TweetFollow Us on Twitter

Compiler Compare
Volume Number:5
Issue Number:8
Column Tag:Jörg's Folder

Compiler Comparison

By Jörg Langowski, MacTutor Editorial Board

Note: Source code files accompanying article are located on MacTech CD-ROM or source code disks.

Code optimization

You will have noticed the change in the column’s title: a recent reader survey has shown that Forth and Basic are the two languages that our readers would most like to see less of in MacTutor. That’s a shame, but we’re responsive (at least we try)

So, from now on my monthly column will have a wider scope. As you might have seen, I have very often used Forth as a vehicle to explain general concepts of Macintosh programming. Since many subscribers don’t seem to be happy with Forth - in fact, people often have asked about things that had been explained in a Forth column, but they just hadn’t read. The column about the Notification Manager in V5#6 is a good example: at the bottom of page 42, one of the ‘ideas to be written’ was explained as: “ An example that places a Notification Manager request in low System memory, and starts a Timer routine ”; this is just the example that was in the Forth Forum that covered the Notification Manager. Anyway, I’ll try to use other vehicles to convey the message from now on. Such as assembly, or maybe even C. Mach2 Forth still is a very good assembly-language development system because of its interactivity. Of course, you cannot create complicated macros, or structures, and have to resort to Forth code for those purposes. I’ll still inform you about interesting things I come across on the Forth scene, but this won’t be an exclusive Forth column anymore. Emphasis will be on two things: basic system-level things such as drivers, trap patches, INITs, network, new managers as they come up; and - on the other side of the spectrum - object-oriented programming in C++.

Apple will ‘Real Soon Now’ release a C++ under MPW, and we’ll hopefully have a pre-release by the time you read this. C++ is a very interesting language, much more than a simple extension of C; reading Stroustrup’s book I got this feeling of ‘yes, that’s how one should have done it in the first place’ that I had 15 years ago when all I knew was Algol 60, and came across the description of Algol 68. Sadly enough, Algol 68 never really caught on; hopefully, C++ will. The C++ column will start with the next issue; including program examples if we get the pre-release soon enough, just ‘dry swimming’ if not.

This month, we’ll talk once more about one of my favorite subjects, number crunching, speed (or the lack of it), and intelligence in compilers (or the lack of it).

A matrix multiplication routine

Since I am doing more and more everyday computation (mostly Fortran) on the MacII, I’m obviously interested in a good optimizing compiler. Now, a standard trick that every decent compiler should have in its repertoire is the elimination of constant expressions from loops, or assignment of array elements that are not dependent on the loop index to intermediate variables.

Imagine my surprise when I found out that (in Language Systems Fortran 1.2.1) I could speed up a loop that looked like this:

 do k=1,n3
 do i=1,n1
 c(i,k) = 0x0
 do j=1,n2
 c(i,k) = c(i,k)+a(i,j)*b(j,k)
 end do
 end do
 end do

by simply writing:

 do k=1,n3
 do i=1,n1
 sum = 0x0
 do j=1,n2
 sum = sum+a(i,j)*b(j,k)
 end do
 c(i,k) = sum
 end do
 end do

Now, in undergraduate programming classes, years ago, we were actually taught to look for constant arithmetical or index expressions in loops and put them outside if possible. Today, almost everybody assumes that the compiler is smart enough to take care of that; incorrectly, as you see. To see how good the compilers available under MPW can do, I wrote a Fortran program (listing 1) that calls several versions of this matrix multiplication loop, written in Fortran (Lang. Sys. 1.2.1), Pascal (Apple 3.0 beta), and C (Apple 3.0 beta). Surprise: none of the compilers was good enough to move the indexing outside of the loop. The following table gives the results (Mac IIx):

Pascal, hand-optimized: 2.7667 seconds

C, register variables, hand-opt.: 3.4667 seconds

Pascal: 4.0333 seconds

Fortran, const. dimensions, opt=3: 4.5000 seconds

Fortran, hand-optimized, opt=3: 4.6500 seconds

C: 4.7167 seconds

Fortran, opt=3: 6.5167 seconds

Fortran, opt=0: 6.6167 seconds

A difference of more than a factor of 2 between the slowest Fortran and the fastest Pascal code. Apple Pascal lived up to its good reputation here, but even that could be improved a lot by eliminating the constant index expression.

