summaryrefslogtreecommitdiffstats
path: root/llvm/lib
diff options
context:
space:
mode:
Diffstat (limited to 'llvm/lib')
-rw-r--r--llvm/lib/Target/AMDGPU/SIInstrFormats.td18
-rw-r--r--llvm/lib/Target/AMDGPU/SIInstrInfo.cpp2
-rw-r--r--llvm/lib/Target/AMDGPU/SIInstrInfo.td141
-rw-r--r--llvm/lib/Target/AMDGPU/SIInstructions.td6
-rw-r--r--llvm/lib/Target/AMDGPU/SIShrinkInstructions.cpp4
-rw-r--r--llvm/lib/Target/AMDGPU/VIInstrFormats.td26
6 files changed, 120 insertions, 77 deletions
diff --git a/llvm/lib/Target/AMDGPU/SIInstrFormats.td b/llvm/lib/Target/AMDGPU/SIInstrFormats.td
index a94212a665e..259638e9e8f 100644
--- a/llvm/lib/Target/AMDGPU/SIInstrFormats.td
+++ b/llvm/lib/Target/AMDGPU/SIInstrFormats.td
@@ -362,8 +362,7 @@ class VOP2_MADKe <bits<6> op> : Enc64 {
let Inst{63-32} = src2;
}
-class VOP3e <bits<9> op> : Enc64 {
- bits<8> vdst;
+class VOP3a <bits<9> op> : Enc64 {
bits<2> src0_modifiers;
bits<9> src0;
bits<2> src1_modifiers;
@@ -373,7 +372,6 @@ class VOP3e <bits<9> op> : Enc64 {
bits<1> clamp;
bits<2> omod;
- let Inst{7-0} = vdst;
let Inst{8} = src0_modifiers{1};
let Inst{9} = src1_modifiers{1};
let Inst{10} = src2_modifiers{1};
@@ -389,6 +387,20 @@ class VOP3e <bits<9> op> : Enc64 {
let Inst{63} = src2_modifiers{0};
}
+class VOP3e <bits<9> op> : VOP3a <op> {
+ bits<8> vdst;
+
+ let Inst{7-0} = vdst;
+}
+
+// Encoding used for VOPC instructions encoded as VOP3
+// Differs from VOP3e by destination name (sdst) as VOPC doesn't have vector dst
+class VOP3ce <bits<9> op> : VOP3a <op> {
+ bits<8> sdst;
+
+ let Inst{7-0} = sdst;
+}
+
class VOP3be <bits<9> op> : Enc64 {
bits<8> vdst;
bits<2> src0_modifiers;
diff --git a/llvm/lib/Target/AMDGPU/SIInstrInfo.cpp b/llvm/lib/Target/AMDGPU/SIInstrInfo.cpp
index 5650098efed..b595582697b 100644
--- a/llvm/lib/Target/AMDGPU/SIInstrInfo.cpp
+++ b/llvm/lib/Target/AMDGPU/SIInstrInfo.cpp
@@ -1244,7 +1244,7 @@ MachineInstr *SIInstrInfo::convertToThreeAddress(MachineFunction::iterator &MBB,
}
}
- const MachineOperand *Dst = getNamedOperand(*MI, AMDGPU::OpName::dst);
+ const MachineOperand *Dst = getNamedOperand(*MI, AMDGPU::OpName::vdst);
const MachineOperand *Src0 = getNamedOperand(*MI, AMDGPU::OpName::src0);
const MachineOperand *Src1 = getNamedOperand(*MI, AMDGPU::OpName::src1);
const MachineOperand *Src2 = getNamedOperand(*MI, AMDGPU::OpName::src2);
diff --git a/llvm/lib/Target/AMDGPU/SIInstrInfo.td b/llvm/lib/Target/AMDGPU/SIInstrInfo.td
index 137d77f485c..e341355274c 100644
--- a/llvm/lib/Target/AMDGPU/SIInstrInfo.td
+++ b/llvm/lib/Target/AMDGPU/SIInstrInfo.td
@@ -1186,8 +1186,8 @@ class getInsDPP <RegisterClass Src0RC, RegisterClass Src1RC, int NumSrcArgs,
// Returns the assembly string for the inputs and outputs of a VOP[12C]
// instruction. This does not add the _e32 suffix, so it can be reused
// by getAsm64.
