i965: Add and use brw_bo_map()

We can encapsulate the logic for choosing the mapping type. This will
also help when we add WC mappings.

A few functional changes are made in this patch. On non-LLC, what were
previously WB mappings are now GTT mappings (in the prefilling debug
code in brw_performance_query.c; the shader_time code in brw_program.c;
and in the case of an RW mapping in intel_buffer_objects.c).

Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>

diff --git a/src/mesa/drivers/dri/i965/brw_bufmgr.c b/src/mesa/drivers/dri/i965/brw_bufmgr.c
index b79f56669068514419b8be86f611bfe66ac3ba62..ec9611feb922f91846a5ef9172289392cf58088b 100644 (file)
--- a/src/mesa/drivers/dri/i965/brw_bufmgr.c
+++ b/src/mesa/drivers/dri/i965/brw_bufmgr.c
@@ -658,7 +658,7 @@ set_domain(struct brw_context *brw, const char *action,
    }
 }
 
-void *
+static void *
 brw_bo_map_cpu(struct brw_context *brw, struct brw_bo *bo, unsigned flags)
 {
    struct brw_bufmgr *bufmgr = bo->bufmgr;
@@ -740,7 +740,7 @@ map_gtt(struct brw_bo *bo)
    return bo->map_gtt;
 }
 
-void *
+static void *
 brw_bo_map_gtt(struct brw_context *brw, struct brw_bo *bo, unsigned flags)
 {
    struct brw_bufmgr *bufmgr = bo->bufmgr;
@@ -814,6 +814,32 @@ brw_bo_map_unsynchronized(struct brw_context *brw, struct brw_bo *bo)
    return map;
 }
 
+static bool
+can_map_cpu(struct brw_bo *bo, unsigned flags)
+{
+   if (bo->cache_coherent)
+      return true;
+
+   if (flags & MAP_PERSISTENT)
+      return false;
+
+   if (flags & MAP_COHERENT)
+      return false;
+
+   return !(flags & MAP_WRITE);
+}
+
+void *
+brw_bo_map(struct brw_context *brw, struct brw_bo *bo, unsigned flags)
+{
+   if (bo->tiling_mode != I915_TILING_NONE && !(flags & MAP_RAW))
+      return brw_bo_map_gtt(brw, bo, flags);
+   else if (can_map_cpu(bo, flags))
+      return brw_bo_map_cpu(brw, bo, flags);
+   else
+      return brw_bo_map_gtt(brw, bo, flags);
+}
+
 int
 brw_bo_unmap(struct brw_bo *bo)
 {

diff --git a/src/mesa/drivers/dri/i965/brw_bufmgr.h b/src/mesa/drivers/dri/i965/brw_bufmgr.h
index 831da6960d3fc90024d0d5bf0d974e92e58ce37b..099afcf4c5756bb469b9944da8ea9240b26de275 100644 (file)
--- a/src/mesa/drivers/dri/i965/brw_bufmgr.h
+++ b/src/mesa/drivers/dri/i965/brw_bufmgr.h
@@ -137,7 +137,7 @@ struct brw_bo {
  *
  * Buffer objects are not necessarily initially mapped into CPU virtual
  * address space or graphics device aperture.  They must be mapped
- * using brw_bo_map_cpu() or brw_bo_map_gtt() to be used by the CPU.
+ * using brw_bo_map() to be used by the CPU.
  */
 struct brw_bo *brw_bo_alloc(struct brw_bufmgr *bufmgr, const char *name,
                             uint64_t size, uint64_t alignment);
@@ -189,7 +189,7 @@ void brw_bo_unreference(struct brw_bo *bo);
  * This function will block waiting for any existing execution on the
  * buffer to complete, first.  The resulting mapping is returned.
  */
-MUST_CHECK void *brw_bo_map_cpu(struct brw_context *brw, struct brw_bo *bo, unsigned flags);
+MUST_CHECK void *brw_bo_map(struct brw_context *brw, struct brw_bo *bo, unsigned flags);
 
