Mesa (main): tu: Make tile stores use a dedicated CS

[GitLab Mirror]

Module: Mesa
Branch: main
Commit: 7948c4b0b43a2a0c15c834f1e066c1e5262146b8
URL:    http://cgit.freedesktop.org/mesa/mesa/commit/?id=7948c4b0b43a2a0c15c834f1e066c1e5262146b8

Author: Connor Abbott <cwabbott0 at gmail.com>
Date:   Wed Jul 28 14:27:59 2021 +0200

tu: Make tile stores use a dedicated CS

We were trying to calculate how much space they need, That was already
difficult and one of the most opaque and hard-to-verify uses of sub_cs,
but it will become even more difficult with the 3D path. What's worse is
that sometimes we have to touch that path when we start touching
registers that would affect rasterization, and there's no indication
that you have to then recalculate the size etc. Just rip this out and
start keeping a separate CS for it instead. Note that this adds a small
amount of memory wastage and extra buffers (at worst one buffer per
command buffer).

Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/12102>

---

 src/freedreno/vulkan/tu_cmd_buffer.c | 33 +++++++++++----------------------
 src/freedreno/vulkan/tu_private.h    |  3 +--
 2 files changed, 12 insertions(+), 24 deletions(-)

diff --git a/src/freedreno/vulkan/tu_cmd_buffer.c b/src/freedreno/vulkan/tu_cmd_buffer.c
index f9bf3e14f61..45eeea2fbdf 100644
--- a/src/freedreno/vulkan/tu_cmd_buffer.c
+++ b/src/freedreno/vulkan/tu_cmd_buffer.c
@@ -1169,7 +1169,7 @@ static void
 tu6_sysmem_render_end(struct tu_cmd_buffer *cmd, struct tu_cs *cs)
 {
    /* Do any resolves of the last subpass. These are handled in the
-    * tile_store_ib in the gmem path.
+    * tile_store_cs in the gmem path.
     */
    tu6_emit_sysmem_resolves(cmd, cs, cmd->state.subpass);
 
@@ -1242,7 +1242,7 @@ tu6_render_tile(struct tu_cmd_buffer *cmd, struct tu_cs *cs)
       tu_cs_emit(cs, A6XX_CP_SET_MARKER_0_MODE(RM6_ENDVIS));
    }
 
-   tu_cs_emit_ib(cs, &cmd->state.tile_store_ib);
+   tu_cs_emit_call(cs, &cmd->tile_store_cs);
 
    tu_cs_sanity_check(cs);
 }
@@ -1299,25 +1299,6 @@ tu_cmd_render_sysmem(struct tu_cmd_buffer *cmd)
    tu6_sysmem_render_end(cmd, &cmd->cs);
 }
 
-static void
-tu_cmd_prepare_tile_store_ib(struct tu_cmd_buffer *cmd)
-{
-   const uint32_t tile_store_space = 7 + (35 * 2) * cmd->state.pass->attachment_count;
-   struct tu_cs sub_cs;
-
-   VkResult result =
-      tu_cs_begin_sub_stream(&cmd->sub_cs, tile_store_space, &sub_cs);
-   if (result != VK_SUCCESS) {
-      cmd->record_result = result;
-      return;
-   }
-
-   /* emit to tile-store sub_cs */
-   tu6_emit_tile_store(cmd, &sub_cs);
-
-   cmd->state.tile_store_ib = tu_cs_end_sub_stream(&cmd->sub_cs, &sub_cs);
-}
-
 static VkResult
 tu_create_cmd_buffer(struct tu_device *device,
                      struct tu_cmd_pool *pool,
@@ -1349,6 +1330,7 @@ tu_create_cmd_buffer(struct tu_device *device,
 
    tu_cs_init(&cmd_buffer->cs, device, TU_CS_MODE_GROW, 4096);
    tu_cs_init(&cmd_buffer->draw_cs, device, TU_CS_MODE_GROW, 4096);
+   tu_cs_init(&cmd_buffer->tile_store_cs, device, TU_CS_MODE_GROW, 2048);
    tu_cs_init(&cmd_buffer->draw_epilogue_cs, device, TU_CS_MODE_GROW, 4096);
    tu_cs_init(&cmd_buffer->sub_cs, device, TU_CS_MODE_SUB_STREAM, 2048);
 
