rights reserved. Presenter: Haichen Shen, Yao Wang Amazon SageMaker Neo, Deep Engine Science Dynamic Model in TVM AWS AI© 2019, Amazon Web Services, Inc. or its Affiliates. All rights reserved. Models with models© 2019, Amazon Web Services, Inc. or its Affiliates. All rights reserved. Support dynamic model in TVM ● Support Any-dim in typing ● Use shape function to compute the type at runtime ● Virtual input_name = "data" input_shape = [tvm.relay.Any(), 3, 224, 224] dtype = "float32" block = get_model('resnet50_v1', pretrained=True) mod, params = relay.frontend.from_mxnet(block, shape={input_name:

Memory Model C++11 – C++23About Me: alex.dathskovsky@speedata.io www.linkedin.com/in/alexdathskovsky https://www.cppnext.comAlex Dathskovsky | alex.dathskovsky@speedata.io | www.linkedin.com/in/a

perspective: > I understand C++, and I kinda get assembly because of compiler explorer. Our typical model of AoT is “what C and C++ do”, and I want to expand the understanding for what other computation models native execution of the workload. Can customize its generated code to include a processor cache model which allows it to compute the cache misses and memory stall time of a workload, at slowdowns of native execution of the workload. Can customize its generated code to include a processor cache model which allows it to compute the cache misses and memory stall time of a workload, at slowdowns of

Change Happening Faster Than Ever? Yes, It Is • AI User + Usage + CapEx Growth = Unprecedented • AI Model Compute Costs High / Rising + Inference Costs Per Token Falling = Performance Converging + Developer 2/24 2/25 4/25 75% 60% 10% 21% 15% 0% Details on Page 293 USA – LLM #1 China USA – LLM #2 AI Model Compute Costs High / Rising + Inference Costs Per Token Falling = Performance Converging + Developer Change Happening Faster Than Ever? Yes, It Is • AI User + Usage + CapEx Growth = Unprecedented • AI Model Compute Costs High / Rising + Inference Costs Per Token Falling = Performance Converging + Developer

· · · · · · · · · · · · · · · · · · · · · · · · · · · · 3375 14.3.15 TiFlash Pipeline Execution Model· · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · 3376 14.4 TiDB Distributed eXecution storage architecture, enabling easy scaling of computing or storage capacity separately. The computing layer supports a maximum of 512 nodes, each node sup- ports a maximum of 1,000 concurrencies, and the maximum 2)) to a �→ = 1 AND b = 2 #56005 @ghazalfamilyusa • Increase the cost of table scans in the cost model for scenarios with a high risk of suboptimal execution plans, making the optimizer prefer indexes

· · · · · · · · · · · · · · · · · · · · · · · · · · · · 3359 14.3.15 TiFlash Pipeline Execution Model· · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · 3360 14.4 TiDB Distributed eXecution storage architecture, enabling easy scaling of computing or storage capacity separately. The computing layer supports a maximum of 512 nodes, each node sup- ports a maximum of 1,000 concurrencies, and the maximum 2)) to a �→ = 1 AND b = 2 #56005 @ghazalfamilyusa • Increase the cost of table scans in the cost model for scenarios with a high risk of suboptimal execution plans, making the optimizer prefer indexes

· · · · · · · · · · · · · · · · · · · · · · · · · · · · 3329 14.3.15 TiFlash Pipeline Execution Model· · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · 3330 14.4 TiDB Distributed eXecution storage architecture, enabling easy scaling of computing or storage capacity separately. The computing layer supports a maximum of 512 nodes, each node sup- ports a maximum of 1,000 concurrencies, and the maximum This feature simplifies TiProxy deployment and �→ reduces the complexity of the database access layer. 2.2.1 Feature details 2.2.1.1 Performance • The optimizer allows pushing the Projection operator

· · · · · · · · · · · · · · · · · · · · · · · · · · · · 3321 14.3.15 TiFlash Pipeline Execution Model· · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · 3322 14.4 TiDB Distributed eXecution storage architecture, enabling easy scaling of computing or storage capacity separately. The computing layer supports a maximum of 512 nodes, each node sup- ports a maximum of 1,000 concurrencies, and the maximum optimizes the process of load- ing statistics from multiple perspectives, such as the concurrency model and memory allocation, to reduce latency, improve throughput, and avoid slow loading of statistics

· · · · · · · · · · · · · · · · · · · · · · · · · · · · 3288 14.3.15 TiFlash Pipeline Execution Model· · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · 3289 14.4 TiDB Distributed eXecution storage architecture, enabling easy scaling of computing or storage capacity separately. The computing layer supports a maximum of 512 nodes, each node sup- ports a maximum of 1,000 concurrencies, and the maximum authentication (introduced in v8.1.0) TiCDC supports client authentication using mutual Transport Layer �→ Security (mTLS) or TiDB username and password. This feature enables �→ CLI or OpenAPI clients