nir/lower_vec_to_movs: Only coalesce if the vec had a SSA destination

[mesa.git] / src / compiler / nir / nir_lower_vec_to_movs.c

diff --git a/src/compiler/nir/nir_lower_vec_to_movs.c b/src/compiler/nir/nir_lower_vec_to_movs.c
index 0ba374937e8cd7a19cb19936b0860ead790617a9..8b24376b0a5befa4982c4b47187fa6f70113fe8f 100644 (file)
--- a/src/compiler/nir/nir_lower_vec_to_movs.c
+++ b/src/compiler/nir/nir_lower_vec_to_movs.c
@@ -118,7 +118,7 @@ has_replicated_dest(nir_alu_instr *alu)
  * can then call insert_mov as normal.
  */
 static unsigned
-try_coalesce(nir_alu_instr *vec, unsigned start_idx, nir_shader *shader)
+try_coalesce(nir_alu_instr *vec, unsigned start_idx)
 {
    assert(start_idx < nir_op_infos[vec->op].num_inputs);
 
@@ -131,7 +131,7 @@ try_coalesce(nir_alu_instr *vec, unsigned start_idx, nir_shader *shader)
    /* If we are going to do a reswizzle, then the vecN operation must be the
     * only use of the source value.  We also can't have any source modifiers.
     */
-   nir_foreach_use(vec->src[start_idx].src.ssa, src) {
+   nir_foreach_use(src, vec->src[start_idx].src.ssa) {
       if (src->parent_instr != &vec->instr)
          return 0;
 
@@ -215,7 +215,7 @@ lower_vec_to_movs_block(nir_block *block, nir_function_impl *impl)
    bool progress = false;
    nir_shader *shader = impl->function->shader;
 
-   nir_foreach_instr_safe(block, instr) {
+   nir_foreach_instr_safe(instr, block) {
       if (instr->type != nir_instr_type_alu)
          continue;
 
@@ -230,6 +230,7 @@ lower_vec_to_movs_block(nir_block *block, nir_function_impl *impl)
          continue; /* The loop */
       }
 
+      bool vec_had_ssa_dest = vec->dest.dest.is_ssa;
       if (vec->dest.dest.is_ssa) {
          /* Since we insert multiple MOVs, we have a register destination. */
          nir_register *reg = nir_local_reg_create(impl);
@@ -263,8 +264,12 @@ lower_vec_to_movs_block(nir_block *block, nir_function_impl *impl)
          if (!(vec->dest.write_mask & (1 << i)))
             continue;
 
-         if (!(finished_write_mask & (1 << i)))
-            finished_write_mask |= try_coalesce(vec, i, shader);
+         /* Coalescing moves the register writes from the vec up to the ALU
+          * instruction in the source.  We can only do this if the original
+          * vecN had an SSA destination.
+          */
+         if (vec_had_ssa_dest && !(finished_write_mask & (1 << i)))
+            finished_write_mask |= try_coalesce(vec, i);
 
          if (!(finished_write_mask & (1 << i)))
             finished_write_mask |= insert_mov(vec, i, shader);
@@ -300,7 +305,7 @@ nir_lower_vec_to_movs(nir_shader *shader)
 {
    bool progress = false;
 
-   nir_foreach_function(shader, function) {
+   nir_foreach_function(function, shader) {
       if (function->impl)
          progress = nir_lower_vec_to_movs_impl(function->impl) || progress;
    }