BuildingBlissOS

Steps

Install necessray packages:

sudo apt install -y openjdk-8-jdk build-essential
sudo apt install git-core gnupg flex bison maven gperf build-essential zip curl zlib1g-dev gcc-multilib g++-multilib libc6-dev-i386 lib32ncurses5-dev x11proto-core-dev libx11-dev lib32z-dev ccache libgl1-mesa-dev libxml2-utils xsltproc unzip squashfs-tools  libssl-dev ninja-build lunzip syslinux syslinux-utils gettext genisoimage gettext bc xorriso libncurses5 xmlstarlet build-essential git imagemagick lib32readline-dev lib32z1-dev liblz4-tool libncurses5-dev libsdl1.2-dev libxml2 lzop pngcrush rsync schedtool  python3-mako libelf-dev$a
sudo apt-get install gcc-10 g++-10
git config --global user.email "you@example.com"
git config --global user.name "Your Name"

Get the repo from ustc.edu.cn, and replace the repo for proxy.

$ curl -sSL  'https://gerrit-googlesource.proxy.ustclug.org/git-repo/+/master/repo?format=TEXT' |base64 -d > ~/bin/repo
$ chmod a+x ~/bin/repo
## 如果提示无法连接到 gerrit.googlesource.com，可以编辑 ~/bin/repo，把 REPO_URL 一行替换成下面的：
## REPO_URL = 'https://gerrit-googlesource.proxy.ustclug.org/git-repo'

Repo init:

repo init -u https://github.com/BlissRoms-x86/manifest.git -b r11-r36

Repo sync(Using redsocks):

repo sync -j8

After repo sync:

cd prebuilts/gcc/linux-x86/host/x86_64-linux-glibc2.11-4.6
git checkout 2078a6bf9e5479104cfe2cbf54e9602672bd89f7

Build via:

. build/envsetup.sh
lunch android_x86_64-userdebug
export NO_KERNEL_CROSS_COMPILE=true
export BLISS_BUILD_VARIANT=foss
mka iso_img -j8

The generated iso is listed as:

out/target/product/x86_64/BlissOS-14.3-x86_64-202207270925_k-_m-r11-r36.iso is built successfully.
#### build completed successfully (02:40:00 (hh:mm:ss)) ####

dash@2204:/media/sda/Code/BlissOS$ 
dash@2204:/media/sda/Code/BlissOS$ ls out/target/product/x86_64/BlissOS-14.3-x86_64-202207270925_k-_m-r11-r36.iso
out/target/product/x86_64/BlissOS-14.3-x86_64-202207270925_k-_m-r11-r36.iso

VfioQEMU

Installation:

sudo qemu-system-x86_64 -name Arch -cdrom /var/lib/libvirt/images/ubuntu-20.04.3-live-server-amd64.iso -drive file=/var/lib/libvirt/images/vfioubuntu.qcow2 -m 8192 -enable-kvm -M q35 -cpu host -smp 4,sockets=1,cores=4,threads=1 -display egl-headless,gl=on -usb -device usb-tablet -bios /usr/share/OVMF/OVMF_CODE.fd -boot order=d -device virtio-vga,virgl=on -spice unix,disable-ticketing,image-compression=off,addr=arch.sock,seamless-migration=on

After installation, start the vm via:

sudo qemu-system-x86_64 -name Arch  -drive file=/var/lib/libvirt/images/vfioubuntu.qcow2,if=virtio  -m 22192 -enable-kvm -M q35 -cpu host -smp 8,sockets=1,cores=8,threads=1  -usb -device usb-tablet -device usb-ehci,id=ehci  -bios /usr/share/OVMF/OVMF_CODE.fd -boot order=c -vga none -device virtio-gpu-pci,virgl=on -display sdl,gl=on  -net nic -net user,hostfwd=tcp::2288-:22,hostfwd=tcp::5555-:5555

VirtIOGPUBenchmarkComparisonCN

1. 硬件信息

内存速度通过sudo dmidecode --type 17来检测
平台 1(i7 9代机器, 4核8线程，32G内存):

CPU: Intel(R) Core(TM) i7-9700K CPU @ 3.60GHz
Memory: MemTotal: 32709264 kB, 2666 MT/s

平台 2(i5 4代机器，4核4线程，32G内存):

CPU: Intel(R) Core(TM) i5-4460  CPU @ 3.20GHz
Memory: MemTotal: 32808212 kB, 1600 MT/s 

测试中使用同一块sata ssd:

Disk 1: Crucial 500GB sata ssd. 

