caio/clang-p2996
1
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity

[NVPTX] Update various intrinsic attributes (#140119)

Browse Source
This commit is contained in:
Alex MacLean
2025-05-16 06:31:07 -07:00
committed by GitHub
parent dd32ad1c37
commit 689a960b69
1 changed files with 116 additions and 107 deletions
Download Patch File Download Diff File Expand all files Collapse all files

223
llvm/include/llvm/IR/IntrinsicsNVVM.td
View File

@@ -831,7 +831,7 @@ let TargetPrefix = "nvvm" in {
//
// Sad
//
let IntrProperties = [IntrNoMem, Commutative, IntrSpeculatable] in {
let IntrProperties = [IntrNoMem, IntrSpeculatable] in {
foreach sign = ["", "u"] in {
def int_nvvm_sad_ # sign # s : NVVMBuiltin,
DefaultAttrsIntrinsic<[llvm_i16_ty], [llvm_i16_ty, llvm_i16_ty, llvm_i16_ty]>;
@@ -1150,41 +1150,40 @@ let TargetPrefix = "nvvm" in {
def int_nvvm_bf2h_rn # ftz : NVVMBuiltin,
DefaultAttrsIntrinsic<[llvm_i16_ty], [llvm_bfloat_ty]>;
}
}
let IntrProperties = [IntrNoMem, IntrNoCallback] in {
foreach rnd = ["rn", "rz"] in {
foreach relu = ["", "_relu"] in {
def int_nvvm_ff2bf16x2_ # rnd # relu : NVVMBuiltin,
Intrinsic<[llvm_v2bf16_ty], [llvm_float_ty, llvm_float_ty]>;
DefaultAttrsIntrinsic<[llvm_v2bf16_ty], [llvm_float_ty, llvm_float_ty]>;
def int_nvvm_ff2f16x2_ # rnd # relu : NVVMBuiltin,
Intrinsic<[llvm_v2f16_ty], [llvm_float_ty, llvm_float_ty]>;
DefaultAttrsIntrinsic<[llvm_v2f16_ty], [llvm_float_ty, llvm_float_ty]>;
def int_nvvm_f2bf16_ # rnd # relu : NVVMBuiltin,
Intrinsic<[llvm_bfloat_ty], [llvm_float_ty]>;
DefaultAttrsIntrinsic<[llvm_bfloat_ty], [llvm_float_ty]>;
}
}
foreach satfinite = ["", "_satfinite"] in {
def int_nvvm_f2tf32_rna # satfinite : NVVMBuiltin,
Intrinsic<[llvm_i32_ty], [llvm_float_ty]>;
DefaultAttrsIntrinsic<[llvm_i32_ty], [llvm_float_ty]>;
foreach rnd = ["rn", "rz"] in
foreach relu = ["", "_relu"] in
def int_nvvm_f2tf32_ # rnd # relu # satfinite : NVVMBuiltin,
Intrinsic<[llvm_i32_ty], [llvm_float_ty]>;
DefaultAttrsIntrinsic<[llvm_i32_ty], [llvm_float_ty]>;
}
foreach type = ["e4m3x2", "e5m2x2"] in {
foreach relu = ["", "_relu"] in {
def int_nvvm_ff_to_ # type # _rn # relu : NVVMBuiltin,
Intrinsic<[llvm_i16_ty], [llvm_float_ty, llvm_float_ty]>;
DefaultAttrsIntrinsic<[llvm_i16_ty], [llvm_float_ty, llvm_float_ty]>;
def int_nvvm_f16x2_to_ # type # _rn # relu : NVVMBuiltin,
Intrinsic<[llvm_i16_ty], [llvm_v2f16_ty]>;
DefaultAttrsIntrinsic<[llvm_i16_ty], [llvm_v2f16_ty]>;
def int_nvvm_ # type # _to_f16x2_rn # relu : NVVMBuiltin,
Intrinsic<[llvm_v2f16_ty], [llvm_i16_ty]>;
DefaultAttrsIntrinsic<[llvm_v2f16_ty], [llvm_i16_ty]>;
}
}
@@ -1222,8 +1221,9 @@ let TargetPrefix = "nvvm" in {
}
def int_nvvm_ue8m0x2_to_bf16x2 : NVVMBuiltin,
Intrinsic<[llvm_v2bf16_ty], [llvm_i16_ty]>;
}
DefaultAttrsIntrinsic<[llvm_v2bf16_ty], [llvm_i16_ty]>;
} // IntrProperties = [IntrNoMem, IntrSpeculatable]
// FNS
def int_nvvm_fns : NVVMBuiltin,
@@ -1440,18 +1440,22 @@ def int_nvvm_internal_addrspace_wrap :
// Move intrinsics, used in nvvm internally
def int_nvvm_move_i16 : Intrinsic<[llvm_i16_ty], [llvm_i16_ty], [IntrNoMem]>;
