gpu

Sep 02, 2026

• gpu
• cuda
• simt
• tensor-cores
• occupancy
• hardware
• llm-systems

Sep 02, 2026

• gpu
• memory-hierarchy
• hbm
• shared-memory
• coalescing
• bank-conflicts
• memory-wall
• performance

Sep 02, 2026

• gpu
• roofline
• arithmetic-intensity
• mfu
• mixed-precision
• fp8
• tensor-cores
• profiling
• performance-modeling

Sep 02, 2026

• gpu
• cuda
• kernels
• vector-add
• matmul
• grid-stride
• cuda-events
• arithmetic-intensity
• llm-systems

Sep 02, 2026

• gpu
• cuda
• gemm
• matmul
• tiling
• coalescing
• shared-memory
• register-blocking
• bank-conflicts
• warptiling
• cublas
• cutlass
• performance

Sep 02, 2026

• gpu
• flashattention
• kernel-fusion
• online-softmax
• memory-wall
• io-aware
• attention
• arithmetic-intensity
• hopper
• blackwell

Sep 02, 2026

• gpu
• triton
• torch-compile
• kernels
• cuda
• autotuning
• llm-systems

Sep 02, 2026

• gpu
• distributed-training
• data-parallelism
• collective-communication
• all-reduce
• nccl
• ddp
• nvlink
• infiniband
• interconnects

Sep 02, 2026

• gpu
• distributed-training
• zero
• fsdp
• fsdp2
• deepspeed
• sharded-data-parallelism
• mixed-precision
• optimizer-state
• offloading
• collectives

Sep 02, 2026

• gpu
• model-parallelism
• tensor-parallelism
• pipeline-parallelism
• sequence-parallelism
• expert-parallelism
• context-parallelism
• megatron
• moe
• distributed-training
• llm-systems

Sep 02, 2026

• gpu
• distributed-training
• 3d-parallelism
• fp8
• transformer-engine
• megatron
• torchtitan
• deepseek-v3
• llama-3
• mfu
• activation-recomputation
• fault-tolerance
• moe

Sep 02, 2026

• gpu
• inference
• serving
• kv-cache
• paged-attention
• vllm
• speculative-decoding
• quantization
• prefill-decode

Sep 02, 2026

• gpu
• careers
• kernels
• distributed-training
• inference
• interviews
• roadmap
• capstone
• gpus