Difference between revisions of "AMX"

From wikieduonline
Jump to navigation Jump to search
Tags: Mobile web edit, Mobile edit
 
(One intermediate revision by the same user not shown)
Line 3: Line 3:
 
==Extensions==
 
==Extensions==
 
AMX was introduced by Intel in June 2020 and first supported by Intel with the [[Sapphire Rapids]] [[microarchitecture]] for [[Xeon]] servers, planned for 2021.<ref>{{Cite web|url=https://www.heise.de/news/Intel-AMX-Erste-Informationen-zur-Advanced-Matrix-Extensions-Architecture-4797415.html|title=Intel AMX: Erste Informationen zur Advanced Matrix Extensions Architecture|first=heise|last=online|website=heise online}}</ref><ref>{{Cite web|url=https://www.anandtech.com/show/16921/intel-sapphire-rapids-nextgen-xeon-scalable-gets-a-tiling-upgrade|title=Intel Xeon Sapphire Rapids: How To Go Monolithic with Tiles|first=Dr Ian|last=Cutress|website=[[AnandTech]]}}</ref> It introduced 2-dimensional [[processor register|registers]] called tiles upon which accelerators can perform operations. It is intended as an extensible architecture, the first accelerator implemented is called tile matrix multiply unit (TMUL).<ref>https://software.intel.com/content/dam/develop/public/us/en/documents/architecture-instruction-set-extensions-programming-reference.pdf</ref> <ref>{{Cite web|url=https://fuse.wikichip.org/news/3600/the-x86-advanced-matrix-extension-amx-brings-matrix-operations-to-debut-with-sapphire-rapids/|title=The x86 Advanced Matrix Extension (AMX) Brings Matrix Operations; To Debut with Sapphire Rapids|first=David|last=Schor|date=June 29, 2020}}</ref>
 
AMX was introduced by Intel in June 2020 and first supported by Intel with the [[Sapphire Rapids]] [[microarchitecture]] for [[Xeon]] servers, planned for 2021.<ref>{{Cite web|url=https://www.heise.de/news/Intel-AMX-Erste-Informationen-zur-Advanced-Matrix-Extensions-Architecture-4797415.html|title=Intel AMX: Erste Informationen zur Advanced Matrix Extensions Architecture|first=heise|last=online|website=heise online}}</ref><ref>{{Cite web|url=https://www.anandtech.com/show/16921/intel-sapphire-rapids-nextgen-xeon-scalable-gets-a-tiling-upgrade|title=Intel Xeon Sapphire Rapids: How To Go Monolithic with Tiles|first=Dr Ian|last=Cutress|website=[[AnandTech]]}}</ref> It introduced 2-dimensional [[processor register|registers]] called tiles upon which accelerators can perform operations. It is intended as an extensible architecture, the first accelerator implemented is called tile matrix multiply unit (TMUL).<ref>https://software.intel.com/content/dam/develop/public/us/en/documents/architecture-instruction-set-extensions-programming-reference.pdf</ref> <ref>{{Cite web|url=https://fuse.wikichip.org/news/3600/the-x86-advanced-matrix-extension-amx-brings-matrix-operations-to-debut-with-sapphire-rapids/|title=The x86 Advanced Matrix Extension (AMX) Brings Matrix Operations; To Debut with Sapphire Rapids|first=David|last=Schor|date=June 29, 2020}}</ref>
 
=== Tile matrix multiply unit ===
 
TMUL unit supports [[bfloat16 floating-point format|BF16]] and [[INT8]] input types<ref>{{Cite web|url=https://en.wikichip.org/wiki/x86/amx|title=Advanced Matrix Extension (AMX) - x86 - WikiChip|website=en.wikichip.org}}</ref>. The register file consists of 8 tiles, each with 16 rows of size 64-byte (32 BF16 or 64 INT8 values). The only supported operation as for now is [[matrix multiplication|matrix multiplication]] <math> C_{nm} = \sum_{k=1}^K A_{nk}B_{km}. </math> <ref name="iaiseaffpr">{{cite web |url=https://software.intel.com/en-us/intel-architecture-instruction-set-extensions-programming-reference |title=Intel Architecture Instruction Set Extensions and Future Features Programming Reference |access-date=2021-09-26 |publisher=Intel}}</ref>
 
  
 
=== Software support ===
 
=== Software support ===
Line 14: Line 11:
 
*Operating system support
 
*Operating system support
 
**[[glibc]] support for detecting AMX feature in CPUs committed at 25 Jun 2020<ref>{{Cite web|url=https://sourceware.org/git/?p=glibc.git;a=commit;h=4fdd4d41a17dda26c854ed935658154a17d4b906|title=x86: Detect Intel Advanced Matrix Extensions|date=2020-07-02|language=en-US|access-date=2020-07-02}}</ref>
 
