GCC AMD Znver3 Compiler Optimization Levels Benchmarks for a future article. -O3 -march=znver3: Processor: AMD Ryzen 9 5950X 16-Core @ 3.40GHz (16 Cores / 32 Threads), Motherboard: ASUS ROG CROSSHAIR VIII HERO (WI-FI) (3003 BIOS), Chipset: AMD Starship/Matisse, Memory: 16GB, Disk: 2000GB Corsair Force MP600, Graphics: AMD Radeon RX 5600 OEM/5600 XT / 5700/5700 8GB (2100/875MHz), Audio: AMD Navi 10 HDMI Audio, Monitor: ASUS MG28U, Network: Realtek RTL8125 2.5GbE + Intel I211 + Intel Wi-Fi 6 AX200 OS: Ubuntu 20.04, Kernel: 5.10.0-051000rc6daily20201205-generic (x86_64) 20201204, Desktop: GNOME Shell 3.36.4, Display Server: X Server 1.20.8, Display Driver: modesetting 1.20.8, OpenGL: 4.6 Mesa 21.0.0-devel (git-1a53572 2020-12-09 focal-oibaf-ppa) (LLVM 11.0.0), Vulkan: 1.2.145, Compiler: GCC 11.0.0 20201213, File-System: ext4, Screen Resolution: 3840x2160 -O3 -march=znver2: Processor: AMD Ryzen 9 5950X 16-Core @ 3.40GHz (16 Cores / 32 Threads), Motherboard: ASUS ROG CROSSHAIR VIII HERO (WI-FI) (3003 BIOS), Chipset: AMD Starship/Matisse, Memory: 16GB, Disk: 2000GB Corsair Force MP600, Graphics: AMD Radeon RX 5600 OEM/5600 XT / 5700/5700 8GB (2100/875MHz), Audio: AMD Navi 10 HDMI Audio, Monitor: ASUS MG28U, Network: Realtek RTL8125 2.5GbE + Intel I211 + Intel Wi-Fi 6 AX200 OS: Ubuntu 20.04, Kernel: 5.10.0-051000rc6daily20201205-generic (x86_64) 20201204, Desktop: GNOME Shell 3.36.4, Display Server: X Server 1.20.8, Display Driver: modesetting 1.20.8, OpenGL: 4.6 Mesa 21.0.0-devel (git-1a53572 2020-12-09 focal-oibaf-ppa) (LLVM 11.0.0), Vulkan: 1.2.145, Compiler: GCC 11.0.0 20201213, File-System: ext4, Screen Resolution: 3840x2160 -O3 -march=x86-64: Processor: AMD Ryzen 9 5950X 16-Core @ 3.40GHz (16 Cores / 32 Threads), Motherboard: ASUS ROG CROSSHAIR VIII HERO (WI-FI) (3003 BIOS), Chipset: AMD Starship/Matisse, Memory: 16GB, Disk: 2000GB Corsair Force MP600, Graphics: AMD Radeon RX 5600 OEM/5600 XT / 5700/5700 8GB (2100/875MHz), Audio: AMD Navi 10 HDMI Audio, Monitor: ASUS MG28U, Network: Realtek RTL8125 2.5GbE + Intel I211 + Intel Wi-Fi 6 AX200 OS: Ubuntu 20.04, Kernel: 5.10.0-051000rc6daily20201205-generic (x86_64) 20201204, Desktop: GNOME Shell 3.36.4, Display Server: X Server 1.20.8, Display Driver: modesetting 1.20.8, OpenGL: 4.6 Mesa 21.0.0-devel (git-1a53572 2020-12-09 focal-oibaf-ppa) (LLVM 11.0.0), Vulkan: 1.2.145, Compiler: GCC 11.0.0 20201213, File-System: ext4, Screen Resolution: 3840x2160 -O3 -maech=znver3 -flto: Processor: AMD Ryzen 9 5950X 16-Core @ 3.40GHz (16 Cores / 32 Threads), Motherboard: ASUS ROG CROSSHAIR VIII HERO (WI-FI) (3003 BIOS), Chipset: AMD Starship/Matisse, Memory: 16GB, Disk: 2000GB Corsair Force MP600, Graphics: AMD Radeon RX 5600 OEM/5600 XT / 5700/5700 8GB (2100/875MHz), Audio: AMD Navi 10 HDMI Audio, Monitor: ASUS MG28U, Network: Realtek RTL8125 2.5GbE + Intel I211 + Intel Wi-Fi 6 AX200 OS: Ubuntu 20.04, Kernel: 5.10.0-051000rc6daily20201205-generic (x86_64) 20201204, Desktop: GNOME Shell 3.36.4, Display Server: X Server 1.20.8, Display Driver: modesetting 1.20.8, OpenGL: 4.6 Mesa 21.0.0-devel (git-1a53572 2020-12-09 focal-oibaf-ppa) (LLVM 11.0.0), Vulkan: 1.2.145, Compiler: GCC 11.0.0 20201213, File-System: ext4, Screen Resolution: 3840x2160 -Ofast -march=znver3 -flto: Processor: AMD Ryzen 9 5950X 16-Core @ 3.40GHz (16 Cores / 32 Threads), Motherboard: ASUS ROG CROSSHAIR VIII HERO (WI-FI) (3003 BIOS), Chipset: AMD Starship/Matisse, Memory: 16GB, Disk: 2000GB Corsair Force MP600, Graphics: AMD Radeon RX 5600 OEM/5600 XT / 5700/5700 8GB (2100/875MHz), Audio: AMD Navi 10 HDMI Audio, Monitor: ASUS MG28U, Network: Realtek RTL8125 2.5GbE + Intel I211 + Intel Wi-Fi 6 AX200 OS: Ubuntu 20.04, Kernel: 5.10.0-051000rc6daily20201205-generic (x86_64) 20201204, Desktop: GNOME Shell 3.36.4, Display Server: X Server 1.20.8, Display Driver: modesetting 1.20.8, OpenGL: 4.6 Mesa 21.0.0-devel (git-1a53572 2020-12-09 focal-oibaf-ppa) (LLVM 11.0.0), Vulkan: 1.2.145, Compiler: GCC 11.0.0 20201213, File-System: ext4, Screen Resolution: 3840x2160 -O2 -march=znver3 -flto: Processor: AMD Ryzen 9 5950X 16-Core @ 3.40GHz (16 Cores / 32 Threads), Motherboard: ASUS ROG CROSSHAIR VIII HERO (WI-FI) (3003 BIOS), Chipset: AMD Starship/Matisse, Memory: 16GB, Disk: 2000GB Corsair Force MP600, Graphics: AMD Radeon RX 5600 OEM/5600 XT / 5700/5700 8GB (2100/875MHz), Audio: AMD Navi 10 HDMI Audio, Monitor: ASUS MG28U, Network: Realtek RTL8125 2.5GbE + Intel I211 + Intel Wi-Fi 6 AX200 OS: Ubuntu 20.04, Kernel: 5.10.0-051000rc6daily20201205-generic (x86_64) 20201204, Desktop: GNOME Shell 3.36.4, Display Server: X Server 1.20.8, Display Driver: modesetting 1.20.8, OpenGL: 4.6 Mesa 21.0.0-devel (git-1a53572 2020-12-09 focal-oibaf-ppa) (LLVM 11.0.0), Vulkan: 1.2.145, Compiler: GCC 11.0.0 20201213, File-System: ext4, Screen Resolution: 3840x2160 -O2 -march=znver3: Processor: AMD Ryzen 9 5950X 16-Core @ 3.40GHz (16 Cores / 32 Threads), Motherboard: ASUS ROG CROSSHAIR VIII HERO (WI-FI) (3003 BIOS), Chipset: AMD Starship/Matisse, Memory: 16GB, Disk: 2000GB Corsair Force MP600, Graphics: AMD Radeon RX 5600 OEM/5600 XT / 5700/5700 8GB (2100/875MHz), Audio: AMD Navi 10 HDMI Audio, Monitor: ASUS MG28U, Network: Realtek RTL8125 2.5GbE + Intel I211 + Intel Wi-Fi 6 AX200 OS: Ubuntu 20.04, Kernel: 5.10.0-051000rc6daily20201205-generic (x86_64) 20201204, Desktop: GNOME Shell 3.36.4, Display Server: X Server 1.20.8, Display Driver: modesetting 1.20.8, OpenGL: 4.6 Mesa 21.0.0-devel (git-1a53572 2020-12-09 focal-oibaf-ppa) (LLVM 11.0.0), Vulkan: 1.2.145, Compiler: GCC 11.0.0 20201213, File-System: ext4, Screen Resolution: 3840x2160 -O2 -march=x86-64: Processor: AMD Ryzen 9 5950X 16-Core @ 3.40GHz (16 Cores / 32 Threads), Motherboard: ASUS ROG CROSSHAIR VIII HERO (WI-FI) (3003 BIOS), Chipset: AMD Starship/Matisse, Memory: 16GB, Disk: 2000GB Corsair Force MP600, Graphics: AMD Radeon RX 5600 OEM/5600 XT / 5700/5700 8GB (2100/875MHz), Audio: AMD Navi 10 HDMI Audio, Monitor: ASUS MG28U, Network: Realtek RTL8125 2.5GbE + Intel I211 + Intel Wi-Fi 6 AX200 OS: Ubuntu 20.04, Kernel: 5.10.0-051000rc6daily20201205-generic (x86_64) 20201204, Desktop: GNOME Shell 3.36.4, Display Server: X Server 1.20.8, Display Driver: modesetting 1.20.8, OpenGL: 4.6 Mesa 21.0.0-devel (git-1a53572 2020-12-09 focal-oibaf-ppa) (LLVM 11.0.0), Vulkan: 1.2.145, Compiler: GCC 11.0.0 20201213, File-System: ext4, Screen Resolution: 3840x2160 Timed ImageMagick Compilation 6.9.0 Time To Compile Seconds < Lower Is Better -O3 -march=znver3 .......... 18.86 |============ -O3 -march=znver2 .......... 19.37 |============ -O3 -march=x86-64 .......... 19.00 |============ -O3 -maech=znver3 -flto .... 63.48 |========================================= -Ofast -march=znver3 -flto . 66.84 |=========================================== -O2 -march=znver3 -flto .... 50.41 |================================ -O2 -march=znver3 .......... 15.80 |========== -O2 -march=x86-64 .......... 16.40 |=========== C-Ray 1.1 Total Time - 4K, 16 Rays Per Pixel Seconds < Lower Is Better -O3 -march=znver3 .......... 25.25 |================== -O3 -march=znver2 .......... 25.25 |================== -O3 -march=x86-64 .......... 31.36 |====================== -O3 -maech=znver3 -flto .... 25.48 |================== -Ofast -march=znver3 -flto . 27.34 |=================== -O2 -march=znver3 -flto .... 46.42 |================================= -O2 -march=znver3 .......... 46.72 |================================= -O2 -march=x86-64 .......... 60.90 |=========================================== Timed MrBayes Analysis 3.2.7 Primate Phylogeny Analysis Seconds < Lower Is Better -O3 -march=znver3 .......... 76.32 |========================================= -O3 -march=znver2 .......... 75.70 |========================================= -O3 -march=x86-64 .......... 77.07 |========================================== -O3 -maech=znver3 -flto .... 73.92 |======================================== -Ofast -march=znver3 -flto . 45.27 |======================== -O2 -march=znver3 -flto .... 79.51 |=========================================== -O2 -march=znver3 .......... 