TMS320F280039C: CRC 32 C calculation problem in the library

Mahmoud

See attached picture below, which describes the sequence of operations to compute a CRC32 vs CRC32-MPEG2.

For CRC32, you will use the flipInputBuf_cpu(), followed by getCRC32_vcu(), followed by __flip32() and finally logical NOT to get the result.

Yes, CRC0 does not support the CRC32 polynomial, so this method is applicable to VCU2 and the VCRC (F28003x).

Thanks,

Sira

thanks for your response 
will the flipInputBuf_cpu() function cause stack overflow problem if it has to flip a very large data "the app data sent to bootloader" because i already implemented code that flips the array elements and caused a problem for large data 

Mahmoud,

If the buffer to be flipped is very large, then the flipping and CRC operations may need to broken down into smaller chunks and called in a loop.

But the bigger question is - if you're using F28003x, it has the VCRC, which has HW support to flip data before computing the CRC. So the external flipping operation is not needed.

See below from the VCRC chapter of spruhs1

Thanks,

Sira

Sira,
well first i looked for the registers of the VCU during debugging but there don't show values, it says "Error Unable to read" 

second i searched the Cware folder for that register you mentioned CRCMSGFLIP and found some assembly files, i don't know how to integrate that assembly code in my project ? the file name is vcrc_configpoly_asm 

i just put the file on my project directory and called the _CRC_runConfigPolyBytesReflected: but it had no effect ? 

what should i include so that the assembly file will work on my project ? 

one more question 
i don't understand the difference between 
VCRC, VCU0, VCU2 for my code i used the VCRC library, could you tell me what is the difference bewteen those libraries ? 

thanks 

Mahmoud,

VCU0, VCU2, VCRC are Hardware IPs on chip. VCU0 came first (with CRC, FFT, etc. capabilities), VCU2 came next (with some enhancements e.g. support for 2 CRC polynomials, flipping prior to compute, etc.). Then came VCRC (with support removed for everything other than CRC, but added CRC configuration capability - like variable length, configurable polynomial etc.

Please refer to the 28003x_vcrc_crc_32 example in C2000Ware, which has a call to the CRC_run32BitPoly2Reflected function of the library. Also look at the definition of this function in the VCRC library (c28x_vcrc_library) in vcu2_crc_32.asm.

The register display issue I'm not so sure about - I'll have to check it.

Thanks,

Sira

thanks but i ran into a second problem which is the limited length the CRC function can take to calculate the CRC, i want to calculate the CRC on a very big chunk of memory  98 K bytes, when calculated it give me a wronf answer after the length of 32 K bytes maybe there is a limit on that function here 

so my proposed solution would be to make the calculations on chunks "if that is true and the function and the CRC hardware has some limitations" then i want to know please how to do it ? let's say i calculated the first 32 k byte i got a value crc for first 32 k byte  how to use that value with the second chunk of memory which parameter to used i know i will add a second input to the function for that  but how to use the previous CRC value for the next chunk to calculate the whole 98 k bytes 

thanks 

i also tried the CRC_run32BitPoly2Reflected  but it didn't work, it returned a value that is not on the online calculator 
if the input to the function is 0x0096  "the value to be calculated by CRC " then what should i type in the online calculator ? 

Please give me a day or two to look into these details.

okay, thanks i will be waiting for your response 
just to confirm about the length limit, i tested it in my code then i read it in one of the header files vcu2_crc,h there is a limit 

Mahmoud, that would be the seedValue. This is what you would pass into the subsequent CRC compute call.

Yes, that's correct, there is indeed a documented upper limit on this.

Typically when the values don't match, it's usually a case of the config parameters and pre and post-processing not matching up to what the online calculator is expecting. There are so many parameters here:

Like:

Seed value

Flip or no Flip

High-Low byte swap

This is what I sent in a picture in one of my earlier mails i.e. what works for CRC32 and what works for CRC32-MPEG2.

Thanks,

Sira

i tried to pass the seed value but it didn't work, have a look here 

result two should match result 3 but it didn't as you can see, result three matches the online calculator with flipping done 

Mahmoud,

I'm a bit confused when I look at your previous picture of the definition of the generate_crc32c function - where it references Polynomial2 (0x1edc6f41) whereas your latest picture is trying to match up to CRC-32C where the polynomial specified is our Polynomial1 (0x04c11db7).

1. So can you please send a picture showing your latest generate_crc32c function?

2. In the online calculator tool, what does RefIn = true, RefOut = true mean? Does it mean the input is Reflected and so is the output?

3. What does XOROUT = 0xFFFFFFFF mean? Does it mean the output (after Reflection) is XORd with 0xFFFFFFFF?

4. What does the CHECK value refer to?

Looking at your previous definition of generate_crc32c, I think the issue may be that your are computing the XOR on the intermediate results (e.g. result1), whereas you should do so only on the final output. So the XOR operation should be outside the generate_crc32c function.

Thanks,

Sira