-class getAsm32 <bit HasDst, int NumSrcArgs> {
- string dst = "$dst";
+class getAsm32 <bit HasDst, int NumSrcArgs, ValueType DstVT = i32> {
+ string dst = !if(!eq(DstVT.Size, 1), "$sdst", "$vdst"); // use $sdst for VOPC
string src0 = ", $src0";
string src1 = ", $src1";
string src2 = ", $src2";
@@ -1199,7 +1199,8 @@ class getAsm32 <bit HasDst, int NumSrcArgs> {
// Returns the assembly string for the inputs and outputs of a VOP3
// instruction.
-class getAsm64 <bit HasDst, int NumSrcArgs, bit HasModifiers> {
+class getAsm64 <bit HasDst, int NumSrcArgs, bit HasModifiers, ValueType DstVT = i32> {
+ string dst = !if(!eq(DstVT.Size, 1), "$sdst", "$vdst"); // use $sdst for VOPC
string src0 = !if(!eq(NumSrcArgs, 1), "$src0_modifiers", "$src0_modifiers,");
string src1 = !if(!eq(NumSrcArgs, 1), "",
!if(!eq(NumSrcArgs, 2), " $src1_modifiers",
@@ -1207,19 +1208,20 @@ class getAsm64 <bit HasDst, int NumSrcArgs, bit HasModifiers> {
string src2 = !if(!eq(NumSrcArgs, 3), " $src2_modifiers", "");
string ret =
!if(!eq(HasModifiers, 0),
- getAsm32<HasDst, NumSrcArgs>.ret,
- "$dst, "#src0#src1#src2#"$clamp"#"$omod");
+ getAsm32<HasDst, NumSrcArgs, DstVT>.ret,
+ dst#", "#src0#src1#src2#"$clamp"#"$omod");
}
-class getAsmDPP <bit HasDst, int NumSrcArgs, bit HasModifiers> {
+class getAsmDPP <bit HasDst, int NumSrcArgs, bit HasModifiers, ValueType DstVT = i32> {
+ string dst = !if(!eq(DstVT.Size, 1), "$sdst", "$vdst"); // use $sdst for VOPC
string src0 = !if(!eq(NumSrcArgs, 1), "$src0_modifiers", "$src0_modifiers,");
string src1 = !if(!eq(NumSrcArgs, 1), "",
!if(!eq(NumSrcArgs, 2), " $src1_modifiers",
" $src1_modifiers,"));
string args = !if(!eq(HasModifiers, 0),
- getAsm32<0, NumSrcArgs>.ret,
+ getAsm32<0, NumSrcArgs, DstVT>.ret,
src0#src1);
- string ret = " $dst"#args#", $dpp_ctrl, "#"$bound_ctrl, "#"$bank_mask, "#"$row_mask";
+ string ret = " "#dst#args#", $dpp_ctrl, "#"$bound_ctrl, "#"$bank_mask, "#"$row_mask";
}
class VOPProfile <list<ValueType> _ArgVT> {
@@ -1245,22 +1247,23 @@ class VOPProfile <list<ValueType> _ArgVT> {
field int NumSrcArgs = getNumSrcArgs<Src0VT, Src1VT, Src2VT>.ret;
field bit HasModifiers = hasModifiers<Src0VT>.ret;
- field dag Outs = !if(HasDst,(outs DstRC:$dst),(outs));
+ field dag Outs = !if(HasDst,(outs DstRC:$vdst),(outs));
// VOP3b instructions are a special case with a second explicit
// output. This is manually overridden for them.
field dag Outs32 = Outs;
field dag Outs64 = Outs;
- field dag OutsDPP = (outs DstRCDPP:$dst);
+ field dag OutsDPP = !if(!eq(DstVT.Size, 1), (outs DstRCDPP:$sdst), // sdst for VOPC
+ (outs DstRCDPP:$vdst));
field dag Ins32 = getIns32<Src0RC32, Src1RC32, NumSrcArgs>.ret;
field dag Ins64 = getIns64<Src0RC64, Src1RC64, Src2RC64, NumSrcArgs,
HasModifiers>.ret;
field dag InsDPP = getInsDPP<Src0DPP, Src1DPP, NumSrcArgs, HasModifiers>.ret;
- field string Asm32 = getAsm32<HasDst, NumSrcArgs>.ret;
- field string Asm64 = getAsm64<HasDst, NumSrcArgs, HasModifiers>.ret;
- field string AsmDPP = getAsmDPP<HasDst, NumSrcArgs, HasModifiers>.ret;
+ field string Asm32 = getAsm32<HasDst, NumSrcArgs, DstVT>.ret;
+ field string Asm64 = getAsm64<HasDst, NumSrcArgs, HasModifiers, DstVT>.ret;
+ field string AsmDPP = getAsmDPP<HasDst, NumSrcArgs, HasModifiers, DstVT>.ret;
}
// FIXME: I think these F16/I16 profiles will need to use f16/i16 types in order
@@ -1295,10 +1298,10 @@ def VOP_I32_I32_I32 : VOPProfile <[i32, i32, i32, untyped]>;
// Write out to vcc or arbitrary SGPR.
def VOP2b_I32_I1_I32_I32 : VOPProfile<[i32, i32, i32, untyped]> {
- let Asm32 = "$dst, vcc, $src0, $src1";
- let Asm64 = "$dst, $sdst, $src0, $src1";
- let Outs32 = (outs DstRC:$dst);
- let Outs64 = (outs DstRC:$dst, SReg_64:$sdst);
+ let Asm32 = "$vdst, vcc, $src0, $src1";
+ let Asm64 = "$vdst, $sdst, $src0, $src1";
+ let Outs32 = (outs DstRC:$vdst);
+ let Outs64 = (outs DstRC:$vdst, SReg_64:$sdst);
}
// Write out to vcc or arbitrary SGPR and read in from vcc or
@@ -1310,10 +1313,10 @@ def VOP2b_I32_I1_I32_I32_I1 : VOPProfile<[i32, i32, i32, i1]> {
// restriction. SGPRs are still allowed because it should
// technically be possible to use VCC again as src0.
let Src0RC32 = VCSrc_32;
- let Asm32 = "$dst, vcc, $src0, $src1, vcc";
- let Asm64 = "$dst, $sdst, $src0, $src1, $src2";
- let Outs32 = (outs DstRC:$dst);
- let Outs64 = (outs DstRC:$dst, SReg_64:$sdst);
+ let Asm32 = "$vdst, vcc, $src0, $src1, vcc";
+ let Asm64 = "$vdst, $sdst, $src0, $src1, $src2";
+ let Outs32 = (outs DstRC:$vdst);
+ let Outs64 = (outs DstRC:$vdst, SReg_64:$sdst);
// Suppress src2 implied by type since the 32-bit encoding uses an
// implicit VCC use.
@@ -1342,11 +1345,12 @@ class VOPC_Profile<ValueType vt0, ValueType vt1 = vt0> : VOPProfile <[i1, vt0, v
let Asm32 = "vcc, $src0, $src1";
// The destination for 32-bit encoding is implicit.
let HasDst32 = 0;
+ let Outs64 = (outs DstRC:$sdst);
}
class VOPC_Class_Profile<ValueType vt> : VOPC_Profile<vt, i32> {
let Ins64 = (ins InputModsNoDefault:$src0_modifiers, Src0RC64:$src0, Src1RC64:$src1);
- let Asm64 = "$dst, $src0_modifiers, $src1";
+ let Asm64 = "$sdst, $src0_modifiers, $src1";
}
def VOPC_I1_F32_F32 : VOPC_Profile<f32>;
@@ -1363,25 +1367,26 @@ def VOP_I64_I64_I64 : VOPProfile <[i64, i64, i64, untyped]>;
def VOP_CNDMASK : VOPProfile <[i32, i32, i32, untyped]> {
let Ins32 = (ins Src0RC32:$src0, Src1RC32:$src1);
let Ins64 = (ins Src0RC64:$src0, Src1RC64:$src1, SSrc_64:$src2);
- let Asm64 = "$dst, $src0, $src1, $src2";
+ let Asm64 = "$vdst, $src0, $src1, $src2";
}
def VOP_F32_F32_F32_F32 : VOPProfile <[f32, f32, f32, f32]>;
def VOP_MADK : VOPProfile <[f32, f32, f32, f32]> {
field dag Ins = (ins VCSrc_32:$src0, VGPR_32:$vsrc1, u32imm:$src2);
- field string Asm = "$dst, $src0, $vsrc1, $src2";
+ field string Asm = "$vdst, $src0, $vsrc1, $src2";
}
def VOP_MAC : VOPProfile <[f32, f32, f32, f32]> {
let Ins32 = (ins Src0RC32:$src0, Src1RC32:$src1, VGPR_32:$src2);
let Ins64 = getIns64<Src0RC64, Src1RC64, RegisterOperand<VGPR_32>, 3,
HasModifiers>.ret;
- let Asm32 = getAsm32<1, 2>.ret;
- let Asm64 = getAsm64<1, 2, HasModifiers>.ret;
+ let Asm32 = getAsm32<1, 2, f32>.ret;
+ let Asm64 = getAsm64<1, 2, HasModifiers, f32>.ret;
}
def VOP_F64_F64_F64_F64 : VOPProfile <[f64, f64, f64, f64]>;
