vc4: Enforce one-uniform-per-instruction after optimization.

This lets us more intelligently decide which uniform values should be put
into temporaries, by choosing the most reused values to push to temps
first.

total uniforms in shared programs: 13457 -> 13433 (-0.18%)
uniforms in affected programs:     1524 -> 1500 (-1.57%)
total instructions in shared programs: 40198 -> 40019 (-0.45%)
instructions in affected programs:     6027 -> 5848 (-2.97%)

I noticed this opportunity because with the NIR work, some programs were
happening to make different uniform copy propagation choices that
significantly increased instruction counts.

1 files changed, 12 insertions, 3 deletions

diff --git a/src/gallium/drivers/vc4/vc4_qir.c b/src/gallium/drivers/vc4/vc4_qir.c
index 9e0ee1f0ae5..5c1fdbddfb6 100644
--- a/src/gallium/drivers/vc4/vc4_qir.c
+++ b/src/gallium/drivers/vc4/vc4_qir.c
@@ -174,6 +174,12 @@ qir_is_multi_instruction(struct qinst *inst)
 }
 
 bool
+qir_is_tex(struct qinst *inst)
+{
+        return inst->op >= QOP_TEX_S && inst->op <= QOP_TEX_DIRECT;
+}
+
+bool
 qir_depends_on_flags(struct qinst *inst)
 {
         switch (inst->op) {
@@ -420,9 +426,12 @@ qir_get_stage_name(enum qstage stage)
 void
 qir_SF(struct vc4_compile *c, struct qreg src)
 {
-        assert(!is_empty_list(&c->instructions));
-        struct qinst *last_inst = (struct qinst *)c->instructions.prev;
-        if (last_inst->dst.file != src.file ||
+        struct qinst *last_inst = NULL;
+        if (!is_empty_list(&c->instructions))
+                last_inst = (struct qinst *)c->instructions.prev;
+
+        if (!last_inst ||
+            last_inst->dst.file != src.file ||
             last_inst->dst.index != src.index ||
             qir_is_multi_instruction(last_inst)) {
                 src = qir_MOV(c, src);