Re: always use runtime checks for CRC-32C instructions

Nathan Bossart <nathandbossart@gmail.com> writes:
> On Mon, Oct 30, 2023 at 10:36:01PM -0500, Nathan Bossart wrote:
>> I tested pg_waldump -z with 50M 65-byte records for the following
>> implementations on an ARM system:
>> 
>> * slicing-by-8                                : ~3.08s
>> * proposed patches applied (runtime check)    : ~2.44s
>> * only CRC intrinsics implementation compiled : ~2.42s
>> * forced inlining                             : ~2.38s
>> 
>> Avoiding the runtime check produced a 0.8% improvement, and forced inlining
>> produced another 1.7% improvement.  In comparison, even the runtime check
>> implementation produced a 20.8% improvement over the slicing-by-8 one.

I find these numbers fairly concerning.  If you can see a
couple-of-percent slowdown on a macroscopic benchmark like pg_waldump,
that implies that the percentage slowdown considering the CRC
operation alone is much worse.  So there may be other use-cases where
we would take a bigger relative hit.

> * From my quick scan of a few dozen machines on the buildfarm, it looks
>   like the runtime checks are already the norm, so the number of systems
>   that would be subject to a regression from v16 to v17 should be pretty
>   small, in theory.  And this regression seems to be on the order of 1%
>   based on the numbers above.

I did a more thorough scrape of the buildfarm results.  Of 161 animals
currently reporting configure output on HEAD, we have

      2 ARMv8 CRC instructions
     36 ARMv8 CRC instructions with runtime check
      2 LoongArch CRCC instructions
      2 SSE 4.2
     52 SSE 4.2 with runtime check
     67 slicing-by-8

While that'd seem to support your conclusion, the two using ARM CRC
*without* a runtime check are my Apple M1 Mac animals (sifaka/indri);
and I see the same selection on my laptop.  So one platform where
we'd clearly be taking a regression is M-series Macs; that's a pretty
popular platform.  The two using SSE without a check are prion and
tayra.  I notice those are using gcc 11; so perhaps the default cflags
have changed to include -msse4.2 recently?  I couldn't see much other
pattern though.  (Scraping results attached in case anybody wants to
look.)

Really this just reinforces my concern that doing a runtime check
all the time is on the wrong side of history.  I grant that we've
got to do that for anything where the availability of the instruction
is really in serious question, but I'm not very convinced that that's
a majority situation on popular platforms.

            regards, tom lane

alimoche,ARMv8 CRC instructions with runtime check,aarch64,gcc (Debian 8.3.0-6) 8.3.0
