New core starting to come together!

News from administrators.
User avatar
MarathonMan
Site Admin
Posts: 692
Joined: Fri Oct 04, 2013 4:49 pm

New core starting to come together!

Post by MarathonMan » Tue Dec 23, 2014 8:03 pm

Still got a pretty bad RSP bug that's lingering around though. :(
Attachments
rampage.png
rampage.png (26.85 KiB) Viewed 23840 times

User avatar
Snowstorm64
Posts: 303
Joined: Sun Oct 20, 2013 8:22 pm

Re: New core starting to come together!

Post by Snowstorm64 » Tue Dec 23, 2014 9:29 pm

Good! :D As for that bug, are you going to fix it before releasing the new RSP in the git?
OS: Debian GNU/Linux Jessie (8.0)
CPU: Intel i7 4770K @ 3.5 GHz
Build: AVX (compiled from git)

User avatar
MarathonMan
Site Admin
Posts: 692
Joined: Fri Oct 04, 2013 4:49 pm

Re: New core starting to come together!

Post by MarathonMan » Wed Dec 24, 2014 12:14 am

The branch I'm working on (rdp-ldst) breaks master, so I haven't merged it yet. But I pushed it just now.

User avatar
OldGnashburg
Posts: 91
Joined: Tue Nov 19, 2013 3:00 pm
Location: Sherwood Park, Alberta, Canada: A place with free universal healthcare, and lots and lots of oil.

Re: New core starting to come together!

Post by OldGnashburg » Wed Dec 24, 2014 12:44 am

This is great news, look forward to actually test CEN64 for the first time on my laptop. Do you have any idea what this bug is (do you have any idea how and what you're going to hammer out)? What's the bug affecting?

PS. By the time anybody reads (AFAIK) it will be Christmas Eve, so Merry Christmas CEN64 Forums (or for you atheists like me, that actually care, Happy Holidays)!

OldGnashburg out!
Gnash, Gnash, Gnash...

User avatar
MarathonMan
Site Admin
Posts: 692
Joined: Fri Oct 04, 2013 4:49 pm

Re: New core starting to come together!

Post by MarathonMan » Wed Dec 24, 2014 1:10 am

Yes, I have a good way of tracking them down. Certainly making a lot of progress.

User avatar
juef
Posts: 31
Joined: Sun Oct 27, 2013 10:19 pm

Re: New core starting to come together!

Post by juef » Wed Dec 24, 2014 10:54 am

Nice! :D And while it's not the most intense scene... That 60.0 VI/s is quite lovely to see :)

ShadowFX
Posts: 86
Joined: Sat Oct 05, 2013 2:08 am
Location: The Netherlands

Re: New core starting to come together!

Post by ShadowFX » Wed Dec 24, 2014 12:58 pm

I'm also anxious to test out the new CEN64 (Windows) with its new cores. Will there be a pre-build binary or does the Windows version need compiling?

Happy holidays, people!
"Change is inevitable; progress is optional"

OS: Windows 10 Pro x64
Specs: Intel Core i7-7700K @ 4.2GHz, 16GB DDR4-RAM, NVIDIA GeForce GTX 1080 Ti
Main build: AVX (official)

User avatar
Breadwinka
Posts: 54
Joined: Fri Oct 04, 2013 11:35 pm

Re: New core starting to come together!

Post by Breadwinka » Wed Dec 24, 2014 1:05 pm


User avatar
MarathonMan
Site Admin
Posts: 692
Joined: Fri Oct 04, 2013 4:49 pm

Re: New core starting to come together!

Post by MarathonMan » Thu Dec 25, 2014 12:08 pm

3D models starting to load. Still missing implementations for some instructions, so that's likely the reason for the corruption. VI/s are very impressive from what I can tell so far -- 60VI/s during times when corruption is not present (when the RDP gets garbage data from the RSP, the VI/s drops very low).

EDIT: Middle image is after an additional instruction got implemented.

EDIT2: Top image is after a (yet another) additional instruction got partially implemented. VI/s still high!
Attachments
chrome3.png
chrome3.png (15.52 KiB) Viewed 23665 times
chrome2.png
chrome2.png (15.72 KiB) Viewed 23668 times
chrome.png
chrome.png (13.2 KiB) Viewed 23671 times

User avatar
Snowstorm64
Posts: 303
Joined: Sun Oct 20, 2013 8:22 pm

Re: New core starting to come together!

