7-branch/gcc/tree-ssa-structalias.c