Mesa (master): i965: updated CURBE allocation code

[Brian Paul]

Module: Mesa
Branch: master
Commit: f8f23e33c21f41756d068b546f2aae37030b5773
URL:    http://cgit.freedesktop.org/mesa/mesa/commit/?id=f8f23e33c21f41756d068b546f2aae37030b5773

Author: Brian Paul <brianp at vmware.com>
Date:   Thu Apr 16 15:43:32 2009 -0600

i965: updated CURBE allocation code

Now that we have real constant buffers, the demands on the CURBE are lessened.
When we use real VS/WM constant buffers we only use the CURBE for clip planes.

---

 src/mesa/drivers/dri/i965/brw_context.h |    3 ++-
 src/mesa/drivers/dri/i965/brw_curbe.c   |   14 +++++++++-----
 src/mesa/drivers/dri/i965/brw_vs_emit.c |    6 ++++--
 3 files changed, 15 insertions(+), 8 deletions(-)

diff --git a/src/mesa/drivers/dri/i965/brw_context.h b/src/mesa/drivers/dri/i965/brw_context.h
index 4c2d3af..a0b3b06 100644
--- a/src/mesa/drivers/dri/i965/brw_context.h
+++ b/src/mesa/drivers/dri/i965/brw_context.h
@@ -187,7 +187,7 @@ struct brw_wm_prog_data {
    GLuint total_grf;
    GLuint total_scratch;
 
-   GLuint nr_params;
+   GLuint nr_params;       /**< number of float params/constants */
    GLboolean error;
 
    /* Pointer to tracked values (only valid once
@@ -226,6 +226,7 @@ struct brw_vs_prog_data {
    GLuint urb_read_length;
    GLuint total_grf;
    GLuint outputs_written;
+   GLuint nr_params;       /**< number of float params/constants */
 
    GLuint inputs_read;
 
diff --git a/src/mesa/drivers/dri/i965/brw_curbe.c b/src/mesa/drivers/dri/i965/brw_curbe.c
index fe1de82..18b187e 100644
--- a/src/mesa/drivers/dri/i965/brw_curbe.c
+++ b/src/mesa/drivers/dri/i965/brw_curbe.c
@@ -45,17 +45,21 @@
 #include "brw_util.h"
 
 
-/* Partition the CURBE between the various users of constant values:
+/**
+ * Partition the CURBE between the various users of constant values:
+ * Note that vertex and fragment shaders can now fetch constants out
+ * of constant buffers.  We no longer allocatea block of the GRF for
+ * constants.  That greatly reduces the demand for space in the CURBE.
+ * Some of the comments within are dated...
  */
 static void calculate_curbe_offsets( struct brw_context *brw )
 {
    GLcontext *ctx = &brw->intel.ctx;
    /* CACHE_NEW_WM_PROG */
-   GLuint nr_fp_regs = (brw->wm.prog_data->nr_params + 15) / 16;
+   const GLuint nr_fp_regs = (brw->wm.prog_data->nr_params + 15) / 16;
    
    /* BRW_NEW_VERTEX_PROGRAM */
-   const struct brw_vertex_program *vp = brw_vertex_program_const(brw->vertex_program);
-   GLuint nr_vp_regs = (vp->program.Base.Parameters->NumParameters * 4 + 15) / 16;
+   const GLuint nr_vp_regs = (brw->vs.prog_data->nr_params + 15) / 16;
    GLuint nr_clip_regs = 0;
    GLuint total_regs;
 
@@ -248,7 +252,7 @@ static void prepare_constant_buffer(struct brw_context *brw)
    /* vertex shader constants */
    if (brw->curbe.vs_size) {
       GLuint offset = brw->curbe.vs_start * 16;
-      GLuint nr = vp->program.Base.Parameters->NumParameters;
+      GLuint nr = brw->vs.prog_data->nr_params / 4;
 
       _mesa_load_state_parameters(ctx, vp->program.Base.Parameters); 
 
diff --git a/src/mesa/drivers/dri/i965/brw_vs_emit.c b/src/mesa/drivers/dri/i965/brw_vs_emit.c
index af0f501..0c86c23 100644
--- a/src/mesa/drivers/dri/i965/brw_vs_emit.c
+++ b/src/mesa/drivers/dri/i965/brw_vs_emit.c
@@ -68,7 +68,6 @@ static void release_tmps( struct brw_vs_compile *c )
 static void brw_vs_alloc_regs( struct brw_vs_compile *c )
 {
    GLuint i, reg = 0, mrf;
-   GLuint nr_params;
 
 #if 0
    if (c->vp->program.Base.Parameters->NumParameters >= 6)
@@ -100,15 +99,18 @@ static void brw_vs_alloc_regs( struct brw_vs_compile *c )
    if (c->use_const_buffer) {
       /* get constants from a real constant buffer */
       c->prog_data.curb_read_length = 0;
+      c->prog_data.nr_params = 4; /* XXX 0 causes a bug elsewhere... */
    }
    else {
       /* use a section of the GRF for constants */
-      nr_params = c->vp->program.Base.Parameters->NumParameters;
+      GLuint nr_params = c->vp->program.Base.Parameters->NumParameters;
       for (i = 0; i < nr_params; i++) {
          c->regs[PROGRAM_STATE_VAR][i] = stride( brw_vec4_grf(reg+i/2, (i%2) * 4), 0, 4, 1);
       }
       reg += (nr_params + 1) / 2;
       c->prog_data.curb_read_length = reg - 1;
+
+      c->prog_data.nr_params = nr_params * 4;
    }
 
    /* Allocate input regs: