Full Site Navigation

Analyzing and Optimizing* Speed**

* Improving, but probably not optimizing
** And boot time and power use

Class #	Topic	Slides and Notes	Assigned Readings and Videos	Supplementary Information
Performance Analysis and Scalar Optimization
19 3/20/25	Memory System	Memory Systems		Drongowski: A short trip through ARM cores Raspberry Pi 4 ARM Cortex-A72 Processor ARM Cortex-A72 tuning: Memory access
20 3/25/25	Performance Analysis	Why So Fast? scans Linux – Performance Analysis (3/27/25)	Why is the RPi4 So Much Faster than the Cortex-M0+? (16:13) Profiling (47:00)	Speed/Scalar/SG1, SG2, SG_gprof DEN0013D_cortex_a_series_programmers_guide (Ch. 16) Drongowski: Performance events on Raspberry Pi 4: Tips
20 3/25/25	Key ARM ISA Features for Performance	Key ARM ISA Features for Performance (updated 3/25/25)	ARM ISA Features for Performance (part a) (0:57:20) ARM ISA Features for Performance (part b) (0:17:34)	Speed/Scalar/compiler DEN0013D_cortex_a_series_programmers_guide (Ch. 1-5, Appendix A) https://en.wikipedia.org/wiki/Comparison_of_ARMv8-A_cores DDI0406B_ARMv7_architecture_reference_ manual_errata_markup_7_0 Chapter A4: The Instruction Sets ARMv8A Arm 64 Quick Reference Chart Armv8-A Instruction Set Architecture
21 3/27/25	Cortex-A72 Microarchitecture	Cortex-A72 Microarchitecture	Cortex-A72 Microarchitecture (0:41:55)	DEN0013D_cortex_a_series_programmers_guide (Ch. 17) ARM C Language Extensions GCC C Language Extensions Inline Assembly Language with GCC Ethernuts ARM GCC Inline Assembler Cookbook
22 4/1/25	Getting Better Object Code Optimizing for the Cortex-A72 Microarchitecture and Memory System	Speed – Getting Better Object Code Optimizing for the Cortex-A72 Optimizing for the Memory System	Performance 3: Getting Better Object Code (0:55:33) Optimizing for the Cortex-A72 (54:21) ERP Ch. 3, 4 Read Cortex-A PG Ch.8, 17	Analysis of A64 do_test1 object with CSEL instruction: CSEL example do_test1 Cortex-A72 Analysis Drongowski ARM Cortex-A72 fetch and branch processing ARM Cortex-A72 tuning: Branch mispredictions Anandtech Tom’s Hardware Cortex-A72 Cortex-A72 MPCore Technical Reference Manual Cortex-A72 software optimization guide, Ch. 1, 2, 3 (3.1-3.20), 4 Arm_cortex_a55_software_optimization_guide_v2 DEN0024A_v8_Arch. Programmers Guide ARMv7-A DDI0406B_ARMv7_Arch Ref Manual Chapter A4: The Instruction Sets DEN0013D_cortex_a_series_programmers_guide (Ch. 8,17)
22 4/1/25	Examining SG1 and SG2 Performance on the Cortex-A72 and Cortex-A53 Comparing SG on Arm32 and Arm64	Optimizing SG slides (4/4/24) SG Optimization spreadsheet (4/2/24) Whiteboard (3/30/23) Cortex-A72 Processor Cortex-A8 Pipelines	SG1 on A72 SG1-1 Introduction, v1 (15:14) SG1-2 Using perf, v1-3 (15:05) SG1-3 More optimizations, v4-9 (4:32) SG2 on A72 SG2-1 Even more optimizations, v11-13 (6:57)	Speed/Scalar/SG1 Speed/Scalar/SG2 Drongowski Perf Tutorial Part 1: Introduction & Hot Spots Part 2: Counting Hardware Events Part 3: Profiling Hardware Events

…

Class #	Topic	Slides and Notes	Assigned Readings and Videos	Supplementary Information
SIMD Optimization
23 4/3/25	Introduction	Introducing NEON Advanced SIMD NEON Advanced SIMD Instructions	Read NEON PG, Ch.1 – 3 Introducing NEON (22:00) Using NEON 1: Documentation, approaches, libraries, generic compiler opts. pp. 1-13. (11:41) Using NEON 2: Loop unrolling example. pp. 14-21. (30:15)	Learn the Architecture – A64 ISA Guide Learn the architecture – Introducing Neon Learn the Architecture – Optimizing C code with Neon Intrinsics (2.1) Armv8 Neon Programmer’s Guide (pdf) Arm Neon Intrinsics Reference DEN0013D_cortex_a_series_programmers_guide (Ch. 7) (old but still useful)
24 4/8/25	Using NEON YUV Color Model, find_chroma_matches	Using NEON Advanced SIMD YUV and ISS slides	Read NEON PG, Ch. 4 – 6 Using NEON 3: Loop unrolling + vectorization details, helping the compiler. pp. 22-35. (23:07) Using NEON 4: Intrinsics. pp. 36-54. (31:58)	Speed/Vector/Neon0 Speed/Vector/Neon1 Learn the Architecture – Compiling for Neon with Auto-Vectorization
25 4/10/25	More on Optimization: Help from the Compiler, SG2, N-Body Simulation			Optimization information: Usability Improvements in GCC 9, -fopt-info Vectorization Optimization in GCC,
TBD	and Scalar Performance	YUV and ISS (24 L26) slides
TBD	Applying NEON: N-Body Gravitational Simulation			Learn the Architecture – Optimizing C code with Neon Intrinsics (2.1) Speed/Vector/NBody
TBD	Applying NEON: find_chroma_matches		Applying NEON Advanced SIMD to find_chroma_matches
N/A	Applying NEON: Fractals and the Mandelbrot Set			Speed/Vector/MandelTime

…

Class #	Topic	Slides and Notes	Assigned Readings and Videos	Supplementary Information
Other Analyses and Optimizations
N/A	RPi Power Basics	RPi Power Analysis Slides		RPi 3 RPi 4 RPi 4B Schematic – rpi_SCH_4b_4p0_reduced mxl7704 PMIC datasheet RPi 4 Firmware Updates Tested PMIC Replacement and Repair
	Basic Multicore Programming
	Examining and Improving Boot Time
	Examining and Improving Power Use