def VOP_I32_I32_I32_I32 : VOPProfile <[i32, i32, i32, i32]>;
def VOP_I64_I32_I32_I64 : VOPProfile <[i64, i32, i32, i64]>;
+// This class is used only with VOPC instructions. Use $sdst for out operand
class SIInstAlias <string asm, Instruction inst, VOPProfile p> :
InstAlias <asm, (inst)>, PredicateControl {
@@ -1392,13 +1397,13 @@ class SIInstAlias <string asm, Instruction inst, VOPProfile p> :
!if (p.HasDst32,
!if (!eq(p.NumSrcArgs, 0),
// 1 dst, 0 src
- (inst p.DstRC:$dst),
+ (inst p.DstRC:$sdst),
!if (!eq(p.NumSrcArgs, 1),
// 1 dst, 1 src
- (inst p.DstRC:$dst, p.Src0RC32:$src0),
+ (inst p.DstRC:$sdst, p.Src0RC32:$src0),
!if (!eq(p.NumSrcArgs, 2),
// 1 dst, 2 src
- (inst p.DstRC:$dst, p.Src0RC32:$src0, p.Src1RC32:$src1),
+ (inst p.DstRC:$sdst, p.Src0RC32:$src0, p.Src1RC32:$src1),
// else - unreachable
(inst)))),
// else
@@ -1481,8 +1486,6 @@ multiclass VOP1_m <vop1 op, string opName, VOPProfile p, list<dag> pattern,
class VOP1_DPP <vop1 op, string opName, VOPProfile p> :
VOP1_DPPe <op.VI>,
VOP_DPP <p.OutsDPP, p.InsDPP, opName#p.AsmDPP, []> {
- // FIXME: remove when we are using the correct names for the encoding fields.
- field bit vdst = 0;
let AssemblerPredicates = [isVI];
let src0_modifiers = !if(p.HasModifiers, ?, 0);
let src1_modifiers = 0;
@@ -1589,6 +1592,22 @@ class VOP3_Real_vi <bits<10> op, dag outs, dag ins, string asm, string opName,
let AssemblerPredicates = [isVI];
}
+class VOP3_C_Real_si <bits<9> op, dag outs, dag ins, string asm, string opName,
+ bit HasMods = 0, bit VOP3Only = 0> :
+ VOP3Common <outs, ins, asm, [], HasMods, VOP3Only>,
+ VOP3ce <op>,
+ SIMCInstr<opName#"_e64", SISubtarget.SI> {
+ let AssemblerPredicates = [isSICI];
+}
+
+class VOP3_C_Real_vi <bits<10> op, dag outs, dag ins, string asm, string opName,
+ bit HasMods = 0, bit VOP3Only = 0> :
+ VOP3Common <outs, ins, asm, [], HasMods, VOP3Only>,
+ VOP3ce_vi <op>,
+ SIMCInstr <opName#"_e64", SISubtarget.VI> {
+ let AssemblerPredicates = [isVI];
+}
+
class VOP3b_Real_si <bits<9> op, dag outs, dag ins, string asm, string opName,
bit HasMods = 0, bit VOP3Only = 0> :
VOP3Common <outs, ins, asm, [], HasMods, VOP3Only>,
@@ -1694,13 +1713,13 @@ multiclass VOP3_C_m <vop op, dag outs, dag ins, string asm,
let SchedRW = sched;
}
- def _si : VOP3_Real_si <op.SI3, outs, ins, asm, opName, HasMods>,
+ def _si : VOP3_C_Real_si <op.SI3, outs, ins, asm, opName, HasMods>,
VOP3DisableFields<1, 0, HasMods> {
let Defs = !if(defExec, [EXEC], []);
let SchedRW = sched;
}
- def _vi : VOP3_Real_vi <op.VI3, outs, ins, asm, opName, HasMods>,
+ def _vi : VOP3_C_Real_vi <op.VI3, outs, ins, asm, opName, HasMods>,
VOP3DisableFields<1, 0, HasMods> {
let Defs = !if(defExec, [EXEC], []);
let SchedRW = sched;
