SIMT
SIMT is a processor execution model in which groups of threads execute a common instruction stream while operating on separate data.
GPU Programming with Triton
The Single Instruction, Multiple Threads (SIMT) execution model runs the same instruction across multiple threads operating on separate data.
- How it is used
- Motivates Triton's block-level programming model by contrasting it with CUDA's thread-level SIMT model.
- How deep it goes
- Explains warp instruction issue, divergent branches, and independent thread progress on GPUs since Volta.
- Examples
- Uses vector addition and conditional ReLU with mixed-sign inputs.
- What you can do after
- Evaluate how thread grouping and divergent branches affect GPU execution.
Grokking GPU Programming
SIMT is a GPU execution model in which groups of threads execute instructions together while operating on separate data.
- How it is used
- Explains how CUDA kernel threads are grouped and executed on GPU hardware.
- How deep it goes
- Traces warp grouping and array execution, then explains divergence and the need for explicit synchronization.
- Examples
- Uses a marching-band analogy, an array-doubling kernel, and a warp-count exercise.
- What you can do after
- Diagnose warp divergence and avoid relying on lockstep when threads exchange data.
CUDA for LLMs
SIMT is a processor execution model in which multiple threads execute the same instruction stream while individual threads may branch independently.
- How it is used
- Identifies data-parallel opportunities by mapping repeated operations on independent values to GPU threads.
- How deep it goes
- Defines SIMT against SIMD and explains independent branching, stopping short of detailing divergence mechanics.
- Examples
- Uses element-wise ReLU and cites convolutions, attention, and gradient updates as data-parallel workloads.
- What you can do after
- Evaluate data-parallel opportunities by spotting repeated operations on independent values and mapping work to GPU threads.
What to know first
- threadA thread is a distinct sequence of execution with its own control flow and state, managed or scheduled by a processor, runtime, or operating system.
- hardware registerA hardware register is a small, rapidly accessible storage location within a processor or processing device that holds data, addresses, or control information during execution.
- thread blockA thread block is a group of GPU threads organized to execute together and cooperate using shared resources.
Related concepts
- SIMD hardwarecontrasts with · SIMD hardware applies the same instruction or operation concurrently to multiple data elements.
- SPMDcontrasts with · SPMD is a parallel programming model in which multiple program instances execute the same code while operating on different data.
- threaduses · A thread is a distinct sequence of execution with its own control flow and state, managed or scheduled by a processor, runtime, or operating system.
- GPUused by · A graphics processing unit (GPU) is a processor designed for highly parallel computation, especially graphics and other data-parallel workloads.
- warpuses · A warp is a group of GPU threads scheduled and issued together for execution, typically processing the same instruction on different data.
- embarrassingly parallelused by · Embarrassingly parallel describes a computation whose independent work units can be executed concurrently with little or no communication, synchronization, or coordination.
Questions about SIMT
What is SIMT?
SIMT is a processor execution model in which groups of threads execute a common instruction stream while operating on separate data.
Which Manning books explain SIMT?
GPU Programming with Triton, Grokking GPU Programming and CUDA for LLMs.
What should I know before learning SIMT?
Manning books that teach SIMT usually introduce thread, hardware register and thread block first.
What is the difference between SIMT and SIMD hardware?
SIMT is a processor execution model in which groups of threads execute a common instruction stream while operating on separate data. SIMD hardware applies the same instruction or operation concurrently to multiple data elements.
These guides are generated from the text of Manning books by Manning's knowledge graph and link to the sections they draw on. The full books are available on liveBook and at manning.com.


