u_upload_mgr: pass alignment to u_upload_alloc manually

The fixed alignment of u_upload_mgr will go away.
This is the first step.

The motivation is that one u_upload_mgr can have multiple users,
each allocating from the same buffer, but requiring a different alignment.

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

21 files changed, 32 insertions, 23 deletions

diff --git a/src/gallium/auxiliary/hud/hud_context.c b/src/gallium/auxiliary/hud/hud_context.c
index c5c33327702..691de81f20a 100644
--- a/src/gallium/auxiliary/hud/hud_context.c
+++ b/src/gallium/auxiliary/hud/hud_context.c
@@ -431,7 +431,7 @@ hud_alloc_vertices(struct hud_context *hud, struct vertex_queue *v,
    v->max_num_vertices = num_vertices;
    v->vbuf.stride = stride;
    u_upload_alloc(hud->uploader, 0, v->vbuf.stride * v->max_num_vertices,
-                  &v->vbuf.buffer_offset, &v->vbuf.buffer,
+                  16, &v->vbuf.buffer_offset, &v->vbuf.buffer,
                   (void**)&v->vertices);
 }
 
diff --git a/src/gallium/auxiliary/indices/u_primconvert.c b/src/gallium/auxiliary/indices/u_primconvert.c
index 70d3e8530b8..c0a31548433 100644
--- a/src/gallium/auxiliary/indices/u_primconvert.c
+++ b/src/gallium/auxiliary/indices/u_primconvert.c
@@ -156,7 +156,7 @@ util_primconvert_draw_vbo(struct primconvert_context *pc,
       pc->upload = u_upload_create(pc->pipe, 4096, 4, PIPE_BIND_INDEX_BUFFER);
    }
 
-   u_upload_alloc(pc->upload, 0, new_ib.index_size * new_info.count,
+   u_upload_alloc(pc->upload, 0, new_ib.index_size * new_info.count, 4,
                   &new_ib.offset, &new_ib.buffer, &dst);
 
    if (info->indexed) {
diff --git a/src/gallium/auxiliary/util/u_upload_mgr.c b/src/gallium/auxiliary/util/u_upload_mgr.c
index 4148bae5ab5..3f790400e40 100644
--- a/src/gallium/auxiliary/util/u_upload_mgr.c
+++ b/src/gallium/auxiliary/util/u_upload_mgr.c
@@ -181,11 +181,11 @@ void
 u_upload_alloc(struct u_upload_mgr *upload,
                unsigned min_out_offset,
                unsigned size,
+               unsigned alignment,
                unsigned *out_offset,
                struct pipe_resource **outbuf,
                void **ptr)
 {
-   unsigned alignment = upload->alignment;
    unsigned buffer_size = upload->buffer ? upload->buffer->width0 : 0;
    unsigned offset;
 
@@ -249,7 +249,7 @@ void u_upload_data(struct u_upload_mgr *upload,
 {
    uint8_t *ptr;
 
-   u_upload_alloc(upload, min_out_offset, size,
+   u_upload_alloc(upload, min_out_offset, size, upload->alignment,
                   out_offset, outbuf,
                   (void**)&ptr);
    if (ptr)
diff --git a/src/gallium/auxiliary/util/u_upload_mgr.h b/src/gallium/auxiliary/util/u_upload_mgr.h
index 67c6daa4e7f..ad7135fc8da 100644
--- a/src/gallium/auxiliary/util/u_upload_mgr.h
+++ b/src/gallium/auxiliary/util/u_upload_mgr.h
@@ -74,6 +74,7 @@ void u_upload_unmap( struct u_upload_mgr *upload );
  * \param upload           Upload manager
  * \param min_out_offset   Minimum offset that should be returned in out_offset.
  * \param size             Size of the allocation.
+ * \param alignment        Alignment of the suballocation within the buffer
  * \param out_offset       Pointer to where the new buffer offset will be returned.
  * \param outbuf           Pointer to where the upload buffer will be returned.
  * \param ptr              Pointer to the allocated memory that is returned.
@@ -81,6 +82,7 @@ void u_upload_unmap( struct u_upload_mgr *upload );
 void u_upload_alloc(struct u_upload_mgr *upload,
                     unsigned min_out_offset,
                     unsigned size,
+                    unsigned alignment,
                     unsigned *out_offset,
                     struct pipe_resource **outbuf,
                     void **ptr);
diff --git a/src/gallium/auxiliary/util/u_vbuf.c b/src/gallium/auxiliary/util/u_vbuf.c
index 54e9e717104..dd64e2d7949 100644
--- a/src/gallium/auxiliary/util/u_vbuf.c
+++ b/src/gallium/auxiliary/util/u_vbuf.c
@@ -454,7 +454,7 @@ u_vbuf_translate_buffers(struct u_vbuf *mgr, struct translate_key *key,
 