@@ -1743,9 +1762,9 @@ multiclass VOP1Inst <vop1 op, string opName, VOPProfile P,
SDPatternOperator node = null_frag> : VOP1_Helper <
op, opName, P, [],
!if(P.HasModifiers,
- [(set P.DstVT:$dst, (node (P.Src0VT (VOP3Mods0 P.Src0VT:$src0,
+ [(set P.DstVT:$vdst, (node (P.Src0VT (VOP3Mods0 P.Src0VT:$src0,
i32:$src0_modifiers, i1:$clamp, i32:$omod))))],
- [(set P.DstVT:$dst, (node P.Src0VT:$src0))])
+ [(set P.DstVT:$vdst, (node P.Src0VT:$src0))])
>;
multiclass VOP1InstSI <vop1 op, string opName, VOPProfile P,
@@ -1755,9 +1774,9 @@ multiclass VOP1InstSI <vop1 op, string opName, VOPProfile P,
defm _e64 : VOP3SI_1_m <op, P.Outs, P.Ins64, opName#P.Asm64,
!if(P.HasModifiers,
- [(set P.DstVT:$dst, (node (P.Src0VT (VOP3Mods0 P.Src0VT:$src0,
+ [(set P.DstVT:$vdst, (node (P.Src0VT (VOP3Mods0 P.Src0VT:$src0,
i32:$src0_modifiers, i1:$clamp, i32:$omod))))],
- [(set P.DstVT:$dst, (node P.Src0VT:$src0))]),
+ [(set P.DstVT:$vdst, (node P.Src0VT:$src0))]),
opName, P.HasModifiers>;
}
@@ -1775,11 +1794,11 @@ multiclass VOP2Inst <vop2 op, string opName, VOPProfile P,
string revOp = opName> : VOP2_Helper <
op, opName, P, [],
!if(P.HasModifiers,
- [(set P.DstVT:$dst,
+ [(set P.DstVT:$vdst,
(node (P.Src0VT (VOP3Mods0 P.Src0VT:$src0, i32:$src0_modifiers,
i1:$clamp, i32:$omod)),
(P.Src1VT (VOP3Mods P.Src1VT:$src1, i32:$src1_modifiers))))],
- [(set P.DstVT:$dst, (node P.Src0VT:$src0, P.Src1VT:$src1))]),
+ [(set P.DstVT:$vdst, (node P.Src0VT:$src0, P.Src1VT:$src1))]),
revOp
>;
@@ -1791,11 +1810,11 @@ multiclass VOP2InstSI <vop2 op, string opName, VOPProfile P,
defm _e64 : VOP3SI_2_m <op, P.Outs, P.Ins64, opName#P.Asm64,
!if(P.HasModifiers,
- [(set P.DstVT:$dst,
+ [(set P.DstVT:$vdst,
(node (P.Src0VT (VOP3Mods0 P.Src0VT:$src0, i32:$src0_modifiers,
i1:$clamp, i32:$omod)),
(P.Src1VT (VOP3Mods P.Src1VT:$src1, i32:$src1_modifiers))))],
- [(set P.DstVT:$dst, (node P.Src0VT:$src0, P.Src1VT:$src1))]),
+ [(set P.DstVT:$vdst, (node P.Src0VT:$src0, P.Src1VT:$src1))]),
opName, revOp, P.HasModifiers>;
}
@@ -1818,11 +1837,11 @@ multiclass VOP2bInst <vop2 op, string opName, VOPProfile P,
string revOp = opName> : VOP2b_Helper <
op, opName, P, [],
!if(P.HasModifiers,
- [(set P.DstVT:$dst,
+ [(set P.DstVT:$vdst,
(node (P.Src0VT (VOP3Mods0 P.Src0VT:$src0, i32:$src0_modifiers,
i1:$clamp, i32:$omod)),
(P.Src1VT (VOP3Mods P.Src1VT:$src1, i32:$src1_modifiers))))],
- [(set P.DstVT:$dst, (node P.Src0VT:$src0, P.Src1VT:$src1))]),
+ [(set P.DstVT:$vdst, (node P.Src0VT:$src0, P.Src1VT:$src1))]),
revOp, !eq(P.NumSrcArgs, 3)
>;
@@ -1842,11 +1861,11 @@ multiclass VOP2_VI3_Inst <vop23 op, string opName, VOPProfile P,
: VOP2_VI3_Helper <
op, opName, P, [],
!if(P.HasModifiers,
- [(set P.DstVT:$dst,
+ [(set P.DstVT:$vdst,
(node (P.Src0VT (VOP3Mods0 P.Src0VT:$src0, i32:$src0_modifiers,
