broadcom/compiler: don't assign registers to unused nodes/temps

In programs with a lot of unused temps, if we don't do this, we may
end up recycling previously used rfs more often, which can be
detrimental to instruction pairing.

total instructions in shared programs: 11464335 -> 11444136 (-0.18%)
instructions in affected programs: 8976743 -> 8956544 (-0.23%)
helped: 33196
HURT: 33778
Inconclusive result

total max-temps in shared programs: 2230150 -> 2229445 (-0.03%)
max-temps in affected programs: 86413 -> 85708 (-0.82%)
helped: 2217
HURT: 1523
Max-temps are helped.

total sfu-stalls in shared programs: 18077 -> 17104 (-5.38%)
sfu-stalls in affected programs: 8669 -> 7696 (-11.22%)
helped: 2657
HURT: 2182
Sfu-stalls are helped.

total inst-and-stalls in shared programs: 11482412 -> 11461240 (-0.18%)
inst-and-stalls in affected programs: 8995697 -> 8974525 (-0.24%)
helped: 33319
HURT: 33708
Inconclusive result

total nops in shared programs: 298140 -> 296185 (-0.66%)
nops in affected programs: 52805 -> 50850 (-3.70%)
helped: 3797
HURT: 2662
Inconclusive result

Reviewed-by: Alejandro Piñeiro <apinheiro@igalia.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/25450>

diff --git a/src/broadcom/compiler/v3d_compiler.h b/src/broadcom/compiler/v3d_compiler.h
index c62f65c..095b33c 100644 (file)
--- a/src/broadcom/compiler/v3d_compiler.h
+++ b/src/broadcom/compiler/v3d_compiler.h
@@ -606,6 +606,7 @@ struct v3d_ra_node_info {
                 uint32_t priority;
                 uint8_t class_bits;
                 bool is_program_end;
+                bool unused;
 
                 /* V3D 7.x */
                 bool is_ldunif_dst;

diff --git a/src/broadcom/compiler/vir_register_allocate.c b/src/broadcom/compiler/vir_register_allocate.c
index a79fb39..a3780a9 100644 (file)
--- a/src/broadcom/compiler/vir_register_allocate.c
+++ b/src/broadcom/compiler/vir_register_allocate.c
@@ -386,6 +386,7 @@ add_node(struct v3d_compile *c, uint32_t temp, uint8_t class_bits)
         c->nodes.info[node].priority = 0;
         c->nodes.info[node].is_ldunif_dst = false;
         c->nodes.info[node].is_program_end = false;
+        c->nodes.info[node].unused = false;
 }
 
 /* The spill offset for this thread takes a bit of setup, so do it once at
@@ -918,6 +919,12 @@ v3d_ra_select_rf(struct v3d_ra_select_callback_data *v3d_ra,
                  BITSET_WORD *regs,
                  unsigned int *out)
 {
+        /* If this node is for an unused temp, ignore. */
+        if (v3d_ra->nodes->info[node].unused) {
+                *out = 0;
+                return true;
+        }
+
         /* In V3D 7.x, try to assign rf0 to temps used as ldunif's dst
          * so we can avoid turning them into ldunifrf (which uses the
          * cond field to encode the dst and would prevent merge with
@@ -1331,6 +1338,7 @@ v3d_register_allocate(struct v3d_compile *c)
         for (uint32_t i = 0; i < num_ra_nodes; i++) {
                 c->nodes.info[i].is_ldunif_dst = false;
                 c->nodes.info[i].is_program_end = false;
+                c->nodes.info[i].unused = false;
                 c->nodes.info[i].priority = 0;
                 c->nodes.info[i].class_bits = 0;
                 if (c->devinfo->has_accumulators && i < ACC_COUNT) {
@@ -1396,6 +1404,12 @@ v3d_register_allocate(struct v3d_compile *c)
 
         /* Add register interferences based on liveness data */
         for (uint32_t i = 0; i < c->num_temps; i++) {
+                /* And while we are here, let's also flag nodes for
+                 * unused temps.
+                 */
+                if (c->temp_start[i] > c->temp_end[i])
+                        c->nodes.info[temp_to_node(c, i)].unused = true;
+
                 for (uint32_t j = i + 1; j < c->num_temps; j++) {
                         if (interferes(c->temp_start[i], c->temp_end[i],
                                        c->temp_start[j], c->temp_end[j])) {