"}}")))),
Float32Regs, Int16Regs, int_nvvm_h2f>;
-def : Pat<(f32 (f16_to_f32 Int16Regs:$a)),
+def : Pat<(f32 (f16_to_fp Int16Regs:$a)),
(CVT_f32_f16 Int16Regs:$a, CvtNONE)>;
-def : Pat<(i16 (f32_to_f16 Float32Regs:$a)),
+def : Pat<(i16 (fp_to_f16 Float32Regs:$a)),
(CVT_f16_f32 Float32Regs:$a, CvtRN_FTZ)>, Requires<[doF32FTZ]>;
-def : Pat<(i16 (f32_to_f16 Float32Regs:$a)),
+def : Pat<(i16 (fp_to_f16 Float32Regs:$a)),
(CVT_f16_f32 Float32Regs:$a, CvtRN)>;
+def : Pat<(f64 (f16_to_fp Int16Regs:$a)),
+ (CVT_f64_f16 Int16Regs:$a, CvtNONE)>;
+def : Pat<(i16 (fp_to_f16 Float64Regs:$a)),
+ (CVT_f16_f64 Float64Regs:$a, CvtRN)>;
+
//
// Bitcast
//
(atomic_load_max_32 node:$a, node:$b)>;
def atomic_load_max_32_gen: ATOMIC_GENERIC_CHK<(ops node:$a, node:$b),
(atomic_load_max_32 node:$a, node:$b)>;
+def atomic_load_max_64_g: ATOMIC_GLOBAL_CHK<(ops node:$a, node:$b)
+ , (atomic_load_max_64 node:$a, node:$b)>;
+def atomic_load_max_64_s: ATOMIC_SHARED_CHK<(ops node:$a, node:$b),
+ (atomic_load_max_64 node:$a, node:$b)>;
+def atomic_load_max_64_gen: ATOMIC_GENERIC_CHK<(ops node:$a, node:$b),
+ (atomic_load_max_64 node:$a, node:$b)>;
def atomic_load_umax_32_g: ATOMIC_GLOBAL_CHK<(ops node:$a, node:$b),
(atomic_load_umax_32 node:$a, node:$b)>;
def atomic_load_umax_32_s: ATOMIC_SHARED_CHK<(ops node:$a, node:$b),
(atomic_load_umax_32 node:$a, node:$b)>;
def atomic_load_umax_32_gen: ATOMIC_GENERIC_CHK<(ops node:$a, node:$b),
(atomic_load_umax_32 node:$a, node:$b)>;
+def atomic_load_umax_64_g: ATOMIC_GLOBAL_CHK<(ops node:$a, node:$b),
+ (atomic_load_umax_64 node:$a, node:$b)>;
+def atomic_load_umax_64_s: ATOMIC_SHARED_CHK<(ops node:$a, node:$b),
+ (atomic_load_umax_64 node:$a, node:$b)>;
+def atomic_load_umax_64_gen: ATOMIC_GENERIC_CHK<(ops node:$a, node:$b),
+ (atomic_load_umax_64 node:$a, node:$b)>;
defm INT_PTX_ATOM_LOAD_MAX_G_32 : F_ATOMIC_2<Int32Regs, ".global", ".s32",
".max", atomic_load_max_32_g, i32imm, imm, hasAtomRedG32>;
atomic_load_max_32_gen, i32imm, imm, hasAtomRedGen32>;
defm INT_PTX_ATOM_LOAD_MAX_GEN_32_USE_G : F_ATOMIC_2<Int32Regs, ".global",
".s32", ".max", atomic_load_max_32_gen, i32imm, imm, useAtomRedG32forGen32>;
+defm INT_PTX_ATOM_LOAD_MAX_G_64 : F_ATOMIC_2<Int64Regs, ".global", ".s64",
+ ".max", atomic_load_max_64_g, i64imm, imm, hasAtomRedG64>;
+defm INT_PTX_ATOM_LOAD_MAX_S_64 : F_ATOMIC_2<Int64Regs, ".shared", ".s64",
+ ".max", atomic_load_max_64_s, i64imm, imm, hasAtomRedS64>;
+defm INT_PTX_ATOM_LOAD_MAX_GEN_64 : F_ATOMIC_2<Int64Regs, "", ".s64", ".max",
+ atomic_load_max_64_gen, i64imm, imm, hasAtomRedGen64>;
+defm INT_PTX_ATOM_LOAD_MAX_GEN_64_USE_G : F_ATOMIC_2<Int64Regs, ".global",
+ ".s64", ".max", atomic_load_max_64_gen, i64imm, imm, useAtomRedG64forGen64>;
defm INT_PTX_ATOM_LOAD_UMAX_G_32 : F_ATOMIC_2<Int32Regs, ".global", ".u32",
".max", atomic_load_umax_32_g, i32imm, imm, hasAtomRedG32>;
defm INT_PTX_ATOM_LOAD_UMAX_S_32 : F_ATOMIC_2<Int32Regs, ".shared", ".u32",
atomic_load_umax_32_gen, i32imm, imm, hasAtomRedGen32>;
defm INT_PTX_ATOM_LOAD_UMAX_GEN_32_USE_G : F_ATOMIC_2<Int32Regs, ".global",
".u32", ".max", atomic_load_umax_32_gen, i32imm, imm, useAtomRedG32forGen32>;
+defm INT_PTX_ATOM_LOAD_UMAX_G_64 : F_ATOMIC_2<Int64Regs, ".global", ".u64",
+ ".max", atomic_load_umax_64_g, i64imm, imm, hasAtomRedG64>;
+defm INT_PTX_ATOM_LOAD_UMAX_S_64 : F_ATOMIC_2<Int64Regs, ".shared", ".u64",
+ ".max", atomic_load_umax_64_s, i64imm, imm, hasAtomRedS64>;
+defm INT_PTX_ATOM_LOAD_UMAX_GEN_64 : F_ATOMIC_2<Int64Regs, "", ".u64", ".max",
+ atomic_load_umax_64_gen, i64imm, imm, hasAtomRedGen64>;
+defm INT_PTX_ATOM_LOAD_UMAX_GEN_64_USE_G : F_ATOMIC_2<Int64Regs, ".global",
+ ".u64", ".max", atomic_load_umax_64_gen, i64imm, imm, useAtomRedG64forGen64>;
// atom_min
(atomic_load_min_32 node:$a, node:$b)>;
def atomic_load_min_32_gen: ATOMIC_GENERIC_CHK<(ops node:$a, node:$b),
(atomic_load_min_32 node:$a, node:$b)>;
+def atomic_load_min_64_g: ATOMIC_GLOBAL_CHK<(ops node:$a, node:$b),
+ (atomic_load_min_64 node:$a, node:$b)>;
+def atomic_load_min_64_s: ATOMIC_SHARED_CHK<(ops node:$a, node:$b),
+ (atomic_load_min_64 node:$a, node:$b)>;
+def atomic_load_min_64_gen: ATOMIC_GENERIC_CHK<(ops node:$a, node:$b),
+ (atomic_load_min_64 node:$a, node:$b)>;
def atomic_load_umin_32_g: ATOMIC_GLOBAL_CHK<(ops node:$a, node:$b),
(atomic_load_umin_32 node:$a, node:$b)>;
def atomic_load_umin_32_s: ATOMIC_SHARED_CHK<(ops node:$a, node:$b),
(atomic_load_umin_32 node:$a, node:$b)>;
def atomic_load_umin_32_gen: ATOMIC_GENERIC_CHK<(ops node:$a, node:$b),
(atomic_load_umin_32 node:$a, node:$b)>;
+def atomic_load_umin_64_g: ATOMIC_GLOBAL_CHK<(ops node:$a, node:$b),
+ (atomic_load_umin_64 node:$a, node:$b)>;
+def atomic_load_umin_64_s: ATOMIC_SHARED_CHK<(ops node:$a, node:$b),
+ (atomic_load_umin_64 node:$a, node:$b)>;
+def atomic_load_umin_64_gen: ATOMIC_GENERIC_CHK<(ops node:$a, node:$b),
+ (atomic_load_umin_64 node:$a, node:$b)>;
defm INT_PTX_ATOM_LOAD_MIN_G_32 : F_ATOMIC_2<Int32Regs, ".global", ".s32",
".min", atomic_load_min_32_g, i32imm, imm, hasAtomRedG32>;
atomic_load_min_32_gen, i32imm, imm, hasAtomRedGen32>;
defm INT_PTX_ATOM_LOAD_MIN_GEN_32_USE_G : F_ATOMIC_2<Int32Regs, ".global",
".s32", ".min", atomic_load_min_32_gen, i32imm, imm, useAtomRedG32forGen32>;
+defm INT_PTX_ATOM_LOAD_MIN_G_64 : F_ATOMIC_2<Int64Regs, ".global", ".s64",
+ ".min", atomic_load_min_64_g, i64imm, imm, hasAtomRedG64>;
+defm INT_PTX_ATOM_LOAD_MIN_S_64 : F_ATOMIC_2<Int64Regs, ".shared", ".s64",
+ ".min", atomic_load_min_64_s, i64imm, imm, hasAtomRedS64>;
+defm INT_PTX_ATOM_LOAD_MIN_GEN_64 : F_ATOMIC_2<Int64Regs, "", ".s64", ".min",
+ atomic_load_min_64_gen, i64imm, imm, hasAtomRedGen64>;
+defm INT_PTX_ATOM_LOAD_MIN_GEN_64_USE_G : F_ATOMIC_2<Int64Regs, ".global",
+ ".s64", ".min", atomic_load_min_64_gen, i64imm, imm, useAtomRedG64forGen64>;
defm INT_PTX_ATOM_LOAD_UMIN_G_32 : F_ATOMIC_2<Int32Regs, ".global", ".u32",
".min", atomic_load_umin_32_g, i32imm, imm, hasAtomRedG32>;
defm INT_PTX_ATOM_LOAD_UMIN_S_32 : F_ATOMIC_2<Int32Regs, ".shared", ".u32",
atomic_load_umin_32_gen, i32imm, imm, hasAtomRedGen32>;
defm INT_PTX_ATOM_LOAD_UMIN_GEN_32_USE_G : F_ATOMIC_2<Int32Regs, ".global",
".u32", ".min", atomic_load_umin_32_gen, i32imm, imm, useAtomRedG32forGen32>;
+defm INT_PTX_ATOM_LOAD_UMIN_G_64 : F_ATOMIC_2<Int64Regs, ".global", ".u64",
+ ".min", atomic_load_umin_64_g, i64imm, imm, hasAtomRedG64>;
+defm INT_PTX_ATOM_LOAD_UMIN_S_64 : F_ATOMIC_2<Int64Regs, ".shared", ".u64",
+ ".min", atomic_load_umin_64_s, i64imm, imm, hasAtomRedS64>;
+defm INT_PTX_ATOM_LOAD_UMIN_GEN_64 : F_ATOMIC_2<Int64Regs, "", ".u64", ".min",
+ atomic_load_umin_64_gen, i64imm, imm, hasAtomRedGen64>;
+defm INT_PTX_ATOM_LOAD_UMIN_GEN_64_USE_G : F_ATOMIC_2<Int64Regs, ".global",
+ ".u64", ".min", atomic_load_umin_64_gen, i64imm, imm, useAtomRedG64forGen64>;
// atom_inc atom_dec
(atomic_load_and_32 node:$a, node:$b)>;
def atomic_load_and_32_gen: ATOMIC_GENERIC_CHK<(ops node:$a, node:$b),
(atomic_load_and_32 node:$a, node:$b)>;
+def atomic_load_and_64_g: ATOMIC_GLOBAL_CHK<(ops node:$a, node:$b),
+ (atomic_load_and_64 node:$a, node:$b)>;
+def atomic_load_and_64_s: ATOMIC_SHARED_CHK<(ops node:$a, node:$b),
+ (atomic_load_and_64 node:$a, node:$b)>;
+def atomic_load_and_64_gen: ATOMIC_GENERIC_CHK<(ops node:$a, node:$b),
+ (atomic_load_and_64 node:$a, node:$b)>;
defm INT_PTX_ATOM_AND_G_32 : F_ATOMIC_2<Int32Regs, ".global", ".b32", ".and",
atomic_load_and_32_g, i32imm, imm, hasAtomRedG32>;
atomic_load_and_32_gen, i32imm, imm, hasAtomRedGen32>;
defm INT_PTX_ATOM_AND_GEN_32_USE_G : F_ATOMIC_2<Int32Regs, ".global", ".b32",
".and", atomic_load_and_32_gen, i32imm, imm, useAtomRedG32forGen32>;
+defm INT_PTX_ATOM_AND_G_64 : F_ATOMIC_2<Int64Regs, ".global", ".b64", ".and",
+ atomic_load_and_64_g, i64imm, imm, hasAtomRedG64>;
+defm INT_PTX_ATOM_AND_S_64 : F_ATOMIC_2<Int64Regs, ".shared", ".b64", ".and",
+ atomic_load_and_64_s, i64imm, imm, hasAtomRedS64>;
+defm INT_PTX_ATOM_AND_GEN_64 : F_ATOMIC_2<Int64Regs, "", ".b64", ".and",
+ atomic_load_and_64_gen, i64imm, imm, hasAtomRedGen64>;
+defm INT_PTX_ATOM_AND_GEN_64_USE_G : F_ATOMIC_2<Int64Regs, ".global", ".b64",
+ ".and", atomic_load_and_64_gen, i64imm, imm, useAtomRedG64forGen64>;
// atom_or
(atomic_load_or_32 node:$a, node:$b)>;
def atomic_load_or_32_gen: ATOMIC_GENERIC_CHK<(ops node:$a, node:$b),
(atomic_load_or_32 node:$a, node:$b)>;
+def atomic_load_or_64_g: ATOMIC_GLOBAL_CHK<(ops node:$a, node:$b),
+ (atomic_load_or_64 node:$a, node:$b)>;
+def atomic_load_or_64_s: ATOMIC_SHARED_CHK<(ops node:$a, node:$b),
+ (atomic_load_or_64 node:$a, node:$b)>;
+def atomic_load_or_64_gen: ATOMIC_GENERIC_CHK<(ops node:$a, node:$b),
+ (atomic_load_or_64 node:$a, node:$b)>;
defm INT_PTX_ATOM_OR_G_32 : F_ATOMIC_2<Int32Regs, ".global", ".b32", ".or",
atomic_load_or_32_g, i32imm, imm, hasAtomRedG32>;
".or", atomic_load_or_32_gen, i32imm, imm, useAtomRedG32forGen32>;
defm INT_PTX_ATOM_OR_S_32 : F_ATOMIC_2<Int32Regs, ".shared", ".b32", ".or",
atomic_load_or_32_s, i32imm, imm, hasAtomRedS32>;
+defm INT_PTX_ATOM_OR_G_64 : F_ATOMIC_2<Int64Regs, ".global", ".b64", ".or",
+ atomic_load_or_64_g, i64imm, imm, hasAtomRedG64>;
+defm INT_PTX_ATOM_OR_GEN_64 : F_ATOMIC_2<Int64Regs, "", ".b64", ".or",
+ atomic_load_or_64_gen, i64imm, imm, hasAtomRedGen64>;
+defm INT_PTX_ATOM_OR_GEN_64_USE_G : F_ATOMIC_2<Int64Regs, ".global", ".b64",
+ ".or", atomic_load_or_64_gen, i64imm, imm, useAtomRedG64forGen64>;
+defm INT_PTX_ATOM_OR_S_64 : F_ATOMIC_2<Int64Regs, ".shared", ".b64", ".or",
+ atomic_load_or_64_s, i64imm, imm, hasAtomRedS64>;
// atom_xor
(atomic_load_xor_32 node:$a, node:$b)>;
def atomic_load_xor_32_gen: ATOMIC_GENERIC_CHK<(ops node:$a, node:$b),
(atomic_load_xor_32 node:$a, node:$b)>;
+def atomic_load_xor_64_g: ATOMIC_GLOBAL_CHK<(ops node:$a, node:$b),
+ (atomic_load_xor_64 node:$a, node:$b)>;
+def atomic_load_xor_64_s: ATOMIC_SHARED_CHK<(ops node:$a, node:$b),
+ (atomic_load_xor_64 node:$a, node:$b)>;
+def atomic_load_xor_64_gen: ATOMIC_GENERIC_CHK<(ops node:$a, node:$b),
+ (atomic_load_xor_64 node:$a, node:$b)>;
defm INT_PTX_ATOM_XOR_G_32 : F_ATOMIC_2<Int32Regs, ".global", ".b32", ".xor",
atomic_load_xor_32_g, i32imm, imm, hasAtomRedG32>;
atomic_load_xor_32_gen, i32imm, imm, hasAtomRedGen32>;
defm INT_PTX_ATOM_XOR_GEN_32_USE_G : F_ATOMIC_2<Int32Regs, ".global", ".b32",
".xor", atomic_load_xor_32_gen, i32imm, imm, useAtomRedG32forGen32>;
+defm INT_PTX_ATOM_XOR_G_64 : F_ATOMIC_2<Int64Regs, ".global", ".b64", ".xor",
+ atomic_load_xor_64_g, i64imm, imm, hasAtomRedG64>;
+defm INT_PTX_ATOM_XOR_S_64 : F_ATOMIC_2<Int64Regs, ".shared", ".b64", ".xor",
+ atomic_load_xor_64_s, i64imm, imm, hasAtomRedS64>;
+defm INT_PTX_ATOM_XOR_GEN_64 : F_ATOMIC_2<Int64Regs, "", ".b64", ".xor",
+ atomic_load_xor_64_gen, i64imm, imm, hasAtomRedGen64>;
+defm INT_PTX_ATOM_XOR_GEN_64_USE_G : F_ATOMIC_2<Int64Regs, ".global", ".b64",
+ ".xor", atomic_load_xor_64_gen, i64imm, imm, useAtomRedG64forGen64>;
// atom_cas
// Support for ldu on sm_20 or later
//-----------------------------------
-def ldu_i8 : PatFrag<(ops node:$ptr), (int_nvvm_ldu_global_i node:$ptr), [{
- MemIntrinsicSDNode *M = cast<MemIntrinsicSDNode>(N);
- return M->getMemoryVT() == MVT::i8;
-}]>;
-
// Scalar
-// @TODO: Revisit this, Changed imemAny to imem
-multiclass LDU_G<string TyStr, NVPTXRegClass regclass, Intrinsic IntOp> {
- def areg: NVPTXInst<(outs regclass:$result), (ins Int32Regs:$src),
- !strconcat("ldu.global.", TyStr),
- [(set regclass:$result, (IntOp Int32Regs:$src))]>, Requires<[hasLDU]>;
- def areg64: NVPTXInst<(outs regclass:$result), (ins Int64Regs:$src),
- !strconcat("ldu.global.", TyStr),
- [(set regclass:$result, (IntOp Int64Regs:$src))]>, Requires<[hasLDU]>;
- def avar: NVPTXInst<(outs regclass:$result), (ins imem:$src),
- !strconcat("ldu.global.", TyStr),
- [(set regclass:$result, (IntOp (Wrapper tglobaladdr:$src)))]>,
- Requires<[hasLDU]>;
- def ari : NVPTXInst<(outs regclass:$result), (ins MEMri:$src),
- !strconcat("ldu.global.", TyStr),
- [(set regclass:$result, (IntOp ADDRri:$src))]>, Requires<[hasLDU]>;
- def ari64 : NVPTXInst<(outs regclass:$result), (ins MEMri64:$src),
- !strconcat("ldu.global.", TyStr),
- [(set regclass:$result, (IntOp ADDRri64:$src))]>, Requires<[hasLDU]>;
-}
-
-multiclass LDU_G_NOINTRIN<string TyStr, NVPTXRegClass regclass, PatFrag IntOp> {
+multiclass LDU_G<string TyStr, NVPTXRegClass regclass> {
def areg: NVPTXInst<(outs regclass:$result), (ins Int32Regs:$src),
!strconcat("ldu.global.", TyStr),
- [(set regclass:$result, (IntOp Int32Regs:$src))]>, Requires<[hasLDU]>;
+ []>, Requires<[hasLDU]>;
def areg64: NVPTXInst<(outs regclass:$result), (ins Int64Regs:$src),
!strconcat("ldu.global.", TyStr),
- [(set regclass:$result, (IntOp Int64Regs:$src))]>, Requires<[hasLDU]>;
- def avar: NVPTXInst<(outs regclass:$result), (ins imem:$src),
+ []>, Requires<[hasLDU]>;
+ def avar: NVPTXInst<(outs regclass:$result), (ins imemAny:$src),
!strconcat("ldu.global.", TyStr),
- [(set regclass:$result, (IntOp (Wrapper tglobaladdr:$src)))]>,
- Requires<[hasLDU]>;
+ []>, Requires<[hasLDU]>;
def ari : NVPTXInst<(outs regclass:$result), (ins MEMri:$src),
!strconcat("ldu.global.", TyStr),
- [(set regclass:$result, (IntOp ADDRri:$src))]>, Requires<[hasLDU]>;
+ []>, Requires<[hasLDU]>;
def ari64 : NVPTXInst<(outs regclass:$result), (ins MEMri64:$src),
!strconcat("ldu.global.", TyStr),
- [(set regclass:$result, (IntOp ADDRri64:$src))]>, Requires<[hasLDU]>;
+ []>, Requires<[hasLDU]>;
}
-defm INT_PTX_LDU_GLOBAL_i8 : LDU_G_NOINTRIN<"u8 \t$result, [$src];", Int16Regs,
- ldu_i8>;
-defm INT_PTX_LDU_GLOBAL_i16 : LDU_G<"u16 \t$result, [$src];", Int16Regs,
-int_nvvm_ldu_global_i>;
-defm INT_PTX_LDU_GLOBAL_i32 : LDU_G<"u32 \t$result, [$src];", Int32Regs,
-int_nvvm_ldu_global_i>;
-defm INT_PTX_LDU_GLOBAL_i64 : LDU_G<"u64 \t$result, [$src];", Int64Regs,
-int_nvvm_ldu_global_i>;
-defm INT_PTX_LDU_GLOBAL_f32 : LDU_G<"f32 \t$result, [$src];", Float32Regs,
-int_nvvm_ldu_global_f>;
-defm INT_PTX_LDU_GLOBAL_f64 : LDU_G<"f64 \t$result, [$src];", Float64Regs,
-int_nvvm_ldu_global_f>;
-defm INT_PTX_LDU_GLOBAL_p32 : LDU_G<"u32 \t$result, [$src];", Int32Regs,
-int_nvvm_ldu_global_p>;
-defm INT_PTX_LDU_GLOBAL_p64 : LDU_G<"u64 \t$result, [$src];", Int64Regs,
-int_nvvm_ldu_global_p>;
+defm INT_PTX_LDU_GLOBAL_i8 : LDU_G<"u8 \t$result, [$src];", Int16Regs>;
+defm INT_PTX_LDU_GLOBAL_i16 : LDU_G<"u16 \t$result, [$src];", Int16Regs>;
+defm INT_PTX_LDU_GLOBAL_i32 : LDU_G<"u32 \t$result, [$src];", Int32Regs>;
+defm INT_PTX_LDU_GLOBAL_i64 : LDU_G<"u64 \t$result, [$src];", Int64Regs>;
+defm INT_PTX_LDU_GLOBAL_f32 : LDU_G<"f32 \t$result, [$src];", Float32Regs>;
+defm INT_PTX_LDU_GLOBAL_f64 : LDU_G<"f64 \t$result, [$src];", Float64Regs>;
+defm INT_PTX_LDU_GLOBAL_p32 : LDU_G<"u32 \t$result, [$src];", Int32Regs>;
+defm INT_PTX_LDU_GLOBAL_p64 : LDU_G<"u64 \t$result, [$src];", Int64Regs>;
// vector
// Support for ldg on sm_35 or later
//-----------------------------------
-def ldg_i8 : PatFrag<(ops node:$ptr), (int_nvvm_ldg_global_i node:$ptr), [{
- MemIntrinsicSDNode *M = cast<MemIntrinsicSDNode>(N);
- return M->getMemoryVT() == MVT::i8;
-}]>;
-
-multiclass LDG_G<string TyStr, NVPTXRegClass regclass, Intrinsic IntOp> {
+multiclass LDG_G<string TyStr, NVPTXRegClass regclass> {
def areg: NVPTXInst<(outs regclass:$result), (ins Int32Regs:$src),
!strconcat("ld.global.nc.", TyStr),
- [(set regclass:$result, (IntOp Int32Regs:$src))]>, Requires<[hasLDG]>;
+ []>, Requires<[hasLDG]>;
def areg64: NVPTXInst<(outs regclass:$result), (ins Int64Regs:$src),
!strconcat("ld.global.nc.", TyStr),
- [(set regclass:$result, (IntOp Int64Regs:$src))]>, Requires<[hasLDG]>;
- def avar: NVPTXInst<(outs regclass:$result), (ins imem:$src),
+ []>, Requires<[hasLDG]>;
+ def avar: NVPTXInst<(outs regclass:$result), (ins imemAny:$src),
!strconcat("ld.global.nc.", TyStr),
- [(set regclass:$result, (IntOp (Wrapper tglobaladdr:$src)))]>,
- Requires<[hasLDG]>;
+ []>, Requires<[hasLDG]>;
def ari : NVPTXInst<(outs regclass:$result), (ins MEMri:$src),
!strconcat("ld.global.nc.", TyStr),
- [(set regclass:$result, (IntOp ADDRri:$src))]>, Requires<[hasLDG]>;
+ []>, Requires<[hasLDG]>;
def ari64 : NVPTXInst<(outs regclass:$result), (ins MEMri64:$src),
!strconcat("ld.global.nc.", TyStr),
- [(set regclass:$result, (IntOp ADDRri64:$src))]>, Requires<[hasLDG]>;
-}
-
-multiclass LDG_G_NOINTRIN<string TyStr, NVPTXRegClass regclass, PatFrag IntOp> {
- def areg: NVPTXInst<(outs regclass:$result), (ins Int32Regs:$src),
- !strconcat("ld.global.nc.", TyStr),
- [(set regclass:$result, (IntOp Int32Regs:$src))]>, Requires<[hasLDG]>;
- def areg64: NVPTXInst<(outs regclass:$result), (ins Int64Regs:$src),
- !strconcat("ld.global.nc.", TyStr),
- [(set regclass:$result, (IntOp Int64Regs:$src))]>, Requires<[hasLDG]>;
- def avar: NVPTXInst<(outs regclass:$result), (ins imem:$src),
- !strconcat("ld.global.nc.", TyStr),
- [(set regclass:$result, (IntOp (Wrapper tglobaladdr:$src)))]>,
- Requires<[hasLDG]>;
- def ari : NVPTXInst<(outs regclass:$result), (ins MEMri:$src),
- !strconcat("ld.global.nc.", TyStr),
- [(set regclass:$result, (IntOp ADDRri:$src))]>, Requires<[hasLDG]>;
- def ari64 : NVPTXInst<(outs regclass:$result), (ins MEMri64:$src),
- !strconcat("ld.global.nc.", TyStr),
- [(set regclass:$result, (IntOp ADDRri64:$src))]>, Requires<[hasLDG]>;
+ []>, Requires<[hasLDG]>;
}
defm INT_PTX_LDG_GLOBAL_i8
- : LDG_G_NOINTRIN<"u8 \t$result, [$src];", Int16Regs, ldg_i8>;
+ : LDG_G<"u8 \t$result, [$src];", Int16Regs>;
defm INT_PTX_LDG_GLOBAL_i16
- : LDG_G<"u16 \t$result, [$src];", Int16Regs, int_nvvm_ldg_global_i>;
+ : LDG_G<"u16 \t$result, [$src];", Int16Regs>;
defm INT_PTX_LDG_GLOBAL_i32
- : LDG_G<"u32 \t$result, [$src];", Int32Regs, int_nvvm_ldg_global_i>;
+ : LDG_G<"u32 \t$result, [$src];", Int32Regs>;
defm INT_PTX_LDG_GLOBAL_i64
- : LDG_G<"u64 \t$result, [$src];", Int64Regs, int_nvvm_ldg_global_i>;
+ : LDG_G<"u64 \t$result, [$src];", Int64Regs>;
defm INT_PTX_LDG_GLOBAL_f32
- : LDG_G<"f32 \t$result, [$src];", Float32Regs, int_nvvm_ldg_global_f>;
+ : LDG_G<"f32 \t$result, [$src];", Float32Regs>;
defm INT_PTX_LDG_GLOBAL_f64
- : LDG_G<"f64 \t$result, [$src];", Float64Regs, int_nvvm_ldg_global_f>;
+ : LDG_G<"f64 \t$result, [$src];", Float64Regs>;
defm INT_PTX_LDG_GLOBAL_p32
- : LDG_G<"u32 \t$result, [$src];", Int32Regs, int_nvvm_ldg_global_p>;
+ : LDG_G<"u32 \t$result, [$src];", Int32Regs>;
defm INT_PTX_LDG_GLOBAL_p64
- : LDG_G<"u64 \t$result, [$src];", Int64Regs, int_nvvm_ldg_global_p>;
+ : LDG_G<"u64 \t$result, [$src];", Int64Regs>;
// vector
[(int_nvvm_compiler_error Int64Regs:$a)]>;
+// isspacep
+
+def ISSPACEP_CONST_32
+ : NVPTXInst<(outs Int1Regs:$d), (ins Int32Regs:$a),
+ "isspacep.const \t$d, $a;",
+ [(set Int1Regs:$d, (int_nvvm_isspacep_const Int32Regs:$a))]>,
+ Requires<[hasPTX31]>;
+def ISSPACEP_CONST_64
+ : NVPTXInst<(outs Int1Regs:$d), (ins Int64Regs:$a),
+ "isspacep.const \t$d, $a;",
+ [(set Int1Regs:$d, (int_nvvm_isspacep_const Int64Regs:$a))]>,
+ Requires<[hasPTX31]>;
+def ISSPACEP_GLOBAL_32
+ : NVPTXInst<(outs Int1Regs:$d), (ins Int32Regs:$a),
+ "isspacep.global \t$d, $a;",
+ [(set Int1Regs:$d, (int_nvvm_isspacep_global Int32Regs:$a))]>;
+def ISSPACEP_GLOBAL_64
+ : NVPTXInst<(outs Int1Regs:$d), (ins Int64Regs:$a),
+ "isspacep.global \t$d, $a;",
+ [(set Int1Regs:$d, (int_nvvm_isspacep_global Int64Regs:$a))]>;
+def ISSPACEP_LOCAL_32
+ : NVPTXInst<(outs Int1Regs:$d), (ins Int32Regs:$a),
+ "isspacep.local \t$d, $a;",
+ [(set Int1Regs:$d, (int_nvvm_isspacep_local Int32Regs:$a))]>;
+def ISSPACEP_LOCAL_64
+ : NVPTXInst<(outs Int1Regs:$d), (ins Int64Regs:$a),
+ "isspacep.local \t$d, $a;",
+ [(set Int1Regs:$d, (int_nvvm_isspacep_local Int64Regs:$a))]>;
+def ISSPACEP_SHARED_32
+ : NVPTXInst<(outs Int1Regs:$d), (ins Int32Regs:$a),
+ "isspacep.shared \t$d, $a;",
+ [(set Int1Regs:$d, (int_nvvm_isspacep_shared Int32Regs:$a))]>;
+def ISSPACEP_SHARED_64
+ : NVPTXInst<(outs Int1Regs:$d), (ins Int64Regs:$a),
+ "isspacep.shared \t$d, $a;",
+ [(set Int1Regs:$d, (int_nvvm_isspacep_shared Int64Regs:$a))]>;
+
+
// Special register reads
def MOV_SPECIAL : NVPTXInst<(outs Int32Regs:$d),
(ins SpecialRegs:$r),
def : Pat<(int_nvvm_read_ptx_sreg_envreg31), (MOV_SPECIAL ENVREG31)>;
+// rotate builtin support
+
+def ROTATE_B32_HW_IMM
+ : NVPTXInst<(outs Int32Regs:$dst),
+ (ins Int32Regs:$src, i32imm:$amt),
+ "shf.l.wrap.b32 \t$dst, $src, $src, $amt;",
+ [(set Int32Regs:$dst,
+ (int_nvvm_rotate_b32 Int32Regs:$src, (i32 imm:$amt)))]>,
+ Requires<[hasHWROT32]> ;
+
+def ROTATE_B32_HW_REG
+ : NVPTXInst<(outs Int32Regs:$dst),
+ (ins Int32Regs:$src, Int32Regs:$amt),
+ "shf.l.wrap.b32 \t$dst, $src, $src, $amt;",
+ [(set Int32Regs:$dst,
+ (int_nvvm_rotate_b32 Int32Regs:$src, Int32Regs:$amt))]>,
+ Requires<[hasHWROT32]> ;
+
+def : Pat<(int_nvvm_rotate_b32 Int32Regs:$src, (i32 imm:$amt)),
+ (ROT32imm_sw Int32Regs:$src, imm:$amt, (SUB_FRM_32 node:$amt))>,
+ Requires<[noHWROT32]> ;
+
+def : Pat<(int_nvvm_rotate_b32 Int32Regs:$src, Int32Regs:$amt),
+ (ROTL32reg_sw Int32Regs:$src, Int32Regs:$amt)>,
+ Requires<[noHWROT32]> ;
+
+def GET_LO_INT64
+ : NVPTXInst<(outs Int32Regs:$dst), (ins Int64Regs:$src),
+ !strconcat("{{\n\t",
+ !strconcat(".reg .b32 %dummy;\n\t",
+ !strconcat("mov.b64 \t{$dst,%dummy}, $src;\n\t",
+ !strconcat("}}", "")))),
+ []> ;
+
+def GET_HI_INT64
+ : NVPTXInst<(outs Int32Regs:$dst), (ins Int64Regs:$src),
+ !strconcat("{{\n\t",
+ !strconcat(".reg .b32 %dummy;\n\t",
+ !strconcat("mov.b64 \t{%dummy,$dst}, $src;\n\t",
+ !strconcat("}}", "")))),
+ []> ;
+
+def PACK_TWO_INT32
+ : NVPTXInst<(outs Int64Regs:$dst), (ins Int32Regs:$lo, Int32Regs:$hi),
+ "mov.b64 \t$dst, {{$lo, $hi}};", []> ;
+
+def : Pat<(int_nvvm_swap_lo_hi_b64 Int64Regs:$src),
+ (PACK_TWO_INT32 (GET_HI_INT64 Int64Regs:$src),
+ (GET_LO_INT64 Int64Regs:$src))> ;
+
+// funnel shift, requires >= sm_32
+def SHF_L_WRAP_B32_IMM
+ : NVPTXInst<(outs Int32Regs:$dst),
+ (ins Int32Regs:$lo, Int32Regs:$hi, i32imm:$amt),
+ "shf.l.wrap.b32 \t$dst, $lo, $hi, $amt;",[]>,
+ Requires<[hasHWROT32]>;
+
+def SHF_L_WRAP_B32_REG
+ : NVPTXInst<(outs Int32Regs:$dst),
+ (ins Int32Regs:$lo, Int32Regs:$hi, Int32Regs:$amt),
+ "shf.l.wrap.b32 \t$dst, $lo, $hi, $amt;",[]>,
+ Requires<[hasHWROT32]>;
+
+def SHF_R_WRAP_B32_IMM
+ : NVPTXInst<(outs Int32Regs:$dst),
+ (ins Int32Regs:$lo, Int32Regs:$hi, i32imm:$amt),
+ "shf.r.wrap.b32 \t$dst, $lo, $hi, $amt;",[]>,
+ Requires<[hasHWROT32]>;
+
+def SHF_R_WRAP_B32_REG
+ : NVPTXInst<(outs Int32Regs:$dst),
+ (ins Int32Regs:$lo, Int32Regs:$hi, Int32Regs:$amt),
+ "shf.r.wrap.b32 \t$dst, $lo, $hi, $amt;",[]>,
+ Requires<[hasHWROT32]>;
+
+// HW version of rotate 64
+def : Pat<(int_nvvm_rotate_b64 Int64Regs:$src, (i32 imm:$amt)),
+ (PACK_TWO_INT32
+ (SHF_L_WRAP_B32_IMM (GET_HI_INT64 Int64Regs:$src),
+ (GET_LO_INT64 Int64Regs:$src), imm:$amt),
+ (SHF_L_WRAP_B32_IMM (GET_LO_INT64 Int64Regs:$src),
+ (GET_HI_INT64 Int64Regs:$src), imm:$amt))>,
+ Requires<[hasHWROT32]>;
+
+def : Pat<(int_nvvm_rotate_b64 Int64Regs:$src, Int32Regs:$amt),
+ (PACK_TWO_INT32
+ (SHF_L_WRAP_B32_REG (GET_HI_INT64 Int64Regs:$src),
+ (GET_LO_INT64 Int64Regs:$src), Int32Regs:$amt),
+ (SHF_L_WRAP_B32_REG (GET_LO_INT64 Int64Regs:$src),
+ (GET_HI_INT64 Int64Regs:$src), Int32Regs:$amt))>,
+ Requires<[hasHWROT32]>;
+
+
+def : Pat<(int_nvvm_rotate_right_b64 Int64Regs:$src, (i32 imm:$amt)),
+ (PACK_TWO_INT32
+ (SHF_R_WRAP_B32_IMM (GET_LO_INT64 Int64Regs:$src),
+ (GET_HI_INT64 Int64Regs:$src), imm:$amt),
+ (SHF_R_WRAP_B32_IMM (GET_HI_INT64 Int64Regs:$src),
+ (GET_LO_INT64 Int64Regs:$src), imm:$amt))>,
+ Requires<[hasHWROT32]>;
+
+def : Pat<(int_nvvm_rotate_right_b64 Int64Regs:$src, Int32Regs:$amt),