 /**
  * Reduces the refcount on the userspace mapping of the buffer
@@ -263,7 +263,6 @@ struct brw_bo *brw_bo_gem_create_from_name(struct brw_bufmgr *bufmgr,
                                            unsigned int handle);
 void brw_bufmgr_enable_reuse(struct brw_bufmgr *bufmgr);
 MUST_CHECK void *brw_bo_map_unsynchronized(struct brw_context *brw, struct brw_bo *bo);
-MUST_CHECK void *brw_bo_map_gtt(struct brw_context *brw, struct brw_bo *bo, unsigned flags);
 
 int brw_bo_wait(struct brw_bo *bo, int64_t timeout_ns);
 

diff --git a/src/mesa/drivers/dri/i965/brw_performance_query.c b/src/mesa/drivers/dri/i965/brw_performance_query.c
index 2ec070bd6f5ea70d89cb43d60393aa1e8a5a8c14..1c9ddf52ea378eea7e5553030f038357e186a907 100644 (file)
--- a/src/mesa/drivers/dri/i965/brw_performance_query.c
+++ b/src/mesa/drivers/dri/i965/brw_performance_query.c
@@ -713,7 +713,7 @@ accumulate_oa_reports(struct brw_context *brw,
    if (!read_oa_samples(brw))
       goto error;
 
-   query_buffer = brw_bo_map_cpu(brw, obj->oa.bo, MAP_READ);
+   query_buffer = brw_bo_map(brw, obj->oa.bo, MAP_READ);
 
    start = last = query_buffer;
    end = query_buffer + (MI_RPC_BO_END_OFFSET_BYTES / sizeof(uint32_t));
@@ -992,7 +992,7 @@ brw_begin_perf_query(struct gl_context *ctx,
                       MI_RPC_BO_SIZE, 64);
 #ifdef DEBUG
       /* Pre-filling the BO helps debug whether writes landed. */
-      void *map = brw_bo_map_cpu(brw, obj->oa.bo, MAP_WRITE);
+      void *map = brw_bo_map(brw, obj->oa.bo, MAP_WRITE);
       memset(map, 0x80, MI_RPC_BO_SIZE);
       brw_bo_unmap(obj->oa.bo);
 #endif
@@ -1214,7 +1214,7 @@ get_pipeline_stats_data(struct brw_context *brw,
    int n_counters = obj->query->n_counters;
    uint8_t *p = data;
 
-   uint64_t *start = brw_bo_map_cpu(brw, obj->pipeline_stats.bo, MAP_READ);
+   uint64_t *start = brw_bo_map(brw, obj->pipeline_stats.bo, MAP_READ);
    uint64_t *end = start + (STATS_BO_END_OFFSET_BYTES / sizeof(uint64_t));
 
    for (int i = 0; i < n_counters; i++) {

diff --git a/src/mesa/drivers/dri/i965/brw_program.c b/src/mesa/drivers/dri/i965/brw_program.c
index 7f87e7302ff27e230efcbcb31d448d3b09447c71..bff3475b3d94e1a10086a1184be172652f1555b3 100644 (file)
--- a/src/mesa/drivers/dri/i965/brw_program.c
+++ b/src/mesa/drivers/dri/i965/brw_program.c
@@ -578,7 +578,7 @@ brw_collect_shader_time(struct brw_context *brw)
     * delaying reading the reports, but it doesn't look like it's a big
     * overhead compared to the cost of tracking the time in the first place.
     */
-   void *bo_map = brw_bo_map_cpu(brw, brw->shader_time.bo, MAP_READ | MAP_WRITE);
+   void *bo_map = brw_bo_map(brw, brw->shader_time.bo, MAP_READ | MAP_WRITE);
 
    for (int i = 0; i < brw->shader_time.num_entries; i++) {
       uint32_t *times = bo_map + i * 3 * BRW_SHADER_TIME_STRIDE;

