glsl: use bitfield_insert instead of and + shift + or for packing

It is fairly tricky to detect the proper conditions for using bitfield
insert, but easy to just use it up front. This removes a lot of
instructions on nvc0 when invoking the packing builtins.

Signed-off-by: Ilia Mirkin <imirkin@alum.mit.edu>
Reviewed-by: Matt Turner <mattst88@gmail.com>

src/glsl/ir_optimization.h

@@ -66,7 +66,9 @@ enum lower_packing_builtins_op {
    LOWER_UNPACK_SNORM_4x8               = 0x0200,
 
    LOWER_PACK_UNORM_4x8                 = 0x0400,
-   LOWER_UNPACK_UNORM_4x8               = 0x0800
+   LOWER_UNPACK_UNORM_4x8               = 0x0800,
+
+   LOWER_PACK_USE_BFI                   = 0x1000,
 };
 
 bool do_common_optimization(exec_list *ir, bool linked,

src/glsl/lower_packing_builtins.cpp

@@ -118,6 +118,7 @@ public:
          *rvalue = split_unpack_half_2x16(op0);
          break;
       case LOWER_PACK_UNPACK_NONE:
+      case LOWER_PACK_USE_BFI:
          assert(!"not reached");
          break;
       }
@@ -222,9 +223,16 @@ private:
 
       /* uvec2 u = UVEC2_RVAL; */
       ir_variable *u = factory.make_temp(glsl_type::uvec2_type,
-                                          "tmp_pack_uvec2_to_uint");
+                                         "tmp_pack_uvec2_to_uint");
       factory.emit(assign(u, uvec2_rval));
 
+      if (op_mask & LOWER_PACK_USE_BFI) {
+         return bitfield_insert(bit_and(swizzle_x(u), constant(0xffffu)),
+                                swizzle_y(u),
+                                constant(16),
+                                constant(16));
+      }
+
       /* return (u.y << 16) | (u.x & 0xffff); */
       return bit_or(lshift(swizzle_y(u), constant(16u)),
                     bit_and(swizzle_x(u), constant(0xffffu)));
@@ -242,9 +250,22 @@ private:
    {
       assert(uvec4_rval->type == glsl_type::uvec4_type);
 
-      /* uvec4 u = UVEC4_RVAL; */
       ir_variable *u = factory.make_temp(glsl_type::uvec4_type,
-                                          "tmp_pack_uvec4_to_uint");
+                                         "tmp_pack_uvec4_to_uint");
+
+      if (op_mask & LOWER_PACK_USE_BFI) {
+         /* uvec4 u = UVEC4_RVAL; */
+         factory.emit(assign(u, uvec4_rval));
+
+         return bitfield_insert(bitfield_insert(
+                                   bitfield_insert(
+                                      bit_and(swizzle_x(u), constant(0xffu)),
+                                      swizzle_y(u), constant(8), constant(8)),
+                                   swizzle_z(u), constant(16), constant(8)),
+                                swizzle_w(u), constant(24), constant(8));
+      }
+
+      /* uvec4 u = UVEC4_RVAL & 0xff */
       factory.emit(assign(u, bit_and(uvec4_rval, constant(0xffu))));
 
       /* return (u.w << 24) | (u.z << 16) | (u.y << 8) | u.x; */

src/mesa/state_tracker/st_glsl_to_tgsi.cpp

@@ -6019,6 +6019,9 @@ st_link_shader(struct gl_context *ctx, struct gl_shader_program *prog)
                                LOWER_PACK_HALF_2x16 |
                                LOWER_UNPACK_HALF_2x16;
 
+         if (ctx->Extensions.ARB_gpu_shader5)
+            lower_inst |= LOWER_PACK_USE_BFI;
+
          lower_packing_builtins(ir, lower_inst);
       }