gallium: split transfer_inline_write into buffer and texture callbacks

to reduce the call indirections with u_resource_vtbl.

The worst call tree you could get was:
  - u_transfer_inline_write_vtbl
    - u_default_transfer_inline_write
      - u_transfer_map_vtbl
        - driver_transfer_map
      - u_transfer_unmap_vtbl
        - driver_transfer_unmap

That's 6 indirect calls. Some drivers only had 5. The goal is to have
1 indirect call for drivers that care. The resource type can be determined
statically at most call sites.

The new interface is:
  pipe_context::buffer_subdata(ctx, resource, usage, offset, size, data)
  pipe_context::texture_subdata(ctx, resource, level, usage, box, data,
                                stride, layer_stride)

v2: fix whitespace, correct ilo's behavior

Reviewed-by: Nicolai Hähnle <[email protected]>
Acked-by: Roland Scheidegger <[email protected]>

10 files changed, 48 insertions, 45 deletions

diff --git a/src/gallium/state_trackers/clover/core/resource.cpp b/src/gallium/state_trackers/clover/core/resource.cpp
index 10a29a94eac..83781d39c01 100644
--- a/src/gallium/state_trackers/clover/core/resource.cpp
+++ b/src/gallium/state_trackers/clover/core/resource.cpp
@@ -161,9 +161,13 @@ root_resource::root_resource(clover::device &dev, memory_obj &obj,
       box rect { {{ 0, 0, 0 }}, {{ info.width0, info.height0, info.depth0 }} };
       unsigned cpp = util_format_get_blocksize(info.format);
 
-      q.pipe->transfer_inline_write(q.pipe, pipe, 0, PIPE_TRANSFER_WRITE,
-                                    rect, data_ptr, cpp * info.width0,
-                                    cpp * info.width0 * info.height0);
+      if (pipe->target == PIPE_BUFFER)
+         q.pipe->buffer_subdata(q.pipe, pipe, PIPE_TRANSFER_WRITE,
+                                0, info.width0, data_ptr);
+      else
+         q.pipe->texture_subdata(q.pipe, pipe, 0, PIPE_TRANSFER_WRITE,
+                                 rect, data_ptr, cpp * info.width0,
+                                 cpp * info.width0 * info.height0);
    }
 }
 
diff --git a/src/gallium/state_trackers/nine/buffer9.h b/src/gallium/state_trackers/nine/buffer9.h
index 8bdb4326a4c..c109cf66140 100644
--- a/src/gallium/state_trackers/nine/buffer9.h
+++ b/src/gallium/state_trackers/nine/buffer9.h
@@ -88,10 +88,10 @@ NineBuffer9_Upload( struct NineBuffer9 *This )
     struct pipe_context *pipe = This->pipe;
 
     assert(This->base.pool == D3DPOOL_MANAGED && This->managed.dirty);
-    pipe->transfer_inline_write(pipe, This->base.resource, 0, 0,
-                                &This->managed.dirty_box,
-                                (char *)This->managed.data + This->managed.dirty_box.x,
-                                This->size, This->size);
+    pipe->buffer_subdata(pipe, This->base.resource, 0,
+                         This->managed.dirty_box.x,
+                         This->managed.dirty_box.width,
+                         (char *)This->managed.data + This->managed.dirty_box.x);
     This->managed.dirty = FALSE;
 }
 
diff --git a/src/gallium/state_trackers/nine/nine_state.c b/src/gallium/state_trackers/nine/nine_state.c
index 6aa63250112..fd098ca68ea 100644
--- a/src/gallium/state_trackers/nine/nine_state.c
+++ b/src/gallium/state_trackers/nine/nine_state.c
@@ -86,8 +86,7 @@ prepare_ps_constants_userbuf(struct NineDevice9 *device);
         DBG("upload ConstantF [%u .. %u]\n", x, (x) + (c) - 1); \
         box.x = (p) * 4 * sizeof(float); \
         box.width = (c) * 4 * sizeof(float); \
-        pipe->transfer_inline_write(pipe, buf, 0, usage, &box, &((d)[p * 4]), \
-                                    0, 0); \
+        pipe->buffer_subdata(pipe, buf, usage, box.x, box.width, &((d)[p * 4])); \
     } while(0)
 
