C2000, CGTools 6.1.3 assembler error E5369

Update: In another project I get the E5369 error message for a C file compiled with CGTools 6.1.3 and the following options (path and file names changed):

-q -pdr -pdv -pden -pds880 -o2 -fr"..." -i"..." -d"_DEBUG" -d"LARGE_MODEL" -d"__TMS320F28XX__"  -ml -v28 --float_support=fpu32 -mf2 --gen_func_subsections -@"../Debug.lkf" "file.c"

Regards,

Johannes

5 delay slots are needed to protect against a hardware silicon exception.  If a memory-mapped register write (MOV32 R0H, ACC) is followed by either FRACF32, UI16TOF32, or F32TOUI32 and there are less than 5 delay cycles in between, R0H may be updated with the result of the latter instruction instead of the data on the write data bus.  These were the only affected instructions.

Only active branches of the compiler are updated so the fix was not applied to v.5.2.

Anna Youssefi

Also, please grab the latest update (currently 6.1.4) because after the initial fix was applied, a second update was made.  In some cases the compiler missed adding the additional delay cycle and the assembler caught it and issued a complaint.  That was fixed shortly before 6.1.3 was released and I don't know if it made it in to the release. If you were getting the warning with compiler-generated code, it is likely to be from that scenario.

If any problem persists with compiler-generated code after the latest compiler version is used, please let me know.

Thank you,

Anna Youssefi

The problem persists with 6.1.4. The following code

extern unsigned long p2(void);
extern float foo(float, float);
void Test(unsigned long k[], float dt) {
  k[0] = p2() ?  0 : (unsigned long) foo(1, dt);
}

triggers the error:

[ui32test.c] "C:\Programme\Texas Instruments\C2000 Code Generation Tools 6.1.4\bin\cl2000" -g -k -pdsw225 -o2 -fr... -d"_DEBUG" -d"LARGE_MODEL" -ml -v28 --float_support=fpu32 -@"Debug.lkf" "ui32test.c"
"ui32test.asm", ERROR!   at line 107: [E5369] Due to FPU pipeline restriction,
                                              this instruction is illegal in
                                              the 5 delay slots of:  [MOV32    
                                              R1H,XAR2              ; [CPU_]
                                              |5|, line 100]
            F32TOUI32 R0H,R0H               ; [CPU_] |5|

1 Assembly Error, No Assembly Warnings

Regards, Johannes

Anna Youssefi said:

If the code is still generating an object file, you can use it safely.

No object file is generated when this error occurs.

When compiled without -o2, the error doesn't occur.

Regards, Johannes

Hi George (and all);

I checked the incident report SDSCM00047353  and it says the status is FIXED, in release 6.1.5 of C2000 Code Generation Tools.

I downloaded and installed C2000 CGTools v6.1.5 from this link: https://www-a.ti.com/downloads/sds_support/TICodegenerationTools/download.htm

I selected Compiler version: TI v6.1.5 and built my project, and got same error [E5369].

Same project, compiler version TI v6.2.3, build fails with same error E5369.

Same project, compiler version TI v5.2.15, builds successfully with no errors.

So... some permutation of this bug seems to persist.  Can I provide any further information which might help to pinpoint the remaining problem?

Thanks very much,

Amos Gaynes

My guess is that this bug is only partially fixed, and you have found another way to trigger it.  I'd appreciate if you would send us a test case so we could investigate.  You probably need to preprocess like this.  Please include the exact version and options you are using.

Thanks and regards,

-George

Hi George,

While compiling the attached file I am getting the error as mentioned in this post. I can reproduce the issue with the compile version 6.2.5.

I have used the below command from the commandline to compile the file:

"C:/TI/ccsv5/tools/compiler/c2000_6.1.6/bin/cl2000"  -ml --float_support=fpu32 -v28 -I"C:/TI/ccsv5/tools/compiler/c2000_6.1.6/include" main.c

/*
 * main.c
 */
int main(void) {
	long rtb_Add;
	  float tmp =7;
	rtb_Add = tmp < 0.0F ? -(long)(unsigned long)-tmp : (long)(unsigned long)tmp;

}

Thank you for submitting a test case.  I can reproduce your result.  While it appears similar to the previous complaints on this this thread, the circumstances are different enough that a I filed a new entry SDSCM00050302 in the SDOWP system.  You are welcome to track it with the SDOWP link below in my signature.

Thanks and regards,

-George