[{return N->get() == ISD::SETOEQ || N->get() == ISD::SETEQ;}]
>;
+def COND_ONE : PatLeaf <
+ (cond),
+ [{return N->get() == ISD::SETONE || N->get() == ISD::SETNE;}]
+>;
+
def COND_OGT : PatLeaf <
(cond),
[{return N->get() == ISD::SETOGT || N->get() == ISD::SETGT;}]
[{return N->get() == ISD::SETOLE || N->get() == ISD::SETLE;}]
>;
-def COND_UNE : PatLeaf <
- (cond),
- [{return N->get() == ISD::SETUNE || N->get() == ISD::SETNE;}]
->;
def COND_O : PatLeaf <(cond), [{return N->get() == ISD::SETO;}]>;
def COND_UO : PatLeaf <(cond), [{return N->get() == ISD::SETUO;}]>;
//===----------------------------------------------------------------------===//
-// PatLeafs for unsigned comparisons
+// PatLeafs for unsigned / unordered comparisons
//===----------------------------------------------------------------------===//
+def COND_UEQ : PatLeaf <(cond), [{return N->get() == ISD::SETUEQ;}]>;
+def COND_UNE : PatLeaf <(cond), [{return N->get() == ISD::SETUNE;}]>;
def COND_UGT : PatLeaf <(cond), [{return N->get() == ISD::SETUGT;}]>;
def COND_UGE : PatLeaf <(cond), [{return N->get() == ISD::SETUGE;}]>;
def COND_ULT : PatLeaf <(cond), [{return N->get() == ISD::SETULT;}]>;
def COND_ULE : PatLeaf <(cond), [{return N->get() == ISD::SETULE;}]>;
+// XXX - For some reason R600 version is preferring to use unordered
+// for setne?
+def COND_UNE_NE : PatLeaf <
+ (cond),
+ [{return N->get() == ISD::SETUNE || N->get() == ISD::SETNE;}]
+>;
+
//===----------------------------------------------------------------------===//
// PatLeafs for signed comparisons
//===----------------------------------------------------------------------===//
(ops node:$value, node:$ptr), (op node:$value, node:$ptr)
>;
-def extloadi8_private : PrivateLoad <extloadi8>;
-def sextloadi8_private : PrivateLoad <sextloadi8>;
-def extloadi16_private : PrivateLoad <extloadi16>;
-def sextloadi16_private : PrivateLoad <sextloadi16>;
def load_private : PrivateLoad <load>;
def truncstorei8_private : PrivateStore <truncstorei8>;
return isLocalLoad(dyn_cast<LoadSDNode>(N));
}]>;
+def extloadi8_private : PrivateLoad <az_extloadi8>;
+def sextloadi8_private : PrivateLoad <sextloadi8>;
+
def az_extloadi16 : PatFrag<(ops node:$ptr), (az_extload node:$ptr), [{
return cast<LoadSDNode>(N)->getMemoryVT() == MVT::i16;
}]>;
return isLocalLoad(dyn_cast<LoadSDNode>(N));
}]>;
+def extloadi16_private : PrivateLoad <az_extloadi16>;
+def sextloadi16_private : PrivateLoad <sextloadi16>;
+
def az_extloadi32 : PatFrag<(ops node:$ptr), (az_extload node:$ptr), [{
return cast<LoadSDNode>(N)->getMemoryVT() == MVT::i32;
}]>;
def mskor_global : PatFrag<(ops node:$val, node:$ptr),
(AMDGPUstore_mskor node:$val, node:$ptr), [{
- return dyn_cast<MemSDNode>(N)->getAddressSpace() == AMDGPUAS::GLOBAL_ADDRESS;
+ return cast<MemSDNode>(N)->getAddressSpace() == AMDGPUAS::GLOBAL_ADDRESS;
}]>;
def mskor_flat : PatFrag<(ops node:$val, node:$ptr),
(AMDGPUstore_mskor node:$val, node:$ptr), [{
- return dyn_cast<MemSDNode>(N)->getAddressSpace() == AMDGPUAS::FLAT_ADDRESS;
+ return cast<MemSDNode>(N)->getAddressSpace() == AMDGPUAS::FLAT_ADDRESS;
}]>;
class global_binary_atomic_op<SDNode atomic_op> : PatFrag<
[{return cast<MemSDNode>(N)->getAddressSpace() == AMDGPUAS::GLOBAL_ADDRESS;}]
>;
+def atomic_swap_global : global_binary_atomic_op<atomic_swap>;
def atomic_add_global : global_binary_atomic_op<atomic_load_add>;
def atomic_and_global : global_binary_atomic_op<atomic_load_and>;
def atomic_max_global : global_binary_atomic_op<atomic_load_max>;
