[RISCV] [1/2] Add intrinsic for Zbc extension

[LevyHsu]

Summary

Head files are included in the second patch in case the name needs to be changed.

RV32 / 64:
clmul
clmulh
clmulr

[craig.topper]

clang/include/clang/Basic/BuiltinsRISCV.def

22

I think we should drop the underscore in clmul_h and clmul_r. The mnemonic doesn't have a period between them. So that's different than orc.b and crc32.w.

llvm/include/llvm/IR/IntrinsicsRISCV.td

82

clmul_h -> clmulh
clmul_r -> clmulr

[LevyHsu]

1. clang/include/clang/Basic/BuiltinsRISCV.def clang/lib/CodeGen/CGBuiltin.cpp llvm/include/llvm/IR/IntrinsicsRISCV.td clang/test/CodeGen/RISCV/rvb-intrinsics/riscv32-zbc.c clang/test/CodeGen/RISCV/rvb-intrinsics/riscv64-zbc.c
   • Renaming clmul_h clmul_r to clmulh clmulr

2. llvm/test/CodeGen/RISCV/rv32zbc-intrinsic.ll llvm/test/CodeGen/RISCV/rv64zbc-intrinsic.ll
   • Renaming clmul.h clmul.r to clmulh clmulr

[craig.topper]

LGTM. We can adjust the builtin names in the future if we need to do something different to match gcc.

clang/include/clang/Basic/BuiltinsRISCV.def

 #if defined(BUILTIN) && !defined(TARGET_BUILTIN)
 #   define TARGET_BUILTIN(ID, TYPE, ATTRS, FEATURE) BUILTIN(ID, TYPE, ATTRS)
 #endif
 
 #include "clang/Basic/riscv_vector_builtins.inc"
 
 // Zbb extension
 TARGET_BUILTIN(__builtin_riscv_orc_b_32, "ZiZi", "nc", "experimental-zbb")
 TARGET_BUILTIN(__builtin_riscv_orc_b_64, "WiWi", "nc", "experimental-zbb")
 
+// Zbc extension
+TARGET_BUILTIN(__builtin_riscv_clmul, "LiLiLi", "nc", "experimental-zbc")
+TARGET_BUILTIN(__builtin_riscv_clmulh, "LiLiLi", "nc", "experimental-zbc")
+TARGET_BUILTIN(__builtin_riscv_clmulr, "LiLiLi", "nc", "experimental-zbc")
+
 // Zbr extension
 TARGET_BUILTIN(__builtin_riscv_crc32_b, "LiLi", "nc", "experimental-zbr")
 TARGET_BUILTIN(__builtin_riscv_crc32_h, "LiLi", "nc", "experimental-zbr")
 TARGET_BUILTIN(__builtin_riscv_crc32_w, "LiLi", "nc", "experimental-zbr")
 TARGET_BUILTIN(__builtin_riscv_crc32c_b, "LiLi", "nc", "experimental-zbr")
 TARGET_BUILTIN(__builtin_riscv_crc32c_h, "LiLi", "nc", "experimental-zbr")
 TARGET_BUILTIN(__builtin_riscv_crc32c_w, "LiLi", "nc", "experimental-zbr")
 TARGET_BUILTIN(__builtin_riscv_crc32_d, "LiLi", "nc", "experimental-zbr")
 TARGET_BUILTIN(__builtin_riscv_crc32c_d, "LiLi", "nc", "experimental-zbr")
 
 #undef BUILTIN
 #undef TARGET_BUILTIN

clang/lib/CodeGen/CGBuiltin.cpp

#include "clang/Basic/riscv_vector_builtin_cg.inc"
 
   // Zbb
   case RISCV::BI__builtin_riscv_orc_b_32:
   case RISCV::BI__builtin_riscv_orc_b_64:
     ID = Intrinsic::riscv_orc_b;
     IntrinsicTypes = {ResultType};
     break;
 
