jwu1980-test
Official@jwu1980-test
High-performance GPU kernel optimization and debugging suite for FlashInfer integration, CUDA trace analysis, and JIT-compiled kernel deployment.
Agent Skills by jwu1980-test
Showing 3 vetted skills indexed across 1 GitHub repositories.
jwu1980-testbenchmark-kernel
Benchmark FlashInfer GPU kernels with CUPTI timing and output median throughput metrics.
jwu1980-testdebug-cuda-crash
Capture CUDA API call traces and inputs to debug crashes.
jwu1980-testadd-cuda-kernel
Add CUDA kernels to FlashInfer with JIT and TVM-FFI bindings.
Frequently Asked Questions About jwu1980-test
FAQPage SchemaWhat specific tasks can be performed with these GPU kernel utilities?▼
These utilities enable precise performance benchmarking of FlashInfer kernels, systematic debugging of GPU crashes through call trace capture, and the seamless integration of custom kernels using JIT compilation and TVM-FFI bindings.
Which technical personas benefit from these GPU kernel capabilities?▼
These capabilities are designed for systems engineers, GPU performance researchers, and low-level infrastructure developers focused on optimizing high-throughput inference kernels and resolving complex memory or execution faults within the CUDA ecosystem.
What are the primary prerequisites for running these kernel utilities?▼
Execution requires a compatible NVIDIA GPU environment with the CUDA toolkit installed. Users must have the FlashInfer library configured and access to CUPTI for timing metrics, alongside a development environment capable of handling TVM-FFI bindings for kernel compilation.