I've been carrying this for a few years. One test in the GCC testsuite is
failing due to a bug in the handling of the v850e3v5 instruction "bins".
When the "bins" instruction specifies a 32bit bitfield size, the simulator
exhibits undefined behavior by trying to shift a 32 bit quantity by 32 bits.
In the case of a 32 bit shift, we know what the resultant mask should be. So
we can just set it.
That seemed better than using 1UL for the constant (on a 32bit host unsigned
long might still just be 32 bits) or needlessly forcing everything to
long long types.
Thankfully the case where this shows up is only bins <src>, 0, 32, <dest>
which would normally be encoded as a simple move.
* testsuite/v850/allinsns.exp: Add v850e3v5.
* testsuite/v850/bins.cgs: New test.
* v850/simops.c (v850_bins): Avoid undefined behavior on left shift.
# All machines.
# Should add more cpus if the testsuite adds coverage for their insns, but
# at the core level, there's no deviation beyond these two.
-set all_machs "v850e v850"
+set all_machs "v850e3v5 v850e v850"
# gas doesn't support any '=' option for v850.
set cpu_option_sep ""
--- /dev/null
+# v850 bins
+# mach: v850e3v5
+# as: -mv850e3v5
+
+ .include "testutils.inc"
+
+ seti 0x7fff, r10
+ seti 0x0, r11
+ bins r10, 0, 32, r11
+ reg r11, 0x7fff
+
+ pass
pos = lsb;
width = (msb - lsb) + 1;
- mask = ~ (-(1 << width));
+ /* A width of 32 exhibits undefined behavior on the shift. The easiest
+ way to make this code safe is to just avoid that case and set the mask
+ to the right value. */
+ if (width >= 32)
+ mask = 0xffffffff;
+ else
+ mask = ~ (-(1 << width));
+
source &= mask;
mask <<= pos;
result = (* dest) & ~ mask;