Surprised, I ran the Fortran benchmark on a Microvax II, and found that even there some speed could be gained by ‘hand-optimizing’ the code:

Fortran, plain: 3.6333 seconds

Fortran, hand-opt.: 3.2833 seconds

Fortran, const. dimensions: 3.1000 seconds

However, the difference between the machine-optimized and the hand-optimized version is not quite as big as for the MPW languages (15% for the VAX vs. 27-30% for MPW). If you compile the VAX code without optimization, you get a bigger difference (23%):

Fortran, plain: 6.2500 seconds

Fortran, hand-opt.: 4.8833 seconds

Fortran, const. dimensions: 4.8000 seconds

Therefore, take-home lesson one: don’t take compiler optimizations for granted.

The machine code behind it

Benchmarks have been run on lots of different machines, using lots of different compilers. I was interested in how the code generated by the MPW compilers actually differed. A job for Nosy, and the results are shown in the last listing. I’ve only printed the innermost loops. Don’t be overwhelmed by the pile of assembly code, just note some important details.

First, for the loop optimization examples discussed here, there seems to be no tradeoff between code length and speed. On the contrary, the fastest code is also the shortest. On the other hand, there are some obvious pieces of code which are clearly redundant. The most blatant example is the Fortran-generated code at the end of the listing, where an index expression is recalculated that was actually in register A1 all the time! 14 extra lines of machine code on each pass through the loop will add up to quite some extra time lost. Another point is that Language System obviously has no great trust in the quality of the 68000/20/30, otherwise how can one explain that they repeat the EXT.L D2 instruction each time it occurs? To make sure it works at least once?

Language Systems Fortran makes other funny assumptions about the machine, for instance it seems to think there are only two floating point registers in the 68881, FP0 and FP7. I have looked at some code which had great potential for optimization by using enough FP registers. Language Systems is, however, known for its responsiveness towards customers, so I hope we won’t have to wait too long until a well-optimized Fortran shows up.

Both Pascal and C like juggling floating point registers. Why generate (like Apple’s C):

FMOVE   FP7,FP1
FADD    FP0,FP1
FMOVE   FP1,FP7

when a simple FADD FP0,FP7 would suffice? Eliminates two floating point instructions per loop. Pascal does

   FADD    FP7,FP0
   FMOVE   FP0,FP7

when a simple inversion of the operands

   FADD FP0,FP7

would give the same result. One floating point instruction per loop eliminated. The timing difference between the Pascal and C routines is partly because of the one extra floating point instruction.

Last remark: I haven’t seen the Absoft MPW 3.0 Fortran yet. If anyone from Absoft is reading this, I’d like an evaluation copy to run the same analysis (since you claim in your ads you have such a great optimizer). If I get enough other languages collected together, we’ll have a follow-up on this article.

Next month

The MacHack is over (thanks, Aimée, Carol, and all the others, for organizing such a good meeting), and I’ll tell you some of my impressions in the next column. Otherwise, we’ll start with an introduction to C++; I hope the compiler will arrive here in time.

Listing 1: Matrix multiplication benchmark

!!S Main
 program matrix
c
c Main program in Language Systems Fortran
c
c Some line breaks in the Fortran program are due to 
c editing.
c
 implicit none
 
 integer i,j,ticks1,ticks2
 extended a(50,50), b(50,50), c(50,50)
 extended time1,time2
 
 integer ticks
 
 type *,’Matrix multiplication benchmark’
 type *,’------------------------------’
 type *
 type *,’This program compares the number crunching power’
 type *,’of some of the popular MPW compilers.’
 type *,’Written under MPW 3.0 by J. Langowski / MacTutor 1989'
 type *
 type *,’Setting up 50x50 matrices...’