+  // Zbc
+  case RISCV::BI__builtin_riscv_clmul:
+    ID = Intrinsic::riscv_clmul;
+    IntrinsicTypes = {ResultType};
+    break;
+  case RISCV::BI__builtin_riscv_clmulh:
+    ID = Intrinsic::riscv_clmulh;
+    IntrinsicTypes = {ResultType};
+    break;
+  case RISCV::BI__builtin_riscv_clmulr:
+    ID = Intrinsic::riscv_clmulr;
+    IntrinsicTypes = {ResultType};
+    break;
+
   // Zbr
   case RISCV::BI__builtin_riscv_crc32_b:
     ID = Intrinsic::riscv_crc32_b;
     IntrinsicTypes = {ResultType};
     break;
   case RISCV::BI__builtin_riscv_crc32_h:
     ID = Intrinsic::riscv_crc32_h;
     IntrinsicTypes = {ResultType};

clang/test/CodeGen/RISCV/rvb-intrinsics/riscv32-zbc.c

+// NOTE: Assertions have been autogenerated by utils/update_cc_test_checks.py
+// RUN: %clang_cc1 -triple riscv32 -target-feature +experimental-zbc -emit-llvm %s -o - \
+// RUN:     | FileCheck %s  -check-prefix=RV32ZBC
+
+// RV32ZBC-LABEL: @clmul(
+// RV32ZBC-NEXT:  entry:
+// RV32ZBC-NEXT:    [[A_ADDR:%.*]] = alloca i32, align 4
+// RV32ZBC-NEXT:    [[B_ADDR:%.*]] = alloca i32, align 4
+// RV32ZBC-NEXT:    store i32 [[A:%.*]], i32* [[A_ADDR]], align 4
+// RV32ZBC-NEXT:    store i32 [[B:%.*]], i32* [[B_ADDR]], align 4
+// RV32ZBC-NEXT:    [[TMP0:%.*]] = load i32, i32* [[A_ADDR]], align 4
+// RV32ZBC-NEXT:    [[TMP1:%.*]] = load i32, i32* [[B_ADDR]], align 4
+// RV32ZBC-NEXT:    [[TMP2:%.*]] = call i32 @llvm.riscv.clmul.i32(i32 [[TMP0]], i32 [[TMP1]])
+// RV32ZBC-NEXT:    ret i32 [[TMP2]]
+//
+long clmul(long a, long b) {
+  return __builtin_riscv_clmul(a, b);
+}
+
+// RV32ZBC-LABEL: @clmulh(
+// RV32ZBC-NEXT:  entry:
+// RV32ZBC-NEXT:    [[A_ADDR:%.*]] = alloca i32, align 4
+// RV32ZBC-NEXT:    [[B_ADDR:%.*]] = alloca i32, align 4
+// RV32ZBC-NEXT:    store i32 [[A:%.*]], i32* [[A_ADDR]], align 4
+// RV32ZBC-NEXT:    store i32 [[B:%.*]], i32* [[B_ADDR]], align 4
+// RV32ZBC-NEXT:    [[TMP0:%.*]] = load i32, i32* [[A_ADDR]], align 4
+// RV32ZBC-NEXT:    [[TMP1:%.*]] = load i32, i32* [[B_ADDR]], align 4
+// RV32ZBC-NEXT:    [[TMP2:%.*]] = call i32 @llvm.riscv.clmulh.i32(i32 [[TMP0]], i32 [[TMP1]])
+// RV32ZBC-NEXT:    ret i32 [[TMP2]]
+//
+long clmulh(long a, long b) {
+  return __builtin_riscv_clmulh(a, b);
+}
+
+// RV32ZBC-LABEL: @clmulr(
+// RV32ZBC-NEXT:  entry:
+// RV32ZBC-NEXT:    [[A_ADDR:%.*]] = alloca i32, align 4
+// RV32ZBC-NEXT:    [[B_ADDR:%.*]] = alloca i32, align 4
+// RV32ZBC-NEXT:    store i32 [[A:%.*]], i32* [[A_ADDR]], align 4
+// RV32ZBC-NEXT:    store i32 [[B:%.*]], i32* [[B_ADDR]], align 4
+// RV32ZBC-NEXT:    [[TMP0:%.*]] = load i32, i32* [[A_ADDR]], align 4
+// RV32ZBC-NEXT:    [[TMP1:%.*]] = load i32, i32* [[B_ADDR]], align 4
+// RV32ZBC-NEXT:    [[TMP2:%.*]] = call i32 @llvm.riscv.clmulr.i32(i32 [[TMP0]], i32 [[TMP1]])
+// RV32ZBC-NEXT:    ret i32 [[TMP2]]
+//
+long clmulr(long a, long b) {
+  return __builtin_riscv_clmulr(a, b);
+}

