03-14-2021, 10:27 AM
Depuis la version 123 Oryaaa a 3 pistes qui l'ont amener à 2 versions qui oeuvent se cumuler.
La 129 et la 130.
La 130 est l'évolution du développement classique avec, en fonction du processeur AVX2 et SSE en version Soap et Body.
La version SSE reste compatible tous processeurs.
Voici une copie du blog d'Oryaaa, reprise également sur le forum JPlay
La 129 et la 130.
La 130 est l'évolution du développement classique avec, en fonction du processeur AVX2 et SSE en version Soap et Body.
La version SSE reste compatible tous processeurs.
Voici une copie du blog d'Oryaaa, reprise également sur le forum JPlay
Citation :"MinorityClean 129 Structural Concepts, Readership High School Graduation
The basic design of Windows 10's time-shared task manager was created by inheriting the basic concept and source code from UNIX, which predates Windows NT, and the specification of distributed processing was used to promote multi-core processing. So the basic concept became a time division schedule, so if there is a specification that affects the physical registers in one EXE, it will affect many cores. However, even if a core is specified, the CPU will not operate with a monopoly of the core as instructed, but for security reasons, it will not be a monopoly specification. Since the processing of a core is always allocated to other processing, all cores will be affected if all cores are operated without specifying threads.
MinorityClean 129 is only implemented by the method of MOV R8, R8 MOV R8w, R8w MOV R8b, R8b, which updates logical registers called general purpose registers and all physical registers. This is an aggressive physical register update program that does it in all threads without waiting for the CPU to automatically update. This is not the same as MinorityClean 130, which is an electronic bit stabilizer process that adjusts to the internal circuitry of the memory chip.
Now, the MinorityClean 130 is structurally designed to update the physical registers with the average value of repeated accesses to the internal circuitry of the memory chip, so that it will not be affected even if it goes through acoustic processing using XMM registers. If you want to make the ultimate ASIO driver, the best way is to make it using only general-purpose and FPU instructions. In fact, the arithmetic precision of SSE2 is not very good even when double precision is used, while FPU has higher arithmetic precision than SSE2. The FPU is more accurate than the SSE2, because the low-pass interpolation is written with this in mind. If you are using SSE2 for format conversion, you can use FPU for format conversion. Huh? I digress. I don't know what I'm talking about... If you stop using SSE2 operation in the XMOS driver and use FPU operation, the sound will be better."
Other recent blog entries from Oryaaaaa:
"Bug head 12.67 release
with Rewrite data 3.93 and MinorityClean 129 130 BODY SOAP
When MC 129 130 is used together, the sound quality is dramatically improved, and when the combo of BODY SOAP is used, the quality is amazingly improved ... Bug head 12.67 allows you to enjoy the best sound quality for your environment, depending on your MC combination.
for Bug head = sexe64.exe "MC129" x 1 / manual operation "MC 130 SOAP AVX2" x 1
for Diretta = sexe64.exe "MC129" x 1 / manual operation "MC 130 BODY AVX2" x 1
for others = sexe64.exe "MC129" x 1 / manual operation "MC 130 BODY SOAP AVX2" x 2"
***
"BODY ... Body, solid impression
SOAP ... To the impression with the body, foaming, and transparency that applied the odd-even law"
! Mon installation !
ROON + HQP / Hdplex H3-i5 > DST-00 Diretta > HOLO Spring 3 > SQM > Benchmark AHB2 / Recital Audio Illumine HEFA // Upload IMG //
ROON + HQP / Hdplex H3-i5 > DST-00 Diretta > HOLO Spring 3 > SQM > Benchmark AHB2 / Recital Audio Illumine HEFA // Upload IMG //
![[Image: image.png]](https://i.postimg.cc/K86SwSvW/image.png)
