On 2/15/22 10:56, Jakub Jelinek wrote:
+bool
+gori_compute::condexpr_adjust (irange &r1, irange &r2, gimple *, tree cond,
+ tree op1, tree op2, fur_source &src)
+{
+ int_range_max tmp, cond_true, cond_false;
+ tree ssa1 = gimple_range_ssa_p (op1);
+ tree ssa2 = g
On Tue, Feb 15, 2022 at 10:46:21AM -0500, Andrew MacLeod wrote:
> Bootstraps on x86_64-pc-linux-gnu with no regressions. But I'm running it
> through again to be sure. OK for trunk, or want to wait for the next
> release?
I'd like to see it in GCC 12 because it is a fix for a P1 regression.
>
As requested in the PR, this patch wires GORI into the evaluation of a
COND_EXPR. This late in the cycle, we'll keep it simple.
given COND ? X : Y
if there is any dependency between COND and X or COND and Y, we use the
GORI engine to determine the range of the SSA_NAME in COND when its TRUE
