//===----------------------------------------------------------------------===//
class AMDGPUReadPreloadRegisterIntrinsic
- : Intrinsic<[llvm_i32_ty], [], [IntrNoMem, IntrSpeculatable, IntrWillReturn]>;
+ : DefaultAttrsIntrinsic<[llvm_i32_ty], [], [IntrNoMem, IntrSpeculatable]>;
class AMDGPUReadPreloadRegisterIntrinsicNamed<string name>
- : Intrinsic<[llvm_i32_ty], [], [IntrNoMem, IntrSpeculatable, IntrWillReturn]>, ClangBuiltin<name>;
+ : DefaultAttrsIntrinsic<[llvm_i32_ty], [], [IntrNoMem, IntrSpeculatable]>, ClangBuiltin<name>;
// Used to tag image and resource intrinsics with information used to generate
// mem operands.
// AS 7 is PARAM_I_ADDRESS, used for kernel arguments
def int_r600_implicitarg_ptr :
ClangBuiltin<"__builtin_r600_implicitarg_ptr">,
- Intrinsic<[LLVMQualPointerType<llvm_i8_ty, 7>], [],
- [IntrNoMem, IntrSpeculatable, IntrWillReturn]>;
+ DefaultAttrsIntrinsic<[LLVMQualPointerType<llvm_i8_ty, 7>], [],
+ [IntrNoMem, IntrSpeculatable]>;
def int_r600_rat_store_typed :
// 1st parameter: Data
// 2nd parameter: Index
// 3rd parameter: Constant RAT ID
- Intrinsic<[], [llvm_v4i32_ty, llvm_v4i32_ty, llvm_i32_ty], [IntrWillReturn]>,
+ DefaultAttrsIntrinsic<[], [llvm_v4i32_ty, llvm_v4i32_ty, llvm_i32_ty], []>,
ClangBuiltin<"__builtin_r600_rat_store_typed">;
-def int_r600_recipsqrt_ieee : Intrinsic<
- [llvm_anyfloat_ty], [LLVMMatchType<0>], [IntrNoMem, IntrSpeculatable, IntrWillReturn]
+def int_r600_recipsqrt_ieee : DefaultAttrsIntrinsic<
+ [llvm_anyfloat_ty], [LLVMMatchType<0>], [IntrNoMem, IntrSpeculatable]
>;
-def int_r600_recipsqrt_clamped : Intrinsic<
- [llvm_anyfloat_ty], [LLVMMatchType<0>], [IntrNoMem, IntrSpeculatable, IntrWillReturn]
+def int_r600_recipsqrt_clamped : DefaultAttrsIntrinsic<
+ [llvm_anyfloat_ty], [LLVMMatchType<0>], [IntrNoMem, IntrSpeculatable]
>;
-def int_r600_cube : Intrinsic<
- [llvm_v4f32_ty], [llvm_v4f32_ty], [IntrNoMem, IntrSpeculatable, IntrWillReturn]
+def int_r600_cube : DefaultAttrsIntrinsic<
+ [llvm_v4f32_ty], [llvm_v4f32_ty], [IntrNoMem, IntrSpeculatable]
>;
-def int_r600_store_stream_output : Intrinsic<
- [], [llvm_v4f32_ty, llvm_i32_ty, llvm_i32_ty, llvm_i32_ty], [IntrWillReturn]
+def int_r600_store_stream_output : DefaultAttrsIntrinsic<
+ [], [llvm_v4f32_ty, llvm_i32_ty, llvm_i32_ty, llvm_i32_ty], []
>;
-class TextureIntrinsicFloatInput : Intrinsic<[llvm_v4f32_ty], [
+class TextureIntrinsicFloatInput : DefaultAttrsIntrinsic<[llvm_v4f32_ty], [
llvm_v4f32_ty, // Coord
llvm_i32_ty, // offset_x
llvm_i32_ty, // offset_y,
llvm_i32_ty, // coord_type_y
llvm_i32_ty, // coord_type_z
llvm_i32_ty], // coord_type_w
- [IntrNoMem, IntrWillReturn]
+ [IntrNoMem]
>;
-class TextureIntrinsicInt32Input : Intrinsic<[llvm_v4i32_ty], [
+class TextureIntrinsicInt32Input : DefaultAttrsIntrinsic<[llvm_v4i32_ty], [
llvm_v4i32_ty, // Coord
llvm_i32_ty, // offset_x
llvm_i32_ty, // offset_y,
llvm_i32_ty, // coord_type_y
llvm_i32_ty, // coord_type_z
llvm_i32_ty], // coord_type_w
- [IntrNoMem, IntrWillReturn]
+ [IntrNoMem]
>;
def int_r600_store_swizzle :
- Intrinsic<[], [llvm_v4f32_ty, llvm_i32_ty, llvm_i32_ty], [IntrWillReturn]
+ Intrinsic<[], [llvm_v4f32_ty, llvm_i32_ty, llvm_i32_ty], [IntrWillReturn, IntrNoCallback, IntrNoFree]
>;
def int_r600_tex : TextureIntrinsicFloatInput;
def int_r600_ddx : TextureIntrinsicFloatInput;
def int_r600_ddy : TextureIntrinsicFloatInput;
-def int_r600_dot4 : Intrinsic<[llvm_float_ty],
- [llvm_v4f32_ty, llvm_v4f32_ty], [IntrNoMem, IntrSpeculatable, IntrWillReturn]
+def int_r600_dot4 : DefaultAttrsIntrinsic<[llvm_float_ty],
+ [llvm_v4f32_ty, llvm_v4f32_ty], [IntrNoMem, IntrSpeculatable]
>;
-def int_r600_kill : Intrinsic<[], [llvm_float_ty], [IntrWillReturn]>;
+def int_r600_kill : DefaultAttrsIntrinsic<[], [llvm_float_ty], []>;
} // End TargetPrefix = "r600"
<"__builtin_amdgcn_workgroup_id">;
def int_amdgcn_dispatch_ptr :
- Intrinsic<[LLVMQualPointerType<llvm_i8_ty, 4>], [],
- [Align<RetIndex, 4>, IntrNoMem, IntrSpeculatable, IntrWillReturn]>;
+ DefaultAttrsIntrinsic<[LLVMQualPointerType<llvm_i8_ty, 4>], [],
+ [Align<RetIndex, 4>, IntrNoMem, IntrSpeculatable]>;
def int_amdgcn_queue_ptr :
ClangBuiltin<"__builtin_amdgcn_queue_ptr">,
- Intrinsic<[LLVMQualPointerType<llvm_i8_ty, 4>], [],
- [Align<RetIndex, 4>, IntrNoMem, IntrSpeculatable, IntrWillReturn]>;
+ DefaultAttrsIntrinsic<[LLVMQualPointerType<llvm_i8_ty, 4>], [],
+ [Align<RetIndex, 4>, IntrNoMem, IntrSpeculatable]>;
def int_amdgcn_kernarg_segment_ptr :
ClangBuiltin<"__builtin_amdgcn_kernarg_segment_ptr">,
- Intrinsic<[LLVMQualPointerType<llvm_i8_ty, 4>], [],
- [Align<RetIndex, 4>, IntrNoMem, IntrSpeculatable, IntrWillReturn]>;
+ DefaultAttrsIntrinsic<[LLVMQualPointerType<llvm_i8_ty, 4>], [],
+ [Align<RetIndex, 4>, IntrNoMem, IntrSpeculatable]>;
def int_amdgcn_implicitarg_ptr :
ClangBuiltin<"__builtin_amdgcn_implicitarg_ptr">,
- Intrinsic<[LLVMQualPointerType<llvm_i8_ty, 4>], [],
- [Align<RetIndex, 4>, IntrNoMem, IntrSpeculatable, IntrWillReturn]>;
+ DefaultAttrsIntrinsic<[LLVMQualPointerType<llvm_i8_ty, 4>], [],
+ [Align<RetIndex, 4>, IntrNoMem, IntrSpeculatable]>;
def int_amdgcn_groupstaticsize :
ClangBuiltin<"__builtin_amdgcn_groupstaticsize">,
- Intrinsic<[llvm_i32_ty], [], [IntrNoMem, IntrSpeculatable, IntrWillReturn]>;
+ DefaultAttrsIntrinsic<[llvm_i32_ty], [], [IntrNoMem, IntrSpeculatable]>;
def int_amdgcn_dispatch_id :
ClangBuiltin<"__builtin_amdgcn_dispatch_id">,
- Intrinsic<[llvm_i64_ty], [], [IntrNoMem, IntrSpeculatable, IntrWillReturn]>;
+ DefaultAttrsIntrinsic<[llvm_i64_ty], [], [IntrNoMem, IntrSpeculatable]>;
// For internal use. Coordinates LDS lowering between IR transform and backend.
def int_amdgcn_lds_kernel_id :
- Intrinsic<[llvm_i32_ty], [], [IntrNoMem, IntrSpeculatable, IntrWillReturn]>;
+ DefaultAttrsIntrinsic<[llvm_i32_ty], [], [IntrNoMem, IntrSpeculatable]>;
def int_amdgcn_implicit_buffer_ptr :
ClangBuiltin<"__builtin_amdgcn_implicit_buffer_ptr">,
- Intrinsic<[LLVMQualPointerType<llvm_i8_ty, 4>], [],
- [Align<RetIndex, 4>, IntrNoMem, IntrSpeculatable, IntrWillReturn]>;
+ DefaultAttrsIntrinsic<[LLVMQualPointerType<llvm_i8_ty, 4>], [],
+ [Align<RetIndex, 4>, IntrNoMem, IntrSpeculatable]>;
// Set EXEC to the 64-bit value given.
// This is always moved to the beginning of the basic block.
def int_amdgcn_wavefrontsize :
ClangBuiltin<"__builtin_amdgcn_wavefrontsize">,
- Intrinsic<[llvm_i32_ty], [], [IntrNoMem, IntrSpeculatable, IntrWillReturn]>;
+ DefaultAttrsIntrinsic<[llvm_i32_ty], [], [IntrNoMem, IntrSpeculatable]>;
//===----------------------------------------------------------------------===//
[ImmArg<ArgIndex<0>>, IntrNoMem, IntrHasSideEffects]>;
def int_amdgcn_s_barrier : ClangBuiltin<"__builtin_amdgcn_s_barrier">,
- Intrinsic<[], [], [IntrNoMem, IntrHasSideEffects, IntrConvergent, IntrWillReturn]>;
+ Intrinsic<[], [], [IntrNoMem, IntrHasSideEffects, IntrConvergent, IntrWillReturn, IntrNoCallback, IntrNoFree]>;
def int_amdgcn_wave_barrier : ClangBuiltin<"__builtin_amdgcn_wave_barrier">,
- Intrinsic<[], [], [IntrNoMem, IntrHasSideEffects, IntrConvergent, IntrWillReturn]>;
+ Intrinsic<[], [], [IntrNoMem, IntrHasSideEffects, IntrConvergent, IntrWillReturn, IntrNoCallback, IntrNoFree]>;
// The 1st parameter is a mask for the types of instructions that may be allowed
// to cross the SCHED_BARRIER during scheduling.
// MASK = 0x0000 0200: ALL DS write instructions may be scheduled across SCHED_BARRIER.
def int_amdgcn_sched_barrier : ClangBuiltin<"__builtin_amdgcn_sched_barrier">,
Intrinsic<[], [llvm_i32_ty], [ImmArg<ArgIndex<0>>, IntrNoMem, IntrHasSideEffects, IntrConvergent,
- IntrWillReturn]>;
+ IntrWillReturn, IntrNoCallback, IntrNoFree]>;
// The first parameter is a mask that determines the types of instructions that
// you would like to synchronize around and add to a scheduling group. The
def int_amdgcn_sched_group_barrier : ClangBuiltin<"__builtin_amdgcn_sched_group_barrier">,
Intrinsic<[], [llvm_i32_ty, llvm_i32_ty, llvm_i32_ty],
[ImmArg<ArgIndex<0>>, ImmArg<ArgIndex<1>>, ImmArg<ArgIndex<2>>, IntrNoMem, IntrHasSideEffects,
- IntrConvergent, IntrWillReturn]>;
+ IntrConvergent, IntrWillReturn, IntrNoCallback, IntrNoFree]>;
// Scheduler optimization hint.
// MASK = 0: Small gemm opt
def int_amdgcn_iglp_opt : ClangBuiltin<"__builtin_amdgcn_iglp_opt">,
Intrinsic<[], [llvm_i32_ty], [ImmArg<ArgIndex<0>>, IntrNoMem, IntrHasSideEffects, IntrConvergent,
- IntrWillReturn]>;
+ IntrWillReturn, IntrNoCallback, IntrNoFree]>;
def int_amdgcn_s_waitcnt : ClangBuiltin<"__builtin_amdgcn_s_waitcnt">,
- Intrinsic<[], [llvm_i32_ty], [ImmArg<ArgIndex<0>>, IntrNoMem, IntrHasSideEffects, IntrWillReturn]>;
+ Intrinsic<[], [llvm_i32_ty], [ImmArg<ArgIndex<0>>, IntrNoMem, IntrHasSideEffects, IntrWillReturn, IntrNoCallback, IntrNoFree]>;
-def int_amdgcn_div_scale : Intrinsic<
+def int_amdgcn_div_scale : DefaultAttrsIntrinsic<
// 1st parameter: Numerator
// 2nd parameter: Denominator
// 3rd parameter: Select quotient. Must equal Numerator or Denominator.
// (0 = Denominator, 1 = Numerator).
