i965/gen6+: Merge VS/GS and WM push constant buffer upload paths.

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

4 files changed, 52 insertions, 66 deletions

diff --git a/src/mesa/drivers/dri/i965/brw_context.h b/src/mesa/drivers/dri/i965/brw_context.h
index 7aa9484..2943a20 100644
--- a/src/mesa/drivers/dri/i965/brw_context.h
+++ b/src/mesa/drivers/dri/i965/brw_context.h
@@ -1780,11 +1780,11 @@ brw_setup_vec4_key_clip_info(struct brw_context *brw,
                              bool program_uses_clip_distance);
 
 void
-gen6_upload_vec4_push_constants(struct brw_context *brw,
-                                const struct gl_program *prog,
-                                const struct brw_vec4_prog_data *prog_data,
-                                struct brw_stage_state *stage_state,
-                                enum aub_state_struct_type type);
+gen6_upload_push_constants(struct brw_context *brw,
+                           const struct gl_program *prog,
+                           const struct brw_stage_prog_data *prog_data,
+                           struct brw_stage_state *stage_state,
+                           enum aub_state_struct_type type);
 
 /* ================================================================
  * From linux kernel i386 header files, copes with odd sizes better
diff --git a/src/mesa/drivers/dri/i965/gen6_vs_state.c b/src/mesa/drivers/dri/i965/gen6_vs_state.c
index 9d3c184..407818e 100644
--- a/src/mesa/drivers/dri/i965/gen6_vs_state.c
+++ b/src/mesa/drivers/dri/i965/gen6_vs_state.c
@@ -32,13 +32,29 @@
 #include "program/prog_parameter.h"
 #include "program/prog_statevars.h"
 #include "intel_batchbuffer.h"
+#include "glsl/glsl_parser_extras.h"
 
+/**
+ * Creates a streamed BO containing the push constants for the VS or GS on
+ * gen6+.
+ *
+ * Push constants are constant values (such as GLSL uniforms) that are
+ * pre-loaded into a shader stage's register space at thread spawn time.
+ *
+ * Not all GLSL uniforms will be uploaded as push constants: The hardware has
+ * a limitation of 32 or 64 EU registers (256 or 512 floats) per stage to be
+ * uploaded as push constants, while GL 4.4 requires at least 1024 components
+ * to be usable for the VS.  Plus, currently we always use pull constants
+ * instead of push constants when doing variable-index array access.
+ *
+ * See brw_curbe.c for the equivalent gen4/5 code.
+ */
 void
-gen6_upload_vec4_push_constants(struct brw_context *brw,
-                                const struct gl_program *prog,
-                                const struct brw_vec4_prog_data *prog_data,
-                                struct brw_stage_state *stage_state,
-                                enum aub_state_struct_type type)
+gen6_upload_push_constants(struct brw_context *brw,
+                           const struct gl_program *prog,
+                           const struct brw_stage_prog_data *prog_data,
+                           struct brw_stage_state *stage_state,
+                           enum aub_state_struct_type type)
 {
    struct gl_context *ctx = &brw->ctx;
 
@@ -48,15 +64,14 @@ gen6_upload_vec4_push_constants(struct brw_context *brw,
    /* XXX: Should this happen somewhere before to get our state flag set? */
    _mesa_load_state_parameters(ctx, prog->Parameters);
 
-   if (prog_data->base.nr_params == 0) {
+   if (prog_data->nr_params == 0) {
       stage_state->push_const_size = 0;
    } else {
-      int params_uploaded;
       float *param;
       int i;
 
       param = brw_state_batch(brw, type,
-			      prog_data->base.nr_params * sizeof(float),
+			      prog_data->nr_params * sizeof(float),
 			      32, &stage_state->push_const_offset);
 
       /* _NEW_PROGRAM_CONSTANTS
@@ -65,21 +80,27 @@ gen6_upload_vec4_push_constants(struct brw_context *brw,
        * side effect of dereferencing uniforms, so _NEW_PROGRAM_CONSTANTS
        * wouldn't be set for them.
       */
-      for (i = 0; i < prog_data->base.nr_params; i++) {
-         param[i] = *prog_data->base.param[i];
+      for (i = 0; i < prog_data->nr_params; i++) {
+         param[i] = *prog_data->param[i];
       }
-      params_uploaded = prog_data->base.nr_params / 4;
 
       if (0) {
-	 fprintf(stderr, "Constant buffer:\n");
-	 for (i = 0; i < params_uploaded; i++) {
-	    float *buf = param + i * 4;
-	    fprintf(stderr, "%d: %f %f %f %f\n",
-                    i, buf[0], buf[1], buf[2], buf[3]);
+	 fprintf(stderr, "%s constants:\n",
+                 _mesa_shader_stage_to_string(stage_state->stage));
+	 for (i = 0; i < prog_data->nr_params; i++) {
+	    if ((i & 7) == 0)
+	       fprintf(stderr, "g%d: ",
+                       prog_data->dispatch_grf_start_reg + i / 8);
+	    fprintf(stderr, "%8f ", param[i]);
+	    if ((i & 7) == 7)
+	       fprintf(stderr, "\n");
 	 }
+	 if ((i & 7) != 0)
+	    fprintf(stderr, "\n");
+	 fprintf(stderr, "\n");
       }
 
