|
|
| General • Portage • Wiki |
Please do not include any flags that aren't 100% safe. Just because it works for you doesn't mean it will for everyone else.
We aren't interested in benchmark scores, personal anecdotes, rumors, things your really smart uncle once told you, random -m flags that made your box dispense root beer, or any other discussion or debate. It will be removed. Take it to the talk page or the appropriate forum. Thanks.[edit] Safe CFLAGS to use for Gentoo
These CFLAGS are for gcc 3.x and 4.x. To check your version:
gcc --version
You can see which CPU you have with the following command:
cat /proc/cpuinfo
If you're running Intel, you can also use x86info to get more detailed info about your chip:
emerge -av x86info
This page is for those who don't want to experiment, and want a stable system, but still optimized for their CPU. But remember, by using these flags, binaries from your system might not work on another one with a different CPU. So if you compiled with optimizations for a Pentium 4 (-march=pentium4), you can't share a hard disk or packages with a friend who only has a Pentium MMX (-march=pentium-mmx). Though the other way around should work, see below.
This is the case if you are using the -march flags. If you use -mtune instead without any -march option, your binaries are backward compatible down to i386, but the scheduling is optimized for your choosen architecture. CPUs are also backward compatible so if you update your system with a new CPU you can still use your old packages. There is one exception though; if your old CPU supports a feature, like AMD's 3dnow, that your new CPU doesn't. So if you have an athlon (-march=athlon) and upgrade to a pentium4 your old packages might not work. You can mix -march and -mtune if you, for example, would like to create binaries that would run on any i686 but are optimized for a pentium4. This would be accomplished by using -march=i686 and -mtune=pentium4.
Note that -O2 is regarded as safer than "-O3", and "-O3" can often be a counter-productive attempt at optimization. On computers with limited cache and/or memory, "-Os" may provide better performance in some cases through smaller binaries, although it is slower when using the OpenSSL library with small keys (DSA keys with less than 2048 bits on VIA C3-2, 1200 MHz and 64 kb on-die cache).
Please note that all the CFLAGS below look like they have a lowercase O. It really is an uppercase O. If you enter a lowercase O in your CFLAGS, then you will break your Portage and you won't be able to install anything. If you see the following error in your emerge output, than it could point to this problem.
checking whether the C compiler works... configure: error: cannot run C compiled programs.
GCC has a number of flags for CPU features like MMX and SSE; -mmx -msse etc. They are implied if you use a correct -march option so you usually don´t need them, with some exceptions. They should not be confused with the similar USE-flags for CPU features. The flag -mfpmath=sse however is not enabled by -march but it usually makes binaries slower due to limitations in the glibc headers. So it's better to not use this flag, even if you have an SSE capable CPU. Also -mfpmath=sse,387 is experimental and unstable.
The flag -fomit-frame-pointer is enabled at -O1, -O2, -O3 and -Os on arches where it doesn't interfere with debugging, such as AMD64, but not x86. So if you're on x86 you should add it to your CFLAGS. Also don't use -momit-leaf-frame-pointer together with -fomit-frame-pointer. It's pointless as -fomit-frame-pointer already omits all frame pointers. In fact if you use both, -momit-leaf-frame-pointer overrides -fomit-frame-pointer creating less optimized code.
If you are, or plan on becoming a programmer/developer, then DO NOT use the -fomit-frame-pointer or -O3 flag as this can make debugging impossible, or at least a lot harder (especially on x86 machines).
This is also important if you find a program with a bug in, which you wish to report. Before you submit a bug report, compile with just "-O2 -march=i686 -pipe" and without "-fomit-frame-pointer" (see http://bugs.gentoo.org/show_bug.cgi?id=68282 for an explanation.) Basically the frame-pointer is needed for stack traces.
[edit] gcc-4.2.0 or newer
If you are using gcc-4.2.0 or newer it is easiest and probably best to just use:
-march=native or -mtune=native.
E.g.:
32-bit
CHOST="i686-pc-linux-gnu"
CFLAGS="-march=native -O2 -pipe"
CXXFLAGS="${CFLAGS}"
64-bit
CHOST="x86_64-pc-linux-gnu"
CFLAGS="-march=native -O2 -pipe"
CXXFLAGS="${CFLAGS}"
[edit] i386
If you would like to use a newer machine (eg. a Pentium III) to compile code that will still run on i386 machines use the following:
CHOST="i386-pc-linux-gnu"
CFLAGS="-mtune=pentium3 -O2 -pipe -fomit-frame-pointer"
CXXFLAGS="${CFLAGS}"
or
CHOST="i386-pc-linux-gnu"
CFLAGS="-march=i386 -O2 -pipe -fomit-frame-pointer"
CXXFLAGS="${CFLAGS}"
The former will create code optimized for a PIII that will still run on a i386, while the latter will be optimized for the i386 alone.
