272 Commits

Author SHA1 Message Date
maxwell hayes
c425e49cb0 [metasound] optimization for mixer node (avoid memset / memzero)
#rb phil.popp
[FYI] Helen.Yang

[CL 32496894 by maxwell hayes in 5.4 branch]
2024-03-26 00:51:13 -04:00
helen yang
bbe3676b9c Biquad filter optimization for mono process audio
#rb phil.popp
#jira UE-208341

[CL 32490568 by helen yang in 5.4 branch]
2024-03-25 18:56:55 -04:00
Aaron McLeran
61c541f80f Adding constructor pin to grain delay node for max delay
#rb Maxwell.Hayes
#jira UE-168960
#fyi sondra.moyls

[CL 31581022 by Aaron McLeran in 5.4 branch]
2024-02-16 16:36:03 -05:00
steve robb
f029468598 Fixed up a lot of bool-taking container resize functions to take EAllowShrinking instead.
[CL 30729174 by steve robb in ue5-main branch]
2024-01-19 16:41:35 -05:00
phil popp
b3db59ff1d Fix remaining failures in metasound automated node tests
#rb Rob.Gay
#jira UE-198649

[CL 30688714 by phil popp in ue5-main branch]
2024-01-18 12:47:05 -05:00
maxwell hayes
dd5caaaf58 Add cvar to allow resetting metasound operators on cache insertion to potentially deallocate memory.
[REVIEW] [at]rob.gay
#rnx

#localization none
#tests local LLM A/B in PIE to confirm regression fixed w/o affecting weapon audio functionality
[FYI] aaron.mcleran

[CL 29596933 by maxwell hayes in ue5-main branch]
2023-11-09 10:54:35 -05:00
jimmy smith
8af9677dec Audio Buffer Diagnostics
#jira UE-197249
#rb jake.burga phil.popp

[CL 29578913 by jimmy smith in ue5-main branch]
2023-11-08 18:33:32 -05:00
maxwell hayes
cd1fdd0c33 [metasound builder] low-hanging fruit pre-allocation of containers and moving Stringbuilding / FName conversion to static init.
#rb Jordan.Cristiano, Aaron.McLeran
#rnx

[CL 29548374 by maxwell hayes in ue5-main branch]
2023-11-07 21:33:28 -05:00
dmytro vovk
ffd73158b9 Audio::ArrayMixIn loop unrolling for SSE optimization
#rb Phil.Popp

[CL 29321309 by dmytro vovk in ue5-main branch]
2023-11-01 12:25:28 -04:00
maxwell hayes
7229618552 [Metasounds] Optimize memory allocations in DSP objects for recycled metasounds.
#rb Aaron.McLeran
#jira UE-150771, UE-199176

#swarm https://p4-swarm.epicgames.net/reviews/29150979

[CL 29298167 by maxwell hayes in ue5-main branch]
2023-10-31 19:12:26 -04:00
phil popp
fc77726340 Optimize submix effect dynamics processor by utilizing deinterleaved codepath
#rnx
#rb Maxwell.Hayes

[CL 29297058 by phil popp in ue5-main branch]
2023-10-31 18:57:52 -04:00
phil popp
36cd56e6c6 [Backout] - CL29276889
[FYI] phil.popp
Original CL Desc
-----------------------------------------------------------------
Optimize submix effect dynamics processor by utilizing deinterleaved codepath
#rnx
#rb Maxwell.Hayes

[CL 29291267 by phil popp in ue5-main branch]
2023-10-31 17:08:03 -04:00
phil popp
320fa71c0a Optimize submix effect dynamics processor by utilizing deinterleaved codepath
#rnx
#rb Maxwell.Hayes

[CL 29278155 by phil popp in ue5-main branch]
2023-10-31 11:58:25 -04:00
miles flanagan
61d74f3ea5 [Metasound] fix filters clamping to bad numbers in lower samplerates
#rnx
#rb aaron.mcleran

[CL 29259226 by miles flanagan in ue5-main branch]
2023-10-30 20:12:10 -04:00
miles flanagan
3af6ab477c defend against overrunning internal wrap buffer in interpolated multi-tap delay
#rb phil.popp, maxwell.hayes
#rnx

[CL 29066834 by miles flanagan in ue5-main branch]
2023-10-25 00:51:33 -04:00
phil popp
b9a2b28833 Fix for crash in multichannel linear resampler.
#rnx
#jira UE-198585
#rb Jimmy.Smith

[CL 28981844 by phil popp in ue5-main branch]
2023-10-20 17:02:30 -04:00
phil popp
fb5af6ce1b Clamp linear resampler to ensure numerical stability.
#rnx
#rb Jimmy.Smith

[CL 28921838 by phil popp in ue5-main branch]
2023-10-19 12:15:16 -04:00
dmytro vovk
e1bacd79a0 Second attempt for vectorized audio conversion from float to pcm16 and vica versa with an optimized version for NEON and added a ~10% faster spicialization of ArrayMixIn for NEON
#rb Phil.Popp

[CL 28706860 by dmytro vovk in ue5-main branch]
2023-10-12 08:21:50 -04:00
miles flanagan
c5e4b0d0e9 strengthen buffer overflow defense in interpolated multi tap delay line to prevent rounding errors on buffer indices
#rb maxwell.hayes
#rnx

[CL 28675781 by miles flanagan in ue5-main branch]
2023-10-11 14:51:35 -04:00
dmytro vovk
382b7ab8f2 Reverted audio vectorization change
#rb none

[CL 28625268 by dmytro vovk in ue5-main branch]
2023-10-10 12:47:41 -04:00
dmytro vovk
7052f813ef Vectorized audio conversion from float to pcm16 and vica versa with an optimized version for NEON and added a ~10% faster spicialization of ArrayMixIn for NEON
#rb Phil.Popp

[CL 28623173 by dmytro vovk in ue5-main branch]
2023-10-10 12:10:28 -04:00
gary gong
bd74210f53 [SignalProcessing] Vectorized process method for DynamicsProcessor
#jira UE-196321

[CL 28455914 by gary gong in ue5-main branch]
2023-10-04 10:41:41 -04:00
gary gong
f52df1926a [SignalProcessing] Per-block process for circular buffer delay
#jira UE-196184

[CL 28197417 by gary gong in ue5-main branch]
2023-09-25 14:34:19 -04:00
charlie huguenard
dc2d480f4a [SignalProcessing] Non-interleaved processing for filters
#jira UE-196187
#rb phil.popp

[CL 28184867 by charlie huguenard in ue5-main branch]
2023-09-25 08:36:45 -04:00
gary gong
4aeee7c74a Add test scenario for ArrayMax in UintTestFloatArrayMath
#jira UE-196009

[CL 28134879 by gary gong in ue5-main branch]
2023-09-22 10:18:37 -04:00