DEV Community

#cuda

Posts

đź‘‹ Sign in for the ability to sort posts by relevant, latest, or top.
Adding GPU backends to a pure-C TTS engine: Metal, CUDA, and the rented-Mac trick

Adding GPU backends to a pure-C TTS engine: Metal, CUDA, and the rented-Mac trick

8
Comments 4
6 min read
Porting a 1,200-line persistent CUDA megakernel to Qwen3-TTS: ~25 ms to first audio chunk

Porting a 1,200-line persistent CUDA megakernel to Qwen3-TTS: ~25 ms to first audio chunk

2
Comments
3 min read
Notes on CUDA Tensor Core GEMM (WMMA)

Notes on CUDA Tensor Core GEMM (WMMA)

Comments
4 min read
Where Tensor-Parallel Inference Hits the NVLink Wall

Where Tensor-Parallel Inference Hits the NVLink Wall

Comments
2 min read
TensorRT `trt.Dims` SIGSEGV inside a GStreamer Python plugin — root cause and fix

TensorRT `trt.Dims` SIGSEGV inside a GStreamer Python plugin — root cause and fix

Comments
4 min read
Bypassing the OS to Run LLMs: What I Learned Building a Firmware-Centric Runtime

Bypassing the OS to Run LLMs: What I Learned Building a Firmware-Centric Runtime

Comments
5 min read
Why CUDA kernels silently corrupt memory and how to catch the bug

Why CUDA kernels silently corrupt memory and how to catch the bug

Comments
5 min read
TensorCircuit-NG vs cuQuantum on H200: JIT compilation beats the "magic GPU library" assumption

TensorCircuit-NG vs cuQuantum on H200: JIT compilation beats the "magic GPU library" assumption

1
Comments
4 min read
How I optimized a Solana vanity address grinder to 44M keys/sec on GPU

How I optimized a Solana vanity address grinder to 44M keys/sec on GPU

Comments
2 min read
From Black Magic to Science: The Evolution of the CUDA Optimization Skill

From Black Magic to Science: The Evolution of the CUDA Optimization Skill

Comments
11 min read
Learning Resources Tech

Learning Resources Tech

Comments
1 min read
The Microsecond Lie: Why your Go timers are lying about the GPU

The Microsecond Lie: Why your Go timers are lying about the GPU

Comments
3 min read
Profiling a CUDA Python Program with GPUFlight

Profiling a CUDA Python Program with GPUFlight

Comments
10 min read
512MiB 512MB — the silent trtexec bug

512MiB 512MB — the silent trtexec bug

Comments
2 min read
Calling CUDA from Go without cgo

Calling CUDA from Go without cgo

1
Comments
2 min read
đź‘‹ Sign in for the ability to sort posts by relevant, latest, or top.