TDA3XEVM: Writing conditional loops in Kernel C for VCOP.

Theodoros Karatapanis

Part Number: TDA3XEVM

Hello,

I have problems porting some algorithm written in C to Kernel C using the BAM framework. I am using only the Programmer's guide (version E) for EVE, currently I want to skip some iteration in a loop but I don't know how. Some sample written in C is this :

for(i1=0;i1<blkHeight;i1++)

for(i2=0; i2<blkWidth; i2++)

{

if (blkHeight==0)

break;

/*CODE FOR LOOP*/

}

How can I achieve the same functionality in Kernel C?

Additionally I saw that conditional moves can't be used with loads or stores or with more than 1 instruction. Is there any way to use if for a block of instructions?

Best regards,

Theodore

over 6 years ago

0 Yordan Kamenov over 6 years ago

TI__Mastermind 42515 points

Hi Theodore,

I have forwarded your question to EVE experts.

Regards,
Yordan

0 Anshu Jain over 6 years ago

TI__Guru 55120 points

Hi Theodore,

VCOP doesnt allow you to have zero loop counts in vloops. In your case you should handle it via ARP32. Basically if blkHeight is zero then ARP32 should not submit the loop to VCOP.

regards,

Anshu

0 Theodoros Karatapanis over 6 years ago in reply to Anshu Jain

Prodigy 195 points

Hi Anshu,

I am sorry i misstyped the question. The condition should be :

If (i1==0)
break;

But also 0 could be replaced by some other value.

Sorry for the confusion.

BR ,
Theodore

0 Anshu Jain over 6 years ago in reply to Theodoros Karatapanis

TI__Guru 55120 points

Hi Theodore,

I am assuming what you need is some kind of early exit mechanism. VCOP does support an early exit mechanism. You can do something like what is shown in the loop below.

for(int i3=0; i3 < 1; i3++)
{
for(int i4=0; i4<1; i4++)
{
__vector V1 = 4;
__vector V2 = 4;
__vector V3 = V1 == V2;
if (V3[0]) goto end_loop1;
...
}
}
end_loop1:;

You can read the details in SPRUHB9B.pdf (section 1.14) documentation which comes as part of ARP32 compiler.

Regards,

Anshu

Processors

Processors forum

TDA3XEVM: Writing conditional loops in Kernel C for VCOP.