>Probably it's the cost of OpenWatcom MT safe CRTL. Harbour does not use
>extensively CRTL. In practice the only one function which may have
>serious performance impact is memory allocation. Maybe I will be able to
>say sth more if I see all speedtst results for Harbour compiled with and
>without -bm flag and check which tests are slower.
Rebuilt without -bm flag
-----------------------
[ total application time: ]....................................55.61
[ total real time: ]...........................................55.45
-----------------------
So -bm cause a very serious performance decrease
Note: Previous (-bm) and this test were using hbrun ST
Below are without and with -bm flag results, both ST
With previous hbrun MT was:
-----------------------
[ total application time: ]...................................198.42
[ total real time: ]..........................................197.09
-----------------------
David Macias
Without -bm flag:
11/06/08 05:35:17 OS/2 4.50
Harbour 1.1.0dev (Rev. 9846) Open Watcom C++ 12.70.8 (32-bit)
THREADS: 0
N_LOOPS: 1000000
[ T000: empty loop overhead ]...................................0.16
====================================================================
[ T001: x := L_C ]..............................................0.13
[ T002: x := L_N ]..............................................0.10
[ T003: x := L_D ]..............................................0.13
[ T004: x := S_C ]..............................................0.10
[ T005: x := S_N ]..............................................0.13
[ T006: x := S_D ]..............................................0.10
[ T007: x := M_C ]..............................................0.16
[ T008: x := M_N ]..............................................0.13
[ T009: x := M_D ]..............................................0.13
[ T010: x := P_C ]..............................................0.13
[ T011: x := P_N ]..............................................0.16
[ T012: x := P_D ]..............................................0.13
[ T013: x := F_C ]..............................................0.26
[ T014: x := F_N ]..............................................0.42
[ T015: x := F_D ]..............................................0.29
[ T016: x := o:GenCode ]........................................0.39
[ T017: x := o[8] ].............................................0.29
[ T018: round( i / 1000, 2 ) ]..................................0.48
[ T019: str( i / 1000 ) ].......................................1.03
[ T020: val( s ) ]..............................................0.48
[ T021: val( a [ i % 16 + 1 ] ) ]...............................0.97
[ T022: dtos( d - i % 10000 ) ].................................1.00
[ T023: eval( { || i % 16 } ) ].................................0.90
[ T024: eval( bc := { || i % 16 } ) ]...........................0.45
[ T025: eval( { |x| x % 16 }, i ) ].............................0.81
[ T026: eval( bc := { |x| x % 16 }, i ) ].......................0.52
[ T027: eval( { |x| f1( x ) }, i ) ]............................1.00
[ T028: eval( bc := { |x| f1( x ) }, i ) ]......................0.71
[ T029: x := &( "f1(" + str(i) + ")" ) ]........................6.16
[ T030: bc := &( "{|x|f1(x)}" ); eval( bc, i ) ]................6.10
[ T031: x := valtype( x ) + valtype( i ) ].....................0.87
[ T032: x := strzero( i % 100, 2 ) $ a[ i % 16 + 1 ] ]..........1.65
[ T033: x := a[ i % 16 + 1 ] == s ].............................0.68
[ T034: x := a[ i % 16 + 1 ] = s ]..............................0.77
[ T035: x := a[ i % 16 + 1 ] >= s ].............................0.74
[ T036: x := a[ i % 16 + 1 ] <= s ].............................0.77
[ T037: x := a[ i % 16 + 1 ] < s ]..............................0.71
[ T038: x := a[ i % 16 + 1 ] > s ]..............................0.77
[ T039: ascan( a, i % 16 ) ]....................................0.65
[ T040: ascan( a, { |x| x == i % 16 } ) ].......................6.52
[ T041: if i%1000==0;a:={};end; aadd(a,{i,1,.T.,s,s2,a2,bc ]....2.87
[ T042: x := a ]................................................0.16
[ T043: x := {} ]...............................................0.29
[ T044: f0() ]..................................................0.23
[ T045: f1( i ) ]...............................................0.35
[ T046: f2( c[1...8] ) ]........................................0.29
[ T047: f2( c[1...40000] ) ]....................................0.32
[ T048: f2( @c[1...40000] ) ]...................................0.32
[ T049: f2( @c[1...40000] ), c2 := c ]..........................0.45
[ T050: f3( a, a2, s, i, s2, bc, i, n, x ) ]....................0.87
[ T051: f2( a ) ]...............................................0.32
[ T052: x := f4() ].............................................1.03
[ T053: x := f5() ].............................................0.58
[ T054: f_prv( c ) ]............................................0.84
====================================================================
[ total application time: ]....................................55.61
[ total real time: ]...........................................55.45
Tests with current Harbour - OS/2 - With OpenWatcom 1.7 (Rev. 9846):
[E:\harbour810\harbour\tests]e:\harbour810\harbourow\bin\hbrun_st.exe
speedtst.prg
[ total application time: ]...................................104.00
[ total real time: ]..........................................103.51
Something changed in between with Harbour code and/or compilation :-(
In compilation I added -bm
Tests with first complete Harbour-OW build showed around 55 seconds,
near gcc335, now show allmost double
Przemek, any hint ?