79.54 |=========================================== -O2 -march=x86-64 .......... 79.34 |=========================================== GraphicsMagick 1.3.33 Operation: Sharpen Iterations Per Minute > Higher Is Better -O3 -march=znver3 .......... 361 |=========================================== -O3 -march=znver2 .......... 377 |============================================= -O3 -march=x86-64 .......... 225 |=========================== -O3 -maech=znver3 -flto .... 370 |============================================ -Ofast -march=znver3 -flto . 328 |======================================= -O2 -march=znver3 -flto .... 377 |============================================= -O2 -march=znver3 .......... 359 |=========================================== -O2 -march=x86-64 .......... 224 |=========================== SciMark 2.0 Computational Test: Composite Mflops > Higher Is Better -O3 -march=znver3 .......... 3958.17 |===================================== -O3 -march=znver2 .......... 4407.32 |========================================= -O3 -march=x86-64 .......... 3660.74 |================================== -O3 -maech=znver3 -flto .... 3855.30 |==================================== -Ofast -march=znver3 -flto . 4101.89 |====================================== -O2 -march=znver3 -flto .... 2924.23 |=========================== -O2 -march=znver3 .......... 2625.59 |======================== -O2 -march=x86-64 .......... 2883.78 |=========================== LibRaw 0.20 Post-Processing Benchmark Mpix/sec > Higher Is Better -O3 -march=znver3 .......... 73.83 |========================================== -O3 -march=znver2 .......... 75.26 |=========================================== -O3 -march=x86-64 .......... 63.90 |===================================== -O3 -maech=znver3 -flto .... 73.73 |========================================== -Ofast -march=znver3 -flto . 75.23 |=========================================== -O2 -march=znver3 -flto .... 49.32 |============================ -O2 -march=znver3 .......... 50.95 |============================= -O2 -march=x86-64 .......... 51.45 |============================= GraphicsMagick 1.3.33 Operation: Resizing Iterations Per Minute > Higher Is Better -O3 -march=znver3 .......... 2135 |============================================ -O3 -march=znver2 .......... 1907 |======================================= -O3 -march=x86-64 .......... 1721 |=================================== -O3 -maech=znver3 -flto .... 1906 |======================================= -Ofast -march=znver3 -flto . 1901 |======================================= -O2 -march=znver3 -flto .... 1854 |====================================== -O2 -march=znver3 .......... 2150 |============================================ -O2 -march=x86-64 .......... 1605 |================================= FFTW 3.3.6 Build: Stock - Size: 1D FFT Size 4096 Mflops > Higher Is Better -O3 -march=znver3 .......... 13702.0 |==================================== -O3 -march=znver2 .......... 14096.0 |===================================== -O3 -march=x86-64 .......... 12436.0 |================================= -O3 -maech=znver3 -flto .... 15584.0 |========================================= -Ofast -march=znver3 -flto . 12149.0 |================================ -O2 -march=znver3 -flto .... 15232.0 |======================================== -O2 -march=znver3 .......... 13771.0 |==================================== -O2 -march=x86-64 .......... 12892.0 |================================== GraphicsMagick 1.3.33 Operation: Enhanced Iterations Per Minute > Higher Is Better -O3 -march=znver3 .......... 432 |========================================= -O3 -march=znver2 .......... 444 |========================================== -O3 -march=x86-64 .......... 382 |==================================== -O3 -maech=znver3 -flto .... 445 |========================================== -Ofast -march=znver3 -flto . 449 |========================================== -O2 -march=znver3 -flto .... 476 |============================================= -O2 -march=znver3 .......... 430 |========================================= -O2 -march=x86-64 .......... 