Backport to 2.8: some FP8 support #3479

bernhardmgruber · 2025-01-22T16:01:14Z

This backports 3 dependent features:

add _CCCL_HAS_NVFP8 macro #3429 (had a wrong definition for the _CCCL_HAS_NVFP8 macro)
Add cuda::is_floating_point supporting half and bfloat #3379 (adds cuda::is_floating_point unit tests)
Specialize __is_extended_floating_point for FP8 types #3470 (fixes the _CCCL_HAS_NVFP8 and adds to cuda::is_floating_point unit tests)

Which is why I created a single backport PR for those.

Co-authored-by: Michael Schellenberger Costa <[email protected]>

Also ensure that we actually can enable FP8 due to FP16 and BF16 requirements Co-authored-by: Michael Schellenberger Costa <[email protected]>

github-actions · 2025-01-22T18:21:30Z

🟩 CI finished in 2h 18m: Pass: 100%/170 | Total: 3d 13h | Avg: 30m 13s | Max: 1h 22m | Hits: 199%/22580

🟩 libcudacxx: Pass: 100%/48 | Total: 17h 40m | Avg: 22m 05s | Max: 1h 01m | Hits: 324%/9876

🟩 cpu
  🟩 amd64              Pass: 100%/46  | Total: 16h 56m | Avg: 22m 05s | Max:  1h 01m | Hits: 324%/9876  
  🟩 arm64              Pass: 100%/2   | Total: 43m 35s | Avg: 21m 47s | Max: 21m 49s
🟩 ctk
  🟩 11.1               Pass: 100%/7   | Total: 53m 24s | Avg:  7m 37s | Max: 36m 03s | Hits: 329%/2286  
  🟩 12.5               Pass: 100%/2   | Total:  1h 13m | Avg: 36m 35s | Max: 37m 03s
  🟩 12.6               Pass: 100%/39  | Total: 15h 33m | Avg: 23m 56s | Max:  1h 01m | Hits: 323%/7590  
🟩 cudacxx
  🟩 ClangCUDA18        Pass: 100%/4   | Total:  1h 05m | Avg: 16m 26s | Max: 20m 29s
  🟩 nvcc11.1           Pass: 100%/7   | Total: 53m 24s | Avg:  7m 37s | Max: 36m 03s | Hits: 329%/2286  
  🟩 nvcc12.5           Pass: 100%/2   | Total:  1h 13m | Avg: 36m 35s | Max: 37m 03s
  🟩 nvcc12.6           Pass: 100%/35  | Total: 14h 27m | Avg: 24m 47s | Max:  1h 01m | Hits: 323%/7590  
🟩 cudacxx_family
  🟩 ClangCUDA          Pass: 100%/4   | Total:  1h 05m | Avg: 16m 26s | Max: 20m 29s
  🟩 nvcc               Pass: 100%/44  | Total: 16h 34m | Avg: 22m 35s | Max:  1h 01m | Hits: 324%/9876  
🟩 cxx
  🟩 Clang9             Pass: 100%/4   | Total: 47m 43s | Avg: 11m 55s | Max: 22m 26s
  🟩 Clang10            Pass: 100%/1   | Total: 26m 25s | Avg: 26m 25s | Max: 26m 25s
  🟩 Clang11            Pass: 100%/1   | Total: 23m 50s | Avg: 23m 50s | Max: 23m 50s
  🟩 Clang12            Pass: 100%/1   | Total: 23m 35s | Avg: 23m 35s | Max: 23m 35s
  🟩 Clang13            Pass: 100%/1   | Total: 23m 41s | Avg: 23m 41s | Max: 23m 41s
  🟩 Clang14            Pass: 100%/1   | Total: 22m 04s | Avg: 22m 04s | Max: 22m 04s
  🟩 Clang15            Pass: 100%/1   | Total: 22m 25s | Avg: 22m 25s | Max: 22m 25s
  🟩 Clang16            Pass: 100%/1   | Total: 22m 50s | Avg: 22m 50s | Max: 22m 50s
  🟩 Clang17            Pass: 100%/1   | Total: 23m 37s | Avg: 23m 37s | Max: 23m 37s
