CPU_FLAGS_X86: aes avx f16c mmx mmxext pclmul popcnt rdrand sse sse2 sse3 sse4_1 sse4_2 ssse3
CPUTYPE?=core-avx-i
CPUTYPE=native achieves.This is make's syntax for "only set the value if there's nothing set yet", so it enables to override it e.g. from the commandline.
OPTIONS_SET+=SSE
OPTIONS_SET+=SSE2
OPTIONS_SET+=SSE3
OPTIONS_SET+=SSE4_1
OPTIONS_SET+=SSE4_2
OPTIONS_SET+=SSE41
OPTIONS_SET+=SSE42
OPTIONS_SET+=SSSE3
OPTIONS_SET+=AVX
OPTIONS_SET+=OPTIMIZED_CFLAGS
4m34.63s real 4m27.83s user 5.15s sys
6m14.65s real 6m8.85s user 3.53s sysThere is a default per architecture. IIRC, for the i386 architecture, this was recently changed from i486 to i686. For most software, the difference won't be measurable indeed. For some "special purpose" tasks OTOH, it can be a substantial difference. Some of this software already does runtime detection of CPU features and picks the best code path. You could of course do the tedious task to identify software not doing that, but still profiting, and enable "optimized" flags there. Whether that's really worthwhile, well, I would say: no.
All
native does is to let the compiler automatically detect the CPU type. So, if you always build on the same machine you're also running the software, it's just the same as explicitly setting some value...Thanks. That's what I imagined.
I never bothered to measure. I just tried to figure the correct setting and put it in place. This is sometimes a bit more difficult, because the building system does not run what the target system runs and vice versa (e.g. Ivybridge/Avoton), and then a common ground needs to be found.
I don't think that's necessary. If there is need for a machine that "just works" (and can easily be swapped), the defaults should do. And if a specific workhorse gets tuned and optimized, then the CPU is known and the code can be built individually for the target.
"native" does not work on my machines:
CPUTYPE?=native# make
===> Building for lsof-4.96.4,8
Constructing version.h
(cd lib; /usr/bin/make DEBUG="-O2" CFGF="-pipe -march=native -fstack-protector-strong -fno-strict-aliasing -DNEEDS_BOOL_TYPEDEF -DHASTASKS -DHAS_DUP2 -DHAS_CLOSEFROM -DHASEFFNLINK=i_effnlink -DHASF_VNODE -DHAS_FILEDESCENT -DHAS_TMPFS -DHASWCTYPE_H -DHASSBSTATE -DHAS_KVM_VNODE -DHAS_UFS1_2 -DHAS_NO_IDEV -DHAS_VM_MEMATTR_T -DNEEDS_DEVICE_T -DHAS_CDEV2PRIV -DHAS_NO_SI_UDEV -DHAS_SYS_SX_H -DHASFUSEFS -DHASMSDOSFS -DHAS_V_LOCKF -DHAS_LOCKF_ENTRY -DHAS_NO_6PORT -DHAS_NO_6PPCB -DNEEDS_BOOLEAN_T -DHAS_SB_CCC -DHAS_FDESCENTTBL -DFREEBSDV=13000 -DHASFDESCFS=2 -DHASPROCFS -DHASPSEUDOFS -DHASNULLFS -DHAS9660FS -DHAS_NO_ISO_DEV -DHASIPv6 -DHASUTMPX -DHAS_XTCPCB_TMAXSEG -DHAS_KF_SOCK_SENDQ -DHAS_STRFTIME -DLSOF_VSTR=\"13.1-RELEASE-p3\"")
cc -pipe -march=native -fstack-protector-strong -fno-strict-aliasing -DNEEDS_BOOL_TYPEDEF -DHASTASKS -DHAS_DUP2 -DHAS_CLOSEFROM -DHASEFFNLINK=i_effnlink -DHASF_VNODE -DHAS_FILEDESCENT -DHAS_TMPFS -DHASWCTYPE_H -DHASSBSTATE -DHAS_KVM_VNODE -DHAS_UFS1_2 -DHAS_NO_IDEV -DHAS_VM_MEMATTR_T -DNEEDS_DEVICE_T -DHAS_CDEV2PRIV -DHAS_NO_SI_UDEV -DHAS_SYS_SX_H -DHASFUSEFS -DHASMSDOSFS -DHAS_V_LOCKF -DHAS_LOCKF_ENTRY -DHAS_NO_6PORT -DHAS_NO_6PPCB -DNEEDS_BOOLEAN_T -DHAS_SB_CCC -DHAS_FDESCENTTBL -DFREEBSDV=13000 -DHASFDESCFS=2 -DHASPROCFS -DHASPSEUDOFS -DHASNULLFS -DHAS9660FS -DHAS_NO_ISO_DEV -DHASIPv6 -DHASUTMPX -DHAS_XTCPCB_TMAXSEG -DHAS_KF_SOCK_SENDQ -DHAS_STRFTIME -DLSOF_VSTR="13.1-RELEASE-p3" -I/usr/src/sys -O2 -c ckkv.c -o ckkv.o
error: unknown target CPU 'athlon-xp'
note: valid target CPU values are: nocona, core2, penryn, bonnell, atom, silvermont, slm, goldmont, goldmont-plus, tremont, nehalem, corei7, westmere, sandybridge, corei7-avx, ivybridge, core-avx-i, haswell, core-avx2, broadwell, skylake, skylake-avx512, skx, cascadelake, cooperlake, cannonlake, icelake-client, rocketlake, icelake-server, tigerlake, sapphirerapids, alderlake, knl, knm, k8, athlon64, athlon-fx, opteron, k8-sse3, athlon64-sse3, opteron-sse3, amdfam10, barcelona, btver1, btver2, bdver1, bdver2, bdver3, bdver4, znver1, znver2, znver3, x86-64, x86-64-v2, x86-64-v3, x86-64-v4
*** [ckkv.o] Error code 1
make[2]: stopped in /usr/ports/sysutils/lsof/work/lsof-4.96.4/lib
1 errorExactly. It's a VIA Samuel with amd64 support and SSE3
CPU: QEMU Virtual CPU version 2.5+ (2994.61-MHz K8-class CPU)
Origin="AuthenticAMD" Id=0x663 Family=0x6 Model=0x6 Stepping=3
Features=0x783fbfd<FPU,DE,PSE,TSC,MSR,PAE,MCE,CX8,APIC,SEP,MTRR,PGE,MCA,CMOV,PAT,PSE36,MMX,FXSR,SSE,SSE2>
Features2=0x80202001<SSE3,CX16,x2APIC,HV>
AMD Features=0x20100800<SYSCALL,NX,LM>
AMD Features2=0x1<LAHF>
qemu.readthedocs.io
Except when it gets it wrong which has occurred numerous occasions using both GCC and LLVM/Clang
What could possibly go wrong?
