Hello,
the title summarizes it.
For the visualization of algorithms, particle systems or molecules, often many objects of the same shape need to be drawn.
The extensions for instanced drawing apparently push OpenGLs ability to draw many objects of the same shape. One test measured a speedup of a factor 2 to 18 on two 2009-vintage video cards from Nvidia and AMD, depending on instance polygon count and the polygon caching capacity of the hardware. The test indicates that the speedup may even improve on newer hardware.
I expect that POGL will get even greater performance benefits, because rendering the same object many times without these extensions requires many API calls, and as POGL is an additional layer on top of the OpenGL API it has additional overhead for API calls.
It looks like an implementation needs to support only 2 - 3 additional functions: ARB_draw_instanced adds DrawArraysInstancedARB() and DrawElementsInstancedARB() and ARB_instanced_arrays adds VertexAttribDivisorARB(). (ARB_instanced_arrays is not strictly needed, but it means that the vertex shader code becomes much simpler as vertex attribute arrays can be used to transmit per-instance data, rather than textures (or uniforms, which also kill performance benefits).)
In my eyes the most important missing feature for visualization.
Note that even Intel HD Graphics since SandyBridge support these extensions. MESA had them since version 9.0, including support in the Intel driver.
Best regards
Actually MESA had them in 7.11 already. Intel support came later.
OpenGL::Modern has support for these: see https://metacpan.org/pod/OpenGL::Modern#glDrawArraysInstanced