diff options
Diffstat (limited to 'lib/Target/R600/AMDGPUInstructions.td')
| -rw-r--r-- | lib/Target/R600/AMDGPUInstructions.td | 97 |
1 files changed, 88 insertions, 9 deletions
diff --git a/lib/Target/R600/AMDGPUInstructions.td b/lib/Target/R600/AMDGPUInstructions.td index 80bdf5b..b86b781 100644 --- a/lib/Target/R600/AMDGPUInstructions.td +++ b/lib/Target/R600/AMDGPUInstructions.td @@ -49,6 +49,11 @@ def u8imm : Operand<i8> { let PrintMethod = "printU8ImmOperand"; } +//===--------------------------------------------------------------------===// +// Custom Operands +//===--------------------------------------------------------------------===// +def brtarget : Operand<OtherVT>; + //===----------------------------------------------------------------------===// // PatLeafs for floating-point comparisons //===----------------------------------------------------------------------===// @@ -127,6 +132,21 @@ def COND_NULL : PatLeaf < // Load/Store Pattern Fragments //===----------------------------------------------------------------------===// +def global_store : PatFrag<(ops node:$val, node:$ptr), + (store node:$val, node:$ptr), [{ + return isGlobalStore(dyn_cast<StoreSDNode>(N)); +}]>; + +// Global address space loads +def global_load : PatFrag<(ops node:$ptr), (load node:$ptr), [{ + return isGlobalLoad(dyn_cast<LoadSDNode>(N)); +}]>; + +// Constant address space loads +def constant_load : PatFrag<(ops node:$ptr), (load node:$ptr), [{ + return isConstantLoad(dyn_cast<LoadSDNode>(N), -1); +}]>; + def az_extload : PatFrag<(ops node:$ptr), (unindexedload node:$ptr), [{ LoadSDNode *L = cast<LoadSDNode>(N); return L->getExtensionType() == ISD::ZEXTLOAD || @@ -232,26 +252,55 @@ def local_load : PatFrag<(ops node:$ptr), (load node:$ptr), [{ return isLocalLoad(dyn_cast<LoadSDNode>(N)); }]>; -def atomic_load_add_local : PatFrag<(ops node:$ptr, node:$value), - (atomic_load_add node:$ptr, node:$value), [{ - return dyn_cast<MemSDNode>(N)->getAddressSpace() == AMDGPUAS::LOCAL_ADDRESS; -}]>; -def atomic_load_sub_local : PatFrag<(ops node:$ptr, node:$value), - (atomic_load_sub node:$ptr, node:$value), [{ - return dyn_cast<MemSDNode>(N)->getAddressSpace() == AMDGPUAS::LOCAL_ADDRESS; +class local_binary_atomic_op<SDNode atomic_op> : + PatFrag<(ops node:$ptr, node:$value), + (atomic_op node:$ptr, node:$value), [{ + return cast<MemSDNode>(N)->getAddressSpace() == AMDGPUAS::LOCAL_ADDRESS; }]>; + +def atomic_swap_local : local_binary_atomic_op<atomic_swap>; +def atomic_load_add_local : local_binary_atomic_op<atomic_load_add>; +def atomic_load_sub_local : local_binary_atomic_op<atomic_load_sub>; +def atomic_load_and_local : local_binary_atomic_op<atomic_load_and>; +def atomic_load_or_local : local_binary_atomic_op<atomic_load_or>; +def atomic_load_xor_local : local_binary_atomic_op<atomic_load_xor>; +def atomic_load_nand_local : local_binary_atomic_op<atomic_load_nand>; +def atomic_load_min_local : local_binary_atomic_op<atomic_load_min>; +def atomic_load_max_local : local_binary_atomic_op<atomic_load_max>; +def atomic_load_umin_local : local_binary_atomic_op<atomic_load_umin>; +def atomic_load_umax_local : local_binary_atomic_op<atomic_load_umax>; + def mskor_global : PatFrag<(ops node:$val, node:$ptr), (AMDGPUstore_mskor node:$val, node:$ptr), [{ return dyn_cast<MemSDNode>(N)->getAddressSpace() == AMDGPUAS::GLOBAL_ADDRESS; }]>; +def atomic_cmp_swap_32_local : + PatFrag<(ops node:$ptr, node:$cmp, node:$swap), + (atomic_cmp_swap node:$ptr, node:$cmp, node:$swap), [{ + AtomicSDNode *AN = cast<AtomicSDNode>(N); + return AN->getMemoryVT() == MVT::i32 && + AN->getAddressSpace() == AMDGPUAS::LOCAL_ADDRESS; +}]>; + +def atomic_cmp_swap_64_local : + PatFrag<(ops node:$ptr, node:$cmp, node:$swap), + (atomic_cmp_swap node:$ptr, node:$cmp, node:$swap), [{ + AtomicSDNode *AN = cast<AtomicSDNode>(N); + return AN->getMemoryVT() == MVT::i64 && + AN->getAddressSpace() == AMDGPUAS::LOCAL_ADDRESS; +}]>; + + class Constants { int TWO_PI = 0x40c90fdb; int PI = 0x40490fdb; int TWO_PI_INV = 0x3e22f983; int FP_UINT_MAX_PLUS_1 = 0x4f800000; // 1 << 32 in floating point encoding +int FP32_NEG_ONE = 0xbf800000; +int FP32_ONE = 0x3f800000; } def CONST : Constants; @@ -273,7 +322,7 @@ class CLAMP <RegisterClass rc> : AMDGPUShaderInst < (outs rc:$dst), (ins rc:$src0), "CLAMP $dst, $src0", - [(set f32:$dst, (int_AMDIL_clamp f32:$src0, (f32 FP_ZERO), (f32 FP_ONE)))] + [(set f32:$dst, (AMDGPUclamp f32:$src0, (f32 FP_ZERO), (f32 FP_ONE)))] >; class FABS <RegisterClass rc> : AMDGPUShaderInst < @@ -363,7 +412,7 @@ class DwordAddrPat<ValueType vt, RegisterClass rc> : Pat < // BFI_INT patterns -multiclass BFIPatterns <Instruction BFI_INT> { +multiclass BFIPatterns <Instruction BFI_INT, Instruction LoadImm32> { // Definition from ISA doc: // (y & x) | (z & ~x) @@ -379,6 +428,19 @@ multiclass BFIPatterns <Instruction BFI_INT> { (BFI_INT $x, $y, $z) >; + def : Pat < + (fcopysign f32:$src0, f32:$src1), + (BFI_INT (LoadImm32 0x7fffffff), $src0, $src1) + >; + + def : Pat < + (f64 (fcopysign f64:$src0, f64:$src1)), + (INSERT_SUBREG (INSERT_SUBREG (f64 (IMPLICIT_DEF)), + (i32 (EXTRACT_SUBREG $src0, sub0)), sub0), + (BFI_INT (LoadImm32 0x7fffffff), + (i32 (EXTRACT_SUBREG $src0, sub1)), + (i32 (EXTRACT_SUBREG $src1, sub1))), sub1) + >; } // SHA-256 Ma patterns @@ -457,6 +519,23 @@ multiclass Expand24UBitOps<Instruction MulInst, Instruction AddInst> { >; } +class RcpPat<Instruction RcpInst, ValueType vt> : Pat < + (fdiv FP_ONE, vt:$src), + (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) + >; +} + include "R600Instructions.td" include "R700Instructions.td" include "EvergreenInstructions.td" |