Przemek:
If you want then you can also try to use -O3 instead of -O2 GCC
optimization flag. It enables some additional optimizations like
automatic normal function inlining. Probably it will give some
additional 2-5% of speed improvement in ST and MT mode.
I made a fresh build using:
CFLAGS = -W -Wall -O3
The improvement was marginal
slightly lower with ST and slightly higher with MT
Both results can be explained by variations in OS tasks too
ST
total application time: 45.81
total real time: 45.81
MT
total application time: 52.78
total real time: 52.78
Previous:
ST
total application time: 47.13
total real time: 47.14
MT
total application time: 51.98
total real time: 51.98
David Macias
Startup loop to increase CPU clock...
09/26/08 06:21:27 Harbour 1.1.0dev (Rev. 9508), OS/2 4.50
ARR_LEN = 16
N_LOOPS = 1000000
empty loops overhead = 0.13
CPU usage -> secondsCPU()
c:=L_C -> 0.12
n:=L_N -> 0.10
d:=L_D -> 0.10
c:=M_C -> 0.12
n:=M_N -> 0.10
d:=M_D -> 0.10
(sh) c:=F_C -> 0.43
(sh) n:=F_N -> 0.36
(sh) d:=F_D -> 0.20
(ex) c:=F_C -> 0.41
(ex) n:=F_N -> 0.37
(ex) d:=F_D -> 0.21
n:=o:GenCode -> 0.32
n:=o[8] -> 0.21
round(i/1000,2) -> 0.47
str(i/1000) -> 1.35
val(a3[i%ARR_LEN+1]) -> 0.84
dtos(j+i%10000-5000) -> 0.98
eval({||i%ARR_LEN}) -> 0.44
eval({|x|x%ARR_LEN},i) -> 0.49
eval({|x|f1(x)},i) -> 0.69
&('f1('+str(i)+')') -> 7.72
eval([&('{|x|f1(x)}')]) -> 0.71
j := valtype(a)+valtype(i) -> 0.88
j := str(i%100,2) $ a2[i%ARR_LEN+1] -> 1.73
j := val(a2[i%ARR_LEN+1]) -> 0.92
j := a2[i%ARR_LEN+1] == s -> 0.61
j := a2[i%ARR_LEN+1] = s -> 0.69
j := a2[i%ARR_LEN+1] >= s -> 0.67
j := a2[i%ARR_LEN+1] < s -> 0.68
aadd(aa,{i,j,s,a,a2,t,bc}) -> 2.21
f0() -> 0.28
f1(i) -> 0.39
f2(c[8]) -> 0.33
f2(c[40000]) -> 0.33
f2(@c[40000]) -> 0.31
f2(c[40000]); c2:=c -> 0.45
f2(@c[40000]); c2:=c -> 0.44
f3(a,a2,c,i,j,t,bc) -> 0.70
f2(a2) -> 0.34
s:=f4() -> 1.25
s:=f5() -> 0.63
ascan(a,i%ARR_LEN) -> 0.60
ascan(a2,c+chr(i%64+64)) -> 2.39
ascan(a,{|x|x==i%ARR_LEN}) -> 6.00
============================================================
total application time: 45.81
total real time: 45.81
Startup loop to increase CPU clock...
09/26/08 06:25:11 Harbour 1.1.0dev (Rev. 9508) (MT), OS/2 4.50
ARR_LEN = 16
N_LOOPS = 1000000
empty loops overhead = 0.15
CPU usage -> secondsCPU()
c:=L_C -> 0.13
n:=L_N -> 0.10
d:=L_D -> 0.11
c:=M_C -> 0.14
n:=M_N -> 0.12
d:=M_D -> 0.11
(sh) c:=F_C -> 0.43
(sh) n:=F_N -> 0.37
(sh) d:=F_D -> 0.20
(ex) c:=F_C -> 0.42
(ex) n:=F_N -> 0.36
(ex) d:=F_D -> 0.20
n:=o:GenCode -> 0.36
n:=o[8] -> 0.23
round(i/1000,2) -> 0.48
str(i/1000) -> 1.43
val(a3[i%ARR_LEN+1]) -> 0.92
dtos(j+i%10000-5000) -> 1.07
eval({||i%ARR_LEN}) -> 0.50
eval({|x|x%ARR_LEN},i) -> 0.54
eval({|x|f1(x)},i) -> 0.78
&('f1('+str(i)+')') -> 8.98
eval([&('{|x|f1(x)}')]) -> 0.81
j := valtype(a)+valtype(i) -> 0.97
j := str(i%100,2) $ a2[i%ARR_LEN+1] -> 1.85
j := val(a2[i%ARR_LEN+1]) -> 0.98
j := a2[i%ARR_LEN+1] == s -> 0.64
j := a2[i%ARR_LEN+1] = s -> 0.70
j := a2[i%ARR_LEN+1] >= s -> 0.70
j := a2[i%ARR_LEN+1] < s -> 0.72
aadd(aa,{i,j,s,a,a2,t,bc}) -> 3.52
f0() -> 0.29
f1(i) -> 0.44
f2(c[8]) -> 0.35
f2(c[40000]) -> 0.35
f2(@c[40000]) -> 0.34
f2(c[40000]); c2:=c -> 0.47
f2(@c[40000]); c2:=c -> 0.47
f3(a,a2,c,i,j,t,bc) -> 0.82
f2(a2) -> 0.35
s:=f4() -> 1.34
s:=f5() -> 0.79
ascan(a,i%ARR_LEN) -> 0.67
ascan(a2,c+chr(i%64+64)) -> 2.54
ascan(a,{|x|x==i%ARR_LEN}) -> 7.74
============================================================
total application time: 52.78
total real time: 52.78
_______________________________________________
Harbour mailing list
Harbour@harbour-project.org
http://lists.harbour-project.org/mailman/listinfo/harbour
