GCC 4.9 vs. LLVM Clang 3.5 On Intel's Haswell-E Platform Under Linux

[phoronix]

Phoronix: GCC 4.9 vs. LLVM Clang 3.5 On Intel's Haswell-E Platform Under Linux

With my Intel Core i7 5960X Haswell-E Linux review out there, one of the quick to be requested extra tests is benchmarking the i7-5960X 16-thread processor with LLVM/Clang against GCC. Here's some initial data comparing the compilers for this $1000+ processor...

http://www.phoronix.com/vr.php?view=MTc4NTc

[AJenbo]

I like these boiled down charts for the less ground breaking benchmarks, but they could be turned slightly more sexy I think

[Michael]

I like these boiled down charts for the less ground breaking benchmarks, but they could be turned slightly more sexy I think

Patches welcome for the Phoronix Test Suite to improve the "sexiness" of the graph.... Or at the very least provide some examples of such rendered charts.

[mendieta]

WTF is going on with John the Ripper under LLVM??

[Michael]

WTF is going on with John the Ripper under LLVM??

No OpenMP support....

[Marc Driftmeyer]

No OpenMP support....

Then test it against Intel's branch.

[Michael]

Then test it against Intel's branch.

This is a comparison just against the mainline compilers since that's what most people use and what's packaged by the distributions. If you're curious about LLVM OpenMP, there's this article focusing on that matter - http://www.phoronix.com/vr.php?view=20488

[samdraz]

Under i7-3630qm for scimark

Under i7-3630qm for scimark

LLVM-3.5

**                                                              **
** SciMark2 Numeric Benchmark, see http://math.nist.gov/scimark **
** for details. (Results can be submitted to [email protected])     **
**                                                              **
Using       2.00 seconds min time per kenel.
Composite Score:         1898.71
FFT             Mflops:  1237.14    (N=1024)
SOR             Mflops:  1573.29    (100 x 100)
MonteCarlo:     Mflops:   500.75
Sparse matmult  Mflops:  2132.57    (N=1000, nz=5000)
LU              Mflops:  4049.82    (M=100, N=100)

GCC-4.9.1

**                                                              **
** SciMark2 Numeric Benchmark, see http://math.nist.gov/scimark **
** for details. (Results can be submitted to [email protected])     **
**                                                              **
Using       2.00 seconds min time per kenel.
Composite Score:         1898.71
FFT             Mflops:  1237.14    (N=1024)
SOR             Mflops:  1573.29    (100 x 100)
MonteCarlo:     Mflops:   500.75
Sparse matmult  Mflops:  2132.57    (N=1000, nz=5000)
LU              Mflops:  4049.82    (M=100, N=100)

[mufasa72]

This is a comparison just against the mainline compilers since that's what most people use and what's packaged by the distributions. If you're curious about LLVM OpenMP, there's this article focusing on that matter - http://www.phoronix.com/vr.php?view=20488

How frequently is the OpenMP branch rebased no top of LLVM TOT? I would actually be interested in seeing new benchmarks, which include John the Ripper.

[dnebdal]

Under i7-3630qm for scimark

LLVM-3.5

**                                                              **
** SciMark2 Numeric Benchmark, see http://math.nist.gov/scimark **
** for details. (Results can be submitted to [email protected])     **
**                                                              **
Using       2.00 seconds min time per kenel.
Composite Score:         1898.71
FFT             Mflops:  1237.14    (N=1024)
SOR             Mflops:  1573.29    (100 x 100)
MonteCarlo:     Mflops:   500.75
Sparse matmult  Mflops:  2132.57    (N=1000, nz=5000)
LU              Mflops:  4049.82    (M=100, N=100)

GCC-4.9.1

**                                                              **
** SciMark2 Numeric Benchmark, see http://math.nist.gov/scimark **
** for details. (Results can be submitted to [email protected])     **
**                                                              **
Using       2.00 seconds min time per kenel.
Composite Score:         1898.71
FFT             Mflops:  1237.14    (N=1024)
SOR             Mflops:  1573.29    (100 x 100)
MonteCarlo:     Mflops:   500.75
Sparse matmult  Mflops:  2132.57    (N=1000, nz=5000)
LU              Mflops:  4049.82    (M=100, N=100)

Uhm, did you accidentally post the same table twice? It seems unlikely that they'd be identical to two decimals.