https://gcc.gnu.org/bugzilla/show_bug.cgi?id=123207
Bug ID: 123207
Summary: [16 Regression] Miscompare of 526.blender_r with
-Ofast on x86_64
Product: gcc
Version: 16.0
Status: UNCONFIRMED
Keywords: needs-bisection
Severity: normal
Priority: P3
Component: tree-optimization
Assignee: unassigned at gcc dot gnu.org
Reporter: pheeck at gcc dot gnu.org
Blocks: 26163
Target Milestone: ---
Host: x86_64-linux
Target: x86_64-linux
I've seen 526.blender_r 2017 SPEC benchmark result in a miscompare. For
example when built with -Ofast -march=x86-64-v3 -flto and run on an AMD Zen5
machine:
****************************************
Contents of imagevalidate_sh3_no_char_0849.err
****************************************
ImageValidator r4696
INFO: threshold set to 0.750000
INFO: max threshold set to 0 absolute, 0.010000 relative
Blocks below threshold: 268 blocks of 144 allowed (14400 total)
AVG SSIM: 0.944458805
MIN SSIM: 0.009791775
SSIM distribution:
[ 1.000]: 2784
[ 0.900, 1.000): 8417
[ 0.800, 0.900): 2624
[ 0.700, 0.800): 442
[ 0.600, 0.700): 97
[ 0.500, 0.600): 19
[ 0.400, 0.500): 3
[ 0.300, 0.400): 6
[ 0.200, 0.300): 3
[ 0.100, 0.200): 2
[ 0.000, 0.100): 3
[-0.100, 0.000): 0
[-0.200, -0.100): 0
[-0.300, -0.200): 0
[-0.400, -0.300): 0
[-0.500, -0.400): 0
[-0.600, -0.500): 0
[-0.700, -0.600): 0
[-0.800, -0.700): 0
[-0.900, -0.800): 0
[-1.000, -0.900): 0
****************************************
*** Miscompare of imagevalidate_sh3_no_char_0849.out; for details see
/home/gcc/buildworker/source/cpu2017/benchspec/CPU/526.blender_r/run/run_peak_refrate_gcc-m64.0000/imagevalidate_sh3_no_char_0849.out.mis
0001: sh3_no_char_0849.tga: AVG SSIM = 1.000000000
sh3_no_char_0849.tga ERROR:
^
'imagevalidate_sh3_no_char_0849.out' long
The miscompare also happens:
- with -Ofast -march=native on an AMD Zen3 machine
- with -Ofast -march=native on an Intel Ice Lake (3rd generation Xeon)
- with -Ofast -march=native -flto on an AMD Zen2 machine
- with -Ofast -march=native -flto on an AMD Zen4 machine
- with -Ofast -march=native -flto on an AMD Zen5 machine
Referenced Bugs:
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=26163
[Bug 26163] [meta-bug] missed optimization in SPEC (2k17, 2k and 2k6 and 95)
