v2.0.19

Fix: correct CUDA regression introduced by FP8-FNUZ fix

Fixed a CUDA regression introduced by the previous FP8-FNUZ fix.

v2.0.18

Fix: resolve FP8-FNUZ bug on AMD CDNA3

On CDNA3, the FP8 format behavior differs from the commonly assumed definition, which could cause NaN generation during upward rounding.
This change avoids that issue.

v2.0.17

Fix: constant in find_max.hpp

v2.0.16

modified test_flops.hpp

v2.0.15

Modified test programs for HIP environments

v2.0.14

Add: execution of cuBLAS Ozaki-I and BF16x9 in test_watt.hpp

Updated the test program for measuring watt and GFLOPS/watt

v2.0.13

Fix an illegal memory access observed in test_flops.hpp after the Ozaki-I path on an NVIDIA B200 SXM 192GB system with cuBLAS 13.1.80.

Reported by T. Yamashita from SAKURA internet Inc.

v2.0.12

Add: test for Zgemm3m in sample program

v2.0.11

Fix: compilation error in test programs using cuBLAS version < 13.1

v2.0.10

Fix: HIP execution via INT8

fixed bugs in INT8-based emulation on HIP