broadcom/vc4: Use a single-entry cached last_hindex value.

Since almost all BOs will be in one CL at a time, this cache will almost
always hit except for the first usage of the BO in each CL.

This didn't show up as statistically significant on the minetest trace
(n=340), but if I lop off the throttled lobe of the bimodal distribution,
it very clearly does (0.74731% +/- 0.162093%, n=269).

1 files changed, 12 insertions, 2 deletions

diff --git a/src/gallium/drivers/vc4/vc4_cl.c b/src/gallium/drivers/vc4/vc4_cl.c
index 508281a27bb..7ae092ebce3 100644
--- a/src/gallium/drivers/vc4/vc4_cl.c
+++ b/src/gallium/drivers/vc4/vc4_cl.c
@@ -61,10 +61,19 @@ vc4_gem_hindex(struct vc4_job *job, struct vc4_bo *bo)
 {
         uint32_t hindex;
         uint32_t *current_handles = job->bo_handles.base;
+        uint32_t cl_hindex_count = cl_offset(&job->bo_handles) / 4;
+        uint32_t last_hindex = bo->last_hindex; /* volatile read! */
 
-        for (hindex = 0; hindex < cl_offset(&job->bo_handles) / 4; hindex++) {
-                if (current_handles[hindex] == bo->handle)
+        if (last_hindex < cl_hindex_count &&
+            current_handles[last_hindex] == bo->handle) {
+                return last_hindex;
+        }
+
+        for (hindex = 0; hindex < cl_hindex_count; hindex++) {
+                if (current_handles[hindex] == bo->handle) {
+                        bo->last_hindex = hindex;
                         return hindex;
+                }
         }
 
         struct vc4_cl_out *out;
@@ -79,5 +88,6 @@ vc4_gem_hindex(struct vc4_job *job, struct vc4_bo *bo)
 
         job->bo_space += bo->size;
 
+        bo->last_hindex = hindex;
         return hindex;
 }