       /* Create and map the output buffer. */
       u_upload_alloc(mgr->uploader, 0,
-                     key->output_stride * num_indices,
+                     key->output_stride * num_indices, 4,
                      &out_offset, &out_buffer,
                      (void**)&out_map);
       if (!out_buffer)
@@ -487,7 +487,7 @@ u_vbuf_translate_buffers(struct u_vbuf *mgr, struct translate_key *key,
       /* Create and map the output buffer. */
       u_upload_alloc(mgr->uploader,
                      key->output_stride * start_vertex,
-                     key->output_stride * num_vertices,
+                     key->output_stride * num_vertices, 4,
                      &out_offset, &out_buffer,
                      (void**)&out_map);
       if (!out_buffer)
diff --git a/src/gallium/auxiliary/vl/vl_compositor.c b/src/gallium/auxiliary/vl/vl_compositor.c
index afe53063b48..f160df63aa5 100644
--- a/src/gallium/auxiliary/vl/vl_compositor.c
+++ b/src/gallium/auxiliary/vl/vl_compositor.c
@@ -716,6 +716,7 @@ gen_vertex_data(struct vl_compositor *c, struct vl_compositor_state *s, struct u
    /* Allocate new memory for vertices. */
    u_upload_alloc(c->upload, 0,
                   c->vertex_buf.stride * VL_COMPOSITOR_MAX_LAYERS * 4, /* size */
+                  4, /* alignment */
                   &c->vertex_buf.buffer_offset, &c->vertex_buf.buffer,
                   (void**)&vb);
 
diff --git a/src/gallium/drivers/freedreno/a3xx/fd3_context.c b/src/gallium/drivers/freedreno/a3xx/fd3_context.c
index 74cbbf2edd8..2413f152f94 100644
--- a/src/gallium/drivers/freedreno/a3xx/fd3_context.c
+++ b/src/gallium/drivers/freedreno/a3xx/fd3_context.c
@@ -172,7 +172,7 @@ fd3_context_create(struct pipe_screen *pscreen, void *priv, unsigned flags)
 	fd3_query_context_init(pctx);
 
 	fd3_ctx->border_color_uploader = u_upload_create(pctx, 4096,
-			2 * PIPE_MAX_SAMPLERS * BORDERCOLOR_SIZE, 0);
+			BORDER_COLOR_UPLOAD_SIZE, 0);
 
 	return pctx;
 }
diff --git a/src/gallium/drivers/freedreno/a3xx/fd3_emit.c b/src/gallium/drivers/freedreno/a3xx/fd3_emit.c
index 24afbc9e956..e65a352e7f6 100644
--- a/src/gallium/drivers/freedreno/a3xx/fd3_emit.c
+++ b/src/gallium/drivers/freedreno/a3xx/fd3_emit.c
@@ -145,7 +145,8 @@ emit_textures(struct fd_context *ctx, struct fd_ringbuffer *ring,
 	void *ptr;
 
 	u_upload_alloc(fd3_ctx->border_color_uploader,
-			0, 2 * PIPE_MAX_SAMPLERS * BORDERCOLOR_SIZE, &off,
+			0, BORDER_COLOR_UPLOAD_SIZE,
+		       BORDER_COLOR_UPLOAD_SIZE, &off,
 			&fd3_ctx->border_color_buf,
 			&ptr);
 
