freedreno/ir3: eliminate unused false-deps

Previously false-dependencies would get flagged as used, even if the
only "use" was a false dep to (for example) prevent a load from being
scheduled after a store.

In addition to being pointless instructions, in some cases they can
cause problems.  For example, ldg (and similar instructions) depend on
an immed arg getting CP'd into the instruction, but this doesn't happen
if an instruction is otherwise unused.  Which can result in undefined
results (overwriting unintended registers).

Signed-off-by: Rob Clark <robdclark@gmail.com>

diff --git a/src/gallium/drivers/freedreno/ir3/ir3_cp.c b/src/gallium/drivers/freedreno/ir3/ir3_cp.c
index 391e94ca44f707c8ec5fd52b50fef41fb2119b84..67a7714b9c5078198770c354a5556e1608b0ccd5 100644 (file)
--- a/src/gallium/drivers/freedreno/ir3/ir3_cp.c
+++ b/src/gallium/drivers/freedreno/ir3/ir3_cp.c
@@ -607,7 +607,7 @@ ir3_cp(struct ir3 *ir, struct ir3_shader_variant *so)
                list_for_each_entry (struct ir3_instruction, instr, &block->instr_list, node) {
                        struct ir3_instruction *src;
 
-                       /* by the way, we don't acount for false-dep's, so the CP
+                       /* by the way, we don't account for false-dep's, so the CP
                         * pass should always happen before false-dep's are inserted
                         */
                        debug_assert(instr->deps_count == 0);

diff --git a/src/gallium/drivers/freedreno/ir3/ir3_depth.c b/src/gallium/drivers/freedreno/ir3/ir3_depth.c
index 270d53dbed7b4fa5122cc6f61a2a6b3026693034..3ece13928f424130cbcd8de2ad033e883118980b 100644 (file)
--- a/src/gallium/drivers/freedreno/ir3/ir3_depth.c
+++ b/src/gallium/drivers/freedreno/ir3/ir3_depth.c
@@ -133,13 +133,18 @@ ir3_insert_by_depth(struct ir3_instruction *instr, struct list_head *list)
 }
 
 static void
-ir3_instr_depth(struct ir3_instruction *instr, unsigned boost)
+ir3_instr_depth(struct ir3_instruction *instr, unsigned boost, bool falsedep)
 {
        struct ir3_instruction *src;
 
        /* if we've already visited this instruction, bail now: */
-       if (ir3_instr_check_mark(instr))
+       if (falsedep) {
+               /* don't mark falsedep's as used, but process them normally: */
+               if (instr->flags & IR3_INSTR_MARK)
+                       return;
+       } else if (ir3_instr_check_mark(instr)) {
                return;
+       }
 
        instr->depth = 0;
 
@@ -147,7 +152,7 @@ ir3_instr_depth(struct ir3_instruction *instr, unsigned boost)
                unsigned sd;
 
                /* visit child to compute it's depth: */
-               ir3_instr_depth(src, boost);
+               ir3_instr_depth(src, boost, __is_false_dep(instr, i));
 
                /* for array writes, no need to delay on previous write: */
                if (i == 0)
@@ -165,9 +170,10 @@ ir3_instr_depth(struct ir3_instruction *instr, unsigned boost)
        ir3_insert_by_depth(instr, &instr->block->instr_list);
 }
 
-static void
+static bool
 remove_unused_by_block(struct ir3_block *block)
 {
+       bool progress = false;
        list_for_each_entry_safe (struct ir3_instruction, instr, &block->instr_list, node) {
                if (!ir3_instr_check_mark(instr)) {
                        if (instr->opc == OPC_END)
@@ -178,32 +184,35 @@ remove_unused_by_block(struct ir3_block *block)
                        instr->flags |= IR3_INSTR_UNUSED;
                        /* and remove from instruction list: */
                        list_delinit(&instr->node);
+                       progress = true;
                }
        }
+       return progress;
 }
 
-void
-ir3_depth(struct ir3 *ir)
+static bool
+compute_depth_and_remove_unused(struct ir3 *ir)
 {
        unsigned i;
+       bool progress = false;
 
        ir3_clear_mark(ir);
        for (i = 0; i < ir->noutputs; i++)
                if (ir->outputs[i])
-                       ir3_instr_depth(ir->outputs[i], 0);
+                       ir3_instr_depth(ir->outputs[i], 0, false);
 
        list_for_each_entry (struct ir3_block, block, &ir->block_list, node) {
                for (i = 0; i < block->keeps_count; i++)
-                       ir3_instr_depth(block->keeps[i], 0);
+                       ir3_instr_depth(block->keeps[i], 0, false);
 
                /* We also need to account for if-condition: */
                if (block->condition)
-                       ir3_instr_depth(block->condition, 6);
+                       ir3_instr_depth(block->condition, 6, false);
        }
 
        /* mark un-used instructions: */
        list_for_each_entry (struct ir3_block, block, &ir->block_list, node) {
-               remove_unused_by_block(block);
+               progress |= remove_unused_by_block(block);
        }
 
        /* note that we can end up with unused indirects, but we should
@@ -221,4 +230,15 @@ ir3_depth(struct ir3 *ir)
                if (in && (in->flags & IR3_INSTR_UNUSED))
                        ir->inputs[i] = NULL;
        }
+
+       return progress;
+}
+
+void
+ir3_depth(struct ir3 *ir)
+{
+       bool progress;
+       do {
+               progress = compute_depth_and_remove_unused(ir);
+       } while (progress);
 }