Rust SIMD on the GPU

# Rust SIMD on the GPU Source: [https://www.vectorware.com/blog/simd-on-gpu/](https://www.vectorware.com/blog/simd-on-gpu/) [VectorWare](https://www.vectorware.com/)[Dispatches](https://www.vectorware.com/blog/) August 10, 202612min read Pedantic mode:Off GPU code can now use Rust's portable SIMD\. We share the implementation approach and what this unlocks for GPU programming\. At[VectorWare](https://www.vectorware.com/), we are building the first[GPU\-native software company](https://www.vectorware.com/blog/announcing-vectorware/)\. Today, we are excited to announce that we can successfully use Rust's portable SIMD \([`core::simd`](https://doc.rust-lang.org/core/simd/index.html)\) on the GPU\. This milestone marks a significant step towards our vision of enabling developers to write complex, high\-performance applications that leverage the full power of GPU hardware using familiar Rust abstractions\. ## Parallelism below the thread When we[brought Rust threads to the GPU](https://www.vectorware.com/blog/threads-on-gpu/), we mapped each[`std::thread`](https://doc.rust-lang.org/std/thread/)to a GPU[warp](https://modal.com/gpu-glossary/device-software/warp)\. This let us run many concurrent threads on the GPU but did not use the parallel[lanes](https://docs.nvidia.com/cuda/cuda-programming-guide/01-introduction/programming-model.html#warps-and-simt)within each thread/warp\. On the CPU, the abstraction for parallelism within a thread is[SIMD](https://en.wikipedia.org/wiki/Single_instruction,_multiple_data)\.A single instruction operates on several data elements packed into a vector unit: where scalar code adds two numbers, a SIMD add takes two vectors of, say, eight`f32`values and produces eight sums at once\. This data parallelism is*inside*a single thread, below the level where the operating system schedules anything\. CPU threadSIMD op012N⋯SIMD lanesCPU thread ## Rust's portable SIMD Historically, writing SIMD in Rust meant reaching for the architecture\-specific vendor intrinsics in[`core::arch`](https://doc.rust-lang.org/core/arch/index.html), such as[`\_mm256\_add\_ps`](https://doc.rust-lang.org/beta/core/arch/x86_64/fn._mm256_add_ps.html)on x86\-64 or[`vaddq\_f32`](https://doc.rust-lang.org/beta/core/arch/arm/fn.vaddq_f32.html)on Arm\. These intrinsics are specific to a single instruction set, so a program that runs on more than one architecture needs a separate implementation for each\. Rust's[portable SIMD](https://doc.rust-lang.org/core/simd/index.html)instead adds a layer of abstraction above these intrinsics\. It provides a single generic type[`Simd<T, N\>`](https://doc.rust-lang.org/core/simd/struct.Simd.html)that represents a vector of`N`elements of type`T`\. A program writes its arithmetic, comparisons, reductions, and lane shuffles once against`Simd`and the compiler lowers them to whatever vector instructions the target CPU has\. **At VectorWare, we realized the GPU is just one more piece of vector hardware for portable SIMD to target\.**As a bonus, portable SIMD lives in`core`rather than`std`and it does not even need the[`std`support we brought to the GPU](https://www.vectorware.com/blog/rust-std-on-gpu)\. ## SIMT is SIMD GPUs execute in a model NVIDIA calls[SIMT](https://en.wikipedia.org/wiki/Single_instruction,_multiple_threads), or Single Instruction, Multiple Thread\. A warp issues one instruction, and each of its 32 lanes runs that instruction on its own data\. One instruction operating on many data elements is*exactly*what SIMD means, and the per\-lane addressing that SIMT adds does not change that\.A warp is a wide vector unit and a portable SIMD vector maps onto that unit directly\. CPU thread012N⋯SIMD lanes≈GPU warp012N⋯warp lanes For example, a`Simd<i16, 32\>`gives one`i16`element to each of the warp's 32 lanes, and adding two such vectors compiles to a single warp instruction in which every lane adds its element at once\. CPUlet a: Simd<i16, 32\> = \[1,1,1,\.\.\.,1\];let b: Simd<i16, 32\> = \[2,2,2,\.\.\.,2\];let c = a \+ b;compiles tovpaddw %zmm2, %zmm1, %zmm0a0\+b0lane 0a1\+b1lane 1a2\+b2lane 2a31\+b31lane 31⋯println\!\("\{c:?\}"\); GPUlet a: Simd<i16, 32\> = \[1,1,1,\.\.\.,1\];let b: Simd<i16, 32\> = \[2,2,2,\.\.\.,2\];let c = a \+ b;compiles toadd\.s16 %rs3, %rs1, %rs2;a0\+b0lane 0a1\+b1lane 1a2\+b2lane 2a31\+b31lane 31⋯println\!\("\{c:?