Allan Bazinet
efbd9f94fc
Spelling
2025-02-12 17:02:40 -08:00
Allan Bazinet
713193be93
Perform decoding at a fixed depth of 2 in all cases
...
Removed depth menu; eliminates use of OSD decoder.
2025-02-12 16:01:07 -08:00
Allan Bazinet
ba897b6fd6
Increase hash entropy
2025-02-11 09:09:55 -08:00
Allan Bazinet
ca01a45caa
Use C++20 default equality operator
2025-02-09 06:01:47 -08:00
Allan Bazinet
5e3c08c5c4
Correct includes for concepts
2025-02-09 05:47:06 -08:00
Allan Bazinet
93c6a2eebd
C++20 version of Kahan summation
2025-02-09 05:44:33 -08:00
Allan Bazinet
8343869ec8
Grammar
2025-02-07 06:14:11 -08:00
Allan Bazinet
e7fba9b8aa
Use more descriptive variable names
2025-02-07 06:12:27 -08:00
Allan Bazinet
df0f435a82
Compute baseline constants at compile time
2025-02-07 03:56:25 -08:00
Allan Bazinet
801f3c1414
3 quick tests for C++20 enablement
2025-02-06 10:01:43 -08:00
Allan Bazinet
7770f48f24
Hoist end computation / usage, note constants
2025-02-05 06:31:02 -08:00
Allan Bazinet
58178211c2
Make index mapping more obvious
2025-02-05 01:58:40 -08:00
Allan Bazinet
79ce7daa90
Commentary
2025-02-04 22:02:15 -08:00
Allan Bazinet
4a180f7461
Experiment 2
2025-02-04 16:40:03 -08:00
Allan Bazinet
d5db781a44
Experiment; baseline with 500-2500 if filter unset
2025-02-04 00:02:57 -08:00
Allan Bazinet
3a0767eeba
Commentary
2025-02-03 09:27:03 -08:00
Allan Bazinet
44a84f2888
Commentary
2025-02-03 09:05:06 -08:00
Allan Bazinet
b2d1c190bd
No-arg constructor need not be explicit
2025-02-03 07:31:40 -08:00
Allan Bazinet
ad78bd1fed
Simplify mode-specific strategy creation
2025-02-03 06:46:08 -08:00
Allan Bazinet
1b1aa24aff
Simplify namespace usage
2025-02-03 02:36:51 -08:00
Allan Bazinet
7545a03343
Simplify worker memory management
2025-02-03 01:50:09 -08:00
Allan Bazinet
425c6637f3
Mitigate data race
2025-02-02 19:44:49 -08:00
Allan Bazinet
ad62061297
Correct missing include
2025-02-02 19:22:00 -08:00
Allan Bazinet
e018ac7c9d
Move FFT initialization off of main thread
2025-02-02 18:59:53 -08:00
Allan Bazinet
3788222206
Formatting
2025-02-02 08:42:28 -08:00
Allan Bazinet
58a962ec25
Formatting
2025-02-01 07:48:43 -08:00
Allan Bazinet
388f5dd9de
Formatting
2025-02-01 07:37:18 -08:00
Allan Bazinet
bf027ee3c0
Document s1 array normalization removal
2025-01-29 08:23:12 -08:00
Allan Bazinet
3bb102b1c5
Strip out s1 normalization code
...
This normalization did not seem to produce any discerable difference in decoder performance, but had the deleterious effect of convering all of s1 to NaN if the media value was zero. Seems to be useless; deleting.
2025-01-29 08:17:26 -08:00
Allan Bazinet
dc18334549
Experiment with disabling s1 normalization
2025-01-28 22:07:14 -08:00
Allan Bazinet
b8766317b1
KahanSum class hardening
2025-01-27 09:40:11 -08:00
Allan Bazinet
f0227182b4
Update commentary
2025-01-27 09:25:34 -08:00
Allan Bazinet
b501d49420
Use naive summation where Fortran does
2025-01-27 09:16:23 -08:00
Allan Bazinet
66b9679e05
Add KahanSum template class
2025-01-27 08:19:41 -08:00
Allan Bazinet
5a61ab8e75
Compensate for accumulation of rounding error
2025-01-27 07:33:40 -08:00
Allan Bazinet
2cbd2c80de
Revise filter creation
2025-01-27 03:52:51 -08:00
Allan Bazinet
7cf53c1249
Ensure alignment for AVX-512 usage in FFTs
2025-01-26 04:26:56 -08:00
Allan Bazinet
068f04d27a
Move to FFTWPlanManager
2025-01-26 04:13:06 -08:00
Allan Bazinet
64811d8d88
Set SNR clamp to -60db for testing
2025-01-25 17:43:48 -08:00
Allan Bazinet
f32a955308
Use mode common macros
2025-01-25 17:41:35 -08:00
Allan Bazinet
dd81e63a7e
Use commons
2025-01-25 17:33:56 -08:00
Allan Bazinet
d6ca6e67f6
Revert apparently crashing xsnr computation
...
Per Joe, this change is the cause of his bad submode exceptions. Frankly, I don’t see how that could be, but trying it out nevertheless.
2025-01-25 13:59:16 -08:00
Allan Bazinet
b22fcb9e23
Cleanup
2025-01-25 06:20:27 -08:00
Allan Bazinet
db2ca62607
Cleanup and commentary
2025-01-25 06:05:18 -08:00
Allan Bazinet
f42f1ff9b2
Remove legacy nsubmode parameter; always -1
...
Use of this parameter went away when the multi-decoder arrived; no longer needed at all.
2025-01-25 05:14:19 -08:00
Allan Bazinet
717a2fd3a0
Kill off write-only npts8 variables
2025-01-25 04:56:34 -08:00
Allan Bazinet
ad63650626
Kill of unused params.nagain
...
This value was invariably set to false; it’s a constant. Dramatically simplifies pass and SNR calculation in decoder by killing it off.
2025-01-25 04:52:54 -08:00
Allan Bazinet
9c171b7300
Remove unused params.ndiskdat and nranera
2025-01-25 04:13:07 -08:00
Allan Bazinet
5423ee61a9
Remove unused params.mode
2025-01-25 04:09:41 -08:00
Allan Bazinet
fc9e45e50b
Remove unused params.mycall and support function
2025-01-25 04:08:05 -08:00