  🟩 Clang18            Pass: 100%/8   | Total:  3h 19m | Avg: 24m 58s | Max:  1h 01m
  🟩 GCC6               Pass: 100%/2   | Total:  5m 04s | Avg:  2m 32s | Max:  2m 45s
  🟩 GCC7               Pass: 100%/2   | Total: 33m 18s | Avg: 16m 39s | Max: 17m 07s
  🟩 GCC8               Pass: 100%/1   | Total: 21m 16s | Avg: 21m 16s | Max: 21m 16s
  🟩 GCC9               Pass: 100%/3   | Total: 30m 38s | Avg: 10m 12s | Max: 25m 16s
  🟩 GCC10              Pass: 100%/1   | Total: 23m 33s | Avg: 23m 33s | Max: 23m 33s
  🟩 GCC11              Pass: 100%/1   | Total: 24m 04s | Avg: 24m 04s | Max: 24m 04s
  🟩 GCC12              Pass: 100%/1   | Total: 27m 33s | Avg: 27m 33s | Max: 27m 33s
  🟩 GCC13              Pass: 100%/10  | Total:  3h 24m | Avg: 20m 26s | Max: 30m 31s
  🟩 Intel2023.2.0      Pass: 100%/1   | Total: 28m 58s | Avg: 28m 58s | Max: 28m 58s
  🟩 MSVC14.16          Pass: 100%/1   | Total: 36m 03s | Avg: 36m 03s | Max: 36m 03s | Hits: 329%/2286  
  🟩 MSVC14.29          Pass: 100%/1   | Total: 37m 48s | Avg: 37m 48s | Max: 37m 48s | Hits: 324%/2481  
  🟩 MSVC14.39          Pass: 100%/2   | Total:  1h 18m | Avg: 39m 08s | Max: 41m 00s | Hits: 322%/5109  
  🟩 NVHPC24.7          Pass: 100%/2   | Total:  1h 13m | Avg: 36m 35s | Max: 37m 03s
🟩 cxx_family
  🟩 Clang              Pass: 100%/20  | Total:  7h 15m | Avg: 21m 47s | Max:  1h 01m
  🟩 GCC                Pass: 100%/21  | Total:  6h 09m | Avg: 17m 36s | Max: 30m 31s
  🟩 Intel              Pass: 100%/1   | Total: 28m 58s | Avg: 28m 58s | Max: 28m 58s
  🟩 MSVC               Pass: 100%/4   | Total:  2h 32m | Avg: 38m 01s | Max: 41m 00s | Hits: 324%/9876  
  🟩 NVHPC              Pass: 100%/2   | Total:  1h 13m | Avg: 36m 35s | Max: 37m 03s
🟩 gpu
  🟩 v100               Pass: 100%/48  | Total: 17h 40m | Avg: 22m 05s | Max:  1h 01m | Hits: 324%/9876  
🟩 jobs
  🟩 Build              Pass: 100%/41  | Total: 14h 37m | Avg: 21m 24s | Max: 41m 00s | Hits: 324%/9876  
  🟩 NVRTC              Pass: 100%/4   | Total:  1h 43m | Avg: 25m 46s | Max: 30m 31s
  🟩 Test               Pass: 100%/2   | Total:  1h 17m | Avg: 38m 41s | Max:  1h 01m
  🟩 VerifyCodegen      Pass: 100%/1   | Total:  1m 59s | Avg:  1m 59s | Max:  1m 59s
🟩 sm
  🟩 90                 Pass: 100%/1   | Total: 12m 44s | Avg: 12m 44s | Max: 12m 44s
  🟩 90a                Pass: 100%/2   | Total: 31m 06s | Avg: 15m 33s | Max: 17m 20s
🟩 std
  🟩 11                 Pass: 100%/6   | Total:  1h 01m | Avg: 10m 11s | Max: 18m 28s
  🟩 14                 Pass: 100%/5   | Total:  1h 47m | Avg: 21m 35s | Max: 36m 03s | Hits: 329%/2286  
  🟩 17                 Pass: 100%/13  | Total:  5h 12m | Avg: 24m 00s | Max: 37m 48s | Hits: 324%/4962  
  🟩 20                 Pass: 100%/23  | Total:  9h 36m | Avg: 25m 04s | Max:  1h 01m | Hits: 321%/2628

🟩 cub: Pass: 100%/47 | Total: 1d 12h | Avg: 46m 52s | Max: 1h 15m | Hits: 27%/3132