clang/test/CodeGen/RISCV/rvb-intrinsics/riscv64-zbc.c

+// NOTE: Assertions have been autogenerated by utils/update_cc_test_checks.py
+// RUN: %clang_cc1 -triple riscv64 -target-feature +experimental-zbc -emit-llvm %s -o - \
+// RUN:     | FileCheck %s  -check-prefix=RV64ZBC
+
+// RV64ZBC-LABEL: @clmul(
+// RV64ZBC-NEXT:  entry:
+// RV64ZBC-NEXT:    [[A_ADDR:%.*]] = alloca i64, align 8
+// RV64ZBC-NEXT:    [[B_ADDR:%.*]] = alloca i64, align 8
+// RV64ZBC-NEXT:    store i64 [[A:%.*]], i64* [[A_ADDR]], align 8
+// RV64ZBC-NEXT:    store i64 [[B:%.*]], i64* [[B_ADDR]], align 8
+// RV64ZBC-NEXT:    [[TMP0:%.*]] = load i64, i64* [[A_ADDR]], align 8
+// RV64ZBC-NEXT:    [[TMP1:%.*]] = load i64, i64* [[B_ADDR]], align 8
+// RV64ZBC-NEXT:    [[TMP2:%.*]] = call i64 @llvm.riscv.clmul.i64(i64 [[TMP0]], i64 [[TMP1]])
+// RV64ZBC-NEXT:    ret i64 [[TMP2]]
+//
+long clmul(long a, long b) {
+  return __builtin_riscv_clmul(a, b);
+}
+
+// RV64ZBC-LABEL: @clmulh(
+// RV64ZBC-NEXT:  entry:
+// RV64ZBC-NEXT:    [[A_ADDR:%.*]] = alloca i64, align 8
+// RV64ZBC-NEXT:    [[B_ADDR:%.*]] = alloca i64, align 8
+// RV64ZBC-NEXT:    store i64 [[A:%.*]], i64* [[A_ADDR]], align 8
+// RV64ZBC-NEXT:    store i64 [[B:%.*]], i64* [[B_ADDR]], align 8
+// RV64ZBC-NEXT:    [[TMP0:%.*]] = load i64, i64* [[A_ADDR]], align 8
+// RV64ZBC-NEXT:    [[TMP1:%.*]] = load i64, i64* [[B_ADDR]], align 8
+// RV64ZBC-NEXT:    [[TMP2:%.*]] = call i64 @llvm.riscv.clmulh.i64(i64 [[TMP0]], i64 [[TMP1]])
+// RV64ZBC-NEXT:    ret i64 [[TMP2]]
+//
+long clmulh(long a, long b) {
+  return __builtin_riscv_clmulh(a, b);
+}
+
+// RV64ZBC-LABEL: @clmulr(
+// RV64ZBC-NEXT:  entry:
+// RV64ZBC-NEXT:    [[A_ADDR:%.*]] = alloca i64, align 8
+// RV64ZBC-NEXT:    [[B_ADDR:%.*]] = alloca i64, align 8
+// RV64ZBC-NEXT:    store i64 [[A:%.*]], i64* [[A_ADDR]], align 8
+// RV64ZBC-NEXT:    store i64 [[B:%.*]], i64* [[B_ADDR]], align 8
+// RV64ZBC-NEXT:    [[TMP0:%.*]] = load i64, i64* [[A_ADDR]], align 8
+// RV64ZBC-NEXT:    [[TMP1:%.*]] = load i64, i64* [[B_ADDR]], align 8
+// RV64ZBC-NEXT:    [[TMP2:%.*]] = call i64 @llvm.riscv.clmulr.i64(i64 [[TMP0]], i64 [[TMP1]])
+// RV64ZBC-NEXT:    ret i64 [[TMP2]]
+//
+long clmulr(long a, long b) {
+  return __builtin_riscv_clmulr(a, b);
+}

llvm/include/llvm/IR/IntrinsicsRISCV.td

 // Bitmanip (Bit Manipulation) Extension
 
 let TargetPrefix = "riscv" in {
 
   class BitManipGPRIntrinsics
       : Intrinsic<[llvm_any_ty],
                   [LLVMMatchType<0>],
                   [IntrNoMem, IntrSpeculatable, IntrWillReturn]>;
+  class BitManipGPRGPRIntrinsics
+      : Intrinsic<[llvm_any_ty],
+                  [LLVMMatchType<0>, LLVMMatchType<0>],
+                  [IntrNoMem, IntrSpeculatable, IntrWillReturn]>;
 
