[Mesa-dev] [PATCH 7/7] radeon/llvm: use trunc intrinsic instead of llvm.AMDGPU.trunc

Vincent Lejeune vljn at ovi.com
Mon Oct 8 07:47:13 PDT 2012


---
 src/gallium/drivers/radeon/AMDGPUISelLowering.cpp   | 3 +--
 src/gallium/drivers/radeon/AMDGPUIntrinsics.td      | 1 -
 src/gallium/drivers/radeon/AMDILIntrinsics.td       | 2 --
 src/gallium/drivers/radeon/R600Instructions.td      | 2 +-
 src/gallium/drivers/radeon/radeon_setup_tgsi_llvm.c | 8 ++++----
 5 files changed, 6 insertions(+), 10 deletions(-)

diff --git a/src/gallium/drivers/radeon/AMDGPUISelLowering.cpp b/src/gallium/drivers/radeon/AMDGPUISelLowering.cpp
index d37df6b..6c69c0b 100644
--- a/src/gallium/drivers/radeon/AMDGPUISelLowering.cpp
+++ b/src/gallium/drivers/radeon/AMDGPUISelLowering.cpp
@@ -39,6 +39,7 @@ AMDGPUTargetLowering::AMDGPUTargetLowering(TargetMachine &TM) :
   setOperationAction(ISD::FABS,   MVT::f32, Legal);
   setOperationAction(ISD::FFLOOR, MVT::f32, Legal);
   setOperationAction(ISD::FRINT,  MVT::f32, Legal);
+  setOperationAction(ISD::FTRUNC, MVT::f32, Legal);
 
   setOperationAction(ISD::UDIV, MVT::i32, Expand);
   setOperationAction(ISD::UDIVREM, MVT::i32, Custom);
@@ -137,8 +138,6 @@ SDValue AMDGPUTargetLowering::LowerINTRINSIC_WO_CHAIN(SDValue Op,
     case AMDGPUIntrinsic::AMDGPU_umin:
       return DAG.getNode(AMDGPUISD::UMIN, DL, VT, Op.getOperand(1),
                                                   Op.getOperand(2));
-    case AMDGPUIntrinsic::AMDIL_round_nearest:
-      return DAG.getNode(ISD::FRINT, DL, VT, Op.getOperand(1));
   }
 }
 
diff --git a/src/gallium/drivers/radeon/AMDGPUIntrinsics.td b/src/gallium/drivers/radeon/AMDGPUIntrinsics.td
index eaca4cf..70bf2af 100644
--- a/src/gallium/drivers/radeon/AMDGPUIntrinsics.td
+++ b/src/gallium/drivers/radeon/AMDGPUIntrinsics.td
@@ -43,7 +43,6 @@ let TargetPrefix = "AMDGPU", isTarget = 1 in {
   def int_AMDGPU_txq : Intrinsic<[llvm_v4f32_ty], [llvm_v4f32_ty, llvm_i32_ty, llvm_i32_ty], [IntrNoMem]>;
   def int_AMDGPU_txd : Intrinsic<[llvm_v4f32_ty], [llvm_v4f32_ty, llvm_v4f32_ty, llvm_v4f32_ty, llvm_i32_ty, llvm_i32_ty], [IntrNoMem]>;
   def int_AMDGPU_txl : Intrinsic<[llvm_v4f32_ty], [llvm_v4f32_ty, llvm_i32_ty, llvm_i32_ty], [IntrNoMem]>;
-  def int_AMDGPU_trunc : Intrinsic<[llvm_float_ty], [llvm_float_ty], [IntrNoMem]>;
   def int_AMDGPU_ddx : Intrinsic<[llvm_v4f32_ty], [llvm_v4f32_ty, llvm_i32_ty, llvm_i32_ty], [IntrNoMem]>;
   def int_AMDGPU_ddy : Intrinsic<[llvm_v4f32_ty], [llvm_v4f32_ty, llvm_i32_ty, llvm_i32_ty], [IntrNoMem]>;
   def int_AMDGPU_imax : Intrinsic<[llvm_i32_ty], [llvm_i32_ty, llvm_i32_ty], [IntrNoMem]>;
diff --git a/src/gallium/drivers/radeon/AMDILIntrinsics.td b/src/gallium/drivers/radeon/AMDILIntrinsics.td
index 3f9e20f..e7a71d4 100644
--- a/src/gallium/drivers/radeon/AMDILIntrinsics.td
+++ b/src/gallium/drivers/radeon/AMDILIntrinsics.td
@@ -142,8 +142,6 @@ let TargetPrefix = "AMDIL", isTarget = 1 in {
           TernaryIntFloat;
   def int_AMDIL_pireduce : GCCBuiltin<"__amdil_pireduce">,
           UnaryIntFloat;
-  def int_AMDIL_round_nearest : GCCBuiltin<"__amdil_round_nearest">,
-          UnaryIntFloat;
   def int_AMDIL_round_neginf : GCCBuiltin<"__amdil_round_neginf">,
           UnaryIntFloat;
   def int_AMDIL_round_zero : GCCBuiltin<"__amdil_round_zero">,
diff --git a/src/gallium/drivers/radeon/R600Instructions.td b/src/gallium/drivers/radeon/R600Instructions.td
index 120a71c..29826fa 100644
--- a/src/gallium/drivers/radeon/R600Instructions.td
+++ b/src/gallium/drivers/radeon/R600Instructions.td
@@ -384,7 +384,7 @@ def FRACT : R600_1OP <
 