 ticks1 = ticks()
 
 do i=1,50
 do j=1,50
 a(i,j) = (i-1) + j-1
 b(j,i) = a(i,j)
 end do
 end do
 
 ticks2 = ticks()
 time1 = (ticks2-ticks1)/60.
 type *
 write (6,’(f8.4,’’ seconds for setting up matrices’’)’) time1
 
 ticks1 = ticks()
 call mat_mult_for3(c,50,a,50,b,50,50,50,50)
 
 ticks2 = ticks()
 time1 = (ticks2-ticks1)/60.
 type *
 write (6,’(f8.4,’’ seconds for multiplying matrices’’,
     *  ‘’ using FORTRAN routine, opt=3'’)’) time1
 type *,’c(25,25) = ‘,c(25,25)
 
 ticks1 = ticks()
 call mat_mult_for(c,50,a,50,b,50,50,50,50)
 
 ticks2 = ticks()
 time1 = (ticks2-ticks1)/60.
 type *
 write (6,’(f8.4,’’ seconds for multiplying matrices’’,
     *  ‘’ using FORTRAN routine, opt=0'’)’) time1
 type *,’c(25,25) = ‘,c(25,25)

 ticks1 = ticks()
 call mat_mult_for1(c,50,a,50,b,50,50,50,50)
 
 ticks2 = ticks()
 time1 = (ticks2-ticks1)/60.
 type *
 write (6,’(f8.4,’’ seconds for multiplying matrices’’,
     *  ‘’ using FORTRAN routine, hand-optimized’’)’) time1
 type *,’c(25,25) = ‘,c(25,25)
 
 ticks1 = ticks()
 call mat_mult_for0(c,50,a,50,b,50,50,50,50)
 
 ticks2 = ticks()
 time1 = (ticks2-ticks1)/60.
 type *
 write (6,’(f8.4,’’ seconds for multiplying matrices’’,
     *  ‘’ using FORTRAN routine, constant dimensions’’)’) time1
 type *,’c(25,25) = ‘,c(25,25)
 
 ticks1 = ticks()
 call mat_mul_pas(c,%val(50),a,%val(50),b,%val(50),%val(50),%val(50),%val(50))
 
 ticks2 = ticks()
 time1 = (ticks2-ticks1)/60.
 type *
 write (6,’(f8.4,’’ seconds for multiplying matrices’’,
     *  ‘’ using PASCAL routine’’)’) time1
 type *,’c(25,25) = ‘,c(25,25)
 
 ticks1 = ticks()
 call mat_mul_pas_opt(c,%val(50),a,%val(50),b,%val(50),%val(50),%val(50),%val(50))
 
 ticks2 = ticks()
 time1 = (ticks2-ticks1)/60.
 type *
 write (6,’(f8.4,’’ seconds for multiplying matrices’’,
     *  ‘’ using PASCAL routine, hand-optimized’’)’) time1
 type *,’c(25,25) = ‘,c(25,25)
 
 ticks1 = ticks()
 call mat_mul_c(c,%val(50),a,%val(50),b,%val(50),%val(50),%val(50),%val(50))
 
 ticks2 = ticks()
 time1 = (ticks2-ticks1)/60.
 type *
 write (6,’(f8.4,’’ seconds for multiplying matrices’’,
     *  ‘’ using C routine’’)’) time1
 type *,’c(25,25) = ‘,c(25,25)
 
 ticks1 = ticks()
 call mat_mul_c_opt(c,%val(50),a,%val(50),b,%val(50),%val(50),%val(50),%val(50))
 
 ticks2 = ticks()
 time1 = (ticks2-ticks1)/60.
 type *
 write (6,’(f8.4,’’ seconds for multiplying matrices’’,
     *  ‘’ using C routine, hand-optimized’’)’) time1
 type *,’c(25,25) = ‘,c(25,25)
 
 end

!!S Main
 subroutine mat_mult_for3(c,nc,a,na,b,nb,n1,n2,n3)
c sets c=a.b
c na,nb,nc are first dimensions
c n1 n2 n3 are problem dimensions
c c is n1xn3
c a    n1 n2
c b    n2 n3
 implicit none
 
 integer na,nb,nc,n1,n2,n3
 integer*2 i,j,k
 extended c(nc,n3),a(na,n2),b(nb,n3)
 
 do k=1,n3
 do i=1,n1
 c(i,k) = 0x0
 do j=1,n2
 c(i,k) = c(i,k)+a(i,j)*b(j,k)
 end do
 end do
 end do
 return
 end

