[APInt] Introduce carry-less multiply primitives #168527

artagnon · 2025-11-18T12:36:39Z

In line with a std proposal to introduce std::clmul, and in preparation to introduce a clmul intrinsic, implement carry-less multiply primitives for APIntOps, clmul[rh].

Ref: https://isocpp.org/files/papers/P3642R3.html

In line with a std proposal to introduce std::clmul, and in preparation to introduce a clmul intrinsic, implement a carry-less multiply for APIntOps. Ref: https://www.open-std.org/jtc1/sc22/wg21/docs/papers/2025/p3642r2.html

llvmbot · 2025-11-18T12:37:12Z

@llvm/pr-subscribers-llvm-support

Author: Ramkumar Ramachandra (artagnon)

Changes

In line with a std proposal to introduce std::clmul, and in preparation to introduce a clmul intrinsic, implement a carry-less multiply for APIntOps.

Ref: https://www.open-std.org/jtc1/sc22/wg21/docs/papers/2025/p3642r2.html

Full diff: https://github.com/llvm/llvm-project/pull/168527.diff

3 Files Affected:

(modified) llvm/include/llvm/ADT/APInt.h (+10)
(modified) llvm/lib/Support/APInt.cpp (+9-1)
(modified) llvm/unittests/ADT/APIntTest.cpp (+28)

diff --git a/llvm/include/llvm/ADT/APInt.h b/llvm/include/llvm/ADT/APInt.h
index fdb3b84b73a1f..0d40d360d96ac 100644
--- a/llvm/include/llvm/ADT/APInt.h
+++ b/llvm/include/llvm/ADT/APInt.h
@@ -2440,6 +2440,16 @@ LLVM_ABI APInt fshl(const APInt &Hi, const APInt &Lo, const APInt &Shift);
 /// (4) fshr(i8 255, i8 0, i8 9)  = fshr(i8 255, i8 0, i8 1) // 9 % 8
 LLVM_ABI APInt fshr(const APInt &Hi, const APInt &Lo, const APInt &Shift);
 
+/// Perform a carry-less multiply, also known as XOR multiplication. All
+/// arguments and result have the same bitwidth.
+///
+/// Examples:
+/// (1) clmul(i4 1, i4 2)   = 2
+/// (2) clmul(i4 5, i4 6)   = 14
+/// (3) clmul(i4 -4, i4 2)  = -8
+/// (4) clmul(i4 -4, i4 -5) = 4
+LLVM_ABI APInt clmul(const APInt &LHS, const APInt &RHS);
+
 } // namespace APIntOps
 
 // See friend declaration above. This additional declaration is required in
diff --git a/llvm/lib/Support/APInt.cpp b/llvm/lib/Support/APInt.cpp
index f6fd5f9ddd633..de17758c8a1cd 100644
--- a/llvm/lib/Support/APInt.cpp
+++ b/llvm/lib/Support/APInt.cpp
@@ -15,10 +15,10 @@
 #include "llvm/ADT/ArrayRef.h"
 #include "llvm/ADT/FoldingSet.h"
 #include "llvm/ADT/Hashing.h"
+#include "llvm/ADT/Sequence.h"
 #include "llvm/ADT/SmallString.h"
 #include "llvm/ADT/StringRef.h"
 #include "llvm/ADT/bit.h"
-#include "llvm/Config/llvm-config.h"
 #include "llvm/Support/Alignment.h"
 #include "llvm/Support/Debug.h"
 #include "llvm/Support/ErrorHandling.h"
@@ -3187,3 +3187,11 @@ APInt llvm::APIntOps::fshr(const APInt &Hi, const APInt &Lo,
     return Lo;
   return Hi.shl(Hi.getBitWidth() - ShiftAmt) | Lo.lshr(ShiftAmt);
 }
+
+APInt llvm::APIntOps::clmul(const APInt &LHS, const APInt &RHS) {
+  assert(LHS.getBitWidth() == RHS.getBitWidth());
+  APInt Result(LHS.getBitWidth(), 0);
+  for (unsigned I : seq<unsigned>(Result.getBitWidth()))
+    Result ^= LHS.shl(I) * (RHS.lshr(I) & 1);
+  return Result;
+}
diff --git a/llvm/unittests/ADT/APIntTest.cpp b/llvm/unittests/ADT/APIntTest.cpp
index ca9f9f17ee112..ddfaaba463a86 100644
--- a/llvm/unittests/ADT/APIntTest.cpp
+++ b/llvm/unittests/ADT/APIntTest.cpp
@@ -3823,4 +3823,32 @@ TEST(APIntTest, Fshr) {
             -8193);
 }
 