diff --git a/src/gallium/drivers/freedreno/a4xx/fd4_context.c b/src/gallium/drivers/freedreno/a4xx/fd4_context.c
index e53e0c56c9a..1037adfebf8 100644
--- a/src/gallium/drivers/freedreno/a4xx/fd4_context.c
+++ b/src/gallium/drivers/freedreno/a4xx/fd4_context.c
@@ -172,7 +172,7 @@ fd4_context_create(struct pipe_screen *pscreen, void *priv, unsigned flags)
 	fd4_query_context_init(pctx);
 
 	fd4_ctx->border_color_uploader = u_upload_create(pctx, 4096,
-			2 * PIPE_MAX_SAMPLERS * BORDERCOLOR_SIZE, 0);
+			BORDER_COLOR_UPLOAD_SIZE, 0);
 
 	return pctx;
 }
diff --git a/src/gallium/drivers/freedreno/a4xx/fd4_emit.c b/src/gallium/drivers/freedreno/a4xx/fd4_emit.c
index b9a28149722..bc62a5d9a4b 100644
--- a/src/gallium/drivers/freedreno/a4xx/fd4_emit.c
+++ b/src/gallium/drivers/freedreno/a4xx/fd4_emit.c
@@ -133,7 +133,8 @@ emit_textures(struct fd_context *ctx, struct fd_ringbuffer *ring,
 	void *ptr;
 
 	u_upload_alloc(fd4_ctx->border_color_uploader,
-			0, 2 * PIPE_MAX_SAMPLERS * BORDERCOLOR_SIZE, &off,
+			0, BORDER_COLOR_UPLOAD_SIZE,
+		       BORDER_COLOR_UPLOAD_SIZE, &off,
 			&fd4_ctx->border_color_buf,
 			&ptr);
 
diff --git a/src/gallium/drivers/freedreno/freedreno_context.h b/src/gallium/drivers/freedreno/freedreno_context.h
index 571c8142bf7..418b71b95de 100644
--- a/src/gallium/drivers/freedreno/freedreno_context.h
+++ b/src/gallium/drivers/freedreno/freedreno_context.h
@@ -40,6 +40,8 @@
 #include "freedreno_gmem.h"
 #include "freedreno_util.h"
 
+#define BORDER_COLOR_UPLOAD_SIZE (2 * PIPE_MAX_SAMPLERS * BORDERCOLOR_SIZE)
+
 struct fd_vertex_stateobj;
 
 struct fd_texture_stateobj {
diff --git a/src/gallium/drivers/r300/r300_render_translate.c b/src/gallium/drivers/r300/r300_render_translate.c
index caeeec05909..7221211deea 100644
--- a/src/gallium/drivers/r300/r300_render_translate.c
+++ b/src/gallium/drivers/r300/r300_render_translate.c
@@ -37,7 +37,7 @@ void r300_translate_index_buffer(struct r300_context *r300,
     switch (*index_size) {
     case 1:
         *out_buffer = NULL;
-        u_upload_alloc(r300->uploader, 0, count * 2,
+        u_upload_alloc(r300->uploader, 0, count * 2, 4,
                        &out_offset, out_buffer, &ptr);
 
         util_shorten_ubyte_elts_to_userptr(
@@ -51,7 +51,7 @@ void r300_translate_index_buffer(struct r300_context *r300,
     case 2:
         if (index_offset) {
             *out_buffer = NULL;
-            u_upload_alloc(r300->uploader, 0, count * 2,
+            u_upload_alloc(r300->uploader, 0, count * 2, 4,
                            &out_offset, out_buffer, &ptr);
 
             util_rebuild_ushort_elts_to_userptr(&r300->context, ib,
@@ -65,7 +65,7 @@ void r300_translate_index_buffer(struct r300_context *r300,
     case 4:
         if (index_offset) {
             *out_buffer = NULL;
-            u_upload_alloc(r300->uploader, 0, count * 4,
+            u_upload_alloc(r300->uploader, 0, count * 4, 4,
                            &out_offset, out_buffer, &ptr);
 
             util_rebuild_uint_elts_to_userptr(&r300->context, ib,
diff --git a/src/gallium/drivers/r600/r600_state_common.c b/src/gallium/drivers/r600/r600_state_common.c
index ca589fa7759..3051c9af09c 100644
--- a/src/gallium/drivers/r600/r600_state_common.c
+++ b/src/gallium/drivers/r600/r600_state_common.c
@@ -1732,7 +1732,7 @@ static void r600_draw_vbo(struct pipe_context *ctx, const struct pipe_draw_info
 				}
 			}
 
