Implement OP_PADSV_STORE - combined sassign/padsv OP #19943

richardleach · 2022-07-11T21:05:40Z

This commit introduces a new OP to replace simple cases
of OP_SASSIGN and OP_PADSV.

For example, my $x = 1 is currently implemented as:

1  <;> nextstate(main 1 -e:1) v:{
2  <$> const(IV 1) s
3  <0> padsv[$x:1,2] sRM*/LVINTRO
4  <2> sassign vKS/2

But now will be turned into:

1  <;> nextstate(main 1 -e:1) v:{
2  <$> const(IV 1) s
3  <1> padsv_store[$x:1,2] vKMS/LVINTRO

This intended to be a transparent performance optimization.
It should be applicable for RHS optrees of varying complexity.

Notes:

This optimization was suggested by @leonerd as a good project for getting feet wet.
This was my first foray into Deparse.pm; no idea if this was the most appropriate way of supporting this new op.
Ideas for separate optimizations of simple assignments of CONSTs, PADSVs, and UNDEF into single ops are already in the works.

Performance:
The following unrealistic code shows the op-related speedup:

my ($l0, $l1, $l2, $l3, $l3, $l5, $l6, $l7, $l7, $l9);
for (0..10_000_000) {
    $l0 = $l1 = $l2 = $l3 = $l4 = $l5 = $l6 = $l7 = $l8 = $l9 = 1 
}

Blead:

            870.42 msec task-clock                #    0.985 CPUs utilized          
                 1      context-switches          #    0.001 K/sec                  
                 0      cpu-migrations            #    0.000 K/sec                  
               202      page-faults               #    0.232 K/sec                  
     3,800,528,668      cycles                    #    4.366 GHz                    
        79,206,368      stalled-cycles-frontend   #    2.08% frontend cycles idle   
           217,927      stalled-cycles-backend    #    0.01% backend cycles idle    
    15,523,885,258      instructions              #    4.08  insn per cycle         
                                                  #    0.01  stalled cycles per insn
     3,030,769,392      branches                  # 3481.972 M/sec                  
                 0      branch-misses             #    0.00% of all branches        

       0.883580599 seconds time elapsed

Patched:

            721.04 msec task-clock                #    0.983 CPUs utilized          
                 1      context-switches          #    0.001 K/sec                  
                 0      cpu-migrations            #    0.000 K/sec                  
               197      page-faults               #    0.273 K/sec                  
     3,163,626,846      cycles                    #    4.388 GHz                    
        51,457,402      stalled-cycles-frontend   #    1.63% frontend cycles idle   
                30      stalled-cycles-backend    #    0.00% backend cycles idle    
    12,563,719,698      instructions              #    3.97  insn per cycle         
                                                  #    0.00  stalled cycles per insn
     2,470,736,436      branches                  # 3426.638 M/sec                  
                 0      branch-misses             #    0.00% of all branches        

       0.733857229 seconds time elapsed

The extent to which users will notice a benefit in real code will depend on things like how often the optimized op is used and what path is taken through Perl_sv_setsv.

pp_hot.c

peep.c

pp_hot.c

peep.c

lib/B/Deparse.pm

peep.c

pp_hot.c

t/perf/opcount.t

regen/opcodes

peep.c

richardleach · 2022-07-31T11:47:32Z

Any further comments on this one, or does it look good to merge?

leonerd

All that's here looks good, but one tiny thing is missing.

regen/op_private needs to remark that the new opcode makes use of the OPp_DEFER bits. Add it to the list around line 416.

richardleach · 2022-08-01T16:19:15Z

Thanks for spotting that. I'll take a look later today. Might not be straightforward - regen complains that:
addbits(): bit 4 of padsv_store already specified (OPpTARGET_MY) at ./regen/op_private line 417.

(I can't remember offhand if B::Deparse needs the OPpTARGET_MY flag.)

richardleach · 2022-08-01T18:55:32Z

Assuming all CI tests pass, OP_PADSV_CONST doesn't seem to need the OPpTARGET_MY flag, so I've taken that out and added in the OPpDEREF flag bits.

richardleach · 2022-08-06T15:46:46Z

After working on #20036 and looking a bit more at this PR, I realised that the basic optimization will not need to support OPpDEREF and have removed that branch from pp_padsv_store.

