i965: Add a lowering pass to convert TXD to TXL by computing the LOD.

Intel hardware doesn't natively support textureGrad with shadow
comparisons.  So we need to generate code to handle it somehow.

Based on the equations of page 205 of the OpenGL 3.0 specification,
it's possible to compute the LOD value that would be selected given the
gradient values.  Then, we can simply convert the TXD to a TXL.

Currently, this passes 34/46 of oglconform's shadow-grad subtests;
four cubemap tests are regressed.  We should investigate this in the
future.

v2: Apply abs() to the scalar case (thanks to Eric).

Signed-off-by: Kenneth Graunke <kenneth@whitecape.org>
Reviewed-by: Eric Anholt <eric@anholt.net>

4 files changed, 157 insertions, 0 deletions

diff --git a/src/mesa/drivers/dri/i965/Makefile.sources b/src/mesa/drivers/dri/i965/Makefile.sources
index 82143ac5b54..334bfd92406 100644
--- a/src/mesa/drivers/dri/i965/Makefile.sources
+++ b/src/mesa/drivers/dri/i965/Makefile.sources
@@ -125,6 +125,7 @@ i965_CXX_FILES = \
 	brw_fs_reg_allocate.cpp \
 	brw_fs_schedule_instructions.cpp \
 	brw_fs_vector_splitting.cpp \
+	brw_lower_texture_gradients.cpp \
 	brw_shader.cpp \
 	brw_vec4.cpp \
 	brw_vec4_emit.cpp \
diff --git a/src/mesa/drivers/dri/i965/brw_context.h b/src/mesa/drivers/dri/i965/brw_context.h
index c1f2314f2f2..b4868fe3c26 100644
--- a/src/mesa/drivers/dri/i965/brw_context.h
+++ b/src/mesa/drivers/dri/i965/brw_context.h
@@ -1276,6 +1276,7 @@ brw_program_reloc(struct brw_context *brw, uint32_t state_offset,
 }
 
 bool brw_do_cubemap_normalize(struct exec_list *instructions);
+bool brw_lower_texture_gradients(struct exec_list *instructions);
 