🟩 cpu
  🟩 amd64              Pass: 100%/45  | Total:  1d 10h | Avg: 46m 13s | Max:  1h 15m | Hits:  27%/3132  
  🟩 arm64              Pass: 100%/2   | Total:  2h 02m | Avg:  1h 01m | Max:  1h 01m
🟩 ctk
  🟩 11.1               Pass: 100%/7   | Total:  1h 30m | Avg: 12m 51s | Max:  1h 03m | Hits:  30%/783   
  🟩 12.5               Pass: 100%/2   | Total:  2h 21m | Avg:  1h 10m | Max:  1h 14m
  🟩 12.6               Pass: 100%/38  | Total:  1d 08h | Avg: 51m 51s | Max:  1h 15m | Hits:  26%/2349  
🟩 cudacxx
  🟩 ClangCUDA18        Pass: 100%/2   | Total:  2h 08m | Avg:  1h 04m | Max:  1h 10m
  🟩 nvcc11.1           Pass: 100%/7   | Total:  1h 30m | Avg: 12m 51s | Max:  1h 03m | Hits:  30%/783   
  🟩 nvcc12.5           Pass: 100%/2   | Total:  2h 21m | Avg:  1h 10m | Max:  1h 14m
  🟩 nvcc12.6           Pass: 100%/36  | Total:  1d 06h | Avg: 51m 09s | Max:  1h 15m | Hits:  26%/2349  
🟩 cudacxx_family
  🟩 ClangCUDA          Pass: 100%/2   | Total:  2h 08m | Avg:  1h 04m | Max:  1h 10m
  🟩 nvcc               Pass: 100%/45  | Total:  1d 10h | Avg: 46m 05s | Max:  1h 15m | Hits:  27%/3132  
🟩 cxx
  🟩 Clang9             Pass: 100%/4   | Total:  2h 06m | Avg: 31m 37s | Max:  1h 05m
  🟩 Clang10            Pass: 100%/1   | Total: 54m 43s | Avg: 54m 43s | Max: 54m 43s
  🟩 Clang11            Pass: 100%/1   | Total: 59m 00s | Avg: 59m 00s | Max: 59m 00s
  🟩 Clang12            Pass: 100%/1   | Total: 59m 02s | Avg: 59m 02s | Max: 59m 02s
  🟩 Clang13            Pass: 100%/1   | Total:  1h 02m | Avg:  1h 02m | Max:  1h 02m
  🟩 Clang14            Pass: 100%/1   | Total:  1h 06m | Avg:  1h 06m | Max:  1h 06m
  🟩 Clang15            Pass: 100%/1   | Total: 57m 22s | Avg: 57m 22s | Max: 57m 22s
  🟩 Clang16            Pass: 100%/1   | Total: 53m 58s | Avg: 53m 58s | Max: 53m 58s
  🟩 Clang17            Pass: 100%/1   | Total: 53m 04s | Avg: 53m 04s | Max: 53m 04s
  🟩 Clang18            Pass: 100%/7   | Total:  5h 38m | Avg: 48m 22s | Max:  1h 10m
  🟩 GCC6               Pass: 100%/2   | Total:  9m 07s | Avg:  4m 33s | Max:  4m 35s
  🟩 GCC7               Pass: 100%/2   | Total:  1h 58m | Avg: 59m 15s | Max:  1h 00m
  🟩 GCC8               Pass: 100%/1   | Total:  1h 05m | Avg:  1h 05m | Max:  1h 05m
  🟩 GCC9               Pass: 100%/3   | Total:  1h 01m | Avg: 20m 22s | Max: 52m 31s
  🟩 GCC10              Pass: 100%/1   | Total: 57m 34s | Avg: 57m 34s | Max: 57m 34s