batta,ARMv8 CRC instructions with runtime check,aarch64,gcc (Debian 10.2.1-6) 10.2.1 20210110
blackneck,ARMv8 CRC instructions with runtime check,aarch64,Debian clang version 13.0.1-6~deb10u4
boiga,ARMv8 CRC instructions with runtime check,aarch64,clang version 14.0.5 (Fedora 14.0.5-2.fc36)
broadbill,ARMv8 CRC instructions with runtime check,aarch64,gcc (GCC) 8.5.0 20210514 (Red Hat 8.5.0-18)
bulbul,ARMv8 CRC instructions with runtime check,aarch64,clang version 15.0.7 (Red Hat
15.0.7-1.module_el8.8.0+3466+dfcbc058)
corzo,ARMv8 CRC instructions with runtime check,aarch64,gcc (GCC) 12.2.1 20221121 (Red Hat 12.2.1-4)
desman,ARMv8 CRC instructions with runtime check,aarch64,gcc (GCC) 13.2.1 20230728 (Red Hat 13.2.1-1)
eelpout,ARMv8 CRC instructions with runtime check,aarch64,gcc (Debian 10.2.1-6) 10.2.1 20210110
gokiburi,ARMv8 CRC instructions with runtime check,aarch64,Debian clang version 11.0.1-2
hachi,ARMv8 CRC instructions with runtime check,aarch64,Debian clang version 11.0.1-2
jackdaw,ARMv8 CRC instructions with runtime check,aarch64,gcc (Debian 10.2.1-6) 10.2.1 20210110
massasauga,ARMv8 CRC instructions with runtime check,aarch64,gcc (GCC) 7.3.1 20180712 (Red Hat 7.3.1-17)
motmot,ARMv8 CRC instructions with runtime check,aarch64,clang version 16.0.6 (Fedora 16.0.6-2.fc38)
oystercatcher,ARMv8 CRC instructions with runtime check,aarch64,clang version 15.0.7 (Red Hat 15.0.7-2.el9)
potoo,ARMv8 CRC instructions with runtime check,aarch64,gcc (GCC) 11.3.1 20221121 (Red Hat 11.3.1-4)
rudd,ARMv8 CRC instructions with runtime check,aarch64,Ubuntu clang version 12.0.0-3ubuntu1~20.04.5
shiner,ARMv8 CRC instructions with runtime check,aarch64,Ubuntu clang version 15.0.7
snakefly,ARMv8 CRC instructions with runtime check,aarch64,gcc (GCC) 7.3.1 20180712 (Red Hat 7.3.1-17)
splitfin,ARMv8 CRC instructions with runtime check,aarch64,gcc (Ubuntu 9.4.0-1ubuntu1~20.04.2) 9.4.0
tanager,ARMv8 CRC instructions with runtime check,aarch64,clang version 7.0.1-8+deb10u2 (tags/RELEASE_701/final)
turbot,ARMv8 CRC instructions with runtime check,aarch64,gcc (Ubuntu 11.4.0-1ubuntu1~22.04) 11.4.0
vimba,ARMv8 CRC instructions with runtime check,aarch64,clang version 10.0.0-4ubuntu1~18.04.2
whinchat,ARMv8 CRC instructions with runtime check,aarch64,Debian clang version 13.0.1-6~deb11u1
whiting,ARMv8 CRC instructions with runtime check,aarch64,gcc (Ubuntu/Linaro 7.5.0-3ubuntu1~18.04) 7.5.0
widowbird,ARMv8 CRC instructions with runtime check,aarch64,gcc (Debian 10.2.1-6) 10.2.1 20210110
ziege,ARMv8 CRC instructions with runtime check,aarch64,gcc (GCC) 8.5.0 20210514 (Red Hat 8.5.0-21)
arowana,slicing-by-8,aarch64,gcc (GCC) 4.8.5 20150623 (Red Hat 4.8.5-44)