 #ifdef __cplusplus
 }
diff --git a/src/mesa/drivers/dri/i965/brw_lower_texture_gradients.cpp b/src/mesa/drivers/dri/i965/brw_lower_texture_gradients.cpp
new file mode 100644
index 00000000000..dfbb983b726
--- /dev/null
+++ b/src/mesa/drivers/dri/i965/brw_lower_texture_gradients.cpp
@@ -0,0 +1,154 @@
+/*
+ * Copyright © 2012 Intel Corporation
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the next
+ * paragraph) shall be included in all copies or substantial portions of the
+ * Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL
+ * THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
+ * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING
+ * FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER
+ * DEALINGS IN THE SOFTWARE.
+ */
+
+/**
+ * \file brw_lower_texture_gradients.cpp
+ */
+
+#include "glsl/ir.h"
+#include "glsl/ir_builder.h"
+
+using namespace ir_builder;
+
+class lower_texture_grad_visitor : public ir_hierarchical_visitor {
+public:
+   lower_texture_grad_visitor()
+   {
+      progress = false;
+   }
+
+   ir_visitor_status visit_leave(ir_texture *ir);
+
+
+   bool progress;
+
+private:
+   void emit(ir_variable *, ir_rvalue *);
+};
+
+/**
+ * Emit a variable declaration and an assignment to initialize it.
+ */
+void
+lower_texture_grad_visitor::emit(ir_variable *var, ir_rvalue *value)
+{
+   base_ir->insert_before(var);
+   base_ir->insert_before(assign(var, value));
+}
+
+static const glsl_type *
+txs_type(const glsl_type *type)
+{
+   unsigned dims;
+   switch (type->sampler_dimensionality) {
+   case GLSL_SAMPLER_DIM_1D:
+      dims = 1;
+      break;
+   case GLSL_SAMPLER_DIM_2D:
+   case GLSL_SAMPLER_DIM_RECT:
+   case GLSL_SAMPLER_DIM_CUBE:
+      dims = 2;
+      break;
+   case GLSL_SAMPLER_DIM_3D:
+      dims = 3;
+      break;
+   default:
+      assert(!"Should not get here: invalid sampler dimensionality");
+   }
+
+   if (type->sampler_array)
+      dims++;
+
+   return glsl_type::get_instance(GLSL_TYPE_INT, dims, 1);
+}
+
+ir_visitor_status
+lower_texture_grad_visitor::visit_leave(ir_texture *ir)
+{
+   /* Only lower textureGrad with shadow samplers */
+   if (ir->op != ir_txd || !ir->shadow_comparitor)
+      return visit_continue;
+
+   /* Cubes are broken.  Avoid assertion failures when swizzling. */
+   if (ir->sampler->type->sampler_dimensionality == GLSL_SAMPLER_DIM_CUBE)
+      return visit_continue;
+
+   void *mem_ctx = ralloc_parent(ir);
+
+   const glsl_type *grad_type = ir->lod_info.grad.dPdx->type;
+
+   /* Use textureSize() to get the width and height of LOD 0; swizzle away
+    * the depth/number of array slices.
+    */
+   ir_texture *txs = new(mem_ctx) ir_texture(ir_txs);
+   txs->set_sampler(ir->sampler->clone(mem_ctx, NULL),
+		    txs_type(ir->sampler->type));
+   txs->lod_info.lod = new(mem_ctx) ir_constant(0);
+   ir_variable *size =
+      new(mem_ctx) ir_variable(grad_type, "size", ir_var_temporary);
+   emit(size, expr(ir_unop_i2f,
+		   swizzle_for_size(txs, grad_type->vector_elements)));
+
+   /* Scale the gradients by width and height.  Effectively, the incoming
+    * gradients are s'(x,y), t'(x,y), and r'(x,y) from equation 3.19 in the
+    * GL 3.0 spec; we want u'(x,y), which is w_t * s'(x,y).
+    */
+   ir_variable *dPdx =
+      new(mem_ctx) ir_variable(grad_type, "dPdx", ir_var_temporary);
+   emit(dPdx, mul(size, ir->lod_info.grad.dPdx));
+
+   ir_variable *dPdy =
+      new(mem_ctx) ir_variable(grad_type, "dPdy", ir_var_temporary);
+   emit(dPdy, mul(size, ir->lod_info.grad.dPdy));
+
+   /* Calculate rho from equation 3.20 of the GL 3.0 specification. */
+   ir_rvalue *rho;
+   if (dPdx->type->is_scalar()) {
+      rho = expr(ir_binop_max, expr(ir_unop_abs, dPdx),
+			       expr(ir_unop_abs, dPdy));
+   } else {
+      rho = expr(ir_binop_max, expr(ir_unop_sqrt, dot(dPdx, dPdx)),
+			       expr(ir_unop_sqrt, dot(dPdy, dPdy)));
+   }
+
+   /* lambda_base = log2(rho).  We're ignoring GL state biases for now. */
+   ir->op = ir_txl;
+   ir->lod_info.lod = expr(ir_unop_log2, rho);
+
+   progress = true;
+   return visit_continue;
+}
+
+extern "C" {
+
+bool
+brw_lower_texture_gradients(struct exec_list *instructions)
+{
+   lower_texture_grad_visitor v;
+
+   visit_list_elements(&v, instructions);
+
+   return v.progress;
+}
+
+}
diff --git a/src/mesa/drivers/dri/i965/brw_shader.cpp b/src/mesa/drivers/dri/i965/brw_shader.cpp
index aa1bfdbd230..140a268cff7 100644
--- a/src/mesa/drivers/dri/i965/brw_shader.cpp
+++ b/src/mesa/drivers/dri/i965/brw_shader.cpp
@@ -157,6 +157,7 @@ brw_link_shader(struct gl_context *ctx, struct gl_shader_program *shProg)
 	 lower_if_to_cond_assign(shader->ir, 16);
 
       do_lower_texture_projection(shader->ir);
+      brw_lower_texture_gradients(shader->ir);
       do_vec_index_to_cond_assign(shader->ir);
       brw_do_cubemap_normalize(shader->ir);
       lower_noise(shader->ir);