fix: harden fast-math reassociation barriers

DiamonDinoia · DiamonDinoia · commit 91b408196ed4 · 2026-03-23T15:14:43.000-04:00
diff --git a/.github/workflows/windows.yml b/.github/workflows/windows.yml
@@ -92,6 +92,70 @@ jobs:
         run: |
           cd _build && ./test/test_xsimd
 
+  build-windows-clang-cl:
+    name: 'clang-cl x64 AVX2'
+    defaults:
+      run:
+        shell: bash {0}
+    runs-on: windows-2025
+    steps:
+    - name: Setup compiler
+      uses: ilammy/msvc-dev-cmd@v1
+      with:
+        arch: amd64
+    - name: Check clang-cl
+      run: |
+        command -v clang-cl
+        clang-cl --version
+    - name: Setup Ninja
+      run: |
+        python3 -m pip install --upgrade pip setuptools wheel
+        python3 -m pip install ninja
+    - name: Checkout xsimd
+      uses: actions/checkout@v3
+    - name: Setup
+      run: |
+        mkdir _build
+        cd _build && cmake .. -DBUILD_TESTS=ON -DDOWNLOAD_DOCTEST=ON -DBUILD_BENCHMARK=ON -DBUILD_EXAMPLES=ON -DCMAKE_BUILD_TYPE=Release -DCMAKE_C_COMPILER=clang-cl -DCMAKE_CXX_COMPILER=clang-cl -DCMAKE_CXX_FLAGS="/arch:AVX2" -G Ninja
+    - name: Build
+      run: |
+        cd _build && cmake --build .
+    - name: Testing xsimd
+      run: |
+        cd _build && ./test/test_xsimd
+
+  build-windows-clang-cl-fast-math:
+    name: 'clang-cl x64 /fp:fast'
+    defaults:
+      run:
+        shell: bash {0}
+    runs-on: windows-2025
+    steps:
+    - name: Setup compiler
+      uses: ilammy/msvc-dev-cmd@v1
+      with:
+        arch: amd64
+    - name: Check clang-cl
+      run: |
+        command -v clang-cl
+        clang-cl --version
+    - name: Setup Ninja
+      run: |
+        python3 -m pip install --upgrade pip setuptools wheel
+        python3 -m pip install ninja
+    - name: Checkout xsimd
+      uses: actions/checkout@v3
+    - name: Setup
+      run: |
+        mkdir _build
+        cd _build && cmake .. -DBUILD_TESTS=ON -DDOWNLOAD_DOCTEST=ON -DBUILD_BENCHMARK=OFF -DBUILD_EXAMPLES=OFF -DCMAKE_BUILD_TYPE=Release -DCMAKE_C_COMPILER=clang-cl -DCMAKE_CXX_COMPILER=clang-cl -DCMAKE_CXX_FLAGS="/fp:fast" -G Ninja
+    - name: Build
+      run: |
+        cd _build && cmake --build .
+    - name: Testing xsimd
+      run: |
+        cd _build && ./test/test_xsimd
+
   build-windows-arm64:
     name: 'MSVC arm64'
     defaults:
diff --git a/include/xsimd/arch/common/xsimd_common_details.hpp b/include/xsimd/arch/common/xsimd_common_details.hpp
@@ -111,6 +111,22 @@ namespace xsimd
 
         namespace detail
         {
+            template <class T>
+            XSIMD_INLINE void reassociation_barrier(T& x, memory_barrier_tag) noexcept
+            {
+#if XSIMD_WITH_INLINE_ASM
+                __asm__ volatile("" : : "r"(&x) : "memory");
+#else
+                (void)x;
+#endif
+            }
+
+            template <class T, class A>
+            XSIMD_INLINE void reassociation_barrier(T& x, A const&) noexcept
+            {
+                detail::reassociation_barrier(x, memory_barrier_tag {});
+            }
+
             template <class F, class A, class T, class... Batches>
             XSIMD_INLINE batch<T, A> apply(F&& func, batch<T, A> const& self, batch<T, A> const& other) noexcept
             {
diff --git a/include/xsimd/arch/common/xsimd_common_math.hpp b/include/xsimd/arch/common/xsimd_common_math.hpp
@@ -1900,17 +1900,9 @@ namespace xsimd
                 batch_type s = bitofsign(self);
                 batch_type v = self ^ s;
                 batch_type t2n = constants::twotonmb<batch_type>();
-                // Under fast-math, reordering is possible and the compiler optimizes d
-                // to v. That's not what we want, so prevent compiler optimization here.
-                // FIXME: it may be better to emit a memory barrier here (?).
-#ifdef __FAST_MATH__
                 batch_type d0 = v + t2n;
-                asm volatile("" ::"r"(&d0) : "memory");
+                detail::reassociation_barrier(d0.data, A {});
                 batch_type d = d0 - t2n;
-#else
-                batch_type d0 = v + t2n;
-                batch_type d = d0 - t2n;
-#endif
                 return s ^ select(v < t2n, d, v);
             }
         }
diff --git a/include/xsimd/arch/xsimd_avx2.hpp b/include/xsimd/arch/xsimd_avx2.hpp
@@ -554,11 +554,8 @@ namespace xsimd
                 __m256d f = _mm256_sub_pd(_mm256_castsi256_pd(xH), _mm256_set1_pd(19342813118337666422669312.)); //  2^84 + 2^52
                 // With -ffast-math, the compiler may reassociate (xH-C)+xL into
                 // xH+(xL-C). Since xL<<C this causes catastrophic cancellation.
-                // The asm barrier forces f into a register before the add, blocking
-                // the reorder. It emits zero instructions.
-#if defined(__GNUC__)
-                __asm__ volatile("" : "+x"(f));
-#endif
+                // Barrier the intermediate before the final add.
+                detail::reassociation_barrier(f, avx2 {});
                 return _mm256_add_pd(f, _mm256_castsi256_pd(xL));
             }
 
