Remember when the Amiga bouncing ball demo was impressive? Ironically 3D graphics ended up being the Amiga's specific achilles heel once Doom and co came on the scene.
It means that, to set the color of a single pixel on the Amiga, you had to manipulate bits at multiple locations in memory (5 in 32 colours), while for the PC each pixel was just one memory location; In chunky mode you could just do something like: videomem[320*y+x]=158 to set the pixel at (x,y) to color 158, where videomem would point directly to the graphics memory (at address 0xa0000) -- It really was the simplest graphics mode to work with!
If you just copied 2D graphics (without scaling/rotating) the Amiga could do it quite will using the blitter/processor, but 3D texture mapping was more challenging because you constantly read and write to individual pixels (each pixel potentially requiring 5 memreads/writes on the Amiga vs. 1 on the PC).
Doom's wall texture mapping was affine, which basically means scaling+rotation operations were involved. The sprites were also scaled. Both operations a problem to the Amiga.
As software based 3D texture mapping games became the new hot thing in 1993-1997, the Amiga was left behind. Probably wouldn't have been a problem if the Amiga has survived until the 3D accelerators in the late 90s.
This is quite well described elsewhere. Google is your friend if you want to know more! :-)
Quake has software full 3D which runs appallingly if you can't do fast FP, it's targeting the new Pentium CPUs which all have fast FPUs onboard, it runs OK on a fast 486DX but it flies on a cheap Pentium.
Doom is just integer calculations, it's fixed point math.
In the general sense though the lack of floating point, as well as flat video addressing seriously hampered Amiga in the 3D ahem space.
EDIT I just remebered there is definitely at least one routine I know of that performs calculations based on IEEE 754 - “fast inverse square” or something. That could be at the root [badum] of my confusion vis-a-vis Doom ...
I am not confused, the opposite. I grew up with that.
With only two axis, it is literally a 2D space, which gives some illusion of 3D as an implementation detail --- not the other way around.
EDIT also, using the term "literally" to talk about 3Dness when it is all rendered onto a 2D screen, is fairly precarious. No matter how many degrees of freedom, or how rendered, it will never be "literally" 3D, in the literal sense of the term.