RaspberryBASIC.org Forum

[AIR]

Last one I'm doing.

NIM

Code: [Select]

proc reverse*(str: string): string =
  result = ""
  for index in countdown(str.high, 0):
    result.add(str[index])

type
    milarray = array[1000001, int]

var
    s = ""
    t = ""
    r:string
    a:milarray

for x in 0..1000000:
    a[x] = x

    s &= chr((x mod 26) + 65)

    if s.len == 26:
        t &= s
        s = ""


r = t.reverse
echo "r Len: ",r.len
echo "Front: ",r[0..25]
echo "Back:  ",r[r.len-26..r.len-1]
echo "UBVal: ",a[1000000]
On my RPI 3B+:

riveraa@dpi:~ $ time ./mil
r Len: 999986
Front: ZYXWVUTSRQPONMLKJIHGFEDCBA
Back:  ZYXWVUTSRQPONMLKJIHGFEDCBA
UBVal: 1000000

real    0m0.123s
user    0m0.104s
sys    0m0.019s

AIR.

[John]

Wow!

Nim is fast.


pi@RPi4B:~/nim-dev/examples $ nim c -d:release onemil3.nim
Hint: used config file '/etc/nim/nim.cfg' [Conf]
Hint: system [Processing]
Hint: onemil3 [Processing]
Hint:  [Link]
Hint: operation successful (12410 lines compiled; 0.566 sec total; 10.754MiB peakmem; Release Build) [SuccessX]
pi@RPi4B:~/nim-dev/examples $ ls -l onemil3
-rwxr-xr-x 1 pi pi 57548 Dec  1 21:21 onemil3
pi@RPi4B:~/nim-dev/examples $ /usr/bin/time ./onemil3
r Len: 999986
Front: ZYXWVUTSRQPONMLKJIHGFEDCBA
Back:  ZYXWVUTSRQPONMLKJIHGFEDCBA
UBVal: 1000000
0.06user 0.05system 0:00.12elapsed 99%CPU (0avgtext+0avgdata 10048maxresident)k
0inputs+0outputs (0major+2294minor)pagefaults 0swaps
pi@RPi4B:~/nim-dev/examples $

[AIR]

Okay, one more, using JADE

Code: C++

1. #include <jade.hpp>
2.  
3. MAIN
4.     STRING s,t,r;
5.     ARRAY<int> a;
6.  
7.     FOR( VAR x = 0 TO x<1000001 STEP x++ ) DO
8.         a.push_back(x);
9.         s += (CHAR)(x%26)+65;
10.  
11.         IF ( LEN(s) == 26 ) THEN
12.             t += s;
13.             s = "";
14.         END
15.     END
16.  
17.     r = REVERSE$(t);
18.  
19.     PRINT("r LEN: ",LEN(t));
20.     PRINT("Front: ", LEFT$(r,26));
21.     PRINT("Back:  ", RIGHT$(r,26));
22.     PRINT("UBVal: ",a[1000000]);
23. END

On my RPI 3B+:

riveraa@dpi:~/Projects/jade $ g++ -std=c++11 mil.cpp -ljade -o mil
riveraa@dpi:~/Projects/jade $ time ./mil
r LEN: 999986
Front: ZYXWVUTSRQPONMLKJIHGFEDCBA
Back:  ZYXWVUTSRQPONMLKJIHGFEDCBA
UBVal: 1000000

real    0m0.194s
user    0m0.165s
sys    0m0.029s

AIR.

JadeLib Repository

EDIT:  Compiled with -O3, this is what I get on my RPI:

riveraa@dpi:~/Projects/jade $ g++ -O3 -std=c++11 mil.cpp -ljade -o mil
riveraa@dpi:~/Projects/jade $ time ./mil
r LEN: 999986
Front: ZYXWVUTSRQPONMLKJIHGFEDCBA
Back:  ZYXWVUTSRQPONMLKJIHGFEDCBA
UBVal: 1000000

real    0m0.070s
user    0m0.060s
sys    0m0.010s

[John]

Nice!

I'll setup a child board for JADE on the Raspberry BASIC forum.

When it comes to strings, C++ has a definite advantage.

