HEART 2024 Memory Centric Tutorial

User Tools

Site Tools

Trace:
start

Differences

This shows you the differences between two versions of the page.

Link to this comparison view

Both sides previous revisionPrevious revision
Next revision		Previous revision
start [2024/04/02 08:35] – geraldodstart [2024/06/20 10:48] (current) – [Tutorial Materials] geraldod
Line 1: Line 1:
 ~~NOCACHE~~ ~~NOCACHE~~
  
-===== ISCA 2024 Tutorial on Memory-Centric Computing Systems (Half Day) =====+===== HEART 2024 Tutorial on Memory-Centric Computing Systems (Half Day) =====
  
 ==== Tutorial Description ==== ==== Tutorial Description ====
Line 12: Line 12:
 Recent PIM products and prototypes place compute units near the memory arrays. New memory interfaces like CXL (Compute Express Link) aid the enablement of compute-capable memories. At the same time, academia and industry are actively exploring other types of PIM by, e.g., exploiting the analog operation of DRAM, SRAM, flash memory, and emerging non-volatile memories, and hybrid PIM architectures that combine processing capabilities of different types and at different parts of the memory/storage hierarchy.  Recent PIM products and prototypes place compute units near the memory arrays. New memory interfaces like CXL (Compute Express Link) aid the enablement of compute-capable memories. At the same time, academia and industry are actively exploring other types of PIM by, e.g., exploiting the analog operation of DRAM, SRAM, flash memory, and emerging non-volatile memories, and hybrid PIM architectures that combine processing capabilities of different types and at different parts of the memory/storage hierarchy. 
    
-{{:memory_centric_comp_banner.jpeg?400 |}}+{{:heart24_banner.jpeg?400 |}}
  
 PIM can improve performance and energy efficiency for many modern applications, enabling a commercially viable way of dealing with huge amounts of data bottlenecking our computing systems, which is especially exacerbated by workloads like AI/ML and genomics. In fact, workloads like large language model training and inference can potentially be “killer applications'' for PIM.  PIM can improve performance and energy efficiency for many modern applications, enabling a commercially viable way of dealing with huge amounts of data bottlenecking our computing systems, which is especially exacerbated by workloads like AI/ML and genomics. In fact, workloads like large language model training and inference can potentially be “killer applications'' for PIM. 
Line 18: Line 18:
 However, there are many open questions spanning the entire computing stack and many challenges for widespread adoption. For example, it is critical to (1) develop programming frameworks and tools that can lower the learning curve and ease the adoption of PIM systems, (2) develop methods to identify what type of PIM would be useful for what workload, and (3) design system and security mechanisms that enable PIM in a wider scale. Implications of PIM on all aspects of computing systems and workloads is a challenging and exciting field of study.   However, there are many open questions spanning the entire computing stack and many challenges for widespread adoption. For example, it is critical to (1) develop programming frameworks and tools that can lower the learning curve and ease the adoption of PIM systems, (2) develop methods to identify what type of PIM would be useful for what workload, and (3) design system and security mechanisms that enable PIM in a wider scale. Implications of PIM on all aspects of computing systems and workloads is a challenging and exciting field of study.  
  
-This tutorial focuses on the latest advances in PIM technology, spanning both hardware and software, including novel PIM ideas, different tools and frameworks to conduct PIM research, and programming techniques and optimization strategies for  PIM kernels. We will (1) provide an introduction to PIM and the taxonomy of PIM systems, (2) give an overview and a rigorous analysis of existing PIM hardware from industry and academia, (3) provide and describe hardware and software infrastructures that can enable new and experienced researchers to conduct research in PIM systems, and (4) shed light on how to improve future PIM systems for emerging memory-bound workloads. The tutorial will also incorporate invited talks from leading industry and academic researchers in PIM systems.  +This tutorial focuses on the latest advances in PIM technology, spanning both hardware and software, including novel PIM ideas, different tools and frameworks to conduct PIM research, and programming techniques and optimization strategies for  PIM kernels. We will (1) provide an introduction to PIM and the taxonomy of PIM systems, (2) give an overview and a rigorous analysis of existing PIM hardware from industry and academia, (3) provide and describe hardware and software infrastructures that can enable new and experienced researchers to conduct research in PIM systems, and (4) shed light on how to improve future PIM systems for emerging memory-bound workloads.
  
