SharedGL

SharedGL provides OpenGL 4.6 acceleration for Windows and Linux guests in QEMU/KVM by streaming commands from a guest client to a host server.

Components

• Server (sglrenderer): Runs on the Linux host. It receives OpenGL commands and renders them in a window on the host.
• Client (ICD / libGL): Runs inside the guest VM. It intercepts OpenGL calls from guest applications and forwards them to the server.

Transports

To use the faster shared memory transport, you must attach an ivshmem device to your VM. Use sglrenderer -v on the host to generate snippets tailored to your current server settings.

libvirt Example

Place inside the <devices> section:

<shmem name="sharedgl_shared_memory">
    <model type="ivshmem-plain"/>
    <size unit="M">??</size>
</shmem>

QEMU Example

qemu-system-x86_64 \
  -object memory-backend-file,size=??M,share,mem-path=/dev/shm/sharedgl_shared_memory,id=sharedgl_shared_memory

<!-- libvirt sample -->
<shmem name="sharedgl_shared_memory">
    <model type="ivshmem-plain"/>
    <size unit="M">??</size>
</shmem>

The server can only be built and run on a Linux host.

Dependencies

• CMake 3.15+
• libepoxy (Latest)
• SDL2 2.24.0+

Build Steps

git clone https://github.com/dmaivel/sharedgl.git
cd sharedgl
mkdir build && cd build
cmake ..
cmake --build . --target sglrenderer --config Release

Build Options

Use -D flags with cmake or ccmake:

• LINUX_LIB32: Set to ON to build the Linux client library as 32-bit (does not affect the server).

When using the default shared memory transport, you must configure your Virtual Machine (libvirt or QEMU) to access the shared memory segment created by sglrenderer.

Use the -v flag with sglrenderer to print the exact configuration strings required for your current memory allocation settings.

Example command:

sglrenderer -m 64 -v

This will output the necessary XML for libvirt (<shmem> block) and the command-line arguments for QEMU (-object memory-backend-file...).

sglrenderer -v

The server must be running on the host before starting the guest. If using a prebuilt tarball, run ./sglrenderer from the root.

CLI Reference

sglrenderer [-h] [-v] [-o] [-n] [-x] [-g MAJOR.MINOR] [-r WIDTHxHEIGHT] [-m SIZE] [-p PORT]

SharedGL provides an automatic initialization mechanism using a constructor attribute. When the client library is loaded, sharedgl_entry is called automatically to initialize the OpenGL implementation (glimpl_init) and, on non-Windows platforms, the GLX implementation (glximpl_init). This ensures the ICD/libGL client is ready to forward OpenGL commands to the server without manual setup in your application code.

// No manual call required if the library is loaded as a plugin/ICD.
// The following is called automatically by the runtime:
void __attribute__((constructor)) sharedgl_entry(void);

SharedGL provides an automatic cleanup mechanism using a destructor attribute. When the client library is unloaded, sharedgl_goodbye is called automatically to perform necessary teardown via glimpl_goodbye().

// No manual call required if the library is unloaded as a plugin/ICD.
// The following is called automatically by the runtime:
void __attribute__((destructor)) sharedgl_goodbye(void);