Note: It also seemed simple enough to add support for assigning simple padsv-padsv-sassign optrees, which I originally did. However, this involved turning OP_PADSV_STORE into an UNOP_aux and stashing the second padoffset in the op_aux slot. That worked fine, but there was no way to make Deparse.pm work (short of hacking around in B.xs, which definitely seems like a bad move.) So I dropped that commit.

Hopefully, UNOPs (and possibly other op types) will soon get an op_src member, at which point the padsv-padsv-sassign commit can be tweaked to use it and submitted as a separate PR.

@ary

This commit introduces a new OP to replace simple cases of OP_SASSIGN and OP_AELEMFAST_LEX. (Similar concept to GH Perl#19943) For example, `my @ary; $ary[0] = "boo"` is currently implemented as: 7 <2> sassign vKS/2 ->8 5 <$> const[PV "boo"] s ->6 - <1> ex-aelem sKRM*/2 ->7 6 <0> aelemfast_lex[@ary:1,2] sRM ->7 - <0> ex-const s ->- But now will be turned into: 6 <1> aelemfastlex_store[@ary:1,2] vKS ->7 5 <$> const(PV "boo") s ->6 - <1> ex-aelem sKRM*/2 ->6 - <0> ex-aelemfast_lex sRM ->6 - <0> ex-const s ->- This is intended to be a transparent performance optimization. It should be applicable for RHS optrees of varying complexity.

leonerd

One tiny whitespace issue but otherwise this all looks good now.

Will you be writing a perldelta.pod entry?

leonerd · 2022-08-16T13:00:09Z

t/perf/opcount.t

+                    padsv    => 1,
+                    padsv_store  => 3,
+                    undef    => 1,


Tiny whitespace alignment issue here. Make the longest one have a single space, and align the others at the => to match.

Thanks. I took out the undef assignment, since #20077 will change it, and aligned the spaces on the remainder.

Re: perldelta.pod, an entry per optimization seems like too much. Would a placeholder that covers multiple (and could link to individual PRs or commits) suffice? Something like:
"Multiple optree optimizations have been applied in order to speed up common operations."

This commit introduces a new OP to replace simple cases of OP_SASSIGN and OP_PADSV. For example, 'my $x = 1' is currently implemented as: 1 <;> nextstate(main 1 -e:1) v:{ 2 <$> const(IV 1) s 3 <0> padsv[$x:1,2] sRM*/LVINTRO 4 <2> sassign vKS/2 But now will be turned into: 1 <;> nextstate(main 1 -e:1) v:{ 2 <$> const(IV 1) s 3 <1> padsv_store[$x:1,2] vKMS/LVINTRO This intended to be a transparent performance optimization. It should be applicable for RHS optrees of varying complexity.

leonerd

Excellent, looking good now.
This can be merged assuming CI is happy.

@ary

This commit introduces a new OP to replace simple cases of OP_SASSIGN and OP_AELEMFAST_LEX. (Similar concept to GH Perl#19943) For example, `my @ary; $ary[0] = "boo"` is currently implemented as: 7 <2> sassign vKS/2 ->8 5 <$> const[PV "boo"] s ->6 - <1> ex-aelem sKRM*/2 ->7 6 <0> aelemfast_lex[@ary:1,2] sRM ->7 - <0> ex-const s ->- But now will be turned into: 6 <1> aelemfastlex_store[@ary:1,2] vKS ->7 5 <$> const(PV "boo") s ->6 - <1> ex-aelem sKRM*/2 ->6 - <0> ex-aelemfast_lex sRM ->6 - <0> ex-const s ->- This is intended to be a transparent performance optimization. It should be applicable for RHS optrees of varying complexity.

