The basic workflow of using f1 instances is as follows:
- Apply to acquire the image.
- Develop the software and generate a qar file.
- Verify the qar file.
- Deploy the image.
1. Acquire the image
After your application for an f1 instance trial is approved, Alibaba Cloud shares an FPGA development image with you as a development goal (currently only surports CentOS 7u2). The user can try Intel’s Quartus full set of development components and the constraint files for the corresponding f1 device on this image, which provides a complete cloud development environment.
Currently only support the requirements of region 1 of East China.
2. Development
We encourage our customers to complete the development, compilation, and simulation operations on the cloud, so that they can use the free Intel Quartus development kit. When it is ready to start, an intermediate qar file can be generated during the compilation stage, and uploaded to an OSS bucket (currently only supports East China region 1), and then you can register the qar related information to Alibaba Cloud through an API.
3. Verify the qar file
Alibaba Cloud verifies when receiving the registration request of a qar file. Once the security verification is passed, we send a notification email to the customer who then initiates the registration, along with an image ID of FPGA.
4. Deploy the image
During the deployment process, the user first initiates an association command for an Elastic Compute Service (ECS) instance and an FPGA image: call the API to associate the two parts by using the instance ID of the f1 ECS and the image ID of FPGA as parameters. This operation can be initiated in any scenario where ECS OpenAPI can be accessed.
- If the f1 instance has never been associated with any FPGA image, you can initiate the association operation directly.
- If the f1 instance has previously been associated with the FPGA image and is loaded with the FPGA image, the FPGA image must be erased from the f1 instance before initiating the association command.
When the association operation is successful, the next step is to load the FPGA image. The load operation must be initiated from the cloud server where the f1 instance is located. By calling the API, the client specifies a newly associated FPGA ID, Alibaba Cloud’s underlying service burns the corresponding FPGA image to the corresponding FPGA device for the instance.
If you want to restore the f1 instance to a clean state, you can use the API provided by ECS to erase a burned FPGA image from the associated f1 instance.
See also
For more information on how to use an f1 instance, see