parula,ARMv8 CRC instructions with runtime check,aarch64/graviton3/c7g.2xl,gcc (GCC) 7.3.1 20180712 (Red Hat 7.3.1-17)
elver,SSE 4.2 with runtime check,amd64,FreeBSD clang version 14.0.5 (https://github.com/llvm/llvm-project.git
llvmorg-14.0.5-0-gc12386ae247c)
gombessa,SSE 4.2 with runtime check,amd64,OpenBSD clang version 10.0.1
hake,SSE 4.2 with runtime check,amd64,gcc (OpenIndiana 9.3.0-oi-4) 9.3.0
mule,SSE 4.2 with runtime check,amd64,gcc (Debian 12.2.0-14) 12.2.0
plover,SSE 4.2 with runtime check,amd64,OpenBSD clang version 10.0.1
pollock,SSE 4.2 with runtime check,amd64,gcc (OmniOS 151046/12.2.0-il-0) 12.2.0
indri,ARMv8 CRC instructions,arm64,Apple clang version 15.0.0 (clang-1500.0.40.1)
sifaka,ARMv8 CRC instructions,arm64,Apple clang version 15.0.0 (clang-1500.0.40.1)
dikkop,ARMv8 CRC instructions with runtime check,arm64,FreeBSD clang version 16.0.6
(https://github.com/llvm/llvm-project.gitllvmorg-16.0.6-0-g7cbf1a259152) 
opaleye,ARMv8 CRC instructions with runtime check,arm64,FreeBSD clang version 13.0.0
(git@github.com:llvm/llvm-project.gitllvmorg-13.0.0-0-gd7b669b3a303) 
chipmunk,ARMv8 CRC instructions with runtime check,armv6l,gcc (Raspbian 4.9.2-10+deb8u2) 4.9.2
grison,ARMv8 CRC instructions with runtime check,armv7,gcc (Raspbian 6.3.0-18+rpi1+deb9u1) 6.3.0 20170516
gull,ARMv8 CRC instructions with runtime check,armv7,clang version 13.0.0
(http://git.linaro.org/toolchain/jenkins-scripts.gitd04b0fafc2d906fd9b2e8e55efb35c9cf7114e68) 
mereswine,ARMv8 CRC instructions with runtime check,armv7,gcc (Debian 10.2.1-6) 10.2.1 20210110
dangomushi,ARMv8 CRC instructions with runtime check,armv7l,clang version 7.0.1-8+deb10u2 (tags/RELEASE_701/final)
turaco,ARMv8 CRC instructions with runtime check,armv7l,gcc (Raspbian 10.2.1-6+rpi1) 10.2.1 20210110
chickadee,slicing-by-8,hppa,gcc (nb1 20220722) 10.4.0
lapwing,SSE 4.2 with runtime check,i686,gcc (Debian 4.7.2-5) 4.7.2
cisticola,LoongArch CRCC instructions,loongarch64,gcc (GCC) 8.3.0 20190222 (Loongson 8.3.0-35 vec)
nuthatch,LoongArch CRCC instructions,loongarch64,gcc (GCC) 13.2.1 20230906
mamba,slicing-by-8,macppc,cc (nb1 20220722) 10.4.0
frogfish,slicing-by-8,mips64eb; -mabi=64,gcc (Debian 4.6.3-14) 4.6.3
topminnow,slicing-by-8,mips64el; -mabi=32,gcc (Debian 4.9.2-10+deb8u1) 4.9.2
hornet,slicing-by-8,ppc64 (power7),xlc_r -D_LARGE_FILES=1
mandrill,slicing-by-8,ppc64 (power7),xlc_r -D_LARGE_FILES=1 -DRANDOMIZE_ALLOCATED_MEMORY
sungazer,slicing-by-8,ppc64 (power7),gcc (GCC) 8.3.0
tern,slicing-by-8,ppc64 (power7),gcc (GCC) 8.3.0
boa,slicing-by-8,ppc64 (power8),gcc (GCC) 4.8.5 20150623 (Red Hat 4.8.5-44)
chub,slicing-by-8,ppc64 (power8),gcc (GCC) 4.8.5 20150623 (Red Hat 4.8.5-44)
hoverfly,slicing-by-8,ppc64 (power8),/opt/IBM/xlc/16.1.0/bin/xlc_r -D_LARGE_FILES=1 -DRANDOMIZE_ALLOCATED_MEMORY
