Loading...

XML

Word

Printable

Type: Bug
Resolution: Fixed
Priority: P3
Fix Version/s: 9
Affects Version/s: 9
Component/s: hotspot
Labels:

Subcomponent:
compiler
Resolved In Build:
b04
CPU:

sparc

Issue	Fix Version	Assignee	Priority	Status	Resolution	Resolved In Build
JDK-8045242	8u25	Poonam Bajaj Parhar	P3	Resolved	Fixed	b01
JDK-8036029	8u20	Poonam Bajaj Parhar	P3	Closed	Fixed	b04
JDK-8037130	8u11	Poonam Bajaj Parhar	P3	Resolved	Fixed	b01
JDK-8036838	8u5	Poonam Bajaj Parhar	P3	Closed	Fixed	b12
JDK-8039589	emb-8u26	Poonam Bajaj Parhar	P3	Resolved	Fixed	b17
JDK-8037668	emb-8u6	Poonam Bajaj Parhar	P3	Closed	Fixed	b10
JDK-8036023	7u80	Poonam Bajaj Parhar	P3	Resolved	Fixed	b01
JDK-8060849	7u79	Poonam Bajaj Parhar	P3	Resolved	Fixed	b01
JDK-8057282	7u76	Poonam Bajaj Parhar	P3	Resolved	Fixed	b01
JDK-8037237	7u65	Poonam Bajaj Parhar	P3	Resolved	Fixed	b03
JDK-8040350	7u60	Poonam Bajaj Parhar	P3	Resolved	Fixed	b15
JDK-8036839	7u55	Poonam Bajaj Parhar	P3	Resolved	Fixed	b12
JDK-8036028	7u51	Poonam Bajaj Parhar	P3	Resolved	Fixed	b34

On SPARC with cbcond we can fail with:
# Internal Error (output.cpp:1576), pid=3601, tid=15
# guarantee((int)(blk_starts[i+1] - blk_starts[i]) >= (current_offset - blk_offset)) failed: shouldn't increase block size

The customer hit the bug with the 32bit version on a virtual machine running on T-series. We didn't identify that machine as niagara so we used the default loop alignment (16 bytes instead of 4). I cannot reproduce the crash on 64bit, but the problem exists there as well.

It very hard to write a unit test for the fix (at least I wasn't able to), since the code must specifically trigger block rotation in loop to make sure that the loop header starts with a compare.

backported by

JDK-8036023 Second phase of branch shortening doesn't account for loop alignment

Resolved

JDK-8036028 Second phase of branch shortening doesn't account for loop alignment

Resolved

JDK-8036839 Second phase of branch shortening doesn't account for loop alignment

Resolved

JDK-8037130 Second phase of branch shortening doesn't account for loop alignment

Resolved

JDK-8037237 Second phase of branch shortening doesn't account for loop alignment

Resolved

JDK-8039589 Second phase of branch shortening doesn't account for loop alignment

Resolved

JDK-8040350 Second phase of branch shortening doesn't account for loop alignment

Resolved

JDK-8045242 Second phase of branch shortening doesn't account for loop alignment

Resolved

JDK-8057282 Second phase of branch shortening doesn't account for loop alignment

Resolved

JDK-8060849 Second phase of branch shortening doesn't account for loop alignment

Resolved

JDK-8036029 Second phase of branch shortening doesn't account for loop alignment

Closed

JDK-8036838 Second phase of branch shortening doesn't account for loop alignment

Closed

JDK-8037668 Second phase of branch shortening doesn't account for loop alignment

Closed

relates to

JDK-8056124 Hotspot should use PICL interface to get cacheline size on SPARC

Resolved

JDK-7063629 use cbcond in C2 generated code on T4

Closed

(8 backported by, 2 relates to)

Assignee:: Igor Veresov

Reporter:: Igor Veresov

Votes:: 0 Vote for this issue

Watchers:: 6 Start watching this issue

Created:: 2014-02-18 16:18

Updated:: 2016-05-01 20:44

Resolved:: 2014-02-26 18:04

Details

Backports

Description

Attachments

Issue Links

Activity

People

Dates