diff --git a/src/mesa/drivers/dri/i965/brw_program_cache.c b/src/mesa/drivers/dri/i965/brw_program_cache.c
index 079e2ae53b5c363ad4b3ce22b78a8c5a5bad25e9..ab03969db91a248a5f09d66bd10ba5e0e84ddf94 100644 (file)
--- a/src/mesa/drivers/dri/i965/brw_program_cache.c
+++ b/src/mesa/drivers/dri/i965/brw_program_cache.c
@@ -227,7 +227,7 @@ brw_cache_new_bo(struct brw_cache *cache, uint32_t new_size)
       if (brw->has_llc) {
          memcpy(llc_map, cache->map, cache->next_offset);
       } else {
-         void *map = brw_bo_map_cpu(brw, cache->bo, MAP_READ);
+         void *map = brw_bo_map(brw, cache->bo, MAP_READ);
          brw_bo_subdata(new_bo, 0, cache->next_offset, map);
          brw_bo_unmap(cache->bo);
       }
@@ -268,7 +268,7 @@ brw_lookup_prog(const struct brw_cache *cache,
 
          void *map;
          if (!brw->has_llc)
-            map = brw_bo_map_cpu(brw, cache->bo, MAP_READ);
+            map = brw_bo_map(brw, cache->bo, MAP_READ);
          else
             map = cache->map;
 
@@ -550,7 +550,7 @@ brw_print_program_cache(struct brw_context *brw)
    void *map;
 
    if (!brw->has_llc)
-      map = brw_bo_map_cpu(brw, cache->bo, MAP_READ);
+      map = brw_bo_map(brw, cache->bo, MAP_READ);
    else
       map = cache->map;
 

diff --git a/src/mesa/drivers/dri/i965/brw_queryobj.c b/src/mesa/drivers/dri/i965/brw_queryobj.c
index 05e23cda5c9a91b3d8ebed69454db416e4c44b0c..a7b896243d6fbf3559e966252d1ae1d984f9f2b7 100644 (file)
--- a/src/mesa/drivers/dri/i965/brw_queryobj.c
+++ b/src/mesa/drivers/dri/i965/brw_queryobj.c
@@ -146,7 +146,7 @@ brw_queryobj_get_results(struct gl_context *ctx,
       }
    }
 
-   results = brw_bo_map_cpu(brw, query->bo, MAP_READ);
+   results = brw_bo_map(brw, query->bo, MAP_READ);
    switch (query->Base.Target) {
    case GL_TIME_ELAPSED_EXT:
       /* The query BO contains the starting and ending timestamps.

diff --git a/src/mesa/drivers/dri/i965/gen6_queryobj.c b/src/mesa/drivers/dri/i965/gen6_queryobj.c
index ab15112d81e9c2b46d8db15fe7fd71013b59325e..4fd30ec06492cd40c88c35a11da1734410b7950e 100644 (file)
--- a/src/mesa/drivers/dri/i965/gen6_queryobj.c
+++ b/src/mesa/drivers/dri/i965/gen6_queryobj.c
@@ -221,7 +221,7 @@ gen6_queryobj_get_results(struct gl_context *ctx,
    if (query->bo == NULL)
       return;
 
-   uint64_t *results = brw_bo_map_cpu(brw, query->bo, MAP_READ);
+   uint64_t *results = brw_bo_map(brw, query->bo, MAP_READ);
    switch (query->Base.Target) {
    case GL_TIME_ELAPSED:
       /* The query BO contains the starting and ending timestamps.

diff --git a/src/mesa/drivers/dri/i965/gen6_sol.c b/src/mesa/drivers/dri/i965/gen6_sol.c
index 5873afdbbea93ca826d9f4ed86ec5af7bb164444..b4824b67e5b25b99902156ce645db6e8ff2e8cf7 100644 (file)
--- a/src/mesa/drivers/dri/i965/gen6_sol.c
+++ b/src/mesa/drivers/dri/i965/gen6_sol.c
@@ -247,7 +247,7 @@ tally_prims_generated(struct brw_context *brw,
    if (unlikely(brw->perf_debug && brw_bo_busy(obj->prim_count_bo)))
       perf_debug("Stalling for # of transform feedback primitives written.\n");
 