  🟩 GCC11              Pass: 100%/1   | Total: 56m 46s | Avg: 56m 46s | Max: 56m 46s
  🟩 GCC12              Pass: 100%/3   | Total:  1h 41m | Avg: 33m 54s | Max: 59m 24s
  🟩 GCC13              Pass: 100%/8   | Total:  5h 13m | Avg: 39m 12s | Max:  1h 01m
  🟩 Intel2023.2.0      Pass: 100%/1   | Total:  1h 03m | Avg:  1h 03m | Max:  1h 03m
  🟩 MSVC14.16          Pass: 100%/1   | Total:  1h 03m | Avg:  1h 03m | Max:  1h 03m | Hits:  30%/783   
  🟩 MSVC14.29          Pass: 100%/1   | Total:  1h 15m | Avg:  1h 15m | Max:  1h 15m | Hits:  26%/783   
  🟩 MSVC14.39          Pass: 100%/2   | Total:  2h 23m | Avg:  1h 11m | Max:  1h 13m | Hits:  26%/1566  
  🟩 NVHPC24.7          Pass: 100%/2   | Total:  2h 21m | Avg:  1h 10m | Max:  1h 14m
🟩 cxx_family
  🟩 Clang              Pass: 100%/19  | Total: 15h 31m | Avg: 49m 00s | Max:  1h 10m
  🟩 GCC                Pass: 100%/21  | Total: 13h 03m | Avg: 37m 19s | Max:  1h 05m
  🟩 Intel              Pass: 100%/1   | Total:  1h 03m | Avg:  1h 03m | Max:  1h 03m
  🟩 MSVC               Pass: 100%/4   | Total:  4h 42m | Avg:  1h 10m | Max:  1h 15m | Hits:  27%/3132  
  🟩 NVHPC              Pass: 100%/2   | Total:  2h 21m | Avg:  1h 10m | Max:  1h 14m
🟩 gpu
  🟩 h100               Pass: 100%/2   | Total: 42m 18s | Avg: 21m 09s | Max: 26m 04s
  🟩 v100               Pass: 100%/45  | Total:  1d 12h | Avg: 48m 00s | Max:  1h 15m | Hits:  27%/3132  
🟩 jobs
  🟩 Build              Pass: 100%/40  | Total:  1d 09h | Avg: 50m 48s | Max:  1h 15m | Hits:  27%/3132  
  🟩 DeviceLaunch       Pass: 100%/1   | Total: 29m 28s | Avg: 29m 28s | Max: 29m 28s
  🟩 GraphCapture       Pass: 100%/1   | Total: 25m 17s | Avg: 25m 17s | Max: 25m 17s
  🟩 HostLaunch         Pass: 100%/3   | Total: 57m 13s | Avg: 19m 04s | Max: 21m 18s
  🟩 TestGPU            Pass: 100%/2   | Total: 58m 21s | Avg: 29m 10s | Max: 35m 50s
🟩 sm
  🟩 90                 Pass: 100%/2   | Total: 42m 18s | Avg: 21m 09s | Max: 26m 04s
  🟩 90a                Pass: 100%/1   | Total: 25m 29s | Avg: 25m 29s | Max: 25m 29s
🟩 std
  🟩 11                 Pass: 100%/5   | Total:  2h 16m | Avg: 27m 16s | Max:  1h 05m
  🟩 14                 Pass: 100%/4   | Total:  3h 00m | Avg: 45m 09s | Max:  1h 03m | Hits:  30%/783   
  🟩 17                 Pass: 100%/12  | Total: 10h 45m | Avg: 53m 45s | Max:  1h 15m | Hits:  26%/1566  
  🟩 20                 Pass: 100%/26  | Total: 20h 40m | Avg: 47m 43s | Max:  1h 13m | Hits:  27%/783