def atomic_min_global : global_binary_atomic_op<atomic_load_min>;
+def atomic_or_global : global_binary_atomic_op<atomic_load_or>;
def atomic_sub_global : global_binary_atomic_op<atomic_load_sub>;
def atomic_umax_global : global_binary_atomic_op<atomic_load_umax>;
def atomic_umin_global : global_binary_atomic_op<atomic_load_umin>;
+def atomic_xor_global : global_binary_atomic_op<atomic_load_xor>;
//===----------------------------------------------------------------------===//
// Misc Pattern Fragments
//===----------------------------------------------------------------------===//
-def fmad : PatFrag <
- (ops node:$src0, node:$src1, node:$src2),
- (fadd (fmul node:$src0, node:$src1), node:$src2)
->;
-
class Constants {
int TWO_PI = 0x40c90fdb;
int PI = 0x40490fdb;
[{return N->isExactlyValue(1.0);}]
>;
+def FP_HALF : PatLeaf <
+ (fpimm),
+ [{return N->isExactlyValue(0.5);}]
+>;
+
let isCodeGenOnly = 1, isPseudo = 1 in {
let usesCustomInserter = 1 in {
// BFI_INT patterns
-multiclass BFIPatterns <Instruction BFI_INT, Instruction LoadImm32> {
-
+multiclass BFIPatterns <Instruction BFI_INT,
+ Instruction LoadImm32,
+ RegisterClass RC64> {
// Definition from ISA doc:
// (y & x) | (z & ~x)
def : Pat <
def : Pat <
(f64 (fcopysign f64:$src0, f64:$src1)),
- (INSERT_SUBREG (INSERT_SUBREG (f64 (IMPLICIT_DEF)),
- (i32 (EXTRACT_SUBREG $src0, sub0)), sub0),
+ (REG_SEQUENCE RC64,
+ (i32 (EXTRACT_SUBREG $src0, sub0)), sub0,
(BFI_INT (LoadImm32 0x7fffffff),
(i32 (EXTRACT_SUBREG $src0, sub1)),
(i32 (EXTRACT_SUBREG $src1, sub1))), sub1)
// Bitfield extract patterns
-/*
-
-XXX: The BFE pattern is not working correctly because the XForm is not being
-applied.
+def IMMZeroBasedBitfieldMask : PatLeaf <(imm), [{
+ return isMask_32(N->getZExtValue());
+}]>;
-def legalshift32 : ImmLeaf <i32, [{return Imm >=0 && Imm < 32;}]>;
-def bfemask : PatLeaf <(imm), [{return isMask_32(N->getZExtValue());}],
- SDNodeXForm<imm, [{ return CurDAG->getTargetConstant(CountTrailingOnes_32(N->getZExtValue()), MVT::i32);}]>>;
+def IMMPopCount : SDNodeXForm<imm, [{
+ return CurDAG->getTargetConstant(countPopulation(N->getZExtValue()),
+ MVT::i32);
+}]>;
-class BFEPattern <Instruction BFE> : Pat <
- (and (srl i32:$x, legalshift32:$y), bfemask:$z),
- (BFE $x, $y, $z)
+class BFEPattern <Instruction BFE, Instruction MOV> : Pat <
+ (i32 (and (i32 (srl i32:$src, i32:$rshift)), IMMZeroBasedBitfieldMask:$mask)),
+ (BFE $src, $rshift, (MOV (i32 (IMMPopCount $mask))))
>;
-*/
-
// rotr pattern
class ROTRPattern <Instruction BIT_ALIGN> : Pat <
(rotr i32:$src0, i32:$src1),
// 24-bit arithmetic patterns
def umul24 : PatFrag <(ops node:$x, node:$y), (mul node:$x, node:$y)>;
+// Special conversion patterns
+
+def cvt_rpi_i32_f32 : PatFrag <
+ (ops node:$src),
+ (fp_to_sint (ffloor (fadd $src, FP_HALF))),
+ [{ (void) N; return TM.Options.NoNaNsFPMath; }]
+>;
+
+def cvt_flr_i32_f32 : PatFrag <
+ (ops node:$src),
+ (fp_to_sint (ffloor $src)),
+ [{ (void)N; return TM.Options.NoNaNsFPMath; }]
+>;
+
/*
class UMUL24Pattern <Instruction UMUL24> : Pat <
(mul U24:$x, U24:$y),
(RcpInst $src)
>;
-multiclass RsqPat<Instruction RsqInst, ValueType vt> {
- def : Pat <
- (fdiv FP_ONE, (fsqrt vt:$src)),
- (RsqInst $src)
- >;
-
- def : Pat <
- (AMDGPUrcp (fsqrt vt:$src)),
- (RsqInst $src)
- >;
-}
+class RsqPat<Instruction RsqInst, ValueType vt> : Pat <
+ (AMDGPUrcp (fsqrt vt:$src)),
+ (RsqInst $src)
+>;
include "R600Instructions.td"
include "R700Instructions.td"