+ (PACK_TWO_INT32
+ (SHF_R_WRAP_B32_REG (GET_LO_INT64 Int64Regs:$src),
+ (GET_HI_INT64 Int64Regs:$src), Int32Regs:$amt),
+ (SHF_R_WRAP_B32_REG (GET_HI_INT64 Int64Regs:$src),
+ (GET_LO_INT64 Int64Regs:$src), Int32Regs:$amt))>,
+ Requires<[hasHWROT32]>;
+
+// SW version of rotate 64
+def : Pat<(int_nvvm_rotate_b64 Int64Regs:$src, (i32 imm:$amt)),
+ (ROT64imm_sw Int64Regs:$src, imm:$amt, (SUB_FRM_32 node:$amt))>,
+ Requires<[noHWROT32]>;
+def : Pat<(int_nvvm_rotate_b64 Int64Regs:$src, Int32Regs:$amt),
+ (ROTL64reg_sw Int64Regs:$src, Int32Regs:$amt)>,
+ Requires<[noHWROT32]>;
+def : Pat<(int_nvvm_rotate_right_b64 Int64Regs:$src, (i32 imm:$amt)),
+ (ROT64imm_sw Int64Regs:$src, (SUB_FRM_64 node:$amt), imm:$amt)>,
+ Requires<[noHWROT32]>;
+def : Pat<(int_nvvm_rotate_right_b64 Int64Regs:$src, Int32Regs:$amt),
+ (ROTR64reg_sw Int64Regs:$src, Int32Regs:$amt)>,
+ Requires<[noHWROT32]>;
+
+
//-----------------------------------
// Texture Intrinsics
//-----------------------------------
// NOTE: For Fermi support, any new texture/surface/sampler intrinsics must be
// also defined in NVPTXReplaceImageHandles.cpp
-
+// texmode_independent
+let IsTex = 1, IsTexModeUnified = 0 in {
// Texture fetch instructions using handles
-def TEX_1D_F32_I32
+def TEX_1D_F32_S32
: NVPTXInst<(outs Float32Regs:$r, Float32Regs:$g,
Float32Regs:$b, Float32Regs:$a),
(ins Int64Regs:$t, Int64Regs:$s, Int32Regs:$x),
"tex.grad.1d.v4.f32.f32\t\\{$r, $g, $b, $a\\}, "
"[$t, $s, \\{$x\\}], \\{$gradx\\}, \\{$grady\\};",
[]>;
-def TEX_1D_I32_I32
+def TEX_1D_S32_S32
: NVPTXInst<(outs Int32Regs:$r, Int32Regs:$g,
Int32Regs:$b, Int32Regs:$a),
(ins Int64Regs:$t, Int64Regs:$s, Int32Regs:$x),
"tex.1d.v4.s32.s32\t\\{$r, $g, $b, $a\\}, [$t, $s, \\{$x\\}];",
[]>;
-def TEX_1D_I32_F32
+def TEX_1D_S32_F32
: NVPTXInst<(outs Int32Regs:$r, Int32Regs:$g,
Int32Regs:$b, Int32Regs:$a),
(ins Int64Regs:$t, Int64Regs:$s, Float32Regs:$x),
"tex.1d.v4.s32.f32\t\\{$r, $g, $b, $a\\}, [$t, $s, \\{$x\\}];",
[]>;
-def TEX_1D_I32_F32_LEVEL
+def TEX_1D_S32_F32_LEVEL
: NVPTXInst<(outs Int32Regs:$r, Int32Regs:$g,
Int32Regs:$b, Int32Regs:$a),
(ins Int64Regs:$t, Int64Regs:$s, Float32Regs:$x,
"tex.level.1d.v4.s32.f32\t\\{$r, $g, $b, $a\\}, "
"[$t, $s, \\{$x\\}], $lod;",
[]>;
-def TEX_1D_I32_F32_GRAD
+def TEX_1D_S32_F32_GRAD
: NVPTXInst<(outs Int32Regs:$r, Int32Regs:$g,
Int32Regs:$b, Int32Regs:$a),
(ins Int64Regs:$t, Int64Regs:$s, Float32Regs:$x,
"tex.grad.1d.v4.s32.f32\t\\{$r, $g, $b, $a\\}, "
"[$t, $s, \\{$x\\}], \\{$gradx\\}, \\{$grady\\};",
[]>;
+def TEX_1D_U32_S32
+ : NVPTXInst<(outs Int32Regs:$r, Int32Regs:$g,
+ Int32Regs:$b, Int32Regs:$a),
+ (ins Int64Regs:$t, Int64Regs:$s, Int32Regs:$x),
+ "tex.1d.v4.u32.s32\t\\{$r, $g, $b, $a\\}, [$t, $s, \\{$x\\}];",
+ []>;
+def TEX_1D_U32_F32
+ : NVPTXInst<(outs Int32Regs:$r, Int32Regs:$g,
+ Int32Regs:$b, Int32Regs:$a),
+ (ins Int64Regs:$t, Int64Regs:$s, Float32Regs:$x),
+ "tex.1d.v4.u32.f32\t\\{$r, $g, $b, $a\\}, [$t, $s, \\{$x\\}];",
+ []>;
+def TEX_1D_U32_F32_LEVEL
+ : NVPTXInst<(outs Int32Regs:$r, Int32Regs:$g,
+ Int32Regs:$b, Int32Regs:$a),
+ (ins Int64Regs:$t, Int64Regs:$s, Float32Regs:$x,
+ Float32Regs:$lod),
+ "tex.level.1d.v4.u32.f32\t\\{$r, $g, $b, $a\\}, "
+ "[$t, $s, \\{$x\\}], $lod;",
+ []>;
+def TEX_1D_U32_F32_GRAD
+ : NVPTXInst<(outs Int32Regs:$r, Int32Regs:$g,
+ Int32Regs:$b, Int32Regs:$a),
+ (ins Int64Regs:$t, Int64Regs:$s, Float32Regs:$x,
+ Float32Regs:$gradx, Float32Regs:$grady),
+ "tex.grad.1d.v4.u32.f32\t\\{$r, $g, $b, $a\\}, "
+ "[$t, $s, \\{$x\\}], \\{$gradx\\}, \\{$grady\\};",
+ []>;
-def TEX_1D_ARRAY_F32_I32
+def TEX_1D_ARRAY_F32_S32
: NVPTXInst<(outs Float32Regs:$r, Float32Regs:$g,
Float32Regs:$b, Float32Regs:$a),
(ins Int64Regs:$t, Int64Regs:$s, Int32Regs:$l, Int32Regs:$x),
"tex.grad.a1d.v4.f32.f32\t\\{$r, $g, $b, $a\\}, "
"[$t, $s, \\{$l, $x\\}], \\{$gradx\\}, \\{$grady\\};",
[]>;
-def TEX_1D_ARRAY_I32_I32
+def TEX_1D_ARRAY_S32_S32
: NVPTXInst<(outs Int32Regs:$r, Int32Regs:$g,
Int32Regs:$b, Int32Regs:$a),
(ins Int64Regs:$t, Int64Regs:$s, Int32Regs:$l, Int32Regs:$x),
"tex.a1d.v4.s32.s32\t\\{$r, $g, $b, $a\\}, "
"[$t, $s, \\{$l, $x\\}];",
[]>;
-def TEX_1D_ARRAY_I32_F32
+def TEX_1D_ARRAY_S32_F32
: NVPTXInst<(outs Int32Regs:$r, Int32Regs:$g,
Int32Regs:$b, Int32Regs:$a),
(ins Int64Regs:$t, Int64Regs:$s, Int32Regs:$l, Float32Regs:$x),
"tex.a1d.v4.s32.f32\t\\{$r, $g, $b, $a\\}, "
"[$t, $s, \\{$l, $x\\}];",
[]>;
-def TEX_1D_ARRAY_I32_F32_LEVEL
+def TEX_1D_ARRAY_S32_F32_LEVEL
: NVPTXInst<(outs Int32Regs:$r, Int32Regs:$g,
Int32Regs:$b, Int32Regs:$a),
(ins Int64Regs:$t, Int64Regs:$s, Int32Regs:$l, Float32Regs:$x,
"tex.level.a1d.v4.s32.f32\t\\{$r, $g, $b, $a\\}, "
"[$t, $s, \\{$l, $x\\}], $lod;",
[]>;
-def TEX_1D_ARRAY_I32_F32_GRAD
+def TEX_1D_ARRAY_S32_F32_GRAD
: NVPTXInst<(outs Int32Regs:$r, Int32Regs:$g,
Int32Regs:$b, Int32Regs:$a),
(ins Int64Regs:$t, Int64Regs:$s, Int32Regs:$l, Float32Regs:$x,
"tex.grad.a1d.v4.s32.f32\t\\{$r, $g, $b, $a\\}, "
"[$t, $s, \\{$l, $x\\}], \\{$gradx\\}, \\{$grady\\};",
[]>;
+def TEX_1D_ARRAY_U32_S32
+ : NVPTXInst<(outs Int32Regs:$r, Int32Regs:$g,
+ Int32Regs:$b, Int32Regs:$a),
+ (ins Int64Regs:$t, Int64Regs:$s, Int32Regs:$l, Int32Regs:$x),
+ "tex.a1d.v4.u32.s32\t\\{$r, $g, $b, $a\\}, "
+ "[$t, $s, \\{$l, $x\\}];",
+ []>;
+def TEX_1D_ARRAY_U32_F32
+ : NVPTXInst<(outs Int32Regs:$r, Int32Regs:$g,
+ Int32Regs:$b, Int32Regs:$a),
+ (ins Int64Regs:$t, Int64Regs:$s, Int32Regs:$l, Float32Regs:$x),
+ "tex.a1d.v4.u32.f32\t\\{$r, $g, $b, $a\\}, "
+ "[$t, $s, \\{$l, $x\\}];",
+ []>;
+def TEX_1D_ARRAY_U32_F32_LEVEL
+ : NVPTXInst<(outs Int32Regs:$r, Int32Regs:$g,
+ Int32Regs:$b, Int32Regs:$a),
+ (ins Int64Regs:$t, Int64Regs:$s, Int32Regs:$l, Float32Regs:$x,
+ Float32Regs:$lod),
+ "tex.level.a1d.v4.u32.f32\t\\{$r, $g, $b, $a\\}, "
+ "[$t, $s, \\{$l, $x\\}], $lod;",
+ []>;
+def TEX_1D_ARRAY_U32_F32_GRAD
+ : NVPTXInst<(outs Int32Regs:$r, Int32Regs:$g,
+ Int32Regs:$b, Int32Regs:$a),
+ (ins Int64Regs:$t, Int64Regs:$s, Int32Regs:$l, Float32Regs:$x,
+ Float32Regs:$gradx, Float32Regs:$grady),
+ "tex.grad.a1d.v4.u32.f32\t\\{$r, $g, $b, $a\\}, "
+ "[$t, $s, \\{$l, $x\\}], \\{$gradx\\}, \\{$grady\\};",
+ []>;
-def TEX_2D_F32_I32
+def TEX_2D_F32_S32
: NVPTXInst<(outs Float32Regs:$r, Float32Regs:$g,
Float32Regs:$b, Float32Regs:$a),
(ins Int64Regs:$t, Int64Regs:$s, Int32Regs:$x, Int32Regs:$y),
"[$t, $s, \\{$x, $y\\}], \\{$gradx0, $gradx1\\}, "
"\\{$grady0, $grady1\\};",
[]>;
-def TEX_2D_I32_I32
+def TEX_2D_S32_S32
: NVPTXInst<(outs Int32Regs:$r, Int32Regs:$g,
Int32Regs:$b, Int32Regs:$a),
(ins Int64Regs:$t, Int64Regs:$s, Int32Regs:$x, Int32Regs:$y),
"tex.2d.v4.s32.s32\t\\{$r, $g, $b, $a\\}, "
"[$t, $s, \\{$x, $y\\}];",
[]>;
-def TEX_2D_I32_F32
+def TEX_2D_S32_F32
: NVPTXInst<(outs Int32Regs:$r, Int32Regs:$g,
Int32Regs:$b, Int32Regs:$a),
(ins Int64Regs:$t, Int64Regs:$s, Float32Regs:$x, Float32Regs:$y),
"tex.2d.v4.s32.f32\t\\{$r, $g, $b, $a\\}, "
"[$t, $s, \\{$x, $y\\}];",
[]>;
-def TEX_2D_I32_F32_LEVEL
+def TEX_2D_S32_F32_LEVEL
: NVPTXInst<(outs Int32Regs:$r, Int32Regs:$g,
Int32Regs:$b, Int32Regs:$a),
(ins Int64Regs:$t, Int64Regs:$s, Float32Regs:$x, Float32Regs:$y,
"tex.level.2d.v4.s32.f32\t\\{$r, $g, $b, $a\\}, "
"[$t, $s, \\{$x, $y\\}], $lod;",
[]>;
-def TEX_2D_I32_F32_GRAD
+def TEX_2D_S32_F32_GRAD
: NVPTXInst<(outs Int32Regs:$r, Int32Regs:$g,
Int32Regs:$b, Int32Regs:$a),
(ins Int64Regs:$t, Int64Regs:$s, Float32Regs:$x, Float32Regs:$y,
"[$t, $s, \\{$x, $y\\}], \\{$gradx0, $gradx1\\}, "
"\\{$grady0, $grady1\\};",
[]>;
+def TEX_2D_U32_S32
+ : NVPTXInst<(outs Int32Regs:$r, Int32Regs:$g,
+ Int32Regs:$b, Int32Regs:$a),
+ (ins Int64Regs:$t, Int64Regs:$s, Int32Regs:$x, Int32Regs:$y),
+ "tex.2d.v4.u32.s32\t\\{$r, $g, $b, $a\\}, "
+ "[$t, $s, \\{$x, $y\\}];",
+ []>;
+def TEX_2D_U32_F32
+ : NVPTXInst<(outs Int32Regs:$r, Int32Regs:$g,
+ Int32Regs:$b, Int32Regs:$a),
+ (ins Int64Regs:$t, Int64Regs:$s, Float32Regs:$x, Float32Regs:$y),
+ "tex.2d.v4.u32.f32\t\\{$r, $g, $b, $a\\}, "
+ "[$t, $s, \\{$x, $y\\}];",
+ []>;
+def TEX_2D_U32_F32_LEVEL
+ : NVPTXInst<(outs Int32Regs:$r, Int32Regs:$g,
+ Int32Regs:$b, Int32Regs:$a),
+ (ins Int64Regs:$t, Int64Regs:$s, Float32Regs:$x, Float32Regs:$y,
+ Float32Regs:$lod),
+ "tex.level.2d.v4.u32.f32\t\\{$r, $g, $b, $a\\}, "
+ "[$t, $s, \\{$x, $y\\}], $lod;",
+ []>;
+def TEX_2D_U32_F32_GRAD
+ : NVPTXInst<(outs Int32Regs:$r, Int32Regs:$g,
+ Int32Regs:$b, Int32Regs:$a),
+ (ins Int64Regs:$t, Int64Regs:$s, Float32Regs:$x, Float32Regs:$y,
+ Float32Regs:$gradx0, Float32Regs:$gradx1,
+ Float32Regs:$grady0, Float32Regs:$grady1),
+ "tex.grad.2d.v4.u32.f32\t\\{$r, $g, $b, $a\\}, "
+ "[$t, $s, \\{$x, $y\\}], \\{$gradx0, $gradx1\\}, "
+ "\\{$grady0, $grady1\\};",
+ []>;
-def TEX_2D_ARRAY_F32_I32
+def TEX_2D_ARRAY_F32_S32
: NVPTXInst<(outs Float32Regs:$r, Float32Regs:$g,
Float32Regs:$b, Float32Regs:$a),
(ins Int64Regs:$t, Int64Regs:$s, Int32Regs:$l, Int32Regs:$x,
"[$t, $s, \\{$l, $x, $y, $y\\}], \\{$gradx0, $gradx1\\}, "
"\\{$grady0, $grady1\\};",
[]>;
-def TEX_2D_ARRAY_I32_I32
+def TEX_2D_ARRAY_S32_S32
: NVPTXInst<(outs Int32Regs:$r, Int32Regs:$g,
Int32Regs:$b, Int32Regs:$a),
(ins Int64Regs:$t, Int64Regs:$s, Int32Regs:$l, Int32Regs:$x,
"tex.a2d.v4.s32.s32\t\\{$r, $g, $b, $a\\}, "
"[$t, $s, \\{$l, $x, $y, $y\\}];",
[]>;
-def TEX_2D_ARRAY_I32_F32
+def TEX_2D_ARRAY_S32_F32
: NVPTXInst<(outs Int32Regs:$r, Int32Regs:$g,
Int32Regs:$b, Int32Regs:$a),
(ins Int64Regs:$t, Int64Regs:$s, Int32Regs:$l, Float32Regs:$x,
"tex.a2d.v4.s32.f32\t\\{$r, $g, $b, $a\\}, "
"[$t, $s, \\{$l, $x, $y, $y\\}];",
[]>;
-def TEX_2D_ARRAY_I32_F32_LEVEL
+def TEX_2D_ARRAY_S32_F32_LEVEL
: NVPTXInst<(outs Int32Regs:$r, Int32Regs:$g,
Int32Regs:$b, Int32Regs:$a),
(ins Int64Regs:$t, Int64Regs:$s, Int32Regs:$l, Float32Regs:$x,
"tex.level.a2d.v4.s32.f32\t\\{$r, $g, $b, $a\\}, "
"[$t, $s, \\{$l, $x, $y, $y\\}], $lod;",
[]>;
-def TEX_2D_ARRAY_I32_F32_GRAD
+def TEX_2D_ARRAY_S32_F32_GRAD
: NVPTXInst<(outs Int32Regs:$r, Int32Regs:$g,
Int32Regs:$b, Int32Regs:$a),
(ins Int64Regs:$t, Int64Regs:$s, Int32Regs:$l, Float32Regs:$x,
"[$t, $s, \\{$l, $x, $y, $y\\}], \\{$gradx0, $gradx1\\}, "
"\\{$grady0, $grady1\\};",
[]>;
+def TEX_2D_ARRAY_U32_S32
+ : NVPTXInst<(outs Int32Regs:$r, Int32Regs:$g,
+ Int32Regs:$b, Int32Regs:$a),
+ (ins Int64Regs:$t, Int64Regs:$s, Int32Regs:$l, Int32Regs:$x,
+ Int32Regs:$y),
+ "tex.a2d.v4.u32.s32\t\\{$r, $g, $b, $a\\}, "
+ "[$t, $s, \\{$l, $x, $y, $y\\}];",
+ []>;
+def TEX_2D_ARRAY_U32_F32
+ : NVPTXInst<(outs Int32Regs:$r, Int32Regs:$g,
+ Int32Regs:$b, Int32Regs:$a),
+ (ins Int64Regs:$t, Int64Regs:$s, Int32Regs:$l, Float32Regs:$x,
+ Float32Regs:$y),
+ "tex.a2d.v4.u32.f32\t\\{$r, $g, $b, $a\\}, "
+ "[$t, $s, \\{$l, $x, $y, $y\\}];",
+ []>;
+def TEX_2D_ARRAY_U32_F32_LEVEL
+ : NVPTXInst<(outs Int32Regs:$r, Int32Regs:$g,
+ Int32Regs:$b, Int32Regs:$a),
+ (ins Int64Regs:$t, Int64Regs:$s, Int32Regs:$l, Float32Regs:$x,
+ Float32Regs:$y, Float32Regs:$lod),
+ "tex.level.a2d.v4.u32.f32\t\\{$r, $g, $b, $a\\}, "
+ "[$t, $s, \\{$l, $x, $y, $y\\}], $lod;",
+ []>;
+def TEX_2D_ARRAY_U32_F32_GRAD
+ : NVPTXInst<(outs Int32Regs:$r, Int32Regs:$g,
+ Int32Regs:$b, Int32Regs:$a),
+ (ins Int64Regs:$t, Int64Regs:$s, Int32Regs:$l, Float32Regs:$x,
+ Float32Regs:$y,
+ Float32Regs:$gradx0, Float32Regs:$gradx1,
+ Float32Regs:$grady0, Float32Regs:$grady1),
+ "tex.grad.a2d.v4.u32.f32\t\\{$r, $g, $b, $a\\}, "
+ "[$t, $s, \\{$l, $x, $y, $y\\}], \\{$gradx0, $gradx1\\}, "
+ "\\{$grady0, $grady1\\};",
+ []>;
-def TEX_3D_F32_I32
+def TEX_3D_F32_S32
: NVPTXInst<(outs Float32Regs:$r, Float32Regs:$g,
Float32Regs:$b, Float32Regs:$a),
(ins Int64Regs:$t, Int64Regs:$s, Int32Regs:$x, Int32Regs:$y,
"\\{$gradx0, $gradx1, $gradx2, $gradx2\\}, "
"\\{$grady0, $grady1, $grady2, $grady2\\};",
[]>;
-def TEX_3D_I32_I32
+def TEX_3D_S32_S32
: NVPTXInst<(outs Int32Regs:$r, Int32Regs:$g,
Int32Regs:$b, Int32Regs:$a),
(ins Int64Regs:$t, Int64Regs:$s, Int32Regs:$x, Int32Regs:$y,
"tex.3d.v4.s32.s32\t\\{$r, $g, $b, $a\\}, "
"[$t, $s, \\{$x, $y, $z, $z\\}];",
[]>;
-def TEX_3D_I32_F32
+def TEX_3D_S32_F32
: NVPTXInst<(outs Int32Regs:$r, Int32Regs:$g,
Int32Regs:$b, Int32Regs:$a),
(ins Int64Regs:$t, Int64Regs:$s, Float32Regs:$x, Float32Regs:$y,
"tex.3d.v4.s32.f32\t\\{$r, $g, $b, $a\\}, "
"[$t, $s, \\{$x, $y, $z, $z\\}];",
[]>;
-def TEX_3D_I32_F32_LEVEL
+def TEX_3D_S32_F32_LEVEL
: NVPTXInst<(outs Int32Regs:$r, Int32Regs:$g,
Int32Regs:$b, Int32Regs:$a),
(ins Int64Regs:$t, Int64Regs:$s, Float32Regs:$x, Float32Regs:$y,
"tex.level.3d.v4.s32.f32\t\\{$r, $g, $b, $a\\}, "
"[$t, $s, \\{$x, $y, $z, $z\\}], $lod;",
[]>;
-def TEX_3D_I32_F32_GRAD
+def TEX_3D_S32_F32_GRAD
: NVPTXInst<(outs Int32Regs:$r, Int32Regs:$g,
Int32Regs:$b, Int32Regs:$a),
(ins Int64Regs:$t, Int64Regs:$s, Float32Regs:$x, Float32Regs:$y,
"\\{$gradx0, $gradx1, $gradx2, $gradx2\\}, "
"\\{$grady0, $grady1, $grady2, $grady2\\};",
[]>;
-
-
-// Surface load instructions
-def SULD_1D_I8_TRAP
- : NVPTXInst<(outs Int16Regs:$r),
- (ins Int64Regs:$s, Int32Regs:$x),
- "suld.b.1d.b8.trap \\{$r\\}, [$s, \\{$x\\}];",
+def TEX_3D_U32_S32
+ : NVPTXInst<(outs Int32Regs:$r, Int32Regs:$g,
+ Int32Regs:$b, Int32Regs:$a),
+ (ins Int64Regs:$t, Int64Regs:$s, Int32Regs:$x, Int32Regs:$y,
+ Int32Regs:$z),
+ "tex.3d.v4.u32.s32\t\\{$r, $g, $b, $a\\}, "
+ "[$t, $s, \\{$x, $y, $z, $z\\}];",
[]>;
-def SULD_1D_I16_TRAP
- : NVPTXInst<(outs Int16Regs:$r),
- (ins Int64Regs:$s, Int32Regs:$x),
- "suld.b.1d.b16.trap \\{$r\\}, [$s, \\{$x\\}];",
+def TEX_3D_U32_F32
+ : NVPTXInst<(outs Int32Regs:$r, Int32Regs:$g,
+ Int32Regs:$b, Int32Regs:$a),
+ (ins Int64Regs:$t, Int64Regs:$s, Float32Regs:$x, Float32Regs:$y,
+ Float32Regs:$z),
+ "tex.3d.v4.u32.f32\t\\{$r, $g, $b, $a\\}, "
+ "[$t, $s, \\{$x, $y, $z, $z\\}];",
[]>;
-def SULD_1D_I32_TRAP
- : NVPTXInst<(outs Int32Regs:$r),
- (ins Int64Regs:$s, Int32Regs:$x),
- "suld.b.1d.b32.trap \\{$r\\}, [$s, \\{$x\\}];",
+def TEX_3D_U32_F32_LEVEL
+ : NVPTXInst<(outs Int32Regs:$r, Int32Regs:$g,
+ Int32Regs:$b, Int32Regs:$a),
+ (ins Int64Regs:$t, Int64Regs:$s, Float32Regs:$x, Float32Regs:$y,
+ Float32Regs:$z, Float32Regs:$lod),
+ "tex.level.3d.v4.u32.f32\t\\{$r, $g, $b, $a\\}, "
+ "[$t, $s, \\{$x, $y, $z, $z\\}], $lod;",
[]>;
-def SULD_1D_V2I8_TRAP
- : NVPTXInst<(outs Int16Regs:$r, Int16Regs:$g),
- (ins Int64Regs:$s, Int32Regs:$x),
- "suld.b.1d.v2.b8.trap \\{$r, $g\\}, [$s, \\{$x\\}];",
+def TEX_3D_U32_F32_GRAD
+ : NVPTXInst<(outs Int32Regs:$r, Int32Regs:$g,
+ Int32Regs:$b, Int32Regs:$a),
+ (ins Int64Regs:$t, Int64Regs:$s, Float32Regs:$x, Float32Regs:$y,
+ Float32Regs:$z,
+ Float32Regs:$gradx0, Float32Regs:$gradx1,
+ Float32Regs:$gradx2, Float32Regs:$grady0,
+ Float32Regs:$grady1, Float32Regs:$grady2),
+ "tex.grad.3d.v4.u32.f32\t\\{$r, $g, $b, $a\\}, "
+ "[$t, $s, \\{$x, $y, $z, $z\\}], "
+ "\\{$gradx0, $gradx1, $gradx2, $gradx2\\}, "
+ "\\{$grady0, $grady1, $grady2, $grady2\\};",
[]>;
-def SULD_1D_V2I16_TRAP
- : NVPTXInst<(outs Int16Regs:$r, Int16Regs:$g),
- (ins Int64Regs:$s, Int32Regs:$x),
- "suld.b.1d.v2.b16.trap \\{$r, $g\\}, [$s, \\{$x\\}];",
+
+def TEX_CUBE_F32_F32
+ : NVPTXInst<(outs Float32Regs:$r, Float32Regs:$g,
+ Float32Regs:$b, Float32Regs:$a),
+ (ins Int64Regs:$t, Int64Regs:$s,
+ Float32Regs:$x, Float32Regs:$y, Float32Regs:$z),
+ "tex.cube.v4.f32.f32\t\\{$r, $g, $b, $a\\}, "
+ "[$t, $s, \\{$x, $y, $z, $z\\}];",
[]>;
-def SULD_1D_V2I32_TRAP
- : NVPTXInst<(outs Int32Regs:$r, Int32Regs:$g),
- (ins Int64Regs:$s, Int32Regs:$x),
- "suld.b.1d.v2.b32.trap \\{$r, $g\\}, [$s, \\{$x\\}];",
+def TEX_CUBE_F32_F32_LEVEL
+ : NVPTXInst<(outs Float32Regs:$r, Float32Regs:$g,
+ Float32Regs:$b, Float32Regs:$a),
+ (ins Int64Regs:$t, Int64Regs:$s,
+ Float32Regs:$x, Float32Regs:$y, Float32Regs:$z,
+ Float32Regs:$lod),
+ "tex.level.cube.v4.f32.f32\t\\{$r, $g, $b, $a\\}, "
+ "[$t, $s, \\{$x, $y, $z, $z\\}], $lod;",
[]>;
-def SULD_1D_V4I8_TRAP
- : NVPTXInst<(outs Int16Regs:$r, Int16Regs:$g, Int16Regs:$b, Int16Regs:$a),
- (ins Int64Regs:$s, Int32Regs:$x),
- "suld.b.1d.v4.b8.trap \\{$r, $g, $b, $a\\}, [$s, \\{$x\\}];",
+def TEX_CUBE_S32_F32
+ : NVPTXInst<(outs Int32Regs:$r, Int32Regs:$g,
+ Int32Regs:$b, Int32Regs:$a),
+ (ins Int64Regs:$t, Int64Regs:$s,
+ Float32Regs:$x, Float32Regs:$y, Float32Regs:$z),
+ "tex.cube.v4.s32.f32\t\\{$r, $g, $b, $a\\}, "
+ "[$t, $s, \\{$x, $y, $z, $z\\}];",
[]>;
-def SULD_1D_V4I16_TRAP
- : NVPTXInst<(outs Int16Regs:$r, Int16Regs:$g, Int16Regs:$b, Int16Regs:$a),
- (ins Int64Regs:$s, Int32Regs:$x),
- "suld.b.1d.v4.b16.trap \\{$r, $g, $b, $a\\}, [$s, \\{$x\\}];",
+def TEX_CUBE_S32_F32_LEVEL
+ : NVPTXInst<(outs Int32Regs:$r, Int32Regs:$g,
+ Int32Regs:$b, Int32Regs:$a),
+ (ins Int64Regs:$t, Int64Regs:$s,
+ Float32Regs:$x, Float32Regs:$y, Float32Regs:$z,
+ Float32Regs:$lod),
+ "tex.level.cube.v4.s32.f32\t\\{$r, $g, $b, $a\\}, "
+ "[$t, $s, \\{$x, $y, $z, $z\\}], $lod;",
[]>;
-def SULD_1D_V4I32_TRAP
- : NVPTXInst<(outs Int32Regs:$r, Int32Regs:$g, Int32Regs:$b, Int32Regs:$a),
- (ins Int64Regs:$s, Int32Regs:$x),
- "suld.b.1d.v4.b32.trap \\{$r, $g, $b, $a\\}, [$s, \\{$x\\}];",
+def TEX_CUBE_U32_F32
+ : NVPTXInst<(outs Int32Regs:$r, Int32Regs:$g,
+ Int32Regs:$b, Int32Regs:$a),
+ (ins Int64Regs:$t, Int64Regs:$s,
+ Float32Regs:$x, Float32Regs:$y, Float32Regs:$z),
+ "tex.cube.v4.u32.f32\t\\{$r, $g, $b, $a\\}, "
+ "[$t, $s, \\{$x, $y, $z, $z\\}];",
+ []>;
+def TEX_CUBE_U32_F32_LEVEL
+ : NVPTXInst<(outs Int32Regs:$r, Int32Regs:$g,
+ Int32Regs:$b, Int32Regs:$a),
+ (ins Int64Regs:$t, Int64Regs:$s,
+ Float32Regs:$x, Float32Regs:$y, Float32Regs:$z,
+ Float32Regs:$lod),
+ "tex.level.cube.v4.u32.f32\t\\{$r, $g, $b, $a\\}, "
+ "[$t, $s, \\{$x, $y, $z, $z\\}], $lod;",
[]>;
-def SULD_1D_ARRAY_I8_TRAP
- : NVPTXInst<(outs Int16Regs:$r),
- (ins Int64Regs:$s, Int32Regs:$l, Int32Regs:$x),
- "suld.b.a1d.b8.trap \\{$r\\}, [$s, \\{$l, $x\\}];",
+def TEX_CUBE_ARRAY_F32_F32
+ : NVPTXInst<(outs Float32Regs:$r, Float32Regs:$g,
+ Float32Regs:$b, Float32Regs:$a),
+ (ins Int64Regs:$t, Int64Regs:$s, Int32Regs:$l,
+ Float32Regs:$x, Float32Regs:$y, Float32Regs:$z),
+ "tex.acube.v4.f32.f32\t\\{$r, $g, $b, $a\\}, "
+ "[$t, $s, \\{$l, $x, $y, $z\\}];",
[]>;
-def SULD_1D_ARRAY_I16_TRAP
- : NVPTXInst<(outs Int16Regs:$r),
- (ins Int64Regs:$s, Int32Regs:$l, Int32Regs:$x),
- "suld.b.a1d.b16.trap \\{$r\\}, [$s, \\{$l, $x\\}];",
+def TEX_CUBE_ARRAY_F32_F32_LEVEL
+ : NVPTXInst<(outs Float32Regs:$r, Float32Regs:$g,
+ Float32Regs:$b, Float32Regs:$a),
+ (ins Int64Regs:$t, Int64Regs:$s, Int32Regs:$l,
+ Float32Regs:$x, Float32Regs:$y, Float32Regs:$z,
+ Float32Regs:$lod),
+ "tex.level.acube.v4.f32.f32\t\\{$r, $g, $b, $a\\}, "
+ "[$t, $s, \\{$l, $x, $y, $z\\}], $lod;",
[]>;
-def SULD_1D_ARRAY_I32_TRAP
- : NVPTXInst<(outs Int32Regs:$r),
- (ins Int64Regs:$s, Int32Regs:$l, Int32Regs:$x),
- "suld.b.a1d.b32.trap \\{$r\\}, [$s, \\{$l, $x\\}];",
+def TEX_CUBE_ARRAY_S32_F32
+ : NVPTXInst<(outs Int32Regs:$r, Int32Regs:$g,
+ Int32Regs:$b, Int32Regs:$a),
+ (ins Int64Regs:$t, Int64Regs:$s, Int32Regs:$l,
+ Float32Regs:$x, Float32Regs:$y, Float32Regs:$z),
+ "tex.acube.v4.s32.f32\t\\{$r, $g, $b, $a\\}, "
+ "[$t, $s, \\{$l, $x, $y, $z\\}];",
[]>;
-def SULD_1D_ARRAY_V2I8_TRAP
- : NVPTXInst<(outs Int16Regs:$r, Int16Regs:$g),
- (ins Int64Regs:$s, Int32Regs:$l, Int32Regs:$x),
- "suld.b.a1d.v2.b8.trap \\{$r, $g\\}, [$s, \\{$l, $x\\}];",
+def TEX_CUBE_ARRAY_S32_F32_LEVEL
+ : NVPTXInst<(outs Int32Regs:$r, Int32Regs:$g,
+ Int32Regs:$b, Int32Regs:$a),
+ (ins Int64Regs:$t, Int64Regs:$s, Int32Regs:$l,
+ Float32Regs:$x, Float32Regs:$y, Float32Regs:$z,
+ Float32Regs:$lod),
+ "tex.level.acube.v4.s32.f32\t\\{$r, $g, $b, $a\\}, "
+ "[$t, $s, \\{$l, $x, $y, $z\\}], $lod;",
[]>;
-def SULD_1D_ARRAY_V2I16_TRAP
- : NVPTXInst<(outs Int16Regs:$r, Int16Regs:$g),
- (ins Int64Regs:$s, Int32Regs:$l, Int32Regs:$x),
- "suld.b.a1d.v2.b16.trap \\{$r, $g\\}, [$s, \\{$l, $x\\}];",
+def TEX_CUBE_ARRAY_U32_F32
+ : NVPTXInst<(outs Int32Regs:$r, Int32Regs:$g,
+ Int32Regs:$b, Int32Regs:$a),
+ (ins Int64Regs:$t, Int64Regs:$s, Int32Regs:$l,
+ Float32Regs:$x, Float32Regs:$y, Float32Regs:$z),
+ "tex.acube.v4.u32.f32\t\\{$r, $g, $b, $a\\}, "
+ "[$t, $s, \\{$l, $x, $y, $z\\}];",
[]>;
-def SULD_1D_ARRAY_V2I32_TRAP
- : NVPTXInst<(outs Int32Regs:$r, Int32Regs:$g),
- (ins Int64Regs:$s, Int32Regs:$l, Int32Regs:$x),
- "suld.b.a1d.v2.b32.trap \\{$r, $g\\}, [$s, \\{$l, $x\\}];",
+def TEX_CUBE_ARRAY_U32_F32_LEVEL
+ : NVPTXInst<(outs Int32Regs:$r, Int32Regs:$g,
+ Int32Regs:$b, Int32Regs:$a),
+ (ins Int64Regs:$t, Int64Regs:$s, Int32Regs:$l,
+ Float32Regs:$x, Float32Regs:$y, Float32Regs:$z,
+ Float32Regs:$lod),
+ "tex.level.acube.v4.u32.f32\t\\{$r, $g, $b, $a\\}, "
+ "[$t, $s, \\{$l, $x, $y, $z\\}], $lod;",
[]>;
-def SULD_1D_ARRAY_V4I8_TRAP
+
+def TLD4_R_2D_F32_F32
+ : NVPTXInst<(outs Float32Regs:$v0, Float32Regs:$v1,
+ Float32Regs:$v2, Float32Regs:$v3),
+ (ins Int64Regs:$t, Int64Regs:$s, Float32Regs:$x, Float32Regs:$y),
+ "tld4.r.2d.v4.f32.f32\t\\{$v0, $v1, $v2, $v3\\}, "
+ "[$t, $s, \\{$x, $y\\}];",
+ []>;
+def TLD4_G_2D_F32_F32
+ : NVPTXInst<(outs Float32Regs:$v0, Float32Regs:$v1,
+ Float32Regs:$v2, Float32Regs:$v3),
+ (ins Int64Regs:$t, Int64Regs:$s, Float32Regs:$x, Float32Regs:$y),
+ "tld4.g.2d.v4.f32.f32\t\\{$v0, $v1, $v2, $v3\\}, "
+ "[$t, $s, \\{$x, $y\\}];",
+ []>;
+def TLD4_B_2D_F32_F32
+ : NVPTXInst<(outs Float32Regs:$v0, Float32Regs:$v1,
+ Float32Regs:$v2, Float32Regs:$v3),
+ (ins Int64Regs:$t, Int64Regs:$s, Float32Regs:$x, Float32Regs:$y),
+ "tld4.b.2d.v4.f32.f32\t\\{$v0, $v1, $v2, $v3\\}, "
+ "[$t, $s, \\{$x, $y\\}];",
+ []>;
+def TLD4_A_2D_F32_F32
+ : NVPTXInst<(outs Float32Regs:$v0, Float32Regs:$v1,
+ Float32Regs:$v2, Float32Regs:$v3),
+ (ins Int64Regs:$t, Int64Regs:$s, Float32Regs:$x, Float32Regs:$y),
+ "tld4.a.2d.v4.f32.f32\t\\{$v0, $v1, $v2, $v3\\}, "
+ "[$t, $s, \\{$x, $y\\}];",
+ []>;
+def TLD4_R_2D_S32_F32
+ : NVPTXInst<(outs Int32Regs:$v0, Int32Regs:$v1,
+ Int32Regs:$v2, Int32Regs:$v3),
+ (ins Int64Regs:$t, Int64Regs:$s, Float32Regs:$x, Float32Regs:$y),
+ "tld4.r.2d.v4.s32.f32\t\\{$v0, $v1, $v2, $v3\\}, "
+ "[$t, $s, \\{$x, $y\\}];",
+ []>;
+def TLD4_G_2D_S32_F32
+ : NVPTXInst<(outs Int32Regs:$v0, Int32Regs:$v1,
+ Int32Regs:$v2, Int32Regs:$v3),
+ (ins Int64Regs:$t, Int64Regs:$s, Float32Regs:$x, Float32Regs:$y),
+ "tld4.g.2d.v4.s32.f32\t\\{$v0, $v1, $v2, $v3\\}, "
+ "[$t, $s, \\{$x, $y\\}];",
+ []>;
+def TLD4_B_2D_S32_F32
+ : NVPTXInst<(outs Int32Regs:$v0, Int32Regs:$v1,
+ Int32Regs:$v2, Int32Regs:$v3),
+ (ins Int64Regs:$t, Int64Regs:$s, Float32Regs:$x, Float32Regs:$y),
+ "tld4.b.2d.v4.s32.f32\t\\{$v0, $v1, $v2, $v3\\}, "
+ "[$t, $s, \\{$x, $y\\}];",
+ []>;
+def TLD4_A_2D_S32_F32
+ : NVPTXInst<(outs Int32Regs:$v0, Int32Regs:$v1,
+ Int32Regs:$v2, Int32Regs:$v3),
+ (ins Int64Regs:$t, Int64Regs:$s, Float32Regs:$x, Float32Regs:$y),
+ "tld4.a.2d.v4.s32.f32\t\\{$v0, $v1, $v2, $v3\\}, "
+ "[$t, $s, \\{$x, $y\\}];",
+ []>;
+def TLD4_R_2D_U32_F32
+ : NVPTXInst<(outs Int32Regs:$v0, Int32Regs:$v1,
+ Int32Regs:$v2, Int32Regs:$v3),
+ (ins Int64Regs:$t, Int64Regs:$s, Float32Regs:$x, Float32Regs:$y),
+ "tld4.r.2d.v4.u32.f32\t\\{$v0, $v1, $v2, $v3\\}, "
+ "[$t, $s, \\{$x, $y\\}];",
+ []>;
+def TLD4_G_2D_U32_F32
+ : NVPTXInst<(outs Int32Regs:$v0, Int32Regs:$v1,
+ Int32Regs:$v2, Int32Regs:$v3),
+ (ins Int64Regs:$t, Int64Regs:$s, Float32Regs:$x, Float32Regs:$y),
+ "tld4.g.2d.v4.u32.f32\t\\{$v0, $v1, $v2, $v3\\}, "
+ "[$t, $s, \\{$x, $y\\}];",
+ []>;
+def TLD4_B_2D_U32_F32
+ : NVPTXInst<(outs Int32Regs:$v0, Int32Regs:$v1,
+ Int32Regs:$v2, Int32Regs:$v3),
+ (ins Int64Regs:$t, Int64Regs:$s, Float32Regs:$x, Float32Regs:$y),
+ "tld4.b.2d.v4.u32.f32\t\\{$v0, $v1, $v2, $v3\\}, "
+ "[$t, $s, \\{$x, $y\\}];",
+ []>;
+def TLD4_A_2D_U32_F32
+ : NVPTXInst<(outs Int32Regs:$v0, Int32Regs:$v1,
