Add AVX10/v2 FP8 to FP6 and vice-versa conversions.
Enables conversions between FP8 and FP6 floating-point formats using AVX10/v2 intrinsics.
This commit introduces support for converting between FP8 (8-bit floating-point) and FP6 (6-bit floating-point) formats using new AVX10/v2 intrinsics. It adds numerous intrinsic functions for various conversion scenarios, including masked and unmasked operations, across different vector sizes (128, 256, and 512 bits). New test cases are included to verify the functionality of these conversions.
In Details
This patch expands the AVX10/v2 intrinsic support in GCC by adding conversions between FP8 (Brain Floating Point 8-bit, BF8) and FP6 (a proposed 6-bit format) floating-point types. It introduces new intrinsics in avx10v2auxintrin.h, updates the i386-builtin.def and sse.md files to handle these new operations, and includes corresponding test cases. These additions enable developers to leverage hardware acceleration for manipulating these newer, lower-precision floating-point formats on supported x86 architectures.
- FP8
- An 8-bit floating-point format, increasingly used in machine learning and AI accelerators for its reduced memory footprint and computational cost compared to FP16 or FP32.
- FP6
- A proposed 6-bit floating-point format, offering even greater data density than FP8, primarily for specialized high-performance computing and AI workloads.
- AVX10/v2
- An extension to the AVX (Advanced Vector Extensions) instruction set for x86 processors, introducing new features and capabilities, including support for newer floating-point formats like FP8.
- Intrinsics
- Special functions or constructs provided by a compiler that allow programmers to access low-level hardware instructions directly, often for performance-critical operations like vector arithmetic.
- BF8
- Brain Floating Point 8-bit format, a specific type of 8-bit floating-point representation designed to balance precision and range, often used in deep learning inference and training.