Post by Snowstorm64 » Thu Dec 25, 2014 1:06 pm

It's good to see the VI/s is still high! :D So far, we have these missing RSP opcodes:
  • NOP
  • CTC2
  • MTC2
  • VMACQ
  • VMACU
  • VMULQ
  • VNOP
  • VRNDN
  • VRNDP
  • LFV
  • LHV
  • LPV
  • LRV
  • LTV
  • LUV
  • SFV
  • SHV
  • SPV
  • SRV
  • STV
  • SUV
  • SWV
Plus VMACF implementation that seems to fail krom's tests.
The total of the missing opcodes is 22 ( and VMACF), right? It's not that far from completing RSP, then. Keep up the good job! :)

(P.S.: Merry Christmas!)
OS: Debian GNU/Linux Jessie (8.0)
CPU: Intel i7 4770K @ 3.5 GHz
Build: AVX (compiled from git)

User avatar
MarathonMan
Site Admin
Posts: 692
Joined: Fri Oct 04, 2013 4:49 pm

Re: New core starting to come together!

Post by MarathonMan » Thu Dec 25, 2014 1:11 pm

Snowstorm64 wrote:Plus VMACF implementation that seems to fail krom's tests.

I think that's failing because VMACU isn't implemented (which causes the accumulator to not get set correctly and VMACF to fail).

Just implemented MTC2. So, LPV/LUV and SUV are the main blockers left. :)

EDIT: Reflections starting to show in demo.
Attachments
chrome4.png
chrome4.png (17.93 KiB) Viewed 23151 times

User avatar
Alegend45
Posts: 11
Joined: Mon Oct 07, 2013 11:24 am

Re: New core starting to come together!

Post by Alegend45 » Thu Dec 25, 2014 3:28 pm

You seem to have an in-progress RDP there. Why isn't it in the git repo yet? I wanna play with it!

User avatar
MarathonMan
Site Admin
Posts: 692
Joined: Fri Oct 04, 2013 4:49 pm

Re: New core starting to come together!

Post by MarathonMan » Thu Dec 25, 2014 11:31 pm

Alegend45 wrote:You seem to have an in-progress RDP there. Why isn't it in the git repo yet? I wanna play with it!
I already released it. I just haven't merged the plugin into the new core publicly:
https://github.com/tj90241/cen64-rdp

It's really a fork of the repository below, with the absolute minimum number of changes to get it to work with CEN64 instead of zilmar spec (and with the VI filters stripped out):
https://code.google.com/p/angrylions-stuff/

User avatar
MarathonMan
Site Admin
Posts: 692
Joined: Fri Oct 04, 2013 4:49 pm

Re: New core starting to come together!

Post by MarathonMan » Thu Dec 25, 2014 11:51 pm

Boots Ridge Racer 64. Granted, the missing/distorted triangle issue makes it look like garbage. But I'm ticked with the (average, in game) 50VI/s on a 3GHz Haswell. :)

I'd be wiling to bet that a i7-4770 will run this at full speed. Just a matter of tuning it down to the point where lower power machines can run it, hopefully.
Attachments
rr64.png
rr64.png (81.63 KiB) Viewed 23074 times

User avatar
Snowstorm64
Posts: 303
Joined: Sun Oct 20, 2013 8:22 pm

Re: New core starting to come together!

Post by Snowstorm64 » Fri Dec 26, 2014 9:46 am

MarathonMan wrote: I'd be wiling to bet that a i7-4770 will run this at full speed. Just a matter of tuning it down to the point where lower power machines can run it, hopefully.
Should we find it out? :P
OS: Debian GNU/Linux Jessie (8.0)
CPU: Intel i7 4770K @ 3.5 GHz
Build: AVX (compiled from git)

User avatar
MarathonMan
Site Admin
Posts: 692
Joined: Fri Oct 04, 2013 4:49 pm

Re: New core starting to come together!

Post by MarathonMan » Fri Dec 26, 2014 2:20 pm

Snowstorm64 wrote:
MarathonMan wrote: I'd be wiling to bet that a i7-4770 will run this at full speed. Just a matter of tuning it down to the point where lower power machines can run it, hopefully.
Should we find it out? :P
Yes, but it's more interesting to test when there's no distortion... don't you agree? ;)

Lost an chunk of performance now that extra triangles are rendering, but that's to be expected I guess. Still seeing > 40VI/s at most times on my end!

I'll try to get things sorted out so that the RDP gets included in the builds shortly.
Attachments
rr64-fixed.png
rr64-fixed.png (77.67 KiB) Viewed 22997 times

User avatar
MarathonMan
Site Admin
Posts: 692
Joined: Fri Oct 04, 2013 4:49 pm

Re: New core starting to come together!

Post by MarathonMan » Fri Dec 26, 2014 3:49 pm

OK: http://downloads.cen64.com now contains builds of the angrylion-rdp branch.

SSE2 builds are broken; I know. I have to verify some things and re-implement them in SSE2 for one last time. Before I do that, I'd like to make sure that my SSSE3+ algorithms are correct (which I have to test on the actual hardware to find out).

Windows builds are still quasi-unstable, especially if you mess around with the Windows UI components (maximize, resize, etc.). I really need to fix up some loose bolts...

Also, the RSP is still very unstable... don't expect it to run too much quite yet.

User avatar
Breadwinka
Posts: 54
Joined: Fri Oct 04, 2013 11:35 pm

Re: New core starting to come together!