419 |======================================== FFTW 3.3.6 Build: Stock - Size: 2D FFT Size 4096 Mflops > Higher Is Better -O3 -march=znver3 .......... 9314.1 |====================================== -O3 -march=znver2 .......... 8927.1 |==================================== -O3 -march=x86-64 .......... 8169.2 |================================= -O3 -maech=znver3 -flto .... 10125.0 |========================================= -Ofast -march=znver3 -flto . 8755.9 |=================================== -O2 -march=znver3 -flto .... 9718.2 |======================================= -O2 -march=znver3 .......... 8823.7 |==================================== -O2 -march=x86-64 .......... 8352.8 |================================== ACES DGEMM 1.0 Sustained Floating-Point Rate GFLOP/s > Higher Is Better -O3 -march=znver3 .......... 7.307093 |====================================== -O3 -march=znver2 .......... 7.512189 |======================================= -O3 -march=x86-64 .......... 6.484557 |================================= -O3 -maech=znver3 -flto .... 7.696089 |======================================== -Ofast -march=znver3 -flto . 7.758952 |======================================== -O2 -march=znver3 -flto .... 7.345638 |====================================== -O2 -march=znver3 .......... 7.391742 |====================================== -O2 -march=x86-64 .......... 7.526567 |======================================= GraphicsMagick 1.3.33 Operation: Noise-Gaussian Iterations Per Minute > Higher Is Better -O3 -march=znver3 .......... 441 |======================================= -O3 -march=znver2 .......... 434 |======================================= -O3 -march=x86-64 .......... 426 |====================================== -O3 -maech=znver3 -flto .... 430 |====================================== -Ofast -march=znver3 -flto . 504 |============================================= -O2 -march=znver3 -flto .... 445 |======================================== -O2 -march=znver3 .......... 452 |======================================== -O2 -march=x86-64 .......... 429 |====================================== Kvazaar 2.0 Video Input: Bosphorus 1080p - Video Preset: Ultra Fast Frames Per Second > Higher Is Better -O3 -march=znver3 .......... 200.12 |====================================== -O3 -march=znver2 .......... 199.35 |====================================== -O3 -march=x86-64 .......... 198.84 |====================================== -O3 -maech=znver3 -flto .... 203.89 |======================================= -Ofast -march=znver3 -flto . 220.99 |========================================== -O2 -march=znver3 -flto .... 201.85 |====================================== -O2 -march=znver3 .......... 195.13 |===================================== -O2 -march=x86-64 .......... 194.52 |===================================== Kvazaar 2.0 Video Input: Bosphorus 1080p - Video Preset: Very Fast Frames Per Second > Higher Is Better -O3 -march=znver3 .......... 111.50 |====================================== -O3 -march=znver2 .......... 111.54 |====================================== -O3 -march=x86-64 .......... 111.77 |======================================= -O3 -maech=znver3 -flto .... 116.07 |======================================== -Ofast -march=znver3 -flto . 121.74 |========================================== -O2 -march=znver3 -flto .... 113.41 |======================================= -O2 -march=znver3 .......... 109.42 |====================================== -O2 -march=x86-64 .......... 108.72 |====================================== Timed HMMer Search 3.3.1 Pfam Database Search Seconds < Lower Is Better -O3 -march=znver3 .......... 85.02 |=========================================== -O3 -march=znver2 .......... 84.79 |========================================== -O3 -march=x86-64 .......... 85.60 |=========================================== -O3 -maech=znver3 -flto .... 84.12 |========================================== -Ofast -march=znver3 -flto . 76.94 |======================================= -O2 -march=znver3 -flto .... 