+TEST(APIntTest, clmul) {
+  EXPECT_EQ(APIntOps::clmul(APInt(4, 1), APInt(4, 2)).getZExtValue(), 2U);
+  EXPECT_EQ(APIntOps::clmul(APInt(4, 5), APInt(4, 6)).getZExtValue(), 14U);
+  EXPECT_EQ(APIntOps::clmul(APInt(4, -4, /*isSigned*/ true),
+                            APInt(4, 2, /*isSigned*/ false))
+                .getSExtValue(),
+            -8);
+  EXPECT_EQ(APIntOps::clmul(APInt(4, -4, /*isSigned*/ true),
+                            APInt(4, -5, /*isSigned*/ true))
+                .getSExtValue(),
+            4);
+  EXPECT_EQ(APIntOps::clmul(APInt(8, 0), APInt(8, 255)).getZExtValue(), 0U);
+  EXPECT_EQ(APIntOps::clmul(APInt(8, 15), APInt(8, 15)).getZExtValue(), 85U);
+  EXPECT_EQ(APIntOps::clmul(APInt(8, 1), APInt(8, 2)).getZExtValue(), 2U);
+  EXPECT_EQ(APIntOps::clmul(APInt(64, 0, /*isSigned*/ true),
+                            APInt(64, 9223372036854775807, /*isSigned*/ true))
+                .getSExtValue(),
+            0);
+  EXPECT_EQ(APIntOps::clmul(APInt(64, 1, /*isSigned*/ true),
+                            APInt(64, 2, /*isSigned*/ true))
+                .getSExtValue(),
+            2);
+  EXPECT_EQ(APIntOps::clmul(APInt(16, -2, /*isSigned*/ true),
+                            APInt(16, -1, /*isSigned*/ true))
+                .getSExtValue(),
+            -21846);
+}
+
 } // end anonymous namespace

llvmbot · 2025-11-18T12:37:13Z

@llvm/pr-subscribers-llvm-adt

Author: Ramkumar Ramachandra (artagnon)

Changes

In line with a std proposal to introduce std::clmul, and in preparation to introduce a clmul intrinsic, implement a carry-less multiply for APIntOps.

Ref: https://www.open-std.org/jtc1/sc22/wg21/docs/papers/2025/p3642r2.html

Full diff: https://github.com/llvm/llvm-project/pull/168527.diff

3 Files Affected:

(modified) llvm/include/llvm/ADT/APInt.h (+10)
(modified) llvm/lib/Support/APInt.cpp (+9-1)
(modified) llvm/unittests/ADT/APIntTest.cpp (+28)

diff --git a/llvm/include/llvm/ADT/APInt.h b/llvm/include/llvm/ADT/APInt.h
index fdb3b84b73a1f..0d40d360d96ac 100644
--- a/llvm/include/llvm/ADT/APInt.h
+++ b/llvm/include/llvm/ADT/APInt.h
@@ -2440,6 +2440,16 @@ LLVM_ABI APInt fshl(const APInt &Hi, const APInt &Lo, const APInt &Shift);
 /// (4) fshr(i8 255, i8 0, i8 9)  = fshr(i8 255, i8 0, i8 1) // 9 % 8
 LLVM_ABI APInt fshr(const APInt &Hi, const APInt &Lo, const APInt &Shift);
 
+/// Perform a carry-less multiply, also known as XOR multiplication. All
+/// arguments and result have the same bitwidth.
+///
+/// Examples:
+/// (1) clmul(i4 1, i4 2)   = 2
+/// (2) clmul(i4 5, i4 6)   = 14
+/// (3) clmul(i4 -4, i4 2)  = -8
+/// (4) clmul(i4 -4, i4 -5) = 4
+LLVM_ABI APInt clmul(const APInt &LHS, const APInt &RHS);
+
 } // namespace APIntOps
 