+ Int32Regs:$v2, Int32Regs:$v3),
+ (ins Int64Regs:$t, Int64Regs:$s, Float32Regs:$x, Float32Regs:$y),
+ "tld4.a.2d.v4.u32.f32\t\\{$v0, $v1, $v2, $v3\\}, "
+ "[$t, $s, \\{$x, $y\\}];",
+ []>;
+}
+
+
+// texmode_unified
+let IsTex = 1, IsTexModeUnified = 1 in {
+// Texture fetch instructions using handles
+def TEX_UNIFIED_1D_F32_S32
+ : NVPTXInst<(outs Float32Regs:$r, Float32Regs:$g,
+ Float32Regs:$b, Float32Regs:$a),
+ (ins Int64Regs:$t, Int32Regs:$x),
+ "tex.1d.v4.f32.s32\t\\{$r, $g, $b, $a\\}, [$t, \\{$x\\}];",
+ []>;
+def TEX_UNIFIED_1D_F32_F32
+ : NVPTXInst<(outs Float32Regs:$r, Float32Regs:$g,
+ Float32Regs:$b, Float32Regs:$a),
+ (ins Int64Regs:$t, Float32Regs:$x),
+ "tex.1d.v4.f32.f32\t\\{$r, $g, $b, $a\\}, [$t, \\{$x\\}];",
+ []>;
+def TEX_UNIFIED_1D_F32_F32_LEVEL
+ : NVPTXInst<(outs Float32Regs:$r, Float32Regs:$g,
+ Float32Regs:$b, Float32Regs:$a),
+ (ins Int64Regs:$t, Float32Regs:$x, Float32Regs:$lod),
+ "tex.level.1d.v4.f32.f32\t\\{$r, $g, $b, $a\\}, "
+ "[$t, \\{$x\\}], $lod;",
+ []>;
+def TEX_UNIFIED_1D_F32_F32_GRAD
+ : NVPTXInst<(outs Float32Regs:$r, Float32Regs:$g,
+ Float32Regs:$b, Float32Regs:$a),
+ (ins Int64Regs:$t, Float32Regs:$x,
+ Float32Regs:$gradx, Float32Regs:$grady),
+ "tex.grad.1d.v4.f32.f32\t\\{$r, $g, $b, $a\\}, "
+ "[$t, \\{$x\\}], \\{$gradx\\}, \\{$grady\\};",
+ []>;
+def TEX_UNIFIED_1D_S32_S32
+ : NVPTXInst<(outs Int32Regs:$r, Int32Regs:$g,
+ Int32Regs:$b, Int32Regs:$a),
+ (ins Int64Regs:$t, Int32Regs:$x),
+ "tex.1d.v4.s32.s32\t\\{$r, $g, $b, $a\\}, [$t, \\{$x\\}];",
+ []>;
+def TEX_UNIFIED_1D_S32_F32
+ : NVPTXInst<(outs Int32Regs:$r, Int32Regs:$g,
+ Int32Regs:$b, Int32Regs:$a),
+ (ins Int64Regs:$t, Float32Regs:$x),
+ "tex.1d.v4.s32.f32\t\\{$r, $g, $b, $a\\}, [$t, \\{$x\\}];",
+ []>;
+def TEX_UNIFIED_1D_S32_F32_LEVEL
+ : NVPTXInst<(outs Int32Regs:$r, Int32Regs:$g,
+ Int32Regs:$b, Int32Regs:$a),
+ (ins Int64Regs:$t, Float32Regs:$x,
+ Float32Regs:$lod),
+ "tex.level.1d.v4.s32.f32\t\\{$r, $g, $b, $a\\}, "
+ "[$t, \\{$x\\}], $lod;",
+ []>;
+def TEX_UNIFIED_1D_S32_F32_GRAD
+ : NVPTXInst<(outs Int32Regs:$r, Int32Regs:$g,
+ Int32Regs:$b, Int32Regs:$a),
+ (ins Int64Regs:$t, Float32Regs:$x,
+ Float32Regs:$gradx, Float32Regs:$grady),
+ "tex.grad.1d.v4.s32.f32\t\\{$r, $g, $b, $a\\}, "
+ "[$t, \\{$x\\}], \\{$gradx\\}, \\{$grady\\};",
+ []>;
+def TEX_UNIFIED_1D_U32_S32
+ : NVPTXInst<(outs Int32Regs:$r, Int32Regs:$g,
+ Int32Regs:$b, Int32Regs:$a),
+ (ins Int64Regs:$t, Int32Regs:$x),
+ "tex.1d.v4.u32.s32\t\\{$r, $g, $b, $a\\}, [$t, \\{$x\\}];",
+ []>;
+def TEX_UNIFIED_1D_U32_F32
+ : NVPTXInst<(outs Int32Regs:$r, Int32Regs:$g,
+ Int32Regs:$b, Int32Regs:$a),
+ (ins Int64Regs:$t, Float32Regs:$x),
+ "tex.1d.v4.u32.f32\t\\{$r, $g, $b, $a\\}, [$t, \\{$x\\}];",
+ []>;
+def TEX_UNIFIED_1D_U32_F32_LEVEL
+ : NVPTXInst<(outs Int32Regs:$r, Int32Regs:$g,
+ Int32Regs:$b, Int32Regs:$a),
+ (ins Int64Regs:$t, Float32Regs:$x,
+ Float32Regs:$lod),
+ "tex.level.1d.v4.u32.f32\t\\{$r, $g, $b, $a\\}, "
+ "[$t, \\{$x\\}], $lod;",
+ []>;
+def TEX_UNIFIED_1D_U32_F32_GRAD
+ : NVPTXInst<(outs Int32Regs:$r, Int32Regs:$g,
+ Int32Regs:$b, Int32Regs:$a),
+ (ins Int64Regs:$t, Float32Regs:$x,
+ Float32Regs:$gradx, Float32Regs:$grady),
+ "tex.grad.1d.v4.u32.f32\t\\{$r, $g, $b, $a\\}, "
+ "[$t, \\{$x\\}], \\{$gradx\\}, \\{$grady\\};",
+ []>;
+
+def TEX_UNIFIED_1D_ARRAY_F32_S32
+ : NVPTXInst<(outs Float32Regs:$r, Float32Regs:$g,
+ Float32Regs:$b, Float32Regs:$a),
+ (ins Int64Regs:$t, Int32Regs:$l, Int32Regs:$x),
+ "tex.a1d.v4.f32.s32\t\\{$r, $g, $b, $a\\}, "
+ "[$t, \\{$l, $x\\}];",
+ []>;
+def TEX_UNIFIED_1D_ARRAY_F32_F32
+ : NVPTXInst<(outs Float32Regs:$r, Float32Regs:$g,
+ Float32Regs:$b, Float32Regs:$a),
+ (ins Int64Regs:$t, Int32Regs:$l, Float32Regs:$x),
+ "tex.a1d.v4.f32.f32\t\\{$r, $g, $b, $a\\}, "
+ "[$t, \\{$l, $x\\}];",
+ []>;
+def TEX_UNIFIED_1D_ARRAY_F32_F32_LEVEL
+ : NVPTXInst<(outs Float32Regs:$r, Float32Regs:$g,
+ Float32Regs:$b, Float32Regs:$a),
+ (ins Int64Regs:$t, Int32Regs:$l, Float32Regs:$x,
+ Float32Regs:$lod),
+ "tex.level.a1d.v4.f32.f32\t\\{$r, $g, $b, $a\\}, "
+ "[$t, \\{$l, $x\\}], $lod;",
+ []>;
+def TEX_UNIFIED_1D_ARRAY_F32_F32_GRAD
+ : NVPTXInst<(outs Float32Regs:$r, Float32Regs:$g,
+ Float32Regs:$b, Float32Regs:$a),
+ (ins Int64Regs:$t, Int32Regs:$l, Float32Regs:$x,
+ Float32Regs:$gradx, Float32Regs:$grady),
+ "tex.grad.a1d.v4.f32.f32\t\\{$r, $g, $b, $a\\}, "
+ "[$t, \\{$l, $x\\}], \\{$gradx\\}, \\{$grady\\};",
+ []>;
+def TEX_UNIFIED_1D_ARRAY_S32_S32
+ : NVPTXInst<(outs Int32Regs:$r, Int32Regs:$g,
+ Int32Regs:$b, Int32Regs:$a),
+ (ins Int64Regs:$t, Int32Regs:$l, Int32Regs:$x),
+ "tex.a1d.v4.s32.s32\t\\{$r, $g, $b, $a\\}, "
+ "[$t, \\{$l, $x\\}];",
+ []>;
+def TEX_UNIFIED_1D_ARRAY_S32_F32
+ : NVPTXInst<(outs Int32Regs:$r, Int32Regs:$g,
+ Int32Regs:$b, Int32Regs:$a),
+ (ins Int64Regs:$t, Int32Regs:$l, Float32Regs:$x),
+ "tex.a1d.v4.s32.f32\t\\{$r, $g, $b, $a\\}, "
+ "[$t, \\{$l, $x\\}];",
+ []>;
+def TEX_UNIFIED_1D_ARRAY_S32_F32_LEVEL
+ : NVPTXInst<(outs Int32Regs:$r, Int32Regs:$g,
+ Int32Regs:$b, Int32Regs:$a),
+ (ins Int64Regs:$t, Int32Regs:$l, Float32Regs:$x,
+ Float32Regs:$lod),
+ "tex.level.a1d.v4.s32.f32\t\\{$r, $g, $b, $a\\}, "
+ "[$t, \\{$l, $x\\}], $lod;",
+ []>;
+def TEX_UNIFIED_1D_ARRAY_S32_F32_GRAD
+ : NVPTXInst<(outs Int32Regs:$r, Int32Regs:$g,
+ Int32Regs:$b, Int32Regs:$a),
+ (ins Int64Regs:$t, Int32Regs:$l, Float32Regs:$x,
+ Float32Regs:$gradx, Float32Regs:$grady),
+ "tex.grad.a1d.v4.s32.f32\t\\{$r, $g, $b, $a\\}, "
+ "[$t, \\{$l, $x\\}], \\{$gradx\\}, \\{$grady\\};",
+ []>;
+def TEX_UNIFIED_1D_ARRAY_U32_S32
+ : NVPTXInst<(outs Int32Regs:$r, Int32Regs:$g,
+ Int32Regs:$b, Int32Regs:$a),
+ (ins Int64Regs:$t, Int32Regs:$l, Int32Regs:$x),
+ "tex.a1d.v4.u32.s32\t\\{$r, $g, $b, $a\\}, "
+ "[$t, \\{$l, $x\\}];",
+ []>;
+def TEX_UNIFIED_1D_ARRAY_U32_F32
+ : NVPTXInst<(outs Int32Regs:$r, Int32Regs:$g,
+ Int32Regs:$b, Int32Regs:$a),
+ (ins Int64Regs:$t, Int32Regs:$l, Float32Regs:$x),
+ "tex.a1d.v4.u32.f32\t\\{$r, $g, $b, $a\\}, "
+ "[$t, \\{$l, $x\\}];",
+ []>;
+def TEX_UNIFIED_1D_ARRAY_U32_F32_LEVEL
+ : NVPTXInst<(outs Int32Regs:$r, Int32Regs:$g,
+ Int32Regs:$b, Int32Regs:$a),
+ (ins Int64Regs:$t, Int32Regs:$l, Float32Regs:$x,
+ Float32Regs:$lod),
+ "tex.level.a1d.v4.u32.f32\t\\{$r, $g, $b, $a\\}, "
+ "[$t, \\{$l, $x\\}], $lod;",
+ []>;
+def TEX_UNIFIED_1D_ARRAY_U32_F32_GRAD
+ : NVPTXInst<(outs Int32Regs:$r, Int32Regs:$g,
+ Int32Regs:$b, Int32Regs:$a),
+ (ins Int64Regs:$t, Int32Regs:$l, Float32Regs:$x,
+ Float32Regs:$gradx, Float32Regs:$grady),
+ "tex.grad.a1d.v4.u32.f32\t\\{$r, $g, $b, $a\\}, "
+ "[$t, \\{$l, $x\\}], \\{$gradx\\}, \\{$grady\\};",
+ []>;
+
+def TEX_UNIFIED_2D_F32_S32
+ : NVPTXInst<(outs Float32Regs:$r, Float32Regs:$g,
+ Float32Regs:$b, Float32Regs:$a),
+ (ins Int64Regs:$t, Int32Regs:$x, Int32Regs:$y),
+ "tex.2d.v4.f32.s32\t\\{$r, $g, $b, $a\\}, "
+ "[$t, \\{$x, $y\\}];",
+ []>;
+def TEX_UNIFIED_2D_F32_F32
+ : NVPTXInst<(outs Float32Regs:$r, Float32Regs:$g,
+ Float32Regs:$b, Float32Regs:$a),
+ (ins Int64Regs:$t, Float32Regs:$x, Float32Regs:$y),
+ "tex.2d.v4.f32.f32\t\\{$r, $g, $b, $a\\}, "
+ "[$t, \\{$x, $y\\}];",
+ []>;
+def TEX_UNIFIED_2D_F32_F32_LEVEL
+ : NVPTXInst<(outs Float32Regs:$r, Float32Regs:$g,
+ Float32Regs:$b, Float32Regs:$a),
+ (ins Int64Regs:$t, Float32Regs:$x, Float32Regs:$y,
+ Float32Regs:$lod),
+ "tex.level.2d.v4.f32.f32\t\\{$r, $g, $b, $a\\}, "
+ "[$t, \\{$x, $y\\}], $lod;",
+ []>;
+def TEX_UNIFIED_2D_F32_F32_GRAD
+ : NVPTXInst<(outs Float32Regs:$r, Float32Regs:$g,
+ Float32Regs:$b, Float32Regs:$a),
+ (ins Int64Regs:$t, Float32Regs:$x, Float32Regs:$y,
+ Float32Regs:$gradx0, Float32Regs:$gradx1,
+ Float32Regs:$grady0, Float32Regs:$grady1),
+ "tex.grad.2d.v4.f32.f32\t\\{$r, $g, $b, $a\\}, "
+ "[$t, \\{$x, $y\\}], \\{$gradx0, $gradx1\\}, "
+ "\\{$grady0, $grady1\\};",
+ []>;
+def TEX_UNIFIED_2D_S32_S32
+ : NVPTXInst<(outs Int32Regs:$r, Int32Regs:$g,
+ Int32Regs:$b, Int32Regs:$a),
+ (ins Int64Regs:$t, Int32Regs:$x, Int32Regs:$y),
+ "tex.2d.v4.s32.s32\t\\{$r, $g, $b, $a\\}, "
+ "[$t, \\{$x, $y\\}];",
+ []>;
+def TEX_UNIFIED_2D_S32_F32
+ : NVPTXInst<(outs Int32Regs:$r, Int32Regs:$g,
+ Int32Regs:$b, Int32Regs:$a),
+ (ins Int64Regs:$t, Float32Regs:$x, Float32Regs:$y),
+ "tex.2d.v4.s32.f32\t\\{$r, $g, $b, $a\\}, "
+ "[$t, \\{$x, $y\\}];",
+ []>;
+def TEX_UNIFIED_2D_S32_F32_LEVEL
+ : NVPTXInst<(outs Int32Regs:$r, Int32Regs:$g,
+ Int32Regs:$b, Int32Regs:$a),
+ (ins Int64Regs:$t, Float32Regs:$x, Float32Regs:$y,
+ Float32Regs:$lod),
+ "tex.level.2d.v4.s32.f32\t\\{$r, $g, $b, $a\\}, "
+ "[$t, \\{$x, $y\\}], $lod;",
+ []>;
+def TEX_UNIFIED_2D_S32_F32_GRAD
+ : NVPTXInst<(outs Int32Regs:$r, Int32Regs:$g,
+ Int32Regs:$b, Int32Regs:$a),
+ (ins Int64Regs:$t, Float32Regs:$x, Float32Regs:$y,
+ Float32Regs:$gradx0, Float32Regs:$gradx1,
+ Float32Regs:$grady0, Float32Regs:$grady1),
+ "tex.grad.2d.v4.s32.f32\t\\{$r, $g, $b, $a\\}, "
+ "[$t, \\{$x, $y\\}], \\{$gradx0, $gradx1\\}, "
+ "\\{$grady0, $grady1\\};",
+ []>;
+def TEX_UNIFIED_2D_U32_S32
+ : NVPTXInst<(outs Int32Regs:$r, Int32Regs:$g,
+ Int32Regs:$b, Int32Regs:$a),
+ (ins Int64Regs:$t, Int32Regs:$x, Int32Regs:$y),
+ "tex.2d.v4.u32.s32\t\\{$r, $g, $b, $a\\}, "
+ "[$t, \\{$x, $y\\}];",
+ []>;
+def TEX_UNIFIED_2D_U32_F32
+ : NVPTXInst<(outs Int32Regs:$r, Int32Regs:$g,
+ Int32Regs:$b, Int32Regs:$a),
+ (ins Int64Regs:$t, Float32Regs:$x, Float32Regs:$y),
+ "tex.2d.v4.u32.f32\t\\{$r, $g, $b, $a\\}, "
+ "[$t, \\{$x, $y\\}];",
+ []>;
+def TEX_UNIFIED_2D_U32_F32_LEVEL
+ : NVPTXInst<(outs Int32Regs:$r, Int32Regs:$g,
+ Int32Regs:$b, Int32Regs:$a),
+ (ins Int64Regs:$t, Float32Regs:$x, Float32Regs:$y,
+ Float32Regs:$lod),
+ "tex.level.2d.v4.u32.f32\t\\{$r, $g, $b, $a\\}, "
+ "[$t, \\{$x, $y\\}], $lod;",
+ []>;
+def TEX_UNIFIED_2D_U32_F32_GRAD
+ : NVPTXInst<(outs Int32Regs:$r, Int32Regs:$g,
+ Int32Regs:$b, Int32Regs:$a),
+ (ins Int64Regs:$t, Float32Regs:$x, Float32Regs:$y,
+ Float32Regs:$gradx0, Float32Regs:$gradx1,
+ Float32Regs:$grady0, Float32Regs:$grady1),
+ "tex.grad.2d.v4.u32.f32\t\\{$r, $g, $b, $a\\}, "
+ "[$t, \\{$x, $y\\}], \\{$gradx0, $gradx1\\}, "
+ "\\{$grady0, $grady1\\};",
+ []>;
+
+def TEX_UNIFIED_2D_ARRAY_F32_S32
+ : NVPTXInst<(outs Float32Regs:$r, Float32Regs:$g,
+ Float32Regs:$b, Float32Regs:$a),
+ (ins Int64Regs:$t, Int32Regs:$l, Int32Regs:$x,
+ Int32Regs:$y),
+ "tex.a2d.v4.f32.s32\t\\{$r, $g, $b, $a\\}, "
+ "[$t, \\{$l, $x, $y, $y\\}];",
+ []>;
+def TEX_UNIFIED_2D_ARRAY_F32_F32
+ : NVPTXInst<(outs Float32Regs:$r, Float32Regs:$g,
+ Float32Regs:$b, Float32Regs:$a),
+ (ins Int64Regs:$t, Int32Regs:$l, Float32Regs:$x,
+ Float32Regs:$y),
+ "tex.a2d.v4.f32.f32\t\\{$r, $g, $b, $a\\}, "
+ "[$t, \\{$l, $x, $y, $y\\}];",
+ []>;
+def TEX_UNIFIED_2D_ARRAY_F32_F32_LEVEL
+ : NVPTXInst<(outs Float32Regs:$r, Float32Regs:$g,
+ Float32Regs:$b, Float32Regs:$a),
+ (ins Int64Regs:$t, Int32Regs:$l, Float32Regs:$x,
+ Float32Regs:$y, Float32Regs:$lod),
+ "tex.level.a2d.v4.f32.f32\t\\{$r, $g, $b, $a\\}, "
+ "[$t, \\{$l, $x, $y, $y\\}], $lod;",
+ []>;
+def TEX_UNIFIED_2D_ARRAY_F32_F32_GRAD
+ : NVPTXInst<(outs Float32Regs:$r, Float32Regs:$g,
+ Float32Regs:$b, Float32Regs:$a),
+ (ins Int64Regs:$t, Int32Regs:$l, Float32Regs:$x,
+ Float32Regs:$y, Float32Regs:$gradx0, Float32Regs:$gradx1,
+ Float32Regs:$grady0, Float32Regs:$grady1),
+ "tex.grad.a2d.v4.f32.f32\t\\{$r, $g, $b, $a\\}, "
+ "[$t, \\{$l, $x, $y, $y\\}], \\{$gradx0, $gradx1\\}, "
+ "\\{$grady0, $grady1\\};",
+ []>;
+def TEX_UNIFIED_2D_ARRAY_S32_S32
+ : NVPTXInst<(outs Int32Regs:$r, Int32Regs:$g,
+ Int32Regs:$b, Int32Regs:$a),
+ (ins Int64Regs:$t, Int32Regs:$l, Int32Regs:$x,
+ Int32Regs:$y),
+ "tex.a2d.v4.s32.s32\t\\{$r, $g, $b, $a\\}, "
+ "[$t, \\{$l, $x, $y, $y\\}];",
+ []>;
+def TEX_UNIFIED_2D_ARRAY_S32_F32
+ : NVPTXInst<(outs Int32Regs:$r, Int32Regs:$g,
+ Int32Regs:$b, Int32Regs:$a),
+ (ins Int64Regs:$t, Int32Regs:$l, Float32Regs:$x,
+ Float32Regs:$y),
+ "tex.a2d.v4.s32.f32\t\\{$r, $g, $b, $a\\}, "
+ "[$t, \\{$l, $x, $y, $y\\}];",
+ []>;
+def TEX_UNIFIED_2D_ARRAY_S32_F32_LEVEL
+ : NVPTXInst<(outs Int32Regs:$r, Int32Regs:$g,
+ Int32Regs:$b, Int32Regs:$a),
+ (ins Int64Regs:$t, Int32Regs:$l, Float32Regs:$x,
+ Float32Regs:$y, Float32Regs:$lod),
+ "tex.level.a2d.v4.s32.f32\t\\{$r, $g, $b, $a\\}, "
+ "[$t, \\{$l, $x, $y, $y\\}], $lod;",
+ []>;
+def TEX_UNIFIED_2D_ARRAY_S32_F32_GRAD
+ : NVPTXInst<(outs Int32Regs:$r, Int32Regs:$g,
+ Int32Regs:$b, Int32Regs:$a),
+ (ins Int64Regs:$t, Int32Regs:$l, Float32Regs:$x,
+ Float32Regs:$y,
+ Float32Regs:$gradx0, Float32Regs:$gradx1,
+ Float32Regs:$grady0, Float32Regs:$grady1),
+ "tex.grad.a2d.v4.s32.f32\t\\{$r, $g, $b, $a\\}, "
+ "[$t, \\{$l, $x, $y, $y\\}], \\{$gradx0, $gradx1\\}, "
+ "\\{$grady0, $grady1\\};",
+ []>;
+def TEX_UNIFIED_2D_ARRAY_U32_S32
+ : NVPTXInst<(outs Int32Regs:$r, Int32Regs:$g,
+ Int32Regs:$b, Int32Regs:$a),
+ (ins Int64Regs:$t, Int32Regs:$l, Int32Regs:$x,
+ Int32Regs:$y),
+ "tex.a2d.v4.u32.s32\t\\{$r, $g, $b, $a\\}, "
+ "[$t, \\{$l, $x, $y, $y\\}];",
+ []>;
+def TEX_UNIFIED_2D_ARRAY_U32_F32
+ : NVPTXInst<(outs Int32Regs:$r, Int32Regs:$g,
+ Int32Regs:$b, Int32Regs:$a),
+ (ins Int64Regs:$t, Int32Regs:$l, Float32Regs:$x,
+ Float32Regs:$y),
+ "tex.a2d.v4.u32.f32\t\\{$r, $g, $b, $a\\}, "
+ "[$t, \\{$l, $x, $y, $y\\}];",
+ []>;
+def TEX_UNIFIED_2D_ARRAY_U32_F32_LEVEL
+ : NVPTXInst<(outs Int32Regs:$r, Int32Regs:$g,
+ Int32Regs:$b, Int32Regs:$a),
+ (ins Int64Regs:$t, Int32Regs:$l, Float32Regs:$x,
+ Float32Regs:$y, Float32Regs:$lod),
+ "tex.level.a2d.v4.u32.f32\t\\{$r, $g, $b, $a\\}, "
+ "[$t, \\{$l, $x, $y, $y\\}], $lod;",
+ []>;
+def TEX_UNIFIED_2D_ARRAY_U32_F32_GRAD
+ : NVPTXInst<(outs Int32Regs:$r, Int32Regs:$g,
+ Int32Regs:$b, Int32Regs:$a),
+ (ins Int64Regs:$t, Int32Regs:$l, Float32Regs:$x,
+ Float32Regs:$y,
+ Float32Regs:$gradx0, Float32Regs:$gradx1,
+ Float32Regs:$grady0, Float32Regs:$grady1),
+ "tex.grad.a2d.v4.u32.f32\t\\{$r, $g, $b, $a\\}, "
+ "[$t, \\{$l, $x, $y, $y\\}], \\{$gradx0, $gradx1\\}, "
+ "\\{$grady0, $grady1\\};",
+ []>;
+
+def TEX_UNIFIED_3D_F32_S32
+ : NVPTXInst<(outs Float32Regs:$r, Float32Regs:$g,
+ Float32Regs:$b, Float32Regs:$a),
+ (ins Int64Regs:$t, Int32Regs:$x, Int32Regs:$y,
+ Int32Regs:$z),
+ "tex.3d.v4.f32.s32\t\\{$r, $g, $b, $a\\}, "
+ "[$t, \\{$x, $y, $z, $z\\}];",
+ []>;
+def TEX_UNIFIED_3D_F32_F32
+ : NVPTXInst<(outs Float32Regs:$r, Float32Regs:$g,
+ Float32Regs:$b, Float32Regs:$a),
+ (ins Int64Regs:$t, Float32Regs:$x, Float32Regs:$y,
+ Float32Regs:$z),
+ "tex.3d.v4.f32.f32\t\\{$r, $g, $b, $a\\}, "
+ "[$t, \\{$x, $y, $z, $z\\}];",
+ []>;
+def TEX_UNIFIED_3D_F32_F32_LEVEL
+ : NVPTXInst<(outs Float32Regs:$r, Float32Regs:$g,
+ Float32Regs:$b, Float32Regs:$a),
+ (ins Int64Regs:$t, Float32Regs:$x, Float32Regs:$y,
+ Float32Regs:$z, Float32Regs:$lod),
+ "tex.level.3d.v4.f32.f32\t\\{$r, $g, $b, $a\\}, "
+ "[$t, \\{$x, $y, $z, $z\\}], $lod;",
+ []>;
+def TEX_UNIFIED_3D_F32_F32_GRAD
+ : NVPTXInst<(outs Float32Regs:$r, Float32Regs:$g,
+ Float32Regs:$b, Float32Regs:$a),
+ (ins Int64Regs:$t, Float32Regs:$x, Float32Regs:$y,
+ Float32Regs:$z,
+ Float32Regs:$gradx0, Float32Regs:$gradx1,
+ Float32Regs:$gradx2, Float32Regs:$grady0,
+ Float32Regs:$grady1, Float32Regs:$grady2),
+ "tex.grad.3d.v4.f32.f32\t\\{$r, $g, $b, $a\\}, "
+ "[$t, \\{$x, $y, $z, $z\\}], "
+ "\\{$gradx0, $gradx1, $gradx2, $gradx2\\}, "
+ "\\{$grady0, $grady1, $grady2, $grady2\\};",
+ []>;
+def TEX_UNIFIED_3D_S32_S32
+ : NVPTXInst<(outs Int32Regs:$r, Int32Regs:$g,
+ Int32Regs:$b, Int32Regs:$a),
+ (ins Int64Regs:$t, Int32Regs:$x, Int32Regs:$y,
+ Int32Regs:$z),
+ "tex.3d.v4.s32.s32\t\\{$r, $g, $b, $a\\}, "
+ "[$t, \\{$x, $y, $z, $z\\}];",
+ []>;
+def TEX_UNIFIED_3D_S32_F32
+ : NVPTXInst<(outs Int32Regs:$r, Int32Regs:$g,
+ Int32Regs:$b, Int32Regs:$a),
+ (ins Int64Regs:$t, Float32Regs:$x, Float32Regs:$y,
+ Float32Regs:$z),
+ "tex.3d.v4.s32.f32\t\\{$r, $g, $b, $a\\}, "
+ "[$t, \\{$x, $y, $z, $z\\}];",
+ []>;
+def TEX_UNIFIED_3D_S32_F32_LEVEL
+ : NVPTXInst<(outs Int32Regs:$r, Int32Regs:$g,
+ Int32Regs:$b, Int32Regs:$a),
+ (ins Int64Regs:$t, Float32Regs:$x, Float32Regs:$y,
+ Float32Regs:$z, Float32Regs:$lod),
+ "tex.level.3d.v4.s32.f32\t\\{$r, $g, $b, $a\\}, "
+ "[$t, \\{$x, $y, $z, $z\\}], $lod;",
+ []>;
+def TEX_UNIFIED_3D_S32_F32_GRAD
+ : NVPTXInst<(outs Int32Regs:$r, Int32Regs:$g,
+ Int32Regs:$b, Int32Regs:$a),
+ (ins Int64Regs:$t, Float32Regs:$x, Float32Regs:$y,
+ Float32Regs:$z,
+ Float32Regs:$gradx0, Float32Regs:$gradx1,
+ Float32Regs:$gradx2, Float32Regs:$grady0,
+ Float32Regs:$grady1, Float32Regs:$grady2),
+ "tex.grad.3d.v4.s32.f32\t\\{$r, $g, $b, $a\\}, "
+ "[$t, \\{$x, $y, $z, $z\\}], "
+ "\\{$gradx0, $gradx1, $gradx2, $gradx2\\}, "
+ "\\{$grady0, $grady1, $grady2, $grady2\\};",
+ []>;
+def TEX_UNIFIED_3D_U32_S32
+ : NVPTXInst<(outs Int32Regs:$r, Int32Regs:$g,
+ Int32Regs:$b, Int32Regs:$a),
+ (ins Int64Regs:$t, Int32Regs:$x, Int32Regs:$y,
+ Int32Regs:$z),
+ "tex.3d.v4.u32.s32\t\\{$r, $g, $b, $a\\}, "
+ "[$t, \\{$x, $y, $z, $z\\}];",
+ []>;
+def TEX_UNIFIED_3D_U32_F32
+ : NVPTXInst<(outs Int32Regs:$r, Int32Regs:$g,
+ Int32Regs:$b, Int32Regs:$a),
+ (ins Int64Regs:$t, Float32Regs:$x, Float32Regs:$y,
+ Float32Regs:$z),
+ "tex.3d.v4.u32.f32\t\\{$r, $g, $b, $a\\}, "
+ "[$t, \\{$x, $y, $z, $z\\}];",
+ []>;
+def TEX_UNIFIED_3D_U32_F32_LEVEL
+ : NVPTXInst<(outs Int32Regs:$r, Int32Regs:$g,
+ Int32Regs:$b, Int32Regs:$a),
+ (ins Int64Regs:$t, Float32Regs:$x, Float32Regs:$y,
+ Float32Regs:$z, Float32Regs:$lod),
+ "tex.level.3d.v4.u32.f32\t\\{$r, $g, $b, $a\\}, "
+ "[$t, \\{$x, $y, $z, $z\\}], $lod;",
+ []>;
+def TEX_UNIFIED_3D_U32_F32_GRAD
+ : NVPTXInst<(outs Int32Regs:$r, Int32Regs:$g,
+ Int32Regs:$b, Int32Regs:$a),
+ (ins Int64Regs:$t, Float32Regs:$x, Float32Regs:$y,
+ Float32Regs:$z,
+ Float32Regs:$gradx0, Float32Regs:$gradx1,
+ Float32Regs:$gradx2, Float32Regs:$grady0,
+ Float32Regs:$grady1, Float32Regs:$grady2),
+ "tex.grad.3d.v4.u32.f32\t\\{$r, $g, $b, $a\\}, "
+ "[$t, \\{$x, $y, $z, $z\\}], "
+ "\\{$gradx0, $gradx1, $gradx2, $gradx2\\}, "
+ "\\{$grady0, $grady1, $grady2, $grady2\\};",
+ []>;
+
+def TEX_UNIFIED_CUBE_F32_F32
+ : NVPTXInst<(outs Float32Regs:$r, Float32Regs:$g,
+ Float32Regs:$b, Float32Regs:$a),
+ (ins Int64Regs:$t,
+ Float32Regs:$x, Float32Regs:$y, Float32Regs:$z),
+ "tex.cube.v4.f32.f32\t\\{$r, $g, $b, $a\\}, "
+ "[$t, \\{$x, $y, $z, $z\\}];",
+ []>;
+def TEX_UNIFIED_CUBE_F32_F32_LEVEL
+ : NVPTXInst<(outs Float32Regs:$r, Float32Regs:$g,
+ Float32Regs:$b, Float32Regs:$a),
+ (ins Int64Regs:$t,
+ Float32Regs:$x, Float32Regs:$y, Float32Regs:$z,
+ Float32Regs:$lod),
+ "tex.level.cube.v4.f32.f32\t\\{$r, $g, $b, $a\\}, "
+ "[$t, \\{$x, $y, $z, $z\\}], $lod;",
+ []>;
+def TEX_UNIFIED_CUBE_S32_F32
+ : NVPTXInst<(outs Int32Regs:$r, Int32Regs:$g,
+ Int32Regs:$b, Int32Regs:$a),
+ (ins Int64Regs:$t,
+ Float32Regs:$x, Float32Regs:$y, Float32Regs:$z),
+ "tex.cube.v4.s32.f32\t\\{$r, $g, $b, $a\\}, "
+ "[$t, \\{$x, $y, $z, $z\\}];",
+ []>;
+def TEX_UNIFIED_CUBE_S32_F32_LEVEL
+ : NVPTXInst<(outs Int32Regs:$r, Int32Regs:$g,
+ Int32Regs:$b, Int32Regs:$a),
+ (ins Int64Regs:$t,
+ Float32Regs:$x, Float32Regs:$y, Float32Regs:$z,
+ Float32Regs:$lod),
+ "tex.level.cube.v4.s32.f32\t\\{$r, $g, $b, $a\\}, "
+ "[$t, \\{$x, $y, $z, $z\\}], $lod;",
+ []>;
+def TEX_UNIFIED_CUBE_U32_F32
+ : NVPTXInst<(outs Int32Regs:$r, Int32Regs:$g,
+ Int32Regs:$b, Int32Regs:$a),
+ (ins Int64Regs:$t,
+ Float32Regs:$x, Float32Regs:$y, Float32Regs:$z),
+ "tex.cube.v4.u32.f32\t\\{$r, $g, $b, $a\\}, "
+ "[$t, \\{$x, $y, $z, $z\\}];",
+ []>;
+def TEX_UNIFIED_CUBE_U32_F32_LEVEL
+ : NVPTXInst<(outs Int32Regs:$r, Int32Regs:$g,
+ Int32Regs:$b, Int32Regs:$a),
+ (ins Int64Regs:$t,
+ Float32Regs:$x, Float32Regs:$y, Float32Regs:$z,
+ Float32Regs:$lod),
+ "tex.level.cube.v4.u32.f32\t\\{$r, $g, $b, $a\\}, "
+ "[$t, \\{$x, $y, $z, $z\\}], $lod;",
+ []>;
+
+def TEX_UNIFIED_CUBE_ARRAY_F32_F32
+ : NVPTXInst<(outs Float32Regs:$r, Float32Regs:$g,
+ Float32Regs:$b, Float32Regs:$a),
+ (ins Int64Regs:$t, Int32Regs:$l,
+ Float32Regs:$x, Float32Regs:$y, Float32Regs:$z),
+ "tex.acube.v4.f32.f32\t\\{$r, $g, $b, $a\\}, "
+ "[$t, \\{$l, $x, $y, $z\\}];",
+ []>;
+def TEX_UNIFIED_CUBE_ARRAY_F32_F32_LEVEL
+ : NVPTXInst<(outs Float32Regs:$r, Float32Regs:$g,
+ Float32Regs:$b, Float32Regs:$a),
+ (ins Int64Regs:$t, Int32Regs:$l,
+ Float32Regs:$x, Float32Regs:$y, Float32Regs:$z,
+ Float32Regs:$lod),
+ "tex.level.acube.v4.f32.f32\t\\{$r, $g, $b, $a\\}, "
+ "[$t, \\{$l, $x, $y, $z\\}], $lod;",
+ []>;
+def TEX_UNIFIED_CUBE_ARRAY_S32_F32
+ : NVPTXInst<(outs Int32Regs:$r, Int32Regs:$g,
+ Int32Regs:$b, Int32Regs:$a),
+ (ins Int64Regs:$t, Int32Regs:$l,
+ Float32Regs:$x, Float32Regs:$y, Float32Regs:$z),
+ "tex.acube.v4.s32.f32\t\\{$r, $g, $b, $a\\}, "
+ "[$t, \\{$l, $x, $y, $z\\}];",
+ []>;
+def TEX_UNIFIED_CUBE_ARRAY_S32_F32_LEVEL
+ : NVPTXInst<(outs Int32Regs:$r, Int32Regs:$g,
+ Int32Regs:$b, Int32Regs:$a),
+ (ins Int64Regs:$t, Int32Regs:$l,
+ Float32Regs:$x, Float32Regs:$y, Float32Regs:$z,
+ Float32Regs:$lod),
+ "tex.level.acube.v4.s32.f32\t\\{$r, $g, $b, $a\\}, "
+ "[$t, \\{$l, $x, $y, $z\\}], $lod;",
+ []>;
+def TEX_UNIFIED_CUBE_ARRAY_U32_F32
+ : NVPTXInst<(outs Int32Regs:$r, Int32Regs:$g,
+ Int32Regs:$b, Int32Regs:$a),
+ (ins Int64Regs:$t, Int32Regs:$l,
+ Float32Regs:$x, Float32Regs:$y, Float32Regs:$z),
+ "tex.acube.v4.u32.f32\t\\{$r, $g, $b, $a\\}, "
+ "[$t, \\{$l, $x, $y, $z\\}];",
+ []>;
+def TEX_UNIFIED_CUBE_ARRAY_U32_F32_LEVEL
+ : NVPTXInst<(outs Int32Regs:$r, Int32Regs:$g,
+ Int32Regs:$b, Int32Regs:$a),
+ (ins Int64Regs:$t, Int32Regs:$l,
+ Float32Regs:$x, Float32Regs:$y, Float32Regs:$z,
+ Float32Regs:$lod),
+ "tex.level.acube.v4.u32.f32\t\\{$r, $g, $b, $a\\}, "
+ "[$t, \\{$l, $x, $y, $z\\}], $lod;",
+ []>;
+
+def TLD4_UNIFIED_R_2D_F32_F32
+ : NVPTXInst<(outs Float32Regs:$v0, Float32Regs:$v1,
+ Float32Regs:$v2, Float32Regs:$v3),
+ (ins Int64Regs:$t, Float32Regs:$x, Float32Regs:$y),
+ "tld4.r.2d.v4.f32.f32\t\\{$v0, $v1, $v2, $v3\\}, "
+ "[$t, \\{$x, $y\\}];",
+ []>;
+def TLD4_UNIFIED_G_2D_F32_F32
+ : NVPTXInst<(outs Float32Regs:$v0, Float32Regs:$v1,
+ Float32Regs:$v2, Float32Regs:$v3),
+ (ins Int64Regs:$t, Float32Regs:$x, Float32Regs:$y),
+ "tld4.g.2d.v4.f32.f32\t\\{$v0, $v1, $v2, $v3\\}, "
+ "[$t, \\{$x, $y\\}];",
+ []>;
+def TLD4_UNIFIED_B_2D_F32_F32
+ : NVPTXInst<(outs Float32Regs:$v0, Float32Regs:$v1,
+ Float32Regs:$v2, Float32Regs:$v3),
+ (ins Int64Regs:$t, Float32Regs:$x, Float32Regs:$y),
+ "tld4.b.2d.v4.f32.f32\t\\{$v0, $v1, $v2, $v3\\}, "
+ "[$t, \\{$x, $y\\}];",
+ []>;
+def TLD4_UNIFIED_A_2D_F32_F32
+ : NVPTXInst<(outs Float32Regs:$v0, Float32Regs:$v1,
+ Float32Regs:$v2, Float32Regs:$v3),
+ (ins Int64Regs:$t, Float32Regs:$x, Float32Regs:$y),
+ "tld4.a.2d.v4.f32.f32\t\\{$v0, $v1, $v2, $v3\\}, "
+ "[$t, \\{$x, $y\\}];",
+ []>;
+def TLD4_UNIFIED_R_2D_S32_F32
+ : NVPTXInst<(outs Int32Regs:$v0, Int32Regs:$v1,
+ Int32Regs:$v2, Int32Regs:$v3),
+ (ins Int64Regs:$t, Float32Regs:$x, Float32Regs:$y),
+ "tld4.r.2d.v4.s32.f32\t\\{$v0, $v1, $v2, $v3\\}, "
+ "[$t, \\{$x, $y\\}];",
+ []>;
+def TLD4_UNIFIED_G_2D_S32_F32
+ : NVPTXInst<(outs Int32Regs:$v0, Int32Regs:$v1,
+ Int32Regs:$v2, Int32Regs:$v3),
+ (ins Int64Regs:$t, Float32Regs:$x, Float32Regs:$y),
+ "tld4.g.2d.v4.s32.f32\t\\{$v0, $v1, $v2, $v3\\}, "
+ "[$t, \\{$x, $y\\}];",
+ []>;
+def TLD4_UNIFIED_B_2D_S32_F32
+ : NVPTXInst<(outs Int32Regs:$v0, Int32Regs:$v1,
+ Int32Regs:$v2, Int32Regs:$v3),
+ (ins Int64Regs:$t, Float32Regs:$x, Float32Regs:$y),
+ "tld4.b.2d.v4.s32.f32\t\\{$v0, $v1, $v2, $v3\\}, "
+ "[$t, \\{$x, $y\\}];",
+ []>;
+def TLD4_UNIFIED_A_2D_S32_F32
+ : NVPTXInst<(outs Int32Regs:$v0, Int32Regs:$v1,
+ Int32Regs:$v2, Int32Regs:$v3),
+ (ins Int64Regs:$t, Float32Regs:$x, Float32Regs:$y),
+ "tld4.a.2d.v4.s32.f32\t\\{$v0, $v1, $v2, $v3\\}, "