 /* OK, this is a bit ugly ... */
@@ -186,7 +185,7 @@ upload_constants(struct NineDevice9 *device, unsigned shader_type)
        box.x = x;
        box.width = c * 4;
        DBG("upload ConstantB [%u .. %u]\n", x, x + c - 1);
-       pipe->transfer_inline_write(pipe, buf, 0, usage, &box, data_b, 0, 0);
+       pipe->buffer_subdata(pipe, buf, usage, box.x, box.width, data_b);
     }
 
     /* int4 */
@@ -203,7 +202,7 @@ upload_constants(struct NineDevice9 *device, unsigned shader_type)
             box.x += x * 4 * sizeof(int);
             box.width = c * 4 * sizeof(int);
             c = 0;
-            pipe->transfer_inline_write(pipe, buf, 0, usage, &box, data, 0, 0);
+            pipe->buffer_subdata(pipe, buf, usage, box.x, box.width, data);
         }
     }
     if (c) {
@@ -212,7 +211,7 @@ upload_constants(struct NineDevice9 *device, unsigned shader_type)
         box.x  = buf->width0 - (NINE_MAX_CONST_I * 4 + NINE_MAX_CONST_B) * 4;
         box.x += x * 4 * sizeof(int);
         box.width = c * 4 * sizeof(int);
-        pipe->transfer_inline_write(pipe, buf, 0, usage, &box, data, 0, 0);
+        pipe->buffer_subdata(pipe, buf, usage, box.x, box.width, data);
     }
 
     /* TODO: only upload these when shader itself changes */
@@ -224,7 +223,7 @@ upload_constants(struct NineDevice9 *device, unsigned shader_type)
             n += r->end - r->bgn;
             box.width = (r->end - r->bgn) * 4 * sizeof(float);
             data = &lconstf_data[4 * n];
-            pipe->transfer_inline_write(pipe, buf, 0, usage, &box, data, 0, 0);
+            pipe->buffer_subdata(pipe, buf, usage, box.x, box.width, data);
             r = r->next;
         }
     }
diff --git a/src/gallium/state_trackers/nine/surface9.c b/src/gallium/state_trackers/nine/surface9.c
index 2606dbfcfa8..6a4a0d9dc74 100644
--- a/src/gallium/state_trackers/nine/surface9.c
+++ b/src/gallium/state_trackers/nine/surface9.c
@@ -673,8 +673,8 @@ NineSurface9_UploadSelf( struct NineSurface9 *This,
 
     ptr = NineSurface9_GetSystemMemPointer(This, box.x, box.y);
 
-    pipe->transfer_inline_write(pipe, res, This->level, 0,
-                                &box, ptr, This->stride, 0);
+    pipe->texture_subdata(pipe, res, This->level, 0,
+                          &box, ptr, This->stride, 0);
 
     return D3D_OK;
 }
diff --git a/src/gallium/state_trackers/nine/volume9.c b/src/gallium/state_trackers/nine/volume9.c
index 1fdc6388197..0302b099931 100644
--- a/src/gallium/state_trackers/nine/volume9.c
+++ b/src/gallium/state_trackers/nine/volume9.c
@@ -496,8 +496,8 @@ NineVolume9_UploadSelf( struct NineVolume9 *This,
 
     ptr = NineVolume9_GetSystemMemPointer(This, box.x, box.y, box.z);
 
