application precision

Author	Message
TRINITAS Send message Joined: 27 Jan 24 Posts: 3 Credit: 64,400 RAC: 0	Message 3637 - Posted: 27 Jan 2024, 22:08:37 UTC What is the calculation precision used by the project application? I would say FP32, but not... Thanks :) ID: 3637 · Rating: 0 · rate: / Reply Quote

Eric Driver Project administrator Project developer Project tester Project scientist Send message Joined: 8 Jul 11 Posts: 1424 Credit: 779,283,209 RAC: 770,399	Message 3638 - Posted: 28 Jan 2024, 5:32:17 UTC - in response to Message 3637. What is the calculation precision used by the project application? I would say FP32, but not... Thanks :) It's integer, arbitrary precision. ID: 3638 · Rating: 0 · rate: / Reply Quote

TRINITAS Send message Joined: 27 Jan 24 Posts: 3 Credit: 64,400 RAC: 0	Message 3639 - Posted: 28 Jan 2024, 10:26:57 UTC - in response to Message 3638. INT32 or INT64? :) ID: 3639 · Rating: 0 · rate: / Reply Quote

Eric Driver Project administrator Project developer Project tester Project scientist Send message Joined: 8 Jul 11 Posts: 1424 Credit: 779,283,209 RAC: 770,399	Message 3640 - Posted: 28 Jan 2024, 16:04:38 UTC - in response to Message 3639. INT32 or INT64? :) Arbitrary, sometimes called multi-precision. It uses the gmp library which allows variables to grow arbitrary large. For the GPU app I had to hard code a size so I used 15 64bit longs which gives 960 bits. ID: 3640 · Rating: 0 · rate: / Reply Quote

TRINITAS Send message Joined: 27 Jan 24 Posts: 3 Credit: 64,400 RAC: 0	Message 3641 - Posted: 28 Jan 2024, 18:22:21 UTC - in response to Message 3640. Ok, i see: a mixed precision INT32/64 This makes Nvidia more suitable, due to their SIMD16 INT32 partitions ID: 3641 · Rating: 0 · rate: / Reply Quote