!!S Main
 subroutine mat_mult_for1(c,nc,a,na,b,nb,n1,n2,n3)
c
csame as before, invariant matrix element eliminated from loop
c
 implicit none
 
 integer na,nb,nc,n1,n2,n3
 integer*2 i,j,k
 extended c(nc,n3),a(na,n2),b(nb,n3)
 extended sum 
 
 do k=1,n3
 do i=1,n1
 sum = 0x0
 do j=1,n2
 sum = sum+a(i,j)*b(j,k)
 end do
 c(i,k) = sum
 end do
 end do
 return
 end

!!S Main
 subroutine mat_mult_for0(c,nc,a,na,b,nb,n1,n2,n3)
c
csame as before, with constant dimensions
c
 implicit none
 
 integer na,nb,nc,n1,n2,n3
 integer*2 i,j,k
 extended c(50,50),a(50,50),b(50,50)
 extended sum 
 
 do k=1,n3
 do i=1,n1
 sum = 0x0
 do j=1,n2
 sum = sum+a(i,j)*b(j,k)
 end do
 c(i,k) = sum
 end do
 end do
 return
 end

!!S Main
 integer function ticks
 ticks = long(362)
 return
 end
Listing 2 : non-optimized Fortran routine           
!!S Main
 subroutine mat_mult_for(c,nc,a,na,b,nb,n1,n2,n3)
c
cduplicate of mat_mul_for3 for compiling without optimization
c
 implicit none
 
 integer na,nb,nc,n1,n2,n3
 integer*2 i,j,k
 extended c(nc,n3),a(na,n2),b(nb,n3)
 
 do k=1,n3
 do i=1,n1
 c(i,k) = 0x0
 do j=1,n2
 c(i,k) = c(i,k)+a(i,j)*b(j,k)
 end do
 end do
 end do
 return
 end
Listing 3 : Pascal routine

{$S Main}
{$R-}
unit matmul;

interface

type matrix = array [1..50,1..50] of extended;

procedure mat_mul_pas
 (var c : matrix; nc : longint;
  var a : matrix; na : longint;
  var b : matrix; nb : longint;
  n1,n2,n3:longint);

procedure mat_mul_pas_opt
 (var c : matrix; nc : longint;
  var a : matrix; na : longint;
  var b : matrix; nb : longint;
  n1,n2,n3:longint);

implementation

procedure mat_mul_pas;
var
 i,j,k:integer;
begin
 for k:=1 to n3 do
 for i:=1 to n1 do
 begin
 c[i,k] := 0;
 for j:=1 to n2 do
 c[i,k] := c[i,k]+a[i,j]*b[j,k];
 end;
end;

procedure mat_mul_pas_opt;
var
 i,j,k:integer; sum:extended;
begin
 for k:=1 to n3 do
 for i:=1 to n1 do
 begin
 sum := 0;
 for j:=1 to n2 do
 sum := sum+a[i,j]*b[j,k];
 c[i,k] := sum;
 end;
end;

end.
Listing 4 : C routine

pascal void mat_mul_c 
 (extended c[50][], long nc,
  extended a[50][], long na,
  extended b[50][], long nb,
  long n1, long n2, long n3)

{
 int i,j,k;
 
 for ( k=1 ; k <= n3; k++ )
 for ( i=1 ; i <= n1 ; i++ )
 {
 c[i][k] = 0.0;
 for ( j=1 ; j <= n2 ; j++ )
 c[i][k] = c[i][k]+a[i][j]*b[j][k];
 }
}

pascal void mat_mul_c_opt
 (extended c[50][], long nc,
  extended a[50][], long na,
  extended b[50][], long nb,
  long n1, long n2, long n3)