-   uint64_t *prim_counts = brw_bo_map_cpu(brw, obj->prim_count_bo, MAP_READ);
+   uint64_t *prim_counts = brw_bo_map(brw, obj->prim_count_bo, MAP_READ);
 
    assert(obj->prim_count_buffer_index % (2 * streams) == 0);
    int pairs = obj->prim_count_buffer_index / (2 * streams);

diff --git a/src/mesa/drivers/dri/i965/intel_batchbuffer.c b/src/mesa/drivers/dri/i965/intel_batchbuffer.c
index de93aeb0b2911259bfbb9b93a46044e42176dce6..62d2fe8ef35b4bb0271b8252a93124706cf9108c 100644 (file)
--- a/src/mesa/drivers/dri/i965/intel_batchbuffer.c
+++ b/src/mesa/drivers/dri/i965/intel_batchbuffer.c
@@ -100,7 +100,7 @@ intel_batchbuffer_reset(struct intel_batchbuffer *batch,
 
    batch->bo = brw_bo_alloc(bufmgr, "batchbuffer", BATCH_SZ, 4096);
    if (has_llc) {
-      batch->map = brw_bo_map_cpu(NULL, batch->bo, MAP_READ | MAP_WRITE);
+      batch->map = brw_bo_map(NULL, batch->bo, MAP_READ | MAP_WRITE);
    }
    batch->map_next = batch->map;
 
@@ -239,7 +239,7 @@ do_batch_dump(struct brw_context *brw)
    if (batch->ring != RENDER_RING)
       return;
 
-   void *map = brw_bo_map_cpu(brw, batch->bo, MAP_READ);
+   void *map = brw_bo_map(brw, batch->bo, MAP_READ);
    if (map == NULL) {
       fprintf(stderr,
              "WARNING: failed to map batchbuffer, "

diff --git a/src/mesa/drivers/dri/i965/intel_buffer_objects.c b/src/mesa/drivers/dri/i965/intel_buffer_objects.c
index cf6382d145e02f0be98f330df962bc43419d0ea1..5813989ce4894f822a1548f2d0b9ae622e19b874 100644 (file)
--- a/src/mesa/drivers/dri/i965/intel_buffer_objects.c
+++ b/src/mesa/drivers/dri/i965/intel_buffer_objects.c
@@ -395,12 +395,7 @@ brw_map_buffer_range(struct gl_context *ctx,
                                                           length +
                                                           intel_obj->map_extra[index],
                                                           alignment);
-      void *map;
-      if (brw->has_llc) {
-         map = brw_bo_map_cpu(brw, intel_obj->range_map_bo[index], access);
-      } else {
-         map = brw_bo_map_gtt(brw, intel_obj->range_map_bo[index], access);
-      }
+      void *map = brw_bo_map(brw, intel_obj->range_map_bo[index], access);
       obj->Mappings[index].Pointer = map + intel_obj->map_extra[index];
       return obj->Mappings[index].Pointer;
    }
@@ -412,12 +407,8 @@ brw_map_buffer_range(struct gl_context *ctx,
          perf_debug("MapBufferRange with GL_MAP_UNSYNCHRONIZED_BIT stalling (it's actually synchronized on non-LLC platforms)\n");
       }
       map = brw_bo_map_unsynchronized(brw, intel_obj->buffer);
-   } else if (!brw->has_llc && (!(access & GL_MAP_READ_BIT) ||
-                              (access & GL_MAP_PERSISTENT_BIT))) {
-      map = brw_bo_map_gtt(brw, intel_obj->buffer, access);
-      mark_buffer_inactive(intel_obj);
    } else {
-      map = brw_bo_map_cpu(brw, intel_obj->buffer, access);
+      map = brw_bo_map(brw, intel_obj->buffer, access);
       mark_buffer_inactive(intel_obj);
    }
 

