gallium/radeon: add a new HUD query for the number of mapped buffers

Useful when debugging applications which map a ton of buffers
and also because we used to run into Linux's limit on the number
of simultaneous mmap() calls.

v2: - update the commit message

Signed-off-by: Samuel Pitoiset <[email protected]>
Reviewed-by: Nicolai Hähnle <[email protected]>

6 files changed, 12 insertions, 0 deletions

diff --git a/src/gallium/winsys/amdgpu/drm/amdgpu_bo.c b/src/gallium/winsys/amdgpu/drm/amdgpu_bo.c
index e8d2c006f36..5ee27b8ede0 100644
--- a/src/gallium/winsys/amdgpu/drm/amdgpu_bo.c
+++ b/src/gallium/winsys/amdgpu/drm/amdgpu_bo.c
@@ -181,6 +181,7 @@ void amdgpu_bo_destroy(struct pb_buffer *_buf)
          bo->ws->mapped_vram -= bo->base.size;
       else if (bo->initial_domain & RADEON_DOMAIN_GTT)
          bo->ws->mapped_gtt -= bo->base.size;
+      bo->ws->num_mapped_buffers--;
    }
 
    FREE(bo);
@@ -308,6 +309,7 @@ static void *amdgpu_bo_map(struct pb_buffer *buf,
          real->ws->mapped_vram += real->base.size;
       else if (real->initial_domain & RADEON_DOMAIN_GTT)
          real->ws->mapped_gtt += real->base.size;
+      real->ws->num_mapped_buffers++;
    }
    return (uint8_t*)cpu + offset;
 }
@@ -327,6 +329,7 @@ static void amdgpu_bo_unmap(struct pb_buffer *buf)
          real->ws->mapped_vram -= real->base.size;
       else if (real->initial_domain & RADEON_DOMAIN_GTT)
          real->ws->mapped_gtt -= real->base.size;
+      real->ws->num_mapped_buffers--;
    }
 
    amdgpu_bo_cpu_unmap(real->bo);
diff --git a/src/gallium/winsys/amdgpu/drm/amdgpu_winsys.c b/src/gallium/winsys/amdgpu/drm/amdgpu_winsys.c
index e944e62f0aa..501f6f980db 100644
--- a/src/gallium/winsys/amdgpu/drm/amdgpu_winsys.c
+++ b/src/gallium/winsys/amdgpu/drm/amdgpu_winsys.c
@@ -424,6 +424,8 @@ static uint64_t amdgpu_query_value(struct radeon_winsys *rws,
       return ws->mapped_gtt;
    case RADEON_BUFFER_WAIT_TIME_NS:
       return ws->buffer_wait_time;
+   case RADEON_NUM_MAPPED_BUFFERS:
+      return ws->num_mapped_buffers;
    case RADEON_TIMESTAMP:
       amdgpu_query_info(ws->dev, AMDGPU_INFO_TIMESTAMP, 8, &retval);
       return retval;
diff --git a/src/gallium/winsys/amdgpu/drm/amdgpu_winsys.h b/src/gallium/winsys/amdgpu/drm/amdgpu_winsys.h
index c56c342a53a..b19d9767dad 100644
--- a/src/gallium/winsys/amdgpu/drm/amdgpu_winsys.h
+++ b/src/gallium/winsys/amdgpu/drm/amdgpu_winsys.h
@@ -64,6 +64,7 @@ struct amdgpu_winsys {
    uint64_t buffer_wait_time; /* time spent in buffer_wait in ns */
    uint64_t num_gfx_IBs;
    uint64_t num_sdma_IBs;
+   uint64_t num_mapped_buffers;
 
    struct radeon_info info;
 
diff --git a/src/gallium/winsys/radeon/drm/radeon_drm_bo.c b/src/gallium/winsys/radeon/drm/radeon_drm_bo.c
index a15d559b015..9872861525c 100644
--- a/src/gallium/winsys/radeon/drm/radeon_drm_bo.c
+++ b/src/gallium/winsys/radeon/drm/radeon_drm_bo.c
@@ -382,6 +382,7 @@ void radeon_bo_destroy(struct pb_buffer *_buf)
             bo->rws->mapped_vram -= bo->base.size;
         else
             bo->rws->mapped_gtt -= bo->base.size;
+        bo->rws->num_mapped_buffers--;
     }
 
     FREE(bo);
@@ -458,6 +459,7 @@ void *radeon_bo_do_map(struct radeon_bo *bo)
        bo->rws->mapped_vram += bo->base.size;
     else
        bo->rws->mapped_gtt += bo->base.size;
+    bo->rws->num_mapped_buffers++;
 
     pipe_mutex_unlock(bo->u.real.map_mutex);
     return (uint8_t*)bo->u.real.ptr + offset;
@@ -570,6 +572,7 @@ static void radeon_bo_unmap(struct pb_buffer *_buf)
        bo->rws->mapped_vram -= bo->base.size;
     else
        bo->rws->mapped_gtt -= bo->base.size;
+    bo->rws->num_mapped_buffers--;
 
     pipe_mutex_unlock(bo->u.real.map_mutex);
 }
diff --git a/src/gallium/winsys/radeon/drm/radeon_drm_winsys.c b/src/gallium/winsys/radeon/drm/radeon_drm_winsys.c
index c85e427a216..f5de66e8bdd 100644
--- a/src/gallium/winsys/radeon/drm/radeon_drm_winsys.c
+++ b/src/gallium/winsys/radeon/drm/radeon_drm_winsys.c
@@ -614,6 +614,8 @@ static uint64_t radeon_query_value(struct radeon_winsys *rws,
        return ws->mapped_gtt;
     case RADEON_BUFFER_WAIT_TIME_NS:
         return ws->buffer_wait_time;
+    case RADEON_NUM_MAPPED_BUFFERS:
+        return ws->num_mapped_buffers;
     case RADEON_TIMESTAMP:
         if (ws->info.drm_minor < 20 || ws->gen < DRV_R600) {
             assert(0);
diff --git a/src/gallium/winsys/radeon/drm/radeon_drm_winsys.h b/src/gallium/winsys/radeon/drm/radeon_drm_winsys.h
index ac623a1b311..7f2d99d372c 100644
--- a/src/gallium/winsys/radeon/drm/radeon_drm_winsys.h
+++ b/src/gallium/winsys/radeon/drm/radeon_drm_winsys.h
@@ -81,6 +81,7 @@ struct radeon_drm_winsys {
     uint64_t buffer_wait_time; /* time spent in buffer_wait in ns */
     uint64_t num_gfx_IBs;
     uint64_t num_sdma_IBs;
+    uint64_t num_mapped_buffers;
     uint32_t next_bo_hash;
 
     enum radeon_generation gen;