Post by Breadwinka » Fri Dec 26, 2014 4:25 pm

Ill test this on my 3570k @ 4.2Ghz in a few. When run in windows it actually plays stuff but wont take any commands. as well if you try clicking on the app you will get not responding.

When its started looks like this
Image

but if you click in focus you get
Image

As well there is no v/i's being recorded.

User avatar
MarathonMan
Site Admin
Posts: 692
Joined: Fri Oct 04, 2013 4:49 pm

Re: New core starting to come together!

Post by MarathonMan » Fri Dec 26, 2014 4:46 pm

I knew the Windows UI part of CEN64 was bad... but I didn't know it was that bad! :oops:

User avatar
Snowstorm64
Posts: 303
Joined: Sun Oct 20, 2013 8:22 pm

Re: New core starting to come together!

Post by Snowstorm64 » Fri Dec 26, 2014 5:09 pm

MarathonMan wrote:
Yes, but it's more interesting to test when there's no distortion... don't you agree? ;)
True, it makes more sense to wait the completion of new RSP for a better evaluation. But I was too much curious about this new RSP. :P

Intel i7-4770K 3.5 GHz, running Ridge Racer 64 (USA) on latest CEN64 build with native support.

With old RSP the game usually runs at 35-40 VI/s, but with the new RSP it runs much faster, at 45-50 VI/s, that is at the levels of the PAL games, which is good enough for me. :)
OS: Debian GNU/Linux Jessie (8.0)
CPU: Intel i7 4770K @ 3.5 GHz
Build: AVX (compiled from git)

User avatar
Nacho
Posts: 66
Joined: Thu Nov 07, 2013 9:25 am

Re: New core starting to come together!

Post by Nacho » Fri Dec 26, 2014 8:38 pm

That's impressive, MarathonMan!

I didn't knew that the old RSP was so unoptimized!
Testing CEN64 on: Intel Core i5 520M 2.4 GHz. SSE2 SSE3 SSE4.1 SSE4.2 SSSE3, but no AVX. Ubuntu Linux

User avatar
MarathonMan
Site Admin
Posts: 692
Joined: Fri Oct 04, 2013 4:49 pm

Re: New core starting to come together!

Post by MarathonMan » Sat Dec 27, 2014 12:10 am

Snowstorm64 wrote:
MarathonMan wrote:With old RSP the game usually runs at 35-40 VI/s, but with the new RSP it runs much faster, at 45-50 VI/s, that is at the levels of the PAL games, which is good enough for me. :)
The new RSP has a good deal of headroom left I think, too. Once I eliminate the excess movdqa instructions used to cache the RSP accumulator into the high SSE registers, it should result in another couple VI/s hopefully (based on some preliminary testing).
Nacho wrote:That's impressive, MarathonMan!

I didn't knew that the old RSP was so unoptimized!
Now just for the RDP, right? :D

User avatar
Alegend45
Posts: 11
Joined: Mon Oct 07, 2013 11:24 am

Re: New core starting to come together!

Post by Alegend45 » Sat Dec 27, 2014 12:14 am

Just one problem. On my Linux x86-64 machine, not even Namco Museum 64 displays anything. The window just displays whatever was under it the whole time. Huh. That's odd. My machine runs the build at downloads.cen64.com just fine. Namco Museum 64 is giving me 60 VI/s on a Core i3-3210.

User avatar
MarathonMan
Site Admin
Posts: 692
Joined: Fri Oct 04, 2013 4:49 pm

Re: New core starting to come together!

Post by MarathonMan » Sat Dec 27, 2014 12:21 am

Alegend45 wrote:Just one problem. On my Linux x86-64 machine, not even Namco Museum 64 displays anything. The window just displays whatever was under it the whole time. Huh. That's odd. My machine runs the build at downloads.cen64.com just fine. Namco Museum 64 is giving me 60 VI/s on a Core i3-3210.
Er... how is it not displaying anything but running at 60VI/s? Can you clarify?

You're likely either have a:
  • Corrupted/incorrect PIF image. The md5sum is 4921d5f2165dee6e2496f4388c4c81da
  • Byteswapped ROM image (ROMs must be in native, unbyteswapped, big-endian format). md5sum for Namco Museum 64 is e61251d2819e3bf3a9c0b95329f60f70

User avatar
iwasaperson
Posts: 49
Joined: Tue Apr 22, 2014 12:50 am

Re: New core starting to come together!

Post by iwasaperson » Sat Dec 27, 2014 12:22 am

Just built it on my 3570K. Getting around 30 VI/s. Also 52 in Dig Dug (Namco Museum), which is up 6 VI/s from the last time I built. Impressive.


EDIT:And adding on what MarathonMan said in the post above, cen64-qt converts n64 to z64, which is unbyteswapped.
Last edited by iwasaperson on Sat Dec 27, 2014 12:35 am, edited 2 times in total.

