i965/bxt: Add basic Broxton infrastructure

[mesa.git] / src / mesa / drivers / dri / i965 / brw_wm_state.c

diff --git a/src/mesa/drivers/dri/i965/brw_wm_state.c b/src/mesa/drivers/dri/i965/brw_wm_state.c
index 6a53a98eec62bf19a1af6e6274d318095734c22a..0cd439046b38796902fac3fa530514a4bda8126c 100644 (file)
--- a/src/mesa/drivers/dri/i965/brw_wm_state.c
+++ b/src/mesa/drivers/dri/i965/brw_wm_state.c
@@ -77,7 +77,7 @@ brw_upload_wm_unit(struct brw_context *brw)
    struct gl_context *ctx = &brw->ctx;
    /* BRW_NEW_FRAGMENT_PROGRAM */
    const struct gl_fragment_program *fp = brw->fragment_program;
-   /* CACHE_NEW_WM_PROG */
+   /* BRW_NEW_FS_PROG_DATA */
    const struct brw_wm_prog_data *prog_data = brw->wm.prog_data;
    struct brw_wm_unit_state *wm;
 
@@ -94,32 +94,44 @@ brw_upload_wm_unit(struct brw_context *brw)
             prog_data->dispatch_grf_start_reg_16);
    }
 
-   /* BRW_NEW_PROGRAM_CACHE | CACHE_NEW_WM_PROG */
-   wm->thread0.grf_reg_count = prog_data->reg_blocks;
-   wm->wm9.grf_reg_count_2 = prog_data->reg_blocks_16;
-
-   wm->thread0.kernel_start_pointer =
-      brw_program_reloc(brw,
-                       brw->wm.base.state_offset +
-                       offsetof(struct brw_wm_unit_state, thread0),
-                       brw->wm.base.prog_offset +
-                       (wm->thread0.grf_reg_count << 1)) >> 6;
-
-   wm->wm9.kernel_start_pointer_2 =
-      brw_program_reloc(brw,
-                       brw->wm.base.state_offset +
-                       offsetof(struct brw_wm_unit_state, wm9),
-                       brw->wm.base.prog_offset +
-                       prog_data->prog_offset_16 +
-                       (wm->wm9.grf_reg_count_2 << 1)) >> 6;
+   /* BRW_NEW_PROGRAM_CACHE | BRW_NEW_FS_PROG_DATA */
+   if (prog_data->no_8) {
+      wm->wm5.enable_16_pix = 1;
+      wm->thread0.grf_reg_count = prog_data->reg_blocks_16;
+      wm->thread0.kernel_start_pointer =
+         brw_program_reloc(brw,
+                           brw->wm.base.state_offset +
+                           offsetof(struct brw_wm_unit_state, thread0),
+                           brw->wm.base.prog_offset +
+                           prog_data->prog_offset_16 +
+                           (prog_data->reg_blocks_16 << 1)) >> 6;
+
+   } else {
+      wm->thread0.grf_reg_count = prog_data->reg_blocks;
+      wm->wm9.grf_reg_count_2 = prog_data->reg_blocks_16;
+
+      wm->wm5.enable_8_pix = 1;
+      if (prog_data->prog_offset_16)
+         wm->wm5.enable_16_pix = 1;
+
+      wm->thread0.kernel_start_pointer =
+         brw_program_reloc(brw,
+                           brw->wm.base.state_offset +
+                           offsetof(struct brw_wm_unit_state, thread0),
+                           brw->wm.base.prog_offset +
+                           (wm->thread0.grf_reg_count << 1)) >> 6;
+
+      wm->wm9.kernel_start_pointer_2 =
+         brw_program_reloc(brw,
+                           brw->wm.base.state_offset +
+                           offsetof(struct brw_wm_unit_state, wm9),
+                           brw->wm.base.prog_offset +
+                           prog_data->prog_offset_16 +
+                           (wm->wm9.grf_reg_count_2 << 1)) >> 6;
+   }
 
    wm->thread1.depth_coef_urb_read_offset = 1;
-   /* Use ALT floating point mode for ARB fragment programs, because they
-    * require 0^0 == 1.  Even though _CurrentFragmentProgram is used for
-    * rendering, CurrentProgram[MESA_SHADER_FRAGMENT] is used for this check
-    * to differentiate between the GLSL and non-GLSL cases.
-    */
-   if (ctx->Shader.CurrentProgram[MESA_SHADER_FRAGMENT] == NULL)
+   if (prog_data->base.use_alt_mode)
       wm->thread1.floating_point_mode = BRW_FLOATING_POINT_NON_IEEE_754;
    else
       wm->thread1.floating_point_mode = BRW_FLOATING_POINT_IEEE_754;
@@ -150,12 +162,11 @@ brw_upload_wm_unit(struct brw_context *brw)
    if (brw->gen == 5)
       wm->wm4.sampler_count = 0; /* hardware requirement */
    else {
-      /* CACHE_NEW_SAMPLER */
       wm->wm4.sampler_count = (brw->wm.base.sampler_count + 1) / 4;
    }
 
    if (brw->wm.base.sampler_count) {
-      /* reloc */
+      /* BRW_NEW_SAMPLER_STATE_TABLE - reloc */
       wm->wm4.sampler_state_pointer = (brw->batch.bo->offset64 +
                                       brw->wm.base.sampler_offset) >> 5;
    } else {
@@ -178,10 +189,6 @@ brw_upload_wm_unit(struct brw_context *brw)
    wm->wm5.program_uses_killpixel =
       prog_data->uses_kill || ctx->Color.AlphaEnabled;
 
-   wm->wm5.enable_8_pix = 1;
-   if (prog_data->prog_offset_16)
-      wm->wm5.enable_16_pix = 1;
-
    wm->wm5.max_threads = brw->max_wm_threads - 1;
 
    /* _NEW_BUFFERS | _NEW_COLOR */
@@ -203,7 +210,7 @@ brw_upload_wm_unit(struct brw_context *brw)
    /* _NEW_POLYGON */
    if (ctx->Polygon.OffsetFill) {
       wm->wm5.depth_offset = 1;
-      /* Something wierd going on with legacy_global_depth_bias,
+      /* Something weird going on with legacy_global_depth_bias,
        * offset_constant, scaling and MRD.  This value passes glean
        * but gives some odd results elsewere (eg. the
        * quad-offset-units test).
@@ -243,7 +250,7 @@ brw_upload_wm_unit(struct brw_context *brw)
                              I915_GEM_DOMAIN_INSTRUCTION, 0);
    }
 
-   brw->state.dirty.brw |= BRW_NEW_GEN4_UNIT_STATE;
+   brw->ctx.NewDriverState |= BRW_NEW_GEN4_UNIT_STATE;
 }
 
 const struct brw_tracked_state brw_wm_unit = {
@@ -256,10 +263,10 @@ const struct brw_tracked_state brw_wm_unit = {
       .brw = BRW_NEW_BATCH |
              BRW_NEW_CURBE_OFFSETS |
              BRW_NEW_FRAGMENT_PROGRAM |
+             BRW_NEW_FS_PROG_DATA |
              BRW_NEW_PROGRAM_CACHE |
+             BRW_NEW_SAMPLER_STATE_TABLE |
              BRW_NEW_STATS_WM,
-      .cache = CACHE_NEW_SAMPLER |
-               CACHE_NEW_WM_PROG,
    },
    .emit = brw_upload_wm_unit,
 };