{
 register int i,j,k;
 register extended sum;
 
 for ( k=1 ; k <= n3; k++ )
 for ( i=1 ; i <= n1 ; i++ )
 {
 sum = 0.0;
 for ( j=1 ; j <= n2 ; j++ )
 sum = sum+a[i][j]*b[j][k];
 c[i][k] = sum;
 }
}
Listing 5 : inner loops compared, Nosy-disassembled

pascal, optimized
lan_3 MOVEA.L  param2(A6),A0
 MOVE D6,D0
 MULS #$258,D0
 MOVE D5,D1
 MULS #12,D1
 ADD  D1,D0
 MOVEA.Lparam3(A6),A1
 MOVE D5,D1
 MULS #$258,D1
 MOVE D7,D2
 MULS #12,D2
 ADD  D2,D1
 LEA  -$264(A0),A0
 FMOVE.X0(A0,D0.W),FP0
 LEA  -$264(A1),A0
 FMUL.X 0(A0,D1.W),FP0
 FADD FP7,FP0  ; could use
 FMOVE  FP0,FP7  ; FADD FP0,FP7 here
 ADDQ #1,D5
 BVS.S  lan_5
lan_4 CMP.W van_1(A6),D5
 BLE  lan_3

c, optimized
lar_1 MOVE.LD7,D0
 MOVE.L D0,D1
 MULU #12,D0
 SWAP D1
 MULU #12,D1
 SWAP D1
 CLR  D1
 ADD.L  D1,D0
 ADD.L  D6,D0
 MOVE.L D5,D1
 MOVE.L D1,D2
 MULU #12,D1
 SWAP D2
 MULU #12,D2
 SWAP D2
 CLR  D2
 ADD.L  D2,D1
 ADD.L  D7,D1
 FMOVE.X  0(A3,D0.L),FP0
 FMUL.X 0(A4,D1.L),FP0
 FMOVE  FP7,FP1
 FADD FP0,FP1
 FMOVE  FP1,FP7
 ADDQ.L #1,D7
lar_2 CMP.L D7,D4
 BGE  lar_1

pascal, plain
lam_3 MOVEA.L  param2(A6),A0
 MOVE D6,D0
 MULS #$258,D0
 MOVE D5,D1
 MULS #12,D1
 ADD  D1,D0
 MOVEA.L  param3(A6),A1
 MOVE D5,D1
 MULS #$258,D1
 MOVE D7,D2
 MULS #12,D2
 ADD  D2,D1
 LEA  -$264(A0),A0
 FMOVE.X 0(A0,D0.W),FP0
 LEA  -$264(A1),A0
 FMUL.X 0(A0,D1.W),FP0
 MOVE D6,D0
 MULS #$258,D0
 MOVE D7,D1
 MULS #12,D1
 ADD  D1,D0
 LEA  -$264(A4),A0
 FADD.X 0(A0,D0.W),FP0
 MOVE D6,D0
 MULS #$258,D0
 MOVE D7,D1
 MULS #12,D1
 ADD  D1,D0
 LEA  -$264(A4),A0
 FMOVE.X FP0,0(A0,D0.W)
 ADDQ #1,D5
 BVS.S  lam_5
lam_4 CMP.W vam_1(A6),D5
 BLE  lam_3

c, plain
lao_3 MOVE.LD5,D0
 MOVE.L D0,D1
 MULU #12,D0
 SWAP D1
 MULU #12,D1
 SWAP D1
 CLR  D1
 ADD.L  D1,D0
 ADD.L  D6,D0
 MOVE.L D7,D1
 MOVE.L D1,D2
 MULU #12,D1
 SWAP D2
 MULU #12,D2
 SWAP D2
 CLR  D2
 ADD.L  D2,D1
 ADD.L  D6,D1
 MOVEA.L  param3(A6),A0
 MOVE.L D5,D2
 MOVE.L D2,D3
 MULU #12,D2
 SWAP D3
 MULU #12,D3
 SWAP D3
 CLR  D3
 ADD.L  D3,D2
 ADD.L  D7,D2
 FMOVE.X  0(A4,D1.L),FP0
 FMUL.X 0(A0,D2.L),FP0
 FADD.X 0(A3,D0.L),FP0
 MOVE.L D5,D0
 MOVE.L D0,D1
 MULU #12,D0
 SWAP D1
 MULU #12,D1
 SWAP D1
 CLR  D1
 ADD.L  D1,D0
 ADD.L  D6,D0
 FMOVE.X FP0,0(A3,D0.L)
 ADDQ.L #1,D7
lao_4 CMP.L D7,D4
 BGE  lao_3