batfish,slicing-by-8,ppc64le,clang version 5.0.0-3~16.04.1 (tags/RELEASE_500/final)
clam,slicing-by-8,ppc64le,"gcc (GCC) 5.2.1 20151016 (Advance-Toolchain-) [ibm/gcc-5-branch, revision: 229493 merged
fromgcc-5-branch, revision 228917]" 
cuon,slicing-by-8,ppc64le,gcc (Ubuntu/IBM 5.4.0-6ubuntu1~16.04.12) 5.4.0 20160609
dhole,slicing-by-8,ppc64le,gcc (GCC) 4.8.5 20150623 (Red Hat 4.8.5-44)
urocryon,slicing-by-8,ppc64le,gcc (Debian 6.3.0-18+deb9u1) 6.3.0 20170516
vulpes,slicing-by-8,ppc64le,gcc (GCC) 7.3.1 20180712 (Red Hat 7.3.1-6)
wobbegong,slicing-by-8,ppc64le,clang version 5.0.2 (tags/RELEASE_502/final)
ayu,slicing-by-8,ppc64le (power8),clang version 4.0.1-10~deb9u2 (tags/RELEASE_401/final)
babbler,slicing-by-8,ppc64le (power9),gcc (GCC) 8.5.0 20210514 (Red Hat 8.5.0-18)
bonito,slicing-by-8,ppc64le (power9),gcc (GCC) 8.3.1 20190223 (Red Hat 8.3.1-2)
buri,slicing-by-8,ppc64le (power9),gcc (GCC) 4.8.5 20150623 (Red Hat 4.8.5-44)
cascabel,slicing-by-8,ppc64le (power9),gcc (Debian 10.2.1-6) 10.2.1 20210110
cavefish,slicing-by-8,ppc64le (power9),gcc (Ubuntu 7.5.0-3ubuntu1~18.04) 7.5.0
chevrotain,slicing-by-8,ppc64le (power9),Debian clang version 13.0.1-6~deb11u1
chimaera,slicing-by-8,ppc64le (power9),gcc (Debian 6.3.0-18+deb9u1) 6.3.0 20170516
demoiselle,slicing-by-8,ppc64le (power9),clang version 5.0.1 (tags/RELEASE_501/final 312548)
elasmobranch,slicing-by-8,ppc64le (power9),gcc (SUSE Linux) 7.3.1 20180323 [gcc-7-branch revision 258812]
habu,slicing-by-8,ppc64le (power9),gcc (GCC) 13.2.1 20230728 (Red Hat 13.2.1-1)
kingsnake,slicing-by-8,ppc64le (power9),clang version 16.0.6 (Fedora 16.0.6-3.fc38)
krait,slicing-by-8,ppc64le (power9),gcc (GCC) 8.5.0 20210514 (Red Hat 8.5.0-21)
lancehead,slicing-by-8,ppc64le (power9),clang version 16.0.6 (Red Hat 16.0.6-2.module_el8+588+6f71ce7b)
nicator,slicing-by-8,ppc64le (power9),gcc (GCC) 11.3.1 20221121 (Red Hat 11.3.1-4)
pytilia,slicing-by-8,ppc64le (power9),clang version 15.0.7 (Red Hat 15.0.7-1.module_el8.8.0+3466+dfcbc058)
boomslang,slicing-by-8,riscv64,gcc (Debian 10.2.1-6) 10.2.1 20210110
copperhead,slicing-by-8,riscv64,gcc (Debian 10.2.1-6) 10.2.1 20210110
aracari,slicing-by-8,s390x (z15),clang version 15.0.7 (Red Hat 15.0.7-1.module+el8.8.0+17939+b58878af)
branta,slicing-by-8,s390x (z15),gcc-10 (Ubuntu 10.5.0-1ubuntu1~20.04) 10.5.0
cotinga,slicing-by-8,s390x (z15),clang version 10.0.0-4ubuntu1~18.04.2
lora,slicing-by-8,s390x (z15),gcc (GCC) 11.3.1 20221121 (Red Hat 11.3.1-4)
mamushi,slicing-by-8,s390x (z15),clang version 15.0.7 (Red Hat 15.0.7-2.el9)
perch,slicing-by-8,s390x (z15),gcc-8 (Ubuntu 8.4.0-1ubuntu1~18.04) 8.4.0
pike,slicing-by-8,s390x (z15),gcc (Debian 10.2.1-6) 10.2.1 20210110