David Macias
With -bm flag:
11/06/08 04:50:32 OS/2 4.50
Harbour 1.1.0dev (Rev. 9846) Open Watcom C++ 12.70.8 (32-bit)
THREADS: 0
N_LOOPS: 1000000
[ T000: empty loop overhead ]...................................0.13
====================================================================
[ T001: x := L_C ]..............................................0.16
[ T002: x := L_N ]..............................................0.16
[ T003: x := L_D ]..............................................0.13
[ T004: x := S_C ]..............................................0.16
[ T005: x := S_N ]..............................................0.13
[ T006: x := S_D ]..............................................0.13
[ T007: x := M_C ]..............................................0.19
[ T008: x := M_N ]..............................................0.16
[ T009: x := M_D ]..............................................0.16
[ T010: x := P_C ]..............................................0.16
[ T011: x := P_N ]..............................................0.16
[ T012: x := P_D ]..............................................0.16
[ T013: x := F_C ]..............................................1.45
[ T014: x := F_N ]..............................................0.42
[ T015: x := F_D ]..............................................0.29
[ T016: x := o:GenCode ]........................................0.42
[ T017: x := o[8] ].............................................0.32
[ T018: round( i / 1000, 2 ) ]..................................0.52
[ T019: str( i / 1000 ) ].......................................2.13
[ T020: val( s ) ]..............................................0.48
[ T021: val( a [ i % 16 + 1 ] ) ]...............................0.97
[ T022: dtos( d - i % 10000 ) ].................................2.19
[ T023: eval( { || i % 16 } ) ].................................4.00
[ T024: eval( bc := { || i % 16 } ) ]...........................0.52
[ T025: eval( { |x| x % 16 }, i ) ].............................2.90
[ T026: eval( bc := { |x| x % 16 }, i ) ].......................0.55
[ T027: eval( { |x| f1( x ) }, i ) ]............................3.13
[ T028: eval( bc := { |x| f1( x ) }, i ) ]......................0.74
[ T029: x := &( "f1(" + str(i) + ")" ) ].......................17.97
[ T030: bc := &( "{|x|f1(x)}" ); eval( bc, i ) ]...............20.87
[ T031: x := valtype( x ) + valtype( i ) ].....................2.03
[ T032: x := strzero( i % 100, 2 ) $ a[ i % 16 + 1 ] ]..........2.68
[ T033: x := a[ i % 16 + 1 ] == s ].............................0.71
[ T034: x := a[ i % 16 + 1 ] = s ]..............................0.74
[ T035: x := a[ i % 16 + 1 ] >= s ].............................0.74
[ T036: x := a[ i % 16 + 1 ] <= s ].............................0.74
[ T037: x := a[ i % 16 + 1 ] < s ]..............................0.74
[ T038: x := a[ i % 16 + 1 ] > s ]..............................0.74
[ T039: ascan( a, i % 16 ) ]....................................0.61
[ T040: ascan( a, { |x| x == i % 16 } ) ].......................9.26
[ T041: if i%1000==0;a:={};end; aadd(a,{i,1,.T.,s,s2,a2,bc ]....5.65
[ T042: x := a ]................................................0.19
[ T043: x := {} ]...............................................1.32
[ T044: f0() ]..................................................0.23
[ T045: f1( i ) ]...............................................0.39
[ T046: f2( c[1...8] ) ]........................................0.35
[ T047: f2( c[1...40000] ) ]....................................0.35
[ T048: f2( @c[1...40000] ) ]...................................0.32
[ T049: f2( @c[1...40000] ), c2 := c ]..........................0.48
[ T050: f3( a, a2, s, i, s2, bc, i, n, x ) ]....................0.94
[ T051: f2( a ) ]...............................................0.39
[ T052: x := f4() ].............................................2.06
[ T053: x := f5() ].............................................1.65
[ T054: f_prv( c ) ]............................................1.94
====================================================================
[ total application time: ]...................................104.00
[ total real time: ]..........................................103.51
_______________________________________________
Harbour mailing list
Harbour@harbour-project.org
http://lists.harbour-project.org/mailman/listinfo/harbour