User avatar
Alegend45
Posts: 11
Joined: Mon Oct 07, 2013 11:24 am

Re: New core starting to come together!

Post by Alegend45 » Sat Dec 27, 2014 12:24 am

MarathonMan wrote:
Alegend45 wrote:Just one problem. On my Linux x86-64 machine, not even Namco Museum 64 displays anything. The window just displays whatever was under it the whole time. Huh. That's odd. My machine runs the build at downloads.cen64.com just fine. Namco Museum 64 is giving me 60 VI/s on a Core i3-3210.
Er... how is it not displaying anything but running at 60VI/s? Can you clarify?

You're likely either have a:
  • Corrupted/incorrect PIF image. The md5sum is 4921d5f2165dee6e2496f4388c4c81da
  • Byteswapped ROM image (ROMs must be in native, unbyteswapped, big-endian format). md5sum for Namco Museum 64 is e61251d2819e3bf3a9c0b95329f60f70
No, I mean Namco Museum 64 doesn't display anything on the build I made myself, but on the precompiled build, it displays just fine at 60 VI/s.

User avatar
MarathonMan
Site Admin
Posts: 692
Joined: Fri Oct 04, 2013 4:49 pm

Re: New core starting to come together!

Post by MarathonMan » Sat Dec 27, 2014 12:25 am

iwasaperson wrote:Just built it on my 3570K. Getting around 30 VI/s. Also 52 in Dig Dug (Namco Museum), which is up 6 VI/s from the last time I built. Impressive.
That's low. :( I have a weaker processor than that and get considerably more VI/s.

Make sure you're specifying the Release build type (and Native or AVX optimizations).

With a 3570K, you should maintain 60VI/s on Dig Dug and see at least 40VI/s in RR64, if not more.

User avatar
MarathonMan
Site Admin
Posts: 692
Joined: Fri Oct 04, 2013 4:49 pm

Re: New core starting to come together!

Post by MarathonMan » Sat Dec 27, 2014 12:27 am

Alegend45 wrote:No, I mean Namco Museum 64 doesn't display anything on the build I made myself, but on the precompiled build, it displays just fine at 60 VI/s.
That's weird. Assuming it's the vanilla repository, if you can specify as much as possible about your build environment (distro, gcc -v/clang -v dump, etc.) or send me a binary with debugging symbols, I can look into it.

User avatar
Alegend45
Posts: 11
Joined: Mon Oct 07, 2013 11:24 am

Re: New core starting to come together!

Post by Alegend45 » Sat Dec 27, 2014 12:33 am

MarathonMan wrote:
Alegend45 wrote:No, I mean Namco Museum 64 doesn't display anything on the build I made myself, but on the precompiled build, it displays just fine at 60 VI/s.
That's weird. Assuming it's the vanilla repository, if you can specify as much as possible about your build environment (distro, gcc -v/clang -v dump, etc.) or send me a binary with debugging symbols, I can look into it.
Here's the gcc -v dump. I'm using Linux Mint 17 with some packages from Debian sid, notably my gcc which works fine for everything else, except maybe Citra (and even that's debatable, because it happened on Linux and Windows with two entirely different compilers). It's a Native Debug build, though changing the Build Type didn't help. My binary is attached. Just rename the zip file to bin to something, because I was too lazy to figure out how to zip it up on the command line.

Code: Select all

Using built-in specs.
COLLECT_GCC=gcc
COLLECT_LTO_WRAPPER=/usr/lib/gcc/x86_64-linux-gnu/4.9/lto-wrapper
Target: x86_64-linux-gnu
Configured with: ../src/configure -v --with-pkgversion='Debian 4.9.2-8' --with-bugurl=file:///usr/share/doc/gcc-4.9/README.Bugs --enable-languages=c,c++,java,go,d,fortran,objc,obj-c++ --prefix=/usr --program-suffix=-4.9 --enable-shared --enable-linker-build-id --libexecdir=/usr/lib --without-included-gettext --enable-threads=posix --with-gxx-include-dir=/usr/include/c++/4.9 --libdir=/usr/lib --enable-nls --with-sysroot=/ --enable-clocale=gnu --enable-libstdcxx-debug --enable-libstdcxx-time=yes --enable-gnu-unique-object --disable-vtable-verify --enable-plugin --with-system-zlib --disable-browser-plugin --enable-java-awt=gtk --enable-gtk-cairo --with-java-home=/usr/lib/jvm/java-1.5.0-gcj-4.9-amd64/jre --enable-java-home --with-jvm-root-dir=/usr/lib/jvm/java-1.5.0-gcj-4.9-amd64 --with-jvm-jar-dir=/usr/lib/jvm-exports/java-1.5.0-gcj-4.9-amd64 --with-arch-directory=amd64 --with-ecj-jar=/usr/share/java/eclipse-ecj.jar --enable-objc-gc --enable-multiarch --with-arch-32=i586 --with-abi=m64 --with-multilib-list=m32,m64,mx32 --enable-multilib --with-tune=generic --enable-checking=release --build=x86_64-linux-gnu --host=x86_64-linux-gnu --target=x86_64-linux-gnu
Thread model: posix
gcc version 4.9.2 (Debian 4.9.2-8) 
Attachments
cen64.zip
(1.54 MiB) Downloaded 319 times

User avatar
iwasaperson
Posts: 49
Joined: Tue Apr 22, 2014 12:50 am

Re: New core starting to come together!

