I added the __sgi thing in there. I really don't want evil.h in there at all. Change the default of EVIL_LEVEL to be NONE_AT_ALL or something for now. That way you have to explicitly change it to get anything.
James
Solomon Boulos wrote:
I noticed there is a good faith effort in evil.h to set the EVIL_LEVEL variable to avoid the ASM if __sgi is defined. These checks should probably just be modified to be more robust towards ensuring that it's a gnu compiler on IA32.
On Aug 5, 2005, at 5:21 PM, Abe Stephens wrote:
Hey,
I've noticed that there is some gnu style IA32 assembly in the trunk.
/store/abe/Manta-trunk/Engine/ImageTraversers/AFR/evil.h(248): error: asm statements not supported in this environment
asm("bsr %1,%0;" : "=a" (eax) : "m" (n));
^
/store/abe/Manta-trunk/Engine/ImageTraversers/AFR/evil.h(248): error: expected a ")"
asm("bsr %1,%0;" : "=a" (eax) : "m" (n));
^
/store/abe/Manta-trunk/Engine/ImageTraversers/AFR/evil.h(303): error: asm statements not supported in this environment
asm ("movl %1,%0;" /* copy f into i */
^
/store/abe/Manta-trunk/Engine/ImageTraversers/AFR/evil.h(306): error: expected a ")"
: "=a" (i) /* output: register int i is in % eax *
This is a problem for two reasons, first we don't have an IA-32 processor and second, Intel doesn't support gnu syntax for assembly instructions.
How necessary is this assembly? I am guessing it doesn't work well on PowerPC either. I suspect we can handle this in a similar way to the SCIRun/Threads code via cmake. Any concerns?
Abe
Archive powered by MHonArc 2.6.16.