🟩 thrust: Pass: 100%/46 | Total: 1d 00h | Avg: 31m 51s | Max: 1h 22m | Hits: 128%/9260

🟩 cmake_options
  🟩 -DTHRUST_DISPATCH_TYPE=Force32bit Pass: 100%/2   | Total: 40m 31s | Avg: 20m 15s | Max: 27m 49s
🟩 cpu
  🟩 amd64              Pass: 100%/44  | Total: 23h 24m | Avg: 31m 55s | Max:  1h 22m | Hits: 128%/9260  
  🟩 arm64              Pass: 100%/2   | Total:  1h 00m | Avg: 30m 29s | Max: 32m 25s
🟩 ctk
  🟩 11.1               Pass: 100%/7   | Total:  1h 28m | Avg: 12m 37s | Max:  1h 03m | Hits:  88%/1852  
  🟩 12.5               Pass: 100%/2   | Total:  2h 23m | Avg:  1h 11m | Max:  1h 16m
  🟩 12.6               Pass: 100%/37  | Total: 20h 34m | Avg: 33m 21s | Max:  1h 22m | Hits: 138%/7408  
🟩 cudacxx
  🟩 ClangCUDA18        Pass: 100%/2   | Total: 56m 39s | Avg: 28m 19s | Max: 28m 32s
  🟩 nvcc11.1           Pass: 100%/7   | Total:  1h 28m | Avg: 12m 37s | Max:  1h 03m | Hits:  88%/1852  
  🟩 nvcc12.5           Pass: 100%/2   | Total:  2h 23m | Avg:  1h 11m | Max:  1h 16m
  🟩 nvcc12.6           Pass: 100%/35  | Total: 19h 37m | Avg: 33m 38s | Max:  1h 22m | Hits: 138%/7408  
🟩 cudacxx_family
  🟩 ClangCUDA          Pass: 100%/2   | Total: 56m 39s | Avg: 28m 19s | Max: 28m 32s
  🟩 nvcc               Pass: 100%/44  | Total: 23h 29m | Avg: 32m 01s | Max:  1h 22m | Hits: 128%/9260  
🟩 cxx
  🟩 Clang9             Pass: 100%/4   | Total:  1h 15m | Avg: 18m 54s | Max: 37m 23s
  🟩 Clang10            Pass: 100%/1   | Total: 34m 13s | Avg: 34m 13s | Max: 34m 13s
  🟩 Clang11            Pass: 100%/1   | Total: 31m 28s | Avg: 31m 28s | Max: 31m 28s
  🟩 Clang12            Pass: 100%/1   | Total: 31m 40s | Avg: 31m 40s | Max: 31m 40s
  🟩 Clang13            Pass: 100%/1   | Total: 37m 20s | Avg: 37m 20s | Max: 37m 20s
  🟩 Clang14            Pass: 100%/1   | Total: 33m 24s | Avg: 33m 24s | Max: 33m 24s
  🟩 Clang15            Pass: 100%/1   | Total: 30m 24s | Avg: 30m 24s | Max: 30m 24s
  🟩 Clang16            Pass: 100%/1   | Total: 34m 08s | Avg: 34m 08s | Max: 34m 08s
  🟩 Clang17            Pass: 100%/1   | Total: 30m 15s | Avg: 30m 15s | Max: 30m 15s
  🟩 Clang18            Pass: 100%/7   | Total:  2h 51m | Avg: 24m 28s | Max: 33m 39s
  🟩 GCC6               Pass: 100%/2   | Total:  8m 12s | Avg:  4m 06s | Max:  4m 10s
  🟩 GCC7               Pass: 100%/2   | Total: 58m 00s | Avg: 29m 00s | Max: 33m 25s
  🟩 GCC8               Pass: 100%/1   | Total: 30m 52s | Avg: 30m 52s | Max: 30m 52s
  🟩 GCC9               Pass: 100%/3   | Total: 40m 10s | Avg: 13m 23s | Max: 31m 51s
  🟩 GCC10              Pass: 100%/1   | Total: 32m 32s | Avg: 32m 32s | Max: 32m 32s
  🟩 GCC11              Pass: 100%/1   | Total: 34m 38s | Avg: 34m 38s | Max: 34m 38s
  🟩 GCC12              Pass: 100%/1   | Total: 34m 21s | Avg: 34m 21s | Max: 34m 21s
  🟩 GCC13              Pass: 100%/8   | Total:  3h 01m | Avg: 22m 40s | Max: 35m 16s
  🟩 Intel2023.2.0      Pass: 100%/1   | Total: 49m 46s | Avg: 49m 46s | Max: 49m 46s
  🟩 MSVC14.16          Pass: 100%/1   | Total:  1h 03m | Avg:  1h 03m | Max:  1h 03m | Hits:  88%/1852  
  🟩 MSVC14.29          Pass: 100%/1   | Total:  1h 22m | Avg:  1h 22m | Max:  1h 22m | Hits:  63%/1852  
  🟩 MSVC14.39          Pass: 100%/3   | Total:  3h 17m | Avg:  1h 05m | Max:  1h 22m | Hits: 164%/5556  
  🟩 NVHPC24.7          Pass: 100%/2   | Total:  2h 23m | Avg:  1h 11m | Max:  1h 16m
🟩 cxx_family
  🟩 Clang              Pass: 100%/19  | Total:  8h 29m | Avg: 26m 49s | Max: 37m 23s
  🟩 GCC                Pass: 100%/19  | Total:  7h 00m | Avg: 22m 06s | Max: 35m 16s
  🟩 Intel              Pass: 100%/1   | Total: 49m 46s | Avg: 49m 46s | Max: 49m 46s
  🟩 MSVC               Pass: 100%/5   | Total:  5h 42m | Avg:  1h 08m | Max:  1h 22m | Hits: 128%/9260  
  🟩 NVHPC              Pass: 100%/2   | Total:  2h 23m | Avg:  1h 11m | Max:  1h 16m
🟩 gpu
  🟩 v100               Pass: 100%/46  | Total:  1d 00h | Avg: 31m 51s | Max:  1h 22m | Hits: 128%/9260  
🟩 jobs
  🟩 Build              Pass: 100%/40  | Total: 22h 55m | Avg: 34m 22s | Max:  1h 22m | Hits:  69%/7408  
  🟩 TestCPU            Pass: 100%/3   | Total: 52m 55s | Avg: 17m 38s | Max: 36m 55s | Hits: 365%/1852  
  🟩 TestGPU            Pass: 100%/3   | Total: 37m 27s | Avg: 12m 29s | Max: 13m 20s
🟩 sm
  🟩 90a                Pass: 100%/1   | Total: 18m 50s | Avg: 18m 50s | Max: 18m 50s
🟩 std
  🟩 11                 Pass: 100%/5   | Total:  1h 05m | Avg: 13m 11s | Max: 29m 35s
  🟩 14                 Pass: 100%/4   | Total:  2h 18m | Avg: 34m 32s | Max:  1h 03m | Hits:  88%/1852  
  🟩 17                 Pass: 100%/12  | Total:  8h 01m | Avg: 40m 09s | Max:  1h 22m | Hits:  63%/3704  
  🟩 20                 Pass: 100%/23  | Total: 12h 19m | Avg: 32m 08s | Max:  1h 17m | Hits: 214%/3704

