Enhance performance of ABAP report when using OpenSQL query and function call - abap

Question

I have an issue with my ABAP report for it being very slow and taking a long time to complete.
My original report had a query to MARA to get all materials into an iTab, loop over it and then query additional tables for each material and store the necessary data and a function call to READ_TEXT in an output iTab.
Since this process takes very long to compete I changed the code to rely more on the database to handle data access instead of making queries for additionoal data each at a time. But still the time to complete is very long for the amount of data to be fetched.
Current code looks like this
report z_test_report.
types:
begin of _output,
matnr type mara-matnr,
text01 type c length 40,
text02 type c length 40,
ntgew type mara-ntgew,
matkl type mara-matkl,
ean11 type mara-ean11,
preis type p length 6 decimals 2,
bstd type mbew-lbkum,
laeng type marm-laeng,
breit type marm-breit,
hoehe type marm-hoehe,
volum type marm-volum,
end of _output,
t_output type _output.
data:
gt_output type table of t_output,
gs_output like line of gt_output,
gr_table type ref to cl_salv_table,
gr_funct type ref to cl_salv_functions,
gr_columns type ref to cl_salv_columns_table,
gr_column type ref to cl_salv_column.
selection-screen begin of block b01 with frame.
parameters: pa_date type d obligatory.
parameters: pa_kschl type t685-kschl obligatory.
parameters: pa_pltyp type t189-pltyp obligatory.
selection-screen end of block b01.
start-of-selection.
select
m~matnr,
m~ntgew,
m~matkl,
m~ean11,
t~maktx,
b~lbkum,
l~laeng,
l~breit,
l~hoehe,
l~volum
from
mara as m
left join
makt as t
on
m~matnr = t~matnr
and
t~spras = 'D'
left join
mbew as b
on
b~matnr = m~matnr
and
b~bwkey = '1100'
left join
marm as l
on
l~matnr = m~matnr
and
l~meinh = m~meins
into
#data(wa_daten)
order by
m~matnr.
gs_output = corresponding #( wa_daten ).
select single
*
from
a908
where
a908~kappl = 'V'
and
a908~kschl = #pa_kschl
and
a908~vkorg = '1100'
and
a908~pltyp = #pa_pltyp
and
a908~matnr = #wa_daten-matnr
and
a908~datab < #sy-datum
and
a908~datbi > #sy-datum
into
#data(gs_a908).
if ( sy-subrc = 0 ).
select single
*
from
konp
where
konp~knumh = #gs_a908-knumh
into
#data(gs_knop).
if ( sy-subrc = 0 ).
gs_output-preis = gs_knop-kbetr.
endif.
endif.
data:
material_name like stxh-tdname,
textlines like tline occurs 0,
textline like line of textlines.
material_name = gs_output-matnr.
call function 'READ_TEXT'
exporting
id = 'GRUN'
language = 'D'
name = material_name
object = 'MATERIAL'
tables
lines = textlines
exceptions
id = 1
language = 2
name = 3
not_found = 4
object = 5
reference_check = 6
wrong_access_to_archive = 7
others = 8.
if ( sy-subrc = 0 ).
loop at textlines into textline.
concatenate gs_output-text02 textline-tdline into gs_output-text02.
endloop.
endif.
condense gs_output-text02.
append gs_output to gt_output.
endselect.
try.
cl_salv_table=>factory(
exporting
list_display = if_salv_c_bool_sap=>false
importing
r_salv_table = gr_table
changing
t_table = gt_output
).
catch cx_salv_msg.
endtry.
gr_funct = gr_table->get_functions( ).
gr_funct->set_all( abap_true ).
gr_table->display( ).
I think this code can be improved since reading 17k entries from MARA and joining them with the other tables should not be the bottle neck. Changing the old code to the current one reduced the runtime from about 45s down to 30s, removing the READ_TEXT only made the report take about 5s less.
If anyone has an idea it would be really appreciated. Or if anyone knows a tool to measure report performance to find out about bottle necks.