Install ispc with Homebrew, apk, apt, dnf, MacPorts, Nix, pacman, scoop, zypper

Project links

• Homepage
• Repository
• Upstream docs
• Package manager page

Commands and aliases

• ispc

Project history

Matt Pharr's project history says he began experimenting with LLVM around 2010 while at Intel and released the code publicly on GitHub on 2011-06-21. The GitHub repository metadata matches that date, and the project site explains the core model: write code that looks mostly serial while multiple program instances execute in parallel on SIMD hardware.

ISPC quickly became both a compiler project and a research artifact. The official documentation points to the InPar 2012 paper by Matt Pharr and William R. Mark, which presents ISPC as a compiler for high-performance CPU programming using SPMD ideas familiar from GPU shaders, CUDA, and OpenCL.

The release notes show continuing evolution with LLVM baselines, new CPU and GPU targets, standard-library work, JIT/library embedding, Python integration experiments, AVX10 and AMX support, and fixes for architecture-specific behavior.

Adoption history

ISPC found adoption among graphics, rendering, simulation, and performance-oriented developers who needed explicit SIMD performance without maintaining separate intrinsics code for every vector width. The project site advertises common speedups on SSE and AVX-width architectures, and official downloads provide binaries for Linux, macOS, and Windows.

Package-manager adoption broadened access beyond Intel-hosted binaries. The input metadata shows ISPC packaged by Homebrew, Debian, Ubuntu, Fedora, MacPorts, Nix, Arch, Scoop, Alpine, and openSUSE.

How it is used

Users write `.ispc` source files, compile them with the `ispc` command, and link the generated objects and headers into C or C++ programs. The documentation emphasizes SPMD concepts, exported functions, task parallelism, vector targets, and examples for integrating generated code with host applications.

In package-manager contexts, ISPC is often installed as a build-time compiler dependency for projects that ship vectorized kernels but want ordinary source builds on developer machines and CI.

Why package nerds care

ISPC is significant because it packages a compiler whose behavior depends on LLVM versions, CPU targets, runtime libraries, and platform-specific binary distribution. That makes it more like a language toolchain than a single-purpose CLI.

It is also a useful case study in how a research-flavored Intel project became a practical dependency for open-source rendering and compute software: source packages must balance reproducible compiler builds, prebuilt binaries, target support, and fast-moving LLVM internals.

Timeline

• 2010: Matt Pharr begins LLVM-based experiments that lead to ISPC.
• 2011: ISPC code goes public on GitHub.
• 2012: Pharr and William R. Mark publish the InPar paper on ISPC.
• 2013: NEON support lands after Pharr's ARM backend experiment and discussion around Intel-maintained support.
• 2025: v1.28.0 adds library/JIT and Python integration work.
• 2026: v1.30.0 and v1.31.0 add AMX, AVX10.2nvl, LLVM baseline, and standard-library work.

Related projects

• ISPC is related to LLVM, C and C++ build systems, SIMD intrinsics, GPU shader languages, CUDA, OpenCL, Intel oneAPI components, rendering engines, and high-performance CPU kernels.

Sources

• GitHub releases and release notes document later target, LLVM, JIT, Python, AVX10, and AMX evolution.
• Matt Pharr's primary-source history posts document the origin and 2011 public release.
• The official site and user guide describe the SPMD language model, CPU/GPU targets, documentation, downloads, and package-manager installation paths.