 // See friend declaration above. This additional declaration is required in
diff --git a/llvm/lib/Support/APInt.cpp b/llvm/lib/Support/APInt.cpp
index f6fd5f9ddd633..de17758c8a1cd 100644
--- a/llvm/lib/Support/APInt.cpp
+++ b/llvm/lib/Support/APInt.cpp
@@ -15,10 +15,10 @@
 #include "llvm/ADT/ArrayRef.h"
 #include "llvm/ADT/FoldingSet.h"
 #include "llvm/ADT/Hashing.h"
+#include "llvm/ADT/Sequence.h"
 #include "llvm/ADT/SmallString.h"
 #include "llvm/ADT/StringRef.h"
 #include "llvm/ADT/bit.h"
-#include "llvm/Config/llvm-config.h"
 #include "llvm/Support/Alignment.h"
 #include "llvm/Support/Debug.h"
 #include "llvm/Support/ErrorHandling.h"
@@ -3187,3 +3187,11 @@ APInt llvm::APIntOps::fshr(const APInt &Hi, const APInt &Lo,
     return Lo;
   return Hi.shl(Hi.getBitWidth() - ShiftAmt) | Lo.lshr(ShiftAmt);
 }
+
+APInt llvm::APIntOps::clmul(const APInt &LHS, const APInt &RHS) {
+  assert(LHS.getBitWidth() == RHS.getBitWidth());
+  APInt Result(LHS.getBitWidth(), 0);
+  for (unsigned I : seq<unsigned>(Result.getBitWidth()))
+    Result ^= LHS.shl(I) * (RHS.lshr(I) & 1);
+  return Result;
+}
diff --git a/llvm/unittests/ADT/APIntTest.cpp b/llvm/unittests/ADT/APIntTest.cpp
index ca9f9f17ee112..ddfaaba463a86 100644
--- a/llvm/unittests/ADT/APIntTest.cpp
+++ b/llvm/unittests/ADT/APIntTest.cpp
@@ -3823,4 +3823,32 @@ TEST(APIntTest, Fshr) {
             -8193);
 }
 
+TEST(APIntTest, clmul) {
+  EXPECT_EQ(APIntOps::clmul(APInt(4, 1), APInt(4, 2)).getZExtValue(), 2U);
+  EXPECT_EQ(APIntOps::clmul(APInt(4, 5), APInt(4, 6)).getZExtValue(), 14U);
+  EXPECT_EQ(APIntOps::clmul(APInt(4, -4, /*isSigned*/ true),
+                            APInt(4, 2, /*isSigned*/ false))
+                .getSExtValue(),
+            -8);
+  EXPECT_EQ(APIntOps::clmul(APInt(4, -4, /*isSigned*/ true),
+                            APInt(4, -5, /*isSigned*/ true))
+                .getSExtValue(),
+            4);
+  EXPECT_EQ(APIntOps::clmul(APInt(8, 0), APInt(8, 255)).getZExtValue(), 0U);
+  EXPECT_EQ(APIntOps::clmul(APInt(8, 15), APInt(8, 15)).getZExtValue(), 85U);
+  EXPECT_EQ(APIntOps::clmul(APInt(8, 1), APInt(8, 2)).getZExtValue(), 2U);
+  EXPECT_EQ(APIntOps::clmul(APInt(64, 0, /*isSigned*/ true),
+                            APInt(64, 9223372036854775807, /*isSigned*/ true))
+                .getSExtValue(),
+            0);
+  EXPECT_EQ(APIntOps::clmul(APInt(64, 1, /*isSigned*/ true),
+                            APInt(64, 2, /*isSigned*/ true))
+                .getSExtValue(),
+            2);
+  EXPECT_EQ(APIntOps::clmul(APInt(16, -2, /*isSigned*/ true),
+                            APInt(16, -1, /*isSigned*/ true))
+                .getSExtValue(),
+            -21846);
+}
+
 } // end anonymous namespace

eisenwave · 2025-11-18T12:59:03Z

https://isocpp.org/files/papers/P3642R3.html#possible-implementation There's a more recent version of the proposal by the way. The naive implementation there is a bit better than the one in R2.

artagnon · 2025-11-18T13:26:41Z

https://isocpp.org/files/papers/P3642R3.html#possible-implementation There's a more recent version of the proposal by the way. The naive implementation there is a bit better than the one in R2.