[llvm_anyfloat_ty, llvm_i1_ty],
[LLVMMatchType<0>, LLVMMatchType<0>, llvm_i1_ty],
- [IntrNoMem, IntrSpeculatable, ImmArg<ArgIndex<2>>, IntrWillReturn]
+ [IntrNoMem, IntrSpeculatable, ImmArg<ArgIndex<2>>]
>;
-def int_amdgcn_div_fmas : Intrinsic<[llvm_anyfloat_ty],
+def int_amdgcn_div_fmas : DefaultAttrsIntrinsic<[llvm_anyfloat_ty],
[LLVMMatchType<0>, LLVMMatchType<0>, LLVMMatchType<0>, llvm_i1_ty],
- [IntrNoMem, IntrSpeculatable, IntrWillReturn]
+ [IntrNoMem, IntrSpeculatable]
>;
-def int_amdgcn_div_fixup : Intrinsic<[llvm_anyfloat_ty],
+def int_amdgcn_div_fixup : DefaultAttrsIntrinsic<[llvm_anyfloat_ty],
[LLVMMatchType<0>, LLVMMatchType<0>, LLVMMatchType<0>],
- [IntrNoMem, IntrSpeculatable, IntrWillReturn]
+ [IntrNoMem, IntrSpeculatable]
>;
// Look Up 2.0 / pi src0 with segment select src1[4:0]
-def int_amdgcn_trig_preop : Intrinsic<
+def int_amdgcn_trig_preop : DefaultAttrsIntrinsic<
[llvm_anyfloat_ty], [LLVMMatchType<0>, llvm_i32_ty],
- [IntrNoMem, IntrSpeculatable, IntrWillReturn]
+ [IntrNoMem, IntrSpeculatable]
>;
-def int_amdgcn_sin : Intrinsic<
+def int_amdgcn_sin : DefaultAttrsIntrinsic<
[llvm_anyfloat_ty], [LLVMMatchType<0>],
- [IntrNoMem, IntrSpeculatable, IntrWillReturn]
+ [IntrNoMem, IntrSpeculatable]
>;
-def int_amdgcn_cos : Intrinsic<
- [llvm_anyfloat_ty], [LLVMMatchType<0>], [IntrNoMem, IntrSpeculatable, IntrWillReturn]
+def int_amdgcn_cos : DefaultAttrsIntrinsic<
+ [llvm_anyfloat_ty], [LLVMMatchType<0>], [IntrNoMem, IntrSpeculatable]
>;
-def int_amdgcn_log_clamp : Intrinsic<
- [llvm_anyfloat_ty], [LLVMMatchType<0>], [IntrNoMem, IntrSpeculatable, IntrWillReturn]
+def int_amdgcn_log_clamp : DefaultAttrsIntrinsic<
+ [llvm_anyfloat_ty], [LLVMMatchType<0>], [IntrNoMem, IntrSpeculatable]
>;
def int_amdgcn_fmul_legacy : ClangBuiltin<"__builtin_amdgcn_fmul_legacy">,
- Intrinsic<[llvm_float_ty], [llvm_float_ty, llvm_float_ty],
- [IntrNoMem, IntrSpeculatable, IntrWillReturn, Commutative]
+ DefaultAttrsIntrinsic<[llvm_float_ty], [llvm_float_ty, llvm_float_ty],
+ [IntrNoMem, IntrSpeculatable, Commutative]
>;
// Fused single-precision multiply-add with legacy behaviour for the multiply,
// v_fmac_legacy_f32 instructions. (Note that v_fma_legacy_f16 is unrelated and
// has a completely different kind of legacy behaviour.)
def int_amdgcn_fma_legacy :
- Intrinsic<[llvm_float_ty], [llvm_float_ty, llvm_float_ty, llvm_float_ty],
- [IntrNoMem, IntrSpeculatable, IntrWillReturn, Commutative]
+ DefaultAttrsIntrinsic<[llvm_float_ty], [llvm_float_ty, llvm_float_ty, llvm_float_ty],
+ [IntrNoMem, IntrSpeculatable, Commutative]
>;
-def int_amdgcn_rcp : Intrinsic<
- [llvm_anyfloat_ty], [LLVMMatchType<0>], [IntrNoMem, IntrSpeculatable, IntrWillReturn]
+def int_amdgcn_rcp : DefaultAttrsIntrinsic<
+ [llvm_anyfloat_ty], [LLVMMatchType<0>], [IntrNoMem, IntrSpeculatable]
>;
def int_amdgcn_rcp_legacy : ClangBuiltin<"__builtin_amdgcn_rcp_legacy">,
- Intrinsic<[llvm_float_ty], [llvm_float_ty],
- [IntrNoMem, IntrSpeculatable, IntrWillReturn]
+ DefaultAttrsIntrinsic<[llvm_float_ty], [llvm_float_ty],
+ [IntrNoMem, IntrSpeculatable]
>;
-def int_amdgcn_sqrt : Intrinsic<
- [llvm_anyfloat_ty], [LLVMMatchType<0>], [IntrNoMem, IntrSpeculatable, IntrWillReturn]
+def int_amdgcn_sqrt : DefaultAttrsIntrinsic<
+ [llvm_anyfloat_ty], [LLVMMatchType<0>], [IntrNoMem, IntrSpeculatable]
>;
-def int_amdgcn_rsq : Intrinsic<
- [llvm_anyfloat_ty], [LLVMMatchType<0>], [IntrNoMem, IntrSpeculatable, IntrWillReturn]
+def int_amdgcn_rsq : DefaultAttrsIntrinsic<
+ [llvm_anyfloat_ty], [LLVMMatchType<0>], [IntrNoMem, IntrSpeculatable]
>;
def int_amdgcn_rsq_legacy : ClangBuiltin<"__builtin_amdgcn_rsq_legacy">,
- Intrinsic<
- [llvm_float_ty], [llvm_float_ty], [IntrNoMem, IntrSpeculatable, IntrWillReturn]
+ DefaultAttrsIntrinsic<
+ [llvm_float_ty], [llvm_float_ty], [IntrNoMem, IntrSpeculatable]
>;
// out = 1.0 / sqrt(a) result clamped to +/- max_float.
-def int_amdgcn_rsq_clamp : Intrinsic<
- [llvm_anyfloat_ty], [LLVMMatchType<0>], [IntrNoMem, IntrSpeculatable, IntrWillReturn]>;
+def int_amdgcn_rsq_clamp : DefaultAttrsIntrinsic<
+ [llvm_anyfloat_ty], [LLVMMatchType<0>], [IntrNoMem, IntrSpeculatable]>;
// For int_amdgcn_ldexp_f16, only the low 16 bits of the i32 src1 operand will used.
-def int_amdgcn_ldexp : Intrinsic<
+def int_amdgcn_ldexp : DefaultAttrsIntrinsic<
[llvm_anyfloat_ty], [LLVMMatchType<0>, llvm_i32_ty],
- [IntrNoMem, IntrSpeculatable, IntrWillReturn]
+ [IntrNoMem, IntrSpeculatable]
>;
-def int_amdgcn_frexp_mant : Intrinsic<
- [llvm_anyfloat_ty], [LLVMMatchType<0>], [IntrNoMem, IntrSpeculatable, IntrWillReturn]
+def int_amdgcn_frexp_mant : DefaultAttrsIntrinsic<
+ [llvm_anyfloat_ty], [LLVMMatchType<0>], [IntrNoMem, IntrSpeculatable]
>;
-def int_amdgcn_frexp_exp : Intrinsic<
- [llvm_anyint_ty], [llvm_anyfloat_ty], [IntrNoMem, IntrSpeculatable, IntrWillReturn]
+def int_amdgcn_frexp_exp : DefaultAttrsIntrinsic<
+ [llvm_anyint_ty], [llvm_anyfloat_ty], [IntrNoMem, IntrSpeculatable]
>;
// v_fract is buggy on SI/CI. It mishandles infinities, may return 1.0
// and always uses rtz, so is not suitable for implementing the OpenCL
// fract function. It should be ok on VI.
-def int_amdgcn_fract : Intrinsic<
- [llvm_anyfloat_ty], [LLVMMatchType<0>], [IntrNoMem, IntrSpeculatable, IntrWillReturn]
+def int_amdgcn_fract : DefaultAttrsIntrinsic<
+ [llvm_anyfloat_ty], [LLVMMatchType<0>], [IntrNoMem, IntrSpeculatable]
>;
def int_amdgcn_cvt_pkrtz : ClangBuiltin<"__builtin_amdgcn_cvt_pkrtz">,
- Intrinsic<[llvm_v2f16_ty], [llvm_float_ty, llvm_float_ty],
- [IntrNoMem, IntrSpeculatable, IntrWillReturn]
+ DefaultAttrsIntrinsic<[llvm_v2f16_ty], [llvm_float_ty, llvm_float_ty],
+ [IntrNoMem, IntrSpeculatable]
>;
def int_amdgcn_cvt_pknorm_i16 :
ClangBuiltin<"__builtin_amdgcn_cvt_pknorm_i16">,
- Intrinsic<[llvm_v2i16_ty], [llvm_float_ty, llvm_float_ty],
- [IntrNoMem, IntrSpeculatable, IntrWillReturn]
+ DefaultAttrsIntrinsic<[llvm_v2i16_ty], [llvm_float_ty, llvm_float_ty],
+ [IntrNoMem, IntrSpeculatable]
>;
def int_amdgcn_cvt_pknorm_u16 :
ClangBuiltin<"__builtin_amdgcn_cvt_pknorm_u16">,
- Intrinsic<[llvm_v2i16_ty], [llvm_float_ty, llvm_float_ty],
- [IntrNoMem, IntrSpeculatable, IntrWillReturn]
+ DefaultAttrsIntrinsic<[llvm_v2i16_ty], [llvm_float_ty, llvm_float_ty],
+ [IntrNoMem, IntrSpeculatable]
>;
def int_amdgcn_cvt_pk_i16 :
ClangBuiltin<"__builtin_amdgcn_cvt_pk_i16">,
- Intrinsic<
+ DefaultAttrsIntrinsic<
[llvm_v2i16_ty], [llvm_i32_ty, llvm_i32_ty],
- [IntrNoMem, IntrSpeculatable, IntrWillReturn]
+ [IntrNoMem, IntrSpeculatable]
>;
def int_amdgcn_cvt_pk_u16 : ClangBuiltin<"__builtin_amdgcn_cvt_pk_u16">,
- Intrinsic<[llvm_v2i16_ty], [llvm_i32_ty, llvm_i32_ty],
- [IntrNoMem, IntrSpeculatable, IntrWillReturn]
+ DefaultAttrsIntrinsic<[llvm_v2i16_ty], [llvm_i32_ty, llvm_i32_ty],
+ [IntrNoMem, IntrSpeculatable]
>;
-def int_amdgcn_class : Intrinsic<
+def int_amdgcn_class : DefaultAttrsIntrinsic<
[llvm_i1_ty], [llvm_anyfloat_ty, llvm_i32_ty],
- [IntrNoMem, IntrSpeculatable, IntrWillReturn]
+ [IntrNoMem, IntrSpeculatable]
>;
def int_amdgcn_fmed3 : ClangBuiltin<"__builtin_amdgcn_fmed3">,
- Intrinsic<[llvm_anyfloat_ty],
+ DefaultAttrsIntrinsic<[llvm_anyfloat_ty],
[LLVMMatchType<0>, LLVMMatchType<0>, LLVMMatchType<0>],
- [IntrNoMem, IntrSpeculatable, IntrWillReturn]
+ [IntrNoMem, IntrSpeculatable]
>;
def int_amdgcn_cubeid : ClangBuiltin<"__builtin_amdgcn_cubeid">,
- Intrinsic<[llvm_float_ty],
+ DefaultAttrsIntrinsic<[llvm_float_ty],
[llvm_float_ty, llvm_float_ty, llvm_float_ty],
- [IntrNoMem, IntrSpeculatable, IntrWillReturn]
+ [IntrNoMem, IntrSpeculatable]
>;
def int_amdgcn_cubema : ClangBuiltin<"__builtin_amdgcn_cubema">,
- Intrinsic<[llvm_float_ty],
+ DefaultAttrsIntrinsic<[llvm_float_ty],
[llvm_float_ty, llvm_float_ty, llvm_float_ty],
- [IntrNoMem, IntrSpeculatable, IntrWillReturn]
+ [IntrNoMem, IntrSpeculatable]
>;
def int_amdgcn_cubesc : ClangBuiltin<"__builtin_amdgcn_cubesc">,
- Intrinsic<[llvm_float_ty],
+ DefaultAttrsIntrinsic<[llvm_float_ty],
[llvm_float_ty, llvm_float_ty, llvm_float_ty],
- [IntrNoMem, IntrSpeculatable, IntrWillReturn]
+ [IntrNoMem, IntrSpeculatable]
>;
def int_amdgcn_cubetc : ClangBuiltin<"__builtin_amdgcn_cubetc">,
- Intrinsic<[llvm_float_ty],
+ DefaultAttrsIntrinsic<[llvm_float_ty],
[llvm_float_ty, llvm_float_ty, llvm_float_ty],
- [IntrNoMem, IntrSpeculatable, IntrWillReturn]
+ [IntrNoMem, IntrSpeculatable]
>;
// v_ffbh_i32, as opposed to v_ffbh_u32. For v_ffbh_u32, llvm.ctlz
// should be used.
def int_amdgcn_sffbh :
- Intrinsic<[llvm_anyint_ty], [LLVMMatchType<0>],
- [IntrNoMem, IntrSpeculatable, IntrWillReturn]
+ DefaultAttrsIntrinsic<[llvm_anyint_ty], [LLVMMatchType<0>],
+ [IntrNoMem, IntrSpeculatable]
>;
// v_mad_f32|f16/v_mac_f32|f16, selected regardless of denorm support.