@@ -1364,6 +1346,7 @@ tu_cmd_buffer_destroy(struct tu_cmd_buffer *cmd_buffer)
 
    tu_cs_finish(&cmd_buffer->cs);
    tu_cs_finish(&cmd_buffer->draw_cs);
+   tu_cs_finish(&cmd_buffer->tile_store_cs);
    tu_cs_finish(&cmd_buffer->draw_epilogue_cs);
    tu_cs_finish(&cmd_buffer->sub_cs);
 
@@ -1377,6 +1360,7 @@ tu_reset_cmd_buffer(struct tu_cmd_buffer *cmd_buffer)
 
    tu_cs_reset(&cmd_buffer->cs);
    tu_cs_reset(&cmd_buffer->draw_cs);
+   tu_cs_reset(&cmd_buffer->tile_store_cs);
    tu_cs_reset(&cmd_buffer->draw_epilogue_cs);
    tu_cs_reset(&cmd_buffer->sub_cs);
 
@@ -1510,6 +1494,7 @@ tu_BeginCommandBuffer(VkCommandBuffer commandBuffer,
 
    tu_cs_begin(&cmd_buffer->cs);
    tu_cs_begin(&cmd_buffer->draw_cs);
+   tu_cs_begin(&cmd_buffer->tile_store_cs);
    tu_cs_begin(&cmd_buffer->draw_epilogue_cs);
 
    /* setup initial configuration into command buffer */
@@ -2051,6 +2036,7 @@ tu_EndCommandBuffer(VkCommandBuffer commandBuffer)
 
    tu_cs_end(&cmd_buffer->cs);
    tu_cs_end(&cmd_buffer->draw_cs);
+   tu_cs_end(&cmd_buffer->tile_store_cs);
    tu_cs_end(&cmd_buffer->draw_epilogue_cs);
 
    cmd_buffer->status = TU_CMD_BUFFER_STATUS_EXECUTABLE;
@@ -2927,7 +2913,7 @@ tu_CmdBeginRenderPass2(VkCommandBuffer commandBuffer,
    cmd->state.framebuffer = fb;
    cmd->state.render_area = pRenderPassBegin->renderArea;
 
-   tu_cmd_prepare_tile_store_ib(cmd);
+   tu6_emit_tile_store(cmd, &cmd->tile_store_cs);
 
    /* Note: because this is external, any flushes will happen before draw_cs
     * gets called. However deferred flushes could have to happen later as part
@@ -4364,6 +4350,7 @@ tu_CmdEndRenderPass2(VkCommandBuffer commandBuffer,
    TU_FROM_HANDLE(tu_cmd_buffer, cmd_buffer, commandBuffer);
 
    tu_cs_end(&cmd_buffer->draw_cs);
+   tu_cs_end(&cmd_buffer->tile_store_cs);
    tu_cs_end(&cmd_buffer->draw_epilogue_cs);
 
    if (use_sysmem_rendering(cmd_buffer))
@@ -4382,6 +4369,8 @@ tu_CmdEndRenderPass2(VkCommandBuffer commandBuffer,
       rendered */
    tu_cs_discard_entries(&cmd_buffer->draw_cs);
    tu_cs_begin(&cmd_buffer->draw_cs);
+   tu_cs_discard_entries(&cmd_buffer->tile_store_cs);
+   tu_cs_begin(&cmd_buffer->tile_store_cs);
    tu_cs_discard_entries(&cmd_buffer->draw_epilogue_cs);
    tu_cs_begin(&cmd_buffer->draw_epilogue_cs);
 
diff --git a/src/freedreno/vulkan/tu_private.h b/src/freedreno/vulkan/tu_private.h
index c0d4e8ef899..153e1880e0a 100644
--- a/src/freedreno/vulkan/tu_private.h
+++ b/src/freedreno/vulkan/tu_private.h
@@ -974,8 +974,6 @@ struct tu_cmd_state
    const struct tu_framebuffer *framebuffer;
    VkRect2D render_area;
 
-   struct tu_cs_entry tile_store_ib;
-
    bool xfb_used;
    bool has_tess;
    bool has_subpass_predication;
@@ -1034,6 +1032,7 @@ struct tu_cmd_buffer
 
    struct tu_cs cs;
    struct tu_cs draw_cs;
+   struct tu_cs tile_store_cs;
    struct tu_cs draw_epilogue_cs;
    struct tu_cs sub_cs;