@@ -575,9 +572,7 @@ namespace xsimd
                 __m256i xL = _mm256_or_si256(_mm256_and_si256(mask, x), _mm256_andnot_si256(mask, _mm256_castpd_si256(_mm256_set1_pd(0x0010000000000000)))); //  2^52
                 __m256d f = _mm256_sub_pd(_mm256_castsi256_pd(xH), _mm256_set1_pd(442726361368656609280.)); //  3*2^67 + 2^52
                 // See above: prevent -ffast-math from reassociating (xH-C)+xL.
-#if defined(__GNUC__)
-                __asm__ volatile("" : "+x"(f));
-#endif
+                detail::reassociation_barrier(f, avx2 {});
                 return _mm256_add_pd(f, _mm256_castsi256_pd(xL));
             }
         }
diff --git a/include/xsimd/arch/xsimd_common_fwd.hpp b/include/xsimd/arch/xsimd_common_fwd.hpp
@@ -23,6 +23,15 @@ namespace xsimd
     class batch;
     template <class T, class A>
     class batch_bool;
+    namespace kernel
+    {
+        namespace detail
+        {
+            struct memory_barrier_tag
+            {
+            };
+        }
+    }
     template <class T, class A, T... Vs>
     struct batch_constant;
     template <class T, class A, bool... Vs>
@@ -101,6 +110,12 @@ namespace xsimd
         // Forward declarations for pack-level helpers
         namespace detail
         {
+            template <class T>
+            XSIMD_INLINE void reassociation_barrier(T& x, memory_barrier_tag) noexcept;
+
+            template <class T, class A>
+            XSIMD_INLINE void reassociation_barrier(T& x, A const&) noexcept;
+
             template <typename T, T... Vs>
             XSIMD_INLINE constexpr bool is_identity() noexcept;
             template <typename T, class A, T... Vs>
@@ -115,7 +130,6 @@ namespace xsimd
             XSIMD_INLINE constexpr bool is_only_from_lo(batch_constant<T, A, Vs...>) noexcept;
             template <typename T, class A, T... Vs>
             XSIMD_INLINE constexpr bool is_only_from_hi(batch_constant<T, A, Vs...>) noexcept;
-
         }
     }
 }
diff --git a/include/xsimd/arch/xsimd_sse2.hpp b/include/xsimd/arch/xsimd_sse2.hpp
@@ -716,6 +716,8 @@ namespace xsimd
                 __m128i mask = _mm_setr_epi16(0xFFFF, 0xFFFF, 0x0000, 0x0000, 0xFFFF, 0xFFFF, 0x0000, 0x0000);
                 __m128i xL = _mm_or_si128(_mm_and_si128(mask, x), _mm_andnot_si128(mask, _mm_castpd_si128(_mm_set1_pd(0x0010000000000000)))); //  2^52
                 __m128d f = _mm_sub_pd(_mm_castsi128_pd(xH), _mm_set1_pd(19342813118337666422669312.)); //  2^84 + 2^52
+                // Prevent -ffast-math from reassociating (xH-C)+xL into xH+(xL-C).
+                detail::reassociation_barrier(f, sse2 {});
                 return _mm_add_pd(f, _mm_castsi128_pd(xL));
             }
 