+ "[$t, \\{$x, $y\\}];",
+ []>;
+def TLD4_UNIFIED_R_2D_U32_F32
+ : NVPTXInst<(outs Int32Regs:$v0, Int32Regs:$v1,
+ Int32Regs:$v2, Int32Regs:$v3),
+ (ins Int64Regs:$t, Float32Regs:$x, Float32Regs:$y),
+ "tld4.r.2d.v4.u32.f32\t\\{$v0, $v1, $v2, $v3\\}, "
+ "[$t, \\{$x, $y\\}];",
+ []>;
+def TLD4_UNIFIED_G_2D_U32_F32
+ : NVPTXInst<(outs Int32Regs:$v0, Int32Regs:$v1,
+ Int32Regs:$v2, Int32Regs:$v3),
+ (ins Int64Regs:$t, Float32Regs:$x, Float32Regs:$y),
+ "tld4.g.2d.v4.u32.f32\t\\{$v0, $v1, $v2, $v3\\}, "
+ "[$t, \\{$x, $y\\}];",
+ []>;
+def TLD4_UNIFIED_B_2D_U32_F32
+ : NVPTXInst<(outs Int32Regs:$v0, Int32Regs:$v1,
+ Int32Regs:$v2, Int32Regs:$v3),
+ (ins Int64Regs:$t, Float32Regs:$x, Float32Regs:$y),
+ "tld4.b.2d.v4.u32.f32\t\\{$v0, $v1, $v2, $v3\\}, "
+ "[$t, \\{$x, $y\\}];",
+ []>;
+def TLD4_UNIFIED_A_2D_U32_F32
+ : NVPTXInst<(outs Int32Regs:$v0, Int32Regs:$v1,
+ Int32Regs:$v2, Int32Regs:$v3),
+ (ins Int64Regs:$t, Float32Regs:$x, Float32Regs:$y),
+ "tld4.a.2d.v4.u32.f32\t\\{$v0, $v1, $v2, $v3\\}, "
+ "[$t, \\{$x, $y\\}];",
+ []>;
+}
+
+
+
+//=== Surface load instructions
+// .clamp variant
+let IsSuld = 1 in {
+def SULD_1D_I8_CLAMP
+ : NVPTXInst<(outs Int16Regs:$r),
+ (ins Int64Regs:$s, Int32Regs:$x),
+ "suld.b.1d.b8.clamp \\{$r\\}, [$s, \\{$x\\}];",
+ []>;
+def SULD_1D_I16_CLAMP
+ : NVPTXInst<(outs Int16Regs:$r),
+ (ins Int64Regs:$s, Int32Regs:$x),
+ "suld.b.1d.b16.clamp \\{$r\\}, [$s, \\{$x\\}];",
+ []>;
+def SULD_1D_I32_CLAMP
+ : NVPTXInst<(outs Int32Regs:$r),
+ (ins Int64Regs:$s, Int32Regs:$x),
+ "suld.b.1d.b32.clamp \\{$r\\}, [$s, \\{$x\\}];",
+ []>;
+def SULD_1D_I64_CLAMP
+ : NVPTXInst<(outs Int64Regs:$r),
+ (ins Int64Regs:$s, Int32Regs:$x),
+ "suld.b.1d.b64.clamp \\{$r\\}, [$s, \\{$x\\}];",
+ []>;
+
+def SULD_1D_ARRAY_I8_CLAMP
+ : NVPTXInst<(outs Int16Regs:$r),
+ (ins Int64Regs:$s, Int32Regs:$l, Int32Regs:$x),
+ "suld.b.a1d.b8.clamp \\{$r\\}, [$s, \\{$l, $x\\}];",
+ []>;
+def SULD_1D_ARRAY_I16_CLAMP
+ : NVPTXInst<(outs Int16Regs:$r),
+ (ins Int64Regs:$s, Int32Regs:$l, Int32Regs:$x),
+ "suld.b.a1d.b16.clamp \\{$r\\}, [$s, \\{$l, $x\\}];",
+ []>;
+def SULD_1D_ARRAY_I32_CLAMP
+ : NVPTXInst<(outs Int32Regs:$r),
+ (ins Int64Regs:$s, Int32Regs:$l, Int32Regs:$x),
+ "suld.b.a1d.b32.clamp \\{$r\\}, [$s, \\{$l, $x\\}];",
+ []>;
+def SULD_1D_ARRAY_I64_CLAMP
+ : NVPTXInst<(outs Int64Regs:$r),
+ (ins Int64Regs:$s, Int32Regs:$l, Int32Regs:$x),
+ "suld.b.a1d.b64.clamp \\{$r\\}, [$s, \\{$l, $x\\}];",
+ []>;
+
+def SULD_2D_I8_CLAMP
+ : NVPTXInst<(outs Int16Regs:$r),
+ (ins Int64Regs:$s, Int32Regs:$x, Int32Regs:$y),
+ "suld.b.2d.b8.clamp \\{$r\\}, [$s, \\{$x, $y\\}];",
+ []>;
+def SULD_2D_I16_CLAMP
+ : NVPTXInst<(outs Int16Regs:$r),
+ (ins Int64Regs:$s, Int32Regs:$x, Int32Regs:$y),
+ "suld.b.2d.b16.clamp \\{$r\\}, [$s, \\{$x, $y\\}];",
+ []>;
+def SULD_2D_I32_CLAMP
+ : NVPTXInst<(outs Int32Regs:$r),
+ (ins Int64Regs:$s, Int32Regs:$x, Int32Regs:$y),
+ "suld.b.2d.b32.clamp \\{$r\\}, [$s, \\{$x, $y\\}];",
+ []>;
+def SULD_2D_I64_CLAMP
+ : NVPTXInst<(outs Int64Regs:$r),
+ (ins Int64Regs:$s, Int32Regs:$x, Int32Regs:$y),
+ "suld.b.2d.b64.clamp \\{$r\\}, [$s, \\{$x, $y\\}];",
+ []>;
+
+def SULD_2D_ARRAY_I8_CLAMP
+ : NVPTXInst<(outs Int16Regs:$r),
+ (ins Int64Regs:$s, Int32Regs:$l, Int32Regs:$x, Int32Regs:$y),
+ "suld.b.a2d.b8.clamp \\{$r\\}, [$s, \\{$l, $x, $y, $y\\}];",
+ []>;
+def SULD_2D_ARRAY_I16_CLAMP
+ : NVPTXInst<(outs Int16Regs:$r),
+ (ins Int64Regs:$s, Int32Regs:$l, Int32Regs:$x, Int32Regs:$y),
+ "suld.b.a2d.b16.clamp \\{$r\\}, [$s, \\{$l, $x, $y, $y\\}];",
+ []>;
+def SULD_2D_ARRAY_I32_CLAMP
+ : NVPTXInst<(outs Int32Regs:$r),
+ (ins Int64Regs:$s, Int32Regs:$l, Int32Regs:$x, Int32Regs:$y),
+ "suld.b.a2d.b32.clamp \\{$r\\}, [$s, \\{$l, $x, $y, $y\\}];",
+ []>;
+def SULD_2D_ARRAY_I64_CLAMP
+ : NVPTXInst<(outs Int64Regs:$r),
+ (ins Int64Regs:$s, Int32Regs:$l, Int32Regs:$x, Int32Regs:$y),
+ "suld.b.a2d.b64.clamp \\{$r\\}, [$s, \\{$l, $x, $y, $y\\}];",
+ []>;
+
+def SULD_3D_I8_CLAMP
+ : NVPTXInst<(outs Int16Regs:$r),
+ (ins Int64Regs:$s, Int32Regs:$x, Int32Regs:$y, Int32Regs:$z),
+ "suld.b.3d.b8.clamp \\{$r\\}, [$s, \\{$x, $y, $z, $z\\}];",
+ []>;
+def SULD_3D_I16_CLAMP
+ : NVPTXInst<(outs Int16Regs:$r),
+ (ins Int64Regs:$s, Int32Regs:$x, Int32Regs:$y, Int32Regs:$z),
+ "suld.b.3d.b16.clamp \\{$r\\}, [$s, \\{$x, $y, $z, $z\\}];",
+ []>;
+def SULD_3D_I32_CLAMP
+ : NVPTXInst<(outs Int32Regs:$r),
+ (ins Int64Regs:$s, Int32Regs:$x, Int32Regs:$y, Int32Regs:$z),
+ "suld.b.3d.b32.clamp \\{$r\\}, [$s, \\{$x, $y, $z, $z\\}];",
+ []>;
+def SULD_3D_I64_CLAMP
+ : NVPTXInst<(outs Int64Regs:$r),
+ (ins Int64Regs:$s, Int32Regs:$x, Int32Regs:$y, Int32Regs:$z),
+ "suld.b.3d.b64.clamp \\{$r\\}, [$s, \\{$x, $y, $z, $z\\}];",
+ []>;
+}
+
+let IsSuld = 2 in {
+def SULD_1D_V2I8_CLAMP
+ : NVPTXInst<(outs Int16Regs:$r, Int16Regs:$g),
+ (ins Int64Regs:$s, Int32Regs:$x),
+ "suld.b.1d.v2.b8.clamp \\{$r, $g\\}, [$s, \\{$x\\}];",
+ []>;
+def SULD_1D_V2I16_CLAMP
+ : NVPTXInst<(outs Int16Regs:$r, Int16Regs:$g),
+ (ins Int64Regs:$s, Int32Regs:$x),
+ "suld.b.1d.v2.b16.clamp \\{$r, $g\\}, [$s, \\{$x\\}];",
+ []>;
+def SULD_1D_V2I32_CLAMP
+ : NVPTXInst<(outs Int32Regs:$r, Int32Regs:$g),
+ (ins Int64Regs:$s, Int32Regs:$x),
+ "suld.b.1d.v2.b32.clamp \\{$r, $g\\}, [$s, \\{$x\\}];",
+ []>;
+def SULD_1D_V2I64_CLAMP
+ : NVPTXInst<(outs Int64Regs:$r, Int64Regs:$g),
+ (ins Int64Regs:$s, Int32Regs:$x),
+ "suld.b.1d.v2.b64.clamp \\{$r, $g\\}, [$s, \\{$x\\}];",
+ []>;
+
+def SULD_1D_ARRAY_V2I8_CLAMP
+ : NVPTXInst<(outs Int16Regs:$r, Int16Regs:$g),
+ (ins Int64Regs:$s, Int32Regs:$l, Int32Regs:$x),
+ "suld.b.a1d.v2.b8.clamp \\{$r, $g\\}, [$s, \\{$l, $x\\}];",
+ []>;
+def SULD_1D_ARRAY_V2I16_CLAMP
+ : NVPTXInst<(outs Int16Regs:$r, Int16Regs:$g),
+ (ins Int64Regs:$s, Int32Regs:$l, Int32Regs:$x),
+ "suld.b.a1d.v2.b16.clamp \\{$r, $g\\}, [$s, \\{$l, $x\\}];",
+ []>;
+def SULD_1D_ARRAY_V2I32_CLAMP
+ : NVPTXInst<(outs Int32Regs:$r, Int32Regs:$g),
+ (ins Int64Regs:$s, Int32Regs:$l, Int32Regs:$x),
+ "suld.b.a1d.v2.b32.clamp \\{$r, $g\\}, [$s, \\{$l, $x\\}];",
+ []>;
+def SULD_1D_ARRAY_V2I64_CLAMP
+ : NVPTXInst<(outs Int64Regs:$r, Int64Regs:$g),
+ (ins Int64Regs:$s, Int32Regs:$l, Int32Regs:$x),
+ "suld.b.a1d.v2.b64.clamp \\{$r, $g\\}, [$s, \\{$l, $x\\}];",
+ []>;
+
+def SULD_2D_V2I8_CLAMP
+ : NVPTXInst<(outs Int16Regs:$r, Int16Regs:$g),
+ (ins Int64Regs:$s, Int32Regs:$x, Int32Regs:$y),
+ "suld.b.2d.v2.b8.clamp \\{$r, $g\\}, [$s, \\{$x, $y\\}];",
+ []>;
+def SULD_2D_V2I16_CLAMP
+ : NVPTXInst<(outs Int16Regs:$r, Int16Regs:$g),
+ (ins Int64Regs:$s, Int32Regs:$x, Int32Regs:$y),
+ "suld.b.2d.v2.b16.clamp \\{$r, $g\\}, [$s, \\{$x, $y\\}];",
+ []>;
+def SULD_2D_V2I32_CLAMP
+ : NVPTXInst<(outs Int32Regs:$r, Int32Regs:$g),
+ (ins Int64Regs:$s, Int32Regs:$x, Int32Regs:$y),
+ "suld.b.2d.v2.b32.clamp \\{$r, $g\\}, [$s, \\{$x, $y\\}];",
+ []>;
+def SULD_2D_V2I64_CLAMP
+ : NVPTXInst<(outs Int64Regs:$r, Int64Regs:$g),
+ (ins Int64Regs:$s, Int32Regs:$x, Int32Regs:$y),
+ "suld.b.2d.v2.b64.clamp \\{$r, $g\\}, [$s, \\{$x, $y\\}];",
+ []>;
+
+def SULD_2D_ARRAY_V2I8_CLAMP
+ : NVPTXInst<(outs Int16Regs:$r, Int16Regs:$g),
+ (ins Int64Regs:$s, Int32Regs:$l, Int32Regs:$x, Int32Regs:$y),
+ "suld.b.a2d.v2.b8.clamp \\{$r, $g\\}, "
+ "[$s, \\{$l, $x, $y, $y\\}];",
+ []>;
+def SULD_2D_ARRAY_V2I16_CLAMP
+ : NVPTXInst<(outs Int16Regs:$r, Int16Regs:$g),
+ (ins Int64Regs:$s, Int32Regs:$l, Int32Regs:$x, Int32Regs:$y),
+ "suld.b.a2d.v2.b16.clamp \\{$r, $g\\}, "
+ "[$s, \\{$l, $x, $y, $y\\}];",
+ []>;
+def SULD_2D_ARRAY_V2I32_CLAMP
+ : NVPTXInst<(outs Int32Regs:$r, Int32Regs:$g),
+ (ins Int64Regs:$s, Int32Regs:$l, Int32Regs:$x, Int32Regs:$y),
+ "suld.b.a2d.v2.b32.clamp \\{$r, $g\\}, "
+ "[$s, \\{$l, $x, $y, $y\\}];",
+ []>;
+def SULD_2D_ARRAY_V2I64_CLAMP
+ : NVPTXInst<(outs Int64Regs:$r, Int64Regs:$g),
+ (ins Int64Regs:$s, Int32Regs:$l, Int32Regs:$x, Int32Regs:$y),
+ "suld.b.a2d.v2.b64.clamp \\{$r, $g\\}, "
+ "[$s, \\{$l, $x, $y, $y\\}];",
+ []>;
+
+def SULD_3D_V2I8_CLAMP
+ : NVPTXInst<(outs Int16Regs:$r, Int16Regs:$g),
+ (ins Int64Regs:$s, Int32Regs:$x, Int32Regs:$y, Int32Regs:$z),
+ "suld.b.3d.v2.b8.clamp \\{$r, $g\\}, [$s, \\{$x, $y, $z, $z\\}];",
+ []>;
+def SULD_3D_V2I16_CLAMP
+ : NVPTXInst<(outs Int16Regs:$r, Int16Regs:$g),
+ (ins Int64Regs:$s, Int32Regs:$x, Int32Regs:$y, Int32Regs:$z),
+ "suld.b.3d.v2.b16.clamp \\{$r, $g\\}, [$s, \\{$x, $y, $z, $z\\}];",
+ []>;
+def SULD_3D_V2I32_CLAMP
+ : NVPTXInst<(outs Int32Regs:$r, Int32Regs:$g),
+ (ins Int64Regs:$s, Int32Regs:$x, Int32Regs:$y, Int32Regs:$z),
+ "suld.b.3d.v2.b32.clamp \\{$r, $g\\}, [$s, \\{$x, $y, $z, $z\\}];",
+ []>;
+def SULD_3D_V2I64_CLAMP
+ : NVPTXInst<(outs Int64Regs:$r, Int64Regs:$g),
+ (ins Int64Regs:$s, Int32Regs:$x, Int32Regs:$y, Int32Regs:$z),
+ "suld.b.3d.v2.b64.clamp \\{$r, $g\\}, [$s, \\{$x, $y, $z, $z\\}];",
+ []>;
+}
+
+let IsSuld = 3 in {
+def SULD_1D_V4I8_CLAMP
+ : NVPTXInst<(outs Int16Regs:$r, Int16Regs:$g, Int16Regs:$b, Int16Regs:$a),
+ (ins Int64Regs:$s, Int32Regs:$x),
+ "suld.b.1d.v4.b8.clamp \\{$r, $g, $b, $a\\}, [$s, \\{$x\\}];",
+ []>;
+def SULD_1D_V4I16_CLAMP
+ : NVPTXInst<(outs Int16Regs:$r, Int16Regs:$g, Int16Regs:$b, Int16Regs:$a),
+ (ins Int64Regs:$s, Int32Regs:$x),
+ "suld.b.1d.v4.b16.clamp \\{$r, $g, $b, $a\\}, [$s, \\{$x\\}];",
+ []>;
+def SULD_1D_V4I32_CLAMP
+ : NVPTXInst<(outs Int32Regs:$r, Int32Regs:$g, Int32Regs:$b, Int32Regs:$a),
+ (ins Int64Regs:$s, Int32Regs:$x),
+ "suld.b.1d.v4.b32.clamp \\{$r, $g, $b, $a\\}, [$s, \\{$x\\}];",
+ []>;
+
+def SULD_1D_ARRAY_V4I8_CLAMP
+ : NVPTXInst<(outs Int16Regs:$r, Int16Regs:$g, Int16Regs:$b, Int16Regs:$a),
+ (ins Int64Regs:$s, Int32Regs:$l, Int32Regs:$x),
+ "suld.b.a1d.v4.b8.clamp \\{$r, $g, $b, $a\\}, "
+ "[$s, \\{$l, $x\\}];",
+ []>;
+def SULD_1D_ARRAY_V4I16_CLAMP
+ : NVPTXInst<(outs Int16Regs:$r, Int16Regs:$g, Int16Regs:$b, Int16Regs:$a),
+ (ins Int64Regs:$s, Int32Regs:$l, Int32Regs:$x),
+ "suld.b.a1d.v4.b16.clamp \\{$r, $g, $b, $a\\}, "
+ "[$s, \\{$l, $x\\}];",
+ []>;
+def SULD_1D_ARRAY_V4I32_CLAMP
+ : NVPTXInst<(outs Int32Regs:$r, Int32Regs:$g, Int32Regs:$b, Int32Regs:$a),
+ (ins Int64Regs:$s, Int32Regs:$l, Int32Regs:$x),
+ "suld.b.a1d.v4.b32.clamp \\{$r, $g, $b, $a\\}, "
+ "[$s, \\{$l, $x\\}];",
+ []>;
+
+def SULD_2D_V4I8_CLAMP
+ : NVPTXInst<(outs Int16Regs:$r, Int16Regs:$g, Int16Regs:$b, Int16Regs:$a),
+ (ins Int64Regs:$s, Int32Regs:$x, Int32Regs:$y),
+ "suld.b.2d.v4.b8.clamp \\{$r, $g, $b, $a\\}, [$s, \\{$x, $y\\}];",
+ []>;
+def SULD_2D_V4I16_CLAMP
+ : NVPTXInst<(outs Int16Regs:$r, Int16Regs:$g, Int16Regs:$b, Int16Regs:$a),
+ (ins Int64Regs:$s, Int32Regs:$x, Int32Regs:$y),
+ "suld.b.2d.v4.b16.clamp \\{$r, $g, $b, $a\\}, [$s, \\{$x, $y\\}];",
+ []>;
+def SULD_2D_V4I32_CLAMP
+ : NVPTXInst<(outs Int32Regs:$r, Int32Regs:$g, Int32Regs:$b, Int32Regs:$a),
+ (ins Int64Regs:$s, Int32Regs:$x, Int32Regs:$y),
+ "suld.b.2d.v4.b32.clamp \\{$r, $g, $b, $a\\}, [$s, \\{$x, $y\\}];",
+ []>;
+
+def SULD_2D_ARRAY_V4I8_CLAMP
+ : NVPTXInst<(outs Int16Regs:$r, Int16Regs:$g, Int16Regs:$b, Int16Regs:$a),
+ (ins Int64Regs:$s, Int32Regs:$l, Int32Regs:$x, Int32Regs:$y),
+ "suld.b.a2d.v4.b8.clamp \\{$r, $g, $b, $a\\}, "
+ "[$s, \\{$l, $x, $y, $y\\}];",
+ []>;
+def SULD_2D_ARRAY_V4I16_CLAMP
+ : NVPTXInst<(outs Int16Regs:$r, Int16Regs:$g, Int16Regs:$b, Int16Regs:$a),
+ (ins Int64Regs:$s, Int32Regs:$l, Int32Regs:$x, Int32Regs:$y),
+ "suld.b.a2d.v4.b16.clamp \\{$r, $g, $b, $a\\}, "
+ "[$s, \\{$l, $x, $y, $y\\}];",
+ []>;
+def SULD_2D_ARRAY_V4I32_CLAMP
+ : NVPTXInst<(outs Int32Regs:$r, Int32Regs:$g, Int32Regs:$b, Int32Regs:$a),
+ (ins Int64Regs:$s, Int32Regs:$l, Int32Regs:$x, Int32Regs:$y),
+ "suld.b.a2d.v4.b32.clamp \\{$r, $g, $b, $a\\}, "
+ "[$s, \\{$l, $x, $y, $y\\}];",
+ []>;
+
+
+def SULD_3D_V4I8_CLAMP
+ : NVPTXInst<(outs Int16Regs:$r, Int16Regs:$g, Int16Regs:$b, Int16Regs:$a),
+ (ins Int64Regs:$s, Int32Regs:$x, Int32Regs:$y, Int32Regs:$z),
+ "suld.b.3d.v4.b8.clamp \\{$r, $g, $b, $a\\}, "
+ "[$s, \\{$x, $y, $z, $z\\}];",
+ []>;
+def SULD_3D_V4I16_CLAMP
+ : NVPTXInst<(outs Int16Regs:$r, Int16Regs:$g, Int16Regs:$b, Int16Regs:$a),
+ (ins Int64Regs:$s, Int32Regs:$x, Int32Regs:$y, Int32Regs:$z),
+ "suld.b.3d.v4.b16.clamp \\{$r, $g, $b, $a\\}, "
+ "[$s, \\{$x, $y, $z, $z\\}];",
+ []>;
+def SULD_3D_V4I32_CLAMP
+ : NVPTXInst<(outs Int32Regs:$r, Int32Regs:$g, Int32Regs:$b, Int32Regs:$a),
+ (ins Int64Regs:$s, Int32Regs:$x, Int32Regs:$y, Int32Regs:$z),
+ "suld.b.3d.v4.b32.clamp \\{$r, $g, $b, $a\\}, "
+ "[$s, \\{$x, $y, $z, $z\\}];",
+ []>;
+}
+
+
+// .trap variant
+let IsSuld = 1 in {
+def SULD_1D_I8_TRAP
+ : NVPTXInst<(outs Int16Regs:$r),
+ (ins Int64Regs:$s, Int32Regs:$x),
+ "suld.b.1d.b8.trap \\{$r\\}, [$s, \\{$x\\}];",
+ []>;
+def SULD_1D_I16_TRAP
+ : NVPTXInst<(outs Int16Regs:$r),
+ (ins Int64Regs:$s, Int32Regs:$x),
+ "suld.b.1d.b16.trap \\{$r\\}, [$s, \\{$x\\}];",
+ []>;
+def SULD_1D_I32_TRAP
+ : NVPTXInst<(outs Int32Regs:$r),
+ (ins Int64Regs:$s, Int32Regs:$x),
+ "suld.b.1d.b32.trap \\{$r\\}, [$s, \\{$x\\}];",
+ []>;
+def SULD_1D_I64_TRAP
+ : NVPTXInst<(outs Int64Regs:$r),
+ (ins Int64Regs:$s, Int32Regs:$x),
+ "suld.b.1d.b64.trap \\{$r\\}, [$s, \\{$x\\}];",
+ []>;
+
+def SULD_1D_ARRAY_I8_TRAP
+ : NVPTXInst<(outs Int16Regs:$r),
+ (ins Int64Regs:$s, Int32Regs:$l, Int32Regs:$x),
+ "suld.b.a1d.b8.trap \\{$r\\}, [$s, \\{$l, $x\\}];",
+ []>;
+def SULD_1D_ARRAY_I16_TRAP
+ : NVPTXInst<(outs Int16Regs:$r),
+ (ins Int64Regs:$s, Int32Regs:$l, Int32Regs:$x),
+ "suld.b.a1d.b16.trap \\{$r\\}, [$s, \\{$l, $x\\}];",
+ []>;
+def SULD_1D_ARRAY_I32_TRAP
+ : NVPTXInst<(outs Int32Regs:$r),
+ (ins Int64Regs:$s, Int32Regs:$l, Int32Regs:$x),
+ "suld.b.a1d.b32.trap \\{$r\\}, [$s, \\{$l, $x\\}];",
+ []>;
+def SULD_1D_ARRAY_I64_TRAP
+ : NVPTXInst<(outs Int64Regs:$r),
+ (ins Int64Regs:$s, Int32Regs:$l, Int32Regs:$x),
+ "suld.b.a1d.b64.trap \\{$r\\}, [$s, \\{$l, $x\\}];",
+ []>;
+
+def SULD_2D_I8_TRAP
+ : NVPTXInst<(outs Int16Regs:$r),
+ (ins Int64Regs:$s, Int32Regs:$x, Int32Regs:$y),
+ "suld.b.2d.b8.trap \\{$r\\}, [$s, \\{$x, $y\\}];",
+ []>;
+def SULD_2D_I16_TRAP
+ : NVPTXInst<(outs Int16Regs:$r),
+ (ins Int64Regs:$s, Int32Regs:$x, Int32Regs:$y),
+ "suld.b.2d.b16.trap \\{$r\\}, [$s, \\{$x, $y\\}];",
+ []>;
+def SULD_2D_I32_TRAP
+ : NVPTXInst<(outs Int32Regs:$r),
+ (ins Int64Regs:$s, Int32Regs:$x, Int32Regs:$y),
+ "suld.b.2d.b32.trap \\{$r\\}, [$s, \\{$x, $y\\}];",
+ []>;
+def SULD_2D_I64_TRAP
+ : NVPTXInst<(outs Int64Regs:$r),
+ (ins Int64Regs:$s, Int32Regs:$x, Int32Regs:$y),
+ "suld.b.2d.b64.trap \\{$r\\}, [$s, \\{$x, $y\\}];",
+ []>;
+
+def SULD_2D_ARRAY_I8_TRAP
+ : NVPTXInst<(outs Int16Regs:$r),
+ (ins Int64Regs:$s, Int32Regs:$l, Int32Regs:$x, Int32Regs:$y),
+ "suld.b.a2d.b8.trap \\{$r\\}, [$s, \\{$l, $x, $y, $y\\}];",
+ []>;
+def SULD_2D_ARRAY_I16_TRAP
+ : NVPTXInst<(outs Int16Regs:$r),
+ (ins Int64Regs:$s, Int32Regs:$l, Int32Regs:$x, Int32Regs:$y),
+ "suld.b.a2d.b16.trap \\{$r\\}, [$s, \\{$l, $x, $y, $y\\}];",
+ []>;
+def SULD_2D_ARRAY_I32_TRAP
+ : NVPTXInst<(outs Int32Regs:$r),
+ (ins Int64Regs:$s, Int32Regs:$l, Int32Regs:$x, Int32Regs:$y),
+ "suld.b.a2d.b32.trap \\{$r\\}, [$s, \\{$l, $x, $y, $y\\}];",
+ []>;
+def SULD_2D_ARRAY_I64_TRAP
+ : NVPTXInst<(outs Int64Regs:$r),
+ (ins Int64Regs:$s, Int32Regs:$l, Int32Regs:$x, Int32Regs:$y),
+ "suld.b.a2d.b64.trap \\{$r\\}, [$s, \\{$l, $x, $y, $y\\}];",
+ []>;
+
+def SULD_3D_I8_TRAP
+ : NVPTXInst<(outs Int16Regs:$r),
+ (ins Int64Regs:$s, Int32Regs:$x, Int32Regs:$y, Int32Regs:$z),
+ "suld.b.3d.b8.trap \\{$r\\}, [$s, \\{$x, $y, $z, $z\\}];",
+ []>;
+def SULD_3D_I16_TRAP
+ : NVPTXInst<(outs Int16Regs:$r),
+ (ins Int64Regs:$s, Int32Regs:$x, Int32Regs:$y, Int32Regs:$z),
+ "suld.b.3d.b16.trap \\{$r\\}, [$s, \\{$x, $y, $z, $z\\}];",
+ []>;
+def SULD_3D_I32_TRAP
+ : NVPTXInst<(outs Int32Regs:$r),
+ (ins Int64Regs:$s, Int32Regs:$x, Int32Regs:$y, Int32Regs:$z),
+ "suld.b.3d.b32.trap \\{$r\\}, [$s, \\{$x, $y, $z, $z\\}];",
+ []>;
+def SULD_3D_I64_TRAP
+ : NVPTXInst<(outs Int64Regs:$r),
+ (ins Int64Regs:$s, Int32Regs:$x, Int32Regs:$y, Int32Regs:$z),
+ "suld.b.3d.b64.trap \\{$r\\}, [$s, \\{$x, $y, $z, $z\\}];",
+ []>;
+}
+
+let IsSuld = 2 in {
+def SULD_1D_V2I8_TRAP
+ : NVPTXInst<(outs Int16Regs:$r, Int16Regs:$g),
+ (ins Int64Regs:$s, Int32Regs:$x),
+ "suld.b.1d.v2.b8.trap \\{$r, $g\\}, [$s, \\{$x\\}];",
+ []>;
+def SULD_1D_V2I16_TRAP
+ : NVPTXInst<(outs Int16Regs:$r, Int16Regs:$g),
+ (ins Int64Regs:$s, Int32Regs:$x),
+ "suld.b.1d.v2.b16.trap \\{$r, $g\\}, [$s, \\{$x\\}];",
+ []>;
+def SULD_1D_V2I32_TRAP
+ : NVPTXInst<(outs Int32Regs:$r, Int32Regs:$g),
+ (ins Int64Regs:$s, Int32Regs:$x),
+ "suld.b.1d.v2.b32.trap \\{$r, $g\\}, [$s, \\{$x\\}];",
+ []>;
+def SULD_1D_V2I64_TRAP
+ : NVPTXInst<(outs Int64Regs:$r, Int64Regs:$g),
+ (ins Int64Regs:$s, Int32Regs:$x),
+ "suld.b.1d.v2.b64.trap \\{$r, $g\\}, [$s, \\{$x\\}];",
+ []>;
+
+def SULD_1D_ARRAY_V2I8_TRAP
+ : NVPTXInst<(outs Int16Regs:$r, Int16Regs:$g),
+ (ins Int64Regs:$s, Int32Regs:$l, Int32Regs:$x),
+ "suld.b.a1d.v2.b8.trap \\{$r, $g\\}, [$s, \\{$l, $x\\}];",
+ []>;
+def SULD_1D_ARRAY_V2I16_TRAP
+ : NVPTXInst<(outs Int16Regs:$r, Int16Regs:$g),
+ (ins Int64Regs:$s, Int32Regs:$l, Int32Regs:$x),
+ "suld.b.a1d.v2.b16.trap \\{$r, $g\\}, [$s, \\{$l, $x\\}];",
+ []>;
+def SULD_1D_ARRAY_V2I32_TRAP
+ : NVPTXInst<(outs Int32Regs:$r, Int32Regs:$g),
+ (ins Int64Regs:$s, Int32Regs:$l, Int32Regs:$x),
+ "suld.b.a1d.v2.b32.trap \\{$r, $g\\}, [$s, \\{$l, $x\\}];",
+ []>;
+def SULD_1D_ARRAY_V2I64_TRAP
+ : NVPTXInst<(outs Int64Regs:$r, Int64Regs:$g),
+ (ins Int64Regs:$s, Int32Regs:$l, Int32Regs:$x),
+ "suld.b.a1d.v2.b64.trap \\{$r, $g\\}, [$s, \\{$l, $x\\}];",
+ []>;
+
+def SULD_2D_V2I8_TRAP
+ : NVPTXInst<(outs Int16Regs:$r, Int16Regs:$g),
+ (ins Int64Regs:$s, Int32Regs:$x, Int32Regs:$y),
+ "suld.b.2d.v2.b8.trap \\{$r, $g\\}, [$s, \\{$x, $y\\}];",
+ []>;
+def SULD_2D_V2I16_TRAP
+ : NVPTXInst<(outs Int16Regs:$r, Int16Regs:$g),
+ (ins Int64Regs:$s, Int32Regs:$x, Int32Regs:$y),
+ "suld.b.2d.v2.b16.trap \\{$r, $g\\}, [$s, \\{$x, $y\\}];",
+ []>;
+def SULD_2D_V2I32_TRAP
+ : NVPTXInst<(outs Int32Regs:$r, Int32Regs:$g),
+ (ins Int64Regs:$s, Int32Regs:$x, Int32Regs:$y),
+ "suld.b.2d.v2.b32.trap \\{$r, $g\\}, [$s, \\{$x, $y\\}];",
+ []>;
+def SULD_2D_V2I64_TRAP
+ : NVPTXInst<(outs Int64Regs:$r, Int64Regs:$g),
+ (ins Int64Regs:$s, Int32Regs:$x, Int32Regs:$y),
+ "suld.b.2d.v2.b64.trap \\{$r, $g\\}, [$s, \\{$x, $y\\}];",
+ []>;
+
+def SULD_2D_ARRAY_V2I8_TRAP
+ : NVPTXInst<(outs Int16Regs:$r, Int16Regs:$g),
+ (ins Int64Regs:$s, Int32Regs:$l, Int32Regs:$x, Int32Regs:$y),
+ "suld.b.a2d.v2.b8.trap \\{$r, $g\\}, "
+ "[$s, \\{$l, $x, $y, $y\\}];",
+ []>;
+def SULD_2D_ARRAY_V2I16_TRAP
+ : NVPTXInst<(outs Int16Regs:$r, Int16Regs:$g),
+ (ins Int64Regs:$s, Int32Regs:$l, Int32Regs:$x, Int32Regs:$y),
+ "suld.b.a2d.v2.b16.trap \\{$r, $g\\}, "
+ "[$s, \\{$l, $x, $y, $y\\}];",
+ []>;
+def SULD_2D_ARRAY_V2I32_TRAP
+ : NVPTXInst<(outs Int32Regs:$r, Int32Regs:$g),
+ (ins Int64Regs:$s, Int32Regs:$l, Int32Regs:$x, Int32Regs:$y),
+ "suld.b.a2d.v2.b32.trap \\{$r, $g\\}, "
+ "[$s, \\{$l, $x, $y, $y\\}];",
+ []>;
+def SULD_2D_ARRAY_V2I64_TRAP
+ : NVPTXInst<(outs Int64Regs:$r, Int64Regs:$g),
+ (ins Int64Regs:$s, Int32Regs:$l, Int32Regs:$x, Int32Regs:$y),
+ "suld.b.a2d.v2.b64.trap \\{$r, $g\\}, "
+ "[$s, \\{$l, $x, $y, $y\\}];",
+ []>;
+
+def SULD_3D_V2I8_TRAP
+ : NVPTXInst<(outs Int16Regs:$r, Int16Regs:$g),
+ (ins Int64Regs:$s, Int32Regs:$x, Int32Regs:$y, Int32Regs:$z),
+ "suld.b.3d.v2.b8.trap \\{$r, $g\\}, [$s, \\{$x, $y, $z, $z\\}];",
+ []>;
+def SULD_3D_V2I16_TRAP
+ : NVPTXInst<(outs Int16Regs:$r, Int16Regs:$g),
+ (ins Int64Regs:$s, Int32Regs:$x, Int32Regs:$y, Int32Regs:$z),
+ "suld.b.3d.v2.b16.trap \\{$r, $g\\}, [$s, \\{$x, $y, $z, $z\\}];",
+ []>;
+def SULD_3D_V2I32_TRAP
+ : NVPTXInst<(outs Int32Regs:$r, Int32Regs:$g),
+ (ins Int64Regs:$s, Int32Regs:$x, Int32Regs:$y, Int32Regs:$z),
+ "suld.b.3d.v2.b32.trap \\{$r, $g\\}, [$s, \\{$x, $y, $z, $z\\}];",
+ []>;
+def SULD_3D_V2I64_TRAP
+ : NVPTXInst<(outs Int64Regs:$r, Int64Regs:$g),
+ (ins Int64Regs:$s, Int32Regs:$x, Int32Regs:$y, Int32Regs:$z),
+ "suld.b.3d.v2.b64.trap \\{$r, $g\\}, [$s, \\{$x, $y, $z, $z\\}];",
+ []>;
+}
+
+let IsSuld = 3 in {
+def SULD_1D_V4I8_TRAP
+ : NVPTXInst<(outs Int16Regs:$r, Int16Regs:$g, Int16Regs:$b, Int16Regs:$a),
+ (ins Int64Regs:$s, Int32Regs:$x),
+ "suld.b.1d.v4.b8.trap \\{$r, $g, $b, $a\\}, [$s, \\{$x\\}];",
+ []>;
+def SULD_1D_V4I16_TRAP
+ : NVPTXInst<(outs Int16Regs:$r, Int16Regs:$g, Int16Regs:$b, Int16Regs:$a),
+ (ins Int64Regs:$s, Int32Regs:$x),
+ "suld.b.1d.v4.b16.trap \\{$r, $g, $b, $a\\}, [$s, \\{$x\\}];",
+ []>;
+def SULD_1D_V4I32_TRAP
+ : NVPTXInst<(outs Int32Regs:$r, Int32Regs:$g, Int32Regs:$b, Int32Regs:$a),
+ (ins Int64Regs:$s, Int32Regs:$x),
+ "suld.b.1d.v4.b32.trap \\{$r, $g, $b, $a\\}, [$s, \\{$x\\}];",
+ []>;
+
+def SULD_1D_ARRAY_V4I8_TRAP
+ : NVPTXInst<(outs Int16Regs:$r, Int16Regs:$g, Int16Regs:$b, Int16Regs:$a),
+ (ins Int64Regs:$s, Int32Regs:$l, Int32Regs:$x),
+ "suld.b.a1d.v4.b8.trap \\{$r, $g, $b, $a\\}, "
+ "[$s, \\{$l, $x\\}];",
+ []>;
+def SULD_1D_ARRAY_V4I16_TRAP
+ : NVPTXInst<(outs Int16Regs:$r, Int16Regs:$g, Int16Regs:$b, Int16Regs:$a),
+ (ins Int64Regs:$s, Int32Regs:$l, Int32Regs:$x),
+ "suld.b.a1d.v4.b16.trap \\{$r, $g, $b, $a\\}, "
+ "[$s, \\{$l, $x\\}];",
+ []>;
+def SULD_1D_ARRAY_V4I32_TRAP
+ : NVPTXInst<(outs Int32Regs:$r, Int32Regs:$g, Int32Regs:$b, Int32Regs:$a),
+ (ins Int64Regs:$s, Int32Regs:$l, Int32Regs:$x),
+ "suld.b.a1d.v4.b32.trap \\{$r, $g, $b, $a\\}, "
+ "[$s, \\{$l, $x\\}];",
+ []>;
+
+def SULD_2D_V4I8_TRAP
+ : NVPTXInst<(outs Int16Regs:$r, Int16Regs:$g, Int16Regs:$b, Int16Regs:$a),
+ (ins Int64Regs:$s, Int32Regs:$x, Int32Regs:$y),
+ "suld.b.2d.v4.b8.trap \\{$r, $g, $b, $a\\}, [$s, \\{$x, $y\\}];",
+ []>;
+def SULD_2D_V4I16_TRAP
+ : NVPTXInst<(outs Int16Regs:$r, Int16Regs:$g, Int16Regs:$b, Int16Regs:$a),
+ (ins Int64Regs:$s, Int32Regs:$x, Int32Regs:$y),
+ "suld.b.2d.v4.b16.trap \\{$r, $g, $b, $a\\}, [$s, \\{$x, $y\\}];",
+ []>;
+def SULD_2D_V4I32_TRAP
+ : NVPTXInst<(outs Int32Regs:$r, Int32Regs:$g, Int32Regs:$b, Int32Regs:$a),
+ (ins Int64Regs:$s, Int32Regs:$x, Int32Regs:$y),
+ "suld.b.2d.v4.b32.trap \\{$r, $g, $b, $a\\}, [$s, \\{$x, $y\\}];",
+ []>;
+
+def SULD_2D_ARRAY_V4I8_TRAP
+ : NVPTXInst<(outs Int16Regs:$r, Int16Regs:$g, Int16Regs:$b, Int16Regs:$a),
+ (ins Int64Regs:$s, Int32Regs:$l, Int32Regs:$x, Int32Regs:$y),
+ "suld.b.a2d.v4.b8.trap \\{$r, $g, $b, $a\\}, "
+ "[$s, \\{$l, $x, $y, $y\\}];",
+ []>;
+def SULD_2D_ARRAY_V4I16_TRAP
+ : NVPTXInst<(outs Int16Regs:$r, Int16Regs:$g, Int16Regs:$b, Int16Regs:$a),
+ (ins Int64Regs:$s, Int32Regs:$l, Int32Regs:$x, Int32Regs:$y),
+ "suld.b.a2d.v4.b16.trap \\{$r, $g, $b, $a\\}, "
+ "[$s, \\{$l, $x, $y, $y\\}];",
+ []>;
+def SULD_2D_ARRAY_V4I32_TRAP
+ : NVPTXInst<(outs Int32Regs:$r, Int32Regs:$g, Int32Regs:$b, Int32Regs:$a),
+ (ins Int64Regs:$s, Int32Regs:$l, Int32Regs:$x, Int32Regs:$y),
+ "suld.b.a2d.v4.b32.trap \\{$r, $g, $b, $a\\}, "
+ "[$s, \\{$l, $x, $y, $y\\}];",
+ []>;
+
+
+def SULD_3D_V4I8_TRAP
+ : NVPTXInst<(outs Int16Regs:$r, Int16Regs:$g, Int16Regs:$b, Int16Regs:$a),
+ (ins Int64Regs:$s, Int32Regs:$x, Int32Regs:$y, Int32Regs:$z),
+ "suld.b.3d.v4.b8.trap \\{$r, $g, $b, $a\\}, "
+ "[$s, \\{$x, $y, $z, $z\\}];",
+ []>;
+def SULD_3D_V4I16_TRAP
+ : NVPTXInst<(outs Int16Regs:$r, Int16Regs:$g, Int16Regs:$b, Int16Regs:$a),
+ (ins Int64Regs:$s, Int32Regs:$x, Int32Regs:$y, Int32Regs:$z),
+ "suld.b.3d.v4.b16.trap \\{$r, $g, $b, $a\\}, "
+ "[$s, \\{$x, $y, $z, $z\\}];",
+ []>;
+def SULD_3D_V4I32_TRAP
+ : NVPTXInst<(outs Int32Regs:$r, Int32Regs:$g, Int32Regs:$b, Int32Regs:$a),
+ (ins Int64Regs:$s, Int32Regs:$x, Int32Regs:$y, Int32Regs:$z),