pipit,slicing-by-8,s390x (z15),gcc (GCC) 8.5.0 20210514 (Red Hat 8.5.0-10)
rinkhals,slicing-by-8,s390x (z15),Debian clang version 13.0.1-6~deb11u1
ruddy,slicing-by-8,s390x (z15),gcc-11 (SUSE Linux) 11.2.1 20210816 [revision 056e324ce46a7924b5cf10f61010cf9dd2ca10e9]
sarus,slicing-by-8,s390x (z15),Ubuntu clang version 14.0.0-1ubuntu1.1
shelduck,slicing-by-8,s390x (z15),gcc (SUSE Linux) 4.8.5
siskin,slicing-by-8,s390x (z15),gcc (GCC) 4.8.5 20150623 (Red Hat 4.8.5-44)
urutau,slicing-by-8,s390x (z15),Ubuntu clang version 12.0.0-3ubuntu1~20.04.5
margay,slicing-by-8,sparc,gcc (GCC) 11.2.0
skate,slicing-by-8,sparc,gcc-4.7 (Debian 4.7.2-5) 4.7.2
snapper,slicing-by-8,sparc,gcc-4.7 (Debian 4.7.2-5) 4.7.2
grackle,slicing-by-8,sparc64,gcc (GCC) 12.2.0
ibisbill,slicing-by-8,sparc64,gcc (Debian 8.3.0-6) 8.3.0
kittiwake,slicing-by-8,sparc64,gcc (Debian 8.3.0-6) 8.3.0
mussurana,slicing-by-8,sparc64,gcc (Debian 6.3.0-18+deb9u1) 6.3.0 20170516
tadarida,slicing-by-8,sparc64,gcc (Debian 6.3.0-18+deb9u1) 6.3.0 20170516
thorntail,slicing-by-8,sparc64,gcc-11 (Debian 11.4.0-1) 11.4.0
waxbill,slicing-by-8,sparc64,gcc (Debian 12.3.0-4) 12.3.0
wrasse,slicing-by-8,sparc64,cc: Studio 12.6 Sun C 5.15 SunOS_sparc 152881-01 2018/03/16
morepork,SSE 4.2 with runtime check,x64,OpenBSD clang version 10.0.1
bushmaster,SSE 4.2 with runtime check,x86-64,clang version 17.0.3 (https://github.com/llvm/llvm-project.git
37b79e779f447f1c714af7f907e7a2ec846d1da0)
calliphoridae,SSE 4.2 with runtime check,x86-64,gcc (Debian 13.2.0-6) 13.2.0
canebrake,SSE 4.2 with runtime check,x86-64,gcc (Debian 13.2.0-6) 13.2.0
culicidae,SSE 4.2 with runtime check,x86-64,gcc (Debian 13.2.0-6) 13.2.0
grassquit,SSE 4.2 with runtime check,x86-64,gcc-12 (Debian 12.3.0-11) 12.3.0
kestrel,SSE 4.2 with runtime check,x86-64,Debian clang version 13.0.1-11+b2
olingo,SSE 4.2 with runtime check,x86-64,Debian clang version 13.0.1-11+b2
skink,SSE 4.2 with runtime check,x86-64,gcc (Debian 13.2.0-6) 13.2.0
taipan,SSE 4.2 with runtime check,x86-64,gcc (Debian 13.2.0-6) 13.2.0
tamandua,SSE 4.2 with runtime check,x86-64,gcc-12 (Debian 12.3.0-11) 12.3.0
urutu,SSE 4.2 with runtime check,x86-64,clang version 16.0.6 (https://github.com/llvm/llvm-project.git
7cbf1a2591520c2491aa35339f227775f4d3adf6)
prion,SSE 4.2,x86_64,gcc (GCC) 11.3.1 20221121 (Red Hat 11.3.1-4)
tayra,SSE 4.2,x86_64,gcc (GCC) 11.3.1 20221121 (Red Hat 11.3.1-4.3.0.4)
alabio,SSE 4.2 with runtime check,x86_64,gcc (Debian 10.2.1-6) 10.2.1 20210110
avocet,SSE 4.2 with runtime check,x86_64,gcc (SUSE Linux) 7.5.0
butterflyfish,SSE 4.2 with runtime check,x86_64,gcc (GCC) 6.3.0
caiman,SSE 4.2 with runtime check,x86_64,gcc (GCC) 13.2.1 20231011 (Red Hat 13.2.1-4)