[John]

AIR,

I get the following errors trying to make JADE.


pi@RPi4B:~/jade-dev/jadelib $ make
Creating regex.o
Creating ltrim.o
Creating rtrim.o
Creating lcase.o
Creating instr.o
Creating grab.o
Creating split.o
Creating right.o
Creating repeat.o
Creating replace.o
Creating pad.o
bin.cc: In function ‘STRING BIN$(long int)’:
bin.cc:4:30: error: ‘CHAR_BIT’ was not declared in this scope
     std::bitset< sizeof(num)*CHAR_BIT > bits( num );
                              ^~~~~~~~
bin.cc:4:30: note: ‘CHAR_BIT’ is defined in header ‘<climits>’; did you forget to ‘#include <climits>’?
bin.cc:2:1:
+#include <climits>
 
bin.cc:4:30:
     std::bitset< sizeof(num)*CHAR_BIT > bits( num );
                              ^~~~~~~~
bin.cc:4:39: error: template argument 1 is invalid
     std::bitset< sizeof(num)*CHAR_BIT > bits( num );
                                       ^
bin.cc:5:17: error: request for member ‘to_string’ in ‘bits’, which is of non-class type ‘int’
     RETURN bits.to_string();
                 ^~~~~~~~~
make: *** [Makefile:38: bin.o] Error 1
pi@RPi4B:~/jade-dev/jadelib $

[AIR]

Do:

make clean

git pull

make

I added the missing header file to the bin.cc file.

[John]

That worked. Posted on RB forum.

[AIR]

Updated C version:

Code: C

1. #include <stdio.h>
2. #include <stdlib.h>
3. #include <string.h>
4.  
5. char *strrev(char *str)
6. {
7.     if (!str || ! *str)
8.         return str;
9.  
10.     int i = strlen(str) - 1, j = 0;
11.  
12.     char ch;
13.     while (i > j) {
14.         ch = str[i];
15.         str[i] = str[j];
16.         str[j] = ch;
17.         i--;
18.         j++;
19.     }
20.     return str;
21. }
22.  
23.  
24. int main() {
25.     int a[1000001] = {0};
26.     char *t=NULL;
27.     char tmp[2] = {0};
28.     char s[27] = {0};
29.     FILE *t_stream;
30.     size_t len;
31.  
32.     t_stream = open_memstream(&t, &len);
33.  
34.     for (int x = 0; x < 1000001; x++) {
35.         a[x] = x;
36.         tmp[0] = (char)(x%26)+65;
37.  
38.         strncat(s,tmp,2);
39.  
40.         if (strlen(s) == 26) {
41.             fprintf(t_stream, "%s",s);
42.             s[0] = '\0';
43.         }
44.  
45.     }
46.  
47.     fclose(t_stream);
48.  
49.     strrev(t);
50.  
51.     printf("r LEN: %lu\n",strlen(t));
52.     printf("Front: %.*s\n", 26, t);
53.     printf("Back:  %s\n", t + strlen(t) - 26);
54.     printf("UBVal: %d\n",a[1000000]);
55.  
56.     free(t);
57.  
58. }
59.  

On my RPI 3B+

riveraa@dpi:~/Projects/C/1mil3 $ gcc -O3 mil.c -o mil
riveraa@dpi:~/Projects/C/1mil3 $ time ./mil
r LEN: 999986
Front: ZYXWVUTSRQPONMLKJIHGFEDCBA
Back:  ZYXWVUTSRQPONMLKJIHGFEDCBA
UBVal: 1000000

real    0m0.201s
user    0m0.190s
sys    0m0.011s

AIR.

[John]

Thanks AIR for the C update. C sometimes needs prosthetics to compete.

[AIR]

Thanks AIR for the C update. C sometimes needs prosthetics to compete.

...and swift needs a full body replacement....

[John]

I'm curious to hear what the Swift maintainer for ARM has to say about my findings.

I'm still willing to give Swift a chance. It's a late comer to open source but seems to have a lot of interest by the industry. I like the transparent wide character / Unicode support, It seems swift on other platforms.

[John]

AIR,

Can you put a Bash 1mil3 together while I work on the Java version? That will finalize the language set and I can create the result table. It would be great if Mike has time to contribute a Scheme version.

[AIR]

Bash version is not worth the effort, would take far too long to complete.

[John]

Okay.

The Java version is posted on the  RaspberryBASIC forum.

[John]

Swift maintainer response.

Yes, that let r = String(t.reversed()) in particular is extremely slow. That reversed() is the generic reversal defined on Collections here, and appears to be executed lazily (has an internal _base String that points to the original t, but the actual reversal is an O(n) swap).
But what adds 5 minutes or so of execution time is the String(Collection) initializer , as we can see running this modified sample using an Array:

It should be an O(n) operation with some additional conversions performed on all 32-bit platforms, but clearly there is something more going on.

Corner case as it may be, I'd open a ticket on https://bugs.swift.org/.

I can't determine if this is a 32 bit or ARM issue. In either case Swift on the RPi is a dead language for me until it's fixed.