-			u_upload_alloc(rctx->b.uploader, start, count * 2,
+			u_upload_alloc(rctx->b.uploader, start, count * 2, 256,
 				       &out_offset, &out_buffer, &ptr);
 
 			util_shorten_ubyte_elts_to_userptr(
diff --git a/src/gallium/drivers/radeon/r600_buffer_common.c b/src/gallium/drivers/radeon/r600_buffer_common.c
index 18925277d2d..484f5c8d5b7 100644
--- a/src/gallium/drivers/radeon/r600_buffer_common.c
+++ b/src/gallium/drivers/radeon/r600_buffer_common.c
@@ -298,7 +298,7 @@ static void *r600_buffer_transfer_map(struct pipe_context *ctx,
 			struct r600_resource *staging = NULL;
 
 			u_upload_alloc(rctx->uploader, 0, box->width + (box->x % R600_MAP_BUFFER_ALIGNMENT),
-				       &offset, (struct pipe_resource**)&staging, (void**)&data);
+				       256, &offset, (struct pipe_resource**)&staging, (void**)&data);
 
 			if (staging) {
 				data += box->x % R600_MAP_BUFFER_ALIGNMENT;
diff --git a/src/gallium/drivers/radeon/r600_pipe_common.c b/src/gallium/drivers/radeon/r600_pipe_common.c
index 9a5e9878176..c044b6130a2 100644
--- a/src/gallium/drivers/radeon/r600_pipe_common.c
+++ b/src/gallium/drivers/radeon/r600_pipe_common.c
@@ -85,7 +85,7 @@ void r600_draw_rectangle(struct blitter_context *blitter,
 	/* Upload vertices. The hw rectangle has only 3 vertices,
 	 * I guess the 4th one is derived from the first 3.
 	 * The vertex specification should match u_blitter's vertex element state. */
-	u_upload_alloc(rctx->uploader, 0, sizeof(float) * 24, &offset, &buf, (void**)&vb);
+	u_upload_alloc(rctx->uploader, 0, sizeof(float) * 24, 256, &offset, &buf, (void**)&vb);
 	if (!buf)
 		return;
 
diff --git a/src/gallium/drivers/radeonsi/si_descriptors.c b/src/gallium/drivers/radeonsi/si_descriptors.c
index b3719dea252..5b0ad8f5622 100644
--- a/src/gallium/drivers/radeonsi/si_descriptors.c
+++ b/src/gallium/drivers/radeonsi/si_descriptors.c
@@ -109,7 +109,7 @@ static bool si_upload_descriptors(struct si_context *sctx,
 	if (!desc->list_dirty)
 		return true;
 
-	u_upload_alloc(sctx->b.uploader, 0, list_size,
+	u_upload_alloc(sctx->b.uploader, 0, list_size, 256,
 		       &desc->buffer_offset,
 		       (struct pipe_resource**)&desc->buffer, &ptr);
 	if (!desc->buffer)
@@ -391,7 +391,7 @@ static bool si_upload_vertex_buffer_descriptors(struct si_context *sctx)
 	 * directly through a staging buffer and don't go through
 	 * the fine-grained upload path.
 	 */
-	u_upload_alloc(sctx->b.uploader, 0, count * 16, &desc->buffer_offset,
+	u_upload_alloc(sctx->b.uploader, 0, count * 16, 256, &desc->buffer_offset,
 		       (struct pipe_resource**)&desc->buffer, (void**)&ptr);
 	if (!desc->buffer)
 		return false;
@@ -465,7 +465,7 @@ void si_upload_const_buffer(struct si_context *sctx, struct r600_resource **rbuf
 {
 	void *tmp;
 