Fortran, optimized
lah_3 MOVE-172(A6),D2
 EXT.L  D2
 EXT.L  D2
 SUB.L  -142(A6),D2
 MULS.L #12,D2 
 MOVE.L D2,D0
 MOVE -170(A6),D2
 EXT.L  D2
 EXT.L  D2
 SUB.L  -130(A6),D2
 MULS.L -134(A6),D2 
 ADD.L  D0,D2
 MOVEA.L  32(A6),A0
 ADDA.L D2,A0
 FMOVE.X  (A0),FP7
 MOVE -170(A6),D2
 EXT.L  D2
 EXT.L  D2
 SUB.L  -118(A6),D2
 MULS.L #12,D2 
 MOVE.L D2,D1
 MOVE -168(A6),D2
 EXT.L  D2
 EXT.L  D2
 SUB.L  -106(A6),D2
 MULS.L -110(A6),D2 
 ADD.L  D1,D2
 MOVEA.L  24(A6),A1
 ADDA.L D2,A1
 FMUL.X (A1),FP7
 FADD.X -94(A6),FP7
 FMOVE.X  FP7,-94(A6)
 ADDQ #1,-170(A6)
 SUBQ.L #1,D5
 BGT  lah_3

Fortran, plain
lae_3 MOVE-164(A6),D2
 EXT.L  D2
 EXT.L  D2
 SUB.L  -134(A6),D2
 MULS.L #12,D2 
 MOVE.L D2,D0
 MOVE -162(A6),D2
 EXT.L  D2
 EXT.L  D2
 SUB.L  -122(A6),D2
 MULS.L -126(A6),D2 
 ADD.L  D0,D2
 MOVEA.L  32(A6),A0
 ADDA.L D2,A0
 FMOVE.X  (A0),FP7 ; get a(i,k)
 MOVE -162(A6),D2
 EXT.L  D2
 EXT.L  D2
 SUB.L  -110(A6),D2
 MULS.L #12,D2 
 MOVE.L D2,D1
 MOVE -160(A6),D2
 EXT.L  D2
 EXT.L  D2
 SUB.L  -98(A6),D2
 MULS.L -102(A6),D2 
 ADD.L  D1,D2
 MOVEA.L  24(A6),A1
 ADDA.L D2,A1
 FMUL.X (A1),FP7 ; multiply by b(i,k)
 MOVE -164(A6),D2
 EXT.L  D2
 EXT.L  D2
 SUB.L  -158(A6),D2
 MULS.L #12,D2 
 MOVE.L D2,D1
 MOVE -160(A6),D2
 EXT.L  D2
 EXT.L  D2
 SUB.L  -146(A6),D2
 MULS.L -150(A6),D2 
 ADD.L  D1,D2
 MOVEA.L  40(A6),A1
 ADDA.L D2,A1
 FADD.X (A1),FP7 ; add c(i,k)
 MOVE -164(A6),D2; this
 EXT.L  D2; whole
 EXT.L  D2; stuff
 SUB.L  -158(A6),D2; is
 MULS.L #12,D2  ;
 MOVE.L D2,D1  ; R
 MOVE -160(A6),D2; E
 EXT.L  D2; D
 EXT.L  D2; U
 SUB.L  -146(A6),D2; N
 MULS.L -150(A6),D2; D
 ADD.L  D1,D2  ; A
 MOVEA.L  40(A6),A1; N
 ADDA.L D2,A1  ; T !!!!
 FMOVE.X  FP7,(A1) ; put back c(i,k)
 ADDQ #1,-162(A6)
 SUBQ.L #1,D5
 BGT  lae_3

 

Community Search:
MacTech Search:

