From 3817fa7a4d1f51c385b28a2e45a1edf227526028 Mon Sep 17 00:00:00 2001 From: =?utf8?q?Daniel=20Sch=C3=BCrmann?= Date: Wed, 17 Jun 2020 12:41:09 +0100 Subject: [PATCH] aco: fix WQM handling in nested loops If on a nested loop - the outer loop needs WQM but - the inner loop doesn't need WQM and - the break condition of the inner loop is computed in the outer loop then it could happen that we transitioned to Exact before entering the inner loop which could create an empty exec mask and lead to an infinite loop. Fixes a GPU hang with RDR2 Cc: 20.1 Reviewed-by: Rhys Perry Part-of: --- src/amd/compiler/aco_insert_exec_mask.cpp | 5 ++++- 1 file changed, 4 insertions(+), 1 deletion(-) diff --git a/src/amd/compiler/aco_insert_exec_mask.cpp b/src/amd/compiler/aco_insert_exec_mask.cpp index fe0dc54ccca..7246fb74e0c 100644 --- a/src/amd/compiler/aco_insert_exec_mask.cpp +++ b/src/amd/compiler/aco_insert_exec_mask.cpp @@ -189,11 +189,14 @@ void get_block_needs(wqm_ctx &ctx, exec_ctx &exec_ctx, Block* block) if (block->kind & block_kind_top_level) { if (ctx.loop && ctx.wqm) { - /* mark all break conditions as WQM */ unsigned block_idx = block->index + 1; while (!(ctx.program->blocks[block_idx].kind & block_kind_top_level)) { + /* flag all break conditions as WQM: + * the conditions might be computed outside the nested CF */ if (ctx.program->blocks[block_idx].kind & block_kind_break) mark_block_wqm(ctx, block_idx); + /* flag all blocks as WQM to ensure we enter all (nested) loops in WQM */ + exec_ctx.info[block_idx].block_needs |= WQM; block_idx++; } } else if (ctx.loop && !ctx.wqm) { -- 2.30.2