def int_amdgcn_fmad_ftz :
- Intrinsic<[llvm_anyfloat_ty],
+ DefaultAttrsIntrinsic<[llvm_anyfloat_ty],
[LLVMMatchType<0>, LLVMMatchType<0>, LLVMMatchType<0>],
- [IntrNoMem, IntrSpeculatable, IntrWillReturn]
+ [IntrNoMem, IntrSpeculatable]
>;
// Fields should mirror atomicrmw
llvm_i32_ty, // scope
llvm_i1_ty], // isVolatile
[IntrArgMemOnly, IntrWillReturn, NoCapture<ArgIndex<0>>,
- ImmArg<ArgIndex<2>>, ImmArg<ArgIndex<3>>, ImmArg<ArgIndex<4>>], "",
+ ImmArg<ArgIndex<2>>, ImmArg<ArgIndex<3>>, ImmArg<ArgIndex<4>>, IntrNoCallback, IntrNoFree], "",
[SDNPMemOperand]
>;
llvm_i32_ty, // scope
llvm_i1_ty], // isVolatile
[IntrArgMemOnly, IntrWillReturn, NoCapture<ArgIndex<0>>,
- ImmArg<ArgIndex<2>>, ImmArg<ArgIndex<3>>, ImmArg<ArgIndex<4>>]
+ ImmArg<ArgIndex<2>>, ImmArg<ArgIndex<3>>, ImmArg<ArgIndex<4>>, IntrNoCallback, IntrNoFree]
>;
// FIXME: The m0 argument should be moved after the normal arguments
llvm_i1_ty], // wave done, set to 1 for the last ordered instruction
[IntrWillReturn, NoCapture<ArgIndex<0>>,
ImmArg<ArgIndex<2>>, ImmArg<ArgIndex<3>>, ImmArg<ArgIndex<4>>,
- ImmArg<ArgIndex<5>>, ImmArg<ArgIndex<6>>, ImmArg<ArgIndex<7>>
+ ImmArg<ArgIndex<5>>, ImmArg<ArgIndex<6>>, ImmArg<ArgIndex<7>>, IntrNoCallback, IntrNoFree
]
>;
[llvm_anyptr_ty, // LDS or GDS ptr
llvm_i1_ty], // isVolatile
[IntrConvergent, IntrWillReturn, IntrArgMemOnly,
- NoCapture<ArgIndex<0>>, ImmArg<ArgIndex<1>>],
+ NoCapture<ArgIndex<0>>, ImmArg<ArgIndex<1>>, IntrNoCallback, IntrNoFree],
"",
[SDNPMemOperand]
>;
defset list<AMDGPURsrcIntrinsic> AMDGPUBufferIntrinsics = {
-class AMDGPUBufferLoad<LLVMType data_ty = llvm_any_ty> : Intrinsic <
+class AMDGPUBufferLoad<LLVMType data_ty = llvm_any_ty> : DefaultAttrsIntrinsic <
[data_ty],
[llvm_v4i32_ty, // rsrc(SGPR)
llvm_i32_ty, // vindex(VGPR)
llvm_i32_ty, // offset(SGPR/VGPR/imm)
llvm_i1_ty, // glc(imm)
llvm_i1_ty], // slc(imm)
- [IntrReadMem, IntrWillReturn,
- ImmArg<ArgIndex<3>>, ImmArg<ArgIndex<4>>], "", [SDNPMemOperand]>,
+ [IntrReadMem, ImmArg<ArgIndex<3>>, ImmArg<ArgIndex<4>>], "", [SDNPMemOperand]>,
AMDGPURsrcIntrinsic<0>;
def int_amdgcn_buffer_load_format : AMDGPUBufferLoad<llvm_anyfloat_ty>;
def int_amdgcn_buffer_load : AMDGPUBufferLoad;
// Generate a buffer_load instruction that may be optimized to s_buffer_load if
// the offset argument is uniform.
-def int_amdgcn_s_buffer_load : Intrinsic <
+def int_amdgcn_s_buffer_load : DefaultAttrsIntrinsic <
[llvm_any_ty],
[llvm_v4i32_ty, // rsrc(SGPR)
llvm_i32_ty, // byte offset
llvm_i32_ty], // cachepolicy(imm; bit 0 = glc, bit 2 = dlc)
- [IntrNoMem, IntrWillReturn, ImmArg<ArgIndex<2>>]>,
+ [IntrNoMem, ImmArg<ArgIndex<2>>]>,
AMDGPURsrcIntrinsic<0>;
-class AMDGPUBufferStore<LLVMType data_ty = llvm_any_ty> : Intrinsic <
+class AMDGPUBufferStore<LLVMType data_ty = llvm_any_ty> : DefaultAttrsIntrinsic <
[],
[data_ty, // vdata(VGPR)
llvm_v4i32_ty, // rsrc(SGPR)
llvm_i32_ty, // offset(SGPR/VGPR/imm)
llvm_i1_ty, // glc(imm)
llvm_i1_ty], // slc(imm)
- [IntrWriteMem, IntrWillReturn,
- ImmArg<ArgIndex<4>>, ImmArg<ArgIndex<5>>], "", [SDNPMemOperand]>,
+ [IntrWriteMem, ImmArg<ArgIndex<4>>, ImmArg<ArgIndex<5>>], "", [SDNPMemOperand]>,
AMDGPURsrcIntrinsic<1>;
def int_amdgcn_buffer_store_format : AMDGPUBufferStore<llvm_anyfloat_ty>;
def int_amdgcn_buffer_store : AMDGPUBufferStore;
// and swizzling changes depending on whether idxen is set in the instruction.
// These new instrinsics also keep the offset and soffset arguments separate as
// they behave differently in bounds checking and swizzling.
-class AMDGPURawBufferLoad<LLVMType data_ty = llvm_any_ty> : Intrinsic <
+class AMDGPURawBufferLoad<LLVMType data_ty = llvm_any_ty> : DefaultAttrsIntrinsic <
[data_ty],
[llvm_v4i32_ty, // rsrc(SGPR)
llvm_i32_ty, // offset(VGPR/imm, included in bounds checking and swizzling)
// bit 1 = slc,
// bit 2 = dlc on gfx10+),
// swizzled buffer (bit 3 = swz))
- [IntrReadMem, IntrWillReturn, ImmArg<ArgIndex<3>>], "", [SDNPMemOperand]>,
+ [IntrReadMem, ImmArg<ArgIndex<3>>], "", [SDNPMemOperand]>,
AMDGPURsrcIntrinsic<0>;
def int_amdgcn_raw_buffer_load_format : AMDGPURawBufferLoad<llvm_anyfloat_ty>;
def int_amdgcn_raw_buffer_load : AMDGPURawBufferLoad;
-class AMDGPUStructBufferLoad<LLVMType data_ty = llvm_any_ty> : Intrinsic <
+class AMDGPUStructBufferLoad<LLVMType data_ty = llvm_any_ty> : DefaultAttrsIntrinsic <
[data_ty],
[llvm_v4i32_ty, // rsrc(SGPR)
llvm_i32_ty, // vindex(VGPR)
// bit 1 = slc,
// bit 2 = dlc on gfx10+),
// swizzled buffer (bit 3 = swz))
- [IntrReadMem, IntrWillReturn, ImmArg<ArgIndex<4>>], "", [SDNPMemOperand]>,
+ [IntrReadMem, ImmArg<ArgIndex<4>>], "", [SDNPMemOperand]>,
AMDGPURsrcIntrinsic<0>;
def int_amdgcn_struct_buffer_load_format : AMDGPUStructBufferLoad;
def int_amdgcn_struct_buffer_load : AMDGPUStructBufferLoad;
-class AMDGPURawBufferStore<LLVMType data_ty = llvm_any_ty> : Intrinsic <
+class AMDGPURawBufferStore<LLVMType data_ty = llvm_any_ty> : DefaultAttrsIntrinsic <
[],
[data_ty, // vdata(VGPR)
llvm_v4i32_ty, // rsrc(SGPR)
// bit 1 = slc,
// bit 2 = dlc on gfx10+),
// swizzled buffer (bit 3 = swz))
- [IntrWriteMem, IntrWillReturn, ImmArg<ArgIndex<4>>], "", [SDNPMemOperand]>,
+ [IntrWriteMem, ImmArg<ArgIndex<4>>], "", [SDNPMemOperand]>,
AMDGPURsrcIntrinsic<1>;
def int_amdgcn_raw_buffer_store_format : AMDGPURawBufferStore<llvm_anyfloat_ty>;
def int_amdgcn_raw_buffer_store : AMDGPURawBufferStore;
-class AMDGPUStructBufferStore<LLVMType data_ty = llvm_any_ty> : Intrinsic <
+class AMDGPUStructBufferStore<LLVMType data_ty = llvm_any_ty> : DefaultAttrsIntrinsic <
[],
[data_ty, // vdata(VGPR)
llvm_v4i32_ty, // rsrc(SGPR)
// bit 1 = slc,
// bit 2 = dlc on gfx10+),
// swizzled buffer (bit 3 = swz))
- [IntrWriteMem, IntrWillReturn, ImmArg<ArgIndex<5>>], "", [SDNPMemOperand]>,
+ [IntrWriteMem, ImmArg<ArgIndex<5>>], "", [SDNPMemOperand]>,
AMDGPURsrcIntrinsic<1>;
def int_amdgcn_struct_buffer_store_format : AMDGPUStructBufferStore;
def int_amdgcn_struct_buffer_store : AMDGPUStructBufferStore;
llvm_i32_ty, // offset(VGPR/imm, included in bounds checking and swizzling)
llvm_i32_ty, // soffset(SGPR/imm, excluded from bounds checking and swizzling)
llvm_i32_ty], // cachepolicy(imm; bit 1 = slc)
- [ImmArg<ArgIndex<4>>, IntrWillReturn], "", [SDNPMemOperand]>,
+ [ImmArg<ArgIndex<4>>, IntrWillReturn, IntrNoCallback, IntrNoFree], "", [SDNPMemOperand]>,
AMDGPURsrcIntrinsic<1, 0>;
def int_amdgcn_raw_buffer_atomic_swap : AMDGPURawBufferAtomic;
def int_amdgcn_raw_buffer_atomic_add : AMDGPURawBufferAtomic;
llvm_i32_ty, // offset(VGPR/imm, included in bounds checking and swizzling)
llvm_i32_ty, // soffset(SGPR/imm, excluded from bounds checking and swizzling)
llvm_i32_ty], // cachepolicy(imm; bit 1 = slc)
- [ImmArg<ArgIndex<5>>, IntrWillReturn], "", [SDNPMemOperand]>,
+ [ImmArg<ArgIndex<5>>, IntrWillReturn, IntrNoCallback, IntrNoFree], "", [SDNPMemOperand]>,
AMDGPURsrcIntrinsic<2, 0>;
// gfx908 intrinsic
llvm_i32_ty, // offset(VGPR/imm, included in bounds checking and swizzling)
llvm_i32_ty, // soffset(SGPR/imm, excluded from bounds checking and swizzling)
llvm_i32_ty], // cachepolicy(imm; bit 1 = slc)
- [ImmArg<ArgIndex<5>>, IntrWillReturn], "", [SDNPMemOperand]>,
+ [ImmArg<ArgIndex<5>>, IntrWillReturn, IntrNoCallback, IntrNoFree], "", [SDNPMemOperand]>,
AMDGPURsrcIntrinsic<1, 0>;
def int_amdgcn_struct_buffer_atomic_swap : AMDGPUStructBufferAtomic;
def int_amdgcn_struct_buffer_atomic_add : AMDGPUStructBufferAtomic;
llvm_i32_ty, // offset(VGPR/imm, included in bounds checking and swizzling)
llvm_i32_ty, // soffset(SGPR/imm, excluded from bounds checking and swizzling)
llvm_i32_ty], // cachepolicy(imm; bit 1 = slc)
- [ImmArg<ArgIndex<6>>, IntrWillReturn], "", [SDNPMemOperand]>,
+ [ImmArg<ArgIndex<6>>, IntrWillReturn, IntrNoCallback, IntrNoFree], "", [SDNPMemOperand]>,
AMDGPURsrcIntrinsic<2, 0>;
// gfx908 intrinsic
// Obsolescent tbuffer intrinsics.
-def int_amdgcn_tbuffer_load : Intrinsic <
+def int_amdgcn_tbuffer_load : DefaultAttrsIntrinsic <
[llvm_any_ty], // overloaded for types f32/i32, v2f32/v2i32, v4f32/v4i32
[llvm_v4i32_ty, // rsrc(SGPR)
llvm_i32_ty, // vindex(VGPR)
llvm_i32_ty, // nfmt(imm)
llvm_i1_ty, // glc(imm)
llvm_i1_ty], // slc(imm)
- [IntrReadMem, IntrWillReturn,
+ [IntrReadMem,
ImmArg<ArgIndex<4>>, ImmArg<ArgIndex<5>>, ImmArg<ArgIndex<6>>,
ImmArg<ArgIndex<7>>, ImmArg<ArgIndex<8>>], "", [SDNPMemOperand]>,
AMDGPURsrcIntrinsic<0>;
-def int_amdgcn_tbuffer_store : Intrinsic <
+def int_amdgcn_tbuffer_store : DefaultAttrsIntrinsic <
[],
[llvm_any_ty, // vdata(VGPR), overloaded for types f32/i32, v2f32/v2i32, v4f32/v4i32
llvm_v4i32_ty, // rsrc(SGPR)
llvm_i32_ty, // nfmt(imm)
llvm_i1_ty, // glc(imm)
llvm_i1_ty], // slc(imm)
- [IntrWriteMem, IntrWillReturn, ImmArg<ArgIndex<5>>,
+ [IntrWriteMem, ImmArg<ArgIndex<5>>,
ImmArg<ArgIndex<6>>, ImmArg<ArgIndex<7>>,
ImmArg<ArgIndex<8>>, ImmArg<ArgIndex<9>>], "", [SDNPMemOperand]>,
AMDGPURsrcIntrinsic<1>;
// - raw and struct variants
// - joint format field
// - joint cachepolicy field
-def int_amdgcn_raw_tbuffer_load : Intrinsic <
+def int_amdgcn_raw_tbuffer_load : DefaultAttrsIntrinsic <
[llvm_any_ty], // overloaded for types f32/i32, v2f32/v2i32, v4f32/v4i32
[llvm_v4i32_ty, // rsrc(SGPR)
llvm_i32_ty, // offset(VGPR/imm, included in bounds checking and swizzling)
// bit 1 = slc,
// bit 2 = dlc on gfx10+),
// swizzled buffer (bit 3 = swz))
- [IntrReadMem, IntrWillReturn,
+ [IntrReadMem,
ImmArg<ArgIndex<3>>, ImmArg<ArgIndex<4>>], "", [SDNPMemOperand]>,
AMDGPURsrcIntrinsic<0>;
-def int_amdgcn_raw_tbuffer_store : Intrinsic <
+def int_amdgcn_raw_tbuffer_store : DefaultAttrsIntrinsic <
[],
[llvm_any_ty, // vdata(VGPR), overloaded for types f32/i32, v2f32/v2i32, v4f32/v4i32
llvm_v4i32_ty, // rsrc(SGPR)
// bit 1 = slc,
// bit 2 = dlc on gfx10+),
// swizzled buffer (bit 3 = swz))
- [IntrWriteMem, IntrWillReturn,
+ [IntrWriteMem,
ImmArg<ArgIndex<4>>, ImmArg<ArgIndex<5>>], "", [SDNPMemOperand]>,
AMDGPURsrcIntrinsic<1>;
-def int_amdgcn_struct_tbuffer_load : Intrinsic <
+def int_amdgcn_struct_tbuffer_load : DefaultAttrsIntrinsic <
[llvm_any_ty], // overloaded for types f32/i32, v2f32/v2i32, v4f32/v4i32
[llvm_v4i32_ty, // rsrc(SGPR)
llvm_i32_ty, // vindex(VGPR)
// bit 1 = slc,
// bit 2 = dlc on gfx10+),
// swizzled buffer (bit 3 = swz))
- [IntrReadMem, IntrWillReturn,
+ [IntrReadMem,
ImmArg<ArgIndex<4>>, ImmArg<ArgIndex<5>>], "", [SDNPMemOperand]>,
AMDGPURsrcIntrinsic<0>;
-def int_amdgcn_struct_tbuffer_store : Intrinsic <
+def int_amdgcn_struct_tbuffer_store : DefaultAttrsIntrinsic <
[],
[llvm_any_ty, // vdata(VGPR), overloaded for types f32/i32, v2f32/v2i32, v4f32/v4i32
llvm_v4i32_ty, // rsrc(SGPR)
// bit 1 = slc,
// bit 2 = dlc on gfx10+),
// swizzled buffer (bit 3 = swz))
- [IntrWriteMem, IntrWillReturn,
+ [IntrWriteMem,
ImmArg<ArgIndex<5>>, ImmArg<ArgIndex<6>>], "", [SDNPMemOperand]>,
AMDGPURsrcIntrinsic<1>;
llvm_i32_ty, // vindex(VGPR)
llvm_i32_ty, // offset(SGPR/VGPR/imm)
llvm_i1_ty], // slc(imm)
- [ImmArg<ArgIndex<4>>, IntrWillReturn], "", [SDNPMemOperand]>,
+ [ImmArg<ArgIndex<4>>, IntrWillReturn, IntrNoCallback, IntrNoFree], "", [SDNPMemOperand]>,
AMDGPURsrcIntrinsic<1, 0>;
def int_amdgcn_buffer_atomic_swap : AMDGPUBufferAtomic;
def int_amdgcn_buffer_atomic_add : AMDGPUBufferAtomic;
llvm_i32_ty, // vindex(VGPR)
llvm_i32_ty, // offset(SGPR/VGPR/imm)
llvm_i1_ty], // slc(imm)
- [ImmArg<ArgIndex<5>>, IntrWillReturn], "", [SDNPMemOperand]>,
+ [ImmArg<ArgIndex<5>>, IntrWillReturn, IntrNoCallback, IntrNoFree], "", [SDNPMemOperand]>,
AMDGPURsrcIntrinsic<2, 0>;
def int_amdgcn_buffer_atomic_csub : AMDGPUBufferAtomic;
llvm_i32_ty, // vindex(VGPR)
llvm_i32_ty, // offset(SGPR/VGPR/imm)
llvm_i1_ty], // slc(imm)
- [ImmArg<ArgIndex<4>>, IntrWillReturn], "", [SDNPMemOperand]>,
+ [ImmArg<ArgIndex<4>>, IntrWillReturn, IntrNoCallback, IntrNoFree], "", [SDNPMemOperand]>,
AMDGPURsrcIntrinsic<1, 0>;
// Legacy form of the intrinsic. raw and struct forms should be preferred.
