[llvm-testresults] Grawp-NO-LTO-PIC i386 nightly tester results
Dale Johannesen
dalej at apple.com
Wed Jan 7 10:59:06 PST 2009
On Jan 7, 2009, at 10:55 AMPST, Evan Cheng wrote:
> I see significant loses (and wins). Is this due to disabling of gcc
> inliner?
>
> To reproduce do something like:
> make DISABLE_LTO=1 TARGET_FLAGS= TARGET_LLCFLAGS="-relocation-
> model=pic -disable-fp-elim"
To experiment with increasing aggressiveness of llvm inliner, use -
inline-threshold=<number> . Values in the 375-400 range got me about
the same amount of inlining as we were getting before on one case I
looked at (default is 200).
> Evan
>
> On Jan 7, 2009, at 8:54 AM, Apache wrote:
>>
>>
>> Significant changes in test results:
>> CBE:
>> singlesource/Benchmarks/BenchmarkGame/fannkuch: -5.88% (4.59 => 4.86)
>> singlesource/Benchmarks/BenchmarkGame/fasta: -19.58% (1.43 => 1.71)
>> singlesource/Benchmarks/BenchmarkGame/recursive: -174.04% (1.04 =>
>> 2.85)
>> singlesource/Benchmarks/Misc-C++/sphereflake: -23.23% (5.38 => 6.63)
>> singlesource/Benchmarks/Shootout-C++/ackermann: -31.89% (1.85 =>
>> 2.44)
>> singlesource/Benchmarks/Shootout-C++/ary3: 12.02% (4.99 => 4.39)
>> singlesource/Benchmarks/Shootout-C++/fibo: -205.34% (2.81 => 8.58)
>> singlesource/Benchmarks/Shootout/fib2: -209.75% (2.77 => 8.58)
>> singlesource/Benchmarks/Shootout/lists: -9.52% (7.35 => 8.05)
>> multisource/Applications/SIBsim4/SIBsim4: -5.32% (6.39 => 6.73)
>> multisource/Applications/aha/aha: -62.85% (3.23 => 5.26)
>> multisource/Applications/lemon/lemon: -10.73% (2.05 => 2.27)
>> multisource/Benchmarks/Olden/bh/bh: -16.86% (3.44 => 4.02)
>> multisource/Benchmarks/Olden/treeadd/treeadd: -90.06% (3.42 => 6.50)
>> multisource/Benchmarks/VersaBench/bmm/bmm: -13.50% (2.74 => 3.11)
>> external/SPEC/CFP2000/177.mesa/177.mesa: -19.50% (3.59 => 4.29)
>> external/SPEC/CINT2000/164.gzip/164.gzip: 12.56% (19.34 => 16.91)
>> external/SPEC/CINT2000/175.vpr/175.vpr: -5.09% (5.89 => 6.19)
>> external/SPEC/CINT2000/197.parser/197.parser: -6.54% (3.21 => 3.42)
>> external/SPEC/CINT2000/256.bzip2/256.bzip2: 11.86% (19.64 => 17.31)
>> LLC:
>> singlesource/Benchmarks/BenchmarkGame/fasta: -28.83% (1.11 => 1.43)
>> singlesource/Benchmarks/BenchmarkGame/recursive: -184.16% (1.01 =>
>> 2.87)
>> singlesource/Benchmarks/Dhrystone/dry: -10.61% (3.30 => 3.65)
>> singlesource/Benchmarks/Dhrystone/fldry: 7.87% (3.94 => 3.63)
>> singlesource/Benchmarks/McGill/queens: -9.97% (3.31 => 3.64)
>> singlesource/Benchmarks/Shootout-C++/ackermann: -54.49% (1.78 =>
>> 2.75)
>> singlesource/Benchmarks/Shootout-C++/fibo: -243.46% (2.60 => 8.93)
>> singlesource/Benchmarks/Shootout/fib2: -238.26% (2.64 => 8.93)
>> singlesource/Benchmarks/Shootout/matrix: 22.58% (4.96 => 3.84)
>> multisource/Applications/aha/aha: -44.85% (3.59 => 5.20)
>> multisource/Applications/lemon/lemon: -10.85% (2.12 => 2.35)
>> multisource/Benchmarks/Fhourstones-3.1/fhourstones3.1: -13.31%
>> (2.48 => 2.81)
>> multisource/Benchmarks/Olden/bh/bh: 36.01% (5.97 => 3.82)
>> multisource/Benchmarks/Olden/treeadd/treeadd: -97.96% (3.43 => 6.79)
>> multisource/Benchmarks/Trimaran/enc-pc1/enc-pc1: 58.26% (3.45 =>
>> 1.44)
>> multisource/Benchmarks/VersaBench/bmm/bmm: -10.53% (2.28 => 2.52)
>> external/SPEC/CFP2000/177.mesa/177.mesa: -28.06% (3.35 => 4.29)
>>
>>
>> _______________________________________________
>> llvm-testresults mailing list
>> llvm-testresults at cs.uiuc.edu
>> http://lists.cs.uiuc.edu/mailman/listinfo/llvm-testresults
>
More information about the llvm-testresults
mailing list