i1:$clamp, i32:$omod)),
(P.Src1VT (VOP3Mods P.Src1VT:$src1, i32:$src1_modifiers))))],
- [(set P.DstVT:$dst, (node P.Src0VT:$src0, P.Src1VT:$src1))]),
+ [(set P.DstVT:$vdst, (node P.Src0VT:$src0, P.Src1VT:$src1))]),
revOp
>;
@@ -1919,7 +1938,7 @@ multiclass VOPC_Helper <vopc op, string opName, list<dag> pat32,
defm _e32 : VOPC_m <op, p.Ins32, p.Asm32, pat32, opName, DefExec, p, sched,
revOp>;
- defm _e64 : VOP3_C_m <op, (outs VOPDstS64:$dst), p.Ins64, opName#p.Asm64, pat64,
+ defm _e64 : VOP3_C_m <op, (outs VOPDstS64:$sdst), p.Ins64, opName#p.Asm64, pat64,
opName, p.HasModifiers, DefExec, revOp, sched>;
}
@@ -1930,7 +1949,7 @@ multiclass VOPC_Class_Helper <vopc op, string opName, list<dag> pat32,
VOPProfile p, list<SchedReadWrite> sched> {
defm _e32 : VOPC_m <op, p.Ins32, p.Asm32, pat32, opName, DefExec, p, sched>;
- defm _e64 : VOP3_C_m <op, (outs VOPDstS64:$dst), p.Ins64, opName#p.Asm64, pat64,
+ defm _e64 : VOP3_C_m <op, (outs VOPDstS64:$sdst), p.Ins64, opName#p.Asm64, pat64,
opName, p.HasModifiers, DefExec, revOp, sched>,
VOP3DisableModFields<1, 0, 0>;
}
@@ -1943,12 +1962,12 @@ multiclass VOPCInst <vopc op, string opName,
VOPC_Helper <
op, opName, [],
!if(P.HasModifiers,
- [(set i1:$dst,
+ [(set i1:$sdst,
(setcc (P.Src0VT (VOP3Mods0 P.Src0VT:$src0, i32:$src0_modifiers,
i1:$clamp, i32:$omod)),
(P.Src1VT (VOP3Mods P.Src1VT:$src1, i32:$src1_modifiers)),
cond))],
- [(set i1:$dst, (setcc P.Src0VT:$src0, P.Src1VT:$src1, cond))]),
+ [(set i1:$sdst, (setcc P.Src0VT:$src0, P.Src1VT:$src1, cond))]),
DefExec, revOp, P, sched
>;
@@ -1957,9 +1976,9 @@ multiclass VOPCClassInst <vopc op, string opName, VOPProfile P,
list<SchedReadWrite> sched> : VOPC_Class_Helper <
op, opName, [],
!if(P.HasModifiers,
- [(set i1:$dst,
+ [(set i1:$sdst,
(AMDGPUfp_class (P.Src0VT (VOP3Mods0Clamp0OMod P.Src0VT:$src0, i32:$src0_modifiers)), P.Src1VT:$src1))],
- [(set i1:$dst, (AMDGPUfp_class P.Src0VT:$src0, P.Src1VT:$src1))]),
+ [(set i1:$sdst, (AMDGPUfp_class P.Src0VT:$src0, P.Src1VT:$src1))]),
DefExec, opName, P, sched
>;
@@ -2016,29 +2035,29 @@ multiclass VOPCX_CLASS_F64 <vopc op, string opName> :
multiclass VOP3Inst <vop3 op, string opName, VOPProfile P,
SDPatternOperator node = null_frag, bit VOP3Only = 0> :
VOP3_Helper <
- op, opName, (outs P.DstRC.RegClass:$dst), P.Ins64, P.Asm64,
+ op, opName, (outs P.DstRC.RegClass:$vdst), P.Ins64, P.Asm64,
!if(!eq(P.NumSrcArgs, 3),
!if(P.HasModifiers,
- [(set P.DstVT:$dst,
+ [(set P.DstVT:$vdst,
(node (P.Src0VT (VOP3Mods0 P.Src0VT:$src0, i32:$src0_modifiers,
i1:$clamp, i32:$omod)),
(P.Src1VT (VOP3Mods P.Src1VT:$src1, i32:$src1_modifiers)),
(P.Src2VT (VOP3Mods P.Src2VT:$src2, i32:$src2_modifiers))))],
- [(set P.DstVT:$dst, (node P.Src0VT:$src0, P.Src1VT:$src1,