\}"\); This new mapping completes the parallelism hierarchy from our earlier work\. On the CPU, a thread contains SIMD lanes, and on the GPU[our`std::thread`is a warp](https://www.vectorware.com/blog/threads-on-gpu/)whose hardware lanes play the same role\. In both cases,`core::simd`drives those lanes\. CPU⋯thread 0012N⋯thread 1012N⋯thread N012N⋯SIMD lanes≈GPU⋯warp 0012N⋯warp 1012N⋯warp N012N⋯warp lanes ## A world first:`core::simd`on the GPU As with our earlier posts, this is hard to show visually because the code is ordinary Rust\. The same`core::simd`types that lower to x86\-64 SIMD on a laptop lower to warp operations on the GPU, with no change to the source\. Here we define a small portable SIMD routine and call it from`main`\. It exercises the core features of the model: elementwise arithmetic, a comparison that produces a lane mask, a`select`driven by that mask, and a horizontal reduction across lanes\. ``` #![feature(portable_simd)] use core::simd::cmp::SimdPartialOrd; use core::simd::num::SimdFloat; use core::simd::{Select, Simd}; // Portable SIMD. This exact function also compiles and runs on the CPU, // where it lowers to x86-64, Arm, or scalar code depending on the target. fn relu_dot(a: Simd<f32, 32>, b: Simd<f32, 32>) -> f32 { // Elementwise multiply: 32 products computed at once. let products = a * b; // Per-lane comparison produces a mask, one boolean per lane. let positive = products.simd_gt(Simd::splat(0.0)); // Keep the positive products, replace the rest with zero. let clamped = positive.select(products, Simd::splat(0.0)); // Horizontal add across all lanes down to a single scalar. clamped.reduce_sum() } fn main() { // Two 32-wide vectors, built with ordinary Rust. let a = Simd::<f32, 32>::splat(2.0); let b = Simd::<f32, 32>::from_array(std::array::from_fn(|i| i as f32 - 16.0)); // Elementwise ops, a comparison mask, a select, and a reduction: // all ordinary portable SIMD, all running on the GPU. let result = relu_dot(a, b); // Printed from the GPU using our std support. println!("relu_dot = {result}"); } ``` The entry point is a normal`fn main`with no GPU\-specific annotations\. Our toolchain compiles it to a GPU kernel, and the result is printed from the device using our[`std`support](https://www.vectorware.com/blog/rust-std-on-gpu)\. Below is a recording of the program running on the GPU, producing the exact same output as[running it on the CPU](https://play.rust-lang.org/?version=nightly&mode=debug&edition=2024&gist=c6fd3bb9bb99b2bb92b2255c3174ac7b)\. ## Implementation As previously mentioned, the mapping rests on a single observation: a warp is a vector unit whose lanes are individually addressable\. Once`Simd<T, N\>`is laid out per lane, each family of operations has a direct warp\-level counterpart\. **SIMD elementwise operations**are the easy case\. Addition, multiplication, comparison, and the other lane\-wise operators come from ordinary Rust trait implementations on`Simd`such as[`Add`](https://doc.rust-lang.org/std/simd/type.f32x32.html#impl-Add%3C%26Simd%3CT,+N%3E%3E-for-Simd%3CT,+N%3E)\. The GPU runs them natively\. **SIMD reductions**such as[`reduce\_sum`](https://doc.rust-lang.org/core/simd/struct.Simd.html#method.reduce_sum)and[`reduce\_max`](https://doc.rust-lang.org/core/simd/struct.Simd.html#method.reduce_max)combine every lane into a scalar\. These use the GPU's warp shuffle instructions to exchange and combine values across lanes, producing the same scalar result in every lane\. **SIMD cross\-lane shuffles**, such as[`simd\_swizzle\!`](https://doc.rust-lang.org/core/simd/macro.simd_swizzle.html)and rotates, move elements between lanes\. Because a SIMD lane is a GPU warp lane, these map onto the same warp shuffle primitives that make GPU lanes so good at exchanging data\. **SIMD masks**map just as cleanly\. A[`Mask<T, N\>`](https://doc.rust-lang.org/core/simd/struct.Mask.html)gives one predicate to each SIMD lane\.[`Mask::select`](https://doc.rust-lang.org/core/simd/struct.Mask.html#method.select)performs a selection in every warp lane\. Horizontal mask queries such as[`any`](https://doc.rust-lang.org/core/simd/struct.Mask.html#method.any)and[`all`](https://doc.rust-lang.org/core/simd/struct.Mask.html#method.all)use GPU[vote and ballot](https://docs.nvidia.com/cuda/parallel-thread-execution/index.html#parallel-synchronization-and-communication-instructions-vote-sync)instructions\. Scalar values in the surrounding code, such as a loop counter or a constant, are computed identically by every lane and so are simply replicated across the warp just like in ordinary CUDA\. This is the same uniform\-versus\-varying distinction that data\-parallel languages like[ISPC](https://ispc.github.io/)make explicit, except here it falls out of Rust's own types: a plain`f32`is uniform, a`Simd<f32, 32\>`is varying\. ## Working with lanes The one place the abstraction and the hardware do not line up is lane count\. On the CPU a`Simd<T, N\>`allows any`N`from 1 through 64, but GPU hardware has a fixed width: 32 lanes on NVIDIA and 32 or 64 on AMD\. The mapping is one to one only when`N`matches that width\. A smaller`N`leaves some lanes idle while a larger`N`gives some or all lanes more than one element to process\. When there is more work than the warp is wide, we need a way to say which lanes do what\. It helps to think of the warp as a small "machine" of its own: a fixed set of primitives for moving and combining data across lanes, plus invariants about which lanes are active and how much data each one holds\. "Programming" it means placing work onto lanes within those rules\. **At VectorWare, we give that machine an IR\.**Rather than a standalone data structure, we encode it in Rust's type system using types, generics, const generics, and trait bounds\. A program is composed of typed operations: ballots, shuffles, reductions, scans, gathers, scatters, atomics, and[strip mining](https://en.wikipedia.org/wiki/Loop_sectioning)for vectors wider than the warp\. Operands, execution shape, and capacity are typed too\. Because the operations carry their shape in the types, many invalid programs cannot be constructed at all\. The IR needs no interpreter on the GPU\. Each operation lowers straight to the corresponding instructions with zero cost over hand\-written PTX\.The same types let us run it on the CPU too\. We built a reference interpreter that executes the IR deterministically, a kind of[Miri](https://github.com/rust-lang/miri)for warp\-lane programming\. We use it to simulate GPU code and for[differential testing](https://en.wikipedia.org/wiki/Differential_testing)\. Our work targets NVIDIA today, but nothing here is CUDA specific\. AMD wavefronts and Vulkan[subgroups](https://docs.vulkan.org/guide/latest/subgroups.html)expose similar primitives and semantics\. The IR itself is architecture\-agnostic Rust\. ## Benefits The same source runs on the CPU and the GPU\. Code and libraries that already use portable SIMD become candidates for GPU execution without a rewrite\. Unmodified CPU code can use GPU lane\-level parallelism\. GPU\-aware code can still go further by using`core::arch`intrinsics that map directly to PTX\. A`Simd<T, N\>`is an ordinary owned value\. The borrow checker, lifetimes, and type checking apply to it exactly as they do on the CPU\. We are not adding a GPU\-specific vector type or a new set of annotations\. We are mapping Rust's existing portable SIMD onto the GPU's native execution model\. At[VectorWare](https://www.vectorware.com/), we are making GPUs behave like a normal Rust platform\. ## Downsides Portable SIMD is still unstable in Rust\. It requires the nightly`\#\!\[feature\(portable\_simd\)\]`, and its surface may change before it stabilizes\. Vectors narrower than the warp leave lanes idle, and vectors wider than the warp turn each operation into more instructions\. The abstraction is only zero cost when the vector width matches the number of warp lanes\. Not every cross\-lane operation maps to an efficient warp instruction\. Shuffles that match the hardware's supported patterns are cheap, but arbitrary permutations may need several instructions or a trip through shared memory\. Horizontal operations like reductions and`all`/`any`also act as synchronization points within the warp, which constrains how freely the scheduler can overlap work\. We had to change the compiler to make the abstraction sound when interacting with other Rust features\. As this is uncharted territory, we are not yet confident we have covered every case\. ## Future work With SIMD,[threads](https://www.vectorware.com/blog/threads-on-gpu), and[async](https://www.vectorware.com/blog/async-await-on-gpu)all mapped onto the GPU, the natural next step is composing them: threads spreading work across warps,`core::simd`spreading data across the lanes within each warp, and async structuring the concurrency between them\. We are also interested in lowering matrix\-shaped SIMD onto the GPU's[tensor cores](https://www.nvidia.com/en-us/data-center/tensor-cores/), and in auto\-vectorizing ordinary scalar Rust loops into`Simd`operations so that code gets warp\-level parallelism without being written against`core::simd`at all\. As[members of the Rust compiler team](https://www.vectorware.com/team), we are keen to explore how much of this can happen in the compiler itself\. A vector representation shared across the CPU and the GPU is valuable, though it is not clear that today's portable SIMD types are the right basis for one\. For one thing, they largely sit in a world of their own within the`core`and`std`APIs\. More exploration is necessary\. ## Is VectorWare only focused on Rust? The speed at which we are able to make progress on the GPU is a testament to the power of Rust's abstractions and ecosystem\. As a company, we understand that not everyone uses Rust\. Our future products will support multiple programming languages and runtimes\. However, we believe Rust is uniquely well suited to building high\-performance, reliable GPU\-native applications and that is what we are most excited about\. ## Follow along Follow us on[X](https://x.com/vectorware),[Bluesky](https://bsky.app/profile/vectorware.com),[LinkedIn](https://www.linkedin.com/company/vectorware/), or subscribe to our[blog](https://www.vectorware.com/blog)to stay updated on our progress\. We will be sharing more about our work in the coming months\. You can also reach us at[hello@vectorware\.com](mailto:[email protected])\.