@@ -730,6 +732,8 @@ namespace xsimd
                 __m128i mask = _mm_setr_epi16(0xFFFF, 0xFFFF, 0xFFFF, 0x0000, 0xFFFF, 0xFFFF, 0xFFFF, 0x0000);
                 __m128i xL = _mm_or_si128(_mm_and_si128(mask, x), _mm_andnot_si128(mask, _mm_castpd_si128(_mm_set1_pd(0x0010000000000000)))); //  2^52
                 __m128d f = _mm_sub_pd(_mm_castsi128_pd(xH), _mm_set1_pd(442726361368656609280.)); //  3*2^67 + 2^52
+                // Prevent -ffast-math from reassociating (xH-C)+xL into xH+(xL-C).
+                detail::reassociation_barrier(f, sse2 {});
                 return _mm_add_pd(f, _mm_castsi128_pd(xL));
             }
 
diff --git a/include/xsimd/arch/xsimd_sse4_1.hpp b/include/xsimd/arch/xsimd_sse4_1.hpp
@@ -64,11 +64,8 @@ namespace xsimd
                 __m128d f = _mm_sub_pd(_mm_castsi128_pd(xH), _mm_set1_pd(442726361368656609280.)); //  3*2^67 + 2^52
                 // With -ffast-math, the compiler may reassociate (xH-C)+xL into
                 // xH+(xL-C). Since xL<<C this causes catastrophic cancellation.
-                // The asm barrier forces f into a register before the add, blocking
-                // the reorder. It emits zero instructions.
-#if defined(__GNUC__)
-                __asm__ volatile("" : "+x"(f));
-#endif
+                // Barrier the intermediate before the final add.
+                detail::reassociation_barrier(f, sse4_1 {});
                 return _mm_add_pd(f, _mm_castsi128_pd(xL));
             }
 
@@ -81,9 +78,7 @@ namespace xsimd
                 __m128i xL = _mm_blend_epi16(x, _mm_castpd_si128(_mm_set1_pd(0x0010000000000000)), 0xcc); //  2^52
                 __m128d f = _mm_sub_pd(_mm_castsi128_pd(xH), _mm_set1_pd(19342813118337666422669312.)); //  2^84 + 2^52
                 // See above: prevent -ffast-math from reassociating (xH-C)+xL.
-#if defined(__GNUC__)
-                __asm__ volatile("" : "+x"(f));
-#endif
+                detail::reassociation_barrier(f, sse4_1 {});
                 return _mm_add_pd(f, _mm_castsi128_pd(xL));
             }
         }
diff --git a/include/xsimd/config/xsimd_config.hpp b/include/xsimd/config/xsimd_config.hpp
@@ -44,6 +44,27 @@
 #define XSIMD_TARGET_X86 0
 #endif
 
+/**
+ * @ingroup xsimd_config_macro
+ *
+ * Set to 1 if GNU-style inline assembly is available, to 0 otherwise.
+ */
+/* Use __clang__ || __GNUC__ for GNU-style inline asm. clang-cl runs in
+ * MSVC-compatibility mode and does not define __GNUC__ by default, but it
+ * still defines __clang__. Clang documents __asm__/__asm__ support and broad
+ * GCC-extension compatibility:
+ * https://clang.llvm.org/docs/LanguageExtensions.html
+ * Clang only emits __GNUC__ when GNUCVersion != 0:
+ * https://raw.githubusercontent.com/llvm/llvm-project/main/clang/lib/Frontend/InitPreprocessor.cpp
+ * and GNUCVersion defaults to 0:
+ * https://raw.githubusercontent.com/llvm/llvm-project/main/clang/include/clang/Basic/LangOptions.def
+ */
+#if defined(__clang__) || defined(__GNUC__)
+#define XSIMD_WITH_INLINE_ASM 1
+#else
+#define XSIMD_WITH_INLINE_ASM 0
+#endif
+
 /**
  * @ingroup xsimd_config_macro
  *
diff --git a/include/xsimd/config/xsimd_cpu_features_x86.hpp b/include/xsimd/config/xsimd_cpu_features_x86.hpp
@@ -533,7 +533,7 @@ namespace xsimd
             __cpuid(buf, leaf);
             std::memcpy(reg.data(), buf, sizeof(buf));
 
-#elif defined(__GNUC__) || defined(__clang__)
+#elif XSIMD_WITH_INLINE_ASM
 
 #if defined(__i386__) && defined(__PIC__)
             // %ebx may be the PIC register
@@ -561,7 +561,7 @@ namespace xsimd
 #error "_MSC_VER < 1400 is not supported"
 #endif
 
-#elif defined(__GNUC__)
+#elif XSIMD_WITH_INLINE_ASM
             x86_reg32_t xcr0 = {};
             __asm__(
                 "xorl %%ecx, %%ecx\n"