[edit] i486
CHOST="i486-pc-linux-gnu"
CFLAGS="-march=i486 -O2 -pipe -fomit-frame-pointer"
CXXFLAGS="${CFLAGS}"
[edit] Intel Processors
[edit] Pentium
vendor_id : GenuineIntel
cpu family : 5
model : 2
model name : Pentium 75 - 200
CHOST="i586-pc-linux-gnu"
CFLAGS="-march=pentium -O2 -pipe -fomit-frame-pointer"
CXXFLAGS="${CFLAGS}"
[edit] Pentium w/ MMX
CHOST="i586-pc-linux-gnu"
CFLAGS="-march=pentium-mmx -O2 -pipe -fomit-frame-pointer"
CXXFLAGS="${CFLAGS}"
[edit] Cyrix MediaGX / AMD Geode GX1
CFLAGS="-march=pentium-mmx -O2 -pipe -fomit-frame-pointer"
CXXFLAGS="${CFLAGS}"
[edit] Pentium Pro
CHOST="i686-pc-linux-gnu"
CFLAGS="-march=pentiumpro -O2 -pipe -fomit-frame-pointer"
CXXFLAGS="${CFLAGS}"
or
CHOST="i686-pc-linux-gnu"
CFLAGS="-march=i686 -O2 -pipe -fomit-frame-pointer"
CXXFLAGS="${CFLAGS}"
(these are identical)
[edit] Pentium II
vendor_id : GenuineIntel
cpu family : 6
model : 3, 5
CHOST="i686-pc-linux-gnu"
CFLAGS="-march=pentium2 -O2 -pipe -fomit-frame-pointer"
CXXFLAGS="${CFLAGS}"
processor : 0 vendor_id : GenuineIntel cpu family : 6 model : 6 model name : Mobile Pentium II stepping : 10 cpu MHz : 397.082 cache size : 256 KB fdiv_bug : no hlt_bug : no f00f_bug : no coma_bug : no fpu : yes fpu_exception : yes cpuid level : 2 wp : yes flags : fpu vme de pse tsc msr pae mce cx8 sep mtrr pge mca cmov pat pse36 mmx fxsr bogomips : 794.75 clflush size : 32
[edit] Celeron (Mendocino), aka Celeron1
vendor_id : GenuineIntel
cpu family : 6
model : 6
stepping : 0
CHOST="i686-pc-linux-gnu"
CFLAGS="-march=pentium2 -O2 -pipe -fomit-frame-pointer"
CXXFLAGS="${CFLAGS}"
processor : 0 vendor_id : GenuineIntel cpu family : 6 model : 6 model name : Celeron (Mendocino) stepping : 0 cpu MHz : 334.108 cache size : 128 KB fdiv_bug : no hlt_bug : no f00f_bug : no coma_bug : no fpu : yes fpu_exception : yes cpuid level : 2 wp : yes flags : fpu vme de pse tsc msr pae mce cx8 sep mtrr pge mca cmov pat pse36 mmx fxsr up bogomips : 686.94 clflush size : 32
[edit] Pentium III
vendor_id : GenuineIntel
cpu family : 6
model : 7, 8
CHOST="i686-pc-linux-gnu"
CFLAGS="-march=pentium3 -O2 -pipe -fomit-frame-pointer"
CXXFLAGS="${CFLAGS}"
processor : 0 vendor_id : GenuineIntel cpu family : 6 model : 8 model name : Pentium III (Coppermine) stepping : 6 cpu MHz : 863.934 cache size : 256 KB fdiv_bug : no hlt_bug : no f00f_bug : no coma_bug : no fpu : yes fpu_exception : yes cpuid level : 2 wp : yes flags : fpu vme de pse tsc msr pae mce cx8 apic sep mtrr pge mca cmov pat pse36 mmx fxsr sse bogomips : 1728.97 clflush size : 32
[edit] Celeron (Coppermine)
vendor_id : GenuineIntel
cpu family : 6
model : 8
stepping : 10
CHOST="i686-pc-linux-gnu"
CFLAGS="-march=pentium3 -O2 -pipe -fomit-frame-pointer"
CXXFLAGS="${CFLAGS}"
[edit] Celeron (Willamette)
vendor_id : GenuineIntel
cpu family : 15
model : 1
model name : Intel(R) Celeron(R) CPU 1.70GHz
stepping : 3
CHOST="i686-pc-linux-gnu"
CFLAGS="-march=pentium4 -O2 -pipe -fomit-frame-pointer"
CXXFLAGS="${CFLAGS}"
[edit] Pentium M(Centrino)/Celeron M
vendor_id : GenuineIntel
cpu family : 6
model : 9 or 13
model name : Intel(R) Pentium(R) M processor XXXXMHz
For gcc 3.3 or older (Warning: dev-java/sun-jdk needs old libstdc++-3.3.so on X86 Systems) :
CHOST="i686-pc-linux-gnu"
CFLAGS="-march=pentium3 -msse2 -O2 -pipe -fomit-frame-pointer"
CXXFLAGS="${CFLAGS}"
For gcc 3.4 and later:
CHOST="i686-pc-linux-gnu"
CFLAGS="-march=pentium-m -O2 -pipe -fomit-frame-pointer"
CXXFLAGS="${CFLAGS}"
The Celeron M is based on the Pentium M but it has half the L2 cache and does not support the SpeedStep technology.