+ "suld.b.3d.v4.b32.trap \\{$r, $g, $b, $a\\}, "
+ "[$s, \\{$x, $y, $z, $z\\}];",
+ []>;
+}
+
+// .zero variant
+let IsSuld = 1 in {
+def SULD_1D_I8_ZERO
+ : NVPTXInst<(outs Int16Regs:$r),
+ (ins Int64Regs:$s, Int32Regs:$x),
+ "suld.b.1d.b8.zero \\{$r\\}, [$s, \\{$x\\}];",
+ []>;
+def SULD_1D_I16_ZERO
+ : NVPTXInst<(outs Int16Regs:$r),
+ (ins Int64Regs:$s, Int32Regs:$x),
+ "suld.b.1d.b16.zero \\{$r\\}, [$s, \\{$x\\}];",
+ []>;
+def SULD_1D_I32_ZERO
+ : NVPTXInst<(outs Int32Regs:$r),
+ (ins Int64Regs:$s, Int32Regs:$x),
+ "suld.b.1d.b32.zero \\{$r\\}, [$s, \\{$x\\}];",
+ []>;
+def SULD_1D_I64_ZERO
+ : NVPTXInst<(outs Int64Regs:$r),
+ (ins Int64Regs:$s, Int32Regs:$x),
+ "suld.b.1d.b64.zero \\{$r\\}, [$s, \\{$x\\}];",
+ []>;
+
+def SULD_1D_ARRAY_I8_ZERO
+ : NVPTXInst<(outs Int16Regs:$r),
+ (ins Int64Regs:$s, Int32Regs:$l, Int32Regs:$x),
+ "suld.b.a1d.b8.zero \\{$r\\}, [$s, \\{$l, $x\\}];",
+ []>;
+def SULD_1D_ARRAY_I16_ZERO
+ : NVPTXInst<(outs Int16Regs:$r),
+ (ins Int64Regs:$s, Int32Regs:$l, Int32Regs:$x),
+ "suld.b.a1d.b16.zero \\{$r\\}, [$s, \\{$l, $x\\}];",
+ []>;
+def SULD_1D_ARRAY_I32_ZERO
+ : NVPTXInst<(outs Int32Regs:$r),
+ (ins Int64Regs:$s, Int32Regs:$l, Int32Regs:$x),
+ "suld.b.a1d.b32.zero \\{$r\\}, [$s, \\{$l, $x\\}];",
+ []>;
+def SULD_1D_ARRAY_I64_ZERO
+ : NVPTXInst<(outs Int64Regs:$r),
+ (ins Int64Regs:$s, Int32Regs:$l, Int32Regs:$x),
+ "suld.b.a1d.b64.zero \\{$r\\}, [$s, \\{$l, $x\\}];",
+ []>;
+
+def SULD_2D_I8_ZERO
+ : NVPTXInst<(outs Int16Regs:$r),
+ (ins Int64Regs:$s, Int32Regs:$x, Int32Regs:$y),
+ "suld.b.2d.b8.zero \\{$r\\}, [$s, \\{$x, $y\\}];",
+ []>;
+def SULD_2D_I16_ZERO
+ : NVPTXInst<(outs Int16Regs:$r),
+ (ins Int64Regs:$s, Int32Regs:$x, Int32Regs:$y),
+ "suld.b.2d.b16.zero \\{$r\\}, [$s, \\{$x, $y\\}];",
+ []>;
+def SULD_2D_I32_ZERO
+ : NVPTXInst<(outs Int32Regs:$r),
+ (ins Int64Regs:$s, Int32Regs:$x, Int32Regs:$y),
+ "suld.b.2d.b32.zero \\{$r\\}, [$s, \\{$x, $y\\}];",
+ []>;
+def SULD_2D_I64_ZERO
+ : NVPTXInst<(outs Int64Regs:$r),
+ (ins Int64Regs:$s, Int32Regs:$x, Int32Regs:$y),
+ "suld.b.2d.b64.zero \\{$r\\}, [$s, \\{$x, $y\\}];",
+ []>;
+
+def SULD_2D_ARRAY_I8_ZERO
+ : NVPTXInst<(outs Int16Regs:$r),
+ (ins Int64Regs:$s, Int32Regs:$l, Int32Regs:$x, Int32Regs:$y),
+ "suld.b.a2d.b8.zero \\{$r\\}, [$s, \\{$l, $x, $y, $y\\}];",
+ []>;
+def SULD_2D_ARRAY_I16_ZERO
+ : NVPTXInst<(outs Int16Regs:$r),
+ (ins Int64Regs:$s, Int32Regs:$l, Int32Regs:$x, Int32Regs:$y),
+ "suld.b.a2d.b16.zero \\{$r\\}, [$s, \\{$l, $x, $y, $y\\}];",
+ []>;
+def SULD_2D_ARRAY_I32_ZERO
+ : NVPTXInst<(outs Int32Regs:$r),
+ (ins Int64Regs:$s, Int32Regs:$l, Int32Regs:$x, Int32Regs:$y),
+ "suld.b.a2d.b32.zero \\{$r\\}, [$s, \\{$l, $x, $y, $y\\}];",
+ []>;
+def SULD_2D_ARRAY_I64_ZERO
+ : NVPTXInst<(outs Int64Regs:$r),
+ (ins Int64Regs:$s, Int32Regs:$l, Int32Regs:$x, Int32Regs:$y),
+ "suld.b.a2d.b64.zero \\{$r\\}, [$s, \\{$l, $x, $y, $y\\}];",
+ []>;
+
+def SULD_3D_I8_ZERO
+ : NVPTXInst<(outs Int16Regs:$r),
+ (ins Int64Regs:$s, Int32Regs:$x, Int32Regs:$y, Int32Regs:$z),
+ "suld.b.3d.b8.zero \\{$r\\}, [$s, \\{$x, $y, $z, $z\\}];",
+ []>;
+def SULD_3D_I16_ZERO
+ : NVPTXInst<(outs Int16Regs:$r),
+ (ins Int64Regs:$s, Int32Regs:$x, Int32Regs:$y, Int32Regs:$z),
+ "suld.b.3d.b16.zero \\{$r\\}, [$s, \\{$x, $y, $z, $z\\}];",
+ []>;
+def SULD_3D_I32_ZERO
+ : NVPTXInst<(outs Int32Regs:$r),
+ (ins Int64Regs:$s, Int32Regs:$x, Int32Regs:$y, Int32Regs:$z),
+ "suld.b.3d.b32.zero \\{$r\\}, [$s, \\{$x, $y, $z, $z\\}];",
+ []>;
+def SULD_3D_I64_ZERO
+ : NVPTXInst<(outs Int64Regs:$r),
+ (ins Int64Regs:$s, Int32Regs:$x, Int32Regs:$y, Int32Regs:$z),
+ "suld.b.3d.b64.zero \\{$r\\}, [$s, \\{$x, $y, $z, $z\\}];",
+ []>;
+}
+
+let IsSuld = 2 in {
+def SULD_1D_V2I8_ZERO
+ : NVPTXInst<(outs Int16Regs:$r, Int16Regs:$g),
+ (ins Int64Regs:$s, Int32Regs:$x),
+ "suld.b.1d.v2.b8.zero \\{$r, $g\\}, [$s, \\{$x\\}];",
+ []>;
+def SULD_1D_V2I16_ZERO
+ : NVPTXInst<(outs Int16Regs:$r, Int16Regs:$g),
+ (ins Int64Regs:$s, Int32Regs:$x),
+ "suld.b.1d.v2.b16.zero \\{$r, $g\\}, [$s, \\{$x\\}];",
+ []>;
+def SULD_1D_V2I32_ZERO
+ : NVPTXInst<(outs Int32Regs:$r, Int32Regs:$g),
+ (ins Int64Regs:$s, Int32Regs:$x),
+ "suld.b.1d.v2.b32.zero \\{$r, $g\\}, [$s, \\{$x\\}];",
+ []>;
+def SULD_1D_V2I64_ZERO
+ : NVPTXInst<(outs Int64Regs:$r, Int64Regs:$g),
+ (ins Int64Regs:$s, Int32Regs:$x),
+ "suld.b.1d.v2.b64.zero \\{$r, $g\\}, [$s, \\{$x\\}];",
+ []>;
+
+def SULD_1D_ARRAY_V2I8_ZERO
+ : NVPTXInst<(outs Int16Regs:$r, Int16Regs:$g),
+ (ins Int64Regs:$s, Int32Regs:$l, Int32Regs:$x),
+ "suld.b.a1d.v2.b8.zero \\{$r, $g\\}, [$s, \\{$l, $x\\}];",
+ []>;
+def SULD_1D_ARRAY_V2I16_ZERO
+ : NVPTXInst<(outs Int16Regs:$r, Int16Regs:$g),
+ (ins Int64Regs:$s, Int32Regs:$l, Int32Regs:$x),
+ "suld.b.a1d.v2.b16.zero \\{$r, $g\\}, [$s, \\{$l, $x\\}];",
+ []>;
+def SULD_1D_ARRAY_V2I32_ZERO
+ : NVPTXInst<(outs Int32Regs:$r, Int32Regs:$g),
+ (ins Int64Regs:$s, Int32Regs:$l, Int32Regs:$x),
+ "suld.b.a1d.v2.b32.zero \\{$r, $g\\}, [$s, \\{$l, $x\\}];",
+ []>;
+def SULD_1D_ARRAY_V2I64_ZERO
+ : NVPTXInst<(outs Int64Regs:$r, Int64Regs:$g),
+ (ins Int64Regs:$s, Int32Regs:$l, Int32Regs:$x),
+ "suld.b.a1d.v2.b64.zero \\{$r, $g\\}, [$s, \\{$l, $x\\}];",
+ []>;
+
+def SULD_2D_V2I8_ZERO
+ : NVPTXInst<(outs Int16Regs:$r, Int16Regs:$g),
+ (ins Int64Regs:$s, Int32Regs:$x, Int32Regs:$y),
+ "suld.b.2d.v2.b8.zero \\{$r, $g\\}, [$s, \\{$x, $y\\}];",
+ []>;
+def SULD_2D_V2I16_ZERO
+ : NVPTXInst<(outs Int16Regs:$r, Int16Regs:$g),
+ (ins Int64Regs:$s, Int32Regs:$x, Int32Regs:$y),
+ "suld.b.2d.v2.b16.zero \\{$r, $g\\}, [$s, \\{$x, $y\\}];",
+ []>;
+def SULD_2D_V2I32_ZERO
+ : NVPTXInst<(outs Int32Regs:$r, Int32Regs:$g),
+ (ins Int64Regs:$s, Int32Regs:$x, Int32Regs:$y),
+ "suld.b.2d.v2.b32.zero \\{$r, $g\\}, [$s, \\{$x, $y\\}];",
+ []>;
+def SULD_2D_V2I64_ZERO
+ : NVPTXInst<(outs Int64Regs:$r, Int64Regs:$g),
+ (ins Int64Regs:$s, Int32Regs:$x, Int32Regs:$y),
+ "suld.b.2d.v2.b64.zero \\{$r, $g\\}, [$s, \\{$x, $y\\}];",
+ []>;
+
+def SULD_2D_ARRAY_V2I8_ZERO
+ : NVPTXInst<(outs Int16Regs:$r, Int16Regs:$g),
+ (ins Int64Regs:$s, Int32Regs:$l, Int32Regs:$x, Int32Regs:$y),
+ "suld.b.a2d.v2.b8.zero \\{$r, $g\\}, "
+ "[$s, \\{$l, $x, $y, $y\\}];",
+ []>;
+def SULD_2D_ARRAY_V2I16_ZERO
+ : NVPTXInst<(outs Int16Regs:$r, Int16Regs:$g),
+ (ins Int64Regs:$s, Int32Regs:$l, Int32Regs:$x, Int32Regs:$y),
+ "suld.b.a2d.v2.b16.zero \\{$r, $g\\}, "
+ "[$s, \\{$l, $x, $y, $y\\}];",
+ []>;
+def SULD_2D_ARRAY_V2I32_ZERO
+ : NVPTXInst<(outs Int32Regs:$r, Int32Regs:$g),
+ (ins Int64Regs:$s, Int32Regs:$l, Int32Regs:$x, Int32Regs:$y),
+ "suld.b.a2d.v2.b32.zero \\{$r, $g\\}, "
+ "[$s, \\{$l, $x, $y, $y\\}];",
+ []>;
+def SULD_2D_ARRAY_V2I64_ZERO
+ : NVPTXInst<(outs Int64Regs:$r, Int64Regs:$g),
+ (ins Int64Regs:$s, Int32Regs:$l, Int32Regs:$x, Int32Regs:$y),
+ "suld.b.a2d.v2.b64.zero \\{$r, $g\\}, "
+ "[$s, \\{$l, $x, $y, $y\\}];",
+ []>;
+
+def SULD_3D_V2I8_ZERO
+ : NVPTXInst<(outs Int16Regs:$r, Int16Regs:$g),
+ (ins Int64Regs:$s, Int32Regs:$x, Int32Regs:$y, Int32Regs:$z),
+ "suld.b.3d.v2.b8.zero \\{$r, $g\\}, [$s, \\{$x, $y, $z, $z\\}];",
+ []>;
+def SULD_3D_V2I16_ZERO
+ : NVPTXInst<(outs Int16Regs:$r, Int16Regs:$g),
+ (ins Int64Regs:$s, Int32Regs:$x, Int32Regs:$y, Int32Regs:$z),
+ "suld.b.3d.v2.b16.zero \\{$r, $g\\}, [$s, \\{$x, $y, $z, $z\\}];",
+ []>;
+def SULD_3D_V2I32_ZERO
+ : NVPTXInst<(outs Int32Regs:$r, Int32Regs:$g),
+ (ins Int64Regs:$s, Int32Regs:$x, Int32Regs:$y, Int32Regs:$z),
+ "suld.b.3d.v2.b32.zero \\{$r, $g\\}, [$s, \\{$x, $y, $z, $z\\}];",
+ []>;
+def SULD_3D_V2I64_ZERO
+ : NVPTXInst<(outs Int64Regs:$r, Int64Regs:$g),
+ (ins Int64Regs:$s, Int32Regs:$x, Int32Regs:$y, Int32Regs:$z),
+ "suld.b.3d.v2.b64.zero \\{$r, $g\\}, [$s, \\{$x, $y, $z, $z\\}];",
+ []>;
+}
+
+let IsSuld = 3 in {
+def SULD_1D_V4I8_ZERO
+ : NVPTXInst<(outs Int16Regs:$r, Int16Regs:$g, Int16Regs:$b, Int16Regs:$a),
+ (ins Int64Regs:$s, Int32Regs:$x),
+ "suld.b.1d.v4.b8.zero \\{$r, $g, $b, $a\\}, [$s, \\{$x\\}];",
+ []>;
+def SULD_1D_V4I16_ZERO
+ : NVPTXInst<(outs Int16Regs:$r, Int16Regs:$g, Int16Regs:$b, Int16Regs:$a),
+ (ins Int64Regs:$s, Int32Regs:$x),
+ "suld.b.1d.v4.b16.zero \\{$r, $g, $b, $a\\}, [$s, \\{$x\\}];",
+ []>;
+def SULD_1D_V4I32_ZERO
+ : NVPTXInst<(outs Int32Regs:$r, Int32Regs:$g, Int32Regs:$b, Int32Regs:$a),
+ (ins Int64Regs:$s, Int32Regs:$x),
+ "suld.b.1d.v4.b32.zero \\{$r, $g, $b, $a\\}, [$s, \\{$x\\}];",
+ []>;
+
+def SULD_1D_ARRAY_V4I8_ZERO
: NVPTXInst<(outs Int16Regs:$r, Int16Regs:$g, Int16Regs:$b, Int16Regs:$a),
(ins Int64Regs:$s, Int32Regs:$l, Int32Regs:$x),
- "suld.b.a1d.v4.b8.trap \\{$r, $g, $b, $a\\}, "
+ "suld.b.a1d.v4.b8.zero \\{$r, $g, $b, $a\\}, "
+ "[$s, \\{$l, $x\\}];",
+ []>;
+def SULD_1D_ARRAY_V4I16_ZERO
+ : NVPTXInst<(outs Int16Regs:$r, Int16Regs:$g, Int16Regs:$b, Int16Regs:$a),
+ (ins Int64Regs:$s, Int32Regs:$l, Int32Regs:$x),
+ "suld.b.a1d.v4.b16.zero \\{$r, $g, $b, $a\\}, "
+ "[$s, \\{$l, $x\\}];",
+ []>;
+def SULD_1D_ARRAY_V4I32_ZERO
+ : NVPTXInst<(outs Int32Regs:$r, Int32Regs:$g, Int32Regs:$b, Int32Regs:$a),
+ (ins Int64Regs:$s, Int32Regs:$l, Int32Regs:$x),
+ "suld.b.a1d.v4.b32.zero \\{$r, $g, $b, $a\\}, "
"[$s, \\{$l, $x\\}];",
[]>;
-def SULD_1D_ARRAY_V4I16_TRAP
- : NVPTXInst<(outs Int16Regs:$r, Int16Regs:$g, Int16Regs:$b, Int16Regs:$a),
- (ins Int64Regs:$s, Int32Regs:$l, Int32Regs:$x),
- "suld.b.a1d.v4.b16.trap \\{$r, $g, $b, $a\\}, "
- "[$s, \\{$l, $x\\}];",
+
+def SULD_2D_V4I8_ZERO
+ : NVPTXInst<(outs Int16Regs:$r, Int16Regs:$g, Int16Regs:$b, Int16Regs:$a),
+ (ins Int64Regs:$s, Int32Regs:$x, Int32Regs:$y),
+ "suld.b.2d.v4.b8.zero \\{$r, $g, $b, $a\\}, [$s, \\{$x, $y\\}];",
+ []>;
+def SULD_2D_V4I16_ZERO
+ : NVPTXInst<(outs Int16Regs:$r, Int16Regs:$g, Int16Regs:$b, Int16Regs:$a),
+ (ins Int64Regs:$s, Int32Regs:$x, Int32Regs:$y),
+ "suld.b.2d.v4.b16.zero \\{$r, $g, $b, $a\\}, [$s, \\{$x, $y\\}];",
+ []>;
+def SULD_2D_V4I32_ZERO
+ : NVPTXInst<(outs Int32Regs:$r, Int32Regs:$g, Int32Regs:$b, Int32Regs:$a),
+ (ins Int64Regs:$s, Int32Regs:$x, Int32Regs:$y),
+ "suld.b.2d.v4.b32.zero \\{$r, $g, $b, $a\\}, [$s, \\{$x, $y\\}];",
+ []>;
+
+def SULD_2D_ARRAY_V4I8_ZERO
+ : NVPTXInst<(outs Int16Regs:$r, Int16Regs:$g, Int16Regs:$b, Int16Regs:$a),
+ (ins Int64Regs:$s, Int32Regs:$l, Int32Regs:$x, Int32Regs:$y),
+ "suld.b.a2d.v4.b8.zero \\{$r, $g, $b, $a\\}, "
+ "[$s, \\{$l, $x, $y, $y\\}];",
+ []>;
+def SULD_2D_ARRAY_V4I16_ZERO
+ : NVPTXInst<(outs Int16Regs:$r, Int16Regs:$g, Int16Regs:$b, Int16Regs:$a),
+ (ins Int64Regs:$s, Int32Regs:$l, Int32Regs:$x, Int32Regs:$y),
+ "suld.b.a2d.v4.b16.zero \\{$r, $g, $b, $a\\}, "
+ "[$s, \\{$l, $x, $y, $y\\}];",
+ []>;
+def SULD_2D_ARRAY_V4I32_ZERO
+ : NVPTXInst<(outs Int32Regs:$r, Int32Regs:$g, Int32Regs:$b, Int32Regs:$a),
+ (ins Int64Regs:$s, Int32Regs:$l, Int32Regs:$x, Int32Regs:$y),
+ "suld.b.a2d.v4.b32.zero \\{$r, $g, $b, $a\\}, "
+ "[$s, \\{$l, $x, $y, $y\\}];",
+ []>;
+
+
+def SULD_3D_V4I8_ZERO
+ : NVPTXInst<(outs Int16Regs:$r, Int16Regs:$g, Int16Regs:$b, Int16Regs:$a),
+ (ins Int64Regs:$s, Int32Regs:$x, Int32Regs:$y, Int32Regs:$z),
+ "suld.b.3d.v4.b8.zero \\{$r, $g, $b, $a\\}, "
+ "[$s, \\{$x, $y, $z, $z\\}];",
+ []>;
+def SULD_3D_V4I16_ZERO
+ : NVPTXInst<(outs Int16Regs:$r, Int16Regs:$g, Int16Regs:$b, Int16Regs:$a),
+ (ins Int64Regs:$s, Int32Regs:$x, Int32Regs:$y, Int32Regs:$z),
+ "suld.b.3d.v4.b16.zero \\{$r, $g, $b, $a\\}, "
+ "[$s, \\{$x, $y, $z, $z\\}];",
+ []>;
+def SULD_3D_V4I32_ZERO
+ : NVPTXInst<(outs Int32Regs:$r, Int32Regs:$g, Int32Regs:$b, Int32Regs:$a),
+ (ins Int64Regs:$s, Int32Regs:$x, Int32Regs:$y, Int32Regs:$z),
+ "suld.b.3d.v4.b32.zero \\{$r, $g, $b, $a\\}, "
+ "[$s, \\{$x, $y, $z, $z\\}];",
+ []>;
+}
+
+//-----------------------------------
+// Texture Query Intrinsics
+//-----------------------------------
+
+let IsSurfTexQuery = 1 in {
+def TXQ_CHANNEL_ORDER
+ : NVPTXInst<(outs Int32Regs:$d), (ins Int64Regs:$a),
+ "txq.channel_order.b32 \t$d, [$a];",
+ []>;
+def TXQ_CHANNEL_DATA_TYPE
+ : NVPTXInst<(outs Int32Regs:$d), (ins Int64Regs:$a),
+ "txq.channel_data_type.b32 \t$d, [$a];",
+ []>;
+def TXQ_WIDTH
+ : NVPTXInst<(outs Int32Regs:$d), (ins Int64Regs:$a),
+ "txq.width.b32 \t$d, [$a];",
+ []>;
+def TXQ_HEIGHT
+ : NVPTXInst<(outs Int32Regs:$d), (ins Int64Regs:$a),
+ "txq.height.b32 \t$d, [$a];",
+ []>;
+def TXQ_DEPTH
+ : NVPTXInst<(outs Int32Regs:$d), (ins Int64Regs:$a),
+ "txq.depth.b32 \t$d, [$a];",
+ []>;
+def TXQ_ARRAY_SIZE
+ : NVPTXInst<(outs Int32Regs:$d), (ins Int64Regs:$a),
+ "txq.array_size.b32 \t$d, [$a];",
+ []>;
+def TXQ_NUM_SAMPLES
+ : NVPTXInst<(outs Int32Regs:$d), (ins Int64Regs:$a),
+ "txq.num_samples.b32 \t$d, [$a];",
+ []>;
+def TXQ_NUM_MIPMAP_LEVELS
+ : NVPTXInst<(outs Int32Regs:$d), (ins Int64Regs:$a),
+ "txq.num_mipmap_levels.b32 \t$d, [$a];",
+ []>;
+}
+
+def : Pat<(int_nvvm_txq_channel_order Int64Regs:$a),
+ (TXQ_CHANNEL_ORDER Int64Regs:$a)>;
+def : Pat<(int_nvvm_txq_channel_data_type Int64Regs:$a),
+ (TXQ_CHANNEL_DATA_TYPE Int64Regs:$a)>;
+def : Pat<(int_nvvm_txq_width Int64Regs:$a),
+ (TXQ_WIDTH Int64Regs:$a)>;
+def : Pat<(int_nvvm_txq_height Int64Regs:$a),
+ (TXQ_HEIGHT Int64Regs:$a)>;
+def : Pat<(int_nvvm_txq_depth Int64Regs:$a),
+ (TXQ_DEPTH Int64Regs:$a)>;
+def : Pat<(int_nvvm_txq_array_size Int64Regs:$a),
+ (TXQ_ARRAY_SIZE Int64Regs:$a)>;
+def : Pat<(int_nvvm_txq_num_samples Int64Regs:$a),
+ (TXQ_NUM_SAMPLES Int64Regs:$a)>;
+def : Pat<(int_nvvm_txq_num_mipmap_levels Int64Regs:$a),
+ (TXQ_NUM_MIPMAP_LEVELS Int64Regs:$a)>;
+
+
+//-----------------------------------
+// Surface Query Intrinsics
+//-----------------------------------
+
+let IsSurfTexQuery = 1 in {
+def SUQ_CHANNEL_ORDER
+ : NVPTXInst<(outs Int32Regs:$d), (ins Int64Regs:$a),
+ "suq.channel_order.b32 \t$d, [$a];",
+ []>;
+def SUQ_CHANNEL_DATA_TYPE
+ : NVPTXInst<(outs Int32Regs:$d), (ins Int64Regs:$a),
+ "suq.channel_data_type.b32 \t$d, [$a];",
+ []>;
+def SUQ_WIDTH
+ : NVPTXInst<(outs Int32Regs:$d), (ins Int64Regs:$a),
+ "suq.width.b32 \t$d, [$a];",
+ []>;
+def SUQ_HEIGHT
+ : NVPTXInst<(outs Int32Regs:$d), (ins Int64Regs:$a),
+ "suq.height.b32 \t$d, [$a];",
+ []>;
+def SUQ_DEPTH
+ : NVPTXInst<(outs Int32Regs:$d), (ins Int64Regs:$a),
+ "suq.depth.b32 \t$d, [$a];",
+ []>;
+def SUQ_ARRAY_SIZE
+ : NVPTXInst<(outs Int32Regs:$d), (ins Int64Regs:$a),
+ "suq.array_size.b32 \t$d, [$a];",
+ []>;
+}
+
+def : Pat<(int_nvvm_suq_channel_order Int64Regs:$a),
+ (SUQ_CHANNEL_ORDER Int64Regs:$a)>;
+def : Pat<(int_nvvm_suq_channel_data_type Int64Regs:$a),
+ (SUQ_CHANNEL_DATA_TYPE Int64Regs:$a)>;
+def : Pat<(int_nvvm_suq_width Int64Regs:$a),
+ (SUQ_WIDTH Int64Regs:$a)>;
+def : Pat<(int_nvvm_suq_height Int64Regs:$a),
+ (SUQ_HEIGHT Int64Regs:$a)>;
+def : Pat<(int_nvvm_suq_depth Int64Regs:$a),
+ (SUQ_DEPTH Int64Regs:$a)>;
+def : Pat<(int_nvvm_suq_array_size Int64Regs:$a),
+ (SUQ_ARRAY_SIZE Int64Regs:$a)>;
+
+
+//===- Handle Query -------------------------------------------------------===//
+
+// TODO: These intrinsics are not yet finalized, pending PTX ISA design work
+def ISTYPEP_SAMPLER
+ : NVPTXInst<(outs Int1Regs:$d), (ins Int64Regs:$a),
+ "istypep.samplerref \t$d, $a;",
+ [(set Int1Regs:$d, (int_nvvm_istypep_sampler Int64Regs:$a))]>;
+def ISTYPEP_SURFACE
+ : NVPTXInst<(outs Int1Regs:$d), (ins Int64Regs:$a),
+ "istypep.surfref \t$d, $a;",
+ [(set Int1Regs:$d, (int_nvvm_istypep_surface Int64Regs:$a))]>;
+def ISTYPEP_TEXTURE
+ : NVPTXInst<(outs Int1Regs:$d), (ins Int64Regs:$a),
+ "istypep.texref \t$d, $a;",
+ [(set Int1Regs:$d, (int_nvvm_istypep_texture Int64Regs:$a))]>;
+
+//===- Surface Stores -----------------------------------------------------===//
+
+let IsSust = 1 in {
+// Unformatted
+// .clamp variant
+def SUST_B_1D_B8_CLAMP
+ : NVPTXInst<(outs),
+ (ins Int64Regs:$s, Int32Regs:$x, Int16Regs:$r),
+ "sust.b.1d.b8.clamp \t[$s, \\{$x\\}], \\{$r\\};",
+ []>;
+def SUST_B_1D_B16_CLAMP
+ : NVPTXInst<(outs),
+ (ins Int64Regs:$s, Int32Regs:$x, Int16Regs:$r),
+ "sust.b.1d.b16.clamp \t[$s, \\{$x\\}], \\{$r\\};",
+ []>;
+def SUST_B_1D_B32_CLAMP
+ : NVPTXInst<(outs),
+ (ins Int64Regs:$s, Int32Regs:$x, Int32Regs:$r),
+ "sust.b.1d.b32.clamp \t[$s, \\{$x\\}], \\{$r\\};",
+ []>;
+def SUST_B_1D_B64_CLAMP
+ : NVPTXInst<(outs),
+ (ins Int64Regs:$s, Int32Regs:$x, Int64Regs:$r),
+ "sust.b.1d.b64.clamp \t[$s, \\{$x\\}], \\{$r\\};",
+ []>;
+def SUST_B_1D_V2B8_CLAMP
+ : NVPTXInst<(outs),
+ (ins Int64Regs:$s, Int32Regs:$x, Int16Regs:$r, Int16Regs:$g),
+ "sust.b.1d.v2.b8.clamp \t[$s, \\{$x\\}], \\{$r, $g\\};",
+ []>;
+def SUST_B_1D_V2B16_CLAMP
+ : NVPTXInst<(outs),
+ (ins Int64Regs:$s, Int32Regs:$x, Int16Regs:$r, Int16Regs:$g),
+ "sust.b.1d.v2.b16.clamp \t[$s, \\{$x\\}], \\{$r, $g\\};",
+ []>;
+def SUST_B_1D_V2B32_CLAMP
+ : NVPTXInst<(outs),
+ (ins Int64Regs:$s, Int32Regs:$x, Int32Regs:$r, Int32Regs:$g),
+ "sust.b.1d.v2.b32.clamp \t[$s, \\{$x\\}], \\{$r, $g\\};",
+ []>;
+def SUST_B_1D_V2B64_CLAMP
+ : NVPTXInst<(outs),
+ (ins Int64Regs:$s, Int32Regs:$x, Int64Regs:$r, Int64Regs:$g),
+ "sust.b.1d.v2.b64.clamp \t[$s, \\{$x\\}], \\{$r, $g\\};",
+ []>;
+def SUST_B_1D_V4B8_CLAMP
+ : NVPTXInst<(outs),
+ (ins Int64Regs:$s, Int32Regs:$x, Int16Regs:$r, Int16Regs:$g,
+ Int16Regs:$b, Int16Regs:$a),
+ "sust.b.1d.v4.b8.clamp \t[$s, \\{$x\\}], \\{$r, $g, $b, $a\\};",
+ []>;
+def SUST_B_1D_V4B16_CLAMP
+ : NVPTXInst<(outs),
+ (ins Int64Regs:$s, Int32Regs:$x, Int16Regs:$r, Int16Regs:$g,
+ Int16Regs:$b, Int16Regs:$a),
+ "sust.b.1d.v4.b16.clamp \t[$s, \\{$x\\}], \\{$r, $g, $b, $a\\};",
+ []>;
+def SUST_B_1D_V4B32_CLAMP
+ : NVPTXInst<(outs),
+ (ins Int64Regs:$s, Int32Regs:$x, Int32Regs:$r, Int32Regs:$g,
+ Int32Regs:$b, Int32Regs:$a),
+ "sust.b.1d.v4.b32.clamp \t[$s, \\{$x\\}], \\{$r, $g, $b, $a\\};",
+ []>;
+
+
+def SUST_B_1D_ARRAY_B8_CLAMP
+ : NVPTXInst<(outs),
+ (ins Int64Regs:$s, Int32Regs:$idx, Int32Regs:$x, Int16Regs:$r),
+ "sust.b.a1d.b8.clamp \t[$s, \\{$idx, $x\\}], \\{$r\\};",
+ []>;
+def SUST_B_1D_ARRAY_B16_CLAMP
+ : NVPTXInst<(outs),
+ (ins Int64Regs:$s, Int32Regs:$idx, Int32Regs:$x, Int16Regs:$r),
+ "sust.b.a1d.b16.clamp \t[$s, \\{$idx, $x\\}], \\{$r\\};",
+ []>;
+def SUST_B_1D_ARRAY_B32_CLAMP
+ : NVPTXInst<(outs),
+ (ins Int64Regs:$s, Int32Regs:$idx, Int32Regs:$x, Int32Regs:$r),
+ "sust.b.a1d.b32.clamp \t[$s, \\{$idx, $x\\}], \\{$r\\};",
+ []>;
+def SUST_B_1D_ARRAY_B64_CLAMP
+ : NVPTXInst<(outs),
+ (ins Int64Regs:$s, Int32Regs:$idx, Int32Regs:$x, Int64Regs:$r),
+ "sust.b.a1d.b64.clamp \t[$s, \\{$idx, $x\\}], \\{$r\\};",
+ []>;
+def SUST_B_1D_ARRAY_V2B8_CLAMP
+ : NVPTXInst<(outs),
+ (ins Int64Regs:$s, Int32Regs:$idx, Int32Regs:$x, Int16Regs:$r,
+ Int16Regs:$g),
+ "sust.b.a1d.v2.b8.clamp \t[$s, \\{$idx, $x\\}], \\{$r, $g\\};",
+ []>;
+def SUST_B_1D_ARRAY_V2B16_CLAMP
+ : NVPTXInst<(outs),
+ (ins Int64Regs:$s, Int32Regs:$idx, Int32Regs:$x, Int16Regs:$r,
+ Int16Regs:$g),
+ "sust.b.a1d.v2.b16.clamp \t[$s, \\{$idx, $x\\}], \\{$r, $g\\};",
+ []>;
+def SUST_B_1D_ARRAY_V2B32_CLAMP
+ : NVPTXInst<(outs),
+ (ins Int64Regs:$s, Int32Regs:$idx, Int32Regs:$x, Int32Regs:$r,
+ Int32Regs:$g),
+ "sust.b.a1d.v2.b32.clamp \t[$s, \\{$idx, $x\\}], \\{$r, $g\\};",
+ []>;
+def SUST_B_1D_ARRAY_V2B64_CLAMP
+ : NVPTXInst<(outs),
+ (ins Int64Regs:$s, Int32Regs:$idx, Int32Regs:$x, Int64Regs:$r,
+ Int64Regs:$g),
+ "sust.b.a1d.v2.b64.clamp \t[$s, \\{$idx, $x\\}], \\{$r, $g\\};",
+ []>;
+def SUST_B_1D_ARRAY_V4B8_CLAMP
+ : NVPTXInst<(outs),
+ (ins Int64Regs:$s, Int32Regs:$idx, Int32Regs:$x, Int16Regs:$r,
+ Int16Regs:$g, Int16Regs:$b, Int16Regs:$a),
+ "sust.b.a1d.v4.b8.clamp \t[$s, \\{$idx, $x\\}], "
+ "\\{$r, $g, $b, $a\\};",
+ []>;
+def SUST_B_1D_ARRAY_V4B16_CLAMP
+ : NVPTXInst<(outs),
+ (ins Int64Regs:$s, Int32Regs:$idx, Int32Regs:$x, Int16Regs:$r,
+ Int16Regs:$g, Int16Regs:$b, Int16Regs:$a),
+ "sust.b.a1d.v4.b16.clamp \t[$s, \\{$idx, $x\\}], "
+ "\\{$r, $g, $b, $a\\};",
+ []>;
+def SUST_B_1D_ARRAY_V4B32_CLAMP
+ : NVPTXInst<(outs),
+ (ins Int64Regs:$s, Int32Regs:$idx, Int32Regs:$x, Int32Regs:$r,
+ Int32Regs:$g, Int32Regs:$b, Int32Regs:$a),
+ "sust.b.a1d.v4.b32.clamp \t[$s, \\{$idx, $x\\}], "
+ "\\{$r, $g, $b, $a\\};",
+ []>;
+
+
+def SUST_B_2D_B8_CLAMP
+ : NVPTXInst<(outs),
+ (ins Int64Regs:$s, Int32Regs:$x, Int32Regs:$y, Int16Regs:$r),
+ "sust.b.2d.b8.clamp \t[$s, \\{$x, $y\\}], \\{$r\\};",
+ []>;
+def SUST_B_2D_B16_CLAMP
+ : NVPTXInst<(outs),
+ (ins Int64Regs:$s, Int32Regs:$x, Int32Regs:$y, Int16Regs:$r),
+ "sust.b.2d.b16.clamp \t[$s, \\{$x, $y\\}], \\{$r\\};",
+ []>;
+def SUST_B_2D_B32_CLAMP
+ : NVPTXInst<(outs),
+ (ins Int64Regs:$s, Int32Regs:$x, Int32Regs:$y, Int32Regs:$r),
+ "sust.b.2d.b32.clamp \t[$s, \\{$x, $y\\}], \\{$r\\};",
+ []>;
+def SUST_B_2D_B64_CLAMP
+ : NVPTXInst<(outs),
+ (ins Int64Regs:$s, Int32Regs:$x, Int32Regs:$y, Int64Regs:$r),
+ "sust.b.2d.b64.clamp \t[$s, \\{$x, $y\\}], \\{$r\\};",
+ []>;
+def SUST_B_2D_V2B8_CLAMP
+ : NVPTXInst<(outs),
+ (ins Int64Regs:$s, Int32Regs:$x, Int32Regs:$y, Int16Regs:$r,
+ Int16Regs:$g),
+ "sust.b.2d.v2.b8.clamp \t[$s, \\{$x, $y\\}], \\{$r, $g\\};",
+ []>;
+def SUST_B_2D_V2B16_CLAMP
+ : NVPTXInst<(outs),
+ (ins Int64Regs:$s, Int32Regs:$x, Int32Regs:$y, Int16Regs:$r,
+ Int16Regs:$g),
+ "sust.b.2d.v2.b16.clamp \t[$s, \\{$x, $y\\}], \\{$r, $g\\};",
+ []>;
+def SUST_B_2D_V2B32_CLAMP
+ : NVPTXInst<(outs),
+ (ins Int64Regs:$s, Int32Regs:$x, Int32Regs:$y, Int32Regs:$r,
+ Int32Regs:$g),
+ "sust.b.2d.v2.b32.clamp \t[$s, \\{$x, $y\\}], \\{$r, $g\\};",
+ []>;
+def SUST_B_2D_V2B64_CLAMP
+ : NVPTXInst<(outs),
+ (ins Int64Regs:$s, Int32Regs:$x, Int32Regs:$y, Int64Regs:$r,
+ Int64Regs:$g),
+ "sust.b.2d.v2.b64.clamp \t[$s, \\{$x, $y\\}], \\{$r, $g\\};",
+ []>;
+def SUST_B_2D_V4B8_CLAMP
+ : NVPTXInst<(outs),
+ (ins Int64Regs:$s, Int32Regs:$x, Int32Regs:$y, Int16Regs:$r,
+ Int16Regs:$g, Int16Regs:$b, Int16Regs:$a),
+ "sust.b.2d.v4.b8.clamp \t[$s, \\{$x, $y\\}], "
+ "\\{$r, $g, $b, $a\\};",
+ []>;
+def SUST_B_2D_V4B16_CLAMP
+ : NVPTXInst<(outs),
+ (ins Int64Regs:$s, Int32Regs:$x, Int32Regs:$y, Int16Regs:$r,
+ Int16Regs:$g, Int16Regs:$b, Int16Regs:$a),
+ "sust.b.2d.v4.b16.clamp \t[$s, \\{$x, $y\\}], "
+ "\\{$r, $g, $b, $a\\};",
+ []>;
+def SUST_B_2D_V4B32_CLAMP
+ : NVPTXInst<(outs),
+ (ins Int64Regs:$s, Int32Regs:$x, Int32Regs:$y, Int32Regs:$r,
+ Int32Regs:$g, Int32Regs:$b, Int32Regs:$a),
+ "sust.b.2d.v4.b32.clamp \t[$s, \\{$x, $y\\}], "
+ "\\{$r, $g, $b, $a\\};",
+ []>;
+
+
+def SUST_B_2D_ARRAY_B8_CLAMP
+ : NVPTXInst<(outs),
+ (ins Int64Regs:$s, Int32Regs:$idx, Int32Regs:$x, Int32Regs:$y,
+ Int16Regs:$r),
+ "sust.b.a2d.b8.clamp \t[$s, \\{$idx, $x, $y, $y\\}], \\{$r\\};",
+ []>;
+def SUST_B_2D_ARRAY_B16_CLAMP
+ : NVPTXInst<(outs),
+ (ins Int64Regs:$s, Int32Regs:$idx, Int32Regs:$x, Int32Regs:$y,
+ Int16Regs:$r),
+ "sust.b.a2d.b16.clamp \t[$s, \\{$idx, $x, $y, $y\\}], \\{$r\\};",
+ []>;
+def SUST_B_2D_ARRAY_B32_CLAMP
+ : NVPTXInst<(outs),
+ (ins Int64Regs:$s, Int32Regs:$idx, Int32Regs:$x, Int32Regs:$y,
+ Int32Regs:$r),
+ "sust.b.a2d.b32.clamp \t[$s, \\{$idx, $x, $y, $y\\}], \\{$r\\};",
+ []>;
+def SUST_B_2D_ARRAY_B64_CLAMP
+ : NVPTXInst<(outs),
+ (ins Int64Regs:$s, Int32Regs:$idx, Int32Regs:$x, Int32Regs:$y,
+ Int64Regs:$r),
+ "sust.b.a2d.b64.clamp \t[$s, \\{$idx, $x, $y, $y\\}], \\{$r\\};",
+ []>;
+def SUST_B_2D_ARRAY_V2B8_CLAMP
+ : NVPTXInst<(outs),
+ (ins Int64Regs:$s, Int32Regs:$idx, Int32Regs:$x, Int32Regs:$y,
+ Int16Regs:$r, Int16Regs:$g),
+ "sust.b.a2d.v2.b8.clamp \t[$s, \\{$idx, $x, $y, $y\\}], "
+ "\\{$r, $g\\};",
+ []>;
+def SUST_B_2D_ARRAY_V2B16_CLAMP
+ : NVPTXInst<(outs),
+ (ins Int64Regs:$s, Int32Regs:$idx, Int32Regs:$x, Int32Regs:$y,
+ Int16Regs:$r, Int16Regs:$g),
+ "sust.b.a2d.v2.b16.clamp \t[$s, \\{$idx, $x, $y, $y\\}], "
+ "\\{$r, $g\\};",
+ []>;
+def SUST_B_2D_ARRAY_V2B32_CLAMP
+ : NVPTXInst<(outs),
+ (ins Int64Regs:$s, Int32Regs:$idx, Int32Regs:$x, Int32Regs:$y,
