agx: fix 16-bit mem swaps

don't clobber r1l

Signed-off-by: Alyssa Rosenzweig <alyssa@rosenzweig.io>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/28483>

src/asahi/compiler/agx_lower_parallel_copy.c

@@ -120,7 +120,8 @@ do_swap(agx_builder *b, const struct agx_copy *copy)
    assert(x.memory == y.memory);
    if (x.memory) {
       agx_index temp1 = agx_register(0, copy->src.size);
-      agx_index temp2 = agx_register(2, copy->src.size);
+      agx_index temp2 =
+         agx_register(agx_size_align_16(copy->src.size), copy->src.size);
 
       agx_index scratch_reg2 = agx_register(0, copy->src.size);
       agx_index scratch_mem2 = scratch_slot(b->shader, copy->src.size);

src/asahi/compiler/test/test-lower-parallel-copy.cpp

@@ -302,6 +302,7 @@ TEST_F(LowerParallelCopy, StackCopies)
       scratch_vec2.size = AGX_SIZE_16;
       temp1.size = AGX_SIZE_16;
       temp2.size = AGX_SIZE_16;
+      temp2.value = 1;
 
       agx_mov_to(b, scratch_vec2, spilled_gpr_vec2);
       agx_mov_to(b, temp1, agx_memory_register(807, AGX_SIZE_16));