i965: Remove never used RSR and RSL opcodes.
[mesa.git] / src / mesa / drivers / dri / i965 / gen6_vs_state.c
index ae1a8412a659bb9eb94d84b4e434dce93caa64c1..569ec8cdc891ef9084c6b34c6eb283ae805af53c 100644 (file)
 #include "program/prog_statevars.h"
 #include "intel_batchbuffer.h"
 
-static void
-gen6_upload_vs_push_constants(struct brw_context *brw)
+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 state_struct_type type)
 {
-   struct intel_context *intel = &brw->intel;
-   struct gl_context *ctx = &intel->ctx;
-   /* _BRW_NEW_VERTEX_PROGRAM */
-   const struct brw_vertex_program *vp =
-      brw_vertex_program_const(brw->vertex_program);
+   struct gl_context *ctx = &brw->ctx;
 
    /* Updates the ParamaterValues[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, vp->program.Base.Parameters);
+   _mesa_load_state_parameters(ctx, prog->Parameters);
 
-   /* CACHE_NEW_VS_PROG */
-   if (brw->vs.prog_data->base.nr_params == 0) {
-      brw->vs.push_const_size = 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, AUB_TRACE_VS_CONSTANTS,
-                             brw->vs.prog_data->base.nr_params * sizeof(float),
-                             32, &brw->vs.push_const_offset);
+      param = brw_state_batch(brw, type,
+                             prog_data->nr_params * sizeof(float),
+                             32, &stage_state->push_const_offset);
 
       /* _NEW_PROGRAM_CONSTANTS
        *
@@ -66,13 +65,13 @@ gen6_upload_vs_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 < brw->vs.prog_data->base.nr_params; i++) {
-         param[i] = *brw->vs.prog_data->base.param[i];
+      for (i = 0; i < prog_data->nr_params; i++) {
+         param[i] = *prog_data->param[i];
       }
-      params_uploaded = brw->vs.prog_data->base.nr_params / 4;
+      params_uploaded = prog_data->nr_params / 4;
 
       if (0) {
-        printf("VS constant buffer:\n");
+        printf("Constant buffer:\n");
         for (i = 0; i < params_uploaded; i++) {
            float *buf = param + i * 4;
            printf("%d: %f %f %f %f\n",
@@ -80,12 +79,27 @@ gen6_upload_vs_push_constants(struct brw_context *brw)
         }
       }
 
-      brw->vs.push_const_size = (params_uploaded + 1) / 2;
+      stage_state->push_const_size = (params_uploaded + 1) / 2;
       /* We can only push 32 registers of constants at a time. */
-      assert(brw->vs.push_const_size <= 32);
+      assert(stage_state->push_const_size <= 32);
    }
 }
 
+static void
+gen6_upload_vs_push_constants(struct brw_context *brw)
+{
+   struct brw_stage_state *stage_state = &brw->vs.base;
+
+   /* _BRW_NEW_VERTEX_PROGRAM */
+   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;
+
+   gen6_upload_vec4_push_constants(brw, &vp->program.Base, prog_data,
+                                   stage_state, AUB_TRACE_VS_CONSTANTS);
+}
+
 const struct brw_tracked_state gen6_vs_push_constants = {
    .dirty = {
       .mesa  = _NEW_TRANSFORM | _NEW_PROGRAM_CONSTANTS,
@@ -99,19 +113,24 @@ const struct brw_tracked_state gen6_vs_push_constants = {
 static void
 upload_vs_state(struct brw_context *brw)
 {
-   struct intel_context *intel = &brw->intel;
+   struct gl_context *ctx = &brw->ctx;
+   const struct brw_stage_state *stage_state = &brw->vs.base;
    uint32_t floating_point_mode = 0;
 
-   /* From the BSpec, Volume 2a, Part 3 "Vertex Shader", Section
+   /* From the BSpec, 3D Pipeline > Geometry > Vertex Shader > State,
     * 3DSTATE_VS, Dword 5.0 "VS Function Enable":
+    *
     *   [DevSNB] A pipeline flush must be programmed prior to a 3DSTATE_VS
     *   command that causes the VS Function Enable to toggle. Pipeline
     *   flush can be executed by sending a PIPE_CONTROL command with CS
     *   stall bit set and a post sync operation.
+    *
+    * Although we don't disable the VS during normal drawing, BLORP sometimes
+    * disables it.  To be safe, do the flush here just in case.
     */
-   intel_emit_post_sync_nonzero_flush(intel);
+   intel_emit_post_sync_nonzero_flush(brw);
 
