is there a way to (re)define true/false to be what I want it to exact?
Hi
true/false is vague in C,but if I want to control things with it,it would be helpful to make it work like lowlevel code
if true=1 and false=0,different definition to use is true=0xffffffff all bits set and false=0,so you can use it as mask out things with AND
x=x+(xd&booleanvariable) ;//AND zero=zero,AND 0xffffffff =x=x+xd
while if true=1 and false=0,multiply with 1
x=x+(xd*booleanvariable);
//gets ridicoulus high results if booleanvariable is 0xfffffff
you tell me you can C,why dont you C your own bugs?
TRUE becomes all bits set in whatever width of data type you end up assigning it to.Code:#define FALSE 0u #define TRUE (~FALSE)
If you dance barefoot on the broken glass of undefined behaviour, you've got to expect the occasional cut.
If at first you don't succeed, try writing your phone number on the exam paper.
Vague? ISO defines the result of boolean expressions to be 0 (false) or 1 (true). But for testing boolean values, 0 is false and non-zero, true.
I would use
if you want to ensure the 0,1 values on expressions:Code:#define FALSE 0 #define TRUE 1
You can use a macro to adjust the boolean value:Code:x = x + ( x & !!booleanvar );
Code:#define BOOLEAN(x) (!!(x)) ... x = x + ( x & BOOLEAN(booleanvar) );
Last edited by flp1969; 02-20-2021 at 07:23 AM.
thanks salem,flp1969
you tell me you can C,why dont you C your own bugs?
Just another tip... notice that there are BOOLEAN operators and BINARY operators for logical operations... The expression above could be writen as:
Code:x += ( x && boolvar );
thanks,flp1969
I am trying something similar awsdert do,but with jumpless conditionals
you tell me you can C,why dont you C your own bugs?
Always a good idea to give a hand to the compiler, but be aware that it's not granted that a C jumpless code is compiled to a binary jumpless one...
Usually, when using optimizations, the compiler will try to eliminate jumps, even with conditional ("if' or "?" operator) and loops... But it always will try to create faster code (with options as -O2 or -O3), using jumps or not.
Here's an example:
The g() function isn't jumpless (in C) and isn't jumpless in assembly:Code:int f( int x, int b ) { return x + ( x && b ); } int g( int x, int b ) { if ( x && b ) x++; return x; }
But g() is faster. If x is 0 there's a penalty for the forward conditional jump, but it is way faster than executing 5 instructions (before ret).Code:f: g: test edi, edi mov eax, edi setne dl test edi, edi xor eax, eax je .L4 test esi, esi test esi, esi setne al setne dl and eax, edx cmp dl, 1 add eax, edi sbb eax, -1 ret .L4: ret
if x isn't zero, the performance is equivalent for 1 less instruction than f().
And... sorry... I made a little mistake... To emulate && with the binary logical operator & you need to transform the values in both variables to boolean (0 or 1), like this:
This will create code exactly asCode:x += ( !!x & !!b );
Code:x += ( x && b );
> Always a good idea to give a hand to the compiler,
Or a bad idea.
If you write overly micro-managed "efficient" code, you typically subvert the ability of the optimiser to see through your cleverness and generate optimal code.
Write clean simple code in the first instance, then measure the performance.
If you dance barefoot on the broken glass of undefined behaviour, you've got to expect the occasional cut.
If at first you don't succeed, try writing your phone number on the exam paper.
all compile options seem like a jungle,"whole program optimization", does what?glances outside the boxes(functions) to use a compatible same optimization algo,instead of optimal suited optimize algo for each one?,remember one advice against inline asm,is it interferes with the whole compiler code optimization?Originally Posted by Salem
"enable fiber-safe optimization", totally cryptic too me
seem like its time to test options like "enable parallel code generation" its only suggesting to do it,not control it todo
you tell me you can C,why dont you C your own bugs?