[edit] Mobile Pentium 4-M (Northwood)
vendor_id : GenuineIntel
cpu family : 15
model : 2
model name : Mobile Intel(R) Pentium(R) 4 - M CPU X.XXGHz
CHOST="i686-pc-linux-gnu"
CFLAGS="-march=pentium4 -O2 -pipe -fomit-frame-pointer"
CXXFLAGS="${CFLAGS}"
[edit] Pentium 4
vendor_id : GenuineIntel
cpu family : 15
model : 0 or 1 or 2
model name : Intel(R) Pentium(R) 4 CPU XXXXMHz
CHOST="i686-pc-linux-gnu"
CFLAGS="-march=pentium4 -O2 -pipe -fomit-frame-pointer"
CXXFLAGS="${CFLAGS}"
[edit] Pentium 4 (Prescott) / Celeron D
vendor_id : GenuineIntel
cpu family : 15
model : 3 or 4
model name : Intel(R) Pentium(R) 4 CPU XXXGHz -or- Intel(R) Celeron(R) CPU XXXGHz
32-bit profile (x86)
CHOST="i686-pc-linux-gnu"
CFLAGS="-march=prescott -O2 -pipe -fomit-frame-pointer"
CXXFLAGS="${CFLAGS}"
64-bit profile (amd64)
CHOST="x86_64-pc-linux-gnu"
CFLAGS="-march=nocona -O2 -pipe -fomit-frame-pointer"
CXXFLAGS="${CFLAGS}"
In 2004, Intel started branding processors with the Prescott core as Intel Celeron D.
[edit] Pentium D 8xx / 9xx
vendor_id : GenuineIntel
cpu family : 15
model : 4, 6
model name : Intel(R) Pentium(R) D CPU x.xxGHz
32-bit profile (x86)
CHOST="i686-pc-linux-gnu"
CFLAGS="-march=prescott -O2 -pipe -fomit-frame-pointer"
CXXFLAGS="${CFLAGS}"
64-bit profile (amd64)
CHOST="x86_64-pc-linux-gnu"
CFLAGS="-march=nocona -O2 -pipe -fomit-frame-pointer"
CXXFLAGS="${CFLAGS}"
[edit] Xeon w/o EM64T
vendor_id : GenuineIntel
cpu family : 15
model : 2
model name : Intel(R) Xeon(R) CPU XXXXMHz
CHOST="i686-pc-linux-gnu"
CFLAGS="-march=pentium4 -O2 -pipe -fomit-frame-pointer"
CXXFLAGS="${CFLAGS}"
[edit] Xeon w/EM64T (also Pentium 4 P6xx)
vendor_id : GenuineIntel
cpu family : 15
model : 4, 6
model name : Intel(R) Xeon(R) CPU XXXXMHz
32-bit profile (x86)
CHOST="i686-pc-linux-gnu"
CFLAGS="-march=prescott -O2 -pipe -fomit-frame-pointer"
CXXFLAGS="${CFLAGS}"
64-bit profile (amd64)
CHOST="x86_64-pc-linux-gnu"
CFLAGS="-march=nocona -O2 -pipe -fomit-frame-pointer"
CXXFLAGS="${CFLAGS}"
[edit] Intel Core Solo/Duo, Pentium Dual-Core Txxxx
vendor_id : GenuineIntel
cpu family : 6
model : 14
model name : Genuine Intel(R) CPU TXXXX @ XXXGHz
CHOST="i686-pc-linux-gnu"
CFLAGS="-march=prescott -O2 -pipe -fomit-frame-pointer"
CXXFLAGS="${CFLAGS}"
Notes:
- It has been confirmed by hongjiu.lu@intel.com that prescott is the correct microarchitecture to use with this CPU. http://article.gmane.org/gmane.comp.gcc.devel/83870
- The newer (eg. 420) Celeron-M processors are Core Solo based, not Pentium-M based. If your processor's family and model are 6 and 14, then you should use the -march=prescott option.
[edit] Intel Core 2 Duo/Quad / Xeon 51xx/53xx, Pentium Dual-Core Exxxx
vendor_id : GenuineIntel
cpu family : 6
model : 15
model name : Intel(R) Core(TM)2 CPU XXXX @ XXXGHz
32 bit profile (x86)
CHOST="i686-pc-linux-gnu"
CFLAGS="-march=prescott -O2 -pipe -fomit-frame-pointer"
CXXFLAGS="${CFLAGS}"
64 bit profile (amd64)
CHOST="x86_64-pc-linux-gnu"
CFLAGS="-march=nocona -O2 -pipe"
CXXFLAGS="${CFLAGS}"
Note:
- GCC 4.3 has -march=core2[1], but Gentoo has nothing above 4.2 in portage.
[edit] VIA Processors
[edit] Esther C5J (Via C7)
vendor_id : CentaurHauls
cpu family : 6
model : 10
model name : VIA Esther processor 2000MHz
CHOST="i686-pc-linux-gnu"
CFLAGS="-march=i686 -mmmx -msse -msse2 -msse3 -O2 -pipe -fomit-frame-pointer"
CXXFLAGS="${CFLAGS}"
As the C7 has the pni flag in /proc/cpuinfo, it supports SSE3. Search for pni on this page to get the explanation.
