radv: pre-calc vertex buffer descriptor size on pipeline object

util_bitcount has a nonzero cost, and calling it like this in a hotpath
generates unnecessary overhead

Reviewed-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/11525>

src/amd/vulkan/radv_cmd_buffer.c

@@ -1053,7 +1053,7 @@ radv_emit_prefetch_L2(struct radv_cmd_buffer *cmd_buffer, struct radv_pipeline *
       radv_emit_shader_prefetch(cmd_buffer, pipeline->shaders[MESA_SHADER_VERTEX]);
 
    if (mask & RADV_PREFETCH_VBO_DESCRIPTORS)
-      si_cp_dma_prefetch(cmd_buffer, state->vb_va, state->vb_size);
+      si_cp_dma_prefetch(cmd_buffer, state->vb_va, pipeline->vb_desc_alloc_size);
 
    if (mask & RADV_PREFETCH_TCS)
       radv_emit_shader_prefetch(cmd_buffer, pipeline->shaders[MESA_SHADER_TESS_CTRL]);
@@ -2837,11 +2837,10 @@ radv_flush_vertex_descriptors(struct radv_cmd_buffer *cmd_buffer, bool pipeline_
       void *vb_ptr;
       unsigned desc_index = 0;
       uint32_t mask = pipeline->vb_desc_usage_mask;
-      uint32_t count = util_bitcount(mask);
       uint64_t va;
 
       /* allocate some descriptor state for vertex buffers */
-      if (!radv_cmd_buffer_upload_alloc(cmd_buffer, count * 16, &vb_offset, &vb_ptr))
+      if (!radv_cmd_buffer_upload_alloc(cmd_buffer, pipeline->vb_desc_alloc_size, &vb_offset, &vb_ptr))
          return;
 
       while (mask) {
@@ -2924,7 +2923,6 @@ radv_flush_vertex_descriptors(struct radv_cmd_buffer *cmd_buffer, bool pipeline_
                                  va);
 
       cmd_buffer->state.vb_va = va;
-      cmd_buffer->state.vb_size = count * 16;
       cmd_buffer->state.prefetch_L2_mask |= RADV_PREFETCH_VBO_DESCRIPTORS;
 
       if (unlikely(cmd_buffer->device->trace_bo))

src/amd/vulkan/radv_pipeline.c

@@ -5305,6 +5305,7 @@ radv_pipeline_init_vertex_input_state(struct radv_pipeline *pipeline,
 
    pipeline->use_per_attribute_vb_descs = info->vs.use_per_attribute_vb_descs;
    pipeline->vb_desc_usage_mask = info->vs.vb_desc_usage_mask;
+   pipeline->vb_desc_alloc_size = util_bitcount(pipeline->vb_desc_usage_mask) * 16;
 }
 
 static struct radv_shader_variant *

src/amd/vulkan/radv_private.h

@@ -1342,7 +1342,6 @@ enum rgp_flush_bits {
 struct radv_cmd_state {
    /* Vertex descriptors */
    uint64_t vb_va;
-   unsigned vb_size;
 
    bool predicating;
    uint64_t dirty;
@@ -1730,6 +1729,7 @@ struct radv_pipeline {
 
    bool use_per_attribute_vb_descs;
    uint32_t vb_desc_usage_mask;
+   uint32_t vb_desc_alloc_size;
 
    uint32_t user_data_0[MESA_SHADER_STAGES];
    union {