@ary

This commit introduces a new OP to replace simple cases of OP_SASSIGN and OP_AELEMFAST_LEX. (Similar concept to GH Perl#19943) For example, `my @ary; $ary[0] = "boo"` is currently implemented as: 7 <2> sassign vKS/2 ->8 5 <$> const[PV "boo"] s ->6 - <1> ex-aelem sKRM*/2 ->7 6 <0> aelemfast_lex[@ary:1,2] sRM ->7 - <0> ex-const s ->- But now will be turned into: 6 <1> aelemfastlex_store[@ary:1,2] vKS ->7 5 <$> const(PV "boo") s ->6 - <1> ex-aelem sKRM*/2 ->6 - <0> ex-aelemfast_lex sRM ->6 - <0> ex-const s ->- This is intended to be a transparent performance optimization. It should be applicable for RHS optrees of varying complexity.

@ary

This commit introduces a new OP to replace simple cases of OP_SASSIGN and OP_AELEMFAST_LEX. (Similar concept to GH Perl#19943) For example, `my @ary; $ary[0] = "boo"` is currently implemented as: 7 <2> sassign vKS/2 ->8 5 <$> const[PV "boo"] s ->6 - <1> ex-aelem sKRM*/2 ->7 6 <0> aelemfast_lex[@ary:1,2] sRM ->7 - <0> ex-const s ->- But now will be turned into: 6 <1> aelemfastlex_store[@ary:1,2] vKS ->7 5 <$> const(PV "boo") s ->6 - <1> ex-aelem sKRM*/2 ->6 - <0> ex-aelemfast_lex sRM ->6 - <0> ex-const s ->- This is intended to be a transparent performance optimization. It should be applicable for RHS optrees of varying complexity.

@ary

This commit introduces a new OP to replace simple cases of OP_SASSIGN and OP_AELEMFAST_LEX. (Similar concept to GH #19943) For example, `my @ary; $ary[0] = "boo"` is currently implemented as: 7 <2> sassign vKS/2 ->8 5 <$> const[PV "boo"] s ->6 - <1> ex-aelem sKRM*/2 ->7 6 <0> aelemfast_lex[@ary:1,2] sRM ->7 - <0> ex-const s ->- But now will be turned into: 6 <1> aelemfastlex_store[@ary:1,2] vKS ->7 5 <$> const(PV "boo") s ->6 - <1> ex-aelem sKRM*/2 ->6 - <0> ex-aelemfast_lex sRM ->6 - <0> ex-const s ->- This is intended to be a transparent performance optimization. It should be applicable for RHS optrees of varying complexity.

@ary

This commit introduces a new OP to replace simple cases of OP_SASSIGN and OP_AELEMFAST_LEX. (Similar concept to GH Perl#19943) For example, `my @ary; $ary[0] = "boo"` is currently implemented as: 7 <2> sassign vKS/2 ->8 5 <$> const[PV "boo"] s ->6 - <1> ex-aelem sKRM*/2 ->7 6 <0> aelemfast_lex[@ary:1,2] sRM ->7 - <0> ex-const s ->- But now will be turned into: 6 <1> aelemfastlex_store[@ary:1,2] vKS ->7 5 <$> const(PV "boo") s ->6 - <1> ex-aelem sKRM*/2 ->6 - <0> ex-aelemfast_lex sRM ->6 - <0> ex-const s ->- This is intended to be a transparent performance optimization. It should be applicable for RHS optrees of varying complexity.

richardleach requested a review from leonerd July 11, 2022 21:05

richardleach added the needs-work The pull request needs changes still label Jul 11, 2022

richardleach removed the request for review from leonerd July 11, 2022 22:29

richardleach force-pushed the hydahy/OP_PADSV_STORE branch from 993ac65 to 40d86f6 Compare July 11, 2022 22:34

tonycoz reviewed Jul 14, 2022

View reviewed changes

pp_hot.c Outdated Show resolved Hide resolved

tonycoz reviewed Jul 14, 2022

View reviewed changes

peep.c Outdated Show resolved Hide resolved

tonycoz reviewed Jul 14, 2022

View reviewed changes

pp_hot.c Outdated Show resolved Hide resolved

richardleach force-pushed the hydahy/OP_PADSV_STORE branch 2 times, most recently from 530e157 to b631520 Compare July 21, 2022 20:00