conchuela,SSE 4.2 with runtime check,x86_64,clang version 14.0.6
culpeo,SSE 4.2 with runtime check,x86_64,gcc (Debian 12.2.0-14) 12.2.0
desmoxytes,SSE 4.2 with runtime check,x86_64,gcc (Debian 13.2.0-6) 13.2.0
dragonet,SSE 4.2 with runtime check,x86_64,clang version 3.9.1
flaviventris,SSE 4.2 with runtime check,x86_64,gcc (Debian 20231017-1) 14.0.0 20231017 (experimental) [master
r14-4678-gce55521bcd1]
guaibasaurus,SSE 4.2 with runtime check,x86_64,gcc (Debian 10.2.1-6) 10.2.1 20210110
hippopotamus,SSE 4.2 with runtime check,x86_64,gcc (SUSE Linux) 7.5.0
idiacanthus,SSE 4.2 with runtime check,x86_64,clang version 5.0.2
jay,SSE 4.2 with runtime check,x86_64,clang version 13.0.1
komodoensis,SSE 4.2 with runtime check,x86_64,gcc (Debian 13.2.0-6) 13.2.0
loach,SSE 4.2 with runtime check,x86_64,clang version 14.0.6
longfin,SSE 4.2 with runtime check,x86_64,Apple clang version 14.0.3 (clang-1403.0.22.14.1)
lorikeet,SSE 4.2 with runtime check,x86_64,gcc (GCC) 10.2.0
malleefowl,SSE 4.2 with runtime check,x86_64,gcc (Alpine 11.2.1_git20220219) 11.2.1 20220219
moonjelly,SSE 4.2 with runtime check,x86_64,gcc (GCC) 14.0.0 20230429 (experimental)
myna,SSE 4.2 with runtime check,x86_64,gcc (GCC) 7.3.0
peripatus,SSE 4.2 with runtime check,x86_64,FreeBSD clang version 14.0.5 (https://github.com/llvm/llvm-project.git
llvmorg-14.0.5-0-gc12386ae247c)
petalura,SSE 4.2 with runtime check,x86_64,clang version 4.0.1
phycodurus,SSE 4.2 with runtime check,x86_64,gcc (Debian 13.2.0-6) 13.2.0
pogona,SSE 4.2 with runtime check,x86_64,gcc (Debian 13.2.0-6) 13.2.0
rhinoceros,SSE 4.2 with runtime check,x86_64,gcc (GCC) 4.8.5 20150623 (Red Hat 4.8.5-44)
schnauzer,SSE 4.2 with runtime check,x86_64,OpenBSD clang version 13.0.0
seawasp,SSE 4.2 with runtime check,x86_64,clang version 17.0.0 (https://github.com/llvm/llvm-project.git
40222ddcf8f54fe523b2d14ab7005ebf412330f1)
serinus,SSE 4.2 with runtime check,x86_64,gcc (Debian 20231017-1) 14.0.0 20231017 (experimental) [master
r14-4678-gce55521bcd1]
sidewinder,SSE 4.2 with runtime check,x86_64,cc (nb4 20200810) 7.5.0
spurfowl,SSE 4.2 with runtime check,x86_64,gcc (Ubuntu 5.4.0-6ubuntu1~16.04.11) 5.4.0 20160609
trilobite,SSE 4.2 with runtime check,x86_64,clang version 13.0.1
xenodermus,SSE 4.2 with runtime check,x86_64,clang version 6.0.1
curculio,slicing-by-8,x86_64,gcc (GCC) 4.2.1 20070719
mantid,SSE 4.2 with runtime check,x86_64 (virtualized),gcc (GCC) 4.8.5 20150623 (Red Hat 4.8.5-44)
blossomcrown,slicing-by-8,,gcc (Ubuntu 7.5.0-3ubuntu1~18.04) 7.5.0
cardinalfish,slicing-by-8,,Debian clang version 14.0.6
cottonmouth,slicing-by-8,,gcc-12 (Ubuntu 12.1.0-2ubuntu1~22.04) 12.1.0
devario,slicing-by-8,,gcc (Debian 12.2.0-9) 12.2.0