   // Zbb
   def int_riscv_orc_b : BitManipGPRIntrinsics;
 
+  // Zbc
+  def int_riscv_clmul  : BitManipGPRGPRIntrinsics;
+  def int_riscv_clmulh : BitManipGPRGPRIntrinsics;
+  def int_riscv_clmulr : BitManipGPRGPRIntrinsics;
+
   // Zbr
   def int_riscv_crc32_b : BitManipGPRIntrinsics;
   def int_riscv_crc32_h : BitManipGPRIntrinsics;
   def int_riscv_crc32_w : BitManipGPRIntrinsics;
   def int_riscv_crc32_d : BitManipGPRIntrinsics;
   def int_riscv_crc32c_b : BitManipGPRIntrinsics;
   def int_riscv_crc32c_h : BitManipGPRIntrinsics;
   def int_riscv_crc32c_w : BitManipGPRIntrinsics;

llvm/lib/Target/RISCV/RISCVInstrInfoB.td

def : Pat<(i64 (or (and (assertsexti32 GPR:$rs2), 0xFFFFFFFFFFFF0000),
                    (srl (and GPR:$rs1, 0xFFFFFFFF), (i64 16)))),
           (PACKUW GPR:$rs1, GPR:$rs2)>;
 } // Predicates = [HasStdExtZbp, IsRV64]
 
 let Predicates = [HasStdExtZbb] in {
 def : PatGpr<int_riscv_orc_b, ORCB>;
 } // Predicates = [HasStdExtZbb]
 
+let Predicates = [HasStdExtZbc] in {
+def : PatGprGpr<int_riscv_clmul, CLMUL>;
+def : PatGprGpr<int_riscv_clmulh, CLMULH>;
+def : PatGprGpr<int_riscv_clmulr, CLMULR>;
+} // Predicates = [HasStdExtZbc]
+
 let Predicates = [HasStdExtZbr] in {
 def : PatGpr<int_riscv_crc32_b, CRC32B>;
 def : PatGpr<int_riscv_crc32_h, CRC32H>;
 def : PatGpr<int_riscv_crc32_w, CRC32W>;
 def : PatGpr<int_riscv_crc32c_b, CRC32CB>;
 def : PatGpr<int_riscv_crc32c_h, CRC32CH>;
 def : PatGpr<int_riscv_crc32c_w, CRC32CW>;
 } // Predicates = [HasStdExtZbr]
 
 let Predicates = [HasStdExtZbr, IsRV64] in {
 def : PatGpr<int_riscv_crc32_d, CRC32D>;
 def : PatGpr<int_riscv_crc32c_d, CRC32CD>;
 } // Predicates = [HasStdExtZbr, IsRV64]

