[
https://issues.apache.org/jira/browse/BEAM-11105?focusedWorklogId=764584&page=com.atlassian.jira.plugin.system.issuetabpanels:worklog-tabpanel#worklog-764584
]
ASF GitHub Bot logged work on BEAM-11105:
-----------------------------------------
Author: ASF GitHub Bot
Created on: 30/Apr/22 01:44
Start Date: 30/Apr/22 01:44
Worklog Time Spent: 10m
Work Description: damccorm commented on code in PR #17475:
URL: https://github.com/apache/beam/pull/17475#discussion_r862276118
##########
sdks/go/pkg/beam/core/runtime/exec/fn_arity.tmpl:
##########
@@ -30,7 +30,7 @@ import (
func (n *invoker) initCall() {
switch fn := n.fn.Fn.(type) {
{{range $out := upto 5}}
-{{range $in := upto 9}}
+{{range $in := upto 10}}
Review Comment:
Yeah, that's a good call - there was not a clear benefit to doing this. I
tried the following benchmarks:
```
func BenchmarkInvokeWith9Args_Call9x4(b *testing.B) {
fn, _ := funcx.New(reflectx.MakeFunc9x4(func(a1 typex.EventTime, a2
bool, a3 string, a4 int, a5 bool, a6 int, a7 string, a8 bool, a9 int) (int,
bool, string, error) {
return a4 + a6 + a9, a2 && a5 && a8, a3 + a7, nil
}))
args := []interface{}{mtime.ZeroTimestamp.Add(2 * time.Millisecond),
true, "hello", 1, true, 3, "world", false, 4}
f := fn.Fn.(reflectx.Func9x4)
for n := 0; n < b.N; n++ {
f.Call9x4(args[0], args[1], args[2], args[3], args[4], args[5],
args[6], args[7], args[8])
}
}
func BenchmarkInvokeWith9Args_Call(b *testing.B) {
fn, _ := funcx.New(reflectx.MakeFunc(func(a1 typex.EventTime, a2 bool,
a3 string, a4 int, a5 bool, a6 int, a7 string, a8 bool, a9 int) (int, bool,
string, error) {
return a4 + a6 + a9, a2 && a5 && a8, a3 + a7, nil
}))
args := []interface{}{mtime.ZeroTimestamp.Add(2 * time.Millisecond),
true, "hello", 1, true, 3, "world", false, 4}
f := fn.Fn
for n := 0; n < b.N; n++ {
f.Call(args)
}
}
```
with the result:
```
BenchmarkInvokeWith9Args_Call9x4-16 722283 1577 ns/op
BenchmarkInvokeWith9Args_Call-16 805590 1447 ns/op
```
So the normal call is actually more performant. It is probably worth
revisiting this more systematically for lower arities as well, but I think that
should fall outside the scope of this PR. For the moment, I've lowered it back
down to 9
Issue Time Tracking
-------------------
Worklog Id: (was: 764584)
Time Spent: 10h 10m (was: 10h)
> [Go SDK] Watermark Estimation
> -----------------------------
>
> Key: BEAM-11105
> URL: https://issues.apache.org/jira/browse/BEAM-11105
> Project: Beam
> Issue Type: Sub-task
> Components: sdk-go
> Reporter: Robert Burke
> Assignee: Danny McCormick
> Priority: P3
> Time Spent: 10h 10m
> Remaining Estimate: 0h
>
> Allow self checkpointing SplittableDoFns to specify a watermark estimator.
> (To be updated once [https://github.com/apache/beam/pull/13160] is merged and
> the programming guide updated with SDF content.)
--
This message was sent by Atlassian Jira
(v8.20.7#820007)