Software Updates via MacUpdate

Espresso 5.1 - Powerful HTML, XML, CSS,...
Note from the developer: For the new Espresso, we changed our versioning and licensing approach with more consistent pricing and a simpler development timeline: "X+1". Each new update would increase... Read more
MacFamilyTree 8.2.7 - Create and explore...
MacFamilyTree gives genealogy a facelift: modern, interactive, convenient and fast. Explore your family tree and your family history in a way generations of chroniclers before you would have loved.... Read more
VueScan 9.6.04 - Scanner software with a...
VueScan is a scanning program that works with most high-quality flatbed and film scanners to produce scans that have excellent color fidelity and color balance. VueScan is easy to use, and has... Read more
WhatsApp 0.2.8000 - Desktop client for W...
WhatsApp is the desktop client for WhatsApp Messenger, a cross-platform mobile messaging app which allows you to exchange messages without having to pay for SMS. WhatsApp Messenger is available for... Read more
Vivaldi 1.13.1008.40 - An advanced brows...
Vivaldi is a browser for our friends. In 1994, two programmers started working on a web browser. Our idea was to make a really fast browser, capable of running on limited hardware, keeping in mind... Read more
TotalFinder 1.10.7 - Adds tabs, hotkeys,...
TotalFinder is a universally acclaimed navigational companion for your Mac. Enhance your Mac's Finder with features so smart and convenient, you won't believe you ever lived without them. Features... Read more
Box Sync 4.0.7886 - Online synchronizati...
Box Sync gives you a hard-drive in the Cloud for online storage. Note: You must first sign up to use Box. What if the files you need are on your laptop -- but you're on the road with your iPhone? No... Read more
iExplorer 4.1.14 - View and transfer fil...
iExplorer is an iPhone browser for Mac lets you view the files on your iOS device. By using a drag and drop interface, you can quickly copy files and folders between your Mac and your iPhone or... Read more
WhatRoute 2.1.1 - Geographically trace o...
WhatRoute is designed to find the names of all the routers an IP packet passes through on its way from your Mac to a destination host. It also measures the round-trip time from your Mac to the router... Read more
TunnelBear 3.5.1 - Subscription-based pr...
TunnelBear is a subscription-based virtual private network (VPN) service and companion app, enabling you to browse the internet privately and securely. Features Browse privately - Secure your data... Read more

Latest Forum Discussions

See All

Everything about Hero Academy 2 - The co...
It's fair to say we've spent a good deal of time on Hero Academy 2. So much so, that we think we're probably in a really good place to give you some advice about how to get the most out of the game. And in this guide, that's exactly what you're... | Read more »
Everything about Hero Academy 2: Part 3...
In the third part of our Hero Academy 2 guide we're going to take a look at the different modes you can play in the game. We'll explain what you need to do in each of them, and tell you why it's important that you do. [Read more] | Read more »
Everything about Hero Academy 2: Part 2...
In this second part of our guide to Hero Academy 2, we're going to have a look at the different card types that you're going to be using in the game. We'll split them up into different sections too, to make sure you're getting the most information... | Read more »
Everything about Hero Academy 2: Part 1...
So you've started playing Hero Academy 2, and you're feeling a little bit lost. Don't worry, we've got your back. So we've come up with a series of guides that are going to help you get to grips with everything that's going on in the game. [Read... | Read more »
What mobile gaming can learn from the Ni...
While Nintendo might not have had things all its own way since it began developing for mobile, one thing it has got right is the release of the Switch. After the disappointment of the WiiU, which I still can't really explain, the Switch felt a... | Read more »
Programmer of Sonic The Hedgehog launche...
Japanese programmer Yuji Naka is best known for leading the team that created the original Sonic The Hedgehog. He’s moved on from the speedy blue hero since then, launching his own company based in Tokyo – Prope Games. Legend of Coin is the... | Read more »
Why doesn't mobile gaming have its...
The Overwatch League is a pretty big deal. It's an attempt to really push eSports into the mainstream, by turning them into, well, regular sports. But slightly less sweaty. It's a lavish affair with teams from all around the world, and more... | Read more »
Give Webzen’s new billiard game PoolTime...
Best known for producing hugely popular MMO titles, South Korean publisher Webzen is now taking aim at a different genre altogether. PoolTime is a realistic eight ball pool simulator, allowing you to compete in real-time matches against players... | Read more »
Let Them Come Guide - How to survive aga...
Let Them Come is all about making it as far as possible against overwhelming odds. Check out some of these tips to help you last a little longer in your unwinnable fight: [Read more] | Read more »
All the best games on sale for iPhone an...
Happy last day of the week. I hope you've been having a good one. I have. I saw ten doggos today. So because I'm in a good mood, I thought I'd round up all of the best games that are currently on sale on the App Store. [Read more] | Read more »

