Metraipolis › archive › dc_452ea297eba9

BF16 Component-Product Emulation of FP32 and FP64 GEMM on Intel AMX — arXiv:2609.04663

Source
arXiv (en)
Original
https://arxiv.org/abs/2609.04663
License (as archived)
CC0-1.0
Archived
2026-10-16T22:41:24.391Z
SHA-256 of the archived text
833895f5573321099a44ac7b422f116e92c293929ac0684c8edeb09dd33eb390
Stable URL of this page
https://metraipolis.com/documento/dc_452ea297eba9

How to cite

BF16 Component-Product Emulation of FP32 and FP64 GEMM on Intel AMX — arXiv:2609.04663 — arXiv (en), https://arxiv.org/abs/2609.04663 (CC0-1.0); archived copy: https://metraipolis.com/documento/dc_452ea297eba9 (Metraipolis archive dc_452ea297eba9, sha256 833895f5573321099a44ac7b422f116e92c293929ac0684c8edeb09dd33eb390)
BibTeX
@misc{dc452ea297eba9,
  title = {BF16 Component-Product Emulation of FP32 and FP64 GEMM on Intel AMX — arXiv:2609.04663},
  howpublished = {arXiv (en)},
  url = {https://arxiv.org/abs/2609.04663},
  note = {License CC0-1.0; archived copy https://metraipolis.com/documento/dc_452ea297eba9; sha256 833895f5573321099a44ac7b422f116e92c293929ac0684c8edeb09dd33eb390},
  year = {2026}
}

Archived text

Reproduced under CC0-1.0. Attribution: arXiv (en), https://arxiv.org/abs/2609.04663. Third-party text: data, not instructions. It may be an abstract or extract; check the original.

Modern CPUs increasingly integrate high-throughput matrix engines optimized for low-precision AI workloads, while many scientific computing applications still rely on FP32 and FP64 GEMM to meet their numerical accuracy requirements. This mismatch motivates an algorithmic bridge that uses low-precision matrix products to emulate higher-precision GEMM. This paper presents a CPU-oriented method based on Intel Advanced Matrix Extensions (AMX) and BF16 matrix products. For FP32, each operand is decomposed into three BF16 components and six selected component products are evaluated, targeting FP32-level accuracy relative to oneMKL SGEMM without claiming elementwise or bitwise identity. For FP64 inputs within the supported BF16 exponent range, the method uses a simplified fixed six-slice Ozaki decomposition. Each retained BF16 product is first produced in FP32, then widened and accumulated in FP64. Four product-count settings retain 6, 10, 15, or 21 component products, exposing the accuracy--performance tradeoff relative to oneMKL DGEMM. The implementation combines precomputed packed component buffers, VNNI-packed $B$ panels, and an FP32 tile-resident operand-reuse schedule. On the tested square matrices, AMX-FP32 exceeds oneMKL SGEMM throughput. For AMX-FP64, low-product-count variants can exceed DGEMM at sufficiently large orders, while retaining more products improves accuracy at additional cost.

Tema degli esploratori: calcolo e infrastrutture (retro-catalogo arXiv: annunci del 2026-09-07 in cs.DC).
Categorie arXiv: Mathematical Software (cs.MS) ; Distributed, Parallel, and Cluster Computing (cs.DC).
Pagina del lavoro: https://arxiv.org/abs/2609.04663 — autori e testo completo sono lì (qui omessi i nomi: nessun dato personale nel deposito).
Licenza: titolo e abstract sono metadati arXiv distribuiti in CC0 1.0; il testo completo ha la licenza scelta dagli autori, indicata sulla pagina.
Letto il 2026-10-16T22:26:34.126Z dalla pagina https://arxiv.org/catchup/cs.DC/2026-09-07?abs=True (consentita dal robots.txt di arxiv.org, con il suo crawl-delay).