[edit] C3 Nehemiah (C5X/C5XL/C5P) (VIA)
vendor_id : CentaurHauls
cpu family : 6
model : 9
model name : VIA Nehemiah
GCC 3.3 and earlier
CHOST="i686-pc-linux-gnu"
CFLAGS="-march=i686 -msse -mmmx -O2 -pipe -fomit-frame-pointer"
CXXFLAGS="${CFLAGS}"
GCC 3.4 and later
CHOST="i686-pc-linux-gnu"
CFLAGS="-march=c3-2 -O2 -pipe -fomit-frame-pointer"
CXXFLAGS="${CFLAGS}"
[edit] C3 Samuel/Ezra (Via EPIA)
CHOST="i586-pc-linux-gnu"
CFLAGS="-march=c3 -m3dnow -O2 -pipe -fomit-frame-pointer"
CXXFLAGS="${CFLAGS}"
The Ezra doesn't have any special instructions that you could optimize for, just consider it a K6-3 - basically a Pentium 2 with 3DNow.
[edit] Transmeta Processors
[edit] Transmeta Crusoe
vendor_id : GenuineTMx86
cpu family : 6
model : 4
model name : Transmeta(tm) Crusoe(tm) Processor TM5800
stepping : 3
CHOST="i686-pc-linux-gnu"
CFLAGS="-march=i686 -Os -mmmx -pipe -fomit-frame-pointer"
CXXFLAGS="${CFLAGS}"
[edit] Transmeta Efficeon
vendor_id : GenuineTMx86
cpu family : 15
model : 2
model name : Transmeta Efficeon(tm) Processor TM8000
stepping : 4
CHOST="i686-pc-linux-gnu"
CFLAGS="-mtune=pentium3 -msse2 -O2 -pipe -falign-functions=0 -falign-jumps=0 -falign-loops=0"
CXXFLAGS="${CFLAGS}"
Note: The Efficeon processor will reorder and realign instructions on translating from x86 to VLIW (Very Long Instruction Word), so not aligning functions/jumps/loops will produce smaller executable without any effect on speed.
[edit] AMD Processors
[edit] K6 (AMD)
CHOST="i586-pc-linux-gnu"
CFLAGS="-march=k6 -O2 -pipe -fomit-frame-pointer"
CXXFLAGS="${CFLAGS}"
[edit] K6-2 (AMD)
vendor_id : AuthenticAMD
cpu family : 5
model : 8
CHOST="i586-pc-linux-gnu"
CFLAGS="-march=k6-2 -O2 -pipe -fomit-frame-pointer"
CXXFLAGS="${CFLAGS}"
# cat /proc/cpuinfo processor : 0 vendor_id : AuthenticAMD cpu family : 5 model : 8 model name : AMD-K6(tm) 3D processor stepping : 12 cpu MHz : 451.031 cache size : 64 KB fdiv_bug : no hlt_bug : no f00f_bug : no coma_bug : no fpu : yes fpu_exception : yes cpuid level : 1 wp : yes flags : fpu vme de pse tsc msr mce cx8 pge mmx syscall 3dnow k6_mtrr up bogomips : 902.84 clflush size : 32
Note: /proc/cpuinfo may be confusing. The model name may say "AMD-K6(tm) 3D Processor" but it is not a AMD K6-3. Rely on the vendor_id, CPU family, and model number.
Note: If you get "Illegal Instruction" errors on compiles (especially long ones), you may be running into a cooling problem, or worse yet, a Sig11 (segmentation fault/bad ram) problem. Bug #24379 has a discussion about it.
[edit] Geode LX (AMD)
vendor_id : AuthenticAMD
cpu family : 5
model : 10
CHOST="i586-pc-linux-gnu"
CFLAGS="-march=k6-2 -Os -pipe -fomit-frame-pointer"
CXXFLAGS="${CFLAGS}"
[edit] K6-3 (AMD)
CHOST="i586-pc-linux-gnu"
CFLAGS="-march=k6-3 -O2 -pipe -fomit-frame-pointer"
CXXFLAGS="${CFLAGS}"
[edit] Duron (AMD)
From 600 to 900 MHz (these models have a tbird-alike core model)
product: AMD Duron(tm) Processor version: 6.3.1
CHOST="i686-pc-linux-gnu"
CFLAGS="-march=athlon-tbird -O2 -pipe -fomit-frame-pointer"
CXXFLAGS="${CFLAGS}"
[edit] Mobile Duron (AMD)
vendor_id : AuthenticAMD
cpu family : 6
model : 3
model name : AMD Duron(tm) Processor
CHOST="i686-pc-linux-gnu"
CFLAGS="-march=athlon-tbird -O2 -pipe -fomit-frame-pointer"
CXXFLAGS="${CFLAGS}"
[edit] Duron Morgan (AMD)
From 900 to 1300 MHz
The Morgan Durons are based on the Palomino core, and hence can be treated as Athlon XP's.
