Suggestions:
- Retest on the latest CUDA version (12.8.1 currently)
- If behavior is unchanged, file a bug.
Related topics
| Topic | Replies | Views | Activity | |
|---|---|---|---|---|
| __float_as_int compiler bug | 8 | 5043 | September 10, 2008 | |
| Changing from float to double generate wrong result Help needed, I am frustrated | 6 | 5785 | August 30, 2009 | |
| Troubles converting float array to float2 array | 11 | 3809 | July 3, 2017 | |
| FP8 WMMA kernel compilation error | 9 | 2278 | March 26, 2023 | |
| Fp8 conversion performance makes it slower than float16 | 7 | 952 | March 3, 2025 | |
| Tackling Floating point differences caused by CUDA tool kits | 2 | 554 | July 6, 2021 | |
| Rs rounding for conversion from FP32 to FP4(E2M1) | 2 | 359 | July 21, 2025 | |
| Cublas fp8 cublasLtMatmulAlgoGetHeuristic returns 0 - nvcc issue | 1 | 750 | July 21, 2023 | |
| Tackling Floating point differences caused by CUDA tool kits | 1 | 678 | July 6, 2021 | |
| Possible bug in cuda behaviour | 3 | 137 | November 29, 2024 |