 ==== Livestream ==== ==== Livestream ====
-[[|Zoom livestream]] +[[https://www.youtube.com/live/0YKI2pVmrik |YouTube livestream]] 
 +{{youtube>0YKI2pVmrik?large}}
  
 ==== Organizers ==== ==== Organizers ====
Line 32: Line 31:
 |[[https://people.inf.ethz.ch/omutlu/index.html|Professor Onur Mutlu]]| <onur.mutlu@safari.ethz.ch> |  |[[https://people.inf.ethz.ch/omutlu/index.html|Professor Onur Mutlu]]| <onur.mutlu@safari.ethz.ch> | 
  
-===== Agenda ===== +===== Agenda (June 21, 2024) ===== 
-==== Lectures (tentative schedule) ====+==== Lectures (tentative schedule, time zone: UTC+1) ==== 
 +  * 10:30 am-11:00 am, Prof. Onur Mutlu / Geraldo F. Oliveira, "Memory-Centric Computing: Introduction to PIM as a Paradigm to Overcome the Data Movement Bottleneck." 
 +    * Introduction: PIM as a paradigm to overcome the data movement bottleneck. Workload analysis and system bottlenecks. 
 +    * PIM taxonomy: technology, location, and nature of computation (e.g., PNM (processing-near-memory) and PUM (processing-using-memory).  
 +    * Advances in different types of PIM at different parts of the memory/storage systems. 
 +    * Research challenges and opportunities in PIM systems, with a focus on enabling adoption in the real world. 
 +  * 11:00 am-11:20 am, Geraldo F. Oliveira, "Real-World PNM Systems." 
 +    * Example real-world PNM systems: UPMEM PIM, Samsung HBM-PIM & CXL-PNM, SK Hynix AiM & CMS 2.0, Samsung AxDIMM, Alibaba PNM, Mythic. 
 +  * 11:30 am-12:00 pm, Geraldo F. Oliveira, "PUM Systems for Bulk Bitwise Operations." 
 +    * PUM systems for bulk bitwise operations in simulated and off-the-shelf memory technologies (DRAM, SRAM, and NVM).
  
-  - Introduction: PIM as a paradigm to overcome the data movement bottleneck.  +    * 12:30 pm-02:00pm, Lunch break. 
-  - Workload analysis and system bottlenecks. + 
-  - PIM taxonomy: technology, location, and nature of computation (e.g., PNM (processing-near-memory) and PUM (processing-using-memory).  +  * 02:00 pm-02:30 pm, Geraldo F. Olveira, "PIM Programming & Infrastructure for PIM Research." 
-  - Advances in different types of PIM at different parts of the memory/storage systems. +    * Programming techniques and tools for PIM systems. 
-  - Example real-world PNM systems: UPMEM PIM, Samsung HBM-PIM & CXL-PNM, SK Hynix AiM & CMS 2.0, Samsung AxDIMM, Alibaba PNM, Mythic. +    * Infrastructures for doing PIM Research (simulation, real systems, FPGA prototypes).  
-  - PUM systems for bulk bitwise operations in simulated and off-the-shelf memory technologies (DRAM, SRAM, and NVM). +  * 02:30 pm-03:00 pm, Geraldo F. Oliveira, “Introduction/Preparation for Hands-on Labs.” 
-  - Programming techniques and tools for PIM systems. +    * Optional - Hands-on Lab: Programming and Understanding a Real PIM Architecture.
-  - Infrastructures for doing PIM Research (simulation, real systems, FPGA prototypes).  +
-  - Research challenges and opportunities in PIM systems, with a focus on enabling adoption in the real world.+
  
