GCC 7.2 vs. GCC 8 vs. LLVM Clang 5 vs. LLVM Clang 6 Znver1 EPYC AMD EPYC 7601 32-Core testing with a TYAN B8026T70AE24HR, testing various compilers. Tests for a future article on Phoronix. GCC 7.2.0: Processor: AMD EPYC 7601 32-Core @ 2.20GHz (32 Cores / 64 Threads), Motherboard: TYAN B8026T70AE24HR, Chipset: AMD Device 1450, Memory: 129024MB, Disk: 120GB Force MP500, Graphics: ASPEED ASPEED Family, Monitor: Acer P243W, Network: Broadcom Limited NetXtreme BCM5720 Gigabit PCIe OS: Ubuntu 17.10, Kernel: 4.13.0-16-generic (x86_64), Desktop: GNOME Shell 3.26.1, Compiler: GCC 7.2.0, File-System: ext4, Screen Resolution: 1920x1200 GCC 8.0.0 20171030: Processor: AMD EPYC 7601 32-Core @ 2.20GHz (32 Cores / 64 Threads), Motherboard: TYAN B8026T70AE24HR, Chipset: AMD Device 1450, Memory: 129024MB, Disk: 120GB Force MP500, Graphics: ASPEED ASPEED Family, Monitor: Acer P243W, Network: Broadcom Limited NetXtreme BCM5720 Gigabit PCIe OS: Ubuntu 17.10, Kernel: 4.13.0-16-generic (x86_64), Desktop: GNOME Shell 3.26.1, Compiler: GCC 8.0.0 20171030, File-System: ext4, Screen Resolution: 1920x1200 LLVM Clang 5.0: Processor: AMD EPYC 7601 32-Core @ 2.20GHz (32 Cores / 64 Threads), Motherboard: TYAN B8026T70AE24HR, Chipset: AMD Device 1450, Memory: 129024MB, Disk: 120GB Force MP500, Graphics: ASPEED ASPEED Family, Monitor: Acer P243W, Network: Broadcom Limited NetXtreme BCM5720 Gigabit PCIe OS: Ubuntu 17.10, Kernel: 4.13.0-16-generic (x86_64), Desktop: GNOME Shell 3.26.1, Compiler: Clang 5.0.0-3, File-System: ext4, Screen Resolution: 1920x1200 LLVM Clang 6.0 svn316959: Processor: AMD EPYC 7601 32-Core @ 2.20GHz (32 Cores / 64 Threads), Motherboard: TYAN B8026T70AE24HR, Chipset: AMD Device 1450, Memory: 129024MB, Disk: 120GB Force MP500, Graphics: ASPEED ASPEED Family, Monitor: Acer P243W, Network: Broadcom Limited NetXtreme BCM5720 Gigabit PCIe OS: Ubuntu 17.10, Kernel: 4.13.0-16-generic (x86_64), Desktop: GNOME Shell 3.26.1, Compiler: Clang 6.0.0-svn316959-1~exp1, File-System: ext4, Screen Resolution: 1920x1200 High Performance Conjugate Gradient 3.0 GFLOP/s > Higher Is Better GCC 7.2.0 ................ 0.73 |============================================= GCC 8.0.0 20171030 ....... 0.72 |============================================= LLVM Clang 5.0 ........... 0.74 |============================================== LLVM Clang 6.0 svn316959 . 0.72 |============================================= FFTW 3.3.6 Build: Stock - Size: 1D FFT Size 1024 Mflops > Higher Is Better GCC 7.2.0 ................ 8539.10 |=========================================== GCC 8.0.0 20171030 ....... 8416.23 |========================================== LLVM Clang 5.0 ........... 7613.47 |====================================== LLVM Clang 6.0 svn316959 . 7580.83 |====================================== FFTW 3.3.6 Build: Stock - Size: 2D FFT Size 1024 Mflops > Higher Is Better GCC 7.2.0 ................ 6420.23 |========================================== GCC 8.0.0 20171030 ....... 6622.90 |=========================================== LLVM Clang 5.0 ........... 6090.37 |======================================== LLVM Clang 6.0 svn316959 . 5852.20 |====================================== FFTW 3.3.6 Build: Float + SSE - Size: 1D FFT Size 1024 Mflops > Higher Is Better GCC 7.2.0 ................ 26763 |============================================= GCC 8.0.0 20171030 ....... 26876 |============================================= LLVM Clang 5.0 ........... 26224 |============================================ LLVM Clang 6.0 svn316959 . 26395 |============================================ FFTW 3.3.6 Build: Float + SSE - Size: 2D FFT Size 1024 Mflops > Higher Is Better GCC 7.2.0 ................ 19834 |========================================== GCC 8.0.0 20171030 ....... 21131 |============================================= LLVM Clang 5.0 ........... 18762 |======================================== LLVM Clang 6.0 svn316959 . 19036 |========================================= GNU GMP GMPbench 6.1.2 Total Time GMPbench Score > Higher Is Better GCC 7.2.0 .......... 3918.80 |================================================= GCC 8.0.0 20171030 . 3926.20 |================================================= SciMark 2.0 Computational Test: Composite Mflops > Higher Is Better GCC 7.2.0 ................ 1925.72 |=========================================== GCC 8.0.0 20171030 ....... 1933.09 |=========================================== LLVM Clang 5.0 ........... 1910.11 |========================================== LLVM Clang 6.0 svn316959 . 1923.63 |=========================================== SciMark 2.0 Computational Test: Monte Carlo Mflops > Higher Is Better GCC 7.2.0 ................ 194.45 |=============== GCC 8.0.0 20171030 ....... 