Hi.
I'm using the GCC compiler for PPC405 as delivered with Xilinx EDK 8.1.
The exact version is: powerpc-eabi-gcc (GCC) 3.4.1 ( Xilinx EDK 8.1.1
Build EDK_I.19.3 190106 )
I found a bug in the code generation.
My code is a serial receiver and uses an uninitialized uchar nData:
uchar nBits;
uchar nData;
for (nBits=0; nBits<8; nBits++) {
nData >>= 1; // the bug is related to this statement
if (function()) nData |= 0x80;
}
Although nData is not initialized, the correct state after the first
"nData >>= 1" statement would be any value between 0x00 and 0x7f.
However, on my target board I found it to be >= 0x80.
Looking at the disassembler output shows that the compiler used the
following assembler instruction to implement the "nData >>= 1" statement:
rlwinm r26,r26,31,24,31 // incorrect implementation, as
generated by gcc
It is a rotate-left by 31 bits (equals a rotate-right by 1 bit), and
subsequent AND-mask with 0xFF (powerpc-eabi is a big endian machine).
This command is not a correct implementation of the statement. It is
only correct if bit 0 of R26 (which will become bit 31 after the rotate)
happens to be 0.
The correct implementation would be to use a different mask value, which
ensures that the "new" bit 31 is 0 (mask 0x7F instead of 0xFF):
rlwinm r26,r26,31,24,30 // corrected implementation
I can work around this bug by initializing nData before the first use,
because GCC will initialize all of r26 (including bit 0) . However I
still consider it a bug that should be addressed.
Keep on good work!
Regards from Spain,
Marc
