Dear experts:
I did some RAM speed tests on both A72 core and R5 core, A72 runs Linux and R5 runs sysbios. RAM performance seems fine on A72 core, but I got a bad speed result on R5 core.
The table below is the R5 core RAM speed result, each test do a 16MB data processing
I wonder if these result is correct. If yes, then could you give me some advice to optimise it? Or is R5 core RAM performance totally fixed to this?
Thanks
Hi Piyali,
My test code is simple:
int test_copy_size = 0x1000000;
void *cache_src = (void *)0x81000000;
void *cache_dst = (void *)0x82000000;
void *noncache_src = (void *)0xa5000000;
void *noncache_dst = (void *)0xa6000000;
Types_FreqHz freq;
Types_Timestamp64 t64, tt64;
Timestamp_getFreq(&freq);
Timestamp_get64(&t64);
memset(cache_src, 0xa5, test_copy_size);
Timestamp_get64(&tt64);
CpswAppUtils_print("--- time64: %lu %09lu\n", (t64.hi << 2) + t64.lo / freq.lo, t64.lo % freq.lo);
CpswAppUtils_print("--- time64: %lu %09lu\n", (tt64.hi << 2) + tt64.lo / freq.lo, tt64.lo % freq.lo);
Repeat memset() and memcpy() with different address areas.
I load the binary through CCS, so there is only one R5 core running, other cores remain halted.
And here is the R5 core MPU settings:
I think the compiler flags is right, since the instructions in disassembly seem no problem. R5 core is busy looping in the pictures below
memset:
memcpy:
Thanks for your confirmation.