FFmpeg screen captures with c66x acceleration enabled

Virtual ffmpeg hardware with c66x accelerator (VMM dialog / VM configuration screen cap)

Side-by-side desktop capture and tablet player, with c66x accelerator low-latency RTP streaming Click here to see a YouTube video of VDI streaming

FFmpeg Accelerator -- FFmpeg Hardware Acceleration

---

• Overview
• Product Information
• Why Use Hardware Acceleration
• FFmpeg Command Line Examples
• FFmpeg in VMs
• HPC Software Model
• CPU vs. GPU Overview
• How To Purchase
• Related Items
• Tech Support

Overview

• Maximize ffmpeg performance with software + hardware solution
• Compatible with Linux servers, all form-factors
• Use ffmpeg standard command line
• Multiple streams with endpoint combinations of file, RTP, and MPEG-TS supported
• H.264 standard settings and some presets supported
• Latency control supported; accelerator cards have their own dedicated NIC to reduce latency
• Add up to 512 CPU cores in a standard server to dramatically increase ffmpeg concurrency
• Run multiple, concurrent ffmpeg instances in multiple VMs, without decreased performance

Product Information

Information for software and 32-core and 64-core accelerators below includes specifications, documentation links, and technical support.

Why Use Hardware Acceleration

Various vendors talk about GPU, DSP, FPGA, etc. but these are all "really hard to use" solutions. These vendors miss the point, trying to prove that servers are not traditionally good at streaming and image processing. Obviously servers are plenty good at running ffmpeg and image processing.

What is fair to say is that trends in server architecture, including virtualization, DPDK, and software-defined networking, have increased performance expectations of individual servers. In the past HPC users were willing to stack boxes; now due to virtualization, they expect one box to be a stack of VMs -- without penalties in performance, latency, and network I/O throughput.

Signalogic's software and hardware acceleration products are designed to take full advantage of server architecture trends. They augment modern servers, providing well integrated functionality and user interface, as opposed to alternative, non-mainstream methods. User interfaces include ffmpeg command line, opencv APIs, OpenMP programming, and VM configuration. Within the latter, our products add cores, DDR3 mem, and NIC resources that can be used both with/without virtualization. The result is improved server performance density, reduced latency, and increased stream concurrency, without having to use proprietary programming languages, study a 300-page chip vendor data sheet, work with low level "SDKs", or other time-consuming development efforts.

FFmpeg Command Line Examples

Below are some example ffmpeg commands. Adding the option "-hwaccel c66x" enables c66x hardware acceleration.

/ffmpeg -y -hwaccel c66x -s 352x288 -i ../video_files/
airshow_352x288p_30fps_420fmt.yuv -c:v libx264-b 1500000 -r 30 ../video_files/test.h264

/ffmpeg -y -hwaccel c66x -s 352x288 -i ../video_files/ airshow_352x288p_30fps_420fmt.yuv
-c:v libx264 -b 1500000 -r 30 -f rtp rtp://10.0.1.65:45056:60-af-6d-75-75-f1

/ffmpeg -y -hwaccel c66x -s 352x288 -i ../video_files/airshow_352x288p_30fps_420fmt.yuv -c:v libx264 -b
1500000 -r 30 ../video_files/test.h264 -f rtp rtp://10.0.1.65:45056:60-af-6d-75-75-f1

FFmpeg in VMs

c66x accelerator resources can be assigned / allocated to VMs, allowing multiple instances of ffmpeg to run concurrently, each with their own accelerator cores, memory, streaming network I/O, independently from the server motherboard. Currently the KVM hypervisor is supported, along with QEMU, libvirt, and virt-manager. Below is an image showing installation of an Ubuntu VM, running on a CentOS host:

VM configuration screen cap, showing c66x accelerator resources

HPC Software Model

Below is a software model diagram that shows how c66x CPUs are combined with x86 CPUs inside high performance computing (HPC) servers. The user interface varies in complexity from left to right, from simple (command line) to complex (OpenMP pragmas). Pragmas are compiled by the CIM^® Hyperpiler^™, which separates C/C++ source into "soure code streams" and augments streams with additional, auto-generated C/C++ source code. Multiple different applications can be run at the same time, as concurrent host processes, concurrent VMs, or a combination. Adding DPDK capability to application user space is optional.