llvm/test/CodeGen/RISCV/rv32zbc-intrinsic.ll

+; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py
+; RUN: llc -mtriple=riscv32 -mattr=+experimental-b -verify-machineinstrs < %s \
+; RUN:   | FileCheck %s -check-prefix=RV32IB
+; RUN: llc -mtriple=riscv32 -mattr=+experimental-zbc -verify-machineinstrs < %s \
+; RUN:   | FileCheck %s -check-prefix=RV32IBC
+
+declare i32 @llvm.riscv.clmul.i32(i32 %a, i32 %b)
+
+define i32 @clmul32(i32 %a, i32 %b) nounwind {
+; RV32IB-LABEL: clmul32:
+; RV32IB:       # %bb.0:
+; RV32IB-NEXT:    clmul a0, a0, a1
+; RV32IB-NEXT:    ret
+;
+; RV32IBC-LABEL: clmul32:
+; RV32IBC:       # %bb.0:
+; RV32IBC-NEXT:    clmul a0, a0, a1
+; RV32IBC-NEXT:    ret
+  %tmp = call i32 @llvm.riscv.clmul.i32(i32 %a, i32 %b)
+ ret i32 %tmp
+}
+
+declare i32 @llvm.riscv.clmulh.i32(i32 %a, i32 %b)
+
+define i32 @clmul32h(i32 %a, i32 %b) nounwind {
+; RV32IB-LABEL: clmul32h:
+; RV32IB:       # %bb.0:
+; RV32IB-NEXT:    clmulh a0, a0, a1
+; RV32IB-NEXT:    ret
+;
+; RV32IBC-LABEL: clmul32h:
+; RV32IBC:       # %bb.0:
+; RV32IBC-NEXT:    clmulh a0, a0, a1
+; RV32IBC-NEXT:    ret
+  %tmp = call i32 @llvm.riscv.clmulh.i32(i32 %a, i32 %b)
+ ret i32 %tmp
+}
+
+declare i32 @llvm.riscv.clmulr.i32(i32 %a, i32 %b)
+
+define i32 @clmul32r(i32 %a, i32 %b) nounwind {
+; RV32IB-LABEL: clmul32r:
+; RV32IB:       # %bb.0:
+; RV32IB-NEXT:    clmulr a0, a0, a1
+; RV32IB-NEXT:    ret
+;
+; RV32IBC-LABEL: clmul32r:
+; RV32IBC:       # %bb.0:
+; RV32IBC-NEXT:    clmulr a0, a0, a1
+; RV32IBC-NEXT:    ret
+  %tmp = call i32 @llvm.riscv.clmulr.i32(i32 %a, i32 %b)
+ ret i32 %tmp
+}

llvm/test/CodeGen/RISCV/rv64zbc-intrinsic.ll

+; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py
+; RUN: llc -mtriple=riscv64 -mattr=+experimental-b -verify-machineinstrs < %s \
+; RUN:   | FileCheck %s -check-prefix=RV64IB
+; RUN: llc -mtriple=riscv64 -mattr=+experimental-zbc -verify-machineinstrs < %s \
+; RUN:   | FileCheck %s -check-prefix=RV64IBC
+
+declare i64 @llvm.riscv.clmul.i64(i64 %a, i64 %b)
+
+define i64 @clmul64(i64 %a, i64 %b) nounwind {
+; RV64IB-LABEL: clmul64:
+; RV64IB:       # %bb.0:
+; RV64IB-NEXT:    clmul a0, a0, a1
+; RV64IB-NEXT:    ret
+;
+; RV64IBC-LABEL: clmul64:
+; RV64IBC:       # %bb.0:
+; RV64IBC-NEXT:    clmul a0, a0, a1
+; RV64IBC-NEXT:    ret
+  %tmp = call i64 @llvm.riscv.clmul.i64(i64 %a, i64 %b)
+ ret i64 %tmp
+}
+
+declare i64 @llvm.riscv.clmulh.i64(i64 %a, i64 %b)
+
+define i64 @clmul64h(i64 %a, i64 %b) nounwind {
+; RV64IB-LABEL: clmul64h:
+; RV64IB:       # %bb.0:
+; RV64IB-NEXT:    clmulh a0, a0, a1
+; RV64IB-NEXT:    ret
+;
+; RV64IBC-LABEL: clmul64h:
+; RV64IBC:       # %bb.0:
+; RV64IBC-NEXT:    clmulh a0, a0, a1
+; RV64IBC-NEXT:    ret
+  %tmp = call i64 @llvm.riscv.clmulh.i64(i64 %a, i64 %b)
+ ret i64 %tmp
+}
+
+declare i64 @llvm.riscv.clmulr.i64(i64 %a, i64 %b)
+
+define i64 @clmul64r(i64 %a, i64 %b) nounwind {
+; RV64IB-LABEL: clmul64r:
+; RV64IB:       # %bb.0:
+; RV64IB-NEXT:    clmulr a0, a0, a1
+; RV64IB-NEXT:    ret
+;
+; RV64IBC-LABEL: clmul64r:
+; RV64IBC:       # %bb.0:
+; RV64IBC-NEXT:    clmulr a0, a0, a1
+; RV64IBC-NEXT:    ret
+  %tmp = call i64 @llvm.riscv.clmulr.i64(i64 %a, i64 %b)
+ ret i64 %tmp
+}