**[[glibc]] support for detecting AMX feature in CPUs committed at 25 Jun 2020<ref>{{Cite web|url=https://sourceware.org/git/?p=glibc.git;a=commit;h=4fdd4d41a17dda26c854ed935658154a17d4b906|title=x86: Detect Intel Advanced Matrix Extensions|date=2020-07-02|language=en-US|access-date=2020-07-02}}</ref>
**[[Linux kernel]] support will not be available until at least 5.16<ref>{{Cite web|url=https://www.phoronix.com/scan.php?page=news_item&px=Linux-FPU-Cleaning-For-AMX|title=Linux x86 FPU Code Getting Reworked In Preparation For Intel AMX - Phoronix|website=[[Phoronix]]}}</ref>
+
**[[Linux kernel]] support is available since [[5.16]]<ref>{{Cite web|url=https://www.phoronix.com/scan.php?page=news_item&px=Linux-FPU-Cleaning-For-AMX|title=Linux x86 FPU Code Getting Reworked In Preparation For Intel AMX - Phoronix|website=[[Phoronix]]}}</ref>
  
 
== References ==
 
== References ==

Latest revision as of 07:31, 30 March 2022

wikipedia:Advanced Matrix Extensions (AMX), also known as Intel Advanced Matrix Extensions (Intel AMX), are extensions to the x86 instruction set architecture (ISA) for microprocessors from Intel and Advanced Micro Devices (AMD) designed to work on matrices to accelerate artificial intelligence (AI) / machine learning (ML) -related workloads.[1]

Extensions[edit]

AMX was introduced by Intel in June 2020 and first supported by Intel with the Sapphire Rapids microarchitecture for Xeon servers, planned for 2021.[2][3] It introduced 2-dimensional registers called tiles upon which accelerators can perform operations. It is intended as an extensible architecture, the first accelerator implemented is called tile matrix multiply unit (TMUL).[4] [5]

Software support[edit]

  • Compiler and assembler support
  • Operating system support

References[edit]

  1. Hemsoth, Nicole (August 19, 2021). "With AMX, Intel Adds AI/ML Sparkle to Sapphire Rapids". The Next Platform.<templatestyles src="Module:Citation/CS1/styles.css"></templatestyles>
  2. online, heise. "Intel AMX: Erste Informationen zur Advanced Matrix Extensions Architecture". heise online.<templatestyles src="Module:Citation/CS1/styles.css"></templatestyles>
  3. Cutress, Dr Ian. "Intel Xeon Sapphire Rapids: How To Go Monolithic with Tiles". AnandTech.<templatestyles src="Module:Citation/CS1/styles.css"></templatestyles>
  4. https://software.intel.com/content/dam/develop/public/us/en/documents/architecture-instruction-set-extensions-programming-reference.pdf
  5. Schor, David (June 29, 2020). "The x86 Advanced Matrix Extension (AMX) Brings Matrix Operations; To Debut with Sapphire Rapids".<templatestyles src="Module:Citation/CS1/styles.css"></templatestyles>
  6. Larabel, Michael (2020-07-02). "Intel AMX Support Begins Landing In LLVM". Phoronix. Retrieved 2020-07-02.<templatestyles src="Module:Citation/CS1/styles.css"></templatestyles>
  7. "[X86-64] Support Intel AMX instructions". 2020-07-02. Retrieved 2020-07-02.<templatestyles src="Module:Citation/CS1/styles.css"></templatestyles>
  8. 8.0 8.1 Larabel, Michael (2020-07-02). "Intel AMX Support Lands In The GNU Assembler". Phoronix. Retrieved 2020-07-02.<templatestyles src="Module:Citation/CS1/styles.css"></templatestyles>
  9. "commits with Intel AMX". 2020-07-02. Retrieved 2020-07-02.<templatestyles src="Module:Citation/CS1/styles.css"></templatestyles>
  10. "[PATCH] Enable GCC support for AMX". 2020-07-06. Retrieved 2020-07-09.<templatestyles src="Module:Citation/CS1/styles.css"></templatestyles>
  11. "x86: Detect Intel Advanced Matrix Extensions". 2020-07-02. Retrieved 2020-07-02.<templatestyles src="Module:Citation/CS1/styles.css"></templatestyles>
  12. "Linux x86 FPU Code Getting Reworked In Preparation For Intel AMX - Phoronix". Phoronix.<templatestyles src="Module:Citation/CS1/styles.css"></templatestyles>

Script error: No such module "Check for unknown parameters".

External links[edit]

See also[edit]

Advertising: