RV1109簡介：

     瑞芯微RV1109是一款用于工控機(jī)或人工智能視覺應(yīng)用的高性能機(jī)器視覺處理器SoC。它基于雙核ARM Cortex-A7 32位內(nèi)核，集成了NEON和FPU。每個(gè)核心有一個(gè)32KB的I-cache和32KB的D-cache和512KB的統(tǒng)一L2緩存。內(nèi)置的NPU支持INT8/INT16混合操作，計(jì)算能力高達(dá)1.2TOPs。此外，基于TensorFlow/MXNet/PyTorch/Caffe等一系列框架的網(wǎng)絡(luò)模型具有很強(qiáng)的兼容性，可以方便地進(jìn)行轉(zhuǎn)換。

   RV1109還引入了新一代完全基于硬件的500萬像素ISP（圖像信號(hào)處理器）和后處理器。實(shí)現(xiàn)了IPC和CVR中常用的算法加速器，如HDR、3A函數(shù)（AE、AF、AWB）、LSC、3DNR、2DNR、銳化、dehaze、魚眼校正、gamma校正、特征點(diǎn)檢測等。所有這些都是實(shí)時(shí)處理。通過兩個(gè)MIPI CSI（或LVDS/subvds）和一個(gè)DVP（BT.601/BT.656/BT.1120）接口，用戶可以建立一個(gè)同時(shí)接收來自3個(gè)攝像機(jī)傳感器的視頻數(shù)據(jù)的系統(tǒng)。

  RV1109中嵌入的視頻編碼器支持5M H.265/H.264編碼。它還支持多流編碼，最高可支持一個(gè)5M30FPS和一個(gè)720P30同時(shí)進(jìn)行。借助于這一功能，攝像機(jī)的視頻可以被編碼成更高的分辨率并存儲(chǔ)在本地存儲(chǔ)器中，同時(shí)將另一個(gè)低分辨率的視頻傳輸?shù)皆拼鎯?chǔ)中。RV1109中的H.264/H.265視頻解碼器支持5M的H.264和H.265。RV1109具有高性能的外部DRAM（DDR3/DDR3L/DDR4/LPDDR3/LPDDR4），能夠維持高要求的內(nèi)存帶寬。

 Features

下面列出的某些特性可能受到第三方許可要求的制約。

Application Processor

Dual-Core Cortex-A7

Full implementation of the ARM architecture v7-A instruction set, ARM Neon Advanced SIMD

Separately Integrated Neon and FPU

32KB L1 I-Cache and 32KB L1 D-Cache per Cortex-A7 CPU

Unified 512KB L2 Cache for Dual-Core Cortex-A7

TrustZone technology supported

Separate power domains for CPU core system to support internal power switch and externally turn on/off based on different application scenario

PD_CPU0: 1st Cortex-A7 + Neon + FPU + L1 I/D Cache

PD_CPU1: 2nd Cortex-A7 + Neon + FPU + L1 I/D Cach

One isolated voltage domain to support DVFS

Video Input Interface

Interface and video input processor

Two MIPI CSI/ LVDS/SubLVDS interfaces, 4 lanes each, 1Gbps per lane

One 8/10/12/16-bit standard DVP interface, up to 150MHz input data

Support 601/BT.656 and BT.1120 VI interfaces

Support the polarity of pixel_clk、hsync、vsync configurable

ISP

Maximum resolution is 5M pixel

DVP input: ITU-R 601/656/1120 with raw8/raw10/raw12/raw16, YUV422

MIPI input: RX data lane x1/x2/x4, raw8/raw10/raw12, YUV422

3A: include AE/Histogram, AF, AWB statistics output

FPN: Fixed Pattern Noise removal

BLC: Black Level Correction

DPCC: Static/Dynamic defect pixel cluster correction

LSC: Lens shading correction

Bayer-NR: Bayer-raw De-noising, 2DNR

HDR: 3-/2-Frame Merge into High-Dynamic Range

TMO: 3-/2-Frame Merge Video Tone mapping

WDR: One Frame Wide-Dynamic Range Tone mapping

Debayer: Advanced Adaptive Demosaic with Chromatic Aberration Correction

CCM/CSM: Color correction matrix; RGB2YUV

Gamma: Gamma out correction

Dehaze/Enhance: Automatic Dehaze and edge enhancement

3DLUT: 3D-Lut Color Palette for Customer

LDCH: Lens-distortion in the horizontal direction

Output Scale*3: support scale down level*3(W0<3264; W1<1280; W2<1280)

Output Scale*2: support scale down level*2(W0<1920; W1<1920)

Output (FBC): support YUV422/420 with Frame Buffer Compression

3DNR: Advanced Temporal Noise reduce in YUV

2DNR: Advanced Spatial Noise reduce in YUV

Sharp: Picture Sharpening & Edge Enhance in YUV

ORB: Oriented Fast and Rotated BRIEF, a method of feature points detection

FEC: the bigger Lens-distortion and Fish Eye Correction

CGC: Color Gamut Compression, YUV full range/limit range convert

Video CODEC

Video Decoder

Real-time decoding of H.264 and 265

Main and Main10 profile for H.265, up to 3072-pixel wide

Video Encoder

Real-time H.265/H.264 video encoding

I-/P-frames and SmartP

Five bit rate control modes (CBR, VBR, FixQp, AVBR, and QpMap)

Up to 100 Mbit/s output bit rate

Support ROI(no limit) encoding;

Support multi-stream encoding/decoding

Ø    3072 x 1728@30 fps+1280 x 720@30 fps

Ø    2688 x 1520@30 fps+1280 x 720@30 fps

Ø    2688 x 1944@30 fps+1280 x 720@30fps

Ø    2688 x 1944@30 encoding + 2688 x 1944@30 fps decoding

Input data format:

YCbCr 4:2:0 planar

YCbCr 4:2:0 semi-planar

YCbYCr 4:2:2

CbYCrY 4:2:2 interleaved

RGB444 and BGR444

RGB555 and BGR555

RGB565 and BGR565

RGB888 and BRG888

Ø    RGB101010 and BRG101010

One isolated voltage domain to support DVFS

JPEG CODEC

JPEG Encoder

Baseline (DCT sequential)

Encoder size is from 96×96 to 8192×8192(67Mpixels)

Up to 90 million pixels per second

PEG Decoder

Decoder size is from 48×48 to 8176×8176(66.8Mpixels)

Up to 76 million pixels per second

Neural Process Unit

Neural network acceleration engine with processing performance up to 1.2 TOPS

Support integer 8, integer 16 convolution operation

Support deeplearning frameworks: TensorFlow, TF-lite, Pytorch, Caffe, ONNX, MXNet, Keras, Darknet

Support OpenVX API

One isolated voltage domain to support DVFS

Memory Organization

Internal on-chip memory

BootRom

SYSTEM_SRAM in the voltage domain of VD_LOGIC

PMU_SRAM in the voltage domain of VD_PMU for low power application

External off-chip memory

DDR3/DDR3L/DDR4/LPDDR3/LPDDR4①

SPI Flash

eMMC

SD Card

Async Nand Flash

Internal Memory

Internal BootRom

Support system boot from the following device:

FSPI Flash interface

eMMC interface

SDMMC interface

Async Nand interface

Support system code download by the following interface:

USB OTG interface (Device mode)

SYSTEM_SRAM

Size: 64KB

PMU_SRAM

Size: 8KB

External Memory or Storage device

Dynamic Memory Interface (DDR3/DDR3L/DDR4/LPDDR3/LPDDR4)

Compatible with JEDEC standards

Compatible with DDR3/DDR3L/ DDR4 /LPDDR3/LPDDR4

Support 32-bit data width, 2 ranks (chip selects), max 4GB addressing space per rank, total addressing space is 4GB (max)

Low power modes, such as power-down and self-refresh for SDRAM

eMMC Interface

Compatible with standard iNAND interface

Compatible with eMMC specification 51

Support three data bus width: 1-bit, 4-bit or 8-bit

Support up to HS200; but not support CMD Queue

SD/MMC Interface

Compatible with SD3.0, MMC 51

Data bus width is 4bits

Flexible Serial Flash Interface(FSPI)

Support transfer data from/to serial flash device

Support x1, x2, x4 data bits mode

Support 2 chips select

Nand Flash Interface

Support async nand flash

Data bus width is 8bits

Support 1 chip select

Support LBA nand flash

Up to 16bits/1KB hardware ECC

Support configurable interface timing

System Component

RISC-V MCU

32bit microcontroller core with RISC-V ISA

Harvard architecture, separate Instruction and Data memories

Instruction set is RV32I with M and C extensions

Integrated Programmable Interrupt Controller (IPIC), all 123 IRQ lines connected to GIC for Cortex-A7 also connect to RISC-V MCU

Integrated Debug Controller with JTAG interface

CRU (clock & reset unit)

Support clock gating control for individual components

One oscillator with 24MHz clock input

Support global soft-reset control for whole chip, also individual soft-reset for each component

PMU (power management unit)

Support 5 separate voltage domains VD_CORE/VD_LOGIC/VD_PMU/VD_NPU/VD_VEPU

Support 14 separate power domains, which can be power up/down by software based on different application scenes

Multiple configurable work modes to save power by different frequency or automatic clock gating control or power domain on/off control

Timer

Support 6 64bit-timers with interrupt-based operation for non-secure application

Support 2 64bit-timers with interrupt-based operation for secure application

Support two operation modes: free-running and user-defined count

Support timer work state checkable

PWM

Support 12 on-chip PWMs (PWM0~PWM11) with interrupt-based operation

Programmable pre-scaled operation to bus clock and then further scaled

Embedded 32-bit timer/counter facility

Support capture mode

Support continuous mode or one-shot mode

Provides reference mode and output various duty-cycle waveform

Optimized for IR application for PWM3, PWM7 and PWM11

Watchdog

32-bit watchdog counter

Counter counts down from a preset value to 0 to indicate the occurrence of a timeout

WDT can perform two types of operations when timeout occurs:

Generate a system reset

First generate an interrupt and if this is not cleared by the service routine by the time a second timeout occurs then generate a system reset

Programmable reset pulse length

Totally 16 defined-ranges of main timeout period

One Watchdog for non-secure application

One Watchdog for secure application

Interrupt Controller

Support 128 SPI interrupt sources input from different components

Support 16 software-triggered interrupts

Two interrupt outputs (nFIQ and nIRQ) separately for each Cortex-A7, both are low-level sensitive

Support different interrupt priority for each interrupt source, and they are always software-programmable

DMAC

Micro-code programming-based DMA

Linked list DMA function is supported to complete scatter-gather transfer

Support data transfer types with memory-to-memory, memory-to-peripheral, peripheral-to-memory

Signals the occurrence of various DMA events using the interrupt output signals

One embedded DMA controller for system

DMAC features:

Support 8 channels

27 hardware requests from peripherals

2 interrupts output

Support TrustZone technology and programmable secure state for each DMA channel

Secure System

Cipher engine

Support SM2/SM3/SM4 cipher

Support SHA-1, SHA-256/224, SHA-512/384, MD5 with hardware padding

Support Link List Item (LLI) DMA transfer

Support AES-128 AES-256 encrypt & decrypt cipher

Support AES ECB/CBC/OFB/CFB/CTR/CTS/XTS mode

Support DES & TDES encrypt & decrypt cipher

Support DES/TDES ECB/CBC/OFB/CFB mode

Support up to 4096 bits PKA mathematical operations for RSA/ECC

Support up to 8-channels configuration

Support Up to 256 bits TRNG output

Support data scrambling for all DDR types

Support secure OTP

Support secure debug

Support secure OS

Mailbox

One Mailbox in SoC to service A7 and RISC-V communication

Support four mailbox elements per mailbox, each element includes one data word, one command word register and one flag bit that can represent one interrupt

Provide 32 lock registers for software to use to indicate whether mailbox is occupied

DECOM

Support for decompressing GZIP files

Support for decompressing LZ4 files, including the General Structure of LZ4 Frame format and the Legacy Frame

Support for decompressing data in Deflate format

Support for decompressing data in ZLIB format

Support complete interrupt and error interrupt output

Support Hash32 check in LZ4 decompression process

Support the limit size function of the decompressed data to prevent the memory from being maliciously destroyed during the decompression process

Support software to stop the decompression process

Graphic Engine

2D Graphics Engine (RGA):

Source formats:

u    ABGR8888, XBGR888, ARGB8888, XRGB888

u    RGB888, RGB565

u    RGBA5551, RGBA4444

YUV420 planar, YUV420 semi-planar

YUV422 planar, YUV422 semi-planar

YUV 10-bit for YUV420/422 semi-planar

BPP8, BPP4, BPP2, BPP1

Destination formats:

u    ABGR8888, XBGR888, ARGB8888, XRGB888

u    RGB888, RGB565

u    RGBA5551, RGBA4444

YUV420 planar, YUV420 semi-planar

YUV422 planar, YUV422 semi-planar

Pixel Format conversion, 601/BT.709

Max resolution: 8192×8192 source, 4096×4096 destination

BitBLT

Two source BitBLT:

A+B=B only BitBLT, A support rotate and scale when B fixed

A+B=C second source (B) has same attribute with (C) plus rotation function

Color fill with gradient fill, and pattern fill

High-performance stretch and shrink

Monochrome expansion for text rendering

New comprehensive per-pixel alpha (color/alpha channel separately)

Alpha blending modes including Java 2 Porter-Duff compositing blending rules, chroma key, pattern mask, fading

Dither operation

0, 90, 180, 270-degree rotation

x-mirror, y-mirror and rotation operation

Image Enhancement Processor (IEP):

Image format

Input data: YUV420/YUV422, semi-planar/planar, UV swap

Output data: YUV420/YUV422, semi-planar, UV swap, Tile mode

YUV down sampling conversion from 422 to 420

Max resolution for dynamic image up to 1920×1080

De-interlace

Display Interface

One up to 24 bits RGB parallel video output interface

One 1120 video output interface

One 4 lane MIPI DSI interface, up to 1Gbps per lane

Up to 1080p@60fps

Video Output Processor (VOP)

n     Up to 1920×1080 @60fps

Multiple layer

Background layer

Win0 layer

Win2 layer

Input format: RGB888, ARGB888, RGB565, YCbCr422, YCbCr420, YCbCr444

1/8 to 8 scaling-down and scaling-up engine

Support virtual display

256 level alpha blending (pre-multiplied alpha support)

Transparency color key

YCbCr2RGB (rec601-mpeg/ rec601-jpeg/rec709)

RGB2YCbCr (BT.601/BT.709)

Support multi-region

Win0 layer and Win2 layer overlay exchangeable

Support RGB or YUV domain overlay

BCSH (Brightness, Contrast, Saturation, Hue adjustment)

BCSH: YCbCr2RGB (rec601-mpeg/ rec601-jpeg/rec709)

BCSH: RGB2YCbCr (BT.601/BT.709)

Support Gamma adjust

Support dither down allegro RGB888to666 RGB888to565 & dither down FRC (configurable) RGB888to666

Blank and black display

Audio Interface

I2S0 with 8 channels

Up to 8 channels TX and 8 channels RX path

Audio resolution from 16bits to 32bits

Sample rate up to 192KHz

Provides master and slave work mode, software configurable

Support 3 I2S formats (normal, left-justified, right-justified)