Post by iwasaperson » Sat Dec 27, 2014 12:35 am

Kind of runs OoT, but nothing past the Nintendo logo, and glitches around. It's a start: https://i.imgur.com/dsWHH3F.png

And I am building Native and Release

User avatar
Alegend45
Posts: 11
Joined: Mon Oct 07, 2013 11:24 am

Re: New core starting to come together!

Post by Alegend45 » Sat Dec 27, 2014 12:37 am

Alegend45 wrote:
MarathonMan wrote:
Alegend45 wrote:No, I mean Namco Museum 64 doesn't display anything on the build I made myself, but on the precompiled build, it displays just fine at 60 VI/s.
That's weird. Assuming it's the vanilla repository, if you can specify as much as possible about your build environment (distro, gcc -v/clang -v dump, etc.) or send me a binary with debugging symbols, I can look into it.
Here's the gcc -v dump. I'm using Linux Mint 17 with some packages from Debian sid, notably my gcc which works fine for everything else, except maybe Citra (and even that's debatable, because it happened on Linux and Windows with two entirely different compilers). It's a Native Debug build, though changing the Build Type didn't help. My binary is attached. Just rename the zip file to bin to something, because I was too lazy to figure out how to zip it up on the command line.

Code: Select all

Using built-in specs.
COLLECT_GCC=gcc
COLLECT_LTO_WRAPPER=/usr/lib/gcc/x86_64-linux-gnu/4.9/lto-wrapper
Target: x86_64-linux-gnu
Configured with: ../src/configure -v --with-pkgversion='Debian 4.9.2-8' --with-bugurl=file:///usr/share/doc/gcc-4.9/README.Bugs --enable-languages=c,c++,java,go,d,fortran,objc,obj-c++ --prefix=/usr --program-suffix=-4.9 --enable-shared --enable-linker-build-id --libexecdir=/usr/lib --without-included-gettext --enable-threads=posix --with-gxx-include-dir=/usr/include/c++/4.9 --libdir=/usr/lib --enable-nls --with-sysroot=/ --enable-clocale=gnu --enable-libstdcxx-debug --enable-libstdcxx-time=yes --enable-gnu-unique-object --disable-vtable-verify --enable-plugin --with-system-zlib --disable-browser-plugin --enable-java-awt=gtk --enable-gtk-cairo --with-java-home=/usr/lib/jvm/java-1.5.0-gcj-4.9-amd64/jre --enable-java-home --with-jvm-root-dir=/usr/lib/jvm/java-1.5.0-gcj-4.9-amd64 --with-jvm-jar-dir=/usr/lib/jvm-exports/java-1.5.0-gcj-4.9-amd64 --with-arch-directory=amd64 --with-ecj-jar=/usr/share/java/eclipse-ecj.jar --enable-objc-gc --enable-multiarch --with-arch-32=i586 --with-abi=m64 --with-multilib-list=m32,m64,mx32 --enable-multilib --with-tune=generic --enable-checking=release --build=x86_64-linux-gnu --host=x86_64-linux-gnu --target=x86_64-linux-gnu
Thread model: posix
gcc version 4.9.2 (Debian 4.9.2-8) 
Wait, never mind, deleting everything and starting again worked.

User avatar
MarathonMan
Site Admin
Posts: 692
Joined: Fri Oct 04, 2013 4:49 pm

Re: New core starting to come together!

Post by MarathonMan » Sat Dec 27, 2014 12:50 am

iwasaperson wrote:Kind of runs OoT, but nothing past the Nintendo logo, and glitches around. It's a start: https://i.imgur.com/dsWHH3F.png

And I am building Native and Release
Couple opcodes still need to be implemented, and a known RSP bug needs fixing before OoT or many other ROMs will boot.

Hm, that's disappointing... are you sure your processor is kicking into turbo boost? Or are you running a lot of background applications?

User avatar
iwasaperson
Posts: 49
Joined: Tue Apr 22, 2014 12:50 am

Re: New core starting to come together!

Post by iwasaperson » Sat Dec 27, 2014 12:55 am

MarathonMan wrote:
iwasaperson wrote:Kind of runs OoT, but nothing past the Nintendo logo, and glitches around. It's a start: https://i.imgur.com/dsWHH3F.png

And I am building Native and Release
Couple opcodes still need to be implemented, and a known RSP bug needs fixing before OoT or many other ROMs will boot.