I get a 500 error on that page: can you kindly post the text if you have access?

llvm/lib/Support/APInt.cpp

dtcxzyw · 2025-11-18T13:31:29Z

https://isocpp.org/files/papers/P3642R3.html#possible-implementation There's a more recent version of the proposal by the way. The naive implementation there is a bit better than the one in R2.

I get a 500 error on that page: can you kindly post the text if you have access?

Cloudflare is down :)

eisenwave · 2025-11-18T13:40:13Z

https://eisenwave.github.io/cpp-proposals/clmul.html also has the latest version, and seems to work despite cloudflare outage.

The solution there is only suitable for small integers though; it relies on multiplication. If the concern is efficiency for large integers, we have a problem.

github-actions · 2025-11-18T13:50:18Z

🐧 Linux x64 Test Results

186295 tests passed
4853 tests skipped

dtcxzyw

LGTM. Thanks.

[APInt] Introduce carry-less multiply operation

0fd708d

In line with a std proposal to introduce std::clmul, and in preparation to introduce a clmul intrinsic, implement a carry-less multiply for APIntOps. Ref: https://www.open-std.org/jtc1/sc22/wg21/docs/papers/2025/p3642r2.html

artagnon requested review from antoniofrighetto, dtcxzyw and pfusik November 18, 2025 12:36

llvmbot added llvm:support llvm:adt labels Nov 18, 2025

artagnon mentioned this pull request Nov 18, 2025

[IR] Add llvm clmul intrinsic #140301

Open

dtcxzyw reviewed Nov 18, 2025

View reviewed changes

llvm/lib/Support/APInt.cpp Outdated Show resolved Hide resolved

llvm/lib/Support/APInt.cpp Outdated Show resolved Hide resolved

[APInt] Avoid a shift with a branch

a1ee9e0

artagnon changed the title ~~[APInt] Introduce carry-less multiply operation~~ [APInt] Introduce carry-less multiply primitives Nov 18, 2025

[APInt] Also introduce clmul[rh]

338998c

dtcxzyw approved these changes Nov 18, 2025

View reviewed changes

artagnon merged commit 727ee7e into llvm:main Nov 18, 2025
10 checks passed

artagnon deleted the apint-clmul branch November 18, 2025 16:54

RKSimon mentioned this pull request Nov 19, 2025

[Clang] VectorExprEvaluator::VisitCallExpr / InterpretBuiltin - allow PCLMULQDQ intrinsics to be used in constexpr #168741

Open

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

Uh oh!

[APInt] Introduce carry-less multiply primitives #168527

[APInt] Introduce carry-less multiply primitives #168527

Uh oh!

artagnon commented Nov 18, 2025 •

edited

Loading

Uh oh!

llvmbot commented Nov 18, 2025

Uh oh!

llvmbot commented Nov 18, 2025

Uh oh!

eisenwave commented Nov 18, 2025

Uh oh!

artagnon commented Nov 18, 2025

Uh oh!

Uh oh!

Uh oh!

dtcxzyw commented Nov 18, 2025

Uh oh!

eisenwave commented Nov 18, 2025 •

edited

Loading

Uh oh!

github-actions bot commented Nov 18, 2025 •

edited

Loading

Uh oh!

dtcxzyw left a comment

Uh oh!

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

4 participants

[APInt] Introduce carry-less multiply primitives #168527

[APInt] Introduce carry-less multiply primitives #168527

Uh oh!

Conversation

artagnon commented Nov 18, 2025 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

llvmbot commented Nov 18, 2025

Uh oh!

llvmbot commented Nov 18, 2025

Uh oh!

eisenwave commented Nov 18, 2025

Uh oh!

artagnon commented Nov 18, 2025

Uh oh!

Uh oh!

Uh oh!

dtcxzyw commented Nov 18, 2025

Uh oh!

eisenwave commented Nov 18, 2025 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

github-actions bot commented Nov 18, 2025 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

🐧 Linux x64 Test Results

Uh oh!

dtcxzyw left a comment

Choose a reason for hiding this comment

Uh oh!

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

4 participants

artagnon commented Nov 18, 2025 •

edited

Loading

eisenwave commented Nov 18, 2025 •

edited

Loading

github-actions bot commented Nov 18, 2025 •

edited

Loading