[INFRA] Add IntelLLVM CI

.github/workflows/ci_linux.yml

@@ -62,6 +62,12 @@ jobs:
             build: unit
             build_type: Release
 
+          - name: "IntelLLVM"
+            compiler: "intel"
+            build: unit
+            build_type: Release
+            cxx_flags: "-fp-model=strict"
+
     steps:
       - name: Checkout SeqAn3
         uses: actions/checkout@v4
@@ -84,6 +90,7 @@ jobs:
           ccache_size: 75M
 
       - name: Install CMake
+        if: contains(matrix.compiler, 'intel') == false
         uses: seqan/actions/setup-cmake@main
         with:
           cmake: 3.16.9

.github/workflows/ci_misc.yml

@@ -48,6 +48,13 @@ jobs:
             build_type: Release
             test_threads: 1 # snippets create and delete files and some separate tests create/delete the same files
 
+          - name: "Snippet IntelLLVM"
+            compiler: "intel"
+            build: snippet
+            build_type: Release
+            test_threads: 1 # snippets create and delete files and some separate tests create/delete the same files
+            cxx_flags: "-fp-model=strict"
+
           - name: "Performance clang17 libc++"
             compiler: "clang-17"
             build: performance
@@ -61,6 +68,13 @@ jobs:
             build_type: Release
             test_threads: 2
 
+          - name: "Performance IntelLLVM"
+            compiler: "intel"
+            build: performance
+            build_type: Release
+            test_threads: 2
+            cxx_flags: "-fp-model=strict"
+
           - name: "Header clang17 libc++"
             compiler: "clang-17"
             build: header
@@ -96,6 +110,7 @@ jobs:
           fetch-depth: 1
 
       - name: Install CMake
+        if: contains(matrix.compiler, 'intel') == false
         uses: seqan/actions/setup-cmake@main
         with:
           cmake: 3.16.9

include/seqan3/io/detail/magic_header.hpp

@@ -10,14 +10,14 @@
 #pragma once
 
 #include <array>
+#include <bit>
 #include <span>
 #include <string>
 #include <type_traits>
 #include <vector>
 
 #include <seqan3/core/detail/template_inspection.hpp>
 #include <seqan3/utility/concept.hpp>
-#include <seqan3/utility/detail/to_little_endian.hpp>
 #include <seqan3/utility/type_pack/traits.hpp>
 
 namespace seqan3::detail
@@ -97,14 +97,16 @@ struct bgzf_compression
         static_assert(std::equality_comparable_with<char_t, char>,
                       "The given char type of the span must be comparable with char.");
 
-        return (header[0] == magic_header[0] &&                                                           // GZ_ID1
-                header[1] == magic_header[1] &&                                                           // GZ_ID2
-                header[2] == magic_header[2] &&                                                           // GZ_CM
-                (header[3] & magic_header[3]) != 0 &&                                                     // FLG_FEXTRA
-                to_little_endian(*reinterpret_cast<uint16_t const *>(&header[10])) == magic_header[10] && // BGZF_ID1
-                header[12] == magic_header[12] &&                                                         // BGZF_ID2
-                header[13] == magic_header[13] &&                                                         // BGZF_SLEN
-                to_little_endian(*reinterpret_cast<uint16_t const *>(&header[14])) == magic_header[14]);  // BGZF_XLEN
+        static constexpr auto id1_pos = std::endian::native == std::endian::little ? 10 : 11;
+
+        return (header[0] == magic_header[0] &&           // GZ_ID1
+                header[1] == magic_header[1] &&           // GZ_ID2
+                header[2] == magic_header[2] &&           // GZ_CM
+                (header[3] & magic_header[3]) != 0 &&     // FLG_FEXTRA
+                header[id1_pos] == magic_header[10] &&    // BGZF_ID1
+                header[12] == magic_header[12] &&         // BGZF_ID2
+                header[13] == magic_header[13] &&         // BGZF_SLEN
+                header[id1_pos + 4] == magic_header[14]); // BGZF_XLEN
     }
 };
 

test/performance/alignment/global_affine_alignment_parallel_benchmark.cpp

@@ -157,8 +157,8 @@ void seqan2_affine_dna4_parallel(benchmark::State & state)
 BENCHMARK_TEMPLATE(seqan2_affine_dna4_parallel, score)->UseRealTime();
 #endif // SEQAN3_HAS_SEQAN2
 
-// Crashes with libc++
-#if defined(SEQAN3_HAS_SEQAN2) && defined(_OPENMP) && !defined(_LIBCPP_VERSION)
+// Crashes with libc++ or IntelLLVM
+#if defined(SEQAN3_HAS_SEQAN2) && defined(_OPENMP) && !defined(_LIBCPP_VERSION) && !defined(__INTEL_LLVM_COMPILER)
 template <typename result_t>
 void seqan2_affine_dna4_omp_for(benchmark::State & state)
 {
@@ -199,7 +199,7 @@ void seqan2_affine_dna4_omp_for(benchmark::State & state)
 
 BENCHMARK_TEMPLATE(seqan2_affine_dna4_omp_for, score)->UseRealTime();
 BENCHMARK_TEMPLATE(seqan2_affine_dna4_omp_for, trace)->UseRealTime();
-#endif // defined(SEQAN3_HAS_SEQAN2) && defined(_OPENMP) && !defined(_LIBCPP_VERSION)
+#endif // defined(SEQAN3_HAS_SEQAN2) && defined(_OPENMP) && !defined(_LIBCPP_VERSION) && !defined(__INTEL_LLVM_COMPILER)
 
 // ============================================================================
 //  instantiate tests

test/seqan3-test.cmake

@@ -11,6 +11,7 @@ cmake_minimum_required (VERSION 3.10)
 # require SeqAn3 package
 find_package (SeqAn3 REQUIRED HINTS ${CMAKE_CURRENT_LIST_DIR}/../build_system)
 
+include (CheckCXXCompilerFlag)
 include (CheckCXXSourceCompiles)
 include (FindPackageHandleStandardArgs)
 include (FindPackageMessage)
@@ -80,7 +81,8 @@ if (NOT TARGET seqan3::test::performance)
     # std::views::join is experimental in libc++
     target_compile_definitions (seqan3_test_performance INTERFACE _LIBCPP_ENABLE_EXPERIMENTAL)
 
-    if (SEQAN3_BENCHMARK_ALIGN_LOOPS)
+    check_cxx_compiler_flag ("-falign-loops=32" SEQAN3_HAS_FALIGN_LOOPS)
+    if (SEQAN3_BENCHMARK_ALIGN_LOOPS AND SEQAN3_HAS_FALIGN_LOOPS)
         target_compile_options (seqan3_test_performance INTERFACE "-falign-loops=32")
     endif ()