GCC optimization question...

[relminator]

Hi, I was wondering of how GCC optimizes equations.

Say I have this:

Code: Select all

int a = (0.3f) * (1 << 12);

Would the right-hand-side be evaluated into a single constant at compile time?

Thanks!!!

[vuurrobin]

AFAIK, it should make it 1 constant, but I don't know how it will treat floats, seeing as the DS has no floating point unit.


asuming of course that you use optimisation level 2 (-O2), which should be standard in the makefile.

[relminator]

So at -O2 even literal floats * literal ints would be converted to a literal constant?

[WinterMute]

If the compiler can tell that an expression resolves to a compile time constant then it will emit that constant rather than the expression it's derived from when optimisation is enabled. It's fairly simple to see this using the -save-temps command and examining the emitted assembly.

Code: Select all

monalisa:test davem$ cat testconst.c 
int testconst() {
	int a = (0.3f) * (1<<12);
	return a;
}

monalisa:test davem$ arm-eabi-gcc -c -O2 -march=armv5te -mtune=arm946e-s -mthumb -save-temps testconst.c 
monalisa:test davem$ cat testconst.s 
	.arch armv5te
	.fpu softvfp
	.eabi_attribute 20, 1
	.eabi_attribute 21, 1
	.eabi_attribute 23, 3
	.eabi_attribute 24, 1
	.eabi_attribute 25, 1
	.eabi_attribute 26, 1
	.eabi_attribute 30, 2
	.eabi_attribute 18, 4
	.code	16
	.file	"testconst.c"
	.text
	.align	2
	.global	testconst
	.code	16
	.thumb_func
	.type	testconst, %function
testconst:
	ldr	r0, .L2
	@ sp needed for prologue
	bx	lr
.L3:
	.align	2
.L2:
	.word	1228
	.size	testconst, .-testconst

[relminator]

Thanks dude!!