🟩 cudax: Pass: 100%/26 | Total: 6h 15m | Avg: 14m 26s | Max: 19m 49s | Hits: 62%/312

🟩 cpu
  🟩 amd64              Pass: 100%/22  | Total:  5h 21m | Avg: 14m 36s | Max: 19m 49s | Hits:  62%/312   
  🟩 arm64              Pass: 100%/4   | Total: 54m 04s | Avg: 13m 31s | Max: 14m 37s
🟩 ctk
  🟩 12.0               Pass: 100%/3   | Total: 40m 49s | Avg: 13m 36s | Max: 15m 00s | Hits:  62%/156   
  🟩 12.5               Pass: 100%/2   | Total: 19m 47s | Avg:  9m 53s | Max: 10m 10s
  🟩 12.6               Pass: 100%/21  | Total:  5h 14m | Avg: 14m 59s | Max: 19m 49s | Hits:  62%/156   
🟩 cudacxx
  🟩 nvcc12.0           Pass: 100%/3   | Total: 40m 49s | Avg: 13m 36s | Max: 15m 00s | Hits:  62%/156   
  🟩 nvcc12.5           Pass: 100%/2   | Total: 19m 47s | Avg:  9m 53s | Max: 10m 10s
  🟩 nvcc12.6           Pass: 100%/21  | Total:  5h 14m | Avg: 14m 59s | Max: 19m 49s | Hits:  62%/156   
🟩 cudacxx_family
  🟩 nvcc               Pass: 100%/26  | Total:  6h 15m | Avg: 14m 26s | Max: 19m 49s | Hits:  62%/312   
🟩 cxx
  🟩 Clang9             Pass: 100%/1   | Total: 13m 35s | Avg: 13m 35s | Max: 13m 35s
  🟩 Clang10            Pass: 100%/1   | Total: 14m 34s | Avg: 14m 34s | Max: 14m 34s
  🟩 Clang11            Pass: 100%/1   | Total: 16m 21s | Avg: 16m 21s | Max: 16m 21s
  🟩 Clang12            Pass: 100%/1   | Total: 13m 37s | Avg: 13m 37s | Max: 13m 37s
  🟩 Clang13            Pass: 100%/1   | Total: 14m 39s | Avg: 14m 39s | Max: 14m 39s
  🟩 Clang14            Pass: 100%/1   | Total: 13m 57s | Avg: 13m 57s | Max: 13m 57s
  🟩 Clang15            Pass: 100%/1   | Total: 17m 29s | Avg: 17m 29s | Max: 17m 29s
  🟩 Clang16            Pass: 100%/1   | Total: 16m 43s | Avg: 16m 43s | Max: 16m 43s
  🟩 Clang17            Pass: 100%/1   | Total: 14m 55s | Avg: 14m 55s | Max: 14m 55s
  🟩 Clang18            Pass: 100%/4   | Total: 59m 11s | Avg: 14m 47s | Max: 16m 30s
  🟩 GCC9               Pass: 100%/1   | Total: 15m 00s | Avg: 15m 00s | Max: 15m 00s
  🟩 GCC10              Pass: 100%/1   | Total: 15m 47s | Avg: 15m 47s | Max: 15m 47s
  🟩 GCC11              Pass: 100%/1   | Total: 16m 43s | Avg: 16m 43s | Max: 16m 43s
  🟩 GCC12              Pass: 100%/2   | Total: 38m 37s | Avg: 19m 18s | Max: 19m 49s
  🟩 GCC13              Pass: 100%/4   | Total: 49m 49s | Avg: 12m 27s | Max: 14m 37s
  🟩 MSVC14.36          Pass: 100%/1   | Total: 12m 14s | Avg: 12m 14s | Max: 12m 14s | Hits:  62%/156   
  🟩 MSVC14.39          Pass: 100%/1   | Total: 12m 23s | Avg: 12m 23s | Max: 12m 23s | Hits:  62%/156   
  🟩 NVHPC24.7          Pass: 100%/2   | Total: 19m 47s | Avg:  9m 53s | Max: 10m 10s
🟩 cxx_family
  🟩 Clang              Pass: 100%/13  | Total:  3h 15m | Avg: 15m 00s | Max: 17m 29s
  🟩 GCC                Pass: 100%/9   | Total:  2h 15m | Avg: 15m 06s | Max: 19m 49s
  🟩 MSVC               Pass: 100%/2   | Total: 24m 37s | Avg: 12m 18s | Max: 12m 23s | Hits:  62%/312   
  🟩 NVHPC              Pass: 100%/2   | Total: 19m 47s | Avg:  9m 53s | Max: 10m 10s
🟩 gpu
  🟩 v100               Pass: 100%/26  | Total:  6h 15m | Avg: 14m 26s | Max: 19m 49s | Hits:  62%/312   
🟩 jobs
  🟩 Build              Pass: 100%/24  | Total:  5h 39m | Avg: 14m 07s | Max: 18m 48s | Hits:  62%/312   
  🟩 Test               Pass: 100%/2   | Total: 36m 19s | Avg: 18m 09s | Max: 19m 49s
🟩 sm
  🟩 90                 Pass: 100%/1   | Total:  9m 52s | Avg:  9m 52s | Max:  9m 52s
  🟩 90a                Pass: 100%/1   | Total: 12m 23s | Avg: 12m 23s | Max: 12m 23s
🟩 std
  🟩 17                 Pass: 100%/6   | Total:  1h 13m | Avg: 12m 17s | Max: 15m 00s
  🟩 20                 Pass: 100%/20  | Total:  5h 01m | Avg: 15m 04s | Max: 19m 49s | Hits:  62%/312

