{"id":3469,"date":"2018-05-10T17:42:01","date_gmt":"2018-05-10T09:42:01","guid":{"rendered":"https:\/\/www.plunify.com\/en\/?page_id=3469"},"modified":"2019-02-08T12:15:46","modified_gmt":"2019-02-08T04:15:46","slug":"fpga-timing-closure-congested-test-measurement-designs-machine-learning","status":"publish","type":"page","link":"https:\/\/www.plunify.com\/en\/fpga-timing-closure-congested-test-measurement-designs-machine-learning\/","title":{"rendered":"FPGA Timing Closure for Congested Test & Measurement Designs – A Machine Learning Solution"},"content":{"rendered":"

[vc_row row_type=”row” stretch_row_type=”no”][vc_column][vc_column_text]FPGAs are widely used in the Test and Measurement (T&M) industry for their capabilities in high-speed signal processing, high precision data processing, and general design flexibility. As market demand for faster and more precise instrumentation increases, T&M FPGA designs also become more complex and more challenging in terms of timing closure – directly impacting Time-to-Market.<\/p>\n

This article discusses how a new design optimization approach known as the\u00a0InTime Timing Closure Methodology<\/a> was recently used to increase the FMax, as well as to optimize and close timing for five large, routing-congested T&M designs within a week<\/span>. Based on machine learning, this new approach is baked into the InTime<\/a> design optimization software tool.<\/p>\n

Design Background<\/h3>\n

Device:\u00a0<\/strong>Xilinx Virtex Ultrascale xcvu190<\/p>\n

Although FPGA resource utilization was relatively low, numerous SLR crossings and high routing congestion make it difficult to reduce the number of failing paths in these designs. The estimated routing congestion for one of the T&M designs:<\/p>\n

INFO: [Route 35-449] Initial Estimated Congestion\r\n________________________________________________________________________\r\n|\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0 | Global Congestion | Long Congestion\u00a0\u00a0 | Short Congestion\u00a0 |\r\n|\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0 |___________________|___________________|___________________|\r\n| Direction | Size\u00a0\u00a0 | % Tiles\u00a0 | Size\u00a0\u00a0 | % Tiles\u00a0 | Size\u00a0\u00a0 | % Tiles\u00a0 |\r\n|___________|________|__________|________|__________|________|__________|\r\n|\u00a0\u00a0\u00a0\u00a0\u00a0 NORTH|\u00a0\u00a0 64x64|\u00a0\u00a0\u00a0\u00a0\u00a0 3.53|\u00a0\u00a0 32x32|\u00a0\u00a0\u00a0\u00a0\u00a0 6.04|\u00a0\u00a0 64x64|\u00a0\u00a0\u00a0\u00a0\u00a0 6.29|\r\n|___________|________|__________|________|__________|________|__________|\r\n|\u00a0\u00a0\u00a0\u00a0\u00a0 SOUTH|\u00a0\u00a0 64x64|\u00a0\u00a0\u00a0\u00a0\u00a0 6.64|\u00a0\u00a0 64x64|\u00a0\u00a0\u00a0\u00a0\u00a0 8.08|\u00a0\u00a0 64x64|\u00a0\u00a0\u00a0\u00a0\u00a0 8.43|\r\n|___________|________|__________|________|__________|________|__________|\r\n|\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0 EAST|\u00a0\u00a0 32x32|\u00a0\u00a0\u00a0\u00a0\u00a0 2.41|\u00a0\u00a0 16x16|\u00a0\u00a0\u00a0\u00a0\u00a0 1.85|\u00a0\u00a0 32x32|\u00a0\u00a0\u00a0\u00a0\u00a0 6.72|\r\n|___________|________|__________|________|__________|________|__________|\r\n|\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0 WEST|\u00a0\u00a0 64x64|\u00a0\u00a0\u00a0\u00a0\u00a0 4.26|\u00a0\u00a0 32x32|\u00a0\u00a0\u00a0\u00a0\u00a0 2.92|\u00a0128x128|\u00a0\u00a0\u00a0\u00a0 14.73|\r\n|___________|________|__________|________|__________|________|__________|<\/pre>\n
Before & After InTime Optimization<\/h3>\n
Here are the FMax Improvements.\u00a0 You can see that there are between 23% to 79.7% improvements in FMax using this method.<\/p>\n
\"FMax_improvements\"<\/p>\n
 <\/p>\n\n\n\n\n\n\n\n\n
Project Name<\/strong><\/td>\nWNS (Before)<\/strong><\/td>\nWNS (After)<\/strong><\/td>\nTNS (Before)<\/strong><\/td>\nTNS (After)<\/strong><\/td>\nWNS Improvements<\/strong><\/td>\nTNS<\/strong>
\n Improvements<\/strong><\/td>\n		FMax (Before)<\/strong><\/td>\nFMax (After)<\/strong><\/td>\n<\/tr>\n
A<\/td>\n-0.854<\/td>\n-0.217<\/td>\n-16961.285<\/td>\n-582.23<\/td>\n74.59%<\/strong><\/td>\n96.57%<\/td>\n298.15<\/td>\n368.05<\/td>\n<\/tr>\n
B<\/td>\n-0.984<\/td>\n-0.141<\/td>\n-9496.21<\/td>\n-127.48<\/td>\n85.67%<\/strong><\/td>\n98.66%<\/td>\n287.03<\/td>\n378.64<\/td>\n<\/tr>\n
C<\/td>\n-1.593<\/td>\n-0.063<\/td>\n-1972.426<\/td>\n-6.406<\/td>\n96.05%<\/strong><\/td>\n99.68%<\/td>\n244.32<\/td>\n390.17<\/td>\n<\/tr>\n
D<\/td>\n-1.679<\/td>\n-0.043<\/td>\n-21123.871<\/td>\n-0.137<\/td>\n97.44%<\/strong><\/td>\n100.00%<\/td>\n239.29<\/td>\n393.24<\/td>\n<\/tr>\n
E<\/td>\n-2.0<\/td>\n-0.006<\/td>\n-72974.359<\/td>\n-0.006<\/td>\n99.70%<\/strong><\/td>\n100.00%<\/td>\n222.22<\/td>\n399.04<\/td>\n<\/tr>\n<\/tbody>\n<\/table>\n
* Rounded up to 3 decimal places<\/p>\n
<\/h3>\n
Massive Productivity Gain – Results in 6 days<\/h3>\n
Although not all designs closed timing, this approach can significantly increase the FMax and improve the timing results in an automated fashion. The builds were accelerated through cloud computing (AWS) and the entire optimization process was driven by the InTime software. All five designs were optimized at the same time<\/span> in only six days. Compared to solely focusing on iterative improvements, this represents a massive productivity gain for the entire organization.<\/p>\n