[PATCH] D40979: [DWARF] Allow duplication of tails with CFI instructions
Francis Visoiu Mistrih via Phabricator via llvm-commits
llvm-commits at lists.llvm.org
Mon Jan 8 12:08:08 PST 2018
thegameg added a comment.
In https://reviews.llvm.org/D40979#969990, @MatzeB wrote:
> Hmm yeah, I guess compact unwind is important for code size. So we would prefer to stay with the current behavior on darwin as you did with this patch. (Though I've never seen actual numbers of how big an impact it has).
I compiled the test suite + SPEC for AArch64 -Os. The diff is between trunk and trunk with generateCompactUnwinding <https://github.com/llvm-mirror/llvm/blob/master/lib/Target/AArch64/MCTargetDesc/AArch64AsmBackend.cpp#L442> returning `UNWIND_ARM64_MODE_DWARF` if there are any CFI directives.
Program diff
cfp2000
183.equake 30.9%
188.ammp 11.5%
177.mesa 6.8%
cfp2006
433.milc 10.9%
482.sphinx3 8.1%
450.soplex 7.9%
447.dealII 7.1%
453.povray 7.0%
cint2000
197.parser 10.3%
254.gap 9.9%
255.vortex 8.1%
253.perlbmk 8.0%
300.twolf 6.9%
176.gcc 5.7%
cint2006
462.libquantum 22.5%
471.omnetpp 11.9%
456.hmmer 11.2%
483.xalancbmk 10.2%
400.perlbench 7.6%
403.gcc 6.4%
464.h264ref 6.1%
445.gobmk 3.0%
cint95
132.ijpeg 17.6%
130.li 16.4%
124.m88ksim 8.9%
147.vortex 8.1%
134.perl 5.7%
099.go 4.8%
test_suite
SingleSour...nchmarks/Adobe-C++/loop_unroll 81.5%
MultiSourc...enchmarks/VersaBench/dbms/dbms 30.1%
SingleSour...arks/Adobe-C++/stepanov_vector 29.0%
MultiSourc.../Reductions-flt/Reductions-flt 24.2%
SingleSour.../simple_types_constant_folding 23.8%
MultiSourc...nchmarks/FreeBench/pifft/pifft 23.5%
MultiSourc...ALAC/encode/alacconvert-encode 21.9%
MultiSourc...ALAC/decode/alacconvert-decode 21.9%
MultiSource/Applications/siod/siod 20.2%
MultiSourc...plications/lambda-0.1.3/lambda 20.0%
MultiSource/Benchmarks/Ptrdist/bc/bc 17.9%
SingleSour...++/simple_types_loop_invariant 14.3%
MultiSourc...Benchmarks/7zip/7zip-benchmark 14.1%
MultiSource/Benchmarks/PAQ8p/paq8p 11.8%
MultiSourc...OE-ProxyApps-C++/miniFE/miniFE 11.0%
MultiSource/Applications/SPASS/SPASS 10.9%
MultiSourc...ch/consumer-jpeg/consumer-jpeg 10.0%
MultiSourc...-ProxyApps-C++/PENNANT/PENNANT 9.8%
MultiSourc.../mediabench/jpeg/jpeg-6a/cjpeg 9.1%
MultiSourc.../MallocBench/espresso/espresso 8.9%
MultiSource/Applications/lua/lua 8.6%
MultiSourc...e/Applications/sqlite3/sqlite3 8.3%
MultiSourc...ks/ASCI_Purple/SMG2000/smg2000 7.9%
MultiSourc...e/Benchmarks/MallocBench/gs/gs 7.2%
MultiSource/Applications/kimwitu++/kc 7.2%
MultiSourc...ch/consumer-lame/consumer-lame 6.9%
MultiSource/Applications/d/make_dparser 5.8%
MultiSourc.../Applications/JM/ldecod/ldecod 5.0%
MultiSourc...e/Applications/ClamAV/clamscan 4.9%
MultiSourc...enchmarks/mafft/pairlocalalign 4.5%
MultiSourc.../DOE-ProxyApps-C++/CLAMR/CLAMR 4.0%
MultiSource/Benchmarks/Bullet/bullet 3.9%
SingleSour.../Benchmarks/Misc-C++-EH/spirit 3.5%
MultiSourc...sumer-typeset/consumer-typeset 3.0%
MultiSourc.../Applications/JM/lencod/lencod 2.5%
MultiSource/Applications/oggenc/oggenc 0.8%
SingleSour...Adobe-C++/stepanov_abstraction 0.1%
Repository:
rL LLVM
https://reviews.llvm.org/D40979
More information about the llvm-commits
mailing list