[Mesa-dev] [PATCH 13/14] i965/gen7-8: Set up early depth/stencil control appropriately for image load/store.

[Francisco Jerez]

---
 src/mesa/drivers/dri/i965/brw_defines.h      |  3 +++
 src/mesa/drivers/dri/i965/gen7_wm_state.c    |  7 +++++++
 src/mesa/drivers/dri/i965/gen8_depth_state.c | 12 ++++++++----
 src/mesa/drivers/dri/i965/gen8_ps_state.c    | 13 +++++++++++++
 4 files changed, 31 insertions(+), 4 deletions(-)

diff --git a/src/mesa/drivers/dri/i965/brw_defines.h b/src/mesa/drivers/dri/i965/brw_defines.h
index 7b151f6..4a3acee 100644
--- a/src/mesa/drivers/dri/i965/brw_defines.h
+++ b/src/mesa/drivers/dri/i965/brw_defines.h
@@ -2220,6 +2220,9 @@ enum brw_wm_barycentric_interp_mode {
 # define GEN7_WM_KILL_ENABLE				(1 << 25)
 # define GEN7_WM_COMPUTED_DEPTH_MODE_SHIFT              23
 # define GEN7_WM_USES_SOURCE_DEPTH			(1 << 20)
+# define GEN7_WM_EARLY_DS_CONTROL_NORMAL                (0 << 21)
+# define GEN7_WM_EARLY_DS_CONTROL_PSEXEC                (1 << 21)
+# define GEN7_WM_EARLY_DS_CONTROL_PREPS                 (2 << 21)
 # define GEN7_WM_USES_SOURCE_W			        (1 << 19)
 # define GEN7_WM_POSITION_ZW_PIXEL			(0 << 17)
 # define GEN7_WM_POSITION_ZW_CENTROID			(2 << 17)
diff --git a/src/mesa/drivers/dri/i965/gen7_wm_state.c b/src/mesa/drivers/dri/i965/gen7_wm_state.c
index 31b1d48..744bba3 100644
--- a/src/mesa/drivers/dri/i965/gen7_wm_state.c
+++ b/src/mesa/drivers/dri/i965/gen7_wm_state.c
@@ -105,6 +105,13 @@ upload_wm_state(struct brw_context *brw)
       dw1 |= GEN7_WM_USES_INPUT_COVERAGE_MASK;
    }
 
+   if (shader) {
+      if (shader->EarlyFragmentTests)
+         dw1 |= GEN7_WM_EARLY_DS_CONTROL_PREPS;
+      else if (shader->NumImages)
+         dw1 |= GEN7_WM_EARLY_DS_CONTROL_PSEXEC;
+   }
+
    BEGIN_BATCH(3);
    OUT_BATCH(_3DSTATE_WM << 16 | (3 - 2));
    OUT_BATCH(dw1);
diff --git a/src/mesa/drivers/dri/i965/gen8_depth_state.c b/src/mesa/drivers/dri/i965/gen8_depth_state.c
index e428089..7b4741a 100644
--- a/src/mesa/drivers/dri/i965/gen8_depth_state.c
+++ b/src/mesa/drivers/dri/i965/gen8_depth_state.c
@@ -241,10 +241,13 @@ pma_fix_enable(const struct brw_context *brw)
     */
    const bool hiz_enabled = depth_irb && intel_renderbuffer_has_hiz(depth_irb);
 
-   /* 3DSTATE_WM::Early Depth/Stencil Control != EDSC_PREPS (2).
-    * We always leave this set to EDSC_NORMAL (0).
+   /* BRW_NEW_FRAGMENT_PROGRAM:
+    * 3DSTATE_WM::Early Depth/Stencil Control != EDSC_PREPS (2).
     */
-   const bool edsc_not_preps = true;
+   struct gl_shader_program *prog = ctx->Shader._CurrentFragmentProgram;
+   struct gl_shader *shader =
+      prog ? prog->_LinkedShaders[MESA_SHADER_FRAGMENT] : NULL;
+   const bool edsc_not_preps = !(shader && shader->EarlyFragmentTests);
 
    /* 3DSTATE_PS_EXTRA::PixelShaderValid is always true. */
    const bool pixel_shader_valid = true;
@@ -369,7 +372,8 @@ const struct brw_tracked_state gen8_pma_fix = {
               _NEW_DEPTH |
               _NEW_MULTISAMPLE |
               _NEW_STENCIL,
-      .brw = BRW_NEW_FS_PROG_DATA,
+      .brw = BRW_NEW_FRAGMENT_PROGRAM |
+             BRW_NEW_FS_PROG_DATA,
    },
    .emit = gen8_emit_pma_stall_workaround
 };
diff --git a/src/mesa/drivers/dri/i965/gen8_ps_state.c b/src/mesa/drivers/dri/i965/gen8_ps_state.c
index b932437..97c2d41 100644
--- a/src/mesa/drivers/dri/i965/gen8_ps_state.c
+++ b/src/mesa/drivers/dri/i965/gen8_ps_state.c
@@ -86,6 +86,10 @@ static void
 upload_wm_state(struct brw_context *brw)
 {
    struct gl_context *ctx = &brw->ctx;
+   /* BRW_NEW_FRAGMENT_PROGRAM */
+   struct gl_shader_program *prog = ctx->Shader._CurrentFragmentProgram;
+   struct gl_shader *shader =
+      prog ? prog->_LinkedShaders[MESA_SHADER_FRAGMENT] : NULL;
    uint32_t dw1 = 0;
 
    dw1 |= GEN7_WM_STATISTICS_ENABLE;
@@ -105,6 +109,14 @@ upload_wm_state(struct brw_context *brw)
    dw1 |= brw->wm.prog_data->barycentric_interp_modes <<
       GEN7_WM_BARYCENTRIC_INTERPOLATION_MODE_SHIFT;
 
+   /* BRW_NEW_FRAGMENT_PROGRAM */
+   if (shader) {
+      if (shader->EarlyFragmentTests)
+         dw1 |= GEN7_WM_EARLY_DS_CONTROL_PREPS;
+      else if (shader->NumImages)
+         dw1 |= GEN7_WM_EARLY_DS_CONTROL_PSEXEC;
+   }
+
    BEGIN_BATCH(2);
    OUT_BATCH(_3DSTATE_WM << 16 | (2 - 2));
    OUT_BATCH(dw1);
@@ -116,6 +128,7 @@ const struct brw_tracked_state gen8_wm_state = {
       .mesa  = _NEW_LINE |
                _NEW_POLYGON,
       .brw   = BRW_NEW_CONTEXT |
+               BRW_NEW_FRAGMENT_PROGRAM |
                BRW_NEW_FS_PROG_DATA,
    },
    .emit = upload_wm_state,
-- 
2.1.3