// bit 2 = dlc on gfx10+))
// swizzled buffer (bit 3 = swz))
[IntrWillReturn, NoCapture<ArgIndex<1>>, ImmArg<ArgIndex<2>>, ImmArg<ArgIndex<5>>,
- ImmArg<ArgIndex<6>>], "", [SDNPMemOperand]>, AMDGPURsrcIntrinsic<0>;
+ ImmArg<ArgIndex<6>>, IntrNoCallback, IntrNoFree], "", [SDNPMemOperand]>, AMDGPURsrcIntrinsic<0>;
def int_amdgcn_raw_buffer_load_lds : AMDGPURawBufferLoadLDS;
class AMDGPUStructBufferLoadLDS : Intrinsic <
// bit 2 = dlc on gfx10+))
// swizzled buffer (bit 3 = swz))
[IntrWillReturn, NoCapture<ArgIndex<1>>, ImmArg<ArgIndex<2>>, ImmArg<ArgIndex<6>>,
- ImmArg<ArgIndex<7>>], "", [SDNPMemOperand]>, AMDGPURsrcIntrinsic<0>;
+ ImmArg<ArgIndex<7>>, IntrNoCallback, IntrNoFree], "", [SDNPMemOperand]>, AMDGPURsrcIntrinsic<0>;
def int_amdgcn_struct_buffer_load_lds : AMDGPUStructBufferLoadLDS;
} // defset AMDGPUBufferIntrinsics
// Uses that do not set the done bit should set IntrWriteMem on the
// call site.
-def int_amdgcn_exp : Intrinsic <[], [
+def int_amdgcn_exp : DefaultAttrsIntrinsic <[], [
llvm_i32_ty, // tgt,
llvm_i32_ty, // en
llvm_any_ty, // src0 (f32 or i32)
llvm_i1_ty // vm (ignored on GFX11+)
],
[ImmArg<ArgIndex<0>>, ImmArg<ArgIndex<1>>, ImmArg<ArgIndex<6>>,
- ImmArg<ArgIndex<7>>, IntrWriteMem, IntrInaccessibleMemOnly,
- IntrWillReturn]
+ ImmArg<ArgIndex<7>>, IntrWriteMem, IntrInaccessibleMemOnly]
>;
// exp with row_en bit set. Only supported on GFX11+.
-def int_amdgcn_exp_row : Intrinsic <[], [
+def int_amdgcn_exp_row : DefaultAttrsIntrinsic <[], [
llvm_i32_ty, // tgt,
llvm_i32_ty, // en
llvm_any_ty, // src0 (f32 or i32)
llvm_i1_ty, // done
llvm_i32_ty], // row number
[ImmArg<ArgIndex<0>>, ImmArg<ArgIndex<1>>, ImmArg<ArgIndex<6>>,
- IntrWriteMem, IntrInaccessibleMemOnly, IntrWillReturn]
+ IntrWriteMem, IntrInaccessibleMemOnly]
>;
// exp with compr bit set. Not supported on GFX11+.
-def int_amdgcn_exp_compr : Intrinsic <[], [
+def int_amdgcn_exp_compr : DefaultAttrsIntrinsic <[], [
llvm_i32_ty, // tgt,
llvm_i32_ty, // en
llvm_anyvector_ty, // src0 (v2f16 or v2i16)
llvm_i1_ty, // done
llvm_i1_ty], // vm
[ImmArg<ArgIndex<0>>, ImmArg<ArgIndex<1>>, ImmArg<ArgIndex<4>>,
- ImmArg<ArgIndex<5>>, IntrWriteMem, IntrInaccessibleMemOnly,
- IntrWillReturn]
+ ImmArg<ArgIndex<5>>, IntrWriteMem, IntrInaccessibleMemOnly]
>;
def int_amdgcn_buffer_wbinvl1_sc :
ClangBuiltin<"__builtin_amdgcn_buffer_wbinvl1_sc">,
- Intrinsic<[], [], [IntrNoMem, IntrHasSideEffects, IntrWillReturn]>;
+ DefaultAttrsIntrinsic<[], [], [IntrNoMem, IntrHasSideEffects]>;
def int_amdgcn_buffer_wbinvl1 :
ClangBuiltin<"__builtin_amdgcn_buffer_wbinvl1">,
- Intrinsic<[], [], [IntrNoMem, IntrHasSideEffects, IntrWillReturn]>;
+ DefaultAttrsIntrinsic<[], [], [IntrNoMem, IntrHasSideEffects]>;
def int_amdgcn_s_dcache_inv :
ClangBuiltin<"__builtin_amdgcn_s_dcache_inv">,
- Intrinsic<[], [], [IntrNoMem, IntrHasSideEffects, IntrWillReturn]>;
+ DefaultAttrsIntrinsic<[], [], [IntrNoMem, IntrHasSideEffects]>;
def int_amdgcn_s_memtime :
ClangBuiltin<"__builtin_amdgcn_s_memtime">,
- Intrinsic<[llvm_i64_ty], [], [IntrNoMem, IntrHasSideEffects, IntrWillReturn]>;
+ DefaultAttrsIntrinsic<[llvm_i64_ty], [], [IntrNoMem, IntrHasSideEffects]>;
def int_amdgcn_s_sleep :
ClangBuiltin<"__builtin_amdgcn_s_sleep">,
- Intrinsic<[], [llvm_i32_ty], [ImmArg<ArgIndex<0>>, IntrNoMem,
- IntrHasSideEffects, IntrWillReturn]> {
+ DefaultAttrsIntrinsic<[], [llvm_i32_ty], [ImmArg<ArgIndex<0>>, IntrNoMem,
+ IntrHasSideEffects]> {
}
def int_amdgcn_s_incperflevel :
ClangBuiltin<"__builtin_amdgcn_s_incperflevel">,
- Intrinsic<[], [llvm_i32_ty], [ImmArg<ArgIndex<0>>, IntrNoMem,
- IntrHasSideEffects, IntrWillReturn]> {
+ DefaultAttrsIntrinsic<[], [llvm_i32_ty], [ImmArg<ArgIndex<0>>, IntrNoMem,
+ IntrHasSideEffects]> {
}
def int_amdgcn_s_decperflevel :
ClangBuiltin<"__builtin_amdgcn_s_decperflevel">,
- Intrinsic<[], [llvm_i32_ty], [ImmArg<ArgIndex<0>>, IntrNoMem,
- IntrHasSideEffects, IntrWillReturn]> {
+ DefaultAttrsIntrinsic<[], [llvm_i32_ty], [ImmArg<ArgIndex<0>>, IntrNoMem,
+ IntrHasSideEffects]> {
}
def int_amdgcn_s_sethalt :
- Intrinsic<[], [llvm_i32_ty], [ImmArg<ArgIndex<0>>, IntrNoMem,
- IntrHasSideEffects, IntrWillReturn]>;
+ DefaultAttrsIntrinsic<[], [llvm_i32_ty], [ImmArg<ArgIndex<0>>, IntrNoMem,
+ IntrHasSideEffects]>;
def int_amdgcn_s_setprio :
ClangBuiltin<"__builtin_amdgcn_s_setprio">,
- Intrinsic<[], [llvm_i16_ty], [ImmArg<ArgIndex<0>>, IntrNoMem,
- IntrHasSideEffects, IntrWillReturn]>;
+ DefaultAttrsIntrinsic<[], [llvm_i16_ty], [ImmArg<ArgIndex<0>>, IntrNoMem,
+ IntrHasSideEffects]>;
// This is IntrHasSideEffects so it can be used to read cycle counters.
def int_amdgcn_s_getreg :
ClangBuiltin<"__builtin_amdgcn_s_getreg">,
- Intrinsic<[llvm_i32_ty], [llvm_i32_ty],
- [IntrNoMem, IntrHasSideEffects, IntrWillReturn, ImmArg<ArgIndex<0>>]
+ DefaultAttrsIntrinsic<[llvm_i32_ty], [llvm_i32_ty],
+ [IntrNoMem, IntrHasSideEffects, ImmArg<ArgIndex<0>>]
>;
// Note this can be used to set FP environment properties that are
// subtarget. llvm.amdgcn.s.setreg(hwmode, value)
def int_amdgcn_s_setreg :
ClangBuiltin<"__builtin_amdgcn_s_setreg">,
- Intrinsic<[], [llvm_i32_ty, llvm_i32_ty],
- [IntrNoMem, IntrHasSideEffects, IntrWillReturn, ImmArg<ArgIndex<0>>]
+ DefaultAttrsIntrinsic<[], [llvm_i32_ty, llvm_i32_ty],
+ [IntrNoMem, IntrHasSideEffects, ImmArg<ArgIndex<0>>]
>;
// int_amdgcn_s_getpc is provided to allow a specific style of position
// especially as we explicitly use IntrNoMem to allow optimizations.
def int_amdgcn_s_getpc :
ClangBuiltin<"__builtin_amdgcn_s_getpc">,
- Intrinsic<[llvm_i64_ty], [], [IntrNoMem, IntrSpeculatable,
+ DefaultAttrsIntrinsic<[llvm_i64_ty], [], [IntrNoMem, IntrSpeculatable,
IntrWillReturn]>;
// __builtin_amdgcn_interp_mov <param>, <attr_chan>, <attr>, <m0>
// param values: 0 = P10, 1 = P20, 2 = P0
def int_amdgcn_interp_mov :
ClangBuiltin<"__builtin_amdgcn_interp_mov">,
- Intrinsic<[llvm_float_ty],
+ DefaultAttrsIntrinsic<[llvm_float_ty],
[llvm_i32_ty, llvm_i32_ty, llvm_i32_ty, llvm_i32_ty],
- [IntrNoMem, IntrSpeculatable, IntrWillReturn,
+ [IntrNoMem, IntrSpeculatable,
ImmArg<ArgIndex<0>>, ImmArg<ArgIndex<1>>, ImmArg<ArgIndex<2>>]>;
// __builtin_amdgcn_interp_p1 <i>, <attr_chan>, <attr>, <m0>
// so it behaves like IntrNoMem.
def int_amdgcn_interp_p1 :
ClangBuiltin<"__builtin_amdgcn_interp_p1">,
- Intrinsic<[llvm_float_ty],
+ DefaultAttrsIntrinsic<[llvm_float_ty],
[llvm_float_ty, llvm_i32_ty, llvm_i32_ty, llvm_i32_ty],
- [IntrNoMem, IntrSpeculatable, IntrWillReturn,
+ [IntrNoMem, IntrSpeculatable,
ImmArg<ArgIndex<1>>, ImmArg<ArgIndex<2>>]>;
// __builtin_amdgcn_interp_p2 <p1>, <j>, <attr_chan>, <attr>, <m0>
def int_amdgcn_interp_p2 :
ClangBuiltin<"__builtin_amdgcn_interp_p2">,
- Intrinsic<[llvm_float_ty],
+ DefaultAttrsIntrinsic<[llvm_float_ty],
[llvm_float_ty, llvm_float_ty, llvm_i32_ty, llvm_i32_ty, llvm_i32_ty],
- [IntrNoMem, IntrSpeculatable, IntrWillReturn,
+ [IntrNoMem, IntrSpeculatable,
ImmArg<ArgIndex<2>>, ImmArg<ArgIndex<3>>]>;
// See int_amdgcn_v_interp_p1 for why this is IntrNoMem.