-   if (brw->vs.push_const_size == 0) {
+   if (stage_state->push_const_size == 0) {
       /* Disable the push constant buffers. */
       BEGIN_BATCH(5);
       OUT_BATCH(_3DSTATE_CONSTANT_VS << 16 | (5 - 2));
@@ -128,8 +147,8 @@ upload_vs_state(struct brw_context *brw)
       /* Pointer to the VS constant buffer.  Covered by the set of
        * state flags from gen6_upload_vs_constants
        */
-      OUT_BATCH(brw->vs.push_const_offset +
-               brw->vs.push_const_size - 1);
+      OUT_BATCH(stage_state->push_const_offset +
+               stage_state->push_const_size - 1);
       OUT_BATCH(0);
       OUT_BATCH(0);
       OUT_BATCH(0);
@@ -139,24 +158,25 @@ upload_vs_state(struct brw_context *brw)
    /* Use ALT floating point mode for ARB vertex programs, because they
     * require 0^0 == 1.
     */
-   if (intel->ctx.Shader.CurrentVertexProgram == NULL)
+   if (ctx->Shader.CurrentVertexProgram == NULL)
       floating_point_mode = GEN6_VS_FLOATING_POINT_MODE_ALT;
 
    BEGIN_BATCH(6);
    OUT_BATCH(_3DSTATE_VS << 16 | (6 - 2));
-   OUT_BATCH(brw->vs.prog_offset);
+   OUT_BATCH(stage_state->prog_offset);
    OUT_BATCH(floating_point_mode |
-            ((ALIGN(brw->sampler.count, 4)/4) << GEN6_VS_SAMPLER_COUNT_SHIFT));
+            ((ALIGN(stage_state->sampler_count, 4)/4) << GEN6_VS_SAMPLER_COUNT_SHIFT));
 
    if (brw->vs.prog_data->base.total_scratch) {
-      OUT_RELOC(brw->vs.scratch_bo,
+      OUT_RELOC(stage_state->scratch_bo,
                I915_GEM_DOMAIN_RENDER, I915_GEM_DOMAIN_RENDER,
                ffs(brw->vs.prog_data->base.total_scratch) - 11);
    } else {
       OUT_BATCH(0);
    }
 
-   OUT_BATCH((1 << GEN6_VS_DISPATCH_START_GRF_SHIFT) |
+   OUT_BATCH((brw->vs.prog_data->base.dispatch_grf_start_reg <<
+              GEN6_VS_DISPATCH_START_GRF_SHIFT) |
             (brw->vs.prog_data->base.urb_read_length << GEN6_VS_URB_READ_LENGTH_SHIFT) |
             (0 << GEN6_VS_URB_ENTRY_READ_OFFSET_SHIFT));
 
@@ -182,7 +202,7 @@ upload_vs_state(struct brw_context *brw)
     * bug reports that led to this workaround, and may be more than
     * what is strictly required to avoid the issue.
     */
-   intel_emit_post_sync_nonzero_flush(intel);
+   intel_emit_post_sync_nonzero_flush(brw);
 
    BEGIN_BATCH(4);
    OUT_BATCH(_3DSTATE_PIPE_CONTROL | (4 - 2));
@@ -199,7 +219,8 @@ const struct brw_tracked_state gen6_vs_state = {
       .mesa  = _NEW_TRANSFORM | _NEW_PROGRAM_CONSTANTS,
       .brw   = (BRW_NEW_CONTEXT |
                BRW_NEW_VERTEX_PROGRAM |
-               BRW_NEW_BATCH),
+               BRW_NEW_BATCH |
+                BRW_NEW_PUSH_CONSTANT_ALLOCATION),
       .cache = CACHE_NEW_VS_PROG | CACHE_NEW_SAMPLER
    },
    .emit = upload_vs_state,