Hm, that's disappointing... are you sure your processor is kicking into turbo boost? Or are you running a lot of background applications?
Checked with CPU-G. 4GHz while running (OC'd). You can see all of my background apps in the OoT pic. Just mpv and Chromium.
https://i.imgur.com/YuSwYyJ.png

EDIT: Just rebooted. Only background processes are xfce4 and Guake. Still the same VI/s.

User avatar
Snowstorm64
Posts: 303
Joined: Sun Oct 20, 2013 8:22 pm

Re: New core starting to come together!

Post by Snowstorm64 » Sat Dec 27, 2014 7:35 am

MarathonMan wrote: The new RSP has a good deal of headroom left I think, too. Once I eliminate the excess movdqa instructions used to cache the RSP accumulator into the high SSE registers, it should result in another couple VI/s hopefully (based on some preliminary testing).
That's awesome. :D
MarathonMan wrote:Now just for the RDP, right? :D
That looks like a lot of work, do you think it will benefit greatly with SSE intrinsic? Finally will we see the much desired 60 VI/s with these RDP optimizations?

iwasaperson wrote: Checked with CPU-G. 4GHz while running (OC'd). You can see all of my background apps in the OoT pic. Just mpv and Chromium.
https://i.imgur.com/YuSwYyJ.png

EDIT: Just rebooted. Only background processes are xfce4 and Guake. Still the same VI/s.
You may want to close mpv and chromium, they consume the CPU a bit. Also you might want to enable VR4300_BUSY_WAIT_DETECTION in the build options.
OS: Debian GNU/Linux Jessie (8.0)
CPU: Intel i7 4770K @ 3.5 GHz
Build: AVX (compiled from git)

User avatar
MarathonMan
Site Admin
Posts: 692
Joined: Fri Oct 04, 2013 4:49 pm

Re: New core starting to come together!

Post by MarathonMan » Sat Dec 27, 2014 9:14 am

iwasaperson wrote:Checked with CPU-G. 4GHz while running (OC'd). You can see all of my background apps in the OoT pic. Just mpv and Chromium.
https://i.imgur.com/YuSwYyJ.png

EDIT: Just rebooted. Only background processes are xfce4 and Guake. Still the same VI/s.
Sorry for the runaround - something was telling my mind that 3570K was Haswell. That's about right for Ivy Bridge.

For some reason, Haswell is like 20%+ faster clock-for-clock at CEN64?
Snowstorm64 wrote:That looks like a lot of work, do you think it will benefit greatly with SSE intrinsic? Finally will we see the much desired 60 VI/s with these RDP optimizations?
If I rewrite the RDP plugin in the same non-cycle-accurate fashion, they probably. I don't think it's possible to do a cycle-accurate RDP and gain VI/s. :(

User avatar
Snowstorm64
Posts: 303
Joined: Sun Oct 20, 2013 8:22 pm

Re: New core starting to come together!

Post by Snowstorm64 » Sat Dec 27, 2014 11:06 am

MarathonMan wrote:
Snowstorm64 wrote:That looks like a lot of work, do you think it will benefit greatly with SSE intrinsic? Finally will we see the much desired 60 VI/s with these RDP optimizations?
If I rewrite the RDP plugin in the same non-cycle-accurate fashion, they probably. I don't think it's possible to do a cycle-accurate RDP and gain VI/s. :(
That's a pity. Well, I'd like to be still able to use the software rasterizer instead of the non-cycle-accurate RDP, maybe with a build option?
OS: Debian GNU/Linux Jessie (8.0)
CPU: Intel i7 4770K @ 3.5 GHz
Build: AVX (compiled from git)

User avatar
MarathonMan
Site Admin
Posts: 692
Joined: Fri Oct 04, 2013 4:49 pm

Re: New core starting to come together!

Post by MarathonMan » Sat Dec 27, 2014 11:56 am

Yes, absolutely. You're not the only one who wants to someday use CEN64 to game! :D

User avatar
Snowstorm64
Posts: 303
Joined: Sun Oct 20, 2013 8:22 pm

Re: New core starting to come together!

Post by Snowstorm64 » Sat Dec 27, 2014 12:18 pm

:)

Benchmark time with another game: Iggy's Reckin' Balls (U), during a level
Old RSP: 25-30 VI/s, with great variations
New RSP: 33-38 VI/s with great variations
OS: Debian GNU/Linux Jessie (8.0)
CPU: Intel i7 4770K @ 3.5 GHz
Build: AVX (compiled from git)

User avatar
iwasaperson
Posts: 49
Joined: Tue Apr 22, 2014 12:50 am

Re: New core starting to come together!

Post by iwasaperson » Sat Dec 27, 2014 2:34 pm

Snowstorm64 wrote: Also you might want to enable VR4300_BUSY_WAIT_DETECTION in the build options.
That just boosted me up to 45 VI/s in game (60 at menus). Lolwut?

