-
If quad-precision takes off, then the current `uint8` for the itemsize cannot represent the `128+128 = 256` bits of a complex quad number. Even now that would fail for most `long double` storage form…
-
Bill Putman recently ran into an issue where DP gave bad results with the stretch provided by Schmidt transform.
The improvement is to use higher precision (where available) and use a different val…
-
When I use instant ID to change faces in the torch2.31+CUDA121 environment, I encounter OutOfCacheError: CUDA out of memory Tried to allocate 256.00 MiB GPU, The image size is 1024px * 1024px, and my …
-
It appears that sigma always assumes double-precision real data. Are there plans to support, for example, complex data? (Support for double-double and/or quad-precision would also be very useful.)
-
- Replace real var types in `nrtype` and throughout.
- Make new underscore variable type (e.g., `_mrk`).
- Make all quad precision the same user-defined (no need for a distinction between quad and dou…
-
Original ticket http://projects.scipy.org/numpy/ticket/2040
Reported 2012-02-07 by atmention:inducer, assigned to unknown.
As is visible from this code:
https://github.com/numpy/numpy/blob/master/nu…
-
### Reproducing code example:
```python
import numpy as np
np.exp(8j * np.arctan(np.array(1, dtype=np.float16))).imag
np.exp…
-
For the first-cut implementation of printing various data types, we relied on upcasting or truncating various non-standard data types. At the moment we have the following (based on my understanding of…
-
For each model, list the range of q, r, etc. for which the model is accurate to within epsilon.
Use the same epsilon across the models (e.g., 5e-5 for single precision, 5e-10 for double precision).…
-
I think this is feasible using float texture, which can be quite useful for high precision simulation.