🟩 cccl_c_parallel: Pass: 100%/2 | Total: 9m 47s | Avg: 4m 53s | Max: 7m 34s

🟩 cpu
  🟩 amd64              Pass: 100%/2   | Total:  9m 47s | Avg:  4m 53s | Max:  7m 34s
🟩 ctk
  🟩 12.6               Pass: 100%/2   | Total:  9m 47s | Avg:  4m 53s | Max:  7m 34s
🟩 cudacxx
  🟩 nvcc12.6           Pass: 100%/2   | Total:  9m 47s | Avg:  4m 53s | Max:  7m 34s
🟩 cudacxx_family
  🟩 nvcc               Pass: 100%/2   | Total:  9m 47s | Avg:  4m 53s | Max:  7m 34s
🟩 cxx
  🟩 GCC13              Pass: 100%/2   | Total:  9m 47s | Avg:  4m 53s | Max:  7m 34s
🟩 cxx_family
  🟩 GCC                Pass: 100%/2   | Total:  9m 47s | Avg:  4m 53s | Max:  7m 34s
🟩 gpu
  🟩 v100               Pass: 100%/2   | Total:  9m 47s | Avg:  4m 53s | Max:  7m 34s
🟩 jobs
  🟩 Build              Pass: 100%/1   | Total:  2m 13s | Avg:  2m 13s | Max:  2m 13s
  🟩 Test               Pass: 100%/1   | Total:  7m 34s | Avg:  7m 34s | Max:  7m 34s

🟩 python: Pass: 100%/1 | Total: 23m 51s | Avg: 23m 51s | Max: 23m 51s

🟩 cpu
  🟩 amd64              Pass: 100%/1   | Total: 23m 51s | Avg: 23m 51s | Max: 23m 51s
🟩 ctk
  🟩 12.6               Pass: 100%/1   | Total: 23m 51s | Avg: 23m 51s | Max: 23m 51s
🟩 cudacxx
  🟩 nvcc12.6           Pass: 100%/1   | Total: 23m 51s | Avg: 23m 51s | Max: 23m 51s
🟩 cudacxx_family
  🟩 nvcc               Pass: 100%/1   | Total: 23m 51s | Avg: 23m 51s | Max: 23m 51s
🟩 cxx
  🟩 GCC13              Pass: 100%/1   | Total: 23m 51s | Avg: 23m 51s | Max: 23m 51s
🟩 cxx_family
  🟩 GCC                Pass: 100%/1   | Total: 23m 51s | Avg: 23m 51s | Max: 23m 51s
🟩 gpu
  🟩 v100               Pass: 100%/1   | Total: 23m 51s | Avg: 23m 51s | Max: 23m 51s
🟩 jobs
  🟩 Test               Pass: 100%/1   | Total: 23m 51s | Avg: 23m 51s | Max: 23m 51s

👃 Inspect Changes

Modifications in project?