+ Int32Regs:$r, Int32Regs:$g),
+ "sust.b.a2d.v2.b32.clamp \t[$s, \\{$idx, $x, $y, $y\\}], "
+ "\\{$r, $g\\};",
+ []>;
+def SUST_B_2D_ARRAY_V2B64_CLAMP
+ : NVPTXInst<(outs),
+ (ins Int64Regs:$s, Int32Regs:$idx, Int32Regs:$x, Int32Regs:$y,
+ Int64Regs:$r, Int64Regs:$g),
+ "sust.b.a2d.v2.b64.clamp \t[$s, \\{$idx, $x, $y, $y\\}], "
+ "\\{$r, $g\\};",
+ []>;
+def SUST_B_2D_ARRAY_V4B8_CLAMP
+ : NVPTXInst<(outs),
+ (ins Int64Regs:$s, Int32Regs:$idx, Int32Regs:$x, Int32Regs:$y,
+ Int16Regs:$r, Int16Regs:$g, Int16Regs:$b, Int16Regs:$a),
+ "sust.b.a2d.v4.b8.clamp \t[$s, \\{$idx, $x, $y, $y\\}], "
+ "\\{$r, $g, $b, $a\\};",
+ []>;
+def SUST_B_2D_ARRAY_V4B16_CLAMP
+ : NVPTXInst<(outs),
+ (ins Int64Regs:$s, Int32Regs:$idx, Int32Regs:$x, Int32Regs:$y,
+ Int16Regs:$r, Int16Regs:$g, Int16Regs:$b, Int16Regs:$a),
+ "sust.b.a2d.v4.b16.clamp \t[$s, \\{$idx, $x, $y, $y\\}], "
+ "\\{$r, $g, $b, $a\\};",
+ []>;
+def SUST_B_2D_ARRAY_V4B32_CLAMP
+ : NVPTXInst<(outs),
+ (ins Int64Regs:$s, Int32Regs:$idx, Int32Regs:$x, Int32Regs:$y,
+ Int32Regs:$r, Int32Regs:$g, Int32Regs:$b, Int32Regs:$a),
+ "sust.b.a2d.v4.b32.clamp \t[$s, \\{$idx, $x, $y, $y\\}], "
+ "\\{$r, $g, $b, $a\\};",
+ []>;
+
+
+def SUST_B_3D_B8_CLAMP
+ : NVPTXInst<(outs),
+ (ins Int64Regs:$s, Int32Regs:$x, Int32Regs:$y, Int32Regs:$z,
+ Int16Regs:$r),
+ "sust.b.3d.b8.clamp \t[$s, \\{$x, $y, $z, $z\\}], \\{$r\\};",
+ []>;
+def SUST_B_3D_B16_CLAMP
+ : NVPTXInst<(outs),
+ (ins Int64Regs:$s, Int32Regs:$x, Int32Regs:$y, Int32Regs:$z,
+ Int16Regs:$r),
+ "sust.b.3d.b16.clamp \t[$s, \\{$x, $y, $z, $z\\}], \\{$r\\};",
+ []>;
+def SUST_B_3D_B32_CLAMP
+ : NVPTXInst<(outs),
+ (ins Int64Regs:$s, Int32Regs:$x, Int32Regs:$y, Int32Regs:$z,
+ Int32Regs:$r),
+ "sust.b.3d.b32.clamp \t[$s, \\{$x, $y, $z, $z\\}], \\{$r\\};",
+ []>;
+def SUST_B_3D_B64_CLAMP
+ : NVPTXInst<(outs),
+ (ins Int64Regs:$s, Int32Regs:$x, Int32Regs:$y, Int32Regs:$z,
+ Int64Regs:$r),
+ "sust.b.3d.b64.clamp \t[$s, \\{$x, $y, $z, $z\\}], \\{$r\\};",
+ []>;
+def SUST_B_3D_V2B8_CLAMP
+ : NVPTXInst<(outs),
+ (ins Int64Regs:$s, Int32Regs:$x, Int32Regs:$y, Int32Regs:$z,
+ Int16Regs:$r, Int16Regs:$g),
+ "sust.b.3d.v2.b8.clamp \t[$s, \\{$x, $y, $z, $z\\}], "
+ "\\{$r, $g\\};",
+ []>;
+def SUST_B_3D_V2B16_CLAMP
+ : NVPTXInst<(outs),
+ (ins Int64Regs:$s, Int32Regs:$x, Int32Regs:$y, Int32Regs:$z,
+ Int16Regs:$r, Int16Regs:$g),
+ "sust.b.3d.v2.b16.clamp \t[$s, \\{$x, $y, $z, $z\\}], "
+ "\\{$r, $g\\};",
+ []>;
+def SUST_B_3D_V2B32_CLAMP
+ : NVPTXInst<(outs),
+ (ins Int64Regs:$s, Int32Regs:$x, Int32Regs:$y, Int32Regs:$z,
+ Int32Regs:$r, Int32Regs:$g),
+ "sust.b.3d.v2.b32.clamp \t[$s, \\{$x, $y, $z, $z\\}], "
+ "\\{$r, $g\\};",
+ []>;
+def SUST_B_3D_V2B64_CLAMP
+ : NVPTXInst<(outs),
+ (ins Int64Regs:$s, Int32Regs:$x, Int32Regs:$y, Int32Regs:$z,
+ Int64Regs:$r, Int64Regs:$g),
+ "sust.b.3d.v2.b64.clamp \t[$s, \\{$x, $y, $z, $z\\}], "
+ "\\{$r, $g\\};",
+ []>;
+def SUST_B_3D_V4B8_CLAMP
+ : NVPTXInst<(outs),
+ (ins Int64Regs:$s, Int32Regs:$x, Int32Regs:$y, Int32Regs:$z,
+ Int16Regs:$r, Int16Regs:$g, Int16Regs:$b, Int16Regs:$a),
+ "sust.b.3d.v4.b8.clamp \t[$s, \\{$x, $y, $z, $z\\}], "
+ "\\{$r, $g, $b, $a\\};",
+ []>;
+def SUST_B_3D_V4B16_CLAMP
+ : NVPTXInst<(outs),
+ (ins Int64Regs:$s, Int32Regs:$x, Int32Regs:$y, Int32Regs:$z,
+ Int16Regs:$r, Int16Regs:$g, Int16Regs:$b, Int16Regs:$a),
+ "sust.b.3d.v4.b16.clamp \t[$s, \\{$x, $y, $z, $z\\}], "
+ "\\{$r, $g, $b, $a\\};",
+ []>;
+def SUST_B_3D_V4B32_CLAMP
+ : NVPTXInst<(outs),
+ (ins Int64Regs:$s, Int32Regs:$x, Int32Regs:$y, Int32Regs:$z,
+ Int32Regs:$r, Int32Regs:$g, Int32Regs:$b, Int32Regs:$a),
+ "sust.b.3d.v4.b32.clamp \t[$s, \\{$x, $y, $z, $z\\}], "
+ "\\{$r, $g, $b, $a\\};",
+ []>;
+
+
+// .trap variant
+def SUST_B_1D_B8_TRAP
+ : NVPTXInst<(outs),
+ (ins Int64Regs:$s, Int32Regs:$x, Int16Regs:$r),
+ "sust.b.1d.b8.trap \t[$s, \\{$x\\}], \\{$r\\};",
+ []>;
+def SUST_B_1D_B16_TRAP
+ : NVPTXInst<(outs),
+ (ins Int64Regs:$s, Int32Regs:$x, Int16Regs:$r),
+ "sust.b.1d.b16.trap \t[$s, \\{$x\\}], \\{$r\\};",
[]>;
-def SULD_1D_ARRAY_V4I32_TRAP
- : NVPTXInst<(outs Int32Regs:$r, Int32Regs:$g, Int32Regs:$b, Int32Regs:$a),
- (ins Int64Regs:$s, Int32Regs:$l, Int32Regs:$x),
- "suld.b.a1d.v4.b32.trap \\{$r, $g, $b, $a\\}, "
- "[$s, \\{$l, $x\\}];",
+def SUST_B_1D_B32_TRAP
+ : NVPTXInst<(outs),
+ (ins Int64Regs:$s, Int32Regs:$x, Int32Regs:$r),
+ "sust.b.1d.b32.trap \t[$s, \\{$x\\}], \\{$r\\};",
+ []>;
+def SUST_B_1D_B64_TRAP
+ : NVPTXInst<(outs),
+ (ins Int64Regs:$s, Int32Regs:$x, Int64Regs:$r),
+ "sust.b.1d.b64.trap \t[$s, \\{$x\\}], \\{$r\\};",
+ []>;
+def SUST_B_1D_V2B8_TRAP
+ : NVPTXInst<(outs),
+ (ins Int64Regs:$s, Int32Regs:$x, Int16Regs:$r, Int16Regs:$g),
+ "sust.b.1d.v2.b8.trap \t[$s, \\{$x\\}], \\{$r, $g\\};",
+ []>;
+def SUST_B_1D_V2B16_TRAP
+ : NVPTXInst<(outs),
+ (ins Int64Regs:$s, Int32Regs:$x, Int16Regs:$r, Int16Regs:$g),
+ "sust.b.1d.v2.b16.trap \t[$s, \\{$x\\}], \\{$r, $g\\};",
+ []>;
+def SUST_B_1D_V2B32_TRAP
+ : NVPTXInst<(outs),
+ (ins Int64Regs:$s, Int32Regs:$x, Int32Regs:$r, Int32Regs:$g),
+ "sust.b.1d.v2.b32.trap \t[$s, \\{$x\\}], \\{$r, $g\\};",
+ []>;
+def SUST_B_1D_V2B64_TRAP
+ : NVPTXInst<(outs),
+ (ins Int64Regs:$s, Int32Regs:$x, Int64Regs:$r, Int64Regs:$g),
+ "sust.b.1d.v2.b64.trap \t[$s, \\{$x\\}], \\{$r, $g\\};",
+ []>;
+def SUST_B_1D_V4B8_TRAP
+ : NVPTXInst<(outs),
+ (ins Int64Regs:$s, Int32Regs:$x, Int16Regs:$r, Int16Regs:$g,
+ Int16Regs:$b, Int16Regs:$a),
+ "sust.b.1d.v4.b8.trap \t[$s, \\{$x\\}], \\{$r, $g, $b, $a\\};",
+ []>;
+def SUST_B_1D_V4B16_TRAP
+ : NVPTXInst<(outs),
+ (ins Int64Regs:$s, Int32Regs:$x, Int16Regs:$r, Int16Regs:$g,
+ Int16Regs:$b, Int16Regs:$a),
+ "sust.b.1d.v4.b16.trap \t[$s, \\{$x\\}], \\{$r, $g, $b, $a\\};",
+ []>;
+def SUST_B_1D_V4B32_TRAP
+ : NVPTXInst<(outs),
+ (ins Int64Regs:$s, Int32Regs:$x, Int32Regs:$r, Int32Regs:$g,
+ Int32Regs:$b, Int32Regs:$a),
+ "sust.b.1d.v4.b32.trap \t[$s, \\{$x\\}], \\{$r, $g, $b, $a\\};",
[]>;
-def SULD_2D_I8_TRAP
- : NVPTXInst<(outs Int16Regs:$r),
- (ins Int64Regs:$s, Int32Regs:$x, Int32Regs:$y),
- "suld.b.2d.b8.trap \\{$r\\}, [$s, \\{$x, $y\\}];",
+
+def SUST_B_1D_ARRAY_B8_TRAP
+ : NVPTXInst<(outs),
+ (ins Int64Regs:$s, Int32Regs:$idx, Int32Regs:$x, Int16Regs:$r),
+ "sust.b.a1d.b8.trap \t[$s, \\{$idx, $x\\}], \\{$r\\};",
[]>;
-def SULD_2D_I16_TRAP
- : NVPTXInst<(outs Int16Regs:$r),
- (ins Int64Regs:$s, Int32Regs:$x, Int32Regs:$y),
- "suld.b.2d.b16.trap \\{$r\\}, [$s, \\{$x, $y\\}];",
+def SUST_B_1D_ARRAY_B16_TRAP
+ : NVPTXInst<(outs),
+ (ins Int64Regs:$s, Int32Regs:$idx, Int32Regs:$x, Int16Regs:$r),
+ "sust.b.a1d.b16.trap \t[$s, \\{$idx, $x\\}], \\{$r\\};",
[]>;
-def SULD_2D_I32_TRAP
- : NVPTXInst<(outs Int32Regs:$r),
- (ins Int64Regs:$s, Int32Regs:$x, Int32Regs:$y),
- "suld.b.2d.b32.trap \\{$r\\}, [$s, \\{$x, $y\\}];",
+def SUST_B_1D_ARRAY_B32_TRAP
+ : NVPTXInst<(outs),
+ (ins Int64Regs:$s, Int32Regs:$idx, Int32Regs:$x, Int32Regs:$r),
+ "sust.b.a1d.b32.trap \t[$s, \\{$idx, $x\\}], \\{$r\\};",
[]>;
-def SULD_2D_V2I8_TRAP
- : NVPTXInst<(outs Int16Regs:$r, Int16Regs:$g),
- (ins Int64Regs:$s, Int32Regs:$x, Int32Regs:$y),
- "suld.b.2d.v2.b8.trap \\{$r, $g\\}, [$s, \\{$x, $y\\}];",
+def SUST_B_1D_ARRAY_B64_TRAP
+ : NVPTXInst<(outs),
+ (ins Int64Regs:$s, Int32Regs:$idx, Int32Regs:$x, Int64Regs:$r),
+ "sust.b.a1d.b64.trap \t[$s, \\{$idx, $x\\}], \\{$r\\};",
[]>;
-def SULD_2D_V2I16_TRAP
- : NVPTXInst<(outs Int16Regs:$r, Int16Regs:$g),
- (ins Int64Regs:$s, Int32Regs:$x, Int32Regs:$y),
- "suld.b.2d.v2.b16.trap \\{$r, $g\\}, [$s, \\{$x, $y\\}];",
+def SUST_B_1D_ARRAY_V2B8_TRAP
+ : NVPTXInst<(outs),
+ (ins Int64Regs:$s, Int32Regs:$idx, Int32Regs:$x, Int16Regs:$r,
+ Int16Regs:$g),
+ "sust.b.a1d.v2.b8.trap \t[$s, \\{$idx, $x\\}], \\{$r, $g\\};",
[]>;
-def SULD_2D_V2I32_TRAP
- : NVPTXInst<(outs Int32Regs:$r, Int32Regs:$g),
- (ins Int64Regs:$s, Int32Regs:$x, Int32Regs:$y),
- "suld.b.2d.v2.b32.trap \\{$r, $g\\}, [$s, \\{$x, $y\\}];",
+def SUST_B_1D_ARRAY_V2B16_TRAP
+ : NVPTXInst<(outs),
+ (ins Int64Regs:$s, Int32Regs:$idx, Int32Regs:$x, Int16Regs:$r,
+ Int16Regs:$g),
+ "sust.b.a1d.v2.b16.trap \t[$s, \\{$idx, $x\\}], \\{$r, $g\\};",
[]>;
-def SULD_2D_V4I8_TRAP
- : NVPTXInst<(outs Int16Regs:$r, Int16Regs:$g, Int16Regs:$b, Int16Regs:$a),
- (ins Int64Regs:$s, Int32Regs:$x, Int32Regs:$y),
- "suld.b.2d.v4.b8.trap \\{$r, $g, $b, $a\\}, [$s, \\{$x, $y\\}];",
+def SUST_B_1D_ARRAY_V2B32_TRAP
+ : NVPTXInst<(outs),
+ (ins Int64Regs:$s, Int32Regs:$idx, Int32Regs:$x, Int32Regs:$r,
+ Int32Regs:$g),
+ "sust.b.a1d.v2.b32.trap \t[$s, \\{$idx, $x\\}], \\{$r, $g\\};",
[]>;
-def SULD_2D_V4I16_TRAP
- : NVPTXInst<(outs Int16Regs:$r, Int16Regs:$g, Int16Regs:$b, Int16Regs:$a),
- (ins Int64Regs:$s, Int32Regs:$x, Int32Regs:$y),
- "suld.b.2d.v4.b16.trap \\{$r, $g, $b, $a\\}, [$s, \\{$x, $y\\}];",
+def SUST_B_1D_ARRAY_V2B64_TRAP
+ : NVPTXInst<(outs),
+ (ins Int64Regs:$s, Int32Regs:$idx, Int32Regs:$x, Int64Regs:$r,
+ Int64Regs:$g),
+ "sust.b.a1d.v2.b64.trap \t[$s, \\{$idx, $x\\}], \\{$r, $g\\};",
[]>;
-def SULD_2D_V4I32_TRAP
- : NVPTXInst<(outs Int32Regs:$r, Int32Regs:$g, Int32Regs:$b, Int32Regs:$a),
- (ins Int64Regs:$s, Int32Regs:$x, Int32Regs:$y),
- "suld.b.2d.v4.b32.trap \\{$r, $g, $b, $a\\}, [$s, \\{$x, $y\\}];",
+def SUST_B_1D_ARRAY_V4B8_TRAP
+ : NVPTXInst<(outs),
+ (ins Int64Regs:$s, Int32Regs:$idx, Int32Regs:$x, Int16Regs:$r,
+ Int16Regs:$g, Int16Regs:$b, Int16Regs:$a),
+ "sust.b.a1d.v4.b8.trap \t[$s, \\{$idx, $x\\}], "
+ "\\{$r, $g, $b, $a\\};",
+ []>;
+def SUST_B_1D_ARRAY_V4B16_TRAP
+ : NVPTXInst<(outs),
+ (ins Int64Regs:$s, Int32Regs:$idx, Int32Regs:$x, Int16Regs:$r,
+ Int16Regs:$g, Int16Regs:$b, Int16Regs:$a),
+ "sust.b.a1d.v4.b16.trap \t[$s, \\{$idx, $x\\}], "
+ "\\{$r, $g, $b, $a\\};",
+ []>;
+def SUST_B_1D_ARRAY_V4B32_TRAP
+ : NVPTXInst<(outs),
+ (ins Int64Regs:$s, Int32Regs:$idx, Int32Regs:$x, Int32Regs:$r,
+ Int32Regs:$g, Int32Regs:$b, Int32Regs:$a),
+ "sust.b.a1d.v4.b32.trap \t[$s, \\{$idx, $x\\}], "
+ "\\{$r, $g, $b, $a\\};",
[]>;
-def SULD_2D_ARRAY_I8_TRAP
- : NVPTXInst<(outs Int16Regs:$r),
- (ins Int64Regs:$s, Int32Regs:$l, Int32Regs:$x, Int32Regs:$y),
- "suld.b.a2d.b8.trap \\{$r\\}, [$s, \\{$l, $x, $y, $y\\}];",
+
+def SUST_B_2D_B8_TRAP
+ : NVPTXInst<(outs),
+ (ins Int64Regs:$s, Int32Regs:$x, Int32Regs:$y, Int16Regs:$r),
+ "sust.b.2d.b8.trap \t[$s, \\{$x, $y\\}], \\{$r\\};",
[]>;
-def SULD_2D_ARRAY_I16_TRAP
- : NVPTXInst<(outs Int16Regs:$r),
- (ins Int64Regs:$s, Int32Regs:$l, Int32Regs:$x, Int32Regs:$y),
- "suld.b.a2d.b16.trap \\{$r\\}, [$s, \\{$l, $x, $y, $y\\}];",
+def SUST_B_2D_B16_TRAP
+ : NVPTXInst<(outs),
+ (ins Int64Regs:$s, Int32Regs:$x, Int32Regs:$y, Int16Regs:$r),
+ "sust.b.2d.b16.trap \t[$s, \\{$x, $y\\}], \\{$r\\};",
[]>;
-def SULD_2D_ARRAY_I32_TRAP
- : NVPTXInst<(outs Int32Regs:$r),
- (ins Int64Regs:$s, Int32Regs:$l, Int32Regs:$x, Int32Regs:$y),
- "suld.b.a2d.b32.trap \\{$r\\}, [$s, \\{$l, $x, $y, $y\\}];",
+def SUST_B_2D_B32_TRAP
+ : NVPTXInst<(outs),
+ (ins Int64Regs:$s, Int32Regs:$x, Int32Regs:$y, Int32Regs:$r),
+ "sust.b.2d.b32.trap \t[$s, \\{$x, $y\\}], \\{$r\\};",
[]>;
-def SULD_2D_ARRAY_V2I8_TRAP
- : NVPTXInst<(outs Int16Regs:$r, Int16Regs:$g),
- (ins Int64Regs:$s, Int32Regs:$l, Int32Regs:$x, Int32Regs:$y),
- "suld.b.a2d.v2.b8.trap \\{$r, $g\\}, "
- "[$s, \\{$l, $x, $y, $y\\}];",
+def SUST_B_2D_B64_TRAP
+ : NVPTXInst<(outs),
+ (ins Int64Regs:$s, Int32Regs:$x, Int32Regs:$y, Int64Regs:$r),
+ "sust.b.2d.b64.trap \t[$s, \\{$x, $y\\}], \\{$r\\};",
[]>;
-def SULD_2D_ARRAY_V2I16_TRAP
- : NVPTXInst<(outs Int16Regs:$r, Int16Regs:$g),
- (ins Int64Regs:$s, Int32Regs:$l, Int32Regs:$x, Int32Regs:$y),
- "suld.b.a2d.v2.b16.trap \\{$r, $g\\}, "
- "[$s, \\{$l, $x, $y, $y\\}];",
+def SUST_B_2D_V2B8_TRAP
+ : NVPTXInst<(outs),
+ (ins Int64Regs:$s, Int32Regs:$x, Int32Regs:$y, Int16Regs:$r,
+ Int16Regs:$g),
+ "sust.b.2d.v2.b8.trap \t[$s, \\{$x, $y\\}], \\{$r, $g\\};",
[]>;
-def SULD_2D_ARRAY_V2I32_TRAP
- : NVPTXInst<(outs Int32Regs:$r, Int32Regs:$g),
- (ins Int64Regs:$s, Int32Regs:$l, Int32Regs:$x, Int32Regs:$y),
- "suld.b.a2d.v2.b32.trap \\{$r, $g\\}, "
- "[$s, \\{$l, $x, $y, $y\\}];",
+def SUST_B_2D_V2B16_TRAP
+ : NVPTXInst<(outs),
+ (ins Int64Regs:$s, Int32Regs:$x, Int32Regs:$y, Int16Regs:$r,
+ Int16Regs:$g),
+ "sust.b.2d.v2.b16.trap \t[$s, \\{$x, $y\\}], \\{$r, $g\\};",
[]>;
-def SULD_2D_ARRAY_V4I8_TRAP
- : NVPTXInst<(outs Int16Regs:$r, Int16Regs:$g, Int16Regs:$b, Int16Regs:$a),
- (ins Int64Regs:$s, Int32Regs:$l, Int32Regs:$x, Int32Regs:$y),
- "suld.b.a2d.v4.b8.trap \\{$r, $g, $b, $a\\}, "
- "[$s, \\{$l, $x, $y, $y\\}];",
+def SUST_B_2D_V2B32_TRAP
+ : NVPTXInst<(outs),
+ (ins Int64Regs:$s, Int32Regs:$x, Int32Regs:$y, Int32Regs:$r,
+ Int32Regs:$g),
+ "sust.b.2d.v2.b32.trap \t[$s, \\{$x, $y\\}], \\{$r, $g\\};",
[]>;
-def SULD_2D_ARRAY_V4I16_TRAP
- : NVPTXInst<(outs Int16Regs:$r, Int16Regs:$g, Int16Regs:$b, Int16Regs:$a),
- (ins Int64Regs:$s, Int32Regs:$l, Int32Regs:$x, Int32Regs:$y),
- "suld.b.a2d.v4.b16.trap \\{$r, $g, $b, $a\\}, "
- "[$s, \\{$l, $x, $y, $y\\}];",
+def SUST_B_2D_V2B64_TRAP
+ : NVPTXInst<(outs),
+ (ins Int64Regs:$s, Int32Regs:$x, Int32Regs:$y, Int64Regs:$r,
+ Int64Regs:$g),
+ "sust.b.2d.v2.b64.trap \t[$s, \\{$x, $y\\}], \\{$r, $g\\};",
[]>;
-def SULD_2D_ARRAY_V4I32_TRAP
- : NVPTXInst<(outs Int32Regs:$r, Int32Regs:$g, Int32Regs:$b, Int32Regs:$a),
- (ins Int64Regs:$s, Int32Regs:$l, Int32Regs:$x, Int32Regs:$y),
- "suld.b.a2d.v4.b32.trap \\{$r, $g, $b, $a\\}, "
- "[$s, \\{$l, $x, $y, $y\\}];",
+def SUST_B_2D_V4B8_TRAP
+ : NVPTXInst<(outs),
+ (ins Int64Regs:$s, Int32Regs:$x, Int32Regs:$y, Int16Regs:$r,
+ Int16Regs:$g, Int16Regs:$b, Int16Regs:$a),
+ "sust.b.2d.v4.b8.trap \t[$s, \\{$x, $y\\}], "
+ "\\{$r, $g, $b, $a\\};",
+ []>;
+def SUST_B_2D_V4B16_TRAP
+ : NVPTXInst<(outs),
+ (ins Int64Regs:$s, Int32Regs:$x, Int32Regs:$y, Int16Regs:$r,
+ Int16Regs:$g, Int16Regs:$b, Int16Regs:$a),
+ "sust.b.2d.v4.b16.trap \t[$s, \\{$x, $y\\}], "
+ "\\{$r, $g, $b, $a\\};",
+ []>;
+def SUST_B_2D_V4B32_TRAP
+ : NVPTXInst<(outs),
+ (ins Int64Regs:$s, Int32Regs:$x, Int32Regs:$y, Int32Regs:$r,
+ Int32Regs:$g, Int32Regs:$b, Int32Regs:$a),
+ "sust.b.2d.v4.b32.trap \t[$s, \\{$x, $y\\}], "
+ "\\{$r, $g, $b, $a\\};",
[]>;
-def SULD_3D_I8_TRAP
- : NVPTXInst<(outs Int16Regs:$r),
- (ins Int64Regs:$s, Int32Regs:$x, Int32Regs:$y, Int32Regs:$z),
- "suld.b.3d.b8.trap \\{$r\\}, [$s, \\{$x, $y, $z, $z\\}];",
+
+def SUST_B_2D_ARRAY_B8_TRAP
+ : NVPTXInst<(outs),
+ (ins Int64Regs:$s, Int32Regs:$idx, Int32Regs:$x, Int32Regs:$y,
+ Int16Regs:$r),
+ "sust.b.a2d.b8.trap \t[$s, \\{$idx, $x, $y, $y\\}], \\{$r\\};",
+ []>;
+def SUST_B_2D_ARRAY_B16_TRAP
+ : NVPTXInst<(outs),
+ (ins Int64Regs:$s, Int32Regs:$idx, Int32Regs:$x, Int32Regs:$y,
+ Int16Regs:$r),
+ "sust.b.a2d.b16.trap \t[$s, \\{$idx, $x, $y, $y\\}], \\{$r\\};",
+ []>;
+def SUST_B_2D_ARRAY_B32_TRAP
+ : NVPTXInst<(outs),
+ (ins Int64Regs:$s, Int32Regs:$idx, Int32Regs:$x, Int32Regs:$y,
+ Int32Regs:$r),
+ "sust.b.a2d.b32.trap \t[$s, \\{$idx, $x, $y, $y\\}], \\{$r\\};",
[]>;
-def SULD_3D_I16_TRAP
- : NVPTXInst<(outs Int16Regs:$r),
- (ins Int64Regs:$s, Int32Regs:$x, Int32Regs:$y, Int32Regs:$z),
- "suld.b.3d.b16.trap \\{$r\\}, [$s, \\{$x, $y, $z, $z\\}];",
+def SUST_B_2D_ARRAY_B64_TRAP
+ : NVPTXInst<(outs),
+ (ins Int64Regs:$s, Int32Regs:$idx, Int32Regs:$x, Int32Regs:$y,
+ Int64Regs:$r),
+ "sust.b.a2d.b64.trap \t[$s, \\{$idx, $x, $y, $y\\}], \\{$r\\};",
[]>;
-def SULD_3D_I32_TRAP
- : NVPTXInst<(outs Int32Regs:$r),
- (ins Int64Regs:$s, Int32Regs:$x, Int32Regs:$y, Int32Regs:$z),
- "suld.b.3d.b32.trap \\{$r\\}, [$s, \\{$x, $y, $z, $z\\}];",
+def SUST_B_2D_ARRAY_V2B8_TRAP
+ : NVPTXInst<(outs),
+ (ins Int64Regs:$s, Int32Regs:$idx, Int32Regs:$x, Int32Regs:$y,
+ Int16Regs:$r, Int16Regs:$g),
+ "sust.b.a2d.v2.b8.trap \t[$s, \\{$idx, $x, $y, $y\\}], "
+ "\\{$r, $g\\};",
[]>;
-def SULD_3D_V2I8_TRAP
- : NVPTXInst<(outs Int16Regs:$r, Int16Regs:$g),
- (ins Int64Regs:$s, Int32Regs:$x, Int32Regs:$y, Int32Regs:$z),
- "suld.b.3d.v2.b8.trap \\{$r, $g\\}, [$s, \\{$x, $y, $z, $z\\}];",
+def SUST_B_2D_ARRAY_V2B16_TRAP
+ : NVPTXInst<(outs),
+ (ins Int64Regs:$s, Int32Regs:$idx, Int32Regs:$x, Int32Regs:$y,
+ Int16Regs:$r, Int16Regs:$g),
+ "sust.b.a2d.v2.b16.trap \t[$s, \\{$idx, $x, $y, $y\\}], "
+ "\\{$r, $g\\};",
[]>;
-def SULD_3D_V2I16_TRAP
- : NVPTXInst<(outs Int16Regs:$r, Int16Regs:$g),
- (ins Int64Regs:$s, Int32Regs:$x, Int32Regs:$y, Int32Regs:$z),
- "suld.b.3d.v2.b16.trap \\{$r, $g\\}, [$s, \\{$x, $y, $z, $z\\}];",
+def SUST_B_2D_ARRAY_V2B32_TRAP
+ : NVPTXInst<(outs),
+ (ins Int64Regs:$s, Int32Regs:$idx, Int32Regs:$x, Int32Regs:$y,
+ Int32Regs:$r, Int32Regs:$g),
+ "sust.b.a2d.v2.b32.trap \t[$s, \\{$idx, $x, $y, $y\\}], "
+ "\\{$r, $g\\};",
[]>;
-def SULD_3D_V2I32_TRAP
- : NVPTXInst<(outs Int32Regs:$r, Int32Regs:$g),
- (ins Int64Regs:$s, Int32Regs:$x, Int32Regs:$y, Int32Regs:$z),
- "suld.b.3d.v2.b32.trap \\{$r, $g\\}, [$s, \\{$x, $y, $z, $z\\}];",
+def SUST_B_2D_ARRAY_V2B64_TRAP
+ : NVPTXInst<(outs),
+ (ins Int64Regs:$s, Int32Regs:$idx, Int32Regs:$x, Int32Regs:$y,
+ Int64Regs:$r, Int64Regs:$g),
+ "sust.b.a2d.v2.b64.trap \t[$s, \\{$idx, $x, $y, $y\\}], "
+ "\\{$r, $g\\};",
[]>;
-def SULD_3D_V4I8_TRAP
- : NVPTXInst<(outs Int16Regs:$r, Int16Regs:$g, Int16Regs:$b, Int16Regs:$a),
- (ins Int64Regs:$s, Int32Regs:$x, Int32Regs:$y, Int32Regs:$z),
- "suld.b.3d.v4.b8.trap \\{$r, $g, $b, $a\\}, "
- "[$s, \\{$x, $y, $z, $z\\}];",
+def SUST_B_2D_ARRAY_V4B8_TRAP
+ : NVPTXInst<(outs),
+ (ins Int64Regs:$s, Int32Regs:$idx, Int32Regs:$x, Int32Regs:$y,
+ Int16Regs:$r, Int16Regs:$g, Int16Regs:$b, Int16Regs:$a),
+ "sust.b.a2d.v4.b8.trap \t[$s, \\{$idx, $x, $y, $y\\}], "
+ "\\{$r, $g, $b, $a\\};",
[]>;
-def SULD_3D_V4I16_TRAP
- : NVPTXInst<(outs Int16Regs:$r, Int16Regs:$g, Int16Regs:$b, Int16Regs:$a),
- (ins Int64Regs:$s, Int32Regs:$x, Int32Regs:$y, Int32Regs:$z),
- "suld.b.3d.v4.b16.trap \\{$r, $g, $b, $a\\}, "
- "[$s, \\{$x, $y, $z, $z\\}];",
+def SUST_B_2D_ARRAY_V4B16_TRAP
+ : NVPTXInst<(outs),
+ (ins Int64Regs:$s, Int32Regs:$idx, Int32Regs:$x, Int32Regs:$y,
+ Int16Regs:$r, Int16Regs:$g, Int16Regs:$b, Int16Regs:$a),
+ "sust.b.a2d.v4.b16.trap \t[$s, \\{$idx, $x, $y, $y\\}], "
+ "\\{$r, $g, $b, $a\\};",
[]>;
-def SULD_3D_V4I32_TRAP
- : NVPTXInst<(outs Int32Regs:$r, Int32Regs:$g, Int32Regs:$b, Int32Regs:$a),
- (ins Int64Regs:$s, Int32Regs:$x, Int32Regs:$y, Int32Regs:$z),
- "suld.b.3d.v4.b32.trap \\{$r, $g, $b, $a\\}, "
- "[$s, \\{$x, $y, $z, $z\\}];",
+def SUST_B_2D_ARRAY_V4B32_TRAP
+ : NVPTXInst<(outs),
+ (ins Int64Regs:$s, Int32Regs:$idx, Int32Regs:$x, Int32Regs:$y,
+ Int32Regs:$r, Int32Regs:$g, Int32Regs:$b, Int32Regs:$a),
+ "sust.b.a2d.v4.b32.trap \t[$s, \\{$idx, $x, $y, $y\\}], "
+ "\\{$r, $g, $b, $a\\};",
[]>;
-//-----------------------------------
-// Texture Query Intrinsics
-//-----------------------------------
-def TXQ_CHANNEL_ORDER
- : NVPTXInst<(outs Int32Regs:$d), (ins Int64Regs:$a),
- "txq.channel_order.b32 \t$d, [$a];",
- []>;
-def TXQ_CHANNEL_DATA_TYPE
- : NVPTXInst<(outs Int32Regs:$d), (ins Int64Regs:$a),
- "txq.channel_data_type.b32 \t$d, [$a];",
- []>;
-def TXQ_WIDTH
- : NVPTXInst<(outs Int32Regs:$d), (ins Int64Regs:$a),
- "txq.width.b32 \t$d, [$a];",
- []>;
-def TXQ_HEIGHT
- : NVPTXInst<(outs Int32Regs:$d), (ins Int64Regs:$a),
- "txq.height.b32 \t$d, [$a];",
+def SUST_B_3D_B8_TRAP
+ : NVPTXInst<(outs),
+ (ins Int64Regs:$s, Int32Regs:$x, Int32Regs:$y, Int32Regs:$z,
+ Int16Regs:$r),
+ "sust.b.3d.b8.trap \t[$s, \\{$x, $y, $z, $z\\}], \\{$r\\};",
[]>;
-def TXQ_DEPTH
- : NVPTXInst<(outs Int32Regs:$d), (ins Int64Regs:$a),
- "txq.depth.b32 \t$d, [$a];",
+def SUST_B_3D_B16_TRAP
+ : NVPTXInst<(outs),
+ (ins Int64Regs:$s, Int32Regs:$x, Int32Regs:$y, Int32Regs:$z,
+ Int16Regs:$r),
+ "sust.b.3d.b16.trap \t[$s, \\{$x, $y, $z, $z\\}], \\{$r\\};",
[]>;
-def TXQ_ARRAY_SIZE
- : NVPTXInst<(outs Int32Regs:$d), (ins Int64Regs:$a),
- "txq.array_size.b32 \t$d, [$a];",
+def SUST_B_3D_B32_TRAP
+ : NVPTXInst<(outs),
+ (ins Int64Regs:$s, Int32Regs:$x, Int32Regs:$y, Int32Regs:$z,
+ Int32Regs:$r),
+ "sust.b.3d.b32.trap \t[$s, \\{$x, $y, $z, $z\\}], \\{$r\\};",
[]>;
-def TXQ_NUM_SAMPLES
- : NVPTXInst<(outs Int32Regs:$d), (ins Int64Regs:$a),
- "txq.num_samples.b32 \t$d, [$a];",
+def SUST_B_3D_B64_TRAP
+ : NVPTXInst<(outs),
+ (ins Int64Regs:$s, Int32Regs:$x, Int32Regs:$y, Int32Regs:$z,
+ Int64Regs:$r),
+ "sust.b.3d.b64.trap \t[$s, \\{$x, $y, $z, $z\\}], \\{$r\\};",
[]>;
-def TXQ_NUM_MIPMAP_LEVELS
- : NVPTXInst<(outs Int32Regs:$d), (ins Int64Regs:$a),
- "txq.num_mipmap_levels.b32 \t$d, [$a];",
+def SUST_B_3D_V2B8_TRAP
+ : NVPTXInst<(outs),
+ (ins Int64Regs:$s, Int32Regs:$x, Int32Regs:$y, Int32Regs:$z,
+ Int16Regs:$r, Int16Regs:$g),
+ "sust.b.3d.v2.b8.trap \t[$s, \\{$x, $y, $z, $z\\}], "
+ "\\{$r, $g\\};",
[]>;
-
-def : Pat<(int_nvvm_txq_channel_order Int64Regs:$a),
- (TXQ_CHANNEL_ORDER Int64Regs:$a)>;
-def : Pat<(int_nvvm_txq_channel_data_type Int64Regs:$a),
- (TXQ_CHANNEL_DATA_TYPE Int64Regs:$a)>;
-def : Pat<(int_nvvm_txq_width Int64Regs:$a),
- (TXQ_WIDTH Int64Regs:$a)>;
-def : Pat<(int_nvvm_txq_height Int64Regs:$a),
- (TXQ_HEIGHT Int64Regs:$a)>;
-def : Pat<(int_nvvm_txq_depth Int64Regs:$a),
- (TXQ_DEPTH Int64Regs:$a)>;
-def : Pat<(int_nvvm_txq_array_size Int64Regs:$a),
- (TXQ_ARRAY_SIZE Int64Regs:$a)>;
-def : Pat<(int_nvvm_txq_num_samples Int64Regs:$a),
- (TXQ_NUM_SAMPLES Int64Regs:$a)>;
-def : Pat<(int_nvvm_txq_num_mipmap_levels Int64Regs:$a),
- (TXQ_NUM_MIPMAP_LEVELS Int64Regs:$a)>;
-
-
-//-----------------------------------
-// Surface Query Intrinsics
-//-----------------------------------
-def SUQ_CHANNEL_ORDER
- : NVPTXInst<(outs Int32Regs:$d), (ins Int64Regs:$a),
- "suq.channel_order.b32 \t$d, [$a];",
+def SUST_B_3D_V2B16_TRAP
+ : NVPTXInst<(outs),
+ (ins Int64Regs:$s, Int32Regs:$x, Int32Regs:$y, Int32Regs:$z,
+ Int16Regs:$r, Int16Regs:$g),
+ "sust.b.3d.v2.b16.trap \t[$s, \\{$x, $y, $z, $z\\}], "
+ "\\{$r, $g\\};",
[]>;
-def SUQ_CHANNEL_DATA_TYPE
- : NVPTXInst<(outs Int32Regs:$d), (ins Int64Regs:$a),
- "suq.channel_data_type.b32 \t$d, [$a];",
+def SUST_B_3D_V2B32_TRAP
+ : NVPTXInst<(outs),
+ (ins Int64Regs:$s, Int32Regs:$x, Int32Regs:$y, Int32Regs:$z,
+ Int32Regs:$r, Int32Regs:$g),
+ "sust.b.3d.v2.b32.trap \t[$s, \\{$x, $y, $z, $z\\}], "
+ "\\{$r, $g\\};",
[]>;
-def SUQ_WIDTH
- : NVPTXInst<(outs Int32Regs:$d), (ins Int64Regs:$a),
- "suq.width.b32 \t$d, [$a];",
+def SUST_B_3D_V2B64_TRAP
+ : NVPTXInst<(outs),
+ (ins Int64Regs:$s, Int32Regs:$x, Int32Regs:$y, Int32Regs:$z,
+ Int64Regs:$r, Int64Regs:$g),
+ "sust.b.3d.v2.b64.trap \t[$s, \\{$x, $y, $z, $z\\}], "
+ "\\{$r, $g\\};",
[]>;
-def SUQ_HEIGHT
- : NVPTXInst<(outs Int32Regs:$d), (ins Int64Regs:$a),
- "suq.height.b32 \t$d, [$a];",
+def SUST_B_3D_V4B8_TRAP
+ : NVPTXInst<(outs),
+ (ins Int64Regs:$s, Int32Regs:$x, Int32Regs:$y, Int32Regs:$z,
+ Int16Regs:$r, Int16Regs:$g, Int16Regs:$b, Int16Regs:$a),
+ "sust.b.3d.v4.b8.trap \t[$s, \\{$x, $y, $z, $z\\}], "
+ "\\{$r, $g, $b, $a\\};",
[]>;
-def SUQ_DEPTH
- : NVPTXInst<(outs Int32Regs:$d), (ins Int64Regs:$a),
- "suq.depth.b32 \t$d, [$a];",
+def SUST_B_3D_V4B16_TRAP
+ : NVPTXInst<(outs),
+ (ins Int64Regs:$s, Int32Regs:$x, Int32Regs:$y, Int32Regs:$z,
+ Int16Regs:$r, Int16Regs:$g, Int16Regs:$b, Int16Regs:$a),
+ "sust.b.3d.v4.b16.trap \t[$s, \\{$x, $y, $z, $z\\}], "
+ "\\{$r, $g, $b, $a\\};",
[]>;
-def SUQ_ARRAY_SIZE
- : NVPTXInst<(outs Int32Regs:$d), (ins Int64Regs:$a),
- "suq.array_size.b32 \t$d, [$a];",
+def SUST_B_3D_V4B32_TRAP
+ : NVPTXInst<(outs),
+ (ins Int64Regs:$s, Int32Regs:$x, Int32Regs:$y, Int32Regs:$z,
+ Int32Regs:$r, Int32Regs:$g, Int32Regs:$b, Int32Regs:$a),
+ "sust.b.3d.v4.b32.trap \t[$s, \\{$x, $y, $z, $z\\}], "
+ "\\{$r, $g, $b, $a\\};",
[]>;
-def : Pat<(int_nvvm_suq_channel_order Int64Regs:$a),