def int_nvvm_move_i32 : Intrinsic<[llvm_i32_ty], [llvm_i32_ty], [IntrNoMem]>;
def int_nvvm_move_i64 : Intrinsic<[llvm_i64_ty], [llvm_i64_ty], [IntrNoMem]>;
def int_nvvm_move_float : Intrinsic<[llvm_float_ty], [llvm_float_ty], [IntrNoMem]>;
def int_nvvm_move_double : Intrinsic<[llvm_double_ty], [llvm_double_ty], [IntrNoMem]>;
def int_nvvm_move_ptr : Intrinsic<[llvm_anyptr_ty], [llvm_anyptr_ty], [IntrNoMem, NoCapture<ArgIndex<0>>]>;
let IntrProperties = [IntrNoMem] in {
def int_nvvm_move_i16 : DefaultAttrsIntrinsic<[llvm_i16_ty], [llvm_i16_ty]>;
def int_nvvm_move_i32 : DefaultAttrsIntrinsic<[llvm_i32_ty], [llvm_i32_ty]>;
def int_nvvm_move_i64 : DefaultAttrsIntrinsic<[llvm_i64_ty], [llvm_i64_ty]>;
def int_nvvm_move_float : DefaultAttrsIntrinsic<[llvm_float_ty], [llvm_float_ty]>;
def int_nvvm_move_double : DefaultAttrsIntrinsic<[llvm_double_ty], [llvm_double_ty]>;
def int_nvvm_move_ptr : DefaultAttrsIntrinsic<[llvm_anyptr_ty], [llvm_anyptr_ty]>;
}
// For getting the handle from a texture or surface variable
def int_nvvm_texsurf_handle
: Intrinsic<[llvm_i64_ty], [llvm_metadata_ty, llvm_anyptr_ty], [IntrNoMem]>;
def int_nvvm_texsurf_handle_internal
: Intrinsic<[llvm_i64_ty], [llvm_anyptr_ty], [IntrNoMem]>;
let IntrProperties = [IntrNoMem, IntrSpeculatable] in {
def int_nvvm_texsurf_handle
: DefaultAttrsIntrinsic<[llvm_i64_ty], [llvm_metadata_ty, llvm_anyptr_ty]>;
def int_nvvm_texsurf_handle_internal
: DefaultAttrsIntrinsic<[llvm_i64_ty], [llvm_anyptr_ty]>;
}
/// Error / Warn
def int_nvvm_compiler_error : Intrinsic<[], [llvm_anyptr_ty]>;
@@ -1472,106 +1476,111 @@ foreach i = 0...31 in
DefaultAttrsIntrinsic<[llvm_i32_ty], [],
[IntrNoMem, IntrSpeculatable, NoUndef<RetIndex>]>;
//
// Texture Fetch
//
let IntrProperties = [IntrReadMem] in {
foreach is_unified = [true, false] in {
defvar mode = !if(is_unified, "_unified", "");
defvar addr_args = !if(is_unified, [llvm_i64_ty], [llvm_i64_ty, llvm_i64_ty]);
foreach is_unified = [true, false] in {
defvar mode = !if(is_unified, "_unified", "");
defvar addr_args = !if(is_unified, [llvm_i64_ty], [llvm_i64_ty, llvm_i64_ty]);
foreach vec = [V4F32, V4S32, V4U32] in {
foreach is_array = [true, false] in {
defvar array = !if(is_array, "_array", "");
defvar array_args = !if(is_array, [llvm_i32_ty], []<LLVMType>);
// Texture Fetch
foreach vec = [V4F32, V4S32, V4U32] in {
foreach is_array = [true, false] in {
defvar array = !if(is_array, "_array", "");
defvar array_args = !if(is_array, [llvm_i32_ty], []<LLVMType>);
def int_nvvm_tex # mode # _1d # array # _ # vec.Name # _s32
: Intrinsic<vec.Types,
!listconcat(addr_args, array_args, !listsplat(llvm_i32_ty, 1))>;
def int_nvvm_tex # mode # _1d # array # _ # vec.Name # _f32
: Intrinsic<vec.Types,
!listconcat(addr_args, array_args, !listsplat(llvm_float_ty, 1))>;
def int_nvvm_tex # mode # _1d # array # _level_ # vec.Name # _f32
: Intrinsic<vec.Types,
!listconcat(addr_args, array_args, !listsplat(llvm_float_ty, 2))>;
def int_nvvm_tex # mode # _1d # array # _grad_ # vec.Name # _f32
