Tír Na HiFi

MOD19 probably is tightest sound + highest ease i can make. improves with realtime, 0.5ms, and close everything, close console.

i did try to push more, my OS cannot handle, tense.

zero agility issues. sharp, fast, accurate. not hard, full vibration here. vocal, piano no problem.

lekt player need improve SQ 200% to reproduce "woman in love" analog as vinyl, still remember how vinyl plays this song. especially bass/vocal/treble, all things like as foam. soft but power/energy like as sea wave.

MOD19 more organic than original, more ease.
think noind is most organic sound, i remember how it compiled, will use this OS environment again. only why a bit terrible, makes sound imbalance. too much power? easy come to hard.
2nd organic is v2.98.17 256 ucx3. maybe make MOD for it, good sound, balance better than noind.

all v2.99.9 256 xx have same code. see noRTCc assembly code have not calling function _RTC_Check_4_to_1 in loop, file size reduced 5kb. more vibration but SQ lesser than /RTCc.

for /RTCc version C compiler makes code:
002cb 48 8b 44 24 20 mov rax, QWORD PTR s$[rsp]
002d0 0f b6 00 movzx eax, BYTE PTR [rax]
002d3 83 f0 07 xor eax, 7
002d6 8b c8 mov ecx, eax
002d8 e8 00 00 00 00 call _RTC_Check_4_to_1
002dd 48 8b 4c 24 20 mov rcx, QWORD PTR s$[rsp]
002e2 88 01 mov BYTE PTR [rcx], al

for what checking after xor operator? here's never overflow with xor oprator. don't understand C compiler.

"You should use /RTC for development builds; /RTC should not be used for a retail build. /RTC cannot be used with compiler optimizations (/O Options (Optimize Code))....."

http://msdn.microsoft.com/en-us/library/8wtf2dfz.aspx

did you /O (optimize the code) with a release build? compiler optimization made SQ worse!!?

need to find out why. no /RTC sounds good

---
debug build usually makes code more easy to understand. so more easy to debug.
debug build makes more simple compile?

uploading (40% right now) "Rolling in the Deep - Adele Acoustic Cover (Jorge and Alexa Narvaez)"
i got it from youtube

father and 5 year old daughter singing.
all kids have the brightest voice.

only MOD19 can make bright voice sounds transparent with air.

99.9% of software makes her voice very hard

maybe i push the stack too low. could be tense for you. you can adjust it
don't change the heap. this should be sweetspot for all situations.

not use any optimization setting. v2.40 xx and older used, file size lesser, sound have shadow.

for example:

use /RTCc: v2.99.9 256 rtcc
unsigned char i;
i=255;
i++; // result 0, no error
i=255;
i=i+1; // error, overflow, player shutdown
i=252;
i+=4 // error, overflow, player shutdown
assembly code calls function _RTC_Check_4_to_1, in loop
----------------------------

don't use /RTCc: v2.99.9 256 nortcc
unsigned char i;
i=255;
i++; // result 0, no error
i=255;
i=i+1; // result 0, no error
i=252;
i+=4 // result 0, no error
----------------------------------------

why is first case better? think system no checking error of overflow. not sure
but nortcc more vibration, JC. reduced many things in loop. i like it. maybe CPU check overflow and doesn't raise error (if overflow than =0). seems this checking make sound a bit blur vs /RTCc.

you right, nortcc has more vibration. note more solid less hard earaches.
but the tune is changed to less sunny.

but hard vibration can make sunny tune.

maybe other area in code needs improve.

noiman MOD19 and noind MOD19 are by far the most correct piano play
fast notes for simple piece like Le Coucou so super clearly, accurate, with full weight, excellent depth/decay and good body. beats all software

unfortunately the bass is compromised -> the actual piano is compromised.

noiman MOD19 is flow focus.
noind MOD19 is weight focus. prefer this

noind MOD19 really good, you correct made all thing. original file imbalance => MOD result can not 100% good. think this MOD method is correct.maybe for all versions?

see few things stripped, removed 2 sections, used NX compatible = false

.pdata section can be removed?

pdata SEGMENT
$pdata$file DD imagerel $LN3
DD imagerel $LN3+986
DD imagerel $unwind$file
$pdata$main DD imagerel $LN12
DD imagerel $LN12+2102
DD imagerel $unwind$main
pdata ENDS

don't like it. it may consist bad matadata as i mentioned??

MOD19 compromises bass to get highest piano correctness. i suspect original bass has noise.
keep original bass = incorrect treble notes. unfortunately.

-commit size is most important for correct piano note body/weight.
-if treble notes is incorrect, then reduce commit size -> tighter sound -> more earaches.
-if treble notes is correct, then reduce commit size -> tighter sound -> more ease.

many things make treble incorrect e.g. OS, PE, software player....

according to my ears, lekt player code is fine. original PE is bottleneck for treble.
(for months i cannot fix the treble... now is much better with PE MOD!!!)

i think many software code is incorrect, coz minimum commit size -> tense tense earaches earaches....
or many different software is different.

in MOD15 you said sound is hard -> i increase the reserve size for MOD19 -> sound is easily vibrating -> clarity lesser little bit -> cannot be too greedy hehe

tring to make MOD20...
i guess original bass can be kept, if i can make large address = false (e.g. s4)
bass may not improve hehe... treble+ease should be better

lekt wrote:noind MOD19 really good, you correct made all thing. original file imbalance => MOD result can not 100% good. think this MOD method is correct.maybe for all versions?

see few things stripped, removed 2 sections, used NX compatible = false

.pdata section can be removed?

pdata SEGMENT
$pdata$file DD imagerel $LN3
DD imagerel $LN3+986
DD imagerel $unwind$file
$pdata$main DD imagerel $LN12
DD imagerel $LN12+2102
DD imagerel $unwind$main
pdata ENDS

don't like it. it may consist bad matadata as i mentioned??

i thought i deleted .pdata with CFF?!

delete section = treble better. bass i don't know.. hehe i am bass amateur

bass testing is very difficult... how do you do it?
i only know good bass makes piano sounds like stewart and son
most beautiful treble notes has true bass


-----
https://downloadcenter.intel.com/Detail ... 9&lang=eng

new hd4000 driver. again, here bass is less fat

-----
now i understand better what large address means!

if large address = disabled
then max heap/stack reserve is automatically smaller too!

don't need use so much memory for 16/44, do we?!
32bits is enough. i hope less memory management = better SQ