[ADT] Add signed and unsigned mulExtended to APInt

[pedroclobo]

Adds mulsExtended and muluExtended methods to APInt, as suggested in #153293.
These are based on the MULDQ and MULUDQ x86 intrinsics.

[llvmbot]

@llvm/pr-subscribers-llvm-adt

@llvm/pr-subscribers-llvm-support

Author: Pedro Lobo (pedroclobo)

Changes

Adds mulxs and mulxu methods to APInt, as suggested in #153293.
These are based on the MULDQ and MULUDQ x86 intrinsics.

---

Full diff: https://github.com/llvm/llvm-project/pull/153399.diff

3 Files Affected:

• (modified) llvm/include/llvm/ADT/APInt.h (+6)
• (modified) llvm/lib/Support/APInt.cpp (+16)
• (modified) llvm/unittests/ADT/APIntTest.cpp (+32)

diff --git a/llvm/include/llvm/ADT/APInt.h b/llvm/include/llvm/ADT/APInt.h
index 295506393a1c4..077d1926e7f35 100644
--- a/llvm/include/llvm/ADT/APInt.h
+++ b/llvm/include/llvm/ADT/APInt.h
@@ -2294,6 +2294,12 @@ LLVM_ABI APInt mulhs(const APInt &C1, const APInt &C2);
 /// Returns the high N bits of the multiplication result.
 LLVM_ABI APInt mulhu(const APInt &C1, const APInt &C2);
 
+/// Performs (2*N)-bit multiplication on sign-extended operands.
+LLVM_ABI APInt mulxs(const APInt &C1, const APInt &C2);
+
+/// Performs (2*N)-bit multiplication on zero-extended operands.
+LLVM_ABI APInt mulxu(const APInt &C1, const APInt &C2);
+
 /// Compute X^N for N>=0.
 /// 0^0 is supported and returns 1.
 LLVM_ABI APInt pow(const APInt &X, int64_t N);
diff --git a/llvm/lib/Support/APInt.cpp b/llvm/lib/Support/APInt.cpp
index 954af7fff92a8..5902bb0449868 100644
--- a/llvm/lib/Support/APInt.cpp
+++ b/llvm/lib/Support/APInt.cpp
@@ -3136,6 +3136,22 @@ APInt APIntOps::mulhu(const APInt &C1, const APInt &C2) {
   return (C1Ext * C2Ext).extractBits(C1.getBitWidth(), C1.getBitWidth());
 }
 
+APInt APIntOps::mulxs(const APInt &C1, const APInt &C2) {
+  assert(C1.getBitWidth() == C2.getBitWidth() && "Unequal bitwidths");
+  unsigned FullWidth = C1.getBitWidth() * 2;
+  APInt C1Ext = C1.sext(FullWidth);
+  APInt C2Ext = C2.sext(FullWidth);
+  return C1Ext * C2Ext;
+}
+
+APInt APIntOps::mulxu(const APInt &C1, const APInt &C2) {
+  assert(C1.getBitWidth() == C2.getBitWidth() && "Unequal bitwidths");
+  unsigned FullWidth = C1.getBitWidth() * 2;
+  APInt C1Ext = C1.zext(FullWidth);
+  APInt C2Ext = C2.zext(FullWidth);
+  return C1Ext * C2Ext;
+}
+
 APInt APIntOps::pow(const APInt &X, int64_t N) {
   assert(N >= 0 && "negative exponents not supported.");
   APInt Acc = APInt(X.getBitWidth(), 1);
diff --git a/llvm/unittests/ADT/APIntTest.cpp b/llvm/unittests/ADT/APIntTest.cpp
index 4741c7bcc140f..d68ae339d1575 100644
--- a/llvm/unittests/ADT/APIntTest.cpp
+++ b/llvm/unittests/ADT/APIntTest.cpp
@@ -3103,6 +3103,38 @@ TEST(APIntOpsTest, Mulh) {
   EXPECT_EQ(APInt(128, "FFEB498812C66C68D4552DB89B8EBF8F", 16), i128Res);
 }
 
+TEST(APIntOpsTest, mulx) {
+  APInt u32a(32, 0x0001'E235);
+  APInt u32b(32, 0xF623'55AD);
+  EXPECT_EQ(0x0001'CFA1'7CA0'76D1, APIntOps::mulxu(u32a, u32b));
+
+  APInt u64a(64, 0x1234'5678'90AB'CDEF);
+  APInt u64b(64, 0xFEDC'BA09'8765'4321);
+  EXPECT_EQ(APInt(128, "121FA000A3723A57C24A442FE55618CF", 16), APIntOps::mulxu(u64a, u64b));
+
+  APInt u128a(128, "1234567890ABCDEF1234567890ABCDEF", 16);
+  APInt u128b(128, "FEDCBA0987654321FEDCBA0987654321", 16);
+  EXPECT_EQ(APInt(256, "121FA000A3723A57E68984312C3A8D7E96B428606E1E6BF5C24A442FE55618CF", 16), APIntOps::mulxu(u128a, u128b));
+
+  APInt s32a(32, 0x1234'5678);
+  APInt s32b(32, 0x10AB'CDEF);
+  APInt s32c(32, 0xFEDC'BA09);
+  EXPECT_EQ(0x012F'7D02'2A42'D208, APIntOps::mulxs(s32a, s32b));
+  EXPECT_EQ(0xFFEB'4988'09CA'3A38, APIntOps::mulxs(s32a, s32c));
+
+  APInt s64a(64, 0x1234'5678'90AB'CDEF);
+  APInt s64b(64, 0x1234'5678'90FE'DCBA);
+  APInt s64c(64, 0xFEDC'BA09'8765'4321);
+  EXPECT_EQ(APInt(128, "014B66DC328E10C1FB99704184EF03A6", 16), APIntOps::mulxs(s64a, s64b));
+  EXPECT_EQ(APInt(128, "FFEB498812C66C68C24A442FE55618CF", 16), APIntOps::mulxs(s64a, s64c));
+
+  APInt s128a(128, "1234567890ABCDEF1234567890ABCDEF", 16);
+  APInt s128b(128, "1234567890FEDCBA1234567890FEDCBA", 16);
+  APInt s128c(128, "FEDCBA0987654321FEDCBA0987654321", 16);
+  EXPECT_EQ(APInt(256, "014B66DC328E10C1FE303DF9EA0B2529F87E475F3C6C180DFB99704184EF03A6", 16), APIntOps::mulxs(s128a, s128b));
+  EXPECT_EQ(APInt(256, "FFEB498812C66C68D4552DB89B8EBF8F96B428606E1E6BF5C24A442FE55618CF", 16), APIntOps::mulxs(s128a, s128c));
+}
+
 TEST(APIntTest, RoundingUDiv) {
   for (uint64_t Ai = 1; Ai <= 255; Ai++) {
     APInt A(8, Ai);

[github-actions]

✅ With the latest revision this PR passed the C/C++ code formatter.

[kuhar]

@pedroclobo remember to update the PR title and description before merging

[pedroclobo]

@kuhar Ready to merge?
Or should I wait for other approvals?

[kuhar]

Maybe also wait for @RKSimon since he made a similar suggestion in the original PR

[RKSimon]

LGTM. Not a fan of the new names, but if there is precedent then so be it.