 ==== Tutorial Materials ==== ==== Tutorial Materials ====
 +
  
 ^ Time ^ Speaker ^ Title ^ Materials ^ ^ Time ^ Speaker ^ Title ^ Materials ^
-| TBA  | TBA | TBA | TBA |+| 10:30am-11:00am  | Prof. Onur Mutlu / Geraldo F. Oliveira | Memory-Centric Computing |{{geraldo-heart24-lecture1-memory-centric-computing-beforelecture.pdf|(PDF)}} {{geraldo-heart24-lecture1-memory-centric-computing-beforelecture.pptx|(PPT)}}| 
 +| 11:00am-11:20am  | Geraldo F. Oliveira | Real-World PNM Systems |{{geraldo-heart24-lecture2-processing-near-memory-beforelecture.pdf|(PDF)}} {{geraldo-heart24-lecture2-processing-near-memory-beforelecture.pptx|(PPT)}}| 
 +| 11:30am-12:00pm  | Geraldo F. Oliveira | PUM Systems for Bulk Bitwise Operations |{{geraldo-heart24-lecture3-processing-using-memory-beforelecture.pdf|(PDF)}} {{geraldo-heart24-lecture3-processing-using-memory-beforelecture.pptx|(PPT)}}| 
 +| 12:30pm-02:00pm  | | Lunch Break| 
 +| 02:00pm-02:30pm  | Geraldo F. Oliveira | PIM Programming & Infrastructure for PIM Research |{{geraldo-heart24-lecture4-adoption-programmability-beforelecture.pdf|(PDF)}} {{geraldo-heart24-lecture4-adoption-programmability-beforelecture.pptx|(PPT)}}| 
 +| 02:30pm-03:30pm  | Geraldo F. Oliveira| Hands-on Lab: Programming and Understanding a Real Processing-in-Memory Architecture |{{heart_2024___pim_tutorial_handout.pdf|(Handout)}} \\ {{geraldo-heart24-processingupmem.pdf|(PDF)}} {{geraldo-heart24-processingupmem.pptx|(PPT)}}| 
  
 ==== Learning Materials ==== ==== Learning Materials ====
Line 75: Line 87:
     * [[https://arxiv.org/pdf/2402.19080.pdf | PDF (arXiv)]]      * [[https://arxiv.org/pdf/2402.19080.pdf | PDF (arXiv)]] 
     * [[https://github.com/CMU-SAFARI/MIMDRAM | Repository (GitHub)]]      * [[https://github.com/CMU-SAFARI/MIMDRAM | Repository (GitHub)]] 
 +  * Hajinazar, N., Oliveira, G. F., Gregorio, S., Ferreira, J. D., Ghiasi, N. M., Patel, M., Alser, M., Ghose, S., Gomez-Luna, J., Mutlu. O., SIMDRAM: An End-to-End Framework for Bit-Serial SIMD Computing in DRAM, in ASPLOS, 2021.
 +    * [[https://arxiv.org/pdf/2105.12839.pdf | PDF (arXiv)]] 
 +    * [[https://www.youtube.com/watch?v=lu3Br4-kySw | Full Talk Video]] 
 +  * Seshadri, V., Lee, D., Mullins, T., Hassan, H., Boroumand, A., Kim, J., Kozuch, M. A., Mutlu, O., Gibbons, P. B., Mowry, T. C., Ambit: In-Memory Accelerator for Bulk Bitwise Operations Using Commodity DRAM Technology, in MICRO, 2017.
 +    * [[https://people.inf.ethz.ch/omutlu/pub/ambit-bulk-bitwise-dram_micro17.pdf | PDF]] 
  
  
start.1712046925.txt.gz · Last modified: 2024/05/22 12:44 (external edit)
Donate Powered by PHP Valid HTML5 Valid CSS Driven by DokuWiki