- (SUQ_CHANNEL_ORDER Int64Regs:$a)>;
-def : Pat<(int_nvvm_suq_channel_data_type Int64Regs:$a),
- (SUQ_CHANNEL_DATA_TYPE Int64Regs:$a)>;
-def : Pat<(int_nvvm_suq_width Int64Regs:$a),
- (SUQ_WIDTH Int64Regs:$a)>;
-def : Pat<(int_nvvm_suq_height Int64Regs:$a),
- (SUQ_HEIGHT Int64Regs:$a)>;
-def : Pat<(int_nvvm_suq_depth Int64Regs:$a),
- (SUQ_DEPTH Int64Regs:$a)>;
-def : Pat<(int_nvvm_suq_array_size Int64Regs:$a),
- (SUQ_ARRAY_SIZE Int64Regs:$a)>;
-
-
-//===- Handle Query -------------------------------------------------------===//
-
-// TODO: These intrinsics are not yet finalized, pending PTX ISA design work
-def ISTYPEP_SAMPLER
- : NVPTXInst<(outs Int1Regs:$d), (ins Int64Regs:$a),
- "istypep.samplerref \t$d, $a;",
- [(set Int1Regs:$d, (int_nvvm_istypep_sampler Int64Regs:$a))]>;
-def ISTYPEP_SURFACE
- : NVPTXInst<(outs Int1Regs:$d), (ins Int64Regs:$a),
- "istypep.surfref \t$d, $a;",
- [(set Int1Regs:$d, (int_nvvm_istypep_surface Int64Regs:$a))]>;
-def ISTYPEP_TEXTURE
- : NVPTXInst<(outs Int1Regs:$d), (ins Int64Regs:$a),
- "istypep.texref \t$d, $a;",
- [(set Int1Regs:$d, (int_nvvm_istypep_texture Int64Regs:$a))]>;
-
-//===- Surface Stores -----------------------------------------------------===//
-
-// Unformatted
-def SUST_B_1D_B8_TRAP
+// .zero variant
+def SUST_B_1D_B8_ZERO
: NVPTXInst<(outs),
(ins Int64Regs:$s, Int32Regs:$x, Int16Regs:$r),
- "sust.b.1d.b8.trap \t[$s, \\{$x\\}], \\{$r\\};",
+ "sust.b.1d.b8.zero \t[$s, \\{$x\\}], \\{$r\\};",
[]>;
-def SUST_B_1D_B16_TRAP
+def SUST_B_1D_B16_ZERO
: NVPTXInst<(outs),
(ins Int64Regs:$s, Int32Regs:$x, Int16Regs:$r),
- "sust.b.1d.b16.trap \t[$s, \\{$x\\}], \\{$r\\};",
+ "sust.b.1d.b16.zero \t[$s, \\{$x\\}], \\{$r\\};",
[]>;
-def SUST_B_1D_B32_TRAP
+def SUST_B_1D_B32_ZERO
: NVPTXInst<(outs),
(ins Int64Regs:$s, Int32Regs:$x, Int32Regs:$r),
- "sust.b.1d.b32.trap \t[$s, \\{$x\\}], \\{$r\\};",
+ "sust.b.1d.b32.zero \t[$s, \\{$x\\}], \\{$r\\};",
[]>;
-def SUST_B_1D_V2B8_TRAP
+def SUST_B_1D_B64_ZERO
+ : NVPTXInst<(outs),
+ (ins Int64Regs:$s, Int32Regs:$x, Int64Regs:$r),
+ "sust.b.1d.b64.zero \t[$s, \\{$x\\}], \\{$r\\};",
+ []>;
+def SUST_B_1D_V2B8_ZERO
: NVPTXInst<(outs),
(ins Int64Regs:$s, Int32Regs:$x, Int16Regs:$r, Int16Regs:$g),
- "sust.b.1d.v2.b8.trap \t[$s, \\{$x\\}], \\{$r, $g\\};",
+ "sust.b.1d.v2.b8.zero \t[$s, \\{$x\\}], \\{$r, $g\\};",
[]>;
-def SUST_B_1D_V2B16_TRAP
+def SUST_B_1D_V2B16_ZERO
: NVPTXInst<(outs),
(ins Int64Regs:$s, Int32Regs:$x, Int16Regs:$r, Int16Regs:$g),
- "sust.b.1d.v2.b16.trap \t[$s, \\{$x\\}], \\{$r, $g\\};",
+ "sust.b.1d.v2.b16.zero \t[$s, \\{$x\\}], \\{$r, $g\\};",
[]>;
-def SUST_B_1D_V2B32_TRAP
+def SUST_B_1D_V2B32_ZERO
: NVPTXInst<(outs),
(ins Int64Regs:$s, Int32Regs:$x, Int32Regs:$r, Int32Regs:$g),
- "sust.b.1d.v2.b32.trap \t[$s, \\{$x\\}], \\{$r, $g\\};",
+ "sust.b.1d.v2.b32.zero \t[$s, \\{$x\\}], \\{$r, $g\\};",
[]>;
-def SUST_B_1D_V4B8_TRAP
+def SUST_B_1D_V2B64_ZERO
+ : NVPTXInst<(outs),
+ (ins Int64Regs:$s, Int32Regs:$x, Int64Regs:$r, Int64Regs:$g),
+ "sust.b.1d.v2.b64.zero \t[$s, \\{$x\\}], \\{$r, $g\\};",
+ []>;
+def SUST_B_1D_V4B8_ZERO
: NVPTXInst<(outs),
(ins Int64Regs:$s, Int32Regs:$x, Int16Regs:$r, Int16Regs:$g,
Int16Regs:$b, Int16Regs:$a),
- "sust.b.1d.v4.b8.trap \t[$s, \\{$x\\}], \\{$r, $g, $b, $a\\};",
+ "sust.b.1d.v4.b8.zero \t[$s, \\{$x\\}], \\{$r, $g, $b, $a\\};",
[]>;
-def SUST_B_1D_V4B16_TRAP
+def SUST_B_1D_V4B16_ZERO
: NVPTXInst<(outs),
(ins Int64Regs:$s, Int32Regs:$x, Int16Regs:$r, Int16Regs:$g,
Int16Regs:$b, Int16Regs:$a),
- "sust.b.1d.v4.b16.trap \t[$s, \\{$x\\}], \\{$r, $g, $b, $a\\};",
+ "sust.b.1d.v4.b16.zero \t[$s, \\{$x\\}], \\{$r, $g, $b, $a\\};",
[]>;
-def SUST_B_1D_V4B32_TRAP
+def SUST_B_1D_V4B32_ZERO
: NVPTXInst<(outs),
(ins Int64Regs:$s, Int32Regs:$x, Int32Regs:$r, Int32Regs:$g,
Int32Regs:$b, Int32Regs:$a),
- "sust.b.1d.v4.b32.trap \t[$s, \\{$x\\}], \\{$r, $g, $b, $a\\};",
+ "sust.b.1d.v4.b32.zero \t[$s, \\{$x\\}], \\{$r, $g, $b, $a\\};",
[]>;
-def SUST_B_1D_ARRAY_B8_TRAP
+def SUST_B_1D_ARRAY_B8_ZERO
: NVPTXInst<(outs),
(ins Int64Regs:$s, Int32Regs:$idx, Int32Regs:$x, Int16Regs:$r),
- "sust.b.a1d.b8.trap \t[$s, \\{$idx, $x\\}], \\{$r\\};",
+ "sust.b.a1d.b8.zero \t[$s, \\{$idx, $x\\}], \\{$r\\};",
[]>;
-def SUST_B_1D_ARRAY_B16_TRAP
+def SUST_B_1D_ARRAY_B16_ZERO
: NVPTXInst<(outs),
(ins Int64Regs:$s, Int32Regs:$idx, Int32Regs:$x, Int16Regs:$r),
- "sust.b.a1d.b16.trap \t[$s, \\{$idx, $x\\}], \\{$r\\};",
+ "sust.b.a1d.b16.zero \t[$s, \\{$idx, $x\\}], \\{$r\\};",
[]>;
-def SUST_B_1D_ARRAY_B32_TRAP
+def SUST_B_1D_ARRAY_B32_ZERO
: NVPTXInst<(outs),
(ins Int64Regs:$s, Int32Regs:$idx, Int32Regs:$x, Int32Regs:$r),
- "sust.b.a1d.b32.trap \t[$s, \\{$idx, $x\\}], \\{$r\\};",
+ "sust.b.a1d.b32.zero \t[$s, \\{$idx, $x\\}], \\{$r\\};",
[]>;
-def SUST_B_1D_ARRAY_V2B8_TRAP
+def SUST_B_1D_ARRAY_B64_ZERO
+ : NVPTXInst<(outs),
+ (ins Int64Regs:$s, Int32Regs:$idx, Int32Regs:$x, Int64Regs:$r),
+ "sust.b.a1d.b64.zero \t[$s, \\{$idx, $x\\}], \\{$r\\};",
+ []>;
+def SUST_B_1D_ARRAY_V2B8_ZERO
: NVPTXInst<(outs),
(ins Int64Regs:$s, Int32Regs:$idx, Int32Regs:$x, Int16Regs:$r,
Int16Regs:$g),
- "sust.b.a1d.v2.b8.trap \t[$s, \\{$idx, $x\\}], \\{$r, $g\\};",
+ "sust.b.a1d.v2.b8.zero \t[$s, \\{$idx, $x\\}], \\{$r, $g\\};",
[]>;
-def SUST_B_1D_ARRAY_V2B16_TRAP
+def SUST_B_1D_ARRAY_V2B16_ZERO
: NVPTXInst<(outs),
(ins Int64Regs:$s, Int32Regs:$idx, Int32Regs:$x, Int16Regs:$r,
Int16Regs:$g),
- "sust.b.a1d.v2.b16.trap \t[$s, \\{$idx, $x\\}], \\{$r, $g\\};",
+ "sust.b.a1d.v2.b16.zero \t[$s, \\{$idx, $x\\}], \\{$r, $g\\};",
[]>;
-def SUST_B_1D_ARRAY_V2B32_TRAP
+def SUST_B_1D_ARRAY_V2B32_ZERO
: NVPTXInst<(outs),
(ins Int64Regs:$s, Int32Regs:$idx, Int32Regs:$x, Int32Regs:$r,
Int32Regs:$g),
- "sust.b.a1d.v2.b32.trap \t[$s, \\{$idx, $x\\}], \\{$r, $g\\};",
+ "sust.b.a1d.v2.b32.zero \t[$s, \\{$idx, $x\\}], \\{$r, $g\\};",
[]>;
-def SUST_B_1D_ARRAY_V4B8_TRAP
+def SUST_B_1D_ARRAY_V2B64_ZERO
+ : NVPTXInst<(outs),
+ (ins Int64Regs:$s, Int32Regs:$idx, Int32Regs:$x, Int64Regs:$r,
+ Int64Regs:$g),
+ "sust.b.a1d.v2.b64.zero \t[$s, \\{$idx, $x\\}], \\{$r, $g\\};",
+ []>;
+def SUST_B_1D_ARRAY_V4B8_ZERO
: NVPTXInst<(outs),
(ins Int64Regs:$s, Int32Regs:$idx, Int32Regs:$x, Int16Regs:$r,
Int16Regs:$g, Int16Regs:$b, Int16Regs:$a),
- "sust.b.a1d.v4.b8.trap \t[$s, \\{$idx, $x\\}], "
+ "sust.b.a1d.v4.b8.zero \t[$s, \\{$idx, $x\\}], "
"\\{$r, $g, $b, $a\\};",
[]>;
-def SUST_B_1D_ARRAY_V4B16_TRAP
+def SUST_B_1D_ARRAY_V4B16_ZERO
: NVPTXInst<(outs),
(ins Int64Regs:$s, Int32Regs:$idx, Int32Regs:$x, Int16Regs:$r,
Int16Regs:$g, Int16Regs:$b, Int16Regs:$a),
- "sust.b.a1d.v4.b16.trap \t[$s, \\{$idx, $x\\}], "
+ "sust.b.a1d.v4.b16.zero \t[$s, \\{$idx, $x\\}], "
"\\{$r, $g, $b, $a\\};",
[]>;
-def SUST_B_1D_ARRAY_V4B32_TRAP
+def SUST_B_1D_ARRAY_V4B32_ZERO
: NVPTXInst<(outs),
(ins Int64Regs:$s, Int32Regs:$idx, Int32Regs:$x, Int32Regs:$r,
Int32Regs:$g, Int32Regs:$b, Int32Regs:$a),
- "sust.b.a1d.v4.b32.trap \t[$s, \\{$idx, $x\\}], "
+ "sust.b.a1d.v4.b32.zero \t[$s, \\{$idx, $x\\}], "
"\\{$r, $g, $b, $a\\};",
[]>;
-def SUST_B_2D_B8_TRAP
+def SUST_B_2D_B8_ZERO
: NVPTXInst<(outs),
(ins Int64Regs:$s, Int32Regs:$x, Int32Regs:$y, Int16Regs:$r),
- "sust.b.2d.b8.trap \t[$s, \\{$x, $y\\}], \\{$r\\};",
+ "sust.b.2d.b8.zero \t[$s, \\{$x, $y\\}], \\{$r\\};",
[]>;
-def SUST_B_2D_B16_TRAP
+def SUST_B_2D_B16_ZERO
: NVPTXInst<(outs),
(ins Int64Regs:$s, Int32Regs:$x, Int32Regs:$y, Int16Regs:$r),
- "sust.b.2d.b16.trap \t[$s, \\{$x, $y\\}], \\{$r\\};",
+ "sust.b.2d.b16.zero \t[$s, \\{$x, $y\\}], \\{$r\\};",
[]>;
-def SUST_B_2D_B32_TRAP
+def SUST_B_2D_B32_ZERO
: NVPTXInst<(outs),
(ins Int64Regs:$s, Int32Regs:$x, Int32Regs:$y, Int32Regs:$r),
- "sust.b.2d.b32.trap \t[$s, \\{$x, $y\\}], \\{$r\\};",
+ "sust.b.2d.b32.zero \t[$s, \\{$x, $y\\}], \\{$r\\};",
[]>;
-def SUST_B_2D_V2B8_TRAP
+def SUST_B_2D_B64_ZERO
+ : NVPTXInst<(outs),
+ (ins Int64Regs:$s, Int32Regs:$x, Int32Regs:$y, Int64Regs:$r),
+ "sust.b.2d.b64.zero \t[$s, \\{$x, $y\\}], \\{$r\\};",
+ []>;
+def SUST_B_2D_V2B8_ZERO
: NVPTXInst<(outs),
(ins Int64Regs:$s, Int32Regs:$x, Int32Regs:$y, Int16Regs:$r,
Int16Regs:$g),
- "sust.b.2d.v2.b8.trap \t[$s, \\{$x, $y\\}], \\{$r, $g\\};",
+ "sust.b.2d.v2.b8.zero \t[$s, \\{$x, $y\\}], \\{$r, $g\\};",
[]>;
-def SUST_B_2D_V2B16_TRAP
+def SUST_B_2D_V2B16_ZERO
: NVPTXInst<(outs),
(ins Int64Regs:$s, Int32Regs:$x, Int32Regs:$y, Int16Regs:$r,
Int16Regs:$g),
- "sust.b.2d.v2.b16.trap \t[$s, \\{$x, $y\\}], \\{$r, $g\\};",
+ "sust.b.2d.v2.b16.zero \t[$s, \\{$x, $y\\}], \\{$r, $g\\};",
[]>;
-def SUST_B_2D_V2B32_TRAP
+def SUST_B_2D_V2B32_ZERO
: NVPTXInst<(outs),
(ins Int64Regs:$s, Int32Regs:$x, Int32Regs:$y, Int32Regs:$r,
Int32Regs:$g),
- "sust.b.2d.v2.b32.trap \t[$s, \\{$x, $y\\}], \\{$r, $g\\};",
+ "sust.b.2d.v2.b32.zero \t[$s, \\{$x, $y\\}], \\{$r, $g\\};",
[]>;
-def SUST_B_2D_V4B8_TRAP
+def SUST_B_2D_V2B64_ZERO
+ : NVPTXInst<(outs),
+ (ins Int64Regs:$s, Int32Regs:$x, Int32Regs:$y, Int64Regs:$r,
+ Int64Regs:$g),
+ "sust.b.2d.v2.b64.zero \t[$s, \\{$x, $y\\}], \\{$r, $g\\};",
+ []>;
+def SUST_B_2D_V4B8_ZERO
: NVPTXInst<(outs),
(ins Int64Regs:$s, Int32Regs:$x, Int32Regs:$y, Int16Regs:$r,
Int16Regs:$g, Int16Regs:$b, Int16Regs:$a),
- "sust.b.2d.v4.b8.trap \t[$s, \\{$x, $y\\}], "
+ "sust.b.2d.v4.b8.zero \t[$s, \\{$x, $y\\}], "
"\\{$r, $g, $b, $a\\};",
[]>;
-def SUST_B_2D_V4B16_TRAP
+def SUST_B_2D_V4B16_ZERO
: NVPTXInst<(outs),
(ins Int64Regs:$s, Int32Regs:$x, Int32Regs:$y, Int16Regs:$r,
Int16Regs:$g, Int16Regs:$b, Int16Regs:$a),
- "sust.b.2d.v4.b16.trap \t[$s, \\{$x, $y\\}], "
+ "sust.b.2d.v4.b16.zero \t[$s, \\{$x, $y\\}], "
"\\{$r, $g, $b, $a\\};",
[]>;
-def SUST_B_2D_V4B32_TRAP
+def SUST_B_2D_V4B32_ZERO
: NVPTXInst<(outs),
(ins Int64Regs:$s, Int32Regs:$x, Int32Regs:$y, Int32Regs:$r,
Int32Regs:$g, Int32Regs:$b, Int32Regs:$a),
- "sust.b.2d.v4.b32.trap \t[$s, \\{$x, $y\\}], "
+ "sust.b.2d.v4.b32.zero \t[$s, \\{$x, $y\\}], "
"\\{$r, $g, $b, $a\\};",
[]>;
-def SUST_B_2D_ARRAY_B8_TRAP
+def SUST_B_2D_ARRAY_B8_ZERO
: NVPTXInst<(outs),
(ins Int64Regs:$s, Int32Regs:$idx, Int32Regs:$x, Int32Regs:$y,
Int16Regs:$r),
- "sust.b.a2d.b8.trap \t[$s, \\{$idx, $x, $y, $y\\}], \\{$r\\};",
+ "sust.b.a2d.b8.zero \t[$s, \\{$idx, $x, $y, $y\\}], \\{$r\\};",
[]>;
-def SUST_B_2D_ARRAY_B16_TRAP
+def SUST_B_2D_ARRAY_B16_ZERO
: NVPTXInst<(outs),
(ins Int64Regs:$s, Int32Regs:$idx, Int32Regs:$x, Int32Regs:$y,
Int16Regs:$r),
- "sust.b.a2d.b16.trap \t[$s, \\{$idx, $x, $y, $y\\}], \\{$r\\};",
+ "sust.b.a2d.b16.zero \t[$s, \\{$idx, $x, $y, $y\\}], \\{$r\\};",
[]>;
-def SUST_B_2D_ARRAY_B32_TRAP
+def SUST_B_2D_ARRAY_B32_ZERO
: NVPTXInst<(outs),
(ins Int64Regs:$s, Int32Regs:$idx, Int32Regs:$x, Int32Regs:$y,
Int32Regs:$r),
- "sust.b.a2d.b32.trap \t[$s, \\{$idx, $x, $y, $y\\}], \\{$r\\};",
+ "sust.b.a2d.b32.zero \t[$s, \\{$idx, $x, $y, $y\\}], \\{$r\\};",
[]>;
-def SUST_B_2D_ARRAY_V2B8_TRAP
+def SUST_B_2D_ARRAY_B64_ZERO
+ : NVPTXInst<(outs),
+ (ins Int64Regs:$s, Int32Regs:$idx, Int32Regs:$x, Int32Regs:$y,
+ Int64Regs:$r),
+ "sust.b.a2d.b64.zero \t[$s, \\{$idx, $x, $y, $y\\}], \\{$r\\};",
+ []>;
+def SUST_B_2D_ARRAY_V2B8_ZERO
: NVPTXInst<(outs),
(ins Int64Regs:$s, Int32Regs:$idx, Int32Regs:$x, Int32Regs:$y,
Int16Regs:$r, Int16Regs:$g),
- "sust.b.a2d.v2.b8.trap \t[$s, \\{$idx, $x, $y, $y\\}], "
+ "sust.b.a2d.v2.b8.zero \t[$s, \\{$idx, $x, $y, $y\\}], "
"\\{$r, $g\\};",
[]>;
-def SUST_B_2D_ARRAY_V2B16_TRAP
+def SUST_B_2D_ARRAY_V2B16_ZERO
: NVPTXInst<(outs),
(ins Int64Regs:$s, Int32Regs:$idx, Int32Regs:$x, Int32Regs:$y,
Int16Regs:$r, Int16Regs:$g),
- "sust.b.a2d.v2.b16.trap \t[$s, \\{$idx, $x, $y, $y\\}], "
+ "sust.b.a2d.v2.b16.zero \t[$s, \\{$idx, $x, $y, $y\\}], "
"\\{$r, $g\\};",
[]>;
-def SUST_B_2D_ARRAY_V2B32_TRAP
+def SUST_B_2D_ARRAY_V2B32_ZERO
: NVPTXInst<(outs),
(ins Int64Regs:$s, Int32Regs:$idx, Int32Regs:$x, Int32Regs:$y,
Int32Regs:$r, Int32Regs:$g),
- "sust.b.a2d.v2.b32.trap \t[$s, \\{$idx, $x, $y, $y\\}], "
+ "sust.b.a2d.v2.b32.zero \t[$s, \\{$idx, $x, $y, $y\\}], "
"\\{$r, $g\\};",
[]>;
-def SUST_B_2D_ARRAY_V4B8_TRAP
+def SUST_B_2D_ARRAY_V2B64_ZERO
+ : NVPTXInst<(outs),
+ (ins Int64Regs:$s, Int32Regs:$idx, Int32Regs:$x, Int32Regs:$y,
+ Int64Regs:$r, Int64Regs:$g),
+ "sust.b.a2d.v2.b64.zero \t[$s, \\{$idx, $x, $y, $y\\}], "
+ "\\{$r, $g\\};",
+ []>;
+def SUST_B_2D_ARRAY_V4B8_ZERO
: NVPTXInst<(outs),
(ins Int64Regs:$s, Int32Regs:$idx, Int32Regs:$x, Int32Regs:$y,
Int16Regs:$r, Int16Regs:$g, Int16Regs:$b, Int16Regs:$a),
- "sust.b.a2d.v4.b8.trap \t[$s, \\{$idx, $x, $y, $y\\}], "
+ "sust.b.a2d.v4.b8.zero \t[$s, \\{$idx, $x, $y, $y\\}], "
"\\{$r, $g, $b, $a\\};",
[]>;
-def SUST_B_2D_ARRAY_V4B16_TRAP
+def SUST_B_2D_ARRAY_V4B16_ZERO
: NVPTXInst<(outs),
(ins Int64Regs:$s, Int32Regs:$idx, Int32Regs:$x, Int32Regs:$y,
Int16Regs:$r, Int16Regs:$g, Int16Regs:$b, Int16Regs:$a),
- "sust.b.a2d.v4.b16.trap \t[$s, \\{$idx, $x, $y, $y\\}], "
+ "sust.b.a2d.v4.b16.zero \t[$s, \\{$idx, $x, $y, $y\\}], "
"\\{$r, $g, $b, $a\\};",
[]>;
-def SUST_B_2D_ARRAY_V4B32_TRAP
+def SUST_B_2D_ARRAY_V4B32_ZERO
: NVPTXInst<(outs),
(ins Int64Regs:$s, Int32Regs:$idx, Int32Regs:$x, Int32Regs:$y,
Int32Regs:$r, Int32Regs:$g, Int32Regs:$b, Int32Regs:$a),
- "sust.b.a2d.v4.b32.trap \t[$s, \\{$idx, $x, $y, $y\\}], "
+ "sust.b.a2d.v4.b32.zero \t[$s, \\{$idx, $x, $y, $y\\}], "
"\\{$r, $g, $b, $a\\};",
[]>;
-def SUST_B_3D_B8_TRAP
+def SUST_B_3D_B8_ZERO
: NVPTXInst<(outs),
(ins Int64Regs:$s, Int32Regs:$x, Int32Regs:$y, Int32Regs:$z,
Int16Regs:$r),
- "sust.b.3d.b8.trap \t[$s, \\{$x, $y, $z, $z\\}], \\{$r\\};",
+ "sust.b.3d.b8.zero \t[$s, \\{$x, $y, $z, $z\\}], \\{$r\\};",
[]>;
-def SUST_B_3D_B16_TRAP
+def SUST_B_3D_B16_ZERO
: NVPTXInst<(outs),
(ins Int64Regs:$s, Int32Regs:$x, Int32Regs:$y, Int32Regs:$z,
Int16Regs:$r),
- "sust.b.3d.b16.trap \t[$s, \\{$x, $y, $z, $z\\}], \\{$r\\};",
+ "sust.b.3d.b16.zero \t[$s, \\{$x, $y, $z, $z\\}], \\{$r\\};",
[]>;
-def SUST_B_3D_B32_TRAP
+def SUST_B_3D_B32_ZERO
: NVPTXInst<(outs),
(ins Int64Regs:$s, Int32Regs:$x, Int32Regs:$y, Int32Regs:$z,
Int32Regs:$r),
- "sust.b.3d.b32.trap \t[$s, \\{$x, $y, $z, $z\\}], \\{$r\\};",
+ "sust.b.3d.b32.zero \t[$s, \\{$x, $y, $z, $z\\}], \\{$r\\};",
[]>;
-def SUST_B_3D_V2B8_TRAP
+def SUST_B_3D_B64_ZERO
+ : NVPTXInst<(outs),
+ (ins Int64Regs:$s, Int32Regs:$x, Int32Regs:$y, Int32Regs:$z,
+ Int64Regs:$r),
+ "sust.b.3d.b64.zero \t[$s, \\{$x, $y, $z, $z\\}], \\{$r\\};",
+ []>;
+def SUST_B_3D_V2B8_ZERO
: NVPTXInst<(outs),
(ins Int64Regs:$s, Int32Regs:$x, Int32Regs:$y, Int32Regs:$z,
Int16Regs:$r, Int16Regs:$g),
- "sust.b.3d.v2.b8.trap \t[$s, \\{$x, $y, $z, $z\\}], "
+ "sust.b.3d.v2.b8.zero \t[$s, \\{$x, $y, $z, $z\\}], "
"\\{$r, $g\\};",
[]>;
-def SUST_B_3D_V2B16_TRAP
+def SUST_B_3D_V2B16_ZERO
: NVPTXInst<(outs),
(ins Int64Regs:$s, Int32Regs:$x, Int32Regs:$y, Int32Regs:$z,
Int16Regs:$r, Int16Regs:$g),
- "sust.b.3d.v2.b16.trap \t[$s, \\{$x, $y, $z, $z\\}], "
+ "sust.b.3d.v2.b16.zero \t[$s, \\{$x, $y, $z, $z\\}], "
"\\{$r, $g\\};",
[]>;
-def SUST_B_3D_V2B32_TRAP
+def SUST_B_3D_V2B32_ZERO
: NVPTXInst<(outs),
(ins Int64Regs:$s, Int32Regs:$x, Int32Regs:$y, Int32Regs:$z,
Int32Regs:$r, Int32Regs:$g),
- "sust.b.3d.v2.b32.trap \t[$s, \\{$x, $y, $z, $z\\}], "
+ "sust.b.3d.v2.b32.zero \t[$s, \\{$x, $y, $z, $z\\}], "
"\\{$r, $g\\};",
[]>;
-def SUST_B_3D_V4B8_TRAP
+def SUST_B_3D_V2B64_ZERO
+ : NVPTXInst<(outs),
+ (ins Int64Regs:$s, Int32Regs:$x, Int32Regs:$y, Int32Regs:$z,
+ Int64Regs:$r, Int64Regs:$g),
+ "sust.b.3d.v2.b64.zero \t[$s, \\{$x, $y, $z, $z\\}], "
+ "\\{$r, $g\\};",
+ []>;
+def SUST_B_3D_V4B8_ZERO
: NVPTXInst<(outs),
(ins Int64Regs:$s, Int32Regs:$x, Int32Regs:$y, Int32Regs:$z,
Int16Regs:$r, Int16Regs:$g, Int16Regs:$b, Int16Regs:$a),
- "sust.b.3d.v4.b8.trap \t[$s, \\{$x, $y, $z, $z\\}], "
+ "sust.b.3d.v4.b8.zero \t[$s, \\{$x, $y, $z, $z\\}], "
"\\{$r, $g, $b, $a\\};",
[]>;
-def SUST_B_3D_V4B16_TRAP
+def SUST_B_3D_V4B16_ZERO
: NVPTXInst<(outs),
(ins Int64Regs:$s, Int32Regs:$x, Int32Regs:$y, Int32Regs:$z,
Int16Regs:$r, Int16Regs:$g, Int16Regs:$b, Int16Regs:$a),
- "sust.b.3d.v4.b16.trap \t[$s, \\{$x, $y, $z, $z\\}], "
+ "sust.b.3d.v4.b16.zero \t[$s, \\{$x, $y, $z, $z\\}], "
"\\{$r, $g, $b, $a\\};",
[]>;
-def SUST_B_3D_V4B32_TRAP
+def SUST_B_3D_V4B32_ZERO
: NVPTXInst<(outs),
(ins Int64Regs:$s, Int32Regs:$x, Int32Regs:$y, Int32Regs:$z,
Int32Regs:$r, Int32Regs:$g, Int32Regs:$b, Int32Regs:$a),
- "sust.b.3d.v4.b32.trap \t[$s, \\{$x, $y, $z, $z\\}], "
+ "sust.b.3d.v4.b32.zero \t[$s, \\{$x, $y, $z, $z\\}], "
"\\{$r, $g, $b, $a\\};",
[]>;
+
+
// Formatted
def SUST_P_1D_B8_TRAP
"sust.p.3d.v4.b32.trap \t[$s, \\{$x, $y, $z, $z\\}], "
"\\{$r, $g, $b, $a\\};",
[]>;
+}
+
+// Surface store instruction patterns
+// I'm not sure why we can't just include these in the instruction definitions,
+// but TableGen complains of type errors :(
+
+// .clamp variant
+def : Pat<(int_nvvm_sust_b_1d_i8_clamp
+ Int64Regs:$s, Int32Regs:$x, Int16Regs:$r),
+ (SUST_B_1D_B8_CLAMP Int64Regs:$s, Int32Regs:$x, Int16Regs:$r)>;
+
+def : Pat<(int_nvvm_sust_b_1d_i16_clamp
+ Int64Regs:$s, Int32Regs:$x, Int16Regs:$r),
+ (SUST_B_1D_B16_CLAMP Int64Regs:$s, Int32Regs:$x, Int16Regs:$r)>;
+
+def : Pat<(int_nvvm_sust_b_1d_i32_clamp
+ Int64Regs:$s, Int32Regs:$x, Int32Regs:$r),
+ (SUST_B_1D_B32_CLAMP Int64Regs:$s, Int32Regs:$x, Int32Regs:$r)>;
+
+def : Pat<(int_nvvm_sust_b_1d_i64_clamp
+ Int64Regs:$s, Int32Regs:$x, Int64Regs:$r),
+ (SUST_B_1D_B64_CLAMP Int64Regs:$s, Int32Regs:$x, Int64Regs:$r)>;
+
+def : Pat<(int_nvvm_sust_b_1d_v2i8_clamp
+ Int64Regs:$s, Int32Regs:$x, Int16Regs:$r, Int16Regs:$g),
+ (SUST_B_1D_V2B8_CLAMP Int64Regs:$s, Int32Regs:$x,
+ Int16Regs:$r, Int16Regs:$g)>;
+
+def : Pat<(int_nvvm_sust_b_1d_v2i16_clamp
+ Int64Regs:$s, Int32Regs:$x, Int16Regs:$r, Int16Regs:$g),
+ (SUST_B_1D_V2B16_CLAMP Int64Regs:$s, Int32Regs:$x,
+ Int16Regs:$r, Int16Regs:$g)>;
+
+def : Pat<(int_nvvm_sust_b_1d_v2i32_clamp
+ Int64Regs:$s, Int32Regs:$x, Int32Regs:$r, Int32Regs:$g),
+ (SUST_B_1D_V2B32_CLAMP Int64Regs:$s, Int32Regs:$x,
+ Int32Regs:$r, Int32Regs:$g)>;
+
+def : Pat<(int_nvvm_sust_b_1d_v2i64_clamp
+ Int64Regs:$s, Int32Regs:$x, Int64Regs:$r, Int64Regs:$g),
+ (SUST_B_1D_V2B64_CLAMP Int64Regs:$s, Int32Regs:$x,
+ Int64Regs:$r, Int64Regs:$g)>;
+
+def : Pat<(int_nvvm_sust_b_1d_v4i8_clamp
+ Int64Regs:$s, Int32Regs:$x,
+ Int16Regs:$r, Int16Regs:$g, Int16Regs:$b, Int16Regs:$a),
+ (SUST_B_1D_V4B8_CLAMP Int64Regs:$s, Int32Regs:$x,
+ Int16Regs:$r, Int16Regs:$g, Int16Regs:$b, Int16Regs:$a)>;
+
+def : Pat<(int_nvvm_sust_b_1d_v4i16_clamp
+ Int64Regs:$s, Int32Regs:$x,
+ Int16Regs:$r, Int16Regs:$g, Int16Regs:$b, Int16Regs:$a),
+ (SUST_B_1D_V4B16_CLAMP Int64Regs:$s, Int32Regs:$x,
+ Int16Regs:$r, Int16Regs:$g, Int16Regs:$b, Int16Regs:$a)>;
+
+def : Pat<(int_nvvm_sust_b_1d_v4i32_clamp
+ Int64Regs:$s, Int32Regs:$x,
+ Int32Regs:$r, Int32Regs:$g, Int32Regs:$b, Int32Regs:$a),
+ (SUST_B_1D_V4B32_CLAMP Int64Regs:$s, Int32Regs:$x,
+ Int32Regs:$r, Int32Regs:$g, Int32Regs:$b, Int32Regs:$a)>;
+
+
+
+def : Pat<(int_nvvm_sust_b_1d_array_i8_clamp
+ Int64Regs:$s, Int32Regs:$l, Int32Regs:$x, Int16Regs:$r),
+ (SUST_B_1D_ARRAY_B8_CLAMP Int64Regs:$s, Int32Regs:$l, Int32Regs:$x,
+ Int16Regs:$r)>;
+
+def : Pat<(int_nvvm_sust_b_1d_array_i16_clamp
+ Int64Regs:$s, Int32Regs:$l, Int32Regs:$x, Int16Regs:$r),
+ (SUST_B_1D_ARRAY_B16_CLAMP Int64Regs:$s, Int32Regs:$l, Int32Regs:$x,
+ Int16Regs:$r)>;
+
+def : Pat<(int_nvvm_sust_b_1d_array_i32_clamp
+ Int64Regs:$s, Int32Regs:$l, Int32Regs:$x, Int32Regs:$r),
+ (SUST_B_1D_ARRAY_B32_CLAMP Int64Regs:$s, Int32Regs:$l, Int32Regs:$x,
+ Int32Regs:$r)>;
+
+def : Pat<(int_nvvm_sust_b_1d_array_i64_clamp
+ Int64Regs:$s, Int32Regs:$l, Int32Regs:$x, Int64Regs:$r),
+ (SUST_B_1D_ARRAY_B64_CLAMP Int64Regs:$s, Int32Regs:$l, Int32Regs:$x,
+ Int64Regs:$r)>;
+
+def : Pat<(int_nvvm_sust_b_1d_array_v2i8_clamp
+ Int64Regs:$s, Int32Regs:$l, Int32Regs:$x, Int16Regs:$r, Int16Regs:$g),
+ (SUST_B_1D_ARRAY_V2B8_CLAMP Int64Regs:$s, Int32Regs:$l, Int32Regs:$x,
+ Int16Regs:$r, Int16Regs:$g)>;
+
+def : Pat<(int_nvvm_sust_b_1d_array_v2i16_clamp
+ Int64Regs:$s, Int32Regs:$l, Int32Regs:$x, Int16Regs:$r, Int16Regs:$g),
+ (SUST_B_1D_ARRAY_V2B16_CLAMP Int64Regs:$s, Int32Regs:$l, Int32Regs:$x,
+ Int16Regs:$r, Int16Regs:$g)>;
+
+def : Pat<(int_nvvm_sust_b_1d_array_v2i32_clamp
+ Int64Regs:$s, Int32Regs:$l, Int32Regs:$x, Int32Regs:$r, Int32Regs:$g),
+ (SUST_B_1D_ARRAY_V2B32_CLAMP Int64Regs:$s, Int32Regs:$l, Int32Regs:$x,
+ Int32Regs:$r, Int32Regs:$g)>;
+
+def : Pat<(int_nvvm_sust_b_1d_array_v2i64_clamp
+ Int64Regs:$s, Int32Regs:$l, Int32Regs:$x, Int64Regs:$r, Int64Regs:$g),
+ (SUST_B_1D_ARRAY_V2B64_CLAMP Int64Regs:$s, Int32Regs:$l, Int32Regs:$x,
+ Int64Regs:$r, Int64Regs:$g)>;
+
+def : Pat<(int_nvvm_sust_b_1d_array_v4i8_clamp
+ Int64Regs:$s, Int32Regs:$l, Int32Regs:$x,
+ Int16Regs:$r, Int16Regs:$g, Int16Regs:$b, Int16Regs:$a),
+ (SUST_B_1D_ARRAY_V4B8_CLAMP Int64Regs:$s, Int32Regs:$l, Int32Regs:$x,
+ Int16Regs:$r, Int16Regs:$g, Int16Regs:$b, Int16Regs:$a)>;
+
+def : Pat<(int_nvvm_sust_b_1d_array_v4i16_clamp
+ Int64Regs:$s, Int32Regs:$l, Int32Regs:$x,
+ Int16Regs:$r, Int16Regs:$g, Int16Regs:$b, Int16Regs:$a),
+ (SUST_B_1D_ARRAY_V4B16_CLAMP Int64Regs:$s, Int32Regs:$l, Int32Regs:$x,
+ Int16Regs:$r, Int16Regs:$g, Int16Regs:$b, Int16Regs:$a)>;
+
+def : Pat<(int_nvvm_sust_b_1d_array_v4i32_clamp
+ Int64Regs:$s, Int32Regs:$l, Int32Regs:$x,
+ Int32Regs:$r, Int32Regs:$g, Int32Regs:$b, Int32Regs:$a),
+ (SUST_B_1D_ARRAY_V4B32_CLAMP Int64Regs:$s, Int32Regs:$l, Int32Regs:$x,
+ Int32Regs:$r, Int32Regs:$g, Int32Regs:$b, Int32Regs:$a)>;
+
+
+
+def : Pat<(int_nvvm_sust_b_2d_i8_clamp
+ Int64Regs:$s, Int32Regs:$x, Int32Regs:$y, Int16Regs:$r),
+ (SUST_B_2D_B8_CLAMP Int64Regs:$s, Int32Regs:$x, Int32Regs:$y,
+ Int16Regs:$r)>;
+
+def : Pat<(int_nvvm_sust_b_2d_i16_clamp
+ Int64Regs:$s, Int32Regs:$x, Int32Regs:$y, Int16Regs:$r),
+ (SUST_B_2D_B16_CLAMP Int64Regs:$s, Int32Regs:$x, Int32Regs:$y,
+ Int16Regs:$r)>;
+
+def : Pat<(int_nvvm_sust_b_2d_i32_clamp
+ Int64Regs:$s, Int32Regs:$x, Int32Regs:$y, Int32Regs:$r),
+ (SUST_B_2D_B32_CLAMP Int64Regs:$s, Int32Regs:$x, Int32Regs:$y,
+ Int32Regs:$r)>;
+
+def : Pat<(int_nvvm_sust_b_2d_i64_clamp
+ Int64Regs:$s, Int32Regs:$x, Int32Regs:$y, Int64Regs:$r),
+ (SUST_B_2D_B64_CLAMP Int64Regs:$s, Int32Regs:$x, Int32Regs:$y,
+ Int64Regs:$r)>;
+
+def : Pat<(int_nvvm_sust_b_2d_v2i8_clamp
+ Int64Regs:$s, Int32Regs:$x, Int32Regs:$y, Int16Regs:$r, Int16Regs:$g),
+ (SUST_B_2D_V2B8_CLAMP Int64Regs:$s, Int32Regs:$x, Int32Regs:$y,
+ Int16Regs:$r, Int16Regs:$g)>;
+
+def : Pat<(int_nvvm_sust_b_2d_v2i16_clamp
+ Int64Regs:$s, Int32Regs:$x, Int32Regs:$y, Int16Regs:$r, Int16Regs:$g),
+ (SUST_B_2D_V2B16_CLAMP Int64Regs:$s, Int32Regs:$x, Int32Regs:$y,
+ Int16Regs:$r, Int16Regs:$g)>;
+
+def : Pat<(int_nvvm_sust_b_2d_v2i32_clamp
+ Int64Regs:$s, Int32Regs:$x, Int32Regs:$y, Int32Regs:$r, Int32Regs:$g),
+ (SUST_B_2D_V2B32_CLAMP Int64Regs:$s, Int32Regs:$x, Int32Regs:$y,
+ Int32Regs:$r, Int32Regs:$g)>;
+
+def : Pat<(int_nvvm_sust_b_2d_v2i64_clamp
+ Int64Regs:$s, Int32Regs:$x, Int32Regs:$y, Int64Regs:$r, Int64Regs:$g),
+ (SUST_B_2D_V2B64_CLAMP Int64Regs:$s, Int32Regs:$x, Int32Regs:$y,
+ Int64Regs:$r, Int64Regs:$g)>;
+
+def : Pat<(int_nvvm_sust_b_2d_v4i8_clamp
+ Int64Regs:$s, Int32Regs:$x, Int32Regs:$y,
+ Int16Regs:$r, Int16Regs:$g, Int16Regs:$b, Int16Regs:$a),
+ (SUST_B_2D_V4B8_CLAMP Int64Regs:$s, Int32Regs:$x, Int32Regs:$y,
+ Int16Regs:$r, Int16Regs:$g, Int16Regs:$b, Int16Regs:$a)>;
+
+def : Pat<(int_nvvm_sust_b_2d_v4i16_clamp
+ Int64Regs:$s, Int32Regs:$x, Int32Regs:$y,
+ Int16Regs:$r, Int16Regs:$g, Int16Regs:$b, Int16Regs:$a),
+ (SUST_B_2D_V4B16_CLAMP Int64Regs:$s, Int32Regs:$x, Int32Regs:$y,
+ Int16Regs:$r, Int16Regs:$g, Int16Regs:$b, Int16Regs:$a)>;
+
+def : Pat<(int_nvvm_sust_b_2d_v4i32_clamp