-    pipe->transfer_inline_write(pipe, res, This->level, 0, &box,
-                                ptr, This->stride, This->layer_stride);
+    pipe->texture_subdata(pipe, res, This->level, 0, &box,
+                          ptr, This->stride, This->layer_stride);
 
     return D3D_OK;
 }
diff --git a/src/gallium/state_trackers/omx/vid_enc.c b/src/gallium/state_trackers/omx/vid_enc.c
index 642238e53f0..0d7ab28fb37 100644
--- a/src/gallium/state_trackers/omx/vid_enc.c
+++ b/src/gallium/state_trackers/omx/vid_enc.c
@@ -905,16 +905,16 @@ static OMX_ERRORTYPE enc_LoadImage(omx_base_PortType *port, OMX_BUFFERHEADERTYPE
       box.width = def->nFrameWidth;
       box.height = def->nFrameHeight;
       box.depth = 1;
-      priv->s_pipe->transfer_inline_write(priv->s_pipe, views[0]->texture, 0,
-                                          PIPE_TRANSFER_WRITE, &box,
-                                          ptr, def->nStride, 0);
+      priv->s_pipe->texture_subdata(priv->s_pipe, views[0]->texture, 0,
+                                    PIPE_TRANSFER_WRITE, &box,
+                                    ptr, def->nStride, 0);
       ptr = ((uint8_t*)buf->pBuffer) + (def->nStride * box.height);
       box.width = def->nFrameWidth / 2;
       box.height = def->nFrameHeight / 2;
       box.depth = 1;
-      priv->s_pipe->transfer_inline_write(priv->s_pipe, views[1]->texture, 0,
-                                          PIPE_TRANSFER_WRITE, &box,
-                                          ptr, def->nStride, 0);
+      priv->s_pipe->texture_subdata(priv->s_pipe, views[1]->texture, 0,
+                                    PIPE_TRANSFER_WRITE, &box,
+                                    ptr, def->nStride, 0);
    } else {
       struct pipe_blit_info blit;
       struct vl_video_buffer *dst_buf = (struct vl_video_buffer *)vbuf;
diff --git a/src/gallium/state_trackers/va/image.c b/src/gallium/state_trackers/va/image.c
index 1b956e36bc1..36b24695edf 100644
--- a/src/gallium/state_trackers/va/image.c
+++ b/src/gallium/state_trackers/va/image.c
@@ -515,10 +515,10 @@ vlVaPutImage(VADriverContextP ctx, VASurfaceID surface, VAImageID image,
       vlVaVideoSurfaceSize(surf, i, &width, &height);
       for (j = 0; j < views[i]->texture->array_size; ++j) {
          struct pipe_box dst_box = {0, 0, j, width, height, 1};
-         drv->pipe->transfer_inline_write(drv->pipe, views[i]->texture, 0,
-            PIPE_TRANSFER_WRITE, &dst_box,
-            data[i] + pitches[i] * j,
-            pitches[i] * views[i]->texture->array_size, 0);
+         drv->pipe->texture_subdata(drv->pipe, views[i]->texture, 0,
+                                    PIPE_TRANSFER_WRITE, &dst_box,
+                                    data[i] + pitches[i] * j,
+                                    pitches[i] * views[i]->texture->array_size, 0);
       }
    }
    pipe_mutex_unlock(drv->mutex);
diff --git a/src/gallium/state_trackers/vdpau/bitmap.c b/src/gallium/state_trackers/vdpau/bitmap.c
index 35c8820433d..fd67a986c61 100644
--- a/src/gallium/state_trackers/vdpau/bitmap.c
+++ b/src/gallium/state_trackers/vdpau/bitmap.c
@@ -201,9 +201,9 @@ vlVdpBitmapSurfacePutBitsNative(VdpBitmapSurface surface,
    vlVdpResolveDelayedRendering(vlsurface->device, NULL, NULL);
 
    dst_box = RectToPipeBox(destination_rect, vlsurface->sampler_view->texture);
-   pipe->transfer_inline_write(pipe, vlsurface->sampler_view->texture, 0,
-                               PIPE_TRANSFER_WRITE, &dst_box, *source_data,
-                               *source_pitches, 0);
+   pipe->texture_subdata(pipe, vlsurface->sampler_view->texture, 0,
+                         PIPE_TRANSFER_WRITE, &dst_box, *source_data,
+                         *source_pitches, 0);
 
    pipe_mutex_unlock(vlsurface->device->mutex);
 
