CCSv5.1.1: compiler issue?

Stefano Sartorelli

Prodigy 180 points

Hi,

I am using CCS 5.1.1 under linux and I am experimenting the following issue.

This is an excerpt of the map file:

text (code area)

..................

                  0c1a6b20    00000010                                                  : PLT_bcp.oe66 ($Tramp$S$$PLT_Bcp_getUlTxQue)
                  0c1a6b30    00000010     --HOLE-- [fill = 0]

.prachBuffers (data area)
*          0    0c1a6b30    0002b000     UNINITIALIZED
                  0c1a6b30    0001e000     phy_phy_phy_ul_debug_e66.ae66 : PHY_TSK_prachReceiver.oe66 (.prachBuffers)
                  0c1c4b30    0000c000                                                  : PHY_UL_prachDecimationFilter.oe66 (.prachBuffers)

It can be seen that an hole is inserted at 0x0c1a6b30, probably to get aligned, but actually there is no hole, because data area start at 0x0c1a6b30 too, overlapping.

Is the above a compiler bug?

Any feedback is welcome.

thanks

Regards

Stefano

over 13 years ago

0 Archaeologist over 13 years ago

TI__Guru* 84285 points

Probably.

This should be moved to the compiler forum.

We're going to need a complete, reproducible test case which demonstrates this problem so that it can be analyzed.

0 Stefano Sartorelli over 13 years ago in reply to Archaeologist

Prodigy 180 points

Hello,

yes, you are right, I posted this on the wrong forum sorry.

I can try to provide you more info (xdc tool/compiler versions, some more details on how we compile, ...), but unfortunately I cannot provide you neither the source code nor the compiling environment here.

I would be pleased if you are able to provide me hints/workarounds to force a different alignment in some way to overcome the problem. For example I noticed that with CCS 5.0.3 the minimum size looks like 32 bytes, this guarantees alignment even with some space waste, or maybe forcing a different areas alignment to linker.

But if the case is of your interest too, I can open a service request.

Thanks

Regards

Stefano

0 Stefano Sartorelli over 13 years ago in reply to Stefano Sartorelli

Prodigy 180 points

Hello,

the case was already reported in this forum here:

http://e2e.ti.com/support/development_tools/compiler/f/343/t/147084.aspx

And a defect was already opened: SDSCM00042408

Solution available in C6000 Code Generation Tools 7.3.6

Workaround: use ">" operator in place of ">>"

Thanks

Regards

Stefano

Code Composer Studio™︎

Code Composer Studio forum

CCSv5.1.1: compiler issue?