Let the instruction contain a field A (an address or displacement) and possibly a register name R. Let M[x] mean "contents of memory at x" and (R) mean "contents of register R".
WHY each rule looks like it does:
Immediate — the constant is baked into the instruction, so no address is needed. Fastest, but the value is fixed at compile time and limited by field width.
Direct — the address bits are the address. Simple, but A must be wide enough to name any location.
Indirect — A names a pointer cell; the CPU reads it to get the real address. Enables following linked structures, but costs an extra memory read.
Displacement (A+(R)) — the workhorse. R holds a base (start of array/struct/stack frame), A is a small offset. This is why array A[i] and struct fields are cheap.
PC-relative — makes code position-independent: a branch stores how far to jump, not an absolute target, so the program runs correctly wherever it's loaded.
Auto inc/dec — bakes pointer-stepping into the access, perfect for stacks (push/pop) and streaming through arrays.
EA = A + (R): base register plus a constant offset.
Why is PC-relative mode used for branches?
It stores a distance/offset (EA = PC + A), making code position-independent/relocatable.
Which mode is ideal for stack push/pop and array streaming?
Auto-increment / auto-decrement.
Which two modes cover ~80% of real instructions?
Register mode and displacement (base+offset) mode.
In LOAD 3(R1) with (R1)=400, what is EA?
403 (EA = 3 + 400).
Trade-off of adding many complex addressing modes?
Fewer instructions but more complex decode/longer critical path (CISC vs RISC trade-off).
Recall Feynman: explain to a 12-year-old
Imagine a treasure map. Sometimes the map says "the gold is RIGHT HERE" (immediate — the answer is written on the map). Sometimes it says "gold is buried at spot #100" (direct — go dig at 100). Sometimes spot #100 has another note saying "actually it's at spot #500" (indirect — one extra trip). Sometimes it says "start at the big tree (in your pocket = a register) and walk 3 steps" (displacement). And "PC-relative" is like "from where you're standing, go 6 steps forward" — so the treasure hunt works no matter where the park sets up the game. Every kind of map is just a different way of saying where to finally dig.
Dekho, addressing mode ka matlab hai: CPU ko kaise pata chalega ki operand (jis number pe kaam karna hai) actually kahan pada hai? Instruction sirf kuch bits ka hota hai, lekin memory bahut badi hoti hai. To har mode ek alag "recipe" hai jo un bits ko real jagah me convert karta hai. Sabse important cheez jo yaad rakhni hai — har mode basically Effective Address (EA) ka ek formula hai, aur jitni baar M[⋅] likhna pade utni memory access lagti hai.
Kuch modes simple: Immediate matlab value instruction ke andar hi likhi hai (zero memory access, sabse fast). Direct matlab address field hi actual address hai. Indirect thoda mehnga — pehle pointer padho, phir uss pointer wali jagah se data padho (do access lagte hain). Register indirect me pointer register me hota hai, to ek hi access. Aur Displacement (EA=A+(R)) sabse zyada use hone wala hai — R me array/struct/stack ka base hota hai aur A chhota offset. Jab bhi array[i] ya struct ka field access karte ho, andar yahi mode chal raha hota hai.
PC-relative ek smart trick hai: branch me actual address store nahi karte, balki "yahan se itne kadam aage jao" store karte hain (EA=PC+A). Isse code position-independent ho jaata hai — OS program ko kahin bhi load kare, sahi chalega. Aur auto-increment/decrement stack ke push/pop aur array streaming ke liye perfect hain, kyunki pointer khud aage-piche ho jaata hai.
Exam aur interview ke liye 80/20 funda: sirf register mode (fastest) aur displacement mode achhe se samajh lo — real programs me yahi sabse zyada aate hain. Aur ek galti se bachna: register-indirect aur memory-indirect ek jaise nahi hain — memory wale me ek extra access lagta hai. Bas EA ka formula likhna seekh lo aur arrows (memory reads) gin lo, sab clear ho jaayega.
