Dr.Hell zaktualizował kod tandemu jednego z najlepszych emulatorów SONY PLAYSTATION - XEBRA/ARBEX i zrobił to już spory czas temu, ale jakoż, że pojawił się changelog i jest to jeden z wierniejszych emulatorów psx1 (bez zbędnych wodotrysków) warto go odnotować bardziej niż tylko wpisem w liście nowych plików.
Dla pocieszenia - pojawiła się również mobilna wersja XEBRA dla maszynek z Androidem dostępna w Google Market.
- Slightly reduce data volume. In GTE, division is performed by reciprocal integration, and its inverse value is inferior in accuracy. It was obvious that the common value for each group was refined. However, the calculation formula for refinement was unknown (There is only the Newton's law or the Gold Schmidt method as the theoretical formula When dealing with integers, how to round numbers is arbitrary by the designer, It can not be theoretically derived). Although it was unknown, it is not a large amount of data for software on PC I wish I could prepare the whole number as an immediate value table (This is effective also in terms of reducing the amount of computation at the time of execution)
When checking the calculation formula of the refinement indicated by nocash this time, since we saw an agreement in total, I decided to adopt the calculation formula.
As an actual utility, compressed data which was 4000 bytes in old algorithm, It is reduced to 257 bytes with the new algorithm. Although the normal distribution data used by the SPU was held for all the sections, Because it is bilaterally symmetrical, we deleted the half section.
- Center display of screen at startup
Final Fantasy IX (Square Electronic Arts L.L.C., Nov 14, 2000)
- The default value of Stall Rate (GPU - in) from 0x14 to 0x8 at the above review. I tried to change it, so it was quite probable that this was an enemy case It returned to 0x14
- Transfer rate (GPU-chain) is unknown from when it is about. It was fixed because it was invalidated. Since the tag part and the data part are too different in the rate, We made them separately, and we used coefficients of GPU-in for the data part.
Xebra (2017/03/25) Changelog:
- CD: Correction of physical position information during reading In a recording medium having a large influence of mechanical elements such as thermal expansion, distortion, eccentricity and vibration Since it is impossible to move the header to the correct position with a single shot. After moving to the position it is estimated that you will wait for the target position to be encountered while leading. Then, for specifying the position, the Q channel of the subcode, that is, the physical position information It is inferred that it will be obtained in advance. Therefore, the physical position information during reading is 2 to 3 sectors It is considered to precede.
Xebra (2017/03/17) Changelog:
- Timer 1 has a mode to count up by the system clock And a mode to count up by HSYNC, Both of them can stop counting up during VBLANK, Because I did not think there was a situation to stop in the system clock mode, For the sake of simplicity, we have adopted a pseudo-reproduction that counts up slowly without stopping. Recently I found out that there is a problematic game. Even in the system clock mode, it stopped properly.
Xebra (2017/01/17) Changelog:
- Correspondence to incorrect display area setting
Xebra (2016/10/25) Changelog:
MoveImage, it did not assume the priority bit processing in LoadImage Fixed.
While reading the copy source in MoveImage was always to be preceded, since it was found that a simple design which is reset for each block line fixed.
In line drawing, it did not assume an illegal drawing area (xmin> xmax)
A phenomenon in which the image is distorted to be due to the fact that the sprite is two dots at a time processing It was to reproduce.
Xebra (2016/06/04) Changelog:
That the partial derivative of the UV coordinates with precision 12-bit fixed decimal Was aware of ago, but was incorrect to take reference points for to suppress the error that did not reproduce the actual Street sare with high precision, which has been running. Relative errors (decided by the convenience of the drawing was adopt unconsciously y point values minimum) Recently finally realized as actual behavior and may be dropped precision. Also what times or serif play a soft failure to come out it was a behavior of CD just a little bit back to the previous ones.
Xebra (2016/05/12) Changelog:
- FIX: because there was a piece in binary conversion data associated with self-modifying disabling more than necessary. Made in setting the coefficient of axial controller [Rate] 0, [Base], [Rate] to calculate.
- The base address is shown in the Show Values.
Xebra/Arbex trafił również na urządzenia przenośne.
Modification of the drawing routines of GPU
It is one bit lower calculation accuracy at the time of lighting the texture in the conventional routine (Results would be devoid of larger when you make a dither to the texture as) such because I feel is
- Since the code on the OS ROM can be performed in a ②Mode 2 was supposed to be executed in Mode 1
- If you are recording a pad history, save and restore execution by the function key image In the (so-called state save and load), it was changed so that the pad history also saved and restored.
- File extension Along with this was also changed slightly.
Xebra OCT 11th 2015 is up!
- Fix the relationship of Setloc and Seek and Read.
- Sound of XEBRA was implemented in the most handy PlaySound () at the beginning. Since this API can not be queued, It took the means of loop playback buffer rewriting. However, since the reproduction position of the buffer is not known It had come out trouble overlap rewriting position and the playback position by the timing. (resulting in crackle..Ahem.) Then, How can was changed to API for waveOut system that can recognize the playback position, This time it queuing has disappeared completely from the head, It was still the loop playback buffer rewriting. WaveOut system in this review is that the queuing can be Since found was, adjusting the position of the buffer was no longer required. (meaning, it will be in perfect sync FOREVER!)
Scale processing correspondence of GDI:
- The Ya case does not contain only emulation driver as OpenGL It feels as vertical synchronization restrictions are increasingly took environment. Because the case is faster of GDI, we have to be able to scale processing even GDI
- Run Mode 2 I've packed only (little fast) some more.
Xebra/Arbex trafił również na urządzenia przenośne.
Xebra/Arbex trafiła na urządzenia przenośne.