// high selects whether high or low 16-bits are loaded from LDS
def int_amdgcn_interp_p1_f16 :
ClangBuiltin<"__builtin_amdgcn_interp_p1_f16">,
- Intrinsic<[llvm_float_ty],
+ DefaultAttrsIntrinsic<[llvm_float_ty],
[llvm_float_ty, llvm_i32_ty, llvm_i32_ty, llvm_i1_ty, llvm_i32_ty],
- [IntrNoMem, IntrSpeculatable, IntrWillReturn,
+ [IntrNoMem, IntrSpeculatable,
ImmArg<ArgIndex<1>>, ImmArg<ArgIndex<2>>, ImmArg<ArgIndex<3>>]>;
// __builtin_amdgcn_interp_p2_f16 <p1>, <j>, <attr_chan>, <attr>, <high>, <m0>
// high selects whether high or low 16-bits are loaded from LDS
def int_amdgcn_interp_p2_f16 :
ClangBuiltin<"__builtin_amdgcn_interp_p2_f16">,
- Intrinsic<[llvm_half_ty],
+ DefaultAttrsIntrinsic<[llvm_half_ty],
[llvm_float_ty, llvm_float_ty, llvm_i32_ty, llvm_i32_ty, llvm_i1_ty, llvm_i32_ty],
- [IntrNoMem, IntrSpeculatable, IntrWillReturn,
+ [IntrNoMem, IntrSpeculatable,
ImmArg<ArgIndex<2>>, ImmArg<ArgIndex<3>>, ImmArg<ArgIndex<4>>]>;
// llvm.amdgcn.lds.direct.load <m0>
// The input argument is m0, which contains a packed combination of address
// offset and flags describing the data type.
def int_amdgcn_lds_direct_load :
- Intrinsic<[llvm_any_ty], // overloaded for types u8, u16, i32/f32, i8, i16
+ DefaultAttrsIntrinsic<[llvm_any_ty], // overloaded for types u8, u16, i32/f32, i8, i16
[llvm_i32_ty],
- [IntrReadMem, IntrSpeculatable, IntrWillReturn]>;
+ [IntrReadMem, IntrSpeculatable]>;
// llvm.amdgcn.lds.param.load <attr_chan>, <attr>, <m0>
// Like interp intrinsics, this reads from lds, but the memory values are constant,
// so it behaves like IntrNoMem.
def int_amdgcn_lds_param_load :
- Intrinsic<[llvm_float_ty],
+ DefaultAttrsIntrinsic<[llvm_float_ty],
[llvm_i32_ty, llvm_i32_ty, llvm_i32_ty],
- [IntrNoMem, IntrSpeculatable, IntrWillReturn,
+ [IntrNoMem, IntrSpeculatable,
ImmArg<ArgIndex<0>>, ImmArg<ArgIndex<1>>]>;
// llvm.amdgcn.interp.inreg.p10 <p>, <i>, <p0>
def int_amdgcn_interp_inreg_p10 :
- Intrinsic<[llvm_float_ty],
+ DefaultAttrsIntrinsic<[llvm_float_ty],
[llvm_float_ty, llvm_float_ty, llvm_float_ty],
- [IntrNoMem, IntrSpeculatable, IntrWillReturn]>;
+ [IntrNoMem, IntrSpeculatable]>;
// llvm.amdgcn.interp.inreg.p2 <p>, <j>, <tmp>
def int_amdgcn_interp_inreg_p2 :
- Intrinsic<[llvm_float_ty],
+ DefaultAttrsIntrinsic<[llvm_float_ty],
[llvm_float_ty, llvm_float_ty, llvm_float_ty],
- [IntrNoMem, IntrSpeculatable, IntrWillReturn]>;
+ [IntrNoMem, IntrSpeculatable]>;
// llvm.amdgcn.interp.inreg.p10.f16 <p>, <i>, <p0>, <high>
// high selects whether high or low 16-bits are used for p and p0 operands
def int_amdgcn_interp_inreg_p10_f16:
- Intrinsic<[llvm_float_ty],
+ DefaultAttrsIntrinsic<[llvm_float_ty],
[llvm_float_ty, llvm_float_ty, llvm_float_ty, llvm_i1_ty],
- [IntrNoMem, IntrSpeculatable, IntrWillReturn,
+ [IntrNoMem, IntrSpeculatable,
ImmArg<ArgIndex<3>>]>;
// llvm.amdgcn.interp.inreg.p2.f16 <p>, <j>, <tmp>, <high>
// high selects whether high or low 16-bits are used for p operand
def int_amdgcn_interp_inreg_p2_f16 :
- Intrinsic<[llvm_half_ty],
+ DefaultAttrsIntrinsic<[llvm_half_ty],
[llvm_float_ty, llvm_float_ty, llvm_float_ty, llvm_i1_ty],
- [IntrNoMem, IntrSpeculatable, IntrWillReturn,
+ [IntrNoMem, IntrSpeculatable,
ImmArg<ArgIndex<3>>]>;
// Deprecated: use llvm.amdgcn.live.mask instead.
-def int_amdgcn_ps_live : Intrinsic <
+def int_amdgcn_ps_live : DefaultAttrsIntrinsic <
[llvm_i1_ty],
[],
- [IntrNoMem, IntrWillReturn]>;
+ [IntrNoMem]>;
// Query currently live lanes.
// Returns true if lane is live (and not a helper lane).
-def int_amdgcn_live_mask : Intrinsic <[llvm_i1_ty],
- [], [IntrReadMem, IntrInaccessibleMemOnly, IntrWillReturn]
+def int_amdgcn_live_mask : DefaultAttrsIntrinsic <[llvm_i1_ty],
+ [], [IntrReadMem, IntrInaccessibleMemOnly]
>;
def int_amdgcn_mbcnt_lo :
ClangBuiltin<"__builtin_amdgcn_mbcnt_lo">,
- Intrinsic<[llvm_i32_ty], [llvm_i32_ty, llvm_i32_ty],
- [IntrNoMem, IntrWillReturn]>;
+ DefaultAttrsIntrinsic<[llvm_i32_ty], [llvm_i32_ty, llvm_i32_ty],
+ [IntrNoMem]>;
def int_amdgcn_mbcnt_hi :
ClangBuiltin<"__builtin_amdgcn_mbcnt_hi">,
- Intrinsic<[llvm_i32_ty], [llvm_i32_ty, llvm_i32_ty],
- [IntrNoMem, IntrWillReturn]>;
+ DefaultAttrsIntrinsic<[llvm_i32_ty], [llvm_i32_ty, llvm_i32_ty],
+ [IntrNoMem]>;
// llvm.amdgcn.ds.swizzle src offset
def int_amdgcn_ds_swizzle :
ClangBuiltin<"__builtin_amdgcn_ds_swizzle">,
Intrinsic<[llvm_i32_ty], [llvm_i32_ty, llvm_i32_ty],
- [IntrNoMem, IntrConvergent, IntrWillReturn,
+ [IntrNoMem, IntrConvergent, IntrWillReturn, IntrNoCallback, IntrNoFree,
ImmArg<ArgIndex<1>>]>;
-def int_amdgcn_ubfe : Intrinsic<[llvm_anyint_ty],
+def int_amdgcn_ubfe : DefaultAttrsIntrinsic<[llvm_anyint_ty],
[LLVMMatchType<0>, llvm_i32_ty, llvm_i32_ty],
- [IntrNoMem, IntrSpeculatable, IntrWillReturn]
+ [IntrNoMem, IntrSpeculatable]
>;
-def int_amdgcn_sbfe : Intrinsic<[llvm_anyint_ty],
+def int_amdgcn_sbfe : DefaultAttrsIntrinsic<[llvm_anyint_ty],
[LLVMMatchType<0>, llvm_i32_ty, llvm_i32_ty],
- [IntrNoMem, IntrSpeculatable, IntrWillReturn]
+ [IntrNoMem, IntrSpeculatable]
>;
def int_amdgcn_lerp :
ClangBuiltin<"__builtin_amdgcn_lerp">,
- Intrinsic<[llvm_i32_ty], [llvm_i32_ty, llvm_i32_ty, llvm_i32_ty],
- [IntrNoMem, IntrSpeculatable, IntrWillReturn]
+ DefaultAttrsIntrinsic<[llvm_i32_ty], [llvm_i32_ty, llvm_i32_ty, llvm_i32_ty],
+ [IntrNoMem, IntrSpeculatable]
>;
def int_amdgcn_sad_u8 :
ClangBuiltin<"__builtin_amdgcn_sad_u8">,
- Intrinsic<[llvm_i32_ty], [llvm_i32_ty, llvm_i32_ty, llvm_i32_ty],
- [IntrNoMem, IntrSpeculatable, IntrWillReturn]
+ DefaultAttrsIntrinsic<[llvm_i32_ty], [llvm_i32_ty, llvm_i32_ty, llvm_i32_ty],
+ [IntrNoMem, IntrSpeculatable]
>;
def int_amdgcn_msad_u8 :
ClangBuiltin<"__builtin_amdgcn_msad_u8">,
- Intrinsic<[llvm_i32_ty], [llvm_i32_ty, llvm_i32_ty, llvm_i32_ty],
- [IntrNoMem, IntrSpeculatable, IntrWillReturn]
+ DefaultAttrsIntrinsic<[llvm_i32_ty], [llvm_i32_ty, llvm_i32_ty, llvm_i32_ty],
+ [IntrNoMem, IntrSpeculatable]
>;
def int_amdgcn_sad_hi_u8 :
ClangBuiltin<"__builtin_amdgcn_sad_hi_u8">,
- Intrinsic<[llvm_i32_ty], [llvm_i32_ty, llvm_i32_ty, llvm_i32_ty],
- [IntrNoMem, IntrSpeculatable, IntrWillReturn]
+ DefaultAttrsIntrinsic<[llvm_i32_ty], [llvm_i32_ty, llvm_i32_ty, llvm_i32_ty],
+ [IntrNoMem, IntrSpeculatable]
>;
def int_amdgcn_sad_u16 :
ClangBuiltin<"__builtin_amdgcn_sad_u16">,
- Intrinsic<[llvm_i32_ty], [llvm_i32_ty, llvm_i32_ty, llvm_i32_ty],
- [IntrNoMem, IntrSpeculatable, IntrWillReturn]
+ DefaultAttrsIntrinsic<[llvm_i32_ty], [llvm_i32_ty, llvm_i32_ty, llvm_i32_ty],
+ [IntrNoMem, IntrSpeculatable]
>;
def int_amdgcn_qsad_pk_u16_u8 :
ClangBuiltin<"__builtin_amdgcn_qsad_pk_u16_u8">,
- Intrinsic<[llvm_i64_ty], [llvm_i64_ty, llvm_i32_ty, llvm_i64_ty],
- [IntrNoMem, IntrSpeculatable, IntrWillReturn]
+ DefaultAttrsIntrinsic<[llvm_i64_ty], [llvm_i64_ty, llvm_i32_ty, llvm_i64_ty],
+ [IntrNoMem, IntrSpeculatable]
>;
def int_amdgcn_mqsad_pk_u16_u8 :
ClangBuiltin<"__builtin_amdgcn_mqsad_pk_u16_u8">,
- Intrinsic<[llvm_i64_ty], [llvm_i64_ty, llvm_i32_ty, llvm_i64_ty],
- [IntrNoMem, IntrSpeculatable, IntrWillReturn]
+ DefaultAttrsIntrinsic<[llvm_i64_ty], [llvm_i64_ty, llvm_i32_ty, llvm_i64_ty],
+ [IntrNoMem, IntrSpeculatable]
>;
def int_amdgcn_mqsad_u32_u8 :
ClangBuiltin<"__builtin_amdgcn_mqsad_u32_u8">,
- Intrinsic<[llvm_v4i32_ty], [llvm_i64_ty, llvm_i32_ty, llvm_v4i32_ty],
- [IntrNoMem, IntrSpeculatable, IntrWillReturn]
+ DefaultAttrsIntrinsic<[llvm_v4i32_ty], [llvm_i64_ty, llvm_i32_ty, llvm_v4i32_ty],
+ [IntrNoMem, IntrSpeculatable]
>;
def int_amdgcn_cvt_pk_u8_f32 :
ClangBuiltin<"__builtin_amdgcn_cvt_pk_u8_f32">,
- Intrinsic<[llvm_i32_ty], [llvm_float_ty, llvm_i32_ty, llvm_i32_ty],
- [IntrNoMem, IntrSpeculatable, IntrWillReturn]
+ DefaultAttrsIntrinsic<[llvm_i32_ty], [llvm_float_ty, llvm_i32_ty, llvm_i32_ty],
+ [IntrNoMem, IntrSpeculatable]
>;
def int_amdgcn_icmp :
Intrinsic<[llvm_anyint_ty], [llvm_anyint_ty, LLVMMatchType<1>, llvm_i32_ty],
- [IntrNoMem, IntrConvergent, IntrWillReturn,
- ImmArg<ArgIndex<2>>]>;
+ [IntrNoMem, IntrConvergent,
+ ImmArg<ArgIndex<2>>, IntrWillReturn, IntrNoCallback, IntrNoFree]>;
def int_amdgcn_fcmp :
Intrinsic<[llvm_anyint_ty], [llvm_anyfloat_ty, LLVMMatchType<1>, llvm_i32_ty],
- [IntrNoMem, IntrConvergent, IntrWillReturn,
- ImmArg<ArgIndex<2>>]>;
+ [IntrNoMem, IntrConvergent,
+ ImmArg<ArgIndex<2>>, IntrWillReturn, IntrNoCallback, IntrNoFree]>;
def int_amdgcn_ballot :
Intrinsic<[llvm_anyint_ty], [llvm_i1_ty],
- [IntrNoMem, IntrConvergent, IntrWillReturn]>;
+ [IntrNoMem, IntrConvergent, IntrWillReturn, IntrNoCallback, IntrNoFree]>;
def int_amdgcn_readfirstlane :
ClangBuiltin<"__builtin_amdgcn_readfirstlane">,
Intrinsic<[llvm_i32_ty], [llvm_i32_ty],
- [IntrNoMem, IntrConvergent, IntrWillReturn]>;
+ [IntrNoMem, IntrConvergent, IntrWillReturn, IntrNoCallback, IntrNoFree]>;
// The lane argument must be uniform across the currently active threads of the
// current wave. Otherwise, the result is undefined.
