ir3/lower_pcopy: Fix bug with "illegal" copies and swaps

If the source and destination were within the same full register, like
hr90.x and hr90.y (which both map to r45.x), then we'd perform the
swap/copy with the wrong register. This broke
dEQP-VK.ssbo.phys.layout.random.16bit.scalar.35 once BDA is enabled.

Fixes: 0ffcb19b9d9 ("ir3: Rewrite register allocation")
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/13818>

1 files changed, 16 insertions, 2 deletions

diff --git a/src/freedreno/ir3/ir3_lower_parallelcopy.c b/src/freedreno/ir3/ir3_lower_parallelcopy.c
index b927e031a17..fbcd6ff62a6 100644
--- a/src/freedreno/ir3/ir3_lower_parallelcopy.c
+++ b/src/freedreno/ir3/ir3_lower_parallelcopy.c
@@ -109,11 +109,18 @@ do_swap(struct ir3_compiler *compiler, struct ir3_instruction *instr,
                     .flags = entry->flags & ~IR3_REG_HALF,
                  });
 
+         /* If src and dst are within the same full register, then swapping src
+          * with tmp above will also move dst to tmp. Account for that here.
+          */
+         unsigned dst =
+            (entry->src.reg & ~1u) == (entry->dst & ~1u) ?
+            tmp + (entry->dst & 1u) : entry->dst;
+
          /* Do the original swap with src replaced with tmp */
          do_swap(compiler, instr,
                  &(struct copy_entry){
                     .src = {.reg = tmp + (entry->src.reg & 1)},
-                    .dst = entry->dst,
+                    .dst = dst,
                     .flags = entry->flags,
                  });
 
@@ -192,9 +199,16 @@ do_copy(struct ir3_compiler *compiler, struct ir3_instruction *instr,
                     .flags = entry->flags & ~IR3_REG_HALF,
                  });
 
+         /* Similar to in do_swap(), account for src being swapped with tmp if
+          * src and dst are in the same register.
+          */
+         struct copy_src src = entry->src;
+         if (!src.flags && (src.reg & ~1u) == (entry->dst & ~1u))
+            src.reg = tmp + (src.reg & 1u);
+
          do_copy(compiler, instr,
                  &(struct copy_entry){
-                    .src = entry->src,
+                    .src = src,
                     .dst = tmp + (entry->dst & 1),
                     .flags = entry->flags,
                  });