8253113: [lworld] [lw3] C1 should avoid copying element of flattened arrays when reading a sub-element #187

fparain · 2020-09-14T13:55:17Z

Please review these changes optimizing access to flattened arrays in C1.

The optimization avoids unnecessary copies of intermediate values when the code accesses a sub-element of a flattened array.

For instance, with the following code:

inline class Point { int x = 0, y = 0; }
void foo() {
Point[] array = new Point[10];
int i = array[1].x;
}

C1 used to create a new instance of Point, copy the content from the array, then reads the x field from this new instance.
With the optimization, C1 directly accesses the x field from the flattened array, and makes no heap allocation.

A simple benchmark on the "int i = array[1].x;" line gives these results:

baseline:
Benchmark Mode Samples Score Score error Units
o.s.MyBenchmark.testArrayReads avgt 200 4.250 0.011 ns/op

optimized:
Benchmark Mode Samples Score Score error Units
o.s.MyBenchmark.testArrayReads avgt 200 2.228 0.004 ns/op

Thank you,

Fred

Progress

Change must not contain extraneous whitespace

Issue

JDK-8253113: [lworld] [lw3] C1 should avoid copying element of flattened arrays when reading a sub-element

Reviewers

Tobias Hartmann (@TobiHartmann - Committer)

Download

$ git fetch https://git.openjdk.java.net/valhalla pull/187/head:pull/187
$ git checkout pull/187

bridgekeeper · 2020-09-14T13:57:34Z

👋 Welcome back fparain! A progress list of the required criteria for merging this PR into lworld will be added to the body of your pull request. There are additional pull request commands available for use with this pull request.

openjdk · 2020-09-14T13:58:44Z

@fparain This change now passes all automated pre-integration checks. In addition to the automated checks, the change must also fulfill all project specific requirements

After integration, the commit message will be:

8253113: [lworld] [lw3] C1 should avoid copying element of flattened arrays when reading a sub-element

Reviewed-by: thartmann

If you would like to add a summary, use the /summary command.
To credit additional contributors, use the /contributor command.
To add additional solved issues, use the /issue command.

Since the source branch of this PR was last updated there has been 1 commit pushed to the lworld branch:

07e2fc5: 8253161: [lworld] C1's substitutability check should use andptr instead of andl for mark word

As there are no conflicts, your changes will automatically be rebased on top of these commits when integrating. If you prefer to avoid automatic rebasing, please merge lworld into your branch, and then specify the current head hash when integrating, like this: /integrate 07e2fc53c1407f6e34e2cb147f5d61190b4227d0.

➡️ To integrate this PR with the above commit message to the lworld branch, type /integrate in a new comment.

mlbridge · 2020-09-14T14:01:49Z

Webrevs

TobiHartmann

Looks good to me, added some minor comments.

TobiHartmann · 2020-09-15T07:07:04Z

src/hotspot/share/c1/c1_LIRGenerator.cpp

  // Find the starting address of the source (inside the array)
  ciType* array_type = array.value()->declared_type();
  ciFlatArrayKlass* flat_array_klass = array_type->as_flat_array_klass();
  assert(flat_array_klass->is_loaded(), "must be");

-  ciInlineKlass* elem_klass = flat_array_klass->element_klass()->as_inline_klass();
+  ciInlineKlass* elem_klass = NULL;


TobiHartmann · 2020-09-15T07:07:58Z

src/hotspot/share/c1/c1_LIRGenerator.cpp

+
+  BasicType subelt_type = field->type()->basic_type();
+
+  #ifndef _LP64


Wrong indentation.

TobiHartmann · 2020-09-15T07:14:22Z

src/hotspot/share/c1/c1_LIRGenerator.cpp

@@ -1595,13 +1595,84 @@ class TempResolvedAddress: public Instruction {
  virtual const char* name() const  { return "TempResolvedAddress"; }
 };

-void LIRGenerator::access_flattened_array(bool is_load, LIRItem& array, LIRItem& index, LIRItem& obj_item) {
+void LIRGenerator::access_sub_element(LIRItem& array, LIRItem& index, LIR_Opr& result, ciField* field, int sub_offset) {


There is quite some code duplication here with access_flattened_array, can we factor it out?

Code has been refactored, and the common code in access_flattened_array() and access_sub_element() is now shared in the new method get_and_load_element_address().

TobiHartmann

Looks good, thanks for making these changes!

TobiHartmann

Nice refactoring. Looks good to me.