diff --git a/src/gallium/state_trackers/vdpau/output.c b/src/gallium/state_trackers/vdpau/output.c
index 8a064e849b6..0b4f081be3e 100644
--- a/src/gallium/state_trackers/vdpau/output.c
+++ b/src/gallium/state_trackers/vdpau/output.c
@@ -257,9 +257,9 @@ vlVdpOutputSurfacePutBitsNative(VdpOutputSurface surface,
    vlVdpResolveDelayedRendering(vlsurface->device, NULL, NULL);
 
    dst_box = RectToPipeBox(destination_rect, vlsurface->sampler_view->texture);
-   pipe->transfer_inline_write(pipe, vlsurface->sampler_view->texture, 0,
-                               PIPE_TRANSFER_WRITE, &dst_box, *source_data,
-                               *source_pitches, 0);
+   pipe->texture_subdata(pipe, vlsurface->sampler_view->texture, 0,
+                         PIPE_TRANSFER_WRITE, &dst_box, *source_data,
+                         *source_pitches, 0);
    pipe_mutex_unlock(vlsurface->device->mutex);
 
    return VDP_STATUS_OK;
@@ -346,9 +346,9 @@ vlVdpOutputSurfacePutBitsIndexed(VdpOutputSurface surface,
    box.height = res->height0;
    box.depth = res->depth0;
 
-   context->transfer_inline_write(context, res, 0, PIPE_TRANSFER_WRITE, &box,
-                                  source_data[0], source_pitch[0],
-                                  source_pitch[0] * res->height0);
+   context->texture_subdata(context, res, 0, PIPE_TRANSFER_WRITE, &box,
+                            source_data[0], source_pitch[0],
+                            source_pitch[0] * res->height0);
 
    memset(&sv_tmpl, 0, sizeof(sv_tmpl));
    u_sampler_view_default_template(&sv_tmpl, res, res->format);
@@ -379,8 +379,8 @@ vlVdpOutputSurfacePutBitsIndexed(VdpOutputSurface surface,
    box.height = res->height0;
    box.depth = res->depth0;
 
-   context->transfer_inline_write(context, res, 0, PIPE_TRANSFER_WRITE, &box, color_table,
-                                  util_format_get_stride(colortbl_format, res->width0), 0);
+   context->texture_subdata(context, res, 0, PIPE_TRANSFER_WRITE, &box, color_table,
+                            util_format_get_stride(colortbl_format, res->width0), 0);
 
    memset(&sv_tmpl, 0, sizeof(sv_tmpl));
    u_sampler_view_default_template(&sv_tmpl, res, res->format);
@@ -485,8 +485,8 @@ vlVdpOutputSurfacePutBitsYCbCr(VdpOutputSurface surface,
          sv->texture->width0, sv->texture->height0, 1
       };
 
-      pipe->transfer_inline_write(pipe, sv->texture, 0, PIPE_TRANSFER_WRITE, &dst_box,
-                                  source_data[i], source_pitches[i], 0);
+      pipe->texture_subdata(pipe, sv->texture, 0, PIPE_TRANSFER_WRITE, &dst_box,
+                            source_data[i], source_pitches[i], 0);
    }
 
    if (!csc_matrix) {
diff --git a/src/gallium/state_trackers/vdpau/surface.c b/src/gallium/state_trackers/vdpau/surface.c
index 7998527b2d1..6dc479a6726 100644
--- a/src/gallium/state_trackers/vdpau/surface.c
+++ b/src/gallium/state_trackers/vdpau/surface.c
@@ -359,11 +359,11 @@ vlVdpVideoSurfacePutBitsYCbCr(VdpVideoSurface surface,
             width, height, 1
          };
 
-         pipe->transfer_inline_write(pipe, sv->texture, 0,
-                                     PIPE_TRANSFER_WRITE, &dst_box,
-                                     source_data[i] + source_pitches[i] * j,
-                                     source_pitches[i] * sv->texture->array_size,
-                                     0);
+         pipe->texture_subdata(pipe, sv->texture, 0,
+                               PIPE_TRANSFER_WRITE, &dst_box,
+                               source_data[i] + source_pitches[i] * j,
+                               source_pitches[i] * sv->texture->array_size,
+                               0);
       }
    }
    pipe_mutex_unlock(p_surf->device->mutex);