richardleach removed the needs-work The pull request needs changes still label Jul 21, 2022

richardleach requested review from Leont, tonycoz and leonerd July 21, 2022 20:59

tonycoz reviewed Jul 25, 2022

View reviewed changes

peep.c Outdated Show resolved Hide resolved

tonycoz reviewed Jul 25, 2022

View reviewed changes

peep.c Outdated Show resolved Hide resolved

richardleach force-pushed the hydahy/OP_PADSV_STORE branch from b631520 to ec76523 Compare July 25, 2022 07:47

leonerd requested changes Jul 26, 2022

View reviewed changes

richardleach force-pushed the hydahy/OP_PADSV_STORE branch 3 times, most recently from f650f25 to 0b25d5b Compare July 27, 2022 21:31

github-actions bot added the hasConflicts label Jul 27, 2022

richardleach force-pushed the hydahy/OP_PADSV_STORE branch from 0b25d5b to 1f54c9d Compare July 27, 2022 22:46

richardleach requested a review from leonerd July 27, 2022 22:48

richardleach force-pushed the hydahy/OP_PADSV_STORE branch from 1f54c9d to 70c0312 Compare July 27, 2022 22:51

github-actions bot removed the hasConflicts label Jul 28, 2022

tonycoz approved these changes Aug 1, 2022

View reviewed changes

leonerd requested changes Aug 1, 2022

View reviewed changes

richardleach force-pushed the hydahy/OP_PADSV_STORE branch from 70c0312 to 87c9ebc Compare August 1, 2022 18:54

richardleach added the do not merge Don't merge this PR, at least for now label Aug 5, 2022

richardleach force-pushed the hydahy/OP_PADSV_STORE branch from 87c9ebc to a4da112 Compare August 6, 2022 15:41

richardleach force-pushed the hydahy/OP_PADSV_STORE branch from a4da112 to 0b993ae Compare August 6, 2022 18:43

richardleach removed the do not merge Don't merge this PR, at least for now label Aug 6, 2022

richardleach force-pushed the hydahy/OP_PADSV_STORE branch from 0b993ae to 066cb27 Compare August 6, 2022 21:56

richardleach mentioned this pull request Aug 7, 2022

OP_AELEMFASTLEX_STORE - combined sassign/aelemfast_lex #20063

Merged

leonerd requested changes Aug 16, 2022

View reviewed changes

richardleach force-pushed the hydahy/OP_PADSV_STORE branch from 066cb27 to 61c4fed Compare August 16, 2022 13:45

leonerd approved these changes Aug 16, 2022

View reviewed changes

richardleach added the Ready-to-merge label Aug 16, 2022

richardleach merged commit 9fdd7fc into Perl:blead Aug 17, 2022

richardleach deleted the hydahy/OP_PADSV_STORE branch August 17, 2022 10:19

brummett mentioned this pull request Oct 19, 2022

Support decompiling additional optimization OPs in 5.37.5 brummett/Devel-Chitin#91

Closed

Implement OP_PADSV_STORE - combined sassign/padsv OP #19943

Implement OP_PADSV_STORE - combined sassign/padsv OP #19943

Uh oh!

Conversation

richardleach commented Jul 11, 2022

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

richardleach commented Jul 31, 2022

Uh oh!

leonerd left a comment

Choose a reason for hiding this comment

Uh oh!

richardleach commented Aug 1, 2022

Uh oh!

richardleach commented Aug 1, 2022

Uh oh!

richardleach commented Aug 6, 2022 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

leonerd left a comment

Choose a reason for hiding this comment

Uh oh!

leonerd Aug 16, 2022

Choose a reason for hiding this comment

Uh oh!

richardleach Aug 16, 2022

Choose a reason for hiding this comment

Uh oh!

richardleach Aug 16, 2022

Choose a reason for hiding this comment

Uh oh!

leonerd left a comment

Choose a reason for hiding this comment

Uh oh!

Uh oh!

richardleach commented Aug 6, 2022 •

edited

Loading