xfcvDataMovers - 2024.2 English

The xfcvDataMovers class provides a high level API abstraction to initiate data transfer from the DDR to the AIE core and vice versa for hw-emulation / hw runs. Because each AIE core has limited local memory which is not sufficient to fit an entire high resolution image (input / output), each image needs to be partitioned into smaller tiles and then send to AIE core for computation. After computation the tiled image at output is stitched back to generate the high resolution image at the output. This process involves complex computation as tiling needs to ensure proper border handling and overlap processing in case of convolution based kernels.

The xfcvDataMovers class object takes some simple, user provided parameters and provides a simple data transaction API where you do not have to consider the complexity. Moreover it provides a template parameter, using which, the application can switch from PL-based data movement to GMIO-based (and vice versa) seamlessly.

The data transfer using the xfcvDataMovers class can be done in one of two ways:

1. PLIO data movers

   This is the default mode for xfcvDataMovers class operation. When this method is used, data is transferred using hardware Tiler / Stitcher IPs provided by AMD. The Makefile provided with design examples shipped with the library provide the locations of the .xo files for these IPs. It also shows how to incorporate them in the Vitis Build System. You need to create an object of xfcvDataMovers class per input / output image as shown in following code.

   Important

   The implementations of Tiler and Stitcher for PLIO are provided as .xo files in the ‘L1/lib/hw’ folder. By using these files, you are agreeing to the terms and conditions specified in the LICENSE.txt file available in the same directory.

   int overlapH = 1;
   int overlapV = 1;
   xF::xfcvDataMovers<xF::TILER, int16_t, MAX_TILE_HEIGHT, MAX_TILE_WIDTH, VECTORIZATION_FACTOR> tiler(overlapH, overlapV);
   xF::xfcvDataMovers<xF::STITCHER, int16_t, MAX_TILE_HEIGHT, MAX_TILE_WIDTH, VECTORIZATION_FACTOR> stitcher;
   

   The choice of MAX_TILE_HEIGHT / MAX_TILE_WIDTH provides constraints on the image tile size which in turn governs local memory usage. The image tile size in bytes can be computed as follows.

   Image tile size = (TILE_HEADER_SIZE_IN_BYTES + MAX_TILE_HEIGHT*MAX_TILE_WIDTH*sizeof(DATA_TYPE))

   Here TILE_HEADER_SIZE_IN_BYTES is 128 bytes for the current version of Tiler / Stitcher. DATA_TYPE in above example is int16_t (2 bytes).o

   Note

   The current version of HW data movers have 8_16 configuration (i.e., an 8-bit image element data type on the host side and a 16-bit image element data type on the AIE kernel side). In future more such configurations will be provided (example: 8_8 / 16_16 etc.).

   Tiler / Stitcher IPs use PL resources available on VCK boards. For 8_16 configuration, the following table illustrates resource utilization numbers for these IPs. The numbers correspond to a single instance of each IP.

   Table 273 Table: Tiler / Stitcher Resource Utilization (8_16 config)

   	LUTs	FFs	BRAMs	DSPs	Fmax
   Tiler	2761	3832	5	13	400 MHz
   Stitcher	2934	3988	5	7	400 MHz
   Total	5695	7820	10	20

2. GMIO data movers

   Transition to GMIO-based data movers can be achieved by using a specialized template implementation of the above class. All above constraints with regard to the image tile size calculation are valid here as well. Sample code is shown below.

   xF::xfcvDataMovers<xF::TILER, int16_t, MAX_TILE_HEIGHT, MAX_TILE_WIDTH, VECTORIZATION_FACTOR, 1, 0, true> tiler(1, 1);
   xF::xfcvDataMovers<xF::STITCHER, int16_t, MAX_TILE_HEIGHT, MAX_TILE_WIDTH, VECTORIZATION_FACTOR, 1, 0, true> stitcher;
   

   Note

   The last template parameter is set to true, implying GMIO specialization.

Once the objects are constructed, simple API calls can be made to initiate the data transfers. Sample code is shown below.

//For PLIO
auto tiles_sz = tiler.host2aie_nb(&src_hndl, srcImageR.size());
stitcher.aie2host_nb(&dst_hndl, dst.size(), tiles_sz);

//For GMIO
auto tiles_sz = tiler.host2aie_nb(srcData.data(), srcImageR.size(), {"gmioIn[0]"});
stitcher.aie2host_nb(dstData.data(), dst.size(), tiles_sz, {"gmioOut[0]"});

Using tile_sz, you can run the graph the appropriate number of times.

filter_graph_hndl.run(tiles_sz[0] * tiles_sz[1]);

After the runs are started, you need to wait for all transactions to complete.

filter_graph_hndl.wait();
tiler.wait();
stitcher.wait();