+ [(set P.DstVT:$vdst, (node P.Src0VT:$src0, P.Src1VT:$src1,
P.Src2VT:$src2))]),
!if(!eq(P.NumSrcArgs, 2),
!if(P.HasModifiers,
- [(set P.DstVT:$dst,
+ [(set P.DstVT:$vdst,
(node (P.Src0VT (VOP3Mods0 P.Src0VT:$src0, i32:$src0_modifiers,
i1:$clamp, i32:$omod)),
(P.Src1VT (VOP3Mods P.Src1VT:$src1, i32:$src1_modifiers))))],
- [(set P.DstVT:$dst, (node P.Src0VT:$src0, P.Src1VT:$src1))])
+ [(set P.DstVT:$vdst, (node P.Src0VT:$src0, P.Src1VT:$src1))])
/* P.NumSrcArgs == 1 */,
!if(P.HasModifiers,
- [(set P.DstVT:$dst,
+ [(set P.DstVT:$vdst,
(node (P.Src0VT (VOP3Mods0 P.Src0VT:$src0, i32:$src0_modifiers,
i1:$clamp, i32:$omod))))],
- [(set P.DstVT:$dst, (node P.Src0VT:$src0))]))),
+ [(set P.DstVT:$vdst, (node P.Src0VT:$src0))]))),
P.NumSrcArgs, P.HasModifiers, VOP3Only
>;
@@ -2048,14 +2067,14 @@ multiclass VOP3_VCC_Inst <vop3 op, string opName,
VOPProfile P,
SDPatternOperator node = null_frag> : VOP3_Helper <
op, opName,
- (outs P.DstRC.RegClass:$dst),
+ (outs P.DstRC.RegClass:$vdst),
(ins InputModsNoDefault:$src0_modifiers, P.Src0RC64:$src0,
InputModsNoDefault:$src1_modifiers, P.Src1RC64:$src1,
InputModsNoDefault:$src2_modifiers, P.Src2RC64:$src2,
ClampMod:$clamp,
omod:$omod),
- "$dst, $src0_modifiers, $src1_modifiers, $src2_modifiers"#"$clamp"#"$omod",
- [(set P.DstVT:$dst,
+ "$vdst, $src0_modifiers, $src1_modifiers, $src2_modifiers"#"$clamp"#"$omod",
+ [(set P.DstVT:$vdst,
(node (P.Src0VT (VOP3Mods0 P.Src0VT:$src0, i32:$src0_modifiers,
i1:$clamp, i32:$omod)),
(P.Src1VT (VOP3Mods P.Src1VT:$src1, i32:$src1_modifiers)),
diff --git a/llvm/lib/Target/AMDGPU/SIInstructions.td b/llvm/lib/Target/AMDGPU/SIInstructions.td
index 251f92988fb..92de6a6eaa7 100644
--- a/llvm/lib/Target/AMDGPU/SIInstructions.td
+++ b/llvm/lib/Target/AMDGPU/SIInstructions.td
@@ -1503,7 +1503,7 @@ defm V_AND_B32 : VOP2Inst <vop2<0x1b, 0x13>, "v_and_b32", VOP_I32_I32_I32>;
defm V_OR_B32 : VOP2Inst <vop2<0x1c, 0x14>, "v_or_b32", VOP_I32_I32_I32>;
defm V_XOR_B32 : VOP2Inst <vop2<0x1d, 0x15>, "v_xor_b32", VOP_I32_I32_I32>;
-let Constraints = "$dst = $src2", DisableEncoding="$src2",
+let Constraints = "$vdst = $src2", DisableEncoding="$src2",
isConvertibleToThreeAddress = 1 in {
defm V_MAC_F32 : VOP2Inst <vop2<0x1f, 0x16>, "v_mac_f32", VOP_MAC>;
}
@@ -1850,14 +1850,14 @@ defm V_ASHRREV_I64 : VOP3Inst <vop3<0, 0x291>, "v_ashrrev_i64",
let isCodeGenOnly = 1, isPseudo = 1 in {
// For use in patterns
-def V_CNDMASK_B64_PSEUDO : VOP3Common <(outs VReg_64:$dst),
+def V_CNDMASK_B64_PSEUDO : VOP3Common <(outs VReg_64:$vdst),
(ins VSrc_64:$src0, VSrc_64:$src1, SSrc_64:$src2), "", []
>;
let hasSideEffects = 0, mayLoad = 0, mayStore = 0, Uses = [EXEC] in {
// 64-bit vector move instruction. This is mainly used by the SIFoldOperands
// pass to enable folding of inline immediates.