554.99 |============================================ LLVM Clang 5.0 ........... 560.13 |============================================ LLVM Clang 6.0 svn316959 . 551.21 |=========================================== SciMark 2.0 Computational Test: Fast Fourier Transform Mflops > Higher Is Better GCC 7.2.0 ................ 223.69 |============================================ GCC 8.0.0 20171030 ....... 223.06 |============================================ LLVM Clang 5.0 ........... 221.42 |============================================ LLVM Clang 6.0 svn316959 . 221.33 |============================================ SciMark 2.0 Computational Test: Sparse Matrix Multiply Mflops > Higher Is Better GCC 7.2.0 ................ 2376.61 |======================================== GCC 8.0.0 20171030 ....... 2444.39 |========================================= LLVM Clang 5.0 ........... 2402.42 |========================================= LLVM Clang 6.0 svn316959 . 2534.37 |=========================================== SciMark 2.0 Computational Test: Dense LU Matrix Factorization Mflops > Higher Is Better GCC 7.2.0 ................ 5150.44 |=========================================== GCC 8.0.0 20171030 ....... 4757.53 |======================================== LLVM Clang 5.0 ........... 4938.62 |========================================= LLVM Clang 6.0 svn316959 . 4883.39 |========================================= SciMark 2.0 Computational Test: Jacobi Successive Over-Relaxation Mflops > Higher Is Better GCC 7.2.0 ................ 1683.40 |=========================================== GCC 8.0.0 20171030 ....... 1685.48 |=========================================== LLVM Clang 5.0 ........... 1427.97 |==================================== LLVM Clang 6.0 svn316959 . 1427.86 |==================================== TSCP 1.81 AI Chess Performance Nodes Per Second > Higher Is Better GCC 7.2.0 ................ 861129 |========================================= GCC 8.0.0 20171030 ....... 853655 |======================================== LLVM Clang 5.0 ........... 895284 |========================================== LLVM Clang 6.0 svn316959 . 933206 |============================================ TTSIOD 3D Renderer 2.3a Phong Rendering With Soft-Shadow Mapping FPS > Higher Is Better GCC 7.2.0 ................ 407.27 |============================================ GCC 8.0.0 20171030 ....... 407.58 |============================================ LLVM Clang 5.0 ........... 41.26 |==== LLVM Clang 6.0 svn316959 . 41.02 |==== GraphicsMagick 1.3.19 Operation: Blur Iterations Per Minute > Higher Is Better GCC 7.2.0 ................ 148 |=============================================== GCC 8.0.0 20171030 ....... 148 |=============================================== LLVM Clang 5.0 ........... 115 |===================================== LLVM Clang 6.0 svn316959 . 116 |===================================== GraphicsMagick 1.3.19 Operation: Sharpen Iterations Per Minute > Higher Is Better GCC 7.2.0 ................ 177 |============================================== GCC 8.0.0 20171030 ....... 179 |=============================================== LLVM Clang 5.0 ........... 145 |====================================== LLVM Clang 6.0 svn316959 . 146 |====================================== GraphicsMagick 1.3.19 Operation: Resizing Iterations Per Minute > Higher Is Better GCC 7.2.0 .......... 174 |==================================================== GCC 8.0.0 20171030 . 176 |===================================================== GraphicsMagick 1.3.19 Operation: HWB Color Space Iterations Per Minute > Higher Is Better GCC 7.2.0 ................ 198 |=============================================== GCC 8.0.0 20171030 ....... 199 |=============================================== LLVM Clang 5.0 ........... 161 |====================================== LLVM Clang 6.0 svn316959 . 162 |====================================== GraphicsMagick 1.3.19 Operation: Local Adaptive Thresholding Iterations Per Minute > Higher Is Better GCC 7.2.0 ................ 109 |=============================================== GCC 8.0.0 20171030 ....... 110 |=============================================== LLVM Clang 5.0 ........... 110 |=============================================== LLVM Clang 6.0 svn316959 . 110 |=============================================== Himeno Benchmark 3.0 Poisson Pressure Solver MFLOPS > Higher Is Better GCC 7.2.0 ................ 936.77 |========================================== GCC 8.0.0 20171030 ....... 942.44 |========================================== LLVM Clang 5.0 ........... 965.54 |=========================================== LLVM Clang 6.0 svn316959 . 