vendor_id : AuthenticAMD
cpu family : 6
model : 7
model name : AMD Duron(tm) Processor
stepping : 1
CHOST="i686-pc-linux-gnu"
CFLAGS="-march=athlon-xp -O2 -pipe -fomit-frame-pointer"
CXXFLAGS="${CFLAGS}"
[edit] Athlon (AMD)
vendor_id : AuthenticAMD
cpu family : 6
model : 4 or 2 or 1
model name : AMD Athlon(TM)Processor or AMD-K7(TM)Processor
stepping : 4 or 2
note: This processor is only available as a Slot A processor; if you have a Socket A/462 Athlon, it's a Thunderbird (see below)
CHOST="i686-pc-linux-gnu"
CFLAGS="-march=athlon -O2 -pipe -fomit-frame-pointer"
CXXFLAGS="${CFLAGS}"
[edit] Athlon-tbird (AMD)
vendor_id : AuthenticAMD
cpu family : 6
model : 4
model name : AMD Athlon(tm) Processor
CHOST="i686-pc-linux-gnu"
CFLAGS="-march=athlon-tbird -O2 -pipe -fomit-frame-pointer"
CXXFLAGS="${CFLAGS}"
[edit] Athlon-tbird XP/Duron (AMD)
vendor_id : AuthenticAMD
cpu family : 6
model : 8
model name : AMD Athlon(tm)
CHOST="i686-pc-linux-gnu"
CFLAGS="-march=athlon-xp -O2 -pipe -fomit-frame-pointer"
CXXFLAGS="${CFLAGS}"
Sempron 2400+ (2000MHz) is also identified as cpu family:6 model:8 model_name:AMD Sempron(TM) 2400+ (256kB cache) The Duron 1800+ is also included in this list.
[edit] Athlon 4 (AMD)
CHOST="i686-pc-linux-gnu"
CFLAGS="-march=athlon-4 -O2 -pipe -fomit-frame-pointer"
CXXFLAGS="${CFLAGS}"
[edit] Athlon XP/Geode NX (AMD)
vendor_id : AuthenticAMD
cpu family : 6
model : 6
stepping : 2
Some CPUs have also been observed with the following:
model : 10
model name : AMD Athlon(tm) XP 2500+
stepping : 0
CHOST="i686-pc-linux-gnu"
CFLAGS="-march=athlon-xp -O2 -pipe -fomit-frame-pointer"
CXXFLAGS="${CFLAGS}"
# cat /proc/cpuinfo processor : 0 vendor_id : AuthenticAMD cpu family : 6 model : 6 model name : AMD Athlon(tm) XP 2000+ stepping : 2 cpu MHz : 1670.597 cache size : 256 KB fdiv_bug : no hlt_bug : no f00f_bug : no coma_bug : no fpu : yes fpu_exception : yes cpuid level : 1 wp : yes flags : fpu vme de pse tsc msr pae mce cx8 apic sep mtrr pge mca cmov pat pse36 mmx fxsr sse syscall mmxext 3dnowext 3dnow up ts bogomips : 3343.54
# cat /proc/cpuinfo processor : 0 vendor_id : AuthenticAMD cpu family : 6 model : 8 model name : AMD Geode NX 1750 stepping : 1 cpu MHz : 1400.115 cache size : 256 KB fdiv_bug : no hlt_bug : no f00f_bug : no coma_bug : no fpu : yes fpu_exception : yes cpuid level : 1 wp : yes flags : fpu vme de pse tsc msr pae mce cx8 apic sep mtrr pge mca cmov pat pse36 mmx fxsr sse syscall mp mmxext 3dnowext 3dnow up ts fid vid bogomips : 2801.55 clflush size : 32
[edit] Athlon MP (AMD)
vendor_id : AuthenticAMD
cpu family : 6
model : 10
model name : AMD Sempron(tm)
stepping : 0
CHOST="i686-pc-linux-gnu"
CFLAGS="-march=athlon-mp -O2 -pipe -fomit-frame-pointer"
CXXFLAGS="${CFLAGS}"
[edit] Athlon 64 (AMD)
32 bit
CHOST="i686-pc-linux-gnu"
CFLAGS="-march=athlon -O2 -pipe -fomit-frame-pointer"
CXXFLAGS="${CFLAGS}"
64 bit
cpu family :15
model :4
model name :AMD Athlon(tm) 64 Processor 3200+
stepping :8
CHOST="x86_64-pc-linux-gnu"
CFLAGS="-march=k8 -O2 -pipe"
CXXFLAGS="${CFLAGS}"
# cat /proc/cpuinfo processor : 0 vendor_id : AuthenticAMD cpu family : 15 model : 63 model name : AMD Athlon(tm) 64 Processor 3200+ stepping : 2 cpu MHz : 1999.821 cache size : 512 KB fpu : yes fpu_exception : yes cpuid level : 1 wp : yes flags : fpu vme de pse tsc msr pae mce cx8 apic sep mtrr pge mca cmov pat pse36 clflush mmx fxsr sse sse2 syscall nx mmxext fxsr_opt lm 3dnowext 3dnow pni lahf_lm bogomips : 4001.79 TLB size : 1024 4K pages clflush size : 64 cache_alignment : 64 address sizes : 40 bits physical, 48 bits virtual power management: ts fid vid ttp
[edit] Athlon 64 X2 (AMD)
vendor_id: AuthenticAMD
cpu family: 15
model: 35, 43, or 75
stepping: 1 or 2
model name: AMD Athlon(tm) 64 X2 Dual Core Processor
32 bit
CHOST="i686-pc-linux-gnu"
CFLAGS="-march=athlon -O2 -pipe -fomit-frame-pointer"
CXXFLAGS="${CFLAGS}"
64 bit
CHOST="x86_64-pc-linux-gnu"
CFLAGS="-march=athlon64 -O2 -pipe"
CXXFLAGS="${CFLAGS}"
If you have a newer Athlon64("Venice" or "San Diego")/Athlon64-X2("Manchester", "Toledo", or "Windsor") (check for pni in cat /proc/cpuinfo) you can also add -msse3 to your CFLAGS to enable SSE3 support. Any chip using the 90nm process except for the "Winchester" class Athlon64 supports SSE3.