Support 4 PCM formats (early, late1, late2, late3)

I2S and PCM mode cannot be used at the same time

I2S1/I2S2 with 2 channels

Up to 2 channels for TX and 2 channels RX path

Audio resolution from 16bits to 32bits

Sample rate up to 192KHz

Provides master and slave work mode, software configurable

Support 3 I2S formats (normal, left-justified, right-justified)

Support 4 PCM formats (early, late1, late2, late3)

I2S and PCM cannot be used at the same time

PDM

Up to 8 channels

Audio resolution from 16bits to 24bits

Sample rate up to 192KHz

Support PDM master receive mode

TDM

Support up to 8 channels for TX and 8 channels RX path

Audio resolution from 16bits to 32bits

Sample rate up to 192KHz

Provides master and slave work mode, software configurable

Support 3 I2S formats (normal, left-justified, right-justified)

Support 4 PCM formats (early, late1, late2, late3)

Audio PWM

Support convert PCM to PWM format

Sample rate up to 16x

Support linear interpolation for 2x/4x/8x/16 oversampling

Support 8/9/10/11 bits maskable L/R channel PWM output

Digital Audio Codec

Support 3-channel digital ADC

Support 2-channel digital DAC

Support I2S/PCM interface

Support I2S/PCM master and slave mode

Support 4-channel audio transmitting in I2S mode

Support 2-channel audio receiving in I2S mode

Support 2-channel audio transmitting or receiving in PCM mode

Support 16~24 bit sample resolution for both digital ADC and digital DAC

Both digital ADC and digital DAC support three groups of sample rates. Group 0 are 8khz/16khz/32kHz/64kHz/128khz, group 1 are 11.025khz/22.05khz/44.1khz/88.2khz/176.4khz and group 2 are 12khz/24khz/48khz/96khz/192khz

The passband of digital ADC filters is 45625*fs

Support digital ADC pass-band ripple within +/-0.1dB

The stop-band of digital ADC filters is 5*fs

Support digital ADC stop-band attenuation at least 60dB

Support volume control for both digital ADC and digital DAC

Support Automatic Level Control (ALC)and noise gate for digital ADC

Support communication with Analog Codec through I2C bus

Connectivity

SDIO Interface

Compatible with SDIO3.0 protocol

4bits data bus widths

GMAC 10/100/1000M ethernet controller

Support 10/100/1000-Mbps data transfer rates with the RGMII interfaces

Support 10/100-Mbps data transfer rates with the RMII interfaces

Support both full-duplex and half-duplex operation

Support for TCP Segmentation Offload (TSO) and UDP Segmentation Offload (USO) network acceleration

USB 2.0 Host

Compatible with USB 2.0 specification

Supports high-speed(480Mbps), full-speed(12Mbps) and low-speed(1.5Mbps) mode

Support Enhanced Host Controller Interface Specification (EHCI), Revision 0

Support Open Host Controller Interface Specification (OHCI), Revision 0a

USB 2.0 OTG

Universal Serial Bus Specification, Revision 0

Extensible Host Controller Interface for Universal Serial Bus (xHCI), Revision 1.1

Compatible Specification

Support Control/Bulk/Interrupt/Isochronous Transfer

SPI Interface

Support 2 SPI Controllers, support two chip-select output

Support serial-master and serial-slave mode, software-configurable

I2C Interface

Support 6 I2C interfaces(I2C0-I2C5)

Support 7bits and 10bits address mode

Software programmable clock frequency

Data on the I2C-bus can be transferred at rates of up to 100k bits/s in the

Standard-mode, up to 400k bits/s in the Fast-mode or up to 1m bits/s in Fast-mode Plus

UART Interface

Support 6 UART interfaces (UART0-UART5)

Support 5bit, 6bit, 7bit, 8bit serial data transmit or receive

Standard asynchronous communication bits such as start, stop and parity

Support different input clock for UART operation to get up to 4Mbps baud rate

Support auto flow control mode(except UART2)