978.76 |============================================ C-Ray 1.1 Total Time Seconds < Lower Is Better GCC 7.2.0 ................ 3.08 |=============================== GCC 8.0.0 20171030 ....... 2.74 |============================ LLVM Clang 5.0 ........... 4.26 |=========================================== LLVM Clang 6.0 svn316959 . 4.55 |============================================== Primesieve 6.2 1e12 Prime Number Generation Seconds < Lower Is Better GCC 7.2.0 ................ 11.91 |============================================= GCC 8.0.0 20171030 ....... 11.94 |============================================= LLVM Clang 5.0 ........... 11.82 |============================================= LLVM Clang 6.0 svn316959 . 11.80 |============================================ Smallpt 1.0 Global Illumination Renderer; 100 Samples Seconds < Lower Is Better GCC 7.2.0 ................ 4 |================================================= GCC 8.0.0 20171030 ....... 4 |================================================= LLVM Clang 5.0 ........... 4 |================================================= LLVM Clang 6.0 svn316959 . 4 |================================================= Stockfish 2014-11-26 Total Time ms < Lower Is Better GCC 7.2.0 ................ 4485 |============================================== GCC 8.0.0 20171030 ....... 4500 |============================================== LLVM Clang 5.0 ........... 4356 |============================================= LLVM Clang 6.0 svn316959 . 4429 |============================================= LAME MP3 Encoding 3.99.5 WAV To MP3 Seconds < Lower Is Better GCC 7.2.0 ................ 11.20 |======================================= GCC 8.0.0 20171030 ....... 10.71 |===================================== LLVM Clang 5.0 ........... 12.81 |============================================= LLVM Clang 6.0 svn316959 . 12.89 |============================================= FFmpeg 3.3.3 H.264 HD To NTSC DV Seconds < Lower Is Better GCC 7.2.0 ................ 10.69 |============================================= GCC 8.0.0 20171030 ....... 10.66 |============================================= LLVM Clang 5.0 ........... 10.41 |============================================ LLVM Clang 6.0 svn316959 . 10.45 |============================================ libjpeg-turbo tjbench 1.5.1 Test: Decompression Throughput Megapixels/sec > Higher Is Better GCC 7.2.0 ................ 141.55 |=========================================== GCC 8.0.0 20171030 ....... 143.79 |============================================ LLVM Clang 5.0 ........... 142.26 |=========================================== LLVM Clang 6.0 svn316959 . 144.46 |============================================ PostgreSQL pgbench 10.0 Scaling: Buffer Test - Test: Normal Load - Mode: Read Only TPS > Higher Is Better GCC 7.2.0 .......... 308134.57 |=============================================== GCC 8.0.0 20171030 . 303543.17 |============================================== LLVM Clang 5.0 ..... 167561.19 |========================== PostgreSQL pgbench 10.0 Scaling: Buffer Test - Test: Single Thread - Mode: Read Only TPS > Higher Is Better GCC 7.2.0 .......... 10784.92 |================================================ GCC 8.0.0 20171030 . 10709.13 |================================================ LLVM Clang 5.0 ..... 10669.81 |=============================================== Redis 3.0.1 Test: LPOP Requests Per Second > Higher Is Better GCC 7.2.0 ................ 202065.84 |==== GCC 8.0.0 20171030 ....... 195542.12 |==== LLVM Clang 5.0 ........... 191613.95 |==== LLVM Clang 6.0 svn316959 . 1823824.96 |======================================== Redis 3.0.1 Test: SADD Requests Per Second > Higher Is Better GCC 7.2.0 ................ 189478.65 |===== GCC 8.0.0 20171030 ....... 193163.62 |===== LLVM Clang 5.0 ........... 194776.06 |===== LLVM Clang 6.0 svn316959 . 1489239.30 |======================================== Redis 3.0.1 Test: LPUSH Requests Per Second > Higher Is Better GCC 7.2.0 ................ 193867.82 |====== GCC 8.0.0 20171030 ....... 192219.37 |====== LLVM Clang 5.0 ........... 203774.66 |======= LLVM Clang 6.0 svn316959 . 1247939.29 |======================================== Redis 3.0.1 Test: GET Requests Per Second > Higher Is Better GCC 7.2.0 ................ 198580.74 |===== GCC 8.0.0 20171030 ....... 198883.33 |===== LLVM Clang 5.0 ........... 197952.09 |===== LLVM Clang 6.0 svn316959 . 1584704.00 |======================================== Redis 3.0.1 Test: SET Requests Per Second > Higher Is Better GCC 7.2.0 ................ 192063.14 |====== GCC 8.0.0 20171030 ....... 193210.49 |====== LLVM Clang 5.0 ........... 199028.19 |====== LLVM Clang 6.0 svn316959 . 1260800.37 |========================================