Also bear in mind that all Athlon64 X2 and Opteron 165, 170, 175, 180, and 185 processors are dual-core CPUs so make sure that you use MAKEOPTS="-j3" in your make.conf. This doesn't affect the code but might reduce compile times.
All socket AM2 processors are SSE3 capable as well.
[edit] 1xx Opteron (AMD)
vendor_id : AuthenticAMD
cpu family : 15
model : 5
model name : AMD Opteron(tm) Processor 1xx
stepping : 8
CHOST="x86_64-pc-linux-gnu"
CFLAGS="-march=opteron -O2 -pipe"
CXXFLAGS="${CFLAGS}"
socket 939 # cat /proc/cpuinfo processor : 0 vendor_id : AuthenticAMD cpu family : 15 model : 39 model name : AMD Opteron(tm) Processor 144 stepping : 1 cpu MHz : 1800.000 cache size : 1024 KB fpu : yes fpu_exception : yes cpuid level : 1 wp : yes flags : fpu vme de pse tsc msr pae mce cx8 apic sep mtrr pge mca cmov pat pse36 clflush mmx fxsr sse sse2 syscall nx mmxext fxsr_opt lm 3dnowext 3dnow up pni lahf_lm bogomips : 3593.50 TLB size : 1024 4K pages clflush size : 64 cache_alignment : 64 address sizes : 40 bits physical, 48 bits virtual power management: ts fid vid ttp
[edit] 2xx/8xx Opteron (AMD)
vendor_id : AuthenticAMD
cpu family : 15
model : 5
model name : AMD Opteron(tm) Processor xxx
stepping : 8
CHOST="x86_64-pc-linux-gnu"
CFLAGS="-march=opteron -O2 -pipe"
CXXFLAGS="${CFLAGS}"
[edit] 22xx Dual-Core Opteron (AMD)
vendor_id : AuthenticAMD
cpu family : 15
model : 65
model name : Dual-Core AMD Opteron(tm) Processor 22xx
stepping : 2
CHOST="x86_64-pc-linux-gnu"
CFLAGS="-march=opteron -O2 -pipe"
CXXFLAGS="${CFLAGS}"
# cat /proc/cpuinfo processor : 0 vendor_id : AuthenticAMD cpu family : 15 model : 65 model name : Dual-Core AMD Opteron(tm) Processor 2212 HE stepping : 3 cpu MHz : 2010.298 cache size : 1024 KB physical id : 0 siblings : 1 core id : 0 cpu cores : 1 fpu : yes fpu_exception : yes cpuid level : 1 wp : yes flags : fpu tsc msr pae mce cx8 apic mca cmov pat pse36 clflush mmx fxsr sse sse2 ht syscall nx mmxext fxsr_opt lm 3dnowext 3dnow pni cx16 lahf_lm cmp_legacy svm cr8_legacy bogomips : 4022.18 TLB size : 1024 4K pages clflush size : 64 cache_alignment : 64 address sizes : 40 bits physical, 48 bits virtual power management: ts fid vid ttp tm stc
[edit] 23xx Quad-Core Barcelona (AMD)
vendor_id : AuthenticAMD
cpu family : 16
model : 2
model name : Quad-Core AMD Opteron(tm) Processor 23xx
stepping : 10
CHOST="x86_64-pc-linux-gnu"
CFLAGS="-march=k8 -O2 -pipe"
CXXFLAGS="${CFLAGS}"
[edit] Sempron/Sempron64 (AMD)
- If it is a Socket A (Socket 462) Sempron you can be sure it is neither 64-bit capable nor SSE2/SSE3 capable.
- If it is a Socket AM2 Sempron you can be sure it is both 64-bit and SSE3 capable.