Price Scanner via MacPrices.net

9.7-inch 2017 WiFi iPads on sale starting at...
B&H Photo has 9.7″ 2017 WiFi #Apple #iPads on sale for $30 off MSRP for a limited time. Shipping is free, and pay sales tax in NY & NJ only: – 32GB iPad WiFi: $299, $30 off – 128GB iPad WiFi... Read more
Wednesday deal: 13″ MacBook Pros for $100-$15...
B&H Photo has 13″ #Apple #MacBook Pros on sale for up to $100-$150 off MSRP. Shipping is free, and B&H charges sales tax for NY & NJ residents only: – 13-inch 2.3GHz/128GB Space Gray... Read more
Apple now offering Certified Refurbished 2017...
Apple has Certified Refurbished 9.7″ WiFi iPads available for $50-$80 off the cost of new models. An Apple one-year warranty is included with each iPad, and shipping is free: – 9″ 32GB WiFi iPad: $... Read more
10″ iPad Pros on sale for $50-$75 off MSRP, n...
B&H Photo has 10″ and #Apple #iPad Pros on sale for up to $75 off MSRP. Shipping is free, and B&H charges sales tax in NY & NJ only. Note that some sale prices are restricted to certain... Read more
Apple refurbished Mac minis available startin...
Apple has restocked Certified Refurbished Mac minis starting at $419. Apple’s one-year warranty is included with each mini, and shipping is free: – 1.4GHz Mac mini: $419 $80 off MSRP – 2.6GHz Mac... Read more
Amazon offers Silver 13″ Apple MacBook Pros f...
Amazon has new Silver 2017 13″ #Apple #MacBook Pros on sale today for up to $150 off MSRP, each including free shipping: – 13″ 2.3GHz/128GB Silver MacBook Pro (MPXR2LL/A): $1199.99 $100 off MSRP – 13... Read more
Sale: 12″ 1.3GHz MacBooks on sale for $1499,...
B&H Photo has Space Gray and Rose Gold 12″ 1.3GHz #Apple MacBooks on sale for $100 off MSRP. Shipping is free, and B&H charges sales tax for NY & NJ residents only: – 12″ 1.3GHz Space... Read more
Apple offers Certified Refurbished 2017 iMacs...
Apple has a full line of Certified Refurbished iMacs available for up to $350 off original MSRP. Apple’s one-year warranty is standard, and shipping is free. The following models are available: – 27... Read more
13″ MacBook Airs on sale for $120-$100 off MS...
B&H Photo has 2017 13″ 128GB MacBook Airs on sale for $120 off MSRP. Shipping is free, and B&H charges sales tax for NY & NJ residents only: – 13″ 1.8GHz/128GB MacBook Air (MQD32LL/A): $... Read more
15″ Touch Bar MacBook Pros on sale for up to...
Adorama has Space Gray 15″ MacBook Pros on sale for $200 off MSRP. Shipping is free, and Adorama charges sales tax in NJ and NY only: – 15″ 2.8GHz MacBook Pro Space Gray (MPTR2LL/A): $2199, $200 off... Read more

Jobs Board

Commerce Engineer, *Apple* Media Products -...
# Commerce Engineer, Apple Media Products Job Number: 113161479 Santa Clara Valley, California, United States Posted: 01-Nov-2017 Weekly Hours: 40.00 **Job Summary** Read more
*Apple* Retail - Multiple Positions - Apple,...
Job Description:SalesSpecialist - Retail Customer Service and SalesTransform Apple Store visitors into loyal Apple customers. When customers enter the store, Read more
Site Reliability Engineer, *Apple* Pay - Ap...
# Site Reliability Engineer, Apple Pay Job Number: 113356036 Santa Clara Valley, California, United States Posted: 12-Jan-2018 Weekly Hours: 40.00 **Job Summary** Read more
UI Tools and Automation Engineer, *Apple* M...
# UI Tools and Automation Engineer, Apple Media Products Job Number: 86351939 Santa Clara Valley, California, United States Posted: 11-Jan-2018 Weekly Hours: 40.00 Read more
*Apple* Retail - Multiple Positions - Apple,...
Job Description: Sales Specialist - Retail Customer Service and Sales Transform Apple Store visitors into loyal Apple customers. When customers enter the store, Read more
All contents are Copyright 1984-2011 by Xplain Corporation. All rights reserved. Theme designed by Icreon.