godotengine
/
godot
1
0
Fork 0
Code Issues Pull Requests Packages Projects Releases Wiki Activity

Don't use -ffast-math or other unsafe math optimizations 

Godot supports many different compilers and for production releases we
have to support 3 currently: GCC8, Clang6, and MSVC2017. These compilers
all do slightly different things with -ffast-math and it is causing
issues now. See #24841, #24540, #10758, #10070. And probably other
complaints about physics differences between release and release_debug
builds.

I've done some performance comparisons on Linux x86_64. All tests are
ran 20 times.

Bunnymark: (higher is better)
(bunnies)    min    max  stdev average
fast-math   7332   7597    71     7432
this pr     7379   7779   108     7621 (102%)

FPBench (gdscript port http://fpbench.org/) (lower is better)
(ms)
fast-math  15441  16127   192    15764
this pr    15671  16855   326    16001  (99%)

Float_add (adding floats in a tight loop) (lower is better)
(sec)
fast-math   5.49   5.78  0.07     5.65
this pr     5.65   5.90  0.06     5.76  (98%)

Float_div (dividing floats in a tight loop) (lower is better)
(sec)
fast-math  11.70  12.36  0.18    11.99
this pr    11.92  12.32  0.12    12.12  (99%)

Float_mul (multiplying floats in a tight loop) (lower is better)
(sec)
fast-math  11.72  12.17  0.12    11.93
this pr    12.01  12.62  0.17    12.26  (97%)

I have also looked at FPS numbers for tps-demo, 3d platformer, 2d
platformer, and sponza and could not find any measurable difference.

I believe that given the issues and oft-reported (physics) glitches on
release builds I believe that the couple of percent of tight-loop
floating point performance regression is well worth it.

This fixes #24540 and fixes #24841

(cherry picked from commit e5b335d367)
This commit is contained in:
Hein-Pieter van Braam 2019-01-09 01:09:56 +01:00 committed by Rémi Verschelde
parent c6cc6c86e5
commit cada03eedb
7 changed files with 9 additions and 15 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

4
modules/etc/SCsub
View File

@ -36,7 +36,3 @@ env_etc.add_source_files(env.modules_sources, "*.cpp")
# upstream uses c++11
if (not env_etc.msvc):
env_etc.Append(CCFLAGS="-std=c++11")
# -ffast-math seems to be incompatible with ec2comp on recent versions of
# GCC and Clang
if '-ffast-math' in env_etc['CCFLAGS']:
env_etc['CCFLAGS'].remove('-ffast-math')

2
platform/android/detect.py
View File

@ -150,7 +150,7 @@ def configure(env):
if (env["target"].startswith("release")):
env.Append(LINKFLAGS=['-O2'])
env.Append(CPPFLAGS=['-O2', '-DNDEBUG', '-ffast-math', '-funsafe-math-optimizations', '-fomit-frame-pointer'])
env.Append(CPPFLAGS=['-O2', '-DNDEBUG', '-fomit-frame-pointer'])
if (can_vectorize):
env.Append(CPPFLAGS=['-ftree-vectorize'])
if (env["target"] == "release_debug"):

4
platform/haiku/detect.py
View File

@ -37,14 +37,14 @@ def configure(env):
## Build type
if (env["target"] == "release"):
env.Prepend(CCFLAGS=['-O3', '-ffast-math'])
env.Prepend(CCFLAGS=['-O3'])
if (env["debug_symbols"] == "yes"):
env.Prepend(CCFLAGS=['-g1'])
if (env["debug_symbols"] == "full"):
env.Prepend(CCFLAGS=['-g2'])
elif (env["target"] == "release_debug"):
env.Prepend(CCFLAGS=['-O2', '-ffast-math', '-DDEBUG_ENABLED'])
env.Prepend(CCFLAGS=['-O2', '-DDEBUG_ENABLED'])
if (env["debug_symbols"] == "yes"):
env.Prepend(CCFLAGS=['-g1'])
if (env["debug_symbols"] == "full"):

2
platform/iphone/detect.py
View File

@ -47,7 +47,7 @@ def configure(env):
if (env["target"].startswith("release")):
env.Append(CPPFLAGS=['-DNDEBUG', '-DNS_BLOCK_ASSERTIONS=1'])
env.Append(CPPFLAGS=['-O2', '-ftree-vectorize', '-fomit-frame-pointer', '-ffast-math', '-funsafe-math-optimizations'])
env.Append(CPPFLAGS=['-O2', '-ftree-vectorize', '-fomit-frame-pointer'])
env.Append(LINKFLAGS=['-O2'])
if env["target"] == "release_debug":

2
platform/osx/detect.py
View File

@ -39,7 +39,7 @@ def configure(env):
## Build type
if (env["target"] == "release"):
env.Prepend(CCFLAGS=['-O3', '-ffast-math', '-fomit-frame-pointer', '-ftree-vectorize', '-msse2'])
env.Prepend(CCFLAGS=['-O3', '-fomit-frame-pointer', '-ftree-vectorize', '-msse2'])
if (env["debug_symbols"] == "yes"):
env.Prepend(CCFLAGS=['-g1'])
if (env["debug_symbols"] == "full"):

4
platform/server/detect.py
View File

@ -39,10 +39,10 @@ def configure(env):
## Build type
if (env["target"] == "release"):
env.Append(CCFLAGS=['-O2', '-ffast-math', '-fomit-frame-pointer'])
env.Append(CCFLAGS=['-O2', '-fomit-frame-pointer'])
elif (env["target"] == "release_debug"):
env.Append(CCFLAGS=['-O2', '-ffast-math', '-DDEBUG_ENABLED'])
env.Append(CCFLAGS=['-O2', '-DDEBUG_ENABLED'])
elif (env["target"] == "debug"):
env.Append(CCFLAGS=['-g2', '-DDEBUG_ENABLED', '-DDEBUG_MEMORY_ENABLED'])

6
platform/x11/detect.py
View File

@ -70,16 +70,14 @@ def configure(env):
## Build type
if (env["target"] == "release"):
# -O3 -ffast-math is identical to -Ofast. We need to split it out so we can selectively disable
# -ffast-math in code for which it generates wrong results.
env.Prepend(CCFLAGS=['-O3', '-ffast-math'])
env.Prepend(CCFLAGS=['-O3'])
if (env["debug_symbols"] == "yes"):
env.Prepend(CCFLAGS=['-g1'])
if (env["debug_symbols"] == "full"):
env.Prepend(CCFLAGS=['-g2'])
elif (env["target"] == "release_debug"):
env.Prepend(CCFLAGS=['-O2', '-ffast-math', '-DDEBUG_ENABLED'])
env.Prepend(CCFLAGS=['-O2', '-DDEBUG_ENABLED'])
if (env["debug_symbols"] == "yes"):
env.Prepend(CCFLAGS=['-g1'])
if (env["debug_symbols"] == "full"):