Mesa (master): i965/fs: Implement texelFetch() on Gen4.

[Kenneth Graunke]

Module: Mesa
Branch: master
Commit: 47b556fbcaea4660b21481e40d89167d883d47f5
URL:    http://cgit.freedesktop.org/mesa/mesa/commit/?id=47b556fbcaea4660b21481e40d89167d883d47f5

Author: Kenneth Graunke <kenneth at whitecape.org>
Date:   Tue Sep  6 16:39:01 2011 -0700

i965/fs: Implement texelFetch() on Gen4.

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

---

 src/mesa/drivers/dri/i965/brw_fs_emit.cpp    |    5 +++++
 src/mesa/drivers/dri/i965/brw_fs_visitor.cpp |   17 +++++++++++++----
 2 files changed, 18 insertions(+), 4 deletions(-)

diff --git a/src/mesa/drivers/dri/i965/brw_fs_emit.cpp b/src/mesa/drivers/dri/i965/brw_fs_emit.cpp
index 906c158..f742e84 100644
--- a/src/mesa/drivers/dri/i965/brw_fs_emit.cpp
+++ b/src/mesa/drivers/dri/i965/brw_fs_emit.cpp
@@ -295,6 +295,11 @@ fs_visitor::generate_tex(fs_inst *inst, struct brw_reg dst, struct brw_reg src)
 	 assert(inst->mlen == 7 || inst->mlen == 10);
 	 msg_type = BRW_SAMPLER_MESSAGE_SIMD8_SAMPLE_GRADIENTS;
 	 break;
+      case FS_OPCODE_TXF:
+	 assert(inst->mlen == 9);
+	 msg_type = BRW_SAMPLER_MESSAGE_SIMD16_LD;
+	 simd_mode = BRW_SAMPLER_SIMD_MODE_SIMD16;
+	 break;
       case FS_OPCODE_TXS:
 	 assert(inst->mlen == 3);
 	 msg_type = BRW_SAMPLER_MESSAGE_SIMD16_RESINFO;
diff --git a/src/mesa/drivers/dri/i965/brw_fs_visitor.cpp b/src/mesa/drivers/dri/i965/brw_fs_visitor.cpp
index 2c8d160..3604231 100644
--- a/src/mesa/drivers/dri/i965/brw_fs_visitor.cpp
+++ b/src/mesa/drivers/dri/i965/brw_fs_visitor.cpp
@@ -678,17 +678,25 @@ fs_visitor::emit_texture_gen4(ir_texture *ir, fs_reg dst, fs_reg coordinate,
        * instructions.  We'll need to do SIMD16 here.
        */
       simd16 = true;
-      assert(ir->op == ir_txb || ir->op == ir_txl);
+      assert(ir->op == ir_txb || ir->op == ir_txl || ir->op == ir_txf);
 
       for (int i = 0; i < ir->coordinate->type->vector_elements; i++) {
 	 fs_inst *inst = emit(BRW_OPCODE_MOV, fs_reg(MRF,
-						     base_mrf + mlen + i * 2),
+						     base_mrf + mlen + i * 2,
+						     coordinate.type),
 			      coordinate);
 	 if (i < 3 && c->key.gl_clamp_mask[i] & (1 << sampler))
 	    inst->saturate = true;
 	 coordinate.reg_offset++;
       }
 
+      /* Initialize the rest of u/v/r with 0.0.  Empirically, this seems to
+       * be necessary for TXF (ld), but seems wise to do for all messages.
+       */
+      for (int i = ir->coordinate->type->vector_elements; i < 3; i++) {
+	 emit(BRW_OPCODE_MOV, fs_reg(MRF, base_mrf + mlen + i * 2), fs_reg(0.0f));
+      }
+
       /* lod/bias appears after u/v/r. */
       mlen += 6;
 
@@ -698,7 +706,8 @@ fs_visitor::emit_texture_gen4(ir_texture *ir, fs_reg dst, fs_reg coordinate,
 	 mlen++;
       } else {
 	 ir->lod_info.lod->accept(this);
-	 emit(BRW_OPCODE_MOV, fs_reg(MRF, base_mrf + mlen), this->result);
+	 emit(BRW_OPCODE_MOV, fs_reg(MRF, base_mrf + mlen, this->result.type),
+			      this->result);
 	 mlen++;
       }
 
@@ -737,7 +746,7 @@ fs_visitor::emit_texture_gen4(ir_texture *ir, fs_reg dst, fs_reg coordinate,
       inst = emit(FS_OPCODE_TXS, dst);
       break;
    case ir_txf:
-      assert(!"GLSL 1.30 features unsupported");
+      inst = emit(FS_OPCODE_TXF, dst);
       break;
    }
    inst->base_mrf = base_mrf;