interpreted vs compiled language Step7

The VERY technical answer is that all the languages are interpreted.
Why? Well because the PLC firmware is actually running a virtual machine which runs "emitted code".
We designed our OEM control system in the early 1980's in a very similar manner.

When you "compile" code on the programming software, regardless of the chosen language, it all ends up as MC7 code. The is code that is interpreted by the virtual machine (VM) in the firmware that is running a RTOS scheduler. The VM has actual machine code instructions for the CPU that handle particular MC7 operations (e.g. variable assignment, data movement / copying, comparison operations function calls etc.). The MC7 code consists of the runtime address of the machine code routine to handle a particular instruction followed by any data / parameters needed for that instruction.
For example, say we have the line IF X > 45.7 THEN X := Y END_IF;

The "emitted" MC7 code would look something like this:

runtime address for 'IF' machine code routine
"greater than" token
the encoded value 45.7 (4 bytes in IEEE754 format)
a jump to the next instruction address after IF statement (for when condition is not satisfied)
runtime address of the assignment machine code routine
the address of  variable 'X'
the address of  variable 'Y'
runtime address of the next instruction

The scheduler calls the machine code routine to handle the instruction. Control then passes to the called machine code routine It will run the instruction, advancing a VM program counter through the list of MC7 emitted code and performing the necessary actions required to implement the instruction. When it gets to the end of the instruction it will pass control back to the scheduler. The scheduler will then move onto the next machine code routine and the process repeats.

I know all this as I actually reverse engineered some of the MC7 code a few years back when I was bored and I could see the emitted code structure. I couldn't fully decode it as I did not have the runtime addresses of the machine code functions or where the symbols (variables) are held. Although given enough time, I could have worked those out.

Sofiane_DZ_Atom

I need information on the difference between the languages (Graph, LAD, FBD, SCL, STL, SFC, CFC) which are interpreted and the languages are compiled

While smiffy gave you all the gory details on what happens once you have your MC7 code (and you may find Claroty's post HERE of interest too on that subject), here's some additional info on how to get to to the MC7 code since your question suggest interest in what Language is the most (runtime) efficient.

S7-300 & S7-400:

• STL: Gets compiled directly to MC7 -> most efficient way of programming 
• LAD/FBD: Is STL under the hood (you can freely change from LAD/FBD to STL) with some display related overhead (e.g. NOP instructions) and thus slightly less efficient
• SCL/GRAPH: Gets compiled to STL code first before compilation to MC7. Considerably less efficient than STL/ or LAD/FBD (if you delete the SCL or GRAPH source file you can see the STL code that is created from it)
• CFC: From memory (don't quote me on this) it compiles to SCL first, then to STL and from there to MC7 and is thus the least efficient language (concerning runtime execution anyhow)

S7-1200 & S7-1500:

All languages are as efficient as each other and directly compiled to MC7 code (according to Siemens anyhow, ref Programming Guideline for S7-1200/1500)