[llvm] 452f624 - AMDGPU: Select llvm.amdgcn.interp.p2.f16 directly

[Matt Arsenault via llvm-commits]

Author: Matt Arsenault
Date: 2020-01-06T20:34:21-05:00
New Revision: 452f6243c9be73a476f796b1425bfc58749b9805

URL: https://github.com/llvm/llvm-project/commit/452f6243c9be73a476f796b1425bfc58749b9805
DIFF: https://github.com/llvm/llvm-project/commit/452f6243c9be73a476f796b1425bfc58749b9805.diff

LOG: AMDGPU: Select llvm.amdgcn.interp.p2.f16 directly

This will enable automatic GlobalISel support in a future commit.

Added: 
    

Modified: 
    llvm/lib/Target/AMDGPU/SIISelLowering.cpp
    llvm/lib/Target/AMDGPU/VOP3Instructions.td

Removed: 
    


################################################################################
diff  --git a/llvm/lib/Target/AMDGPU/SIISelLowering.cpp b/llvm/lib/Target/AMDGPU/SIISelLowering.cpp
index aa8cf7971197..7bacdf2e8a49 100644
--- a/llvm/lib/Target/AMDGPU/SIISelLowering.cpp
+++ b/llvm/lib/Target/AMDGPU/SIISelLowering.cpp
@@ -5951,22 +5951,6 @@ SDValue SITargetLowering::LowerINTRINSIC_WO_CHAIN(SDValue Op,
       return DAG.getNode(AMDGPUISD::INTERP_P1LL_F16, DL, MVT::f32, Ops);
     }
   }
-  case Intrinsic::amdgcn_interp_p2_f16: {
-    SDValue ToM0 = DAG.getCopyToReg(DAG.getEntryNode(), DL, AMDGPU::M0,
-                                    Op.getOperand(6), SDValue());
-    SDValue Ops[] = {
-      Op.getOperand(2), // Src0
-      Op.getOperand(3), // Attrchan
-      Op.getOperand(4), // Attr
-      DAG.getTargetConstant(0, DL, MVT::i32), // $src0_modifiers
-      Op.getOperand(1), // Src2
-      DAG.getTargetConstant(0, DL, MVT::i32), // $src2_modifiers
-      Op.getOperand(5), // high
-      DAG.getTargetConstant(0, DL, MVT::i1), // $clamp
-      ToM0.getValue(1)
-    };
-    return DAG.getNode(AMDGPUISD::INTERP_P2_F16, DL, MVT::f16, Ops);
-  }
   case Intrinsic::amdgcn_sin:
     return DAG.getNode(AMDGPUISD::SIN_HW, DL, VT, Op.getOperand(1));
 

diff  --git a/llvm/lib/Target/AMDGPU/VOP3Instructions.td b/llvm/lib/Target/AMDGPU/VOP3Instructions.td
index d8e28cadd623..67c8b926302d 100644
--- a/llvm/lib/Target/AMDGPU/VOP3Instructions.td
+++ b/llvm/lib/Target/AMDGPU/VOP3Instructions.td
@@ -258,15 +258,15 @@ class getInterp16Ins <bit HasSrc2, bit HasOMod,
                     (ins Src0Mod:$src0_modifiers, VRegSrc_32:$src0,
                          Attr:$attr, AttrChan:$attrchan,
                          Src2Mod:$src2_modifiers, VRegSrc_32:$src2,
-                         highmod:$high, clampmod:$clamp, omod:$omod),
+                         highmod:$high, clampmod0:$clamp, omod0:$omod),
                     (ins Src0Mod:$src0_modifiers, VRegSrc_32:$src0,
                          Attr:$attr, AttrChan:$attrchan,
                          Src2Mod:$src2_modifiers, VRegSrc_32:$src2,
-                         highmod:$high, clampmod:$clamp)
+                         highmod:$high, clampmod0:$clamp)
                 ),
                 (ins Src0Mod:$src0_modifiers, VRegSrc_32:$src0,
                      Attr:$attr, AttrChan:$attrchan,
-                     highmod:$high, clampmod:$clamp, omod:$omod)
+                     highmod:$high, clampmod0:$clamp, omod0:$omod)
             );
 }
 
@@ -452,14 +452,16 @@ def V_MAD_I16 : VOP3Inst <"v_mad_i16", VOP3_Profile<VOP_I16_I16_I16_I16, VOP3_CL
 let FPDPRounding = 1 in {
 def V_MAD_F16 : VOP3Inst <"v_mad_f16", VOP3_Profile<VOP_F16_F16_F16_F16>, fmad>;
 let Uses = [M0, EXEC] in {
+// For some reason the intrinsic operands are in a 
diff erent order
+// from the instruction operands.
 def V_INTERP_P2_F16 : VOP3Interp <"v_interp_p2_f16", VOP3_INTERP16<[f16, f32, i32, f32]>,
-       [(set f16:$vdst, (AMDGPUinterp_p2_f16 f32:$src0, (i32 timm:$attrchan),
-                                                        (i32 timm:$attr),
-                                                        (i32 timm:$src0_modifiers),
-                                                        (f32 VRegSrc_32:$src2),
-                                                        (i32 timm:$src2_modifiers),
-                                                        (i1 timm:$high),
-                                                        (i1 timm:$clamp)))]>;
+       [(set f16:$vdst,
+         (int_amdgcn_interp_p2_f16 (VOP3Mods f32:$src2, i32:$src2_modifiers),
+                                   (VOP3Mods f32:$src0, i32:$src0_modifiers),
+                                   (i32 timm:$attrchan),
+                                   (i32 timm:$attr),
+                                   (i1 timm:$high),
+                                   M0))]>;
 } // End Uses = [M0, EXEC]
 } // End FPDPRounding = 1
 } // End renamedInGFX9 = 1