-	u_upload_alloc(sctx->b.uploader, 0, size, const_offset,
+	u_upload_alloc(sctx->b.uploader, 0, size, 256, const_offset,
 		       (struct pipe_resource**)rbuffer, &tmp);
 	if (rbuffer)
 		util_memcpy_cpu_to_le32(tmp, ptr, size);
diff --git a/src/gallium/drivers/radeonsi/si_state_draw.c b/src/gallium/drivers/radeonsi/si_state_draw.c
index e5500111f43..d5540bec71d 100644
--- a/src/gallium/drivers/radeonsi/si_state_draw.c
+++ b/src/gallium/drivers/radeonsi/si_state_draw.c
@@ -818,7 +818,7 @@ void si_draw_vbo(struct pipe_context *ctx, const struct pipe_draw_info *info)
 			si_get_draw_start_count(sctx, info, &start, &count);
 			start_offset = start * ib.index_size;
 
-			u_upload_alloc(sctx->b.uploader, start_offset, count * 2,
+			u_upload_alloc(sctx->b.uploader, start_offset, count * 2, 256,
 				       &out_offset, &out_buffer, &ptr);
 			if (!out_buffer) {
 				pipe_resource_reference(&ib.buffer, NULL);
diff --git a/src/gallium/drivers/svga/svga_context.c b/src/gallium/drivers/svga/svga_context.c
index d407785ddd9..97e649e38ba 100644
--- a/src/gallium/drivers/svga/svga_context.c
+++ b/src/gallium/drivers/svga/svga_context.c
@@ -46,7 +46,6 @@
 #include "svga_winsys.h"
 
 #define CONST0_UPLOAD_DEFAULT_SIZE 65536
-#define CONST0_UPLOAD_ALIGNMENT 256
 
 DEBUG_GET_ONCE_BOOL_OPTION(no_swtnl, "SVGA_NO_SWTNL", FALSE)
 DEBUG_GET_ONCE_BOOL_OPTION(force_swtnl, "SVGA_FORCE_SWTNL", FALSE);
diff --git a/src/gallium/drivers/svga/svga_context.h b/src/gallium/drivers/svga/svga_context.h
index 78e346a92b9..c282932cb18 100644
--- a/src/gallium/drivers/svga/svga_context.h
+++ b/src/gallium/drivers/svga/svga_context.h
@@ -74,6 +74,8 @@
  */
 #define SVGA_MAX_CONST_BUF_SIZE (4096 * 4 * sizeof(int))
 
+#define CONST0_UPLOAD_ALIGNMENT 256
+
 struct draw_vertex_shader;
 struct draw_fragment_shader;
 struct svga_shader_variant;
diff --git a/src/gallium/drivers/svga/svga_state_constants.c b/src/gallium/drivers/svga/svga_state_constants.c
index 2cf41134bd6..8ab1693088a 100644
--- a/src/gallium/drivers/svga/svga_state_constants.c
+++ b/src/gallium/drivers/svga/svga_state_constants.c
@@ -613,7 +613,8 @@ emit_constbuf_vgpu10(struct svga_context *svga, unsigned shader)
     */
    new_buf_size = align(new_buf_size, 16);
 
-   u_upload_alloc(svga->const0_upload, 0, new_buf_size, &offset,
+   u_upload_alloc(svga->const0_upload, 0, new_buf_size,
+                  CONST0_UPLOAD_ALIGNMENT, &offset,
                   &dst_buffer, &dst_map);
    if (!dst_map) {
       if (src_map)
diff --git a/src/gallium/drivers/vc4/vc4_resource.c b/src/gallium/drivers/vc4/vc4_resource.c
index 9e6678a0625..308fb9fc77b 100644
--- a/src/gallium/drivers/vc4/vc4_resource.c
+++ b/src/gallium/drivers/vc4/vc4_resource.c
@@ -921,7 +921,7 @@ vc4_get_shadow_index_buffer(struct pipe_context *pctx,
 
         void *data;
         struct pipe_resource *shadow_rsc = NULL;
-        u_upload_alloc(vc4->uploader, 0, count * 2,
+        u_upload_alloc(vc4->uploader, 0, count * 2, 4,;
                        shadow_offset, &shadow_rsc, &data);
         uint16_t *dst = data;