84.93 |=========================================== -O2 -march=znver3 .......... 85.13 |=========================================== -O2 -march=x86-64 .......... 85.79 |=========================================== Smallpt 1.0 Global Illumination Renderer; 128 Samples Seconds < Lower Is Better -O3 -march=znver3 .......... 4.673 |======================================== -O3 -march=znver2 .......... 4.677 |======================================== -O3 -march=x86-64 .......... 5.068 |=========================================== -O3 -maech=znver3 -flto .... 4.656 |======================================== -Ofast -march=znver3 -flto . 4.602 |======================================= -O2 -march=znver3 -flto .... 4.607 |======================================= -O2 -march=znver3 .......... 4.567 |======================================= -O2 -march=x86-64 .......... 4.829 |========================================= Kvazaar 2.0 Video Input: Bosphorus 4K - Video Preset: Ultra Fast Frames Per Second > Higher Is Better -O3 -march=znver3 .......... 54.73 |======================================== -O3 -march=znver2 .......... 54.75 |======================================== -O3 -march=x86-64 .......... 54.43 |======================================== -O3 -maech=znver3 -flto .... 56.12 |========================================= -Ofast -march=znver3 -flto . 58.69 |=========================================== -O2 -march=znver3 -flto .... 54.91 |======================================== -O2 -march=znver3 .......... 53.83 |======================================= -O2 -march=x86-64 .......... 53.13 |======================================= Coremark 1.0 CoreMark Size 666 - Iterations Per Second Iterations/Sec > Higher Is Better -O3 -march=znver3 .......... 716026.62 |====================================== -O3 -march=znver2 .......... 721620.07 |====================================== -O3 -march=x86-64 .......... 717480.24 |====================================== -O3 -maech=znver3 -flto .... 736850.43 |======================================= -Ofast -march=znver3 -flto . 728279.79 |======================================= -O2 -march=znver3 -flto .... 681281.71 |==================================== -O2 -march=znver3 .......... 669924.92 |=================================== -O2 -march=x86-64 .......... 700520.67 |===================================== Kvazaar 2.0 Video Input: Bosphorus 4K - Video Preset: Very Fast Frames Per Second > Higher Is Better -O3 -march=znver3 .......... 30.31 |======================================== -O3 -march=znver2 .......... 30.34 |======================================== -O3 -march=x86-64 .......... 30.48 |======================================== -O3 -maech=znver3 -flto .... 31.44 |========================================== -Ofast -march=znver3 -flto . 32.40 |=========================================== -O2 -march=znver3 -flto .... 30.76 |========================================= -O2 -march=znver3 .......... 29.87 |======================================== -O2 -march=x86-64 .......... 29.55 |======================================= AOM AV1 2.0 Encoder Mode: Speed 8 Realtime Frames Per Second > Higher Is Better -O3 -march=znver3 .......... 62.33 |========================================== -O3 -march=znver2 .......... 62.28 |========================================== -O3 -march=x86-64 .......... 60.07 |========================================= -O3 -maech=znver3 -flto .... 62.52 |========================================== -Ofast -march=znver3 -flto . 63.76 |=========================================== -O2 -march=znver3 -flto .... 62.08 |========================================== -O2 -march=znver3 .......... 62.10 |========================================== -O2 -march=x86-64 .......... 60.88 |========================================= Kvazaar 2.0 Video Input: Bosphorus 4K - Video Preset: Slow Frames Per Second > Higher Is Better -O3 -march=znver3 .......... 11.58 |========================================= -O3 -march=znver2 .......... 11.59 |========================================= -O3 -march=x86-64 .......... 