	Project
	CCCL Infrastructure
+/-	libcu++
	CUB
	Thrust
	CUDA Experimental
	python
	CCCL C Parallel Library
	Catch2Helper

Modifications in project or dependencies?

	Project
	CCCL Infrastructure
+/-	libcu++
+/-	CUB
+/-	Thrust
+/-	CUDA Experimental
+/-	python
+/-	CCCL C Parallel Library
+/-	Catch2Helper

🏃‍ Runner counts (total jobs: 170)

#	Runner
125	`linux-amd64-cpu16`
19	`linux-amd64-gpu-v100-latest-1`
15	`windows-amd64-cpu16`
10	`linux-arm64-cpu16`
1	`linux-amd64-gpu-h100-latest-1-testing`

fbusato and others added 3 commits January 22, 2025 16:52

add _CCCL_HAS_NVFP8 macro (NVIDIA#3429)

d2ca2b6

Add cuda::is_floating_point supporting half and bfloat (NVIDIA#3379)

432a060

Co-authored-by: Michael Schellenberger Costa <[email protected]>

Specialize __is_extended_floating_point for FP8 types (NVIDIA#3470)

43e80f7

Also ensure that we actually can enable FP8 due to FP16 and BF16 requirements Co-authored-by: Michael Schellenberger Costa <[email protected]>

bernhardmgruber requested review from a team as code owners January 22, 2025 16:01

bernhardmgruber requested review from ericniebler and gonidelis January 22, 2025 16:01

miscco approved these changes Jan 22, 2025

View reviewed changes

miscco merged commit 51b08b0 into NVIDIA:branch/2.8.x Jan 22, 2025
186 checks passed

bernhardmgruber deleted the backport_fp8_macro branch January 22, 2025 19:26

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

Backport to 2.8: some FP8 support #3479

Backport to 2.8: some FP8 support #3479

bernhardmgruber commented Jan 22, 2025

github-actions bot commented Jan 22, 2025

🟩 libcudacxx: Pass: 100%/48 | Total: 17h 40m | Avg: 22m 05s | Max: 1h 01m | Hits: 324%/9876

🟩 cub: Pass: 100%/47 | Total: 1d 12h | Avg: 46m 52s | Max: 1h 15m | Hits: 27%/3132

🟩 thrust: Pass: 100%/46 | Total: 1d 00h | Avg: 31m 51s | Max: 1h 22m | Hits: 128%/9260

🟩 cudax: Pass: 100%/26 | Total: 6h 15m | Avg: 14m 26s | Max: 19m 49s | Hits: 62%/312

🟩 cccl_c_parallel: Pass: 100%/2 | Total: 9m 47s | Avg: 4m 53s | Max: 7m 34s

🟩 python: Pass: 100%/1 | Total: 23m 51s | Avg: 23m 51s | Max: 23m 51s

👃 Inspect Changes

Modifications in project?

Modifications in project or dependencies?

🏃‍ Runner counts (total jobs: 170)

Backport to 2.8: some FP8 support #3479

Backport to 2.8: some FP8 support #3479

Conversation

bernhardmgruber commented Jan 22, 2025

github-actions bot commented Jan 22, 2025

🟩 libcudacxx: Pass: 100%/48 | Total: 17h 40m | Avg: 22m 05s | Max: 1h 01m | Hits: 324%/9876

🟩 cub: Pass: 100%/47 | Total: 1d 12h | Avg: 46m 52s | Max: 1h 15m | Hits: 27%/3132

🟩 thrust: Pass: 100%/46 | Total: 1d 00h | Avg: 31m 51s | Max: 1h 22m | Hits: 128%/9260

🟩 cudax: Pass: 100%/26 | Total: 6h 15m | Avg: 14m 26s | Max: 19m 49s | Hits: 62%/312

🟩 cccl_c_parallel: Pass: 100%/2 | Total: 9m 47s | Avg: 4m 53s | Max: 7m 34s

🟩 python: Pass: 100%/1 | Total: 23m 51s | Avg: 23m 51s | Max: 23m 51s

👃 Inspect Changes

Modifications in project?

Modifications in project or dependencies?

🏃‍ Runner counts (total jobs: 170)