diff --git a/src/mesa/drivers/dri/i965/intel_mipmap_tree.c b/src/mesa/drivers/dri/i965/intel_mipmap_tree.c
index f81e087ec71f0007913d422a62e914384a80ec09..07e9ecf4f4fd68d44fab11477210f24530e2dfb3 100644 (file)
--- a/src/mesa/drivers/dri/i965/intel_mipmap_tree.c
+++ b/src/mesa/drivers/dri/i965/intel_mipmap_tree.c
@@ -1386,7 +1386,7 @@ intel_miptree_init_mcs(struct brw_context *brw,
     *
     * Note: the clear value for MCS buffers is all 1's, so we memset to 0xff.
     */
-   void *map = brw_bo_map_gtt(brw, mt->mcs_buf->bo, MAP_WRITE);
+   void *map = brw_bo_map(brw, mt->mcs_buf->bo, MAP_WRITE);
    if (unlikely(map == NULL)) {
       fprintf(stderr, "Failed to map mcs buffer into GTT\n");
       brw_bo_unreference(mt->mcs_buf->bo);
@@ -2426,22 +2426,7 @@ intel_miptree_map_raw(struct brw_context *brw,
    if (brw_batch_references(&brw->batch, bo))
       intel_batchbuffer_flush(brw);
 
-   /* brw_bo_map_cpu() uses a WB mmaping of the buffer's backing storage. It
-    * will utilize the CPU cache even if the buffer is incoherent with the
-    * GPU (i.e. any writes will be stored in the cache and not flushed to
-    * memory and so will be invisible to the GPU or display engine). This
-    * is the majority of buffers on a !llc machine, but even on a llc
-    * almost all scanouts are incoherent with the CPU. A WB write into the
-    * backing storage of the current scanout will not be immediately
-    * visible on the screen. The transfer from cache to screen is slow and
-    * indeterministic causing visible glitching on the screen. Never use
-    * this WB mapping for writes to an active scanout (reads are fine, so
-    * long as cache consistency is maintained).
-    */
-   if (mt->tiling != I915_TILING_NONE || mt->is_scanout)
-      return brw_bo_map_gtt(brw, bo, mode);
-   else
-      return brw_bo_map_cpu(brw, bo, mode);
+   return brw_bo_map(brw, bo, mode);
 }
 
 static void

diff --git a/src/mesa/drivers/dri/i965/intel_pixel_read.c b/src/mesa/drivers/dri/i965/intel_pixel_read.c
index 9d8c6f4793416da7b3c68b0feadf934a0257a8f8..4614279c6434c0e64e9916fa6d65a70748772a22 100644 (file)
--- a/src/mesa/drivers/dri/i965/intel_pixel_read.c
+++ b/src/mesa/drivers/dri/i965/intel_pixel_read.c
@@ -145,7 +145,7 @@ intel_readpixels_tiled_memcpy(struct gl_context * ctx,
       intel_batchbuffer_flush(brw);
    }
 
-   void *map = brw_bo_map_cpu(brw, bo, MAP_READ);
+   void *map = brw_bo_map(brw, bo, MAP_READ | MAP_RAW);
    if (map == NULL) {
       DBG("%s: failed to map bo\n", __func__);
       return false;

diff --git a/src/mesa/drivers/dri/i965/intel_screen.c b/src/mesa/drivers/dri/i965/intel_screen.c
index 90832ae6a7b003bfab07033a3b466e2f87caa9d8..9354cd53ba9569a41c14194b71dae64ad5afea1d 100644 (file)
--- a/src/mesa/drivers/dri/i965/intel_screen.c
+++ b/src/mesa/drivers/dri/i965/intel_screen.c
@@ -1420,7 +1420,7 @@ intel_detect_pipelined_register(struct intel_screen *screen,
    if (bo == NULL)
       goto err_results;
 