-      stage_state->push_const_size = (params_uploaded + 1) / 2;
+      stage_state->push_const_size = ALIGN(prog_data->nr_params, 8) / 8;
       /* We can only push 32 registers of constants at a time. */
       assert(stage_state->push_const_size <= 32);
    }
@@ -94,10 +115,10 @@ gen6_upload_vs_push_constants(struct brw_context *brw)
    const struct brw_vertex_program *vp =
       brw_vertex_program_const(brw->vertex_program);
    /* CACHE_NEW_VS_PROG */
-   const struct brw_vec4_prog_data *prog_data = &brw->vs.prog_data->base;
+   const struct brw_stage_prog_data *prog_data = &brw->vs.prog_data->base.base;
 
-   gen6_upload_vec4_push_constants(brw, &vp->program.Base, prog_data,
-                                   stage_state, AUB_TRACE_VS_CONSTANTS);
+   gen6_upload_push_constants(brw, &vp->program.Base, prog_data,
+                              stage_state, AUB_TRACE_VS_CONSTANTS);
 
    if (brw->gen >= 7) {
       if (brw->gen == 7 && !brw->is_haswell && !brw->is_baytrail)
diff --git a/src/mesa/drivers/dri/i965/gen6_wm_state.c b/src/mesa/drivers/dri/i965/gen6_wm_state.c
index 042b47d..db0e3aa 100644
--- a/src/mesa/drivers/dri/i965/gen6_wm_state.c
+++ b/src/mesa/drivers/dri/i965/gen6_wm_state.c
@@ -38,50 +38,15 @@
 static void
 gen6_upload_wm_push_constants(struct brw_context *brw)
 {
-   struct gl_context *ctx = &brw->ctx;
+   struct brw_stage_state *stage_state = &brw->wm.base;
    /* BRW_NEW_FRAGMENT_PROGRAM */
    const struct brw_fragment_program *fp =
       brw_fragment_program_const(brw->fragment_program);
    /* CACHE_NEW_WM_PROG */
    const struct brw_wm_prog_data *prog_data = brw->wm.prog_data;
 
-   /* Updates the ParameterValues[i] pointers for all parameters of the
-    * basic type of PROGRAM_STATE_VAR.
-    */
-   /* XXX: Should this happen somewhere before to get our state flag set? */
-   _mesa_load_state_parameters(ctx, fp->program.Base.Parameters);
-
-   if (prog_data->base.nr_params == 0) {
-      brw->wm.base.push_const_size = 0;
-   } else {
-      float *constants;
-      unsigned int i;
-
-      constants = brw_state_batch(brw, AUB_TRACE_WM_CONSTANTS,
-				  prog_data->base.nr_params * sizeof(float),
-				  32, &brw->wm.base.push_const_offset);
-
-      for (i = 0; i < prog_data->base.nr_params; i++) {
-	 constants[i] = *prog_data->base.param[i];
-      }
-
-      if (0) {
-	 fprintf(stderr, "WM constants:\n");
-	 for (i = 0; i < prog_data->base.nr_params; i++) {
-	    if ((i & 7) == 0)
-	       fprintf(stderr, "g%d: ",
-                       prog_data->base.dispatch_grf_start_reg + i / 8);
-	    fprintf(stderr, "%8f ", constants[i]);
-	    if ((i & 7) == 7)
-	       fprintf(stderr, "\n");
-	 }
-	 if ((i & 7) != 0)
-	    fprintf(stderr, "\n");
-	 fprintf(stderr, "\n");
-      }
-
-      brw->wm.base.push_const_size = ALIGN(prog_data->base.nr_params, 8) / 8;
-   }
+   gen6_upload_push_constants(brw, &fp->program.Base, &prog_data->base,
+                              stage_state, AUB_TRACE_WM_CONSTANTS);
 
    if (brw->gen >= 7) {
       gen7_upload_constant_state(brw, &brw->wm.base, true,
diff --git a/src/mesa/drivers/dri/i965/gen7_gs_state.c b/src/mesa/drivers/dri/i965/gen7_gs_state.c
index 5f66bc6..93f48f6 100644
--- a/src/mesa/drivers/dri/i965/gen7_gs_state.c
+++ b/src/mesa/drivers/dri/i965/gen7_gs_state.c
@@ -37,11 +37,11 @@ gen7_upload_gs_push_constants(struct brw_context *brw)
 
    if (gp) {
       /* CACHE_NEW_GS_PROG */
-      const struct brw_vec4_prog_data *prog_data = &brw->gs.prog_data->base;
+      const struct brw_stage_prog_data *prog_data = &brw->gs.prog_data->base.base;
       struct brw_stage_state *stage_state = &brw->gs.base;
 
-      gen6_upload_vec4_push_constants(brw, &gp->program.Base, prog_data,
-                                      stage_state, AUB_TRACE_VS_CONSTANTS);
+      gen6_upload_push_constants(brw, &gp->program.Base, prog_data,
+                                 stage_state, AUB_TRACE_VS_CONSTANTS);
    }
 
    gen7_upload_constant_state(brw, stage_state, gp, _3DSTATE_CONSTANT_GS);