Arvind, your comments are supposed to tell us what we don't know!

[pre]
acall DELAY ; calls the 30 MS delay subroutine
mov a, p1 ; mov the value of p1 to acc
push acc ; value of acc is preserved
anl a, #0fh ; higher nibble is made low ie ROW
cjne a, #0fh, AHEAD ;if the Lower Nibble ie Coulmn reading is not high then
; it means any column has been pressed
; jump to subroutine AHEAD
sjmp SCAN_MATRIX_KEYBOARD ; otherwise LOOP
[/pre]

Your comments are meant to explain the basic logic of what the code is doing not reword what the instruction is. I think Jan and me would've preferred you try to rewrite your code with the hints we gave you. Please don't expect us to rewrite your code - you won't learn anything. A Jan mentions - its all bad! You've got the basic concept, but the execution is flawed. Try writing code to decode just one row and use the LCD to output the results - once you have that going, then try all four. Break the problem down and you'll solve it in less time it took to post this message and wait for responses. A little persistance goes a long way. I see you didn't solve the problem for the 4bit LCD.