-   map = brw_bo_map_cpu(NULL, bo, MAP_WRITE);
+   map = brw_bo_map(NULL, bo, MAP_WRITE);
    if (!map)
       goto err_batch;
 
@@ -1477,7 +1477,7 @@ intel_detect_pipelined_register(struct intel_screen *screen,
    drmIoctl(dri_screen->fd, DRM_IOCTL_I915_GEM_EXECBUFFER2, &execbuf);
 
    /* Check whether the value got written. */
-   void *results_map = brw_bo_map_cpu(NULL, results, MAP_READ);
+   void *results_map = brw_bo_map(NULL, results, MAP_READ);
    if (results_map) {
       success = *((uint32_t *)results_map + offset) == expected_value;
       brw_bo_unmap(results);

diff --git a/src/mesa/drivers/dri/i965/intel_tex_image.c b/src/mesa/drivers/dri/i965/intel_tex_image.c
index c34f73b5364efcf6fed3ef9feff6bee9858e1c84..828f20b7b6725bc987719ba31a8bbbabb299b8f3 100644 (file)
--- a/src/mesa/drivers/dri/i965/intel_tex_image.c
+++ b/src/mesa/drivers/dri/i965/intel_tex_image.c
@@ -534,7 +534,7 @@ intel_gettexsubimage_tiled_memcpy(struct gl_context *ctx,
       intel_batchbuffer_flush(brw);
    }
 
-   void *map = brw_bo_map_cpu(brw, bo, MAP_READ);
+   void *map = brw_bo_map(brw, bo, MAP_READ | MAP_RAW);
    if (map == NULL) {
       DBG("%s: failed to map bo\n", __func__);
       return false;

diff --git a/src/mesa/drivers/dri/i965/intel_tex_subimage.c b/src/mesa/drivers/dri/i965/intel_tex_subimage.c
index b0b1a0ff68e03939bb5dcbb48ae01d3df23f5325..40569e17a338d281ae3b7a81d744510e892c357b 100644 (file)
--- a/src/mesa/drivers/dri/i965/intel_tex_subimage.c
+++ b/src/mesa/drivers/dri/i965/intel_tex_subimage.c
@@ -146,7 +146,7 @@ intel_texsubimage_tiled_memcpy(struct gl_context * ctx,
       intel_batchbuffer_flush(brw);
    }
 
-   void *map = brw_bo_map_cpu(brw, bo, MAP_WRITE);
+   void *map = brw_bo_map(brw, bo, MAP_WRITE | MAP_RAW);
    if (map == NULL) {
       DBG("%s: failed to map bo\n", __func__);
       return false;

diff --git a/src/mesa/drivers/dri/i965/intel_upload.c b/src/mesa/drivers/dri/i965/intel_upload.c
index fc5b946fc0876ea857876e225bf8ad7ff93721ff..4b5d880a5fe5b28e775d22e8b132a2ef85f5e0d7 100644 (file)
--- a/src/mesa/drivers/dri/i965/intel_upload.c
+++ b/src/mesa/drivers/dri/i965/intel_upload.c
@@ -100,10 +100,7 @@ intel_upload_space(struct brw_context *brw,
    if (!brw->upload.bo) {
       brw->upload.bo = brw_bo_alloc(brw->bufmgr, "streamed data",
                                     MAX2(INTEL_UPLOAD_SIZE, size), 4096);
-      if (brw->has_llc)
-         brw->upload.map = brw_bo_map_cpu(brw, brw->upload.bo, MAP_READ | MAP_WRITE);
-      else
-         brw->upload.map = brw_bo_map_gtt(brw, brw->upload.bo, MAP_READ | MAP_WRITE);
+      brw->upload.map = brw_bo_map(brw, brw->upload.bo, MAP_READ | MAP_WRITE);
    }
 
    brw->upload.next_offset = offset + size;