MKL-DNN GCC 9 Cascadelake Compiler Tuning
2 x Intel Xeon Platinum 8280 testing with a GIGABYTE MD61-SC2-00 v01000100 (T15 BIOS) and ASPEED Family on Ubuntu 18.04 via the Phoronix Test Suite.


-O3 -march=skylake-avx512: 

	Processor: 2 x Intel Xeon Platinum 8280 @ 4.00GHz (56 Cores / 112 Threads), Motherboard: GIGABYTE MD61-SC2-00 v01000100 (T15 BIOS), Chipset: Intel Sky Lake-E DMI3 Registers, Memory: 386048MB, Disk: Samsung SSD 970 PRO 512GB, Graphics: ASPEED Family, Monitor: VE228, Network: 2 x Intel X722 for 1GbE + 2 x QLogic FastLinQ QL41000 10/25/40/50GbE

	OS: Ubuntu 18.04, Kernel: 5.1.0-999-generic (x86_64) 20190416, Desktop: GNOME Shell 3.28.3, Display Server: X Server 1.20.1, Display Driver: modesetting 1.20.1, Compiler: GCC 9.0.1 20190414, File-System: ext4, Screen Resolution: 1920x1080

-O3 -march=cascadelake: 

	Processor: 2 x Intel Xeon Platinum 8280 @ 4.00GHz (56 Cores / 112 Threads), Motherboard: GIGABYTE MD61-SC2-00 v01000100 (T15 BIOS), Chipset: Intel Sky Lake-E DMI3 Registers, Memory: 386048MB, Disk: Samsung SSD 970 PRO 512GB, Graphics: ASPEED Family, Monitor: VE228, Network: 2 x Intel X722 for 1GbE + 2 x QLogic FastLinQ QL41000 10/25/40/50GbE

	OS: Ubuntu 18.04, Kernel: 5.1.0-999-generic (x86_64) 20190416, Desktop: GNOME Shell 3.28.3, Display Server: X Server 1.20.1, Display Driver: modesetting 1.20.1, Compiler: GCC 9.0.1 20190414, File-System: ext4, Screen Resolution: 1920x1080


MKL-DNN 2019-04-16
Harness: IP Batch 1D - Data Type: f32
ms < Lower Is Better
-O3 -march=skylake-avx512 . 17.55 |============================================
-O3 -march=cascadelake .... 17.44 |============================================


MKL-DNN 2019-04-16
Harness: IP Batch All - Data Type: f32
ms < Lower Is Better
-O3 -march=skylake-avx512 . 249 |==============================================
-O3 -march=cascadelake .... 249 |==============================================


MKL-DNN 2019-04-16
Harness: IP Batch 1D - Data Type: u8s8u8s32
ms < Lower Is Better
-O3 -march=skylake-avx512 . 4.59 |=============================================
-O3 -march=cascadelake .... 4.58 |=============================================


MKL-DNN 2019-04-16
Harness: IP Batch All - Data Type: u8s8u8s32
ms < Lower Is Better
-O3 -march=skylake-avx512 . 72.49 |============================================
-O3 -march=cascadelake .... 72.38 |============================================


MKL-DNN 2019-04-16
Harness: Convolution Batch conv_3d - Data Type: f32
ms < Lower Is Better
-O3 -march=skylake-avx512 . 85.67 |============================================
-O3 -march=cascadelake .... 85.40 |============================================


MKL-DNN 2019-04-16
Harness: Convolution Batch conv_all - Data Type: f32
ms < Lower Is Better
-O3 -march=skylake-avx512 . 13218 |============================================
-O3 -march=cascadelake .... 13215 |============================================


MKL-DNN 2019-04-16
Harness: Deconvolution Batch deconv_1d - Data Type: f32
ms < Lower Is Better
-O3 -march=skylake-avx512 . 23.77 |============================================
-O3 -march=cascadelake .... 23.53 |============================================


MKL-DNN 2019-04-16
Harness: Deconvolution Batch deconv_3d - Data Type: f32
ms < Lower Is Better
-O3 -march=skylake-avx512 . 27.41 |===========================================
-O3 -march=cascadelake .... 27.77 |============================================


MKL-DNN 2019-04-16
Harness: Convolution Batch conv_alexnet - Data Type: f32
ms < Lower Is Better
-O3 -march=skylake-avx512 . 1756 |=============================================
-O3 -march=cascadelake .... 1753 |=============================================


MKL-DNN 2019-04-16
Harness: Deconvolution Batch deconv_all - Data Type: f32
ms < Lower Is Better
-O3 -march=skylake-avx512 . 10271 |============================================
-O3 -march=cascadelake .... 10257 |============================================


MKL-DNN 2019-04-16
Harness: Convolution Batch conv_3d - Data Type: u8s8u8s32
ms < Lower Is Better
-O3 -march=skylake-avx512 . 125838 |==========================================
-O3 -march=cascadelake .... 128646 |===========================================


MKL-DNN 2019-04-16
Harness: Convolution Batch conv_all - Data Type: u8s8u8s32
ms < Lower Is Better
-O3 -march=skylake-avx512 . 59581 |============================================
-O3 -march=cascadelake .... 59454 |============================================


MKL-DNN 2019-04-16
Harness: Convolution Batch conv_googlenet_v3 - Data Type: f32
ms < Lower Is Better
-O3 -march=skylake-avx512 . 727 |==============================================
-O3 -march=cascadelake .... 727 |==============================================


MKL-DNN 2019-04-16
Harness: Deconvolution Batch deconv_1d - Data Type: u8s8u8s32
ms < Lower Is Better
-O3 -march=skylake-avx512 . 5.40 |=============================================
-O3 -march=cascadelake .... 5.41 |=============================================


MKL-DNN 2019-04-16
Harness: Deconvolution Batch deconv_3d - Data Type: u8s8u8s32
ms < Lower Is Better
-O3 -march=skylake-avx512 . 79735 |============================================
-O3 -march=cascadelake .... 79948 |============================================


MKL-DNN 2019-04-16
Harness: Convolution Batch conv_alexnet - Data Type: u8s8u8s32
ms < Lower Is Better
-O3 -march=skylake-avx512 . 452 |==============================================
-O3 -march=cascadelake .... 453 |==============================================


MKL-DNN 2019-04-16
Harness: Deconvolution Batch deconv_all - Data Type: u8s8u8s32
ms < Lower Is Better
-O3 -march=skylake-avx512 . 126456 |===========================================
-O3 -march=cascadelake .... 126386 |===========================================


MKL-DNN 2019-04-16
Harness: Convolution Batch conv_googlenet_v3 - Data Type: u8s8u8s32
ms < Lower Is Better
-O3 -march=skylake-avx512 . 177 |==============================================
-O3 -march=cascadelake .... 177 |==============================================