Skip to content

Commit

Permalink
amd/compiler: fix early-exit for some p_end_linear_vgpr insertion code
Browse files Browse the repository at this point in the history
Uniform merge blocks don't have block_kind_merge.
  • Loading branch information
pendingchaos committed Jul 8, 2019
1 parent 47187f7 commit f918de3
Showing 1 changed file with 2 additions and 2 deletions.
4 changes: 2 additions & 2 deletions src/amd/compiler/aco_spill.cpp
Expand Up @@ -1530,8 +1530,8 @@ void assign_spill_slots(spill_ctx& ctx, unsigned spills_to_vgpr) {
* TODO: Moving the spills and reloads to before p_logical_end might produce
* slightly better code. */
for (Block& block : ctx.program->blocks) {
/* only merge blocks, loop headers and loop exits have logical phis */
if (!(block.kind & (block_kind_merge | block_kind_loop_header | block_kind_loop_exit)))
/* loops exits are already handled */
if (block.logical_preds.size() <= 1)
continue;

bool has_logical_phis = false;
Expand Down

0 comments on commit f918de3

Please sign in to comment.