def int_amdgcn_readlane :
ClangBuiltin<"__builtin_amdgcn_readlane">,
Intrinsic<[llvm_i32_ty], [llvm_i32_ty, llvm_i32_ty],
- [IntrNoMem, IntrConvergent, IntrWillReturn]>;
+ [IntrNoMem, IntrConvergent, IntrWillReturn, IntrNoCallback, IntrNoFree]>;
// The value to write and lane select arguments must be uniform across the
// currently active threads of the current wave. Otherwise, the result is
llvm_i32_ty, // uniform lane select
llvm_i32_ty // returned by all lanes other than the selected one
],
- [IntrNoMem, IntrConvergent, IntrWillReturn]
+ [IntrNoMem, IntrConvergent, IntrWillReturn, IntrNoCallback, IntrNoFree]
>;
def int_amdgcn_alignbyte : ClangBuiltin<"__builtin_amdgcn_alignbyte">,
- Intrinsic<[llvm_i32_ty], [llvm_i32_ty, llvm_i32_ty, llvm_i32_ty],
- [IntrNoMem, IntrSpeculatable, IntrWillReturn]
+ DefaultAttrsIntrinsic<[llvm_i32_ty], [llvm_i32_ty, llvm_i32_ty, llvm_i32_ty],
+ [IntrNoMem, IntrSpeculatable]
>;
-def int_amdgcn_mul_i24 : Intrinsic<[llvm_i32_ty],
+def int_amdgcn_mul_i24 : DefaultAttrsIntrinsic<[llvm_i32_ty],
[llvm_i32_ty, llvm_i32_ty],
- [IntrNoMem, IntrSpeculatable, IntrWillReturn]
+ [IntrNoMem, IntrSpeculatable]
>;
-def int_amdgcn_mul_u24 : Intrinsic<[llvm_i32_ty],
+def int_amdgcn_mul_u24 : DefaultAttrsIntrinsic<[llvm_i32_ty],
[llvm_i32_ty, llvm_i32_ty],
- [IntrNoMem, IntrSpeculatable, IntrWillReturn]
+ [IntrNoMem, IntrSpeculatable]
>;
-def int_amdgcn_mulhi_i24 : Intrinsic<[llvm_i32_ty],
+def int_amdgcn_mulhi_i24 : DefaultAttrsIntrinsic<[llvm_i32_ty],
[llvm_i32_ty, llvm_i32_ty],
- [IntrNoMem, IntrSpeculatable, IntrWillReturn]
+ [IntrNoMem, IntrSpeculatable]
>;
-def int_amdgcn_mulhi_u24 : Intrinsic<[llvm_i32_ty],
+def int_amdgcn_mulhi_u24 : DefaultAttrsIntrinsic<[llvm_i32_ty],
[llvm_i32_ty, llvm_i32_ty],
- [IntrNoMem, IntrSpeculatable, IntrWillReturn]
+ [IntrNoMem, IntrSpeculatable]
>;
// llvm.amdgcn.ds.gws.init(i32 bar_val, i32 resource_id)
Intrinsic<[],
[llvm_i32_ty, llvm_i32_ty],
[IntrConvergent, IntrWriteMem,
- IntrInaccessibleMemOnly, IntrWillReturn], "",
+ IntrInaccessibleMemOnly, IntrWillReturn, IntrNoCallback, IntrNoFree], "",
[SDNPMemOperand]
>;
ClangBuiltin<"__builtin_amdgcn_ds_gws_barrier">,
Intrinsic<[],
[llvm_i32_ty, llvm_i32_ty],
- [IntrConvergent, IntrInaccessibleMemOnly, IntrWillReturn], "",
+ [IntrConvergent, IntrInaccessibleMemOnly, IntrWillReturn, IntrNoCallback, IntrNoFree], "",
[SDNPMemOperand]
>;
ClangBuiltin<"__builtin_amdgcn_ds_gws_sema_v">,
Intrinsic<[],
[llvm_i32_ty],
- [IntrConvergent, IntrInaccessibleMemOnly, IntrWillReturn], "",
+ [IntrConvergent, IntrInaccessibleMemOnly, IntrWillReturn, IntrNoCallback, IntrNoFree], "",
[SDNPMemOperand]
>;
ClangBuiltin<"__builtin_amdgcn_ds_gws_sema_br">,
Intrinsic<[],
[llvm_i32_ty, llvm_i32_ty],
- [IntrConvergent, IntrInaccessibleMemOnly, IntrWillReturn], "",
+ [IntrConvergent, IntrInaccessibleMemOnly, IntrWillReturn, IntrNoCallback, IntrNoFree], "",
[SDNPMemOperand]
>;
ClangBuiltin<"__builtin_amdgcn_ds_gws_sema_p">,
Intrinsic<[],
[llvm_i32_ty],
- [IntrConvergent, IntrInaccessibleMemOnly, IntrWillReturn], "",
+ [IntrConvergent, IntrInaccessibleMemOnly, IntrWillReturn, IntrNoCallback, IntrNoFree], "",
[SDNPMemOperand]
>;
ClangBuiltin<"__builtin_amdgcn_ds_gws_sema_release_all">,
Intrinsic<[],
[llvm_i32_ty],
- [IntrConvergent, IntrInaccessibleMemOnly, IntrWillReturn], "",
+ [IntrConvergent, IntrInaccessibleMemOnly, IntrWillReturn, IntrNoCallback, IntrNoFree], "",
[SDNPMemOperand]
>;
// Copies the source value to the destination value, with the guarantee that
// the source value is computed as if the entire program were executed in WQM.
def int_amdgcn_wqm : Intrinsic<[llvm_any_ty],
- [LLVMMatchType<0>], [IntrNoMem, IntrSpeculatable, IntrWillReturn]
+ [LLVMMatchType<0>], [IntrNoMem, IntrSpeculatable, IntrWillReturn, IntrNoCallback, IntrNoFree]
>;
// Copies the source value to the destination value, such that the source
// is computed as if the entire program were executed in WQM if any other
// program code executes in WQM.
def int_amdgcn_softwqm : Intrinsic<[llvm_any_ty],
- [LLVMMatchType<0>], [IntrNoMem, IntrSpeculatable, IntrWillReturn]
+ [LLVMMatchType<0>], [IntrNoMem, IntrSpeculatable, IntrWillReturn, IntrNoCallback, IntrNoFree]
>;
// Return true if at least one thread within the pixel quad passes true into
// the function.
def int_amdgcn_wqm_vote : Intrinsic<[llvm_i1_ty],
- [llvm_i1_ty], [IntrNoMem, IntrConvergent, IntrWillReturn]
+ [llvm_i1_ty], [IntrNoMem, IntrConvergent, IntrWillReturn, IntrNoCallback, IntrNoFree]
>;
// If false, set EXEC=0 for the current thread until the end of program.
// FIXME: Should this be IntrNoMem, IntrHasSideEffects, or IntrWillReturn?
-def int_amdgcn_kill : Intrinsic<[], [llvm_i1_ty], []>;
+def int_amdgcn_kill : Intrinsic<[], [llvm_i1_ty], [IntrNoCallback, IntrNoFree]>;
def int_amdgcn_endpgm : ClangBuiltin<"__builtin_amdgcn_endpgm">,
- Intrinsic<[], [], [IntrNoReturn, IntrCold, IntrNoMem, IntrHasSideEffects]
+ Intrinsic<[], [], [IntrNoReturn, IntrCold, IntrNoMem, IntrHasSideEffects, IntrNoCallback, IntrNoFree]
>;
// If false, mark all active lanes as helper lanes until the end of program.
def int_amdgcn_wqm_demote : Intrinsic<[],
- [llvm_i1_ty], [IntrWriteMem, IntrInaccessibleMemOnly]
+ [llvm_i1_ty], [IntrWriteMem, IntrInaccessibleMemOnly, IntrNoCallback, IntrNoFree]
>;
// Copies the active channels of the source value to the destination value,
// undefined value.
def int_amdgcn_strict_wwm : Intrinsic<[llvm_any_ty],
[LLVMMatchType<0>], [IntrNoMem, IntrSpeculatable,
- IntrConvergent, IntrWillReturn]
+ IntrConvergent, IntrWillReturn, IntrNoCallback, IntrNoFree]
>;
// Deprecated. Use int_amdgcn_strict_wwm instead.
def int_amdgcn_wwm : Intrinsic<[llvm_any_ty],
[LLVMMatchType<0>], [IntrNoMem, IntrSpeculatable,
- IntrConvergent, IntrWillReturn]
+ IntrConvergent, IntrWillReturn, IntrNoCallback, IntrNoFree]
>;
def int_amdgcn_strict_wqm : Intrinsic<[llvm_any_ty],
[LLVMMatchType<0>], [IntrNoMem, IntrSpeculatable,
- IntrConvergent, IntrWillReturn]
+ IntrConvergent, IntrWillReturn, IntrNoCallback, IntrNoFree]
>;
// Given a value, copies it while setting all the inactive lanes to a given
Intrinsic<[llvm_anyint_ty],
[LLVMMatchType<0>, // value to be copied
LLVMMatchType<0>], // value for the inactive lanes to take
- [IntrNoMem, IntrConvergent, IntrWillReturn]>;
+ [IntrNoMem, IntrConvergent, IntrWillReturn, IntrNoCallback, IntrNoFree]>;
// Return if the given flat pointer points to a local memory address.
def int_amdgcn_is_shared : ClangBuiltin<"__builtin_amdgcn_is_shared">,
- Intrinsic<[llvm_i1_ty], [llvm_ptr_ty],
- [IntrNoMem, IntrSpeculatable, NoCapture<ArgIndex<0>>, IntrWillReturn]
+ DefaultAttrsIntrinsic<[llvm_i1_ty], [llvm_ptr_ty],
+ [IntrNoMem, IntrSpeculatable, NoCapture<ArgIndex<0>>]
>;
// Return if the given flat pointer points to a prvate memory address.