EDIT: Dig Dug is still the same though :(

beannaich
Posts: 149
Joined: Mon Oct 21, 2013 2:43 pm

Re: New core starting to come together!

Post by beannaich » Sat Dec 27, 2014 8:13 pm

iwasaperson wrote:
Snowstorm64 wrote: Also you might want to enable VR4300_BUSY_WAIT_DETECTION in the build options.
That just boosted me up to 45 VI/s in game (60 at menus). Lolwut?

EDIT: Dig Dug is still the same though :(
busy wait detection is used to skip idle periods for the cpu. when the RCP is busy, the cpu will most likely "spin wait" for it to not be busy. This results in pointless emulation, and can safely be skipped for a nice speed-up. All modern emulators employ some form of wait detection.

User avatar
iwasaperson
Posts: 49
Joined: Tue Apr 22, 2014 12:50 am

Re: New core starting to come together!

Post by iwasaperson » Sat Dec 27, 2014 8:43 pm

beannaich wrote:
iwasaperson wrote:
Snowstorm64 wrote: Also you might want to enable VR4300_BUSY_WAIT_DETECTION in the build options.
That just boosted me up to 45 VI/s in game (60 at menus). Lolwut?

EDIT: Dig Dug is still the same though :(
busy wait detection is used to skip idle periods for the cpu. when the RCP is busy, the cpu will most likely "spin wait" for it to not be busy. This results in pointless emulation, and can safely be skipped for a nice speed-up. All modern emulators employ some form of wait detection.
Good to know.

User avatar
MarathonMan
Site Admin
Posts: 692
Joined: Fri Oct 04, 2013 4:49 pm

Re: New core starting to come together!

Post by MarathonMan » Mon Dec 29, 2014 3:31 pm

Here's what it looks like running on my Sandy Bridge system:

https://www.youtube.com/watch?v=-oPN3_N_31s

User avatar
Nintendo Maniac 64
Posts: 185
Joined: Fri Oct 04, 2013 11:37 pm

Re: New core starting to come together!

Post by Nintendo Maniac 64 » Mon Dec 29, 2014 5:01 pm

Last month I got a fancy new Pentium G3258, but, uh, is a gamepad actually required to send any button presses? I figured there'd at least be super-basic keyboard support, like A button = A key, B button = B key, etc.
MarathonMan wrote:For some reason, Haswell is like 20%+ faster clock-for-clock at CEN64?
This is perfectly in line with other emulators - the very same performance increase on Haswell occurs with Dolphin, PCSX2, and MAME.


Speaking of performance, does anyone know how much of a performance difference there is between the AVX build and the SSE4.1 build? Pentiums don't have AVX, so I can't test it myself. :P
CEN64 Forum's resident straight-male kuutsundere
(just "tsundere" makes people think of "Shana clones" *shivers*)

CPU+iGPU: Pentium G3258 @ 4.6GHz/1.281v
dGPU: Radeon HD5870 1GB
RAM: Vengeance 1600 4x4GB
OS: Windows 7

User avatar
Snowstorm64
Posts: 303
Joined: Sun Oct 20, 2013 8:22 pm

Re: New core starting to come together!

Post by Snowstorm64 » Mon Dec 29, 2014 5:24 pm

Nintendo Maniac 64 wrote:Last month I got a fancy new Pentium G3258, but, uh, is a gamepad actually required to send any button presses? I figured there'd at least be super-basic keyboard support, like A button = A key, B button = B key, etc.
There's no gamepad support for now, but keyboard is available. See here
Nintendo Maniac 64 wrote:Speaking of performance, does anyone know how much of a performance difference there is between the AVX build and the SSE4.1 build? Pentiums don't have AVX, so I can't test it myself. :P
See this thread, less lines = code is faster.
OS: Debian GNU/Linux Jessie (8.0)
CPU: Intel i7 4770K @ 3.5 GHz
Build: AVX (compiled from git)

User avatar
MarathonMan
Site Admin
Posts: 692
Joined: Fri Oct 04, 2013 4:49 pm

Re: New core starting to come together!

Post by MarathonMan » Mon Dec 29, 2014 6:10 pm

Nintendo Maniac 64 wrote:I figured there'd at least be super-basic keyboard support, like A button = A key, B button = B key, etc.
The Windows build is pretty buggered at the moment. There is/should be keyboard support with statically mapped buttons.
Nintendo Maniac 64 wrote:Last month I got a fancy new Pentium G3258, but,

...

Pentiums don't have AVX, so I can't test it myself. :P
I just got myself a G3258 that I was going to O/C for CEN64. I wasn't aware of this until now. :(

User avatar
Nintendo Maniac 64
Posts: 185
Joined: Fri Oct 04, 2013 11:37 pm

Re: New core starting to come together!

Post by Nintendo Maniac 64 » Mon Dec 29, 2014 6:14 pm

MarathonMan wrote:The Windows build is pretty buggered at the moment. There is/should be keyboard support with statically mapped buttons.
Yeah, that doesn't seem to work - I'm guessing I have to click the windows first in order for the keyboard to work? However, as you already know, clicking the window results in it not responding, so...