 def TRUNC : R600_1OP <
   0x11, "TRUNC",
-  [(set R600_Reg32:$dst, (int_AMDGPU_trunc R600_Reg32:$src))]
+  [(set R600_Reg32:$dst, (ftrunc R600_Reg32:$src))]
 >;
 
 def CEIL : R600_1OP <
diff --git a/src/gallium/drivers/radeon/radeon_setup_tgsi_llvm.c b/src/gallium/drivers/radeon/radeon_setup_tgsi_llvm.c
index 57ec372..1a8154a 100644
--- a/src/gallium/drivers/radeon/radeon_setup_tgsi_llvm.c
+++ b/src/gallium/drivers/radeon/radeon_setup_tgsi_llvm.c
@@ -1099,8 +1099,8 @@ void radeon_llvm_context_init(struct radeon_llvm_context * ctx)
 	bld_base->op_actions[TGSI_OPCODE_USNE].emit = emit_icmp;
 	bld_base->op_actions[TGSI_OPCODE_ISGE].emit = emit_icmp;
 	bld_base->op_actions[TGSI_OPCODE_ISLT].emit = emit_icmp;
-	bld_base->op_actions[TGSI_OPCODE_ROUND].emit = build_tgsi_intrinsic_nomem;
-	bld_base->op_actions[TGSI_OPCODE_ROUND].intr_name = "llvm.AMDIL.round.nearest.";
+	bld_base->op_actions[TGSI_OPCODE_ROUND].emit = build_tgsi_intrinsic_readonly;
+	bld_base->op_actions[TGSI_OPCODE_ROUND].intr_name = "rint";
 	bld_base->op_actions[TGSI_OPCODE_MIN].emit = build_tgsi_intrinsic_nomem;
 	bld_base->op_actions[TGSI_OPCODE_MIN].intr_name = "llvm.AMDIL.min.";
 	bld_base->op_actions[TGSI_OPCODE_MAX].emit = build_tgsi_intrinsic_nomem;
@@ -1187,8 +1187,8 @@ void radeon_llvm_context_init(struct radeon_llvm_context * ctx)
 	bld_base->op_actions[TGSI_OPCODE_TXL].intr_name = "llvm.AMDGPU.txl";
 	bld_base->op_actions[TGSI_OPCODE_TXP].fetch_args = txp_fetch_args;
 	bld_base->op_actions[TGSI_OPCODE_TXP].intr_name = "llvm.AMDGPU.tex";
-	bld_base->op_actions[TGSI_OPCODE_TRUNC].emit = build_tgsi_intrinsic_nomem;
-	bld_base->op_actions[TGSI_OPCODE_TRUNC].intr_name = "llvm.AMDGPU.trunc";
+	bld_base->op_actions[TGSI_OPCODE_TRUNC].emit = build_tgsi_intrinsic_readonly;
+	bld_base->op_actions[TGSI_OPCODE_TRUNC].intr_name = "trunc";
 
 	bld_base->rsq_action.emit = build_tgsi_intrinsic_nomem;
 	bld_base->rsq_action.intr_name = "llvm.AMDGPU.rsq";
-- 
1.7.11.4



More information about the mesa-dev mailing list