TobiHartmann · 2020-09-15T15:26:42Z

src/hotspot/share/c1/c1_LIRGenerator.cpp

+void LIRGenerator::access_sub_element(LIRItem& array, LIRItem& index, LIR_Opr& result, ciField* field, int sub_offset) {
+  assert(field != NULL, "Need a subelement type specified");
+
+// Find the starting address of the source (inside the array)


Indentation is wrong.

TobiHartmann · 2020-09-15T15:27:03Z

src/hotspot/share/c1/c1_LIRGenerator.cpp

+                                          ciField* field, int sub_offset) {
+  assert(sub_offset == 0 || field != NULL, "Sanity check");
+
+// Find the starting address of the source (inside the array)


Indentation is wrong.

fparain · 2020-09-15T15:32:25Z

Thanks Tobias for your review.
I've fixed the wrong indentations.

Fred

fparain · 2020-09-15T15:34:08Z

/integrate

openjdk · 2020-09-15T15:36:20Z

@fparain Since your change was applied there has been 1 commit pushed to the lworld branch:

07e2fc5: 8253161: [lworld] C1's substitutability check should use andptr instead of andl for mark word

Your commit was automatically rebased without conflicts.

Pushed as commit 1a52191.

💡 You may see a message that your pull request was closed with unmerged commits. This can be safely ignored.

fparain added 7 commits September 11, 2020 09:33

First implementation

ee34315

Remove debugging code

c87c830

Extend CFG format with information related to sub-element accesses

804a348

Fix white space

75c2b18

Fix flattened sub-element access, add unit tests

ae99725

Renaming and cleanup

02da68d

Remove debug message in unit test

c67a81a

openjdk bot added ready rfr labels Sep 14, 2020

TobiHartmann approved these changes Sep 15, 2020

View reviewed changes

Fixes from review

c748044

TobiHartmann approved these changes Sep 15, 2020

View reviewed changes

More refactoring

fcde384

TobiHartmann approved these changes Sep 15, 2020

View reviewed changes

Fix indentation

8d97e01

TobiHartmann approved these changes Sep 15, 2020

View reviewed changes

openjdk bot closed this Sep 15, 2020

openjdk bot added integrated and removed ready rfr labels Sep 15, 2020

fparain deleted the c1_aaload_opt branch September 28, 2020 14:48

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

8253113: [lworld] [lw3] C1 should avoid copying element of flattened arrays when reading a sub-element #187

8253113: [lworld] [lw3] C1 should avoid copying element of flattened arrays when reading a sub-element #187

fparain commented Sep 14, 2020 •

edited by openjdk bot

Loading

bridgekeeper bot commented Sep 14, 2020

openjdk bot commented Sep 14, 2020 •

edited

Loading

mlbridge bot commented Sep 14, 2020 •

edited

Loading

TobiHartmann left a comment

TobiHartmann Sep 15, 2020

fparain Sep 15, 2020

TobiHartmann Sep 15, 2020

fparain Sep 15, 2020

TobiHartmann Sep 15, 2020

fparain Sep 15, 2020

TobiHartmann left a comment

TobiHartmann left a comment

TobiHartmann Sep 15, 2020

TobiHartmann Sep 15, 2020

fparain commented Sep 15, 2020

fparain commented Sep 15, 2020

openjdk bot commented Sep 15, 2020


		BasicType subelt_type = field->type()->basic_type();

		#ifndef _LP64

8253113: [lworld] [lw3] C1 should avoid copying element of flattened arrays when reading a sub-element #187

8253113: [lworld] [lw3] C1 should avoid copying element of flattened arrays when reading a sub-element #187

Conversation

fparain commented Sep 14, 2020 • edited by openjdk bot Loading

Progress

Issue

Reviewers

Download

bridgekeeper bot commented Sep 14, 2020

openjdk bot commented Sep 14, 2020 • edited Loading

mlbridge bot commented Sep 14, 2020 • edited Loading

Webrevs

TobiHartmann left a comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

TobiHartmann left a comment

Choose a reason for hiding this comment

TobiHartmann left a comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

fparain commented Sep 15, 2020

fparain commented Sep 15, 2020

openjdk bot commented Sep 15, 2020

fparain commented Sep 14, 2020 •

edited by openjdk bot

Loading

openjdk bot commented Sep 14, 2020 •

edited

Loading

mlbridge bot commented Sep 14, 2020 •

edited

Loading