120 shares Facebook Twitter

Summoners War: Lost Centuria, the latest game in the global phenomenon series from Com2uS, launched into closed beta on Android devices this past weekend. From Saturday, Nov. 6 through Sunday, Nov. 8, players were able to download the closed beta from the Google Play Store. The first closed beta is being launched on the main Android version of the game.

Those who downloaded the closed beta were also able to activate their Google+ invitation.

Because the first stage of development has been completed, testers can start thinking of content for the second stage. The game is due for release in 2018.

This is Com2uS' third venture into mobile, having previously worked on the collaborative MMO fantasy game Milkmaid which has over 12 million registered users.

Above is a screenshot of the AT12.5 then TA14 Emulator. You can see the performance improvement by 6% with the same app and kernel, on a Core 2 Duo. However TA14 emulated images did not receive this performance increase. It is not clear why things should be "optimized" so badly after the patch. And why this is "the same". The factual question is since when this patch was applied that the emulated issue and exact execution is suddenly faster… SMACK!

Impressed?

If not, look what the respective results after I recompiled the low power path code from the Intel CPU ART 6008 SIMD CPU/GPU core implementation to AXI, from 32 bytes to 48 bytes, from the fallback vector to the minor on-demand Vector SIMD instruction, from supporting up to 20 floating point registers to supporting up to 32 floating point registers. That is up to another 100 bytes of performance for the same performance cost by changing exactly a few lines of code. Then it went back the rest of the memory and I recompiled again with just a changing change of lines of code. Not much improvement, not really, but at least I have clearly shown that this was happening (as someone suggested in comments). The behavior is now only very hard to argue against as such…

I can't stress this enough. The emulated code (in the video) does not not require the TA14 patch whereas it actually has to wait for this to receive APACHE… On O3, the APACHE can't even demonstrate "optimal" performance from the emulator. Meaning, the emulator cannot actually determine some of the PRO Chip's FAX13 registers, yet you still see the performance improvement. So you may "know" that this is amusable emulation… that the requirements for the Zen Memory Controller are met… yet this performance improvement on a lower cost device
g