-def V_MOV_B64_PSEUDO : InstSI <(outs VReg_64:$dst), (ins VSrc_64:$src0), "", []>;
+def V_MOV_B64_PSEUDO : InstSI <(outs VReg_64:$vdst), (ins VSrc_64:$src0), "", []>;
} // End let hasSideEffects = 0, mayLoad = 0, mayStore = 0
let hasSideEffects = 1, SALU = 1 in {
diff --git a/llvm/lib/Target/AMDGPU/SIShrinkInstructions.cpp b/llvm/lib/Target/AMDGPU/SIShrinkInstructions.cpp
index 4f0913fe62f..c8cac805abf 100644
--- a/llvm/lib/Target/AMDGPU/SIShrinkInstructions.cpp
+++ b/llvm/lib/Target/AMDGPU/SIShrinkInstructions.cpp
@@ -287,9 +287,9 @@ bool SIShrinkInstructions::runOnMachineFunction(MachineFunction &MF) {
MachineInstrBuilder Inst32 =
BuildMI(MBB, I, MI.getDebugLoc(), TII->get(Op32));
- // Add the dst operand if the 32-bit encoding also has an explicit $dst.
+ // Add the dst operand if the 32-bit encoding also has an explicit $vdst.
// For VOPC instructions, this is replaced by an implicit def of vcc.
- int Op32DstIdx = AMDGPU::getNamedOperandIdx(Op32, AMDGPU::OpName::dst);
+ int Op32DstIdx = AMDGPU::getNamedOperandIdx(Op32, AMDGPU::OpName::vdst);
if (Op32DstIdx != -1) {
// dst
Inst32.addOperand(MI.getOperand(0));
diff --git a/llvm/lib/Target/AMDGPU/VIInstrFormats.td b/llvm/lib/Target/AMDGPU/VIInstrFormats.td
index 7a476bc87ff..c21137233e2 100644
--- a/llvm/lib/Target/AMDGPU/VIInstrFormats.td
+++ b/llvm/lib/Target/AMDGPU/VIInstrFormats.td
@@ -104,8 +104,7 @@ class SMEMe_vi <bits<8> op, bit imm> : Enc64 {
let Inst{51-32} = offset;
}
-class VOP3e_vi <bits<10> op> : Enc64 {
- bits<8> vdst;
+class VOP3a_vi <bits<10> op> : Enc64 {
bits<2> src0_modifiers;
bits<9> src0;
bits<2> src1_modifiers;
@@ -115,7 +114,6 @@ class VOP3e_vi <bits<10> op> : Enc64 {
bits<1> clamp;
bits<2> omod;
- let Inst{7-0} = vdst;
let Inst{8} = src0_modifiers{1};
let Inst{9} = src1_modifiers{1};
let Inst{10} = src2_modifiers{1};
@@ -131,6 +129,20 @@ class VOP3e_vi <bits<10> op> : Enc64 {
let Inst{63} = src2_modifiers{0};
}
+class VOP3e_vi <bits<10> op> : VOP3a_vi <op> {
+ bits<8> vdst;
+
+ let Inst{7-0} = vdst;
+}
+
+// Encoding used for VOPC instructions encoded as VOP3
+// Differs from VOP3e by destination name (sdst) as VOPC doesn't have vector dst
+class VOP3ce_vi <bits<10> op> : VOP3a_vi <op> {
+ bits<8> sdst;
+
+ let Inst{7-0} = sdst;
+}
+
class VOP3be_vi <bits<10> op> : Enc64 {
bits<8> vdst;
bits<2> src0_modifiers;
@@ -183,21 +195,21 @@ class VOP_DPPe : Enc64 {
}
class VOP1_DPPe <bits<8> op> : VOP_DPPe {
- bits<8> dst;
+ bits<8> vdst;
let Inst{8-0} = 0xfa; // dpp
let Inst{16-9} = op;
- let Inst{24-17} = dst;
+ let Inst{24-17} = vdst;
let Inst{31-25} = 0x3f; //encoding
}
class VOP2_DPPe <bits<6> op> : Enc32 {
- bits<8> dst;
+ bits<8> vdst;
bits<8> src1;
let Inst{8-0} = 0xfa; //dpp
let Inst{16-9} = src1;
- let Inst{24-17} = dst;
+ let Inst{24-17} = vdst;
let Inst{30-25} = op;
let Inst{31} = 0x0; //encoding
}
OpenPOWER on IntegriCloud