def int_amdgcn_is_private : ClangBuiltin<"__builtin_amdgcn_is_private">,
- Intrinsic<[llvm_i1_ty], [llvm_ptr_ty],
- [IntrNoMem, IntrSpeculatable, NoCapture<ArgIndex<0>>, IntrWillReturn]
+ DefaultAttrsIntrinsic<[llvm_i1_ty], [llvm_ptr_ty],
+ [IntrNoMem, IntrSpeculatable, NoCapture<ArgIndex<0>>]
>;
//===----------------------------------------------------------------------===//
def int_amdgcn_s_dcache_inv_vol :
ClangBuiltin<"__builtin_amdgcn_s_dcache_inv_vol">,
- Intrinsic<[], [], [IntrNoMem, IntrHasSideEffects, IntrWillReturn]>;
+ DefaultAttrsIntrinsic<[], [], [IntrNoMem, IntrHasSideEffects]>;
def int_amdgcn_buffer_wbinvl1_vol :
ClangBuiltin<"__builtin_amdgcn_buffer_wbinvl1_vol">,
- Intrinsic<[], [], [IntrNoMem, IntrHasSideEffects, IntrWillReturn]>;
+ DefaultAttrsIntrinsic<[], [], [IntrNoMem, IntrHasSideEffects]>;
//===----------------------------------------------------------------------===//
// VI Intrinsics
llvm_i1_ty],
[IntrNoMem, IntrConvergent, IntrWillReturn,
ImmArg<ArgIndex<1>>, ImmArg<ArgIndex<2>>,
- ImmArg<ArgIndex<3>>, ImmArg<ArgIndex<4>>]>;
+ ImmArg<ArgIndex<3>>, ImmArg<ArgIndex<4>>, IntrNoCallback, IntrNoFree]>;
// llvm.amdgcn.update.dpp.i32 <old> <src> <dpp_ctrl> <row_mask> <bank_mask> <bound_ctrl>
// Should be equivalent to:
llvm_i32_ty, llvm_i32_ty, llvm_i1_ty],
[IntrNoMem, IntrConvergent, IntrWillReturn,
ImmArg<ArgIndex<2>>, ImmArg<ArgIndex<3>>,
- ImmArg<ArgIndex<4>>, ImmArg<ArgIndex<5>>]>;
+ ImmArg<ArgIndex<4>>, ImmArg<ArgIndex<5>>, IntrNoCallback, IntrNoFree]>;
def int_amdgcn_s_dcache_wb :
ClangBuiltin<"__builtin_amdgcn_s_dcache_wb">,
- Intrinsic<[], [], [IntrNoMem, IntrHasSideEffects, IntrWillReturn]>;
+ Intrinsic<[], [], [IntrNoMem, IntrHasSideEffects, IntrWillReturn, IntrNoCallback, IntrNoFree]>;
def int_amdgcn_s_dcache_wb_vol :
ClangBuiltin<"__builtin_amdgcn_s_dcache_wb_vol">,
- Intrinsic<[], [], [IntrNoMem, IntrHasSideEffects, IntrWillReturn]>;
+ Intrinsic<[], [], [IntrNoMem, IntrHasSideEffects, IntrWillReturn, IntrNoCallback, IntrNoFree]>;
def int_amdgcn_s_memrealtime :
ClangBuiltin<"__builtin_amdgcn_s_memrealtime">,
- Intrinsic<[llvm_i64_ty], [], [IntrNoMem, IntrHasSideEffects, IntrWillReturn]>;
+ Intrinsic<[llvm_i64_ty], [], [IntrNoMem, IntrHasSideEffects, IntrWillReturn, IntrNoCallback, IntrNoFree]>;
// llvm.amdgcn.ds.permute <index> <src>
def int_amdgcn_ds_permute :
ClangBuiltin<"__builtin_amdgcn_ds_permute">,
Intrinsic<[llvm_i32_ty], [llvm_i32_ty, llvm_i32_ty],
- [IntrNoMem, IntrConvergent, IntrWillReturn]>;
+ [IntrNoMem, IntrConvergent, IntrWillReturn, IntrNoCallback, IntrNoFree]>;
// llvm.amdgcn.ds.bpermute <index> <src>
def int_amdgcn_ds_bpermute :
ClangBuiltin<"__builtin_amdgcn_ds_bpermute">,
Intrinsic<[llvm_i32_ty], [llvm_i32_ty, llvm_i32_ty],
- [IntrNoMem, IntrConvergent, IntrWillReturn]>;
+ [IntrNoMem, IntrConvergent, IntrWillReturn, IntrNoCallback, IntrNoFree]>;
// llvm.amdgcn.perm <src0> <src1> <selector>
def int_amdgcn_perm :
ClangBuiltin<"__builtin_amdgcn_perm">,
Intrinsic<[llvm_i32_ty], [llvm_i32_ty, llvm_i32_ty, llvm_i32_ty],
- [IntrNoMem, IntrSpeculatable, IntrWillReturn]>;
+ [IntrNoMem, IntrSpeculatable, IntrWillReturn, IntrNoCallback, IntrNoFree]>;
//===----------------------------------------------------------------------===//
// GFX9 Intrinsics
// bit 2 = dlc on gfx10+))
// bit 4 = scc/nt on gfx90a+))
[IntrWillReturn, NoCapture<ArgIndex<0>>, NoCapture<ArgIndex<1>>,
- ImmArg<ArgIndex<2>>, ImmArg<ArgIndex<3>>, ImmArg<ArgIndex<4>>],
+ ImmArg<ArgIndex<2>>, ImmArg<ArgIndex<3>>, ImmArg<ArgIndex<4>>, IntrNoCallback, IntrNoFree],
"", [SDNPMemOperand]>;
def int_amdgcn_global_load_lds : AMDGPUGlobalLoadLDS;
Intrinsic<[llvm_i32_ty],
[llvm_i32_ty, llvm_i32_ty, llvm_i32_ty, llvm_i32_ty, llvm_i1_ty, llvm_i1_ty],
[IntrNoMem, IntrConvergent, IntrWillReturn,
- ImmArg<ArgIndex<4>>, ImmArg<ArgIndex<5>>]>;
+ ImmArg<ArgIndex<4>>, ImmArg<ArgIndex<5>>, IntrNoCallback, IntrNoFree]>;
// llvm.amdgcn.permlanex16 <old> <src0> <src1> <src2> <fi> <bound_control>
def int_amdgcn_permlanex16 : ClangBuiltin<"__builtin_amdgcn_permlanex16">,
Intrinsic<[llvm_i32_ty],
[llvm_i32_ty, llvm_i32_ty, llvm_i32_ty, llvm_i32_ty, llvm_i1_ty, llvm_i1_ty],
[IntrNoMem, IntrConvergent, IntrWillReturn,
- ImmArg<ArgIndex<4>>, ImmArg<ArgIndex<5>>]>;
+ ImmArg<ArgIndex<4>>, ImmArg<ArgIndex<5>>, IntrNoCallback, IntrNoFree]>;
// llvm.amdgcn.mov.dpp8.i32 <src> <sel>
// <sel> is a 32-bit constant whose high 8 bits must be zero which selects
Intrinsic<[llvm_anyint_ty],
[LLVMMatchType<0>, llvm_i32_ty],
[IntrNoMem, IntrConvergent, IntrWillReturn,
- ImmArg<ArgIndex<1>>]>;
+ ImmArg<ArgIndex<1>>, IntrNoCallback, IntrNoFree]>;
def int_amdgcn_s_get_waveid_in_workgroup :
ClangBuiltin<"__builtin_amdgcn_s_get_waveid_in_workgroup">,
Intrinsic<[llvm_i32_ty], [],
- [IntrNoMem, IntrHasSideEffects, IntrWillReturn]>;
+ [IntrNoMem, IntrHasSideEffects, IntrWillReturn, IntrNoCallback, IntrNoFree]>;
class AMDGPUGlobalAtomicRtn<LLVMType vt> : Intrinsic <
[vt],
[llvm_anyptr_ty, // vaddr
vt], // vdata(VGPR)
- [IntrArgMemOnly, IntrWillReturn, NoCapture<ArgIndex<0>>], "",
+ [IntrArgMemOnly, IntrWillReturn, NoCapture<ArgIndex<0>>, IntrNoCallback, IntrNoFree], "",
[SDNPMemOperand]>;
def int_amdgcn_global_atomic_csub : AMDGPUGlobalAtomicRtn<llvm_i32_ty>;
// <node_ptr> is i32 or i64.
// <ray_dir> and <ray_inv_dir> are both v3f16 or both v3f32.
def int_amdgcn_image_bvh_intersect_ray :
- Intrinsic<[llvm_v4i32_ty],
+ DefaultAttrsIntrinsic<[llvm_v4i32_ty],
[llvm_anyint_ty, llvm_float_ty, llvm_v3f32_ty, llvm_anyvector_ty,
LLVMMatchType<1>, llvm_v4i32_ty],
- [IntrReadMem, IntrWillReturn]>;
+ [IntrReadMem, IntrWillReturn, IntrNoCallback, IntrNoFree]>;
//===----------------------------------------------------------------------===//
// GFX11 Intrinsics
def int_amdgcn_permlane64 :
ClangBuiltin<"__builtin_amdgcn_permlane64">,
Intrinsic<[llvm_i32_ty], [llvm_i32_ty],
- [IntrNoMem, IntrConvergent, IntrWillReturn]>;
+ [IntrNoMem, IntrConvergent, IntrWillReturn, IntrNoCallback, IntrNoFree]>;
def int_amdgcn_ds_add_gs_reg_rtn :
ClangBuiltin<"__builtin_amdgcn_ds_add_gs_reg_rtn">,
Intrinsic<[llvm_anyint_ty], [llvm_i32_ty, llvm_i32_ty],
- [ImmArg<ArgIndex<1>>, IntrHasSideEffects, IntrWillReturn]>;
+ [ImmArg<ArgIndex<1>>, IntrHasSideEffects, IntrWillReturn, IntrNoCallback, IntrNoFree]>;
def int_amdgcn_ds_sub_gs_reg_rtn :
ClangBuiltin<"__builtin_amdgcn_ds_sub_gs_reg_rtn">,
Intrinsic<[llvm_anyint_ty], [llvm_i32_ty, llvm_i32_ty],
- [ImmArg<ArgIndex<1>>, IntrHasSideEffects, IntrWillReturn]>;
+ [ImmArg<ArgIndex<1>>, IntrHasSideEffects, IntrWillReturn, IntrNoCallback, IntrNoFree]>;
def int_amdgcn_ds_bvh_stack_rtn :
Intrinsic<
llvm_v4i32_ty, // %data1
llvm_i32_ty, // %offset
],
- [ImmArg<ArgIndex<3>>, IntrWillReturn]
+ [ImmArg<ArgIndex<3>>, IntrWillReturn, IntrNoCallback, IntrNoFree]
>;
// WMMA (Wave Matrix Multiply-Accumulate) intrinsics
AB, // %B
LLVMMatchType<0>, // %C
],
- [IntrNoMem, IntrConvergent, IntrWillReturn]
+ [IntrNoMem, IntrConvergent, IntrWillReturn, IntrNoCallback, IntrNoFree]
>;
class AMDGPUWmmaIntrinsicOPSEL<LLVMType AB, LLVMType CD> :
LLVMMatchType<0>, // %C
llvm_i1_ty, // %high
],
- [IntrNoMem, IntrConvergent, IntrWillReturn, ImmArg<ArgIndex<3>>]
+ [IntrNoMem, IntrConvergent, ImmArg<ArgIndex<3>>, IntrWillReturn, IntrNoCallback, IntrNoFree]
>;
class AMDGPUWmmaIntrinsicIU<LLVMType AB, LLVMType CD> :
LLVMMatchType<0>, // %C
llvm_i1_ty, // %clamp
],
- [IntrNoMem, IntrConvergent, IntrWillReturn, ImmArg<ArgIndex<0>>, ImmArg<ArgIndex<2>>, ImmArg<ArgIndex<5>>]
+ [IntrNoMem, IntrConvergent, ImmArg<ArgIndex<0>>, ImmArg<ArgIndex<2>>, ImmArg<ArgIndex<5>>, IntrWillReturn, IntrNoCallback, IntrNoFree]
>;
def int_amdgcn_wmma_f32_16x16x16_f16 : AMDGPUWmmaIntrinsic<llvm_v16f16_ty, llvm_anyfloat_ty>;
// %r = %a[0] * %b[0] + %a[1] * %b[1] + %c
def int_amdgcn_fdot2 :
ClangBuiltin<"__builtin_amdgcn_fdot2">,
- Intrinsic<
+ DefaultAttrsIntrinsic<
[llvm_float_ty], // %r
[
llvm_v2f16_ty, // %a
llvm_float_ty, // %c
llvm_i1_ty // %clamp
],
- [IntrNoMem, IntrSpeculatable, IntrWillReturn, ImmArg<ArgIndex<3>>]
+ [IntrNoMem, IntrSpeculatable, ImmArg<ArgIndex<3>>]
>;
// f16 %r = llvm.amdgcn.fdot2.f16.f16(v2f16 %a, v2f16 %b, f16 %c)
// %r = %a[0] * %b[0] + %a[1] * %b[1] + %c
def int_amdgcn_fdot2_f16_f16 :
ClangBuiltin<"__builtin_amdgcn_fdot2_f16_f16">,
- Intrinsic<
+ DefaultAttrsIntrinsic<
[llvm_half_ty], // %r
[
llvm_v2f16_ty, // %a
llvm_v2f16_ty, // %b
llvm_half_ty // %c
],
- [IntrNoMem, IntrSpeculatable, IntrWillReturn]
+ [IntrNoMem, IntrSpeculatable]
>;
// bf16 %r = llvm.amdgcn.fdot2.bf16.bf16(v2bf16 %a, v2bf16 %b, bf16 %c)
// %r = %a[0] * %b[0] + %a[1] * %b[1] + %c
def int_amdgcn_fdot2_bf16_bf16 :
ClangBuiltin<"__builtin_amdgcn_fdot2_bf16_bf16">,
- Intrinsic<
+ DefaultAttrsIntrinsic<
[llvm_i16_ty], // %r
[
llvm_v2i16_ty, // %a
llvm_v2i16_ty, // %b
llvm_i16_ty // %c
],
- [IntrNoMem, IntrSpeculatable, IntrWillReturn]
+ [IntrNoMem, IntrSpeculatable]
>;
// f32 %r = llvm.amdgcn.fdot2.f32.bf16(v2bf16 %a, v2bf16 %b, f32 %c, i1 %clamp)
// %r = %a[0] * %b[0] + %a[1] * %b[1] + %c
def int_amdgcn_fdot2_f32_bf16 :
ClangBuiltin<"__builtin_amdgcn_fdot2_f32_bf16">,
- Intrinsic<
+ DefaultAttrsIntrinsic<
[llvm_float_ty], // %r
[
llvm_v2i16_ty, // %a
llvm_float_ty, // %c
llvm_i1_ty // %clamp
],
- [IntrNoMem, IntrSpeculatable, IntrWillReturn, ImmArg<ArgIndex<3>>]
+ [IntrNoMem, IntrSpeculatable, ImmArg<ArgIndex<3>>]
>;
// i32 %r = llvm.amdgcn.sdot2(v2i16 %a, v2i16 %b, i32 %c, i1 %clamp)
// %r = %a[0] * %b[0] + %a[1] * %b[1] + %c
def int_amdgcn_sdot2 :
ClangBuiltin<"__builtin_amdgcn_sdot2">,
- Intrinsic<
+ DefaultAttrsIntrinsic<
[llvm_i32_ty], // %r
[
llvm_v2i16_ty, // %a
llvm_i32_ty, // %c
llvm_i1_ty // %clamp
],
- [IntrNoMem, IntrSpeculatable, IntrWillReturn, ImmArg<ArgIndex<3>>]
+ [IntrNoMem, IntrSpeculatable, ImmArg<ArgIndex<3>>]
>;
// u32 %r = llvm.amdgcn.udot2(v2u16 %a, v2u16 %b, u32 %c, i1 %clamp)
// %r = %a[0] * %b[0] + %a[1] * %b[1] + %c
