SELECT MKPF~BUDAT MKPF~CPUTM
     MSEG~MATNR MSEG~WERKS MSEG~EBELN 
      MSEG~EBELP  MSEG~ERFMG
           INTO CORRESPONDING FIELDS OF TABLE 
           W_DTL_INVOICE
           FROM MKPF INNER JOIN MSEG
                  ON  MKPF~MBLNR = MSEG~MBLNR
                  AND MKPF~MJAHR = MSEG~MJAHR
           WHERE MKPF~BUDAT > '20040721' AND
                 MSEG~BWART = '101'       AND
                 MSEG~BUKRS = '1733'.

It does not have much records.
Can somebody let me know how to fine tune this statement.

Kasi

I have gone through the same problem last year. This is the solution I made. My program spended more than an hour to 
select 5000 records from mkpf inner mseg(From a range of 100000 records) . 
Now it needs less than 15 seconds to perform the same task. 
Try this changes 

** in global data add these two hashed tables

data: begin of wa_mkpfmseg,
      mblnr like mkpf-mblnr,
      mjahr like mkpf-mjahr,
      zeile like mseg-zeile,
      bukrs like mseg-bukrs
        bwart like mseg-bwart,
        budat like mkpf-budat,
      cputm like mkpf-cputm,
        matnr like mseg-matnr,
     werks like mseg-werks,
        ebeln like mseg-ebeln,
        ebelp like mseg-ebelp,
        erfmg like mseg-erfmg,
        end of wa_mkpfmseg.

data ht_mkpfmseg like hashed table of wa_mkpfmseg
         with unique key mblnr mjahr zeile
       with header line.

data: begin of wa_mkpfmsegSel,
        budat like mkpf-budat,
      cputm like mkpf-cputm,
        matnr like mseg-matnr,
      werks like mseg-werks,
        ebeln like mseg-ebeln,
        ebelp like mseg-ebelp,
        erfmg like mseg-erfmg,
        end of wa_mkpfmseg.


data ht_mkpfmsegSel like hashed table of wa_mkpfmsegSel
         with unique key budat cputm matnr werks ebeln ebelp 
       with header line.

** change your select sentence to look like this:
  select mkpf~mblnr
              mkpf~mjahr
              mseg~zeile
            mseg~bukrs 
           mseg~bwart
        MKPF~BUDAT MKPF~CPUTM
             MSEG~MATNR MSEG~WERKS MSEG~EBELN 
             MSEG~EBELP  MSEG~ERFMG
                  INTO  TABLE ht_mkpfmseg
                  FROM MKPF INNER JOIN MSEG
                       ON   mkpf~mandt = mseg~mandt 
                  AND    MKPF~MBLNR = MSEG~MBLNR
                  AND MKPF~MJAHR = MSEG~MJAHR
                 where mkpf~budat > '20040721'.

  loop at ht_mkpfmseg.
                check ht_mkpfmseg-bukrs = '1733' and ht_mkpfmseg-bwart = '101'          
                read table ht_mkpfmsegsel with table key   
                     budat = ht_mkpfmseg-budat
                     cputm = ht_mkpfmseg-cputm                                                          

                     matnr  = ht_mkpfmseg-matnr          
                     werks  = ht_mkpfmseg-werks                                                   

                     ebeln = ht_mkpfmseg-ebeln   
                     ebelp  = ht_mkpfmseg-ebelp
                transporting erfmg.
                if sy-subrc <> 0.
                   move-corresponding ht_mkpfmseg to ht_mkpfmsegsel.
                   insert table ht_mkpfmsegsel.
                else.
                        ht_mkpfmsegSel-budat = ht_mkpfmseg-budat.
                        ht_mkpfmsegSel-cputm = ht_mkpfmseg-cputm.
                        ht_mkpfmsegSel-matnr = ht_mkpfmseg-matnr,
                        ht_mkpfmsegSel-werks = ht_mkpfmseg-werks.
                        ht_mkpfmsegSel-ebeln = ht_mkpfmseg-ebeln.
                        ht_mkpfmsegSel-ebelp = ht_mkpfmseg-ebelp.
                        add ht_mkpfmseg-erfmg to ht_mkpfmsegSel-erfmg.
                        modify table ht_mkpfmsegSel transporting erfmg.
                endif.
  endloop.
  " at this point ht_mkpfmsegSel has the data collected that you want. 
  loop at ht_mkpfmsegSel.
        .... Here put the code between your select ... endselect.
   endloop.

2. the data is selected into a hashed table which are the faster access tables provided by SAP from 4.0 up to 4.6c 
(I don't know about 4.7) 

3. As the select isn't restricting bukrs and bwart (but there is a good reason on this: it's faster to read 100000 records into 
a hashed table and then filter the 80000 unwanted than to select the 20000 records via non index fields in the where clause. 
I tested it in my own programs).

I hope this help you. If not, just let us know. 

Kind regards

Horacio.

Fast Links:
Using Select-Options and Ranges
Difference Between Select-Options & Ranges

Get help for your ABAP problems
Do you have a ABAP Question?

SAP Books
SAP Certification, Functional, Basis Administration and ABAP Programming Reference Books

ABAP Tips
ABAP Forum for Discussion and Samples Program Codes for Abapers

Best regards,
SAP Basis, ABAP Programming and Other IMG Stuff
http://www.erpgreat.com