Software model for combination of c66x and x86 CPUs in HPC servers

Notes

¹CIM^® = Compute Intensive Multicore
²RTAF = Real-time Algorithm Framework
³HPMN = High Performance Multicore Network stack

CPU vs. GPU Overview

CPU and GPU devices are constructed from fundamentally different chip architectures. Both are very good at certain things, and both are not so good at some things -- and these strengths and weaknesses are mostly opposites, or complementary. In general:

• CPUs tend to be good at complex algorithms that require random memory access, non-uniform treatment of data sets, unpredictable decision paths, and interface to peripherals (network I/O, PCIe, USB, etc).
• GPUs tend to be good at well-defined algorithms that operate uniformly on large data sets, accurate and very fast math, and graphics applications of all types

Neither of these devices is a "do-all" or panacea for complex computing, or somehow fundamentally superior to the other, as the marketing hype would have you believe. People tend to forget that top global semiconductor manufacturers are all on the same technology curve -- which of course makes sense, as they all use the state-of-the-art semiconductor manufacturing technology of our time. If you look closely at the two key factors that form the basis for Moore's Law, performance and power consumption (the chip metric is GFlops / Watt), you won't see much much difference between Intel, Nvidia, Texas Instruments, Xilinx, etc. What you will find are differences in corporate practice and marketing culture, ingrained over very long periods of time -- 30 years or more -- that makes one manufacturer or another more adept at serving certain market segments, with advantages (or disadvantages!) in package size, memory bandwidth, onchip integrated peripherals, programmability, etc.

Below are simplified diagrams of CPU and GPU accelerators.

c66x multicore CPU accelerator diagram (shown for a CIM-64 card, with 8 cores per CPU and 2 GB mem per CPU). All CPU cores have NIC access	GPU accelerator diagram (shown for a Kepler 80 card, with 13 Streaming Multiprocessors (SMs) per GPU, 192 CUDA cores per SM, and 12 GB mem per GPU). A GPU can have literally 1000s of "CUDA cores"

In comparing the diagrams above, some obvious differences and similarities stand out:

• GPU cores are sometimes called "CUDA cores", in reference to Nvidia's programming language. It's not easy to compare CPU and GPU cores (apples and oranges). Maybe the easiest way to think about it is (i) for any given math calculation, a GPU core can almost always do it much faster, and (ii) GPU cores do not run arbitrary C/C++, Java, or Python code, so they're not programmable in the conventional sense
• A GPU accelerator can bring far more processing force to bear on massively parallel problems. Examples include graphics, bitcoin mining, climate simulations, DNA sequencing -- any problem where the data set can be subdivided into "regions", such that the results of one region do not depend on others
• A CPU accelerator typically has its own NIC (or more than one), which can provide advantages in reduced latency and "data localization" -- bringing the compute cores closer to the data. Onboard NICs are typically not found on GPU accelerators as GPU cores are not designed to run device drivers, TCP/IP stack, etc.
• Both types of accelerators take full advantage of high performance PCIe interfaces found in modern servers, including multiple PCIe slots and risers, accessibility to DPDK cores, and excellent software support in Linux

How to Purchase Enter the fields shown above, including: Required Number of Streams Required Number of VMs Max Frame rate Max Resolution Qty -- number of cards required Your discount category Then click on "Get Quote / Info" to receive pricing via e-mail.

Related Items Video Algorithms Video Analytics

Tech Support

Signalogic's engineering staff designs, develops, maintains, operates, and tests software and hardware in the company's in-house lab, using servers from HP, Dell, Supermicro, Artesyn, Advantech, and others, and Linux installations including CentOS, Ubuntu, Red Hat, Wind River, Gentoo, and more. Streaming output is tested using Android and Microsoft Surface tablets with players including VLC, MPlayer, and others. Customers can submit technical questions via e-mail, phone, or Skype chat.

Signalogic engineers are experts in server and embedded system development. Unlike retailers and distributors who offer analytics and video-related products, Signalogic can also perform contract development. Signalogic is a member of third-party programs for HP, Dell, Intel, and Texas Instruments. A high level of expert tech support is a distinct advantage when purchasing products from Signalogic.