Intel Xeon Cascade Lake Compiler Performance - GCC 9/10 vs. LLVM Clang 8/9

Written by Michael Larabel in Software on 27 June 2019 at 01:00 PM EDT.
Cascade Lake Compiler Benchmarks
Cascade Lake Compiler Benchmarks

Something we've seen on other systems in past compiler comparisons is the Intel MKL-DNN math kernel library benchmark performing much better when built on Clang than GCC.

Cascade Lake Compiler Benchmarks
Cascade Lake Compiler Benchmarks
Cascade Lake Compiler Benchmarks

For Intel's SVT video encoder projects meanwhile the AV1 encoder came out about even while GCC 9 was the fastest for the HEVC and VP9 CPU-based encoders, but in HEVC/VP9 those compilers were doing some odd behavior with regards to the compiler flags that may explain some of the difference.

Cascade Lake Compiler Benchmarks

The VP9 vpxenc performance came out slightly ahead on the LLVM Clang compilers.

Cascade Lake Compiler Benchmarks
Cascade Lake Compiler Benchmarks
Cascade Lake Compiler Benchmarks
Cascade Lake Compiler Benchmarks
Cascade Lake Compiler Benchmarks

While the LLVM Clang compiler has increasingly good OpenMP support these days, in OMP workloads like GraphicsMagick the performance still tends to be better with the GNU Compiler Collection.

Cascade Lake Compiler Benchmarks

GCC also led in the Himeno pressure solver tests.


Related Articles