def int_amdgcn_udot2 :
ClangBuiltin<"__builtin_amdgcn_udot2">,
- Intrinsic<
+ DefaultAttrsIntrinsic<
[llvm_i32_ty], // %r
[
llvm_v2i16_ty, // %a
llvm_i32_ty, // %c
llvm_i1_ty // %clamp
],
- [IntrNoMem, IntrSpeculatable, IntrWillReturn, ImmArg<ArgIndex<3>>]
+ [IntrNoMem, IntrSpeculatable, ImmArg<ArgIndex<3>>]
>;
// i32 %r = llvm.amdgcn.sdot4(v4i8 (as i32) %a, v4i8 (as i32) %b, i32 %c, i1 %clamp)
// %r = %a[0] * %b[0] + %a[1] * %b[1] + %a[2] * %b[2] + %a[3] * %b[3] + %c
def int_amdgcn_sdot4 :
ClangBuiltin<"__builtin_amdgcn_sdot4">,
- Intrinsic<
+ DefaultAttrsIntrinsic<
[llvm_i32_ty], // %r
[
llvm_i32_ty, // %a
llvm_i32_ty, // %c
llvm_i1_ty // %clamp
],
- [IntrNoMem, IntrSpeculatable, IntrWillReturn, ImmArg<ArgIndex<3>>]
+ [IntrNoMem, IntrSpeculatable, ImmArg<ArgIndex<3>>]
>;
// u32 %r = llvm.amdgcn.udot4(v4u8 (as u32) %a, v4u8 (as u32) %b, u32 %c, i1 %clamp)
// %r = %a[0] * %b[0] + %a[1] * %b[1] + %a[2] * %b[2] + %a[3] * %b[3] + %c
def int_amdgcn_udot4 :
ClangBuiltin<"__builtin_amdgcn_udot4">,
- Intrinsic<
+ DefaultAttrsIntrinsic<
[llvm_i32_ty], // %r
[
llvm_i32_ty, // %a
llvm_i32_ty, // %c
llvm_i1_ty // %clamp
],
- [IntrNoMem, IntrSpeculatable, IntrWillReturn, ImmArg<ArgIndex<3>>]
+ [IntrNoMem, IntrSpeculatable, ImmArg<ArgIndex<3>>]
>;
// i32 %r = llvm.amdgcn.sudot4(i1 %a_sign, v4i8 (as i32) %a, i1 %b_sign, v4i8 (as i32) %b, i32 %c, i1 %clamp)
// %r = %a[0] * %b[0] + %a[1] * %b[1] + %a[2] * %b[2] + %a[3] * %b[3] + %c
def int_amdgcn_sudot4 :
ClangBuiltin<"__builtin_amdgcn_sudot4">,
- Intrinsic<
+ DefaultAttrsIntrinsic<
[llvm_i32_ty], // %r
[
llvm_i1_ty, // %a_sign
llvm_i32_ty, // %c
llvm_i1_ty // %clamp
],
- [IntrNoMem, IntrSpeculatable, IntrWillReturn,
+ [IntrNoMem, IntrSpeculatable,
ImmArg<ArgIndex<0>>, ImmArg<ArgIndex<2>>, ImmArg<ArgIndex<5>>]
>;
// %a[4] * %b[4] + %a[5] * %b[5] + %a[6] * %b[6] + %a[7] * %b[7] + %c
def int_amdgcn_sdot8 :
ClangBuiltin<"__builtin_amdgcn_sdot8">,
- Intrinsic<
+ DefaultAttrsIntrinsic<
[llvm_i32_ty], // %r
[
llvm_i32_ty, // %a
llvm_i32_ty, // %c
llvm_i1_ty // %clamp
],
- [IntrNoMem, IntrSpeculatable, IntrWillReturn, ImmArg<ArgIndex<3>>]
+ [IntrNoMem, IntrSpeculatable, ImmArg<ArgIndex<3>>]
>;
// u32 %r = llvm.amdgcn.udot8(v8u4 (as u32) %a, v8u4 (as u32) %b, u32 %c, i1 %clamp)
// %a[4] * %b[4] + %a[5] * %b[5] + %a[6] * %b[6] + %a[7] * %b[7] + %c
def int_amdgcn_udot8 :
ClangBuiltin<"__builtin_amdgcn_udot8">,
- Intrinsic<
+ DefaultAttrsIntrinsic<
[llvm_i32_ty], // %r
[
llvm_i32_ty, // %a
llvm_i32_ty, // %c
llvm_i1_ty // %clamp
],
- [IntrNoMem, IntrSpeculatable, IntrWillReturn, ImmArg<ArgIndex<3>>]
+ [IntrNoMem, IntrSpeculatable, ImmArg<ArgIndex<3>>]
>;
// i32 %r = llvm.amdgcn.sudot8(i1 %a_sign, v8i4 (as i32) %a, i1 %b_sign, v8i4 (as i32) %b, i32 %c, i1 %clamp)
// %a[4] * %b[4] + %a[5] * %b[5] + %a[6] * %b[6] + %a[7] * %b[7] + %c
def int_amdgcn_sudot8 :
ClangBuiltin<"__builtin_amdgcn_sudot8">,
- Intrinsic<
+ DefaultAttrsIntrinsic<
[llvm_i32_ty], // %r
[
llvm_i1_ty, // %a_sign
llvm_i32_ty, // %c
llvm_i1_ty // %clamp
],
- [IntrNoMem, IntrSpeculatable, IntrWillReturn,
+ [IntrNoMem, IntrSpeculatable,
ImmArg<ArgIndex<0>>, ImmArg<ArgIndex<2>>, ImmArg<ArgIndex<5>>]
>;
// llvm.amdgcn.mfma.*.* vdst, srcA, srcB, srcC, cbsz, abid, blgp
class AMDGPUMfmaIntrinsic<LLVMType DestTy, LLVMType SrcABTy> :
ClangBuiltin<!subst("int", "__builtin", NAME)>,
- Intrinsic<[DestTy],
+ DefaultAttrsIntrinsic<[DestTy],
[SrcABTy, SrcABTy, DestTy,
llvm_i32_ty, llvm_i32_ty, llvm_i32_ty],
- [IntrConvergent, IntrNoMem, IntrWillReturn,
+ [IntrConvergent, IntrNoMem,
ImmArg<ArgIndex<3>>, ImmArg<ArgIndex<4>>, ImmArg<ArgIndex<5>>]>;
def int_amdgcn_mfma_f32_32x32x1f32 : AMDGPUMfmaIntrinsic<llvm_v32f32_ty, llvm_float_ty>;
// bf16 atomics use v2i16 argument since there is no bf16 data type in the llvm.
def int_amdgcn_global_atomic_fadd_v2bf16 : AMDGPUGlobalAtomicRtn<llvm_v2i16_ty>;
def int_amdgcn_flat_atomic_fadd_v2bf16 : AMDGPUGlobalAtomicRtn<llvm_v2i16_ty>;
-def int_amdgcn_ds_fadd_v2bf16 : Intrinsic<
+def int_amdgcn_ds_fadd_v2bf16 : DefaultAttrsIntrinsic<
[llvm_v2i16_ty],
[LLVMQualPointerType<llvm_v2i16_ty, 3>, llvm_v2i16_ty],
- [IntrArgMemOnly, IntrWillReturn, NoCapture<ArgIndex<0>>]>,
+ [IntrArgMemOnly, NoCapture<ArgIndex<0>>]>,
ClangBuiltin<"__builtin_amdgcn_ds_atomic_fadd_v2bf16">;
def int_amdgcn_mfma_i32_16x16x32_i8 : AMDGPUMfmaIntrinsic<llvm_v4i32_ty, llvm_i64_ty>;
// llvm.amdgcn.smfmac.?32.* vdst, srcA, srcB, srcC, index, cbsz, abid
class AMDGPUMSmfmacIntrinsic<LLVMType DestTy, LLVMType SrcA, LLVMType SrcB> :
ClangBuiltin<!subst("int", "__builtin", NAME)>,
- Intrinsic<[DestTy],
+ DefaultAttrsIntrinsic<[DestTy],
[SrcA, SrcB, DestTy, llvm_i32_ty,
llvm_i32_ty, llvm_i32_ty],
- [IntrConvergent, IntrNoMem, IntrWillReturn,
+ [IntrConvergent, IntrNoMem,
ImmArg<ArgIndex<4>>, ImmArg<ArgIndex<5>>]>;
def int_amdgcn_smfmac_f32_16x16x32_f16 : AMDGPUMSmfmacIntrinsic<llvm_v4f32_ty, llvm_v4f16_ty, llvm_v8f16_ty>;
// llvm.amdgcn.cvt.f32.bf8 float vdst, int srcA, imm byte_sel [0..3]
// byte_sel selects byte from srcA.
def int_amdgcn_cvt_f32_bf8 : ClangBuiltin<"__builtin_amdgcn_cvt_f32_bf8">,
- Intrinsic<[llvm_float_ty],
+ DefaultAttrsIntrinsic<[llvm_float_ty],
[llvm_i32_ty, llvm_i32_ty],
- [IntrNoMem, IntrWillReturn, ImmArg<ArgIndex<1>>]>;
+ [IntrNoMem, ImmArg<ArgIndex<1>>]>;
// llvm.amdgcn.cvt.f32.fp8 float vdst, int srcA, imm byte_sel [0..3]
def int_amdgcn_cvt_f32_fp8 : ClangBuiltin<"__builtin_amdgcn_cvt_f32_fp8">,
- Intrinsic<[llvm_float_ty],
+ DefaultAttrsIntrinsic<[llvm_float_ty],
[llvm_i32_ty, llvm_i32_ty],
- [IntrNoMem, IntrWillReturn, ImmArg<ArgIndex<1>>]>;
+ [IntrNoMem, ImmArg<ArgIndex<1>>]>;
// llvm.amdgcn.cvt.pk.f32.bf8 float2 vdst, int srcA, imm word_sel
// word_sel = 1 selects 2 high bytes, 0 selects 2 low bytes.
def int_amdgcn_cvt_pk_f32_bf8 : ClangBuiltin<"__builtin_amdgcn_cvt_pk_f32_bf8">,
- Intrinsic<[llvm_v2f32_ty],
+ DefaultAttrsIntrinsic<[llvm_v2f32_ty],
[llvm_i32_ty, llvm_i1_ty],
- [IntrNoMem, IntrWillReturn, ImmArg<ArgIndex<1>>]>;
+ [IntrNoMem, ImmArg<ArgIndex<1>>]>;
// llvm.amdgcn.cvt.pk.f32.fp8 float2 vdst, int srcA, imm word_sel.
def int_amdgcn_cvt_pk_f32_fp8 : ClangBuiltin<"__builtin_amdgcn_cvt_pk_f32_fp8">,
- Intrinsic<[llvm_v2f32_ty],
+ DefaultAttrsIntrinsic<[llvm_v2f32_ty],
[llvm_i32_ty, llvm_i1_ty],
- [IntrNoMem, IntrWillReturn, ImmArg<ArgIndex<1>>]>;
+ [IntrNoMem, ImmArg<ArgIndex<1>>]>;
// llvm.amdgcn.cvt.pk.bf8.f32 int vdst, float srcA, float srcB, int old, imm word_sel
// word_sel = 1 selects 2 high bytes in the vdst, 0 selects 2 low bytes.
def int_amdgcn_cvt_pk_bf8_f32 : ClangBuiltin<"__builtin_amdgcn_cvt_pk_bf8_f32">,
- Intrinsic<[llvm_i32_ty],
+ DefaultAttrsIntrinsic<[llvm_i32_ty],
[llvm_float_ty, llvm_float_ty, llvm_i32_ty, llvm_i1_ty],
- [IntrNoMem, IntrWillReturn, ImmArg<ArgIndex<3>>]>;
+ [IntrNoMem, ImmArg<ArgIndex<3>>]>;
// llvm.amdgcn.cvt.pk.fp8.f32 int vdst, float srcA, float srcB, int old, imm word_sel
def int_amdgcn_cvt_pk_fp8_f32 : ClangBuiltin<"__builtin_amdgcn_cvt_pk_fp8_f32">,
- Intrinsic<[llvm_i32_ty],
+ DefaultAttrsIntrinsic<[llvm_i32_ty],
[llvm_float_ty, llvm_float_ty, llvm_i32_ty, llvm_i1_ty],
- [IntrNoMem, IntrWillReturn, ImmArg<ArgIndex<3>>]>;
+ [IntrNoMem, ImmArg<ArgIndex<3>>]>;
// llvm.amdgcn.cvt.sr.bf8.f32 int vdst, float srcA, int srcB, int old, imm byte_sel [0..3]
// byte_sel selects byte to write into vdst.
def int_amdgcn_cvt_sr_bf8_f32 : ClangBuiltin<"__builtin_amdgcn_cvt_sr_bf8_f32">,
- Intrinsic<[llvm_i32_ty],
+ DefaultAttrsIntrinsic<[llvm_i32_ty],
[llvm_float_ty, llvm_i32_ty, llvm_i32_ty, llvm_i32_ty],
- [IntrNoMem, IntrWillReturn, ImmArg<ArgIndex<3>>]>;
+ [IntrNoMem, ImmArg<ArgIndex<3>>]>;
// llvm.amdgcn.cvt.sr.fp8.f32 int vdst, float srcA, int srcB, int old, imm byte_sel [0..3]
def int_amdgcn_cvt_sr_fp8_f32 : ClangBuiltin<"__builtin_amdgcn_cvt_sr_fp8_f32">,
- Intrinsic<[llvm_i32_ty],
+ DefaultAttrsIntrinsic<[llvm_i32_ty],
[llvm_float_ty, llvm_i32_ty, llvm_i32_ty, llvm_i32_ty],
- [IntrNoMem, IntrWillReturn, ImmArg<ArgIndex<3>>]>;
+ [IntrNoMem, ImmArg<ArgIndex<3>>]>;
//===----------------------------------------------------------------------===//
// Special Intrinsics for backend internal use only. No frontend
// should emit calls to these.
// ===----------------------------------------------------------------------===//
def int_amdgcn_if : Intrinsic<[llvm_i1_ty, llvm_anyint_ty],
- [llvm_i1_ty], [IntrConvergent, IntrWillReturn]
+ [llvm_i1_ty], [IntrConvergent, IntrWillReturn, IntrNoCallback, IntrNoFree]
>;
def int_amdgcn_else : Intrinsic<[llvm_i1_ty, llvm_anyint_ty],
- [llvm_anyint_ty], [IntrConvergent, IntrWillReturn]
+ [llvm_anyint_ty], [IntrConvergent, IntrWillReturn, IntrNoCallback, IntrNoFree]
>;
def int_amdgcn_if_break : Intrinsic<[llvm_anyint_ty],
[llvm_i1_ty, LLVMMatchType<0>],
- [IntrNoMem, IntrConvergent, IntrWillReturn]
+ [IntrNoMem, IntrConvergent, IntrWillReturn, IntrNoCallback, IntrNoFree]
>;
def int_amdgcn_loop : Intrinsic<[llvm_i1_ty],
- [llvm_anyint_ty], [IntrConvergent, IntrWillReturn]
+ [llvm_anyint_ty], [IntrConvergent, IntrWillReturn, IntrNoCallback, IntrNoFree]
>;
def int_amdgcn_end_cf : Intrinsic<[], [llvm_anyint_ty],
- [IntrConvergent, IntrWillReturn]>;
+ [IntrConvergent, IntrWillReturn, IntrNoCallback, IntrNoFree]>;
// Represent unreachable in a divergent region.
def int_amdgcn_unreachable : Intrinsic<[], [], [IntrConvergent]>;
// Emit 2.5 ulp, no denormal division. Should only be inserted by
// pass based on !fpmath metadata.
-def int_amdgcn_fdiv_fast : Intrinsic<
+def int_amdgcn_fdiv_fast : DefaultAttrsIntrinsic<
[llvm_float_ty], [llvm_float_ty, llvm_float_ty],
- [IntrNoMem, IntrSpeculatable, IntrWillReturn]
+ [IntrNoMem, IntrSpeculatable]
>;
// Represent a relocation constant.
-def int_amdgcn_reloc_constant : Intrinsic<
+def int_amdgcn_reloc_constant : DefaultAttrsIntrinsic<
[llvm_i32_ty], [llvm_metadata_ty],
- [IntrNoMem, IntrSpeculatable, IntrWillReturn]
+ [IntrNoMem, IntrSpeculatable]
>;
}