Bug 59611 - std::copy performs worse than naive implementation when copying a single known byte
Summary: std::copy performs worse than naive implementation when copying a single know...
Status: RESOLVED FIXED
Alias: None
Product: gcc
Classification: Unclassified
Component: tree-optimization (show other bugs)
Version: 4.9.0
: P3 normal
Target Milestone: ---
Assignee: Richard Biener
URL:
Keywords: missed-optimization
Depends on: 61473
Blocks:
  Show dependency tree / graph
 
Reported: 2013-12-27 16:55 UTC by Eelis
Modified: 2014-07-14 11:50 UTC (History)
2 users (show)

See Also:
Host:
Target:
Build:
Known to work:
Known to fail:
Last reconfirmed: 2013-12-27 00:00:00

Attachments
Testcase (240 bytes, text/x-c++src)
2013-12-27 16:55 UTC, Eelis 		Details
View All Add an attachment (proposed patch, testcase, etc.)

Note You need to log in before you can comment on or make changes to this bug.
Description Eelis 2013-12-27 16:55:37 UTC 
Created attachment 31524 [details]
Testcase

The attached testcase shows how using a naive implementation (mycopy) causes gcc to generate better code than using std::copy.

(Clang /does/ generate optimal code when using std::copy in this case.)
Comment 1 Eelis 2013-12-27 16:57:47 UTC 
(With -O3)
Comment 2 Marc Glisse 2013-12-27 22:29:42 UTC 
The call to memcpy is folded rather early to:
  MEM[(char * {ref-all})p_2(D)] = MEM[(char * {ref-all})&a];

and then we don't touch it anymore, maybe because we don't go through an SSA_NAME. We thus miss that the RHS is a constant.
Comment 3 Richard Biener 2014-06-26 14:49:03 UTC 
Mine.
Comment 4 Richard Biener 2014-06-27 10:40:37 UTC 
Fixed by the patch for PR61473.
Comment 5 Richard Biener 2014-07-14 11:50:14 UTC 
Verified that the 61473 fix fixed this as well.