EDMA3 Intermediate chaining (for breaking up large transfers)

Hi,

My setup :

C6678 EVM, CCS 5.1.1.00031, Compiler v3.3, mcsdk_2_00_05_17, pdk_C6678_1_0_0_17, edma3_lld_02_11_03_02

Background :

I've been playing around EDMA3 for a few days now.

I've successfully tested the edma_test.c examples that comes with the CSL.

I've also tested several simple cases that worked well such as : simple block transfer using global regions, simple block transfer using shadow regions, chained block transfer with shadow regions, 2D block transfer.

My problem :

I'm now trying to do the 'Breaking Up Large Transfers with Intermediate Chaining' of p99/173 of the EDMA3 controller guide

I want to transfer 10240 short elements (2bytes) as 10*1024 transfers. These elements are in external memory and will be transfered to L2 memory

Right now, only the first transfer of 1024 elements works, the self-chaining doesn't work

I use the instance 1, channel 1, shadow region 5

myParamSetup is as follows :

    myParamSetup.option = CSL_EDMA3_OPT_MAKE( CSL_EDMA3_ITCCH_EN, \
                                              CSL_EDMA3_TCCH_DIS, \
                                              CSL_EDMA3_ITCINT_DIS, \
                                              CSL_EDMA3_TCINT_EN, \
                                              channelNum, \
                                              CSL_EDMA3_TCC_NORMAL, \
                                              CSL_EDMA3_FIFOWIDTH_NONE, \
                                              CSL_EDMA3_STATIC_DIS, \
                                              CSL_EDMA3_SYNC_A, \
                                              CSL_EDMA3_ADDRMODE_INCR, \
                                              CSL_EDMA3_ADDRMODE_INCR
                                             );

Additionnal info : channelNum = 1, ITCCH_EN because I want the self chaining,
 TCCH_DIS because it's a one time transfer, TCINT_EN 
because at the end of the 10 transfers I want to poll regionIntr.intr & (1<<channelNum) to know when it's finished.

    myParamSetup.srcAddr    = (Uint32)block_src;
    myParamSetup.aCntbCnt   = CSL_EDMA3_CNT_MAKE(ACNT,BCNT);
    myParamSetup.dstAddr    = (Uint32)block_dst + L2Offset;      //  global adress
    myParamSetup.srcDstBidx = CSL_EDMA3_BIDX_MAKE(SRCBIDX,DSTBIDX); // 0,0 in this case
    myParamSetup.linkBcntrld= CSL_EDMA3_LINKBCNTRLD_MAKE(CSL_EDMA3_LINK_NULL,0);
    myParamSetup.srcDstCidx = CSL_EDMA3_CIDX_MAKE(0,0);
    myParamSetup.cCnt = CCNT;

with

        #define TYPE_BYTE_SIZE 2      //using 'short'

	#define NB_ELEMENTS_PER_TRANSFER 1024
        #define NB_TRANSFERS 10

	#define ACNT (NB_ELEMENTS_PER_TRANSFER*TYPE_BYTE_SIZE)
	#define BCNT NB_TRANSFERS
	#define CCNT 1
	#define SRCBIDX 0 								 	   // default
	#define DSTBIDX 0

and 

	coreNum = 0;
	L2Offset = 0x10000000 + coreNum*0x01000000;

I also have in my cmd file

    .block_src_DATA	: > 0x80000000     // source in external memory
    .block_dst_DATA	: > 0x00800000     // destination in L2 memory

with block_src and block_dst of 10240 size.

It is supposed to work like this :

"Upon completion of the first array, intermediate transfer complete code chaining 
generates a synchronization event to channel 1, a value specified by the TCC field. 
This intermediate transfer completion chaining event causes EDMA3 channel 1 to transfer the next array"

but it seems like it doesn't decrement BCNT and only do it once. It doesn't self-chain.

Can you explain to me what I'm doing wrong ?

I can provide the files if needed.

Thank you,

CM