- If it is a Socket 754 it depends on the stepping (part number):
- Steppings: CG (Part No.: *AX) or D0 (Part No.: *BA) have neither SSE3 nor AMD64
- Stepping: E3 (Part No.: *BO) has SSE3
- Stepping: E6 (Part No.: *BX) has SSE3 and AMD64
- Anyway you should check /proc/cpuinfo for pni (SSE3) string before using -msse3.
vendor_id : AuthenticAMD
cpu family : 15
model : 44
model name : AMD Sempron(tm) Processor
Note: The model number may differ.
32 bit (Socket A)
CHOST="i686-pc-linux-gnu"
CFLAGS="-march=athlon-xp -O2 -pipe -fomit-frame-pointer"
CXXFLAGS="${CFLAGS}"
64 bit
CHOST="x86_64-pc-linux-gnu"
CFLAGS="-march=k8 -O2 -pipe"
CXXFLAGS="${CFLAGS}"
64 bit (SSE3 capable)
CHOST="x86_64-pc-linux-gnu"
CFLAGS="-march=k8 -msse3 -O2 -pipe"
CXXFLAGS="${CFLAGS}"
Sempron64 - 32 bit
CHOST="i686-pc-linux-gnu"
CFLAGS="-march=k8 -msse3 -O2 -pipe -fomit-frame-pointer"
CXXFLAGS="${CFLAGS}"
Sempron64 - 64 bit
CHOST="x86_64-pc-linux-gnu"
CFLAGS="-march=k8 -msse3 -O2 -pipe"
CXXFLAGS="${CFLAGS}"
[edit] Turion64 (AMD)
32 bit
CHOST="i686-pc-linux-gnu"
CFLAGS="-march=k8 -msse3 -O2 -pipe -fomit-frame-pointer"
CXXFLAGS="${CFLAGS}"
64 bit
CHOST="x86_64-pc-linux-gnu"
CFLAGS="-march=k8 -msse3 -O2 -pipe"
CXXFLAGS="${CFLAGS}"
SSE3 is called pni in /proc/cpuinfo for this Central Processing Unit.
[edit] Mobile Sempron (AMD)
vendor_id : AuthenticAMD
cpu family : 15
model : 8
model name : Mobile AMD Sempron Processor 2800+
CHOST="i686-pc-linux-gnu"
CFLAGS="-march=athlon-xp -msse3 -O2 -pipe -fomit-frame-pointer"
CXXFLAGS="${CFLAGS}"
If you have a Sempron without SSE3 (earlier chips), omit "-msse3".
SSE3 is called pni (prescott new instuction) by cpuinfo (cat /proc/cpuinfo).
Check it out, there are many different sempron mobile CPUs on sale, those that work at 1.8GHz are K8 (i.e. AMD SM 3000+), some others, older, aren't yet k8 but they're still k7 so you must specify athlon-xp instead of athlon64, anyway AMD announced in 2006 there will be 64bit Sempron Mobile CPUs so be carefull and pay attention, run a cat /proc/cpuinfo and then ask on AMD forums if you don't know what to do.
[edit] PowerPC Processors
[edit] PowerPC 601
The PowerPC 601 CPU is a mixture between POWER and PowerPC architectures and it is a must you specify -mcpu=601 for taking advantage of the POWER part of the processor and do not use unimplemented PowerPC instructions (implementation of both ISAs is incomplete on 601).
CHOST="powerpc-unknown-linux-gnu"
CFLAGS="-mcpu=601 -O2 -pipe"
CXXFLAGS="${CFLAGS}"
NOTE: Code created with this flags will not run in any other CPU but PowerPC 601, if you want to run on 601 and others use -mcpu=common and code will run in ANY POWER or PowerPC CPU.
[edit] PowerPC 603
CHOST="powerpc-unknown-linux-gnu"
CFLAGS="-mcpu=603 -O2 -pipe"
CXXFLAGS="${CFLAGS}"
[edit] PowerPC 603e
CHOST="powerpc-unknown-linux-gnu"
CFLAGS="-mcpu=603e -O2 -pipe"
CXXFLAGS="${CFLAGS}"
[edit] PowerPC 604
CHOST="powerpc-unknown-linux-gnu"
CFLAGS="-mcpu=604 -O2 -pipe"
CXXFLAGS="${CFLAGS}"
[edit] PowerPC 604e
CHOST="powerpc-unknown-linux-gnu"
CFLAGS="-mcpu=604e -O2 -pipe"
CXXFLAGS="${CFLAGS}"
-Os may be beneficial on older processors (of any kind, not just PowerPCs).
[edit] PowerPC 604e
CHOST="powerpc-unknown-linux-gnu"
CFLAGS="-mcpu=604e -O2 -pipe -fno-strict-aliasing"
CXXFLAGS="${CFLAGS}"
[edit] PowerPC 750 (G3)
CHOST="powerpc-unknown-linux-gnu"
CFLAGS="-mcpu=750 -Os -pipe -fno-strict-aliasing"
CXXFLAGS="${CFLAGS}"
Given the smaller cache sizes, and the fact that the L2 cache is off-die on the G3, better performance may be achieved with -Os rather than -O2.
