# Explore the PIE capabilities on the ESP32-P4

DevFeed: [Explore the PIE capabilities on the ESP32-P4](<https://devfeed.tech/articles/explore-the-pie-capabilities-on-the-esp32-p4-13674.md>)

Original publisher: [Read original article](<https://developer.espressif.com/blog/2024/12/pie-introduction/>)

Author: John Lee

Published: 2024-12-05T00:00:00Z

Content type: article

Language: en

Sources: [Blog on Developer Portal](<https://devfeed.tech/sources/blog-on-developer-portal.md>)

Topics: [ESP32-P4](<https://devfeed.tech/topics/esp32-p4.md>), [On-device AI](<https://devfeed.tech/topics/on-device-ai.md>), [tinyml](<https://devfeed.tech/topics/tinyml.md>), [performance-engineering](<https://devfeed.tech/topics/performance-engineering.md>), [real-time](<https://devfeed.tech/topics/real-time.md>)

Tags: [ai](<https://devfeed.tech/tags/ai.md>), [algorithms](<https://devfeed.tech/tags/algorithms.md>), [architecture](<https://devfeed.tech/tags/architecture.md>), [assembly](<https://devfeed.tech/tags/assembly.md>), [blog](<https://devfeed.tech/tags/blog.md>), [data-movement](<https://devfeed.tech/tags/data-movement.md>), [dsp](<https://devfeed.tech/tags/dsp.md>), [efficiency](<https://devfeed.tech/tags/efficiency.md>), [esp32](<https://devfeed.tech/tags/esp32.md>), [esp32-p4](<https://devfeed.tech/tags/esp32-p4.md>), [esp32-s3](<https://devfeed.tech/tags/esp32-s3.md>), [execution](<https://devfeed.tech/tags/execution.md>), [extensions](<https://devfeed.tech/tags/extensions.md>), [hardware](<https://devfeed.tech/tags/hardware.md>), [inference](<https://devfeed.tech/tags/inference.md>), [loops](<https://devfeed.tech/tags/loops.md>), [mac](<https://devfeed.tech/tags/mac.md>), [model-inference](<https://devfeed.tech/tags/model-inference.md>), [natural-language-processing](<https://devfeed.tech/tags/natural-language-processing.md>), [pie](<https://devfeed.tech/tags/pie.md>)

## AI overview

PIE adds SIMD instructions to ESP32-S3 and ESP32-P4 to accelerate selected AI and DSP workloads, supporting 8-, 16- and 32-bit vector operations and combining data movement with multiplication, shifting and accumulation. The excerpt identifies several chip differences: ESP32-S3 uses a TIE-based extension, while ESP32-P4 adds a hardware-loop extension; their two MAC accumulators are 160 bits and 256 bits, respectively. Instruction formats are largely similar, though prefixes and some instruction behavior or usage differ. Consult each chip's technical reference manual for those details. The excerpt says the article will analyze and test PIE, but provides no test results.

## Source excerpt

The recent breakthroughs in artificial intelligence technology in such fields as image recognition, speech recognition, and natural language processing have opened up more possibilities for embedded system applications. When attempting to deploy AI model inference on embedded devices such as the ESP32-P4, we always strive to minimize inference time as much as possible to meet real-time requirements.