EDIT: Hey, Cruis'n USA boots! But it hangs on the initial copyright screen.

EDIT 2: Cruis'n Exotica also boots, but similarly it hangs right after the initial copyright screen.

EDIT 3: The World is Not Enough boots, but like Ridge Racer it needs controller input to get past the memory card screen, and you know...

EDIT 4: Dr. Mario 64 boots and shows the title screen perfectly, but it hangs when it tried to switch over to the game demo.

BTW, I noticed that there isn't a VI/s counter on Windows...
Last edited by Nintendo Maniac 64 on Mon Dec 29, 2014 6:31 pm, edited 1 time in total.
CEN64 Forum's resident straight-male kuutsundere
(just "tsundere" makes people think of "Shana clones" *shivers*)

CPU+iGPU: Pentium G3258 @ 4.6GHz/1.281v
dGPU: Radeon HD5870 1GB
RAM: Vengeance 1600 4x4GB
OS: Windows 7

User avatar
MarathonMan
Site Admin
Posts: 692
Joined: Fri Oct 04, 2013 4:49 pm

Re: New core starting to come together!

Post by MarathonMan » Mon Dec 29, 2014 6:30 pm

Most games not booting is due to a stray bug in the RSP somewhere. I'm not too hung up about it ATM. Once I fix it, well over half of the commercial carts will boot and play just fine.

Missing VI/s counter is part of the Windows build being buggered.

User avatar
Nintendo Maniac 64
Posts: 185
Joined: Fri Oct 04, 2013 11:37 pm

Re: New core starting to come together!

Post by Nintendo Maniac 64 » Mon Dec 29, 2014 6:58 pm

Well I found several more games that seem to boot fine but I can't get any farther because they want controller input:
  • Hydro Thunder
  • Quake II
  • San Francisco Rush 2049
And games that seem to boot and run correctly all the way to the title screen but then eventually crash when trying to switch over to the game demo:
  • Star Wars: Episode 1 - Racer
Lastly, I know what you said, but for completion sake here are the games that boot but then hang or crash after an initial copyright screen, N64 logo, or similar:
  • Lego Racers
  • Sin and Punishment / Tsumi to Batsu
  • Star Wars: Shadow's of the Empire
  • Wave Race 64
CEN64 Forum's resident straight-male kuutsundere
(just "tsundere" makes people think of "Shana clones" *shivers*)

CPU+iGPU: Pentium G3258 @ 4.6GHz/1.281v
dGPU: Radeon HD5870 1GB
RAM: Vengeance 1600 4x4GB
OS: Windows 7

User avatar
Breadwinka
Posts: 54
Joined: Fri Oct 04, 2013 11:35 pm

Re: New core starting to come together!

Post by Breadwinka » Mon Dec 29, 2014 8:48 pm

MarathonMan wrote:Most games not booting is due to a stray bug in the RSP somewhere. I'm not too hung up about it ATM. Once I fix it, well over half of the commercial carts will boot and play just fine.

Missing VI/s counter is part of the Windows build being buggered.
I really need to get my linux distro back up and running.

Presence
Posts: 51
Joined: Fri Oct 18, 2013 9:27 am

Re: New core starting to come together!

Post by Presence » Tue Dec 30, 2014 7:09 pm

I'm getting an error when compiling:

Code: Select all

[ 88%] Building C object CMakeFiles/cen64arch.dir/arch/x86_64/rsp/rsp.c.o
/home/dan/Temp/cen64/arch/x86_64/rsp/rsp.c: In function ‘rsp_vect_load_and_shuffle_operand’:
/home/dan/Temp/cen64/arch/x86_64/rsp/rsp.c:203:22: error: ‘src’ undeclared (first use in this function)
     memcpy(&word_lo, src + (element - 8), sizeof(word_lo));
                      ^
/home/dan/Temp/cen64/arch/x86_64/rsp/rsp.c:203:22: note: each undeclared identifier is reported only once for each function it appears in
/home/dan/Temp/cen64/arch/x86_64/rsp/rsp.c:241:1: warning: control reaches end of non-void function [-Wreturn-type]
 }
 ^
CMakeFiles/cen64arch.dir/build.make:326: recipe for target 'CMakeFiles/cen64arch.dir/arch/x86_64/rsp/rsp.c.o' failed
make[2]: *** [CMakeFiles/cen64arch.dir/arch/x86_64/rsp/rsp.c.o] Error 1
CMakeFiles/Makefile2:140: recipe for target 'CMakeFiles/cen64arch.dir/all' failed
make[1]: *** [CMakeFiles/cen64arch.dir/all] Error 2
Makefile:76: recipe for target 'all' failed
make: *** [all] Error 2

Post Reply

Who is online

Users browsing this forum: No registered users and 0 guests