[edit] PowerPC 750cx (G3e)
CHOST="powerpc-unknown-linux-gnu"
CFLAGS="-mcpu=750 -O2 -pipe -fno-strict-aliasing"
CXXFLAGS="${CFLAGS}"
The 750cx offers more L2 cache than its predecessor, and moves it on-die for a performance boost. Better performance may be achieved with -O2.
[edit] PowerPC 7400 (G4)
CHOST="powerpc-unknown-linux-gnu"
CFLAGS="-mcpu=7400 -O2 -pipe -maltivec -mabi=altivec"
CXXFLAGS="${CFLAGS}"
Note: -O3 is unstable on G4
[edit] PowerPC 7450 (G4 second generation)
CHOST="powerpc-unknown-linux-gnu"
CFLAGS="-mcpu=7450 -O2 -pipe -maltivec -mabi=altivec"
CXXFLAGS="${CFLAGS}"
[edit] PowerPC 970 (G5)
CHOST="powerpc64-unknown-linux-gnu"
CFLAGS="-mcpu=G5 -O2 -pipe -maltivec -mabi=altivec -fno-strict-aliasing"
CXXFLAGS="${CFLAGS}"
LDFLAGS="-Wl,-O1"
[edit] Cell Broadband Engine (Playstation 3)
Don't use -mcpu=970 it will make the PS3 slower due to the different pipeline (after all it is a CELL B.E and not a G5). GCC 4.3 or SDK 2.1+ is recommended. A second hint: Unlike x86-64 architecture, it is better to use 32bit user land on a PowerPC with a 64bit kernel. The benefit is, that pointers are 32bits so memory usage goes down slightly. In fact unlike x86_64, the number of registers don't change and the register size does not change (though the Linux kernel has issues with interrupts) and the kernel will save all 64bits of the register across context switches.
CHOST="powerpc64-unknown-linux-gnu"
CFLAGS="-O2 -pipe -mcpu=cell -mabi=altivec"
CXXFLAGS="${CFLAGS}"
LDFLAGS=""
[edit] PowerPC (If you don't know which one)
CHOST="powerpc-unknown-linux-gnu"
CFLAGS="-O2 -pipe"
CXXFLAGS="${CFLAGS}"
[edit] Embedded PowerPC CPUs
You should check /proc/cpuinfo for the CPU model and then for a correct -mcpu flag on man gcc. If you do not find it, use powerpc.
CHOST="powerpc-unknown-linux-gnu"
CFLAGS="-mcpu=<flag> -Os -pipe"
CXXFLAGS="${CFLAGS}"
[edit] Other Processors
[edit] Sparc
CHOST="sparc-unknown-linux-gnu"
CFLAGS="-O2 -pipe"
CXXFLAGS="${CFLAGS}"
[edit] Sparc 64
CHOST="sparc-unknown-linux-gnu"
CFLAGS="-mcpu=ultrasparc -mtune=ultrasparc -O2 -pipe"
CXXFLAGS="${CFLAGS}"
Be careful! The CHOST for Sparc64 is still sparc-unknown-linux-gnu, not sparc64-!
[edit] HPPA 1.1
CHOST="hppa1.1-unknown-linux-gnu"
CFLAGS="-O2 -pipe -mschedule=7100LC -march=1.1 -fomit-frame-pointer"
CXXFLAGS="${CFLAGS}"
[edit] HPPA 2.0
CHOST="hppa2.0-unknown-linux-gnu"
CFLAGS="-O2 -pipe -mschedule=8000 -march=2.0 -fomit-frame-pointer"
CXXFLAGS="${CFLAGS}"
[edit] Alpha ev56
CHOST="alpha-unknown-linux-gnu"
CFLAGS="-mcpu=ev56 -mieee -O2 -pipe"
CXXFLAGS="${CFLAGS}"
[edit] Alpha ev67
CHOST="alpha-unknown-linux-gnu"
CFLAGS="-mieee -mcpu=ev67 -Wa,-mev6 -O2 -pipe "
CXXFLAGS="${CFLAGS}"
This applies to ev68 (Compaq Alpha DS25) machines too. Without -mcpu=ev67 -Wa,-mev6 flags it cannot bootstrap at all.
[edit] See Also
[edit] gcccpuopt
gcccpuopt is a script which reports the optimum (32 bit) CPU specific options for your CPU and gcc version.
[edit] Relevant files in gcc source code
These files may be interesting if you want to know what does exactly your -march= mean.
| Arch | File Location | Search for |
|---|---|---|
| x86/x86-64 | gcc/config/i386/i386.c | processor_alias_table[] |
| POWER | gcc/config/rs6000/rs6000.c | processor_target_table[] |
| SPARC | gcc/config/sparc/sparc.c | cpu_table[] |
| Alpha | gcc/config/alpha/alpha.c | cpu_table[] |
| HPPA | gcc/config/pa/pa.c | pa_handle_option (size_t code |
| ARM | gcc/config/arm/arm-cores.def |
[edit] Other links
- A guide to setting and choosing cflags
- CFLAGS matrix - table of many CFLAGS and people experience when using them
- Choosing safe LDFLAGS
- Setting package-specific CFLAGS
- Official Documentation: Compilation Optimization Guide