测试显卡信息, 一块为WX5100 8G, 一块为5700 xt 8G(检测命令为 lspci | grep -i vga):

Card1: VGA compatible controller: Advanced Micro Devices, Inc. [AMD/ATI] Navi 10 [Radeon RX 5600 OEM/5600 XT / 5700/5700 XT] (rev c4)
Card2: VGA compatible controller: Advanced Micro Devices, Inc. [AMD/ATI] Ellesmere [Radeon Pro WX 5100]

2. 系统/软件信息

列举如下:

$ cat /etc/issue
Ubuntu 20.04.4 LTS \n \l
$ uname -a
Linux hope 5.4.0-120-generic #136-Ubuntu SMP Fri Jun 10 13:40:48 UTC 2022 x86_64 x86_64 x86_64 GNU/Linux
$ qemu-system-x86_64 --version
QEMU emulator version 6.0.0 (Debian 1:6.0+dfsg-2expubuntu1~focal1.0)
Copyright (c) 2003-2021 Fabrice Bellard and the QEMU Project developers
$ dpkg -l | grep -i virgl
ii  libvirglrenderer-dev:amd64                 0.8.2-1ubuntu1.1                      amd64        virtual GPU for KVM virtualization - headers
ii  libvirglrenderer1:amd64                    0.8.2-1ubuntu1.1                      amd64        virtual GPU for KVM virtualization

软件信息:

Unigine_Heaven-4.0
Unigine_Valley-1.0
glmark2
Redroid(Android 12)
scrcpy

3. 桌面场景测试

3.1 裸机对比(WX5100)

结论:

1. i5/i7平台上，WX5100表现基本一致。
2. 因WX5100算力比较一案，i5-4代没有成为测试中的瓶颈位。

3.2 裸机对比(5700xt)

结论:

1. 5700xt在i7平台上运行远比i5平台上流畅，4代i5平台无法发挥5700xt的算力特性，具体表现为：glmark2跑分差距较大，Unigine valley帧率相差较大(1080场景下掉了约70帧)，4K下表现一致。
2. 显卡选型时需充分考虑CPU/内存等平台算力特性。

3.3 Virgl On i5-4460(WX5100)

结论:

1. glmark2: Virgl跑分约为裸机的31.3%。
2. Unigine场景跑分: Virgl性能并未出现明显折损，因为WX5100本身算力一般, CPU足够支撑virgl转码开销。

3.4 Virgl On i7-9700K(WX5100)

结论:

1. glmark2: Virgl 约为裸机跑分的44.2%。
2. Unigine场景跑分: Virgl性能并未出现明显折损，因为WX5100本身算力一般, CPU足够支撑virgl转码开销。
3. 提升平台算力(CPU,内存等)有助于提升opengl转码效率(对比3.3场景下glmark2跑分).

3.5 Virgl On i5-4460(5700xt)

结论:

1. glmark2: Virgl约为裸机的21%。
2. Unigine Valley/Unigine Heaven: Virgl约为裸机跑分的55%~60%, 但在4K场景下，Virgl为裸机的87%。

3.6 Virgl On i7-9700K(5700xt)

结论:

1. glmark2: Virgl约为裸机的19%。
2. Unigine Valley/Unigine Heaven: Virgl约为裸机跑分的55%~60%, 但在4K场景下，Virgl为裸机的83%。

4. Android 场景

结论:

1. 鲁大师跑分极限: 约为60+万分，跑分时支持的最大分辨率为1080p(待确认)?
2. scrcpy开启与否会影响跑分结果: 尤其在低端CPU上跑分下降尤为明显。
3. 当GPU算力不至于成为瓶颈的场景下，virgl在安卓下渲染的跑分可以达到裸机极限跑分的: 90% (i7-9700K + 5700xt)
4. WX5100可能成为跑分的性能瓶颈的场景下， virgl安卓渲染跑分在i7下可达到裸机极限的: 85% (i7-9700K + WX5100), i5下可达到裸机极限的: 77% (i5-4460 + WX5100)

VirtIOGPUBenchmarkComparison

1. HardWare Info

Detect the memory speed via sudo dmidecode --type 17.
Hardware 1(i7 9th machine):

CPU: Intel(R) Core(TM) i7-9700K CPU @ 3.60GHz
Memory: MemTotal: 32709264 kB, 2666 MT/s

Hardware 2(i5 machine):

CPU: Intel(R) Core(TM) i5-4460  CPU @ 3.20GHz
Memory: MemTotal: 32808212 kB, 1600 MT/s 

Use the same ssd disk:

Disk 1: Crucial 500GB sata ssd. 
Disk 2: nvme ssd 256GB. 

Graphical Card Info(Detected via lspci | grep -i vga):

Card1: VGA compatible controller: Advanced Micro Devices, Inc. [AMD/ATI] Navi 10 [Radeon RX 5600 OEM/5600 XT / 5700/5700 XT] (rev c4)
Card2: VGA compatible controller: Advanced Micro Devices, Inc. [AMD/ATI] Ellesmere [Radeon Pro WX 5100]

2. System/SoftWare Info

System and Software info listed as:

500GB Crucial ssd:

$ cat /etc/issue
Ubuntu 20.04.4 LTS \n \l
$ uname -a
Linux hope 5.4.0-120-generic #136-Ubuntu SMP Fri Jun 10 13:40:48 UTC 2022 x86_64 x86_64 x86_64 GNU/Linux
$ qemu-system-x86_64 --version
QEMU emulator version 6.0.0 (Debian 1:6.0+dfsg-2expubuntu1~focal1.0)
Copyright (c) 2003-2021 Fabrice Bellard and the QEMU Project developers
$ dpkg -l | grep -i virgl
ii  libvirglrenderer-dev:amd64                 0.8.2-1ubuntu1.1                      amd64        virtual GPU for KVM virtualization - headers
ii  libvirglrenderer1:amd64                    0.8.2-1ubuntu1.1                      amd64        virtual GPU for KVM virtualization

nvme ssd 256GB:

$ cat /etc/issue
Pop!_OS 21.10 \n \l
$ uname -a
Linux pop-os 5.17.15-76051715-generic #202206141358~1655919116~21.10~1db9e34 SMP PREEMPT Wed Jun 22 19 x86_64 x86_64 x86_64 GNU/Linux
$ qemu-system-x86_64 --version
QEMU emulator version 6.0.0 (Debian 1:6.0+dfsg-2expubuntu1.3)
Copyright (c) 2003-2021 Fabrice Bellard and the QEMU Project developers
$ dpkg -l | grep -i virgl
ii  libvirglrenderer1:amd64                 0.8.2-5ubuntu0.21.10.1                                            amd64        virtual GPU for KVM virtualization

Software:

Unigine_Heaven-4.0
Unigine_Valley-1.0
glmark2
Redroid(Android 12)

3. Desktop Scenario

3.1 BareMetal Comparison(WX5100)

Conclusion:

1. WX5100 behaves almost the same on both i5/i7 platforms.
2. i5’s cpu power won’t be a bottleneck in WX5100 scenario.

3.2 BareMetal Comparison(5700xt)

Conclusion:

1. 5700xt runs smoother in i7 than on i5, perhaps i5 4th is too old for working together with 5700xt?(4 cores, bus width, etc).
2. i5’s CPU/Memory becomes a bottleneck in 5700xt testing.

3.3 Virgl On i5-4460(WX5100)

Conclusion:

1. glmark2: Virgl is 31.3% of the baremetal performance.
2. Unigine Valley/Unigine Heaven: Virgl performance almost the same as baremetal.

3.4 Virgl On i7-9700K(WX5100)

Conclusion:

1. glmark2: Virgl is 44.2% of the baremetal performance.
2. Unigine Valley/Unigine Heaven: Virgl performance almost the same as baremetal.
3. Comparing to 3.3: glmark2 will be improved using powerful CPU/Memory.

3.5 Virgl On i5-4460(5700xt)

Conclusion:

1. glmark2: Virgl is 21% of the baremetal performance.
2. Unigine Valley/Unigine Heaven: Virgl is around 55%~60% of the baremetal performance, while in 4k behavior, 87% of the baremetal performance.

3.6 Virgl On i7-9700K(5700xt)

Conclusion:

1. glmark2: Virgl is 19% of the baremetal performance.
2. Unigine Valley/Unigine Heaven: Virgl is around 55%~60% of the baremetal performance, while in 4k behavior, 83% of the baremetal performance.

4. Android Scenario

Conclusion:

1. Ludashi’s top score: around 61W, the maximum performance would be 1080p?
2. scrcpy will greatly impact the performance when using Virgl under low-end CPU.
3. virgl will be 85% for the best of bare metal on WX5100 on i7, 91% of bare metal on 5700xt on i7

UsingGimpForPhotoEditing

Create the new image:

Adjust the little image via:

160x200 px, resolution is 160 px/in.

Then select/copy the image into the big images, save as jpg, then we could print out the image.