Bug 30179 - Generic fmod and remainder are slower than x87 implementations
Summary: Generic fmod and remainder are slower than x87 implementations
Status: NEW
Alias: None
Product: glibc
Classification: Unclassified
Component: math (show other bugs)
Version: 2.38
: P2 normal
Target Milestone: ---
Assignee: H.J. Lu
URL:
Keywords:
Depends on:
Blocks:
 
Reported: 2023-02-27 16:35 UTC by H.J. Lu
Modified: 2023-02-27 16:35 UTC (History)
0 users

See Also:
Host:
Target: x86-64
Build:
Last reconfirmed:

Attachments
A testcase (747 bytes, application/octet-stream)
2023-02-27 16:35 UTC, H.J. Lu 		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 H.J. Lu 2023-02-27 16:35:12 UTC 
Created attachment 14722 [details]
A testcase

Performance of generic fmod and remainder highly depends on input values. Performance
of x87 (fprem/fprem1) implementations can be much faster.  On Intel Coffee Lake,

[hjl@gnu-cfl-3 fmod-2]$ make
gcc -O2   -c -o test.o test.c
gcc    -c -o x87.o x87.S
gcc -static -o x87 test.o x87.o
gcc -static -o sse test.o -lm
time ./sse
3.15user 0.00system 0:03.15elapsed 99%CPU (0avgtext+0avgdata 684maxresident)k
0inputs+0outputs (0major+39minor)pagefaults 0swaps
time ./x87
0.25user 0.00system 0:00.25elapsed 99%CPU (0avgtext+0avgdata 680maxresident)k
0inputs+0outputs (0major+37minor)pagefaults 0swaps
[hjl@gnu-cfl-3 fmod-2]$