i965/fs: Add support for bit instructions.

Don't bother scalarizing ir_binop_bfm, since its results are
identical for all channels.

v2: Subtract result of FBH from 31 (unless an error) to convert
    MSB counts to LSB counts.
v3: Use op0->clone() in ir_triop_bfi to prevent (var_ref
    channel_expressions) from appearing multiple times in the IR.

Reviewed-by: Chris Forbes <chrisf@ijw.co.nz> [v2]

1 files changed, 37 insertions, 0 deletions

diff --git a/src/mesa/drivers/dri/i965/brw_fs_channel_expressions.cpp b/src/mesa/drivers/dri/i965/brw_fs_channel_expressions.cpp
index 30d8d9b..0f3d4ab 100644
--- a/src/mesa/drivers/dri/i965/brw_fs_channel_expressions.cpp
+++ b/src/mesa/drivers/dri/i965/brw_fs_channel_expressions.cpp
@@ -216,6 +216,10 @@ ir_channel_expressions_visitor::visit_leave(ir_assignment *ir)
    case ir_unop_cos_reduced:
    case ir_unop_dFdx:
    case ir_unop_dFdy:
+   case ir_unop_bitfield_reverse:
+   case ir_unop_bit_count:
+   case ir_unop_find_msb:
+   case ir_unop_find_lsb:
       for (i = 0; i < vector_elements; i++) {
 	 ir_rvalue *op0 = get_element(op_var[0], i);
 
@@ -338,11 +342,26 @@ ir_channel_expressions_visitor::visit_leave(ir_assignment *ir)
       assert(!"noise should have been broken down to function call");
       break;
 
+   case ir_binop_bfm: {
+      /* Does not need to be scalarized, since its result will be identical
+       * for all channels.
+       */
+      ir_rvalue *op0 = get_element(op_var[0], 0);
+      ir_rvalue *op1 = get_element(op_var[1], 0);
+
+      assign(ir, 0, new(mem_ctx) ir_expression(expr->operation,
+                                               element_type,
+                                               op0,
+                                               op1));
+      break;
+   }
+
    case ir_binop_ubo_load:
       assert(!"not yet supported");
       break;
 
    case ir_triop_lrp:
+   case ir_triop_bitfield_extract:
       for (i = 0; i < vector_elements; i++) {
 	 ir_rvalue *op0 = get_element(op_var[0], i);
 	 ir_rvalue *op1 = get_element(op_var[1], i);
@@ -356,6 +375,23 @@ ir_channel_expressions_visitor::visit_leave(ir_assignment *ir)
       }
       break;
 
+   case ir_triop_bfi: {
+      /* Only a single BFM is needed for multiple BFIs. */
+      ir_rvalue *op0 = get_element(op_var[0], 0);
+
+      for (i = 0; i < vector_elements; i++) {
+         ir_rvalue *op1 = get_element(op_var[1], i);
+         ir_rvalue *op2 = get_element(op_var[2], i);
+
+         assign(ir, i, new(mem_ctx) ir_expression(expr->operation,
+                                                  element_type,
+                                                  op0->clone(mem_ctx, NULL),
+                                                  op1,
+                                                  op2));
+      }
+      break;
+   }
+
    case ir_unop_pack_snorm_2x16:
    case ir_unop_pack_snorm_4x8:
    case ir_unop_pack_unorm_2x16:
@@ -366,6 +402,7 @@ ir_channel_expressions_visitor::visit_leave(ir_assignment *ir)
    case ir_unop_unpack_unorm_2x16:
    case ir_unop_unpack_unorm_4x8:
    case ir_unop_unpack_half_2x16:
+   case ir_quadop_bitfield_insert:
    case ir_quadop_vector:
       assert(!"should have been lowered");
       break;