1. [Publications](/publications)
2. Providing Accuracy Guarantees for the Fixed-Point Emulation of Matrix Multiplication

 # Providing Accuracy Guarantees for the Fixed-Point Emulation of Matrix Multiplication

  ![Publication image](/sites/default/files/styles/wide/public/default_images/default.jpeg?itok=TfIobf92 "Publication image")

 Fixed-point methods such as Ozaki-II emulate double-precision matrix multiplication (GEMM) on fast low-precision hardware. Software that chooses the mantissa bit count automatically is numerically safe when the emulated product keeps the error guarantee of the floating-point GEMM it replaces. We analyze the worst-case and average-case error of fixed-point and floating-point dot products, define exponent span capacities that determine the extra mantissa bits an input requires, and approximate them with a low-precision GEMM. We prove that 53 bits plus such a capacity and a small constant preserve the classical worst-case error bound for a floating-point dot product, and we demonstrate this guarantee in software on several test cases. In a developer build of cuBLAS on an NVIDIA B200, each implemented bit-count rule earns grade~A on Test~2c of ``Grading the BLAS’' at every size and exponent range tested, falling back to double precision at the widest ranges; in exact arithmetic, the bound holds on every adversarial and random input tested.

 ## Authors

Cole Brower (NVIDIA)

Peter Caday (NVIDIA)

John Gunnels (NVIDIA)

 ## Publication Date

Friday, November 13, 2026

 ## Research Area

[Algorithms and Numerical Methods](/research-area/algorithms)

[High Performance Computing](/research-area/high-performance-computing)

 ## Uploaded Files

[ESC-paper.pdf](https://d1qx31qr3h6wln.cloudfront.net/publications/ESC-paper.pdf?VersionId=HACrCI7HiLJI9bygLOzrlQ1ZHJ05dvGp "Open file in new window")610.91 KB