11.69 |========================================= -O3 -maech=znver3 -flto .... 11.88 |========================================== -Ofast -march=znver3 -flto . 12.12 |=========================================== -O2 -march=znver3 -flto .... 11.68 |========================================= -O2 -march=znver3 .......... 11.52 |========================================= -O2 -march=x86-64 .......... 11.48 |========================================= Darmstadt Automotive Parallel Heterogeneous Suite Backend: OpenMP - Kernel: NDT Mapping Test Cases Per Minute > Higher Is Better -O3 -march=znver3 .......... 971.31 |========================================== -O3 -march=znver2 .......... 964.72 |========================================== -O3 -march=x86-64 .......... 931.03 |======================================== -O3 -maech=znver3 -flto .... 971.45 |========================================== -Ofast -march=znver3 -flto . 969.91 |========================================== -O2 -march=znver3 -flto .... 970.46 |========================================== -O2 -march=znver3 .......... 966.99 |========================================== -O2 -march=x86-64 .......... 921.71 |======================================== Kvazaar 2.0 Video Input: Bosphorus 1080p - Video Preset: Slow Frames Per Second > Higher Is Better -O3 -march=znver3 .......... 44.91 |========================================= -O3 -march=znver2 .......... 44.97 |========================================= -O3 -march=x86-64 .......... 45.75 |========================================== -O3 -maech=znver3 -flto .... 46.30 |========================================== -Ofast -march=znver3 -flto . 47.25 |=========================================== -O2 -march=znver3 -flto .... 45.68 |========================================== -O2 -march=znver3 .......... 44.89 |========================================= -O2 -march=x86-64 .......... 45.09 |========================================= Kvazaar 2.0 Video Input: Bosphorus 4K - Video Preset: Medium Frames Per Second > Higher Is Better -O3 -march=znver3 .......... 11.74 |========================================= -O3 -march=znver2 .......... 11.72 |========================================= -O3 -march=x86-64 .......... 11.86 |========================================== -O3 -maech=znver3 -flto .... 12.03 |========================================== -Ofast -march=znver3 -flto . 12.27 |=========================================== -O2 -march=znver3 -flto .... 11.84 |========================================= -O2 -march=znver3 .......... 11.73 |========================================= -O2 -march=x86-64 .......... 11.68 |========================================= Kvazaar 2.0 Video Input: Bosphorus 1080p - Video Preset: Medium Frames Per Second > Higher Is Better -O3 -march=znver3 .......... 46.42 |========================================= -O3 -march=znver2 .......... 46.38 |========================================= -O3 -march=x86-64 .......... 46.74 |========================================= -O3 -maech=znver3 -flto .... 47.32 |========================================== -Ofast -march=znver3 -flto . 48.54 |=========================================== -O2 -march=znver3 -flto .... 47.04 |========================================== -O2 -march=znver3 .......... 46.31 |========================================= -O2 -march=x86-64 .......... 46.32 |========================================= AOBench Size: 2048 x 2048 - Total Time Seconds < Lower Is Better -O3 -march=znver3 .......... 25.07 |=============================== -O3 -march=znver2 .......... 25.02 |=============================== -O3 -march=x86-64 .......... 29.99 |===================================== -O3 -maech=znver3 -flto .... 25.85 |================================ -Ofast -march=znver3 -flto . 24.51 |============================== -O2 -march=znver3 -flto .... 27.36 |================================== -O2 -march=znver3 .......... 30.71 |====================================== -O2 -march=x86-64 .......... 35.04 |===========================================