+ Int64Regs:$s, Int32Regs:$x, Int32Regs:$y,
+ Int32Regs:$r, Int32Regs:$g, Int32Regs:$b, Int32Regs:$a),
+ (SUST_B_2D_V4B32_CLAMP Int64Regs:$s, Int32Regs:$x, Int32Regs:$y,
+ Int32Regs:$r, Int32Regs:$g, Int32Regs:$b, Int32Regs:$a)>;
+
+
+
+def : Pat<(int_nvvm_sust_b_2d_array_i8_clamp
+ Int64Regs:$s, Int32Regs:$l, Int32Regs:$x, Int32Regs:$y, Int16Regs:$r),
+ (SUST_B_2D_ARRAY_B8_CLAMP Int64Regs:$s,
+ Int32Regs:$l, Int32Regs:$x, Int32Regs:$y,
+ Int16Regs:$r)>;
+
+def : Pat<(int_nvvm_sust_b_2d_array_i16_clamp
+ Int64Regs:$s, Int32Regs:$l, Int32Regs:$x, Int32Regs:$y, Int16Regs:$r),
+ (SUST_B_2D_ARRAY_B16_CLAMP Int64Regs:$s,
+ Int32Regs:$l, Int32Regs:$x, Int32Regs:$y,
+ Int16Regs:$r)>;
+
+def : Pat<(int_nvvm_sust_b_2d_array_i32_clamp
+ Int64Regs:$s, Int32Regs:$l, Int32Regs:$x, Int32Regs:$y, Int32Regs:$r),
+ (SUST_B_2D_ARRAY_B32_CLAMP Int64Regs:$s,
+ Int32Regs:$l, Int32Regs:$x, Int32Regs:$y,
+ Int32Regs:$r)>;
+
+def : Pat<(int_nvvm_sust_b_2d_array_i64_clamp
+ Int64Regs:$s, Int32Regs:$l, Int32Regs:$x, Int32Regs:$y, Int64Regs:$r),
+ (SUST_B_2D_ARRAY_B64_CLAMP Int64Regs:$s,
+ Int32Regs:$l, Int32Regs:$x, Int32Regs:$y,
+ Int64Regs:$r)>;
+
+def : Pat<(int_nvvm_sust_b_2d_array_v2i8_clamp
+ Int64Regs:$s, Int32Regs:$l, Int32Regs:$x, Int32Regs:$y,
+ Int16Regs:$r, Int16Regs:$g),
+ (SUST_B_2D_ARRAY_V2B8_CLAMP Int64Regs:$s, Int32Regs:$l,
+ Int32Regs:$x, Int32Regs:$y,
+ Int16Regs:$r, Int16Regs:$g)>;
+
+def : Pat<(int_nvvm_sust_b_2d_array_v2i16_clamp
+ Int64Regs:$s, Int32Regs:$l, Int32Regs:$x, Int32Regs:$y,
+ Int16Regs:$r, Int16Regs:$g),
+ (SUST_B_2D_ARRAY_V2B16_CLAMP Int64Regs:$s, Int32Regs:$l,
+ Int32Regs:$x, Int32Regs:$y,
+ Int16Regs:$r, Int16Regs:$g)>;
+
+def : Pat<(int_nvvm_sust_b_2d_array_v2i32_clamp
+ Int64Regs:$s, Int32Regs:$l, Int32Regs:$x, Int32Regs:$y, Int32Regs:$r,
+ Int32Regs:$g),
+ (SUST_B_2D_ARRAY_V2B32_CLAMP Int64Regs:$s, Int32Regs:$l,
+ Int32Regs:$x, Int32Regs:$y, Int32Regs:$r, Int32Regs:$g)>;
+
+def : Pat<(int_nvvm_sust_b_2d_array_v2i64_clamp
+ Int64Regs:$s, Int32Regs:$l, Int32Regs:$x, Int32Regs:$y, Int64Regs:$r,
+ Int64Regs:$g),
+ (SUST_B_2D_ARRAY_V2B64_CLAMP Int64Regs:$s, Int32Regs:$l,
+ Int32Regs:$x, Int32Regs:$y, Int64Regs:$r, Int64Regs:$g)>;
+
+def : Pat<(int_nvvm_sust_b_2d_array_v4i8_clamp
+ Int64Regs:$s, Int32Regs:$l, Int32Regs:$x, Int32Regs:$y,
+ Int16Regs:$r, Int16Regs:$g, Int16Regs:$b, Int16Regs:$a),
+ (SUST_B_2D_ARRAY_V4B8_CLAMP Int64Regs:$s,
+ Int32Regs:$l, Int32Regs:$x, Int32Regs:$y,
+ Int16Regs:$r, Int16Regs:$g, Int16Regs:$b, Int16Regs:$a)>;
+
+def : Pat<(int_nvvm_sust_b_2d_array_v4i16_clamp
+ Int64Regs:$s, Int32Regs:$l, Int32Regs:$x, Int32Regs:$y,
+ Int16Regs:$r, Int16Regs:$g, Int16Regs:$b, Int16Regs:$a),
+ (SUST_B_2D_ARRAY_V4B16_CLAMP Int64Regs:$s,
+ Int32Regs:$l, Int32Regs:$x, Int32Regs:$y,
+ Int16Regs:$r, Int16Regs:$g, Int16Regs:$b, Int16Regs:$a)>;
+
+def : Pat<(int_nvvm_sust_b_2d_array_v4i32_clamp
+ Int64Regs:$s, Int32Regs:$l, Int32Regs:$x, Int32Regs:$y,
+ Int32Regs:$r, Int32Regs:$g, Int32Regs:$b, Int32Regs:$a),
+ (SUST_B_2D_ARRAY_V4B32_CLAMP Int64Regs:$s, Int32Regs:$l,
+ Int32Regs:$x, Int32Regs:$y,
+ Int32Regs:$r, Int32Regs:$g, Int32Regs:$b, Int32Regs:$a)>;
+
+
+
+def : Pat<(int_nvvm_sust_b_3d_i8_clamp
+ Int64Regs:$s, Int32Regs:$x, Int32Regs:$y, Int32Regs:$z,
+ Int16Regs:$r),
+ (SUST_B_3D_B8_CLAMP Int64Regs:$s,
+ Int32Regs:$x, Int32Regs:$y, Int32Regs:$z,
+ Int16Regs:$r)>;
+
+def : Pat<(int_nvvm_sust_b_3d_i16_clamp
+ Int64Regs:$s, Int32Regs:$x, Int32Regs:$y, Int32Regs:$z,
+ Int16Regs:$r),
+ (SUST_B_3D_B16_CLAMP Int64Regs:$s,
+ Int32Regs:$x, Int32Regs:$y, Int32Regs:$z,
+ Int16Regs:$r)>;
+def : Pat<(int_nvvm_sust_b_3d_i32_clamp
+ Int64Regs:$s, Int32Regs:$x, Int32Regs:$y, Int32Regs:$z,
+ Int32Regs:$r),
+ (SUST_B_3D_B32_CLAMP Int64Regs:$s,
+ Int32Regs:$x, Int32Regs:$y, Int32Regs:$z,
+ Int32Regs:$r)>;
+
+def : Pat<(int_nvvm_sust_b_3d_i64_clamp
+ Int64Regs:$s, Int32Regs:$x, Int32Regs:$y, Int32Regs:$z,
+ Int64Regs:$r),
+ (SUST_B_3D_B64_CLAMP Int64Regs:$s,
+ Int32Regs:$x, Int32Regs:$y, Int32Regs:$z,
+ Int64Regs:$r)>;
+
+def : Pat<(int_nvvm_sust_b_3d_v2i8_clamp
+ Int64Regs:$s, Int32Regs:$x, Int32Regs:$y, Int32Regs:$z,
+ Int16Regs:$r, Int16Regs:$g),
+ (SUST_B_3D_V2B8_CLAMP Int64Regs:$s,
+ Int32Regs:$x, Int32Regs:$y, Int32Regs:$z,
+ Int16Regs:$r, Int16Regs:$g)>;
+
+def : Pat<(int_nvvm_sust_b_3d_v2i16_clamp
+ Int64Regs:$s, Int32Regs:$x, Int32Regs:$y, Int32Regs:$z,
+ Int16Regs:$r, Int16Regs:$g),
+ (SUST_B_3D_V2B16_CLAMP Int64Regs:$s,
+ Int32Regs:$x, Int32Regs:$y, Int32Regs:$z,
+ Int16Regs:$r, Int16Regs:$g)>;
+
+def : Pat<(int_nvvm_sust_b_3d_v2i32_clamp
+ Int64Regs:$s, Int32Regs:$x, Int32Regs:$y, Int32Regs:$z,
+ Int32Regs:$r, Int32Regs:$g),
+ (SUST_B_3D_V2B32_CLAMP Int64Regs:$s,
+ Int32Regs:$x, Int32Regs:$y, Int32Regs:$z,
+ Int32Regs:$r, Int32Regs:$g)>;
+
+def : Pat<(int_nvvm_sust_b_3d_v2i64_clamp
+ Int64Regs:$s, Int32Regs:$x, Int32Regs:$y, Int32Regs:$z,
+ Int64Regs:$r, Int64Regs:$g),
+ (SUST_B_3D_V2B64_CLAMP Int64Regs:$s,
+ Int32Regs:$x, Int32Regs:$y, Int32Regs:$z,
+ Int64Regs:$r, Int64Regs:$g)>;
+
+def : Pat<(int_nvvm_sust_b_3d_v4i8_clamp
+ Int64Regs:$s, Int32Regs:$x, Int32Regs:$y, Int32Regs:$z,
+ Int16Regs:$r, Int16Regs:$g, Int16Regs:$b, Int16Regs:$a),
+ (SUST_B_3D_V4B8_CLAMP Int64Regs:$s,
+ Int32Regs:$x, Int32Regs:$y, Int32Regs:$z,
+ Int16Regs:$r, Int16Regs:$g, Int16Regs:$b, Int16Regs:$a)>;
+
+def : Pat<(int_nvvm_sust_b_3d_v4i16_clamp
+ Int64Regs:$s, Int32Regs:$x, Int32Regs:$y, Int32Regs:$z,
+ Int16Regs:$r, Int16Regs:$g, Int16Regs:$b, Int16Regs:$a),
+ (SUST_B_3D_V4B16_CLAMP Int64Regs:$s,
+ Int32Regs:$x, Int32Regs:$y, Int32Regs:$z,
+ Int16Regs:$r, Int16Regs:$g, Int16Regs:$b, Int16Regs:$a)>;
+
+def : Pat<(int_nvvm_sust_b_3d_v4i32_clamp
+ Int64Regs:$s, Int32Regs:$x, Int32Regs:$y, Int32Regs:$z,
+ Int32Regs:$r, Int32Regs:$g, Int32Regs:$b, Int32Regs:$a),
+ (SUST_B_3D_V4B32_CLAMP Int64Regs:$s,
+ Int32Regs:$x, Int32Regs:$y, Int32Regs:$z,
+ Int32Regs:$r, Int32Regs:$g, Int32Regs:$b, Int32Regs:$a)>;
-// Surface store instruction patterns
-// I'm not sure why we can't just include these in the instruction definitions,
-// but TableGen complains of type errors :(
+// .trap variant
def : Pat<(int_nvvm_sust_b_1d_i8_trap
Int64Regs:$s, Int32Regs:$x, Int16Regs:$r),
(SUST_B_1D_B8_TRAP Int64Regs:$s, Int32Regs:$x, Int16Regs:$r)>;
Int64Regs:$s, Int32Regs:$x, Int32Regs:$r),
(SUST_B_1D_B32_TRAP Int64Regs:$s, Int32Regs:$x, Int32Regs:$r)>;
+def : Pat<(int_nvvm_sust_b_1d_i64_trap
+ Int64Regs:$s, Int32Regs:$x, Int64Regs:$r),
+ (SUST_B_1D_B64_TRAP Int64Regs:$s, Int32Regs:$x, Int64Regs:$r)>;
+
def : Pat<(int_nvvm_sust_b_1d_v2i8_trap
Int64Regs:$s, Int32Regs:$x, Int16Regs:$r, Int16Regs:$g),
(SUST_B_1D_V2B8_TRAP Int64Regs:$s, Int32Regs:$x,
(SUST_B_1D_V2B32_TRAP Int64Regs:$s, Int32Regs:$x,
Int32Regs:$r, Int32Regs:$g)>;
+def : Pat<(int_nvvm_sust_b_1d_v2i64_trap
+ Int64Regs:$s, Int32Regs:$x, Int64Regs:$r, Int64Regs:$g),
+ (SUST_B_1D_V2B64_TRAP Int64Regs:$s, Int32Regs:$x,
+ Int64Regs:$r, Int64Regs:$g)>;
+
def : Pat<(int_nvvm_sust_b_1d_v4i8_trap
Int64Regs:$s, Int32Regs:$x,
Int16Regs:$r, Int16Regs:$g, Int16Regs:$b, Int16Regs:$a),
(SUST_B_1D_ARRAY_B32_TRAP Int64Regs:$s, Int32Regs:$l, Int32Regs:$x,
Int32Regs:$r)>;
+def : Pat<(int_nvvm_sust_b_1d_array_i64_trap
+ Int64Regs:$s, Int32Regs:$l, Int32Regs:$x, Int64Regs:$r),
+ (SUST_B_1D_ARRAY_B64_TRAP Int64Regs:$s, Int32Regs:$l, Int32Regs:$x,
+ Int64Regs:$r)>;
+
def : Pat<(int_nvvm_sust_b_1d_array_v2i8_trap
Int64Regs:$s, Int32Regs:$l, Int32Regs:$x, Int16Regs:$r, Int16Regs:$g),
(SUST_B_1D_ARRAY_V2B8_TRAP Int64Regs:$s, Int32Regs:$l, Int32Regs:$x,
(SUST_B_1D_ARRAY_V2B32_TRAP Int64Regs:$s, Int32Regs:$l, Int32Regs:$x,
Int32Regs:$r, Int32Regs:$g)>;
+def : Pat<(int_nvvm_sust_b_1d_array_v2i64_trap
+ Int64Regs:$s, Int32Regs:$l, Int32Regs:$x, Int64Regs:$r, Int64Regs:$g),
+ (SUST_B_1D_ARRAY_V2B64_TRAP Int64Regs:$s, Int32Regs:$l, Int32Regs:$x,
+ Int64Regs:$r, Int64Regs:$g)>;
+
def : Pat<(int_nvvm_sust_b_1d_array_v4i8_trap
Int64Regs:$s, Int32Regs:$l, Int32Regs:$x,
Int16Regs:$r, Int16Regs:$g, Int16Regs:$b, Int16Regs:$a),
(SUST_B_2D_B32_TRAP Int64Regs:$s, Int32Regs:$x, Int32Regs:$y,
Int32Regs:$r)>;
+def : Pat<(int_nvvm_sust_b_2d_i64_trap
+ Int64Regs:$s, Int32Regs:$x, Int32Regs:$y, Int64Regs:$r),
+ (SUST_B_2D_B64_TRAP Int64Regs:$s, Int32Regs:$x, Int32Regs:$y,
+ Int64Regs:$r)>;
+
def : Pat<(int_nvvm_sust_b_2d_v2i8_trap
Int64Regs:$s, Int32Regs:$x, Int32Regs:$y, Int16Regs:$r, Int16Regs:$g),
(SUST_B_2D_V2B8_TRAP Int64Regs:$s, Int32Regs:$x, Int32Regs:$y,
(SUST_B_2D_V2B32_TRAP Int64Regs:$s, Int32Regs:$x, Int32Regs:$y,
Int32Regs:$r, Int32Regs:$g)>;
+def : Pat<(int_nvvm_sust_b_2d_v2i64_trap
+ Int64Regs:$s, Int32Regs:$x, Int32Regs:$y, Int64Regs:$r, Int64Regs:$g),
+ (SUST_B_2D_V2B64_TRAP Int64Regs:$s, Int32Regs:$x, Int32Regs:$y,
+ Int64Regs:$r, Int64Regs:$g)>;
+
def : Pat<(int_nvvm_sust_b_2d_v4i8_trap
Int64Regs:$s, Int32Regs:$x, Int32Regs:$y,
Int16Regs:$r, Int16Regs:$g, Int16Regs:$b, Int16Regs:$a),
Int32Regs:$l, Int32Regs:$x, Int32Regs:$y,
Int32Regs:$r)>;
+def : Pat<(int_nvvm_sust_b_2d_array_i64_trap
+ Int64Regs:$s, Int32Regs:$l, Int32Regs:$x, Int32Regs:$y, Int64Regs:$r),
+ (SUST_B_2D_ARRAY_B64_TRAP Int64Regs:$s,
+ Int32Regs:$l, Int32Regs:$x, Int32Regs:$y,
+ Int64Regs:$r)>;
+
def : Pat<(int_nvvm_sust_b_2d_array_v2i8_trap
Int64Regs:$s, Int32Regs:$l, Int32Regs:$x, Int32Regs:$y,
Int16Regs:$r, Int16Regs:$g),
(SUST_B_2D_ARRAY_V2B32_TRAP Int64Regs:$s, Int32Regs:$l,
Int32Regs:$x, Int32Regs:$y, Int32Regs:$r, Int32Regs:$g)>;
+def : Pat<(int_nvvm_sust_b_2d_array_v2i64_trap
+ Int64Regs:$s, Int32Regs:$l, Int32Regs:$x, Int32Regs:$y, Int64Regs:$r,
+ Int64Regs:$g),
+ (SUST_B_2D_ARRAY_V2B64_TRAP Int64Regs:$s, Int32Regs:$l,
+ Int32Regs:$x, Int32Regs:$y, Int64Regs:$r, Int64Regs:$g)>;
+
def : Pat<(int_nvvm_sust_b_2d_array_v4i8_trap
Int64Regs:$s, Int32Regs:$l, Int32Regs:$x, Int32Regs:$y,
Int16Regs:$r, Int16Regs:$g, Int16Regs:$b, Int16Regs:$a),
Int32Regs:$x, Int32Regs:$y, Int32Regs:$z,
Int32Regs:$r)>;
+def : Pat<(int_nvvm_sust_b_3d_i64_trap
+ Int64Regs:$s, Int32Regs:$x, Int32Regs:$y, Int32Regs:$z,
+ Int64Regs:$r),
+ (SUST_B_3D_B64_TRAP Int64Regs:$s,
+ Int32Regs:$x, Int32Regs:$y, Int32Regs:$z,
+ Int64Regs:$r)>;
+
def : Pat<(int_nvvm_sust_b_3d_v2i8_trap
Int64Regs:$s, Int32Regs:$x, Int32Regs:$y, Int32Regs:$z,
Int16Regs:$r, Int16Regs:$g),
Int32Regs:$x, Int32Regs:$y, Int32Regs:$z,
Int32Regs:$r, Int32Regs:$g)>;
+def : Pat<(int_nvvm_sust_b_3d_v2i64_trap
+ Int64Regs:$s, Int32Regs:$x, Int32Regs:$y, Int32Regs:$z,
+ Int64Regs:$r, Int64Regs:$g),
+ (SUST_B_3D_V2B64_TRAP Int64Regs:$s,
+ Int32Regs:$x, Int32Regs:$y, Int32Regs:$z,
+ Int64Regs:$r, Int64Regs:$g)>;
+
def : Pat<(int_nvvm_sust_b_3d_v4i8_trap
Int64Regs:$s, Int32Regs:$x, Int32Regs:$y, Int32Regs:$z,
Int16Regs:$r, Int16Regs:$g, Int16Regs:$b, Int16Regs:$a),
Int32Regs:$r, Int32Regs:$g, Int32Regs:$b, Int32Regs:$a)>;
+// .zero variant
+def : Pat<(int_nvvm_sust_b_1d_i8_zero
+ Int64Regs:$s, Int32Regs:$x, Int16Regs:$r),
+ (SUST_B_1D_B8_ZERO Int64Regs:$s, Int32Regs:$x, Int16Regs:$r)>;
+
+def : Pat<(int_nvvm_sust_b_1d_i16_zero
+ Int64Regs:$s, Int32Regs:$x, Int16Regs:$r),
+ (SUST_B_1D_B16_ZERO Int64Regs:$s, Int32Regs:$x, Int16Regs:$r)>;
+
+def : Pat<(int_nvvm_sust_b_1d_i32_zero
+ Int64Regs:$s, Int32Regs:$x, Int32Regs:$r),
+ (SUST_B_1D_B32_ZERO Int64Regs:$s, Int32Regs:$x, Int32Regs:$r)>;
+
+def : Pat<(int_nvvm_sust_b_1d_i64_zero
+ Int64Regs:$s, Int32Regs:$x, Int64Regs:$r),
+ (SUST_B_1D_B64_ZERO Int64Regs:$s, Int32Regs:$x, Int64Regs:$r)>;
+
+def : Pat<(int_nvvm_sust_b_1d_v2i8_zero
+ Int64Regs:$s, Int32Regs:$x, Int16Regs:$r, Int16Regs:$g),
+ (SUST_B_1D_V2B8_ZERO Int64Regs:$s, Int32Regs:$x,
+ Int16Regs:$r, Int16Regs:$g)>;
+
+def : Pat<(int_nvvm_sust_b_1d_v2i16_zero
+ Int64Regs:$s, Int32Regs:$x, Int16Regs:$r, Int16Regs:$g),
+ (SUST_B_1D_V2B16_ZERO Int64Regs:$s, Int32Regs:$x,
+ Int16Regs:$r, Int16Regs:$g)>;
+
+def : Pat<(int_nvvm_sust_b_1d_v2i32_zero
+ Int64Regs:$s, Int32Regs:$x, Int32Regs:$r, Int32Regs:$g),
+ (SUST_B_1D_V2B32_ZERO Int64Regs:$s, Int32Regs:$x,
+ Int32Regs:$r, Int32Regs:$g)>;
+
+def : Pat<(int_nvvm_sust_b_1d_v2i64_zero
+ Int64Regs:$s, Int32Regs:$x, Int64Regs:$r, Int64Regs:$g),
+ (SUST_B_1D_V2B64_ZERO Int64Regs:$s, Int32Regs:$x,
+ Int64Regs:$r, Int64Regs:$g)>;
+
+def : Pat<(int_nvvm_sust_b_1d_v4i8_zero
+ Int64Regs:$s, Int32Regs:$x,
+ Int16Regs:$r, Int16Regs:$g, Int16Regs:$b, Int16Regs:$a),
+ (SUST_B_1D_V4B8_ZERO Int64Regs:$s, Int32Regs:$x,
+ Int16Regs:$r, Int16Regs:$g, Int16Regs:$b, Int16Regs:$a)>;
+
+def : Pat<(int_nvvm_sust_b_1d_v4i16_zero
+ Int64Regs:$s, Int32Regs:$x,
+ Int16Regs:$r, Int16Regs:$g, Int16Regs:$b, Int16Regs:$a),
+ (SUST_B_1D_V4B16_ZERO Int64Regs:$s, Int32Regs:$x,
+ Int16Regs:$r, Int16Regs:$g, Int16Regs:$b, Int16Regs:$a)>;
+
+def : Pat<(int_nvvm_sust_b_1d_v4i32_zero
+ Int64Regs:$s, Int32Regs:$x,
+ Int32Regs:$r, Int32Regs:$g, Int32Regs:$b, Int32Regs:$a),
+ (SUST_B_1D_V4B32_ZERO Int64Regs:$s, Int32Regs:$x,
+ Int32Regs:$r, Int32Regs:$g, Int32Regs:$b, Int32Regs:$a)>;
+
+
+
+def : Pat<(int_nvvm_sust_b_1d_array_i8_zero
+ Int64Regs:$s, Int32Regs:$l, Int32Regs:$x, Int16Regs:$r),
+ (SUST_B_1D_ARRAY_B8_ZERO Int64Regs:$s, Int32Regs:$l, Int32Regs:$x,
+ Int16Regs:$r)>;
+
+def : Pat<(int_nvvm_sust_b_1d_array_i16_zero
+ Int64Regs:$s, Int32Regs:$l, Int32Regs:$x, Int16Regs:$r),
+ (SUST_B_1D_ARRAY_B16_ZERO Int64Regs:$s, Int32Regs:$l, Int32Regs:$x,
+ Int16Regs:$r)>;
+
+def : Pat<(int_nvvm_sust_b_1d_array_i32_zero
+ Int64Regs:$s, Int32Regs:$l, Int32Regs:$x, Int32Regs:$r),
+ (SUST_B_1D_ARRAY_B32_ZERO Int64Regs:$s, Int32Regs:$l, Int32Regs:$x,
+ Int32Regs:$r)>;
+
+def : Pat<(int_nvvm_sust_b_1d_array_i64_zero
+ Int64Regs:$s, Int32Regs:$l, Int32Regs:$x, Int64Regs:$r),
+ (SUST_B_1D_ARRAY_B64_ZERO Int64Regs:$s, Int32Regs:$l, Int32Regs:$x,
+ Int64Regs:$r)>;
+
+def : Pat<(int_nvvm_sust_b_1d_array_v2i8_zero
+ Int64Regs:$s, Int32Regs:$l, Int32Regs:$x, Int16Regs:$r, Int16Regs:$g),
+ (SUST_B_1D_ARRAY_V2B8_ZERO Int64Regs:$s, Int32Regs:$l, Int32Regs:$x,
+ Int16Regs:$r, Int16Regs:$g)>;
+
+def : Pat<(int_nvvm_sust_b_1d_array_v2i16_zero
+ Int64Regs:$s, Int32Regs:$l, Int32Regs:$x, Int16Regs:$r, Int16Regs:$g),
+ (SUST_B_1D_ARRAY_V2B16_ZERO Int64Regs:$s, Int32Regs:$l, Int32Regs:$x,
+ Int16Regs:$r, Int16Regs:$g)>;
+
+def : Pat<(int_nvvm_sust_b_1d_array_v2i32_zero
+ Int64Regs:$s, Int32Regs:$l, Int32Regs:$x, Int32Regs:$r, Int32Regs:$g),
+ (SUST_B_1D_ARRAY_V2B32_ZERO Int64Regs:$s, Int32Regs:$l, Int32Regs:$x,
+ Int32Regs:$r, Int32Regs:$g)>;
+
+def : Pat<(int_nvvm_sust_b_1d_array_v2i64_zero
+ Int64Regs:$s, Int32Regs:$l, Int32Regs:$x, Int64Regs:$r, Int64Regs:$g),
+ (SUST_B_1D_ARRAY_V2B64_ZERO Int64Regs:$s, Int32Regs:$l, Int32Regs:$x,
+ Int64Regs:$r, Int64Regs:$g)>;
+
+def : Pat<(int_nvvm_sust_b_1d_array_v4i8_zero
+ Int64Regs:$s, Int32Regs:$l, Int32Regs:$x,
+ Int16Regs:$r, Int16Regs:$g, Int16Regs:$b, Int16Regs:$a),
+ (SUST_B_1D_ARRAY_V4B8_ZERO Int64Regs:$s, Int32Regs:$l, Int32Regs:$x,
+ Int16Regs:$r, Int16Regs:$g, Int16Regs:$b, Int16Regs:$a)>;
+
+def : Pat<(int_nvvm_sust_b_1d_array_v4i16_zero
+ Int64Regs:$s, Int32Regs:$l, Int32Regs:$x,
+ Int16Regs:$r, Int16Regs:$g, Int16Regs:$b, Int16Regs:$a),
+ (SUST_B_1D_ARRAY_V4B16_ZERO Int64Regs:$s, Int32Regs:$l, Int32Regs:$x,
+ Int16Regs:$r, Int16Regs:$g, Int16Regs:$b, Int16Regs:$a)>;
+
+def : Pat<(int_nvvm_sust_b_1d_array_v4i32_zero
+ Int64Regs:$s, Int32Regs:$l, Int32Regs:$x,
+ Int32Regs:$r, Int32Regs:$g, Int32Regs:$b, Int32Regs:$a),
+ (SUST_B_1D_ARRAY_V4B32_ZERO Int64Regs:$s, Int32Regs:$l, Int32Regs:$x,
+ Int32Regs:$r, Int32Regs:$g, Int32Regs:$b, Int32Regs:$a)>;
+
+
+
+def : Pat<(int_nvvm_sust_b_2d_i8_zero
+ Int64Regs:$s, Int32Regs:$x, Int32Regs:$y, Int16Regs:$r),
+ (SUST_B_2D_B8_ZERO Int64Regs:$s, Int32Regs:$x, Int32Regs:$y,
+ Int16Regs:$r)>;
+
+def : Pat<(int_nvvm_sust_b_2d_i16_zero
+ Int64Regs:$s, Int32Regs:$x, Int32Regs:$y, Int16Regs:$r),
+ (SUST_B_2D_B16_ZERO Int64Regs:$s, Int32Regs:$x, Int32Regs:$y,
+ Int16Regs:$r)>;
+
+def : Pat<(int_nvvm_sust_b_2d_i32_zero
+ Int64Regs:$s, Int32Regs:$x, Int32Regs:$y, Int32Regs:$r),
+ (SUST_B_2D_B32_ZERO Int64Regs:$s, Int32Regs:$x, Int32Regs:$y,
+ Int32Regs:$r)>;
+
+def : Pat<(int_nvvm_sust_b_2d_i64_zero
+ Int64Regs:$s, Int32Regs:$x, Int32Regs:$y, Int64Regs:$r),
+ (SUST_B_2D_B64_ZERO Int64Regs:$s, Int32Regs:$x, Int32Regs:$y,
+ Int64Regs:$r)>;
+
+def : Pat<(int_nvvm_sust_b_2d_v2i8_zero
+ Int64Regs:$s, Int32Regs:$x, Int32Regs:$y, Int16Regs:$r, Int16Regs:$g),
+ (SUST_B_2D_V2B8_ZERO Int64Regs:$s, Int32Regs:$x, Int32Regs:$y,
+ Int16Regs:$r, Int16Regs:$g)>;
+
+def : Pat<(int_nvvm_sust_b_2d_v2i16_zero
+ Int64Regs:$s, Int32Regs:$x, Int32Regs:$y, Int16Regs:$r, Int16Regs:$g),
+ (SUST_B_2D_V2B16_ZERO Int64Regs:$s, Int32Regs:$x, Int32Regs:$y,
+ Int16Regs:$r, Int16Regs:$g)>;
+
+def : Pat<(int_nvvm_sust_b_2d_v2i32_zero
+ Int64Regs:$s, Int32Regs:$x, Int32Regs:$y, Int32Regs:$r, Int32Regs:$g),
+ (SUST_B_2D_V2B32_ZERO Int64Regs:$s, Int32Regs:$x, Int32Regs:$y,
+ Int32Regs:$r, Int32Regs:$g)>;
+
+def : Pat<(int_nvvm_sust_b_2d_v2i64_zero
+ Int64Regs:$s, Int32Regs:$x, Int32Regs:$y, Int64Regs:$r, Int64Regs:$g),
+ (SUST_B_2D_V2B64_ZERO Int64Regs:$s, Int32Regs:$x, Int32Regs:$y,
+ Int64Regs:$r, Int64Regs:$g)>;
+
+def : Pat<(int_nvvm_sust_b_2d_v4i8_zero
+ Int64Regs:$s, Int32Regs:$x, Int32Regs:$y,
+ Int16Regs:$r, Int16Regs:$g, Int16Regs:$b, Int16Regs:$a),
+ (SUST_B_2D_V4B8_ZERO Int64Regs:$s, Int32Regs:$x, Int32Regs:$y,
+ Int16Regs:$r, Int16Regs:$g, Int16Regs:$b, Int16Regs:$a)>;
+
+def : Pat<(int_nvvm_sust_b_2d_v4i16_zero
+ Int64Regs:$s, Int32Regs:$x, Int32Regs:$y,
+ Int16Regs:$r, Int16Regs:$g, Int16Regs:$b, Int16Regs:$a),
+ (SUST_B_2D_V4B16_ZERO Int64Regs:$s, Int32Regs:$x, Int32Regs:$y,
+ Int16Regs:$r, Int16Regs:$g, Int16Regs:$b, Int16Regs:$a)>;
+
+def : Pat<(int_nvvm_sust_b_2d_v4i32_zero
+ Int64Regs:$s, Int32Regs:$x, Int32Regs:$y,
+ Int32Regs:$r, Int32Regs:$g, Int32Regs:$b, Int32Regs:$a),
+ (SUST_B_2D_V4B32_ZERO Int64Regs:$s, Int32Regs:$x, Int32Regs:$y,
+ Int32Regs:$r, Int32Regs:$g, Int32Regs:$b, Int32Regs:$a)>;
+
+
+
+def : Pat<(int_nvvm_sust_b_2d_array_i8_zero
+ Int64Regs:$s, Int32Regs:$l, Int32Regs:$x, Int32Regs:$y, Int16Regs:$r),
+ (SUST_B_2D_ARRAY_B8_ZERO Int64Regs:$s,
+ Int32Regs:$l, Int32Regs:$x, Int32Regs:$y,
+ Int16Regs:$r)>;
+
+def : Pat<(int_nvvm_sust_b_2d_array_i16_zero
+ Int64Regs:$s, Int32Regs:$l, Int32Regs:$x, Int32Regs:$y, Int16Regs:$r),
+ (SUST_B_2D_ARRAY_B16_ZERO Int64Regs:$s,
+ Int32Regs:$l, Int32Regs:$x, Int32Regs:$y,
+ Int16Regs:$r)>;
+
+def : Pat<(int_nvvm_sust_b_2d_array_i32_zero
+ Int64Regs:$s, Int32Regs:$l, Int32Regs:$x, Int32Regs:$y, Int32Regs:$r),
+ (SUST_B_2D_ARRAY_B32_ZERO Int64Regs:$s,
+ Int32Regs:$l, Int32Regs:$x, Int32Regs:$y,
+ Int32Regs:$r)>;
+
+def : Pat<(int_nvvm_sust_b_2d_array_i64_zero
+ Int64Regs:$s, Int32Regs:$l, Int32Regs:$x, Int32Regs:$y, Int64Regs:$r),
+ (SUST_B_2D_ARRAY_B64_ZERO Int64Regs:$s,
+ Int32Regs:$l, Int32Regs:$x, Int32Regs:$y,
+ Int64Regs:$r)>;
+
+def : Pat<(int_nvvm_sust_b_2d_array_v2i8_zero
+ Int64Regs:$s, Int32Regs:$l, Int32Regs:$x, Int32Regs:$y,
+ Int16Regs:$r, Int16Regs:$g),
+ (SUST_B_2D_ARRAY_V2B8_ZERO Int64Regs:$s, Int32Regs:$l,
+ Int32Regs:$x, Int32Regs:$y,
+ Int16Regs:$r, Int16Regs:$g)>;
+
+def : Pat<(int_nvvm_sust_b_2d_array_v2i16_zero
+ Int64Regs:$s, Int32Regs:$l, Int32Regs:$x, Int32Regs:$y,
+ Int16Regs:$r, Int16Regs:$g),
+ (SUST_B_2D_ARRAY_V2B16_ZERO Int64Regs:$s, Int32Regs:$l,
+ Int32Regs:$x, Int32Regs:$y,
+ Int16Regs:$r, Int16Regs:$g)>;
+
+def : Pat<(int_nvvm_sust_b_2d_array_v2i32_zero
+ Int64Regs:$s, Int32Regs:$l, Int32Regs:$x, Int32Regs:$y, Int32Regs:$r,
+ Int32Regs:$g),
+ (SUST_B_2D_ARRAY_V2B32_ZERO Int64Regs:$s, Int32Regs:$l,
+ Int32Regs:$x, Int32Regs:$y, Int32Regs:$r, Int32Regs:$g)>;
+
+def : Pat<(int_nvvm_sust_b_2d_array_v2i64_zero
+ Int64Regs:$s, Int32Regs:$l, Int32Regs:$x, Int32Regs:$y, Int64Regs:$r,
+ Int64Regs:$g),
+ (SUST_B_2D_ARRAY_V2B64_ZERO Int64Regs:$s, Int32Regs:$l,
+ Int32Regs:$x, Int32Regs:$y, Int64Regs:$r, Int64Regs:$g)>;
+
+def : Pat<(int_nvvm_sust_b_2d_array_v4i8_zero
+ Int64Regs:$s, Int32Regs:$l, Int32Regs:$x, Int32Regs:$y,
+ Int16Regs:$r, Int16Regs:$g, Int16Regs:$b, Int16Regs:$a),
+ (SUST_B_2D_ARRAY_V4B8_ZERO Int64Regs:$s,
+ Int32Regs:$l, Int32Regs:$x, Int32Regs:$y,
+ Int16Regs:$r, Int16Regs:$g, Int16Regs:$b, Int16Regs:$a)>;
+
+def : Pat<(int_nvvm_sust_b_2d_array_v4i16_zero
+ Int64Regs:$s, Int32Regs:$l, Int32Regs:$x, Int32Regs:$y,
+ Int16Regs:$r, Int16Regs:$g, Int16Regs:$b, Int16Regs:$a),
+ (SUST_B_2D_ARRAY_V4B16_ZERO Int64Regs:$s,
+ Int32Regs:$l, Int32Regs:$x, Int32Regs:$y,
+ Int16Regs:$r, Int16Regs:$g, Int16Regs:$b, Int16Regs:$a)>;
+
+def : Pat<(int_nvvm_sust_b_2d_array_v4i32_zero
+ Int64Regs:$s, Int32Regs:$l, Int32Regs:$x, Int32Regs:$y,
+ Int32Regs:$r, Int32Regs:$g, Int32Regs:$b, Int32Regs:$a),
+ (SUST_B_2D_ARRAY_V4B32_ZERO Int64Regs:$s, Int32Regs:$l,
+ Int32Regs:$x, Int32Regs:$y,
+ Int32Regs:$r, Int32Regs:$g, Int32Regs:$b, Int32Regs:$a)>;
+
+
+
+def : Pat<(int_nvvm_sust_b_3d_i8_zero
+ Int64Regs:$s, Int32Regs:$x, Int32Regs:$y, Int32Regs:$z,
+ Int16Regs:$r),
+ (SUST_B_3D_B8_ZERO Int64Regs:$s,
+ Int32Regs:$x, Int32Regs:$y, Int32Regs:$z,
+ Int16Regs:$r)>;
+
+def : Pat<(int_nvvm_sust_b_3d_i16_zero
+ Int64Regs:$s, Int32Regs:$x, Int32Regs:$y, Int32Regs:$z,
+ Int16Regs:$r),
+ (SUST_B_3D_B16_ZERO Int64Regs:$s,
+ Int32Regs:$x, Int32Regs:$y, Int32Regs:$z,
+ Int16Regs:$r)>;
+
+def : Pat<(int_nvvm_sust_b_3d_i32_zero
+ Int64Regs:$s, Int32Regs:$x, Int32Regs:$y, Int32Regs:$z,
+ Int32Regs:$r),
+ (SUST_B_3D_B32_ZERO Int64Regs:$s,
+ Int32Regs:$x, Int32Regs:$y, Int32Regs:$z,
+ Int32Regs:$r)>;
+
+def : Pat<(int_nvvm_sust_b_3d_i64_zero
+ Int64Regs:$s, Int32Regs:$x, Int32Regs:$y, Int32Regs:$z,
+ Int64Regs:$r),
+ (SUST_B_3D_B64_ZERO Int64Regs:$s,
+ Int32Regs:$x, Int32Regs:$y, Int32Regs:$z,
+ Int64Regs:$r)>;
+
+def : Pat<(int_nvvm_sust_b_3d_v2i8_zero
+ Int64Regs:$s, Int32Regs:$x, Int32Regs:$y, Int32Regs:$z,
+ Int16Regs:$r, Int16Regs:$g),
+ (SUST_B_3D_V2B8_ZERO Int64Regs:$s,
+ Int32Regs:$x, Int32Regs:$y, Int32Regs:$z,
+ Int16Regs:$r, Int16Regs:$g)>;
+
+def : Pat<(int_nvvm_sust_b_3d_v2i16_zero
+ Int64Regs:$s, Int32Regs:$x, Int32Regs:$y, Int32Regs:$z,
+ Int16Regs:$r, Int16Regs:$g),
+ (SUST_B_3D_V2B16_ZERO Int64Regs:$s,
+ Int32Regs:$x, Int32Regs:$y, Int32Regs:$z,
+ Int16Regs:$r, Int16Regs:$g)>;
+
+def : Pat<(int_nvvm_sust_b_3d_v2i32_zero
+ Int64Regs:$s, Int32Regs:$x, Int32Regs:$y, Int32Regs:$z,
+ Int32Regs:$r, Int32Regs:$g),
+ (SUST_B_3D_V2B32_ZERO Int64Regs:$s,
+ Int32Regs:$x, Int32Regs:$y, Int32Regs:$z,
+ Int32Regs:$r, Int32Regs:$g)>;
+
+def : Pat<(int_nvvm_sust_b_3d_v2i64_zero
+ Int64Regs:$s, Int32Regs:$x, Int32Regs:$y, Int32Regs:$z,
+ Int64Regs:$r, Int64Regs:$g),
+ (SUST_B_3D_V2B64_ZERO Int64Regs:$s,
+ Int32Regs:$x, Int32Regs:$y, Int32Regs:$z,
+ Int64Regs:$r, Int64Regs:$g)>;
+
+def : Pat<(int_nvvm_sust_b_3d_v4i8_zero
+ Int64Regs:$s, Int32Regs:$x, Int32Regs:$y, Int32Regs:$z,
+ Int16Regs:$r, Int16Regs:$g, Int16Regs:$b, Int16Regs:$a),
+ (SUST_B_3D_V4B8_ZERO Int64Regs:$s,
+ Int32Regs:$x, Int32Regs:$y, Int32Regs:$z,
+ Int16Regs:$r, Int16Regs:$g, Int16Regs:$b, Int16Regs:$a)>;
+
+def : Pat<(int_nvvm_sust_b_3d_v4i16_zero
+ Int64Regs:$s, Int32Regs:$x, Int32Regs:$y, Int32Regs:$z,
+ Int16Regs:$r, Int16Regs:$g, Int16Regs:$b, Int16Regs:$a),
+ (SUST_B_3D_V4B16_ZERO Int64Regs:$s,
+ Int32Regs:$x, Int32Regs:$y, Int32Regs:$z,
+ Int16Regs:$r, Int16Regs:$g, Int16Regs:$b, Int16Regs:$a)>;
+
+def : Pat<(int_nvvm_sust_b_3d_v4i32_zero
+ Int64Regs:$s, Int32Regs:$x, Int32Regs:$y, Int32Regs:$z,
+ Int32Regs:$r, Int32Regs:$g, Int32Regs:$b, Int32Regs:$a),
+ (SUST_B_3D_V4B32_ZERO Int64Regs:$s,
+ Int32Regs:$x, Int32Regs:$y, Int32Regs:$z,
+ Int32Regs:$r, Int32Regs:$g, Int32Regs:$b, Int32Regs:$a)>;
+
+
def : Pat<(int_nvvm_sust_p_1d_i8_trap
projects / oota-llvm.git / blobdiff