: Intrinsic<vec.Types,
!listconcat(addr_args, array_args, !listsplat(llvm_float_ty, 3))>;
def int_nvvm_tex # mode # _2d # array # _ # vec.Name # _s32
: Intrinsic<vec.Types,
!listconcat(addr_args, array_args, !listsplat(llvm_i32_ty, 2))>;
def int_nvvm_tex # mode # _2d # array # _ # vec.Name # _f32
: Intrinsic<vec.Types,
!listconcat(addr_args, array_args, !listsplat(llvm_float_ty, 2))>;
def int_nvvm_tex # mode # _2d # array # _level_ # vec.Name # _f32
: Intrinsic<vec.Types,
!listconcat(addr_args, array_args, !listsplat(llvm_float_ty, 3))>;
def int_nvvm_tex # mode # _2d # array # _grad_ # vec.Name # _f32
: Intrinsic<vec.Types,
!listconcat(addr_args, array_args, !listsplat(llvm_float_ty, 6))>;
if !not(is_array) then {
def int_nvvm_tex # mode # _3d_ # vec.Name # _s32
def int_nvvm_tex # mode # _1d # array # _ # vec.Name # _s32
: Intrinsic<vec.Types,
!listconcat(addr_args, !listsplat(llvm_i32_ty, 3))>;
def int_nvvm_tex # mode # _3d_ # vec.Name # _f32
!listconcat(addr_args, array_args, !listsplat(llvm_i32_ty, 1))>;
def int_nvvm_tex # mode # _1d # array # _ # vec.Name # _f32
: Intrinsic<vec.Types,
!listconcat(addr_args, !listsplat(llvm_float_ty, 3))>;
def int_nvvm_tex # mode # _3d_level_ # vec.Name # _f32
!listconcat(addr_args, array_args, !listsplat(llvm_float_ty, 1))>;
def int_nvvm_tex # mode # _1d # array # _level_ # vec.Name # _f32
: Intrinsic<vec.Types,
!listconcat(addr_args, !listsplat(llvm_float_ty, 4))>;
def int_nvvm_tex # mode # _3d_grad_ # vec.Name # _f32
!listconcat(addr_args, array_args, !listsplat(llvm_float_ty, 2))>;
def int_nvvm_tex # mode # _1d # array # _grad_ # vec.Name # _f32
: Intrinsic<vec.Types,
!listconcat(addr_args, !listsplat(llvm_float_ty, 9))>;
}
!listconcat(addr_args, array_args, !listsplat(llvm_float_ty, 3))>;
def int_nvvm_tex # mode # _cube # array # _ # vec.Name # _f32
: Intrinsic<vec.Types,
!listconcat(addr_args, array_args, !listsplat(llvm_float_ty, 3))>;
def int_nvvm_tex # mode # _cube # array # _level_ # vec.Name # _f32
: Intrinsic<vec.Types,
!listconcat(addr_args, array_args, !listsplat(llvm_float_ty, 4))>;
if is_unified then
def int_nvvm_tex # mode # _cube # array # _grad_ # vec.Name # _f32
def int_nvvm_tex # mode # _2d # array # _ # vec.Name # _s32
: Intrinsic<vec.Types,
!listconcat(addr_args, array_args, !listsplat(llvm_float_ty, 9))>;
} // is_array
!listconcat(addr_args, array_args, !listsplat(llvm_i32_ty, 2))>;
def int_nvvm_tex # mode # _2d # array # _ # vec.Name # _f32
: Intrinsic<vec.Types,
!listconcat(addr_args, array_args, !listsplat(llvm_float_ty, 2))>;
def int_nvvm_tex # mode # _2d # array # _level_ # vec.Name # _f32
: Intrinsic<vec.Types,
!listconcat(addr_args, array_args, !listsplat(llvm_float_ty, 3))>;
def int_nvvm_tex # mode # _2d # array # _grad_ # vec.Name # _f32
: Intrinsic<vec.Types,
!listconcat(addr_args, array_args, !listsplat(llvm_float_ty, 6))>;
foreach comp = ["r", "g", "b", "a"] in {
def int_nvvm_tld4 # mode # _ # comp # _2d_ # vec.Name # _f32
: Intrinsic<vec.Types,
!listconcat(addr_args, !listsplat(llvm_float_ty, 2))>;
} // comp
} // vec
} // is_unified
if !not(is_array) then {
def int_nvvm_tex # mode # _3d_ # vec.Name # _s32
: Intrinsic<vec.Types,
!listconcat(addr_args, !listsplat(llvm_i32_ty, 3))>;
def int_nvvm_tex # mode # _3d_ # vec.Name # _f32
: Intrinsic<vec.Types,
!listconcat(addr_args, !listsplat(llvm_float_ty, 3))>;
def int_nvvm_tex # mode # _3d_level_ # vec.Name # _f32
: Intrinsic<vec.Types,
!listconcat(addr_args, !listsplat(llvm_float_ty, 4))>;
def int_nvvm_tex # mode # _3d_grad_ # vec.Name # _f32
: Intrinsic<vec.Types,
!listconcat(addr_args, !listsplat(llvm_float_ty, 9))>;
}
def int_nvvm_tex # mode # _cube # array # _ # vec.Name # _f32
: Intrinsic<vec.Types,
!listconcat(addr_args, array_args, !listsplat(llvm_float_ty, 3))>;
def int_nvvm_tex # mode # _cube # array # _level_ # vec.Name # _f32
: Intrinsic<vec.Types,
!listconcat(addr_args, array_args, !listsplat(llvm_float_ty, 4))>;
if is_unified then
def int_nvvm_tex # mode # _cube # array # _grad_ # vec.Name # _f32
: Intrinsic<vec.Types,
!listconcat(addr_args, array_args, !listsplat(llvm_float_ty, 9))>;
} // is_array
foreach comp = ["r", "g", "b", "a"] in {
def int_nvvm_tld4 # mode # _ # comp # _2d_ # vec.Name # _f32
: Intrinsic<vec.Types,
!listconcat(addr_args, !listsplat(llvm_float_ty, 2))>;
} // comp
} // vec
} // is_unified
} // IntrProperties = [IntrReadMem]
//=== Surface Load
foreach clamp = ["clamp", "trap", "zero"] in {
foreach vec = [TV_I8, TV_I16, TV_I32, TV_I64,
TV_V2I8, TV_V2I16, TV_V2I32, TV_V2I64,
TV_V4I8, TV_V4I16, TV_V4I32] in {
let IntrProperties = [IntrReadMem] in {
foreach clamp = ["clamp", "trap", "zero"] in {
foreach vec = [TV_I8, TV_I16, TV_I32, TV_I64,
TV_V2I8, TV_V2I16, TV_V2I32, TV_V2I64,
TV_V4I8, TV_V4I16, TV_V4I32] in {
def int_nvvm_suld_1d_ # vec.Name # _ # clamp
: Intrinsic<vec.Types,
[llvm_i64_ty, llvm_i32_ty]>;
def int_nvvm_suld_1d_ # vec.Name # _ # clamp
: Intrinsic<vec.Types,
[llvm_i64_ty, llvm_i32_ty]>;
def int_nvvm_suld_1d_array_ # vec.Name # _ # clamp
: Intrinsic<vec.Types,
[llvm_i64_ty, llvm_i32_ty, llvm_i32_ty]>;
def int_nvvm_suld_1d_array_ # vec.Name # _ # clamp
: Intrinsic<vec.Types,
[llvm_i64_ty, llvm_i32_ty, llvm_i32_ty]>;
def int_nvvm_suld_2d_ # vec.Name # _ # clamp
: Intrinsic<vec.Types,
[llvm_i64_ty, llvm_i32_ty, llvm_i32_ty]>;
def int_nvvm_suld_2d_ # vec.Name # _ # clamp
: Intrinsic<vec.Types,
[llvm_i64_ty, llvm_i32_ty, llvm_i32_ty]>;
def int_nvvm_suld_2d_array_ # vec.Name # _ # clamp
: Intrinsic<vec.Types,
[llvm_i64_ty, llvm_i32_ty, llvm_i32_ty, llvm_i32_ty]>;
def int_nvvm_suld_2d_array_ # vec.Name # _ # clamp
: Intrinsic<vec.Types,
[llvm_i64_ty, llvm_i32_ty, llvm_i32_ty, llvm_i32_ty]>;
def int_nvvm_suld_3d_ # vec.Name # _ # clamp
: Intrinsic<vec.Types,
[llvm_i64_ty, llvm_i32_ty, llvm_i32_ty, llvm_i32_ty]>;
} // vec
} // clamp
def int_nvvm_suld_3d_ # vec.Name # _ # clamp
: Intrinsic<vec.Types,
[llvm_i64_ty, llvm_i32_ty, llvm_i32_ty, llvm_i32_ty]>;
} // vec
} // clamp
} // IntrProperties = [IntrReadMem]
//===- Texture Query ------------------------------------------------------===//