Merge pull request #51 from szhorvat/lbfgs-update

ntamas · Nov 20, 2024 · ffa9ea7 · ffa9ea7
2 parents c98d8be + 9151d55
commit ffa9ea7
Show file tree

Hide file tree

Showing 17 changed files with 306 additions and 87 deletions.
diff --git a/.github/workflows/build.yml b/.github/workflows/build.yml
@@ -2,6 +2,10 @@ name: Build and run tests
 
 on: [push, pull_request]
 
+env:
+  CMAKE_COLOR_DIAGNOSTICS: ON
+  OMP_NUM_THREADS: 1
+
 jobs:
   build:
     name: ${{ matrix.config.name }}
@@ -12,21 +16,45 @@ jobs:
       matrix:
         config:
         - {
-            name: "Windows",
+            name: "Windows static",
+            os: windows-latest,
+            build_type: "Release",
+            shared_libs: false,
+            generators: "Visual Studio 17 2022"
+          }
+        - {
+            name: "Ubuntu Linux static",
+            os: ubuntu-latest,
+            build_type: "Release",
+            shared_libs: false,
+            generators: "Ninja"
+          }
+        - {
+            name: "macOS static",
+            os: macos-latest,
+            build_type: "Release",
+            shared_libs: false,
+            generators: "Ninja"
+          }
+        - {
+            name: "Windows shared",
             os: windows-latest,
             build_type: "Release",
+            shared_libs: true,
             generators: "Visual Studio 17 2022"
           }
         - {
-            name: "Ubuntu Linux",
+            name: "Ubuntu Linux shared",
             os: ubuntu-latest,
             build_type: "Release",
+            shared_libs: true,
             generators: "Ninja"
           }
         - {
-            name: "macOS",
+            name: "macOS shared",
             os: macos-latest,
             build_type: "Release",
+            shared_libs: true,
             generators: "Ninja"
           }
 
@@ -56,6 +84,7 @@ jobs:
           cmake \
             -S . \
             -B build \
+            -DBUILD_SHARED_LIBS=${{ matrix.config.shared_libs }} \
             -DCMAKE_BUILD_TYPE=${{ matrix.config.build_type }} \
             -G "${{ matrix.config.generators }}" \
             -DCMAKE_INSTALL_PREFIX:PATH=instdir
@@ -70,13 +99,13 @@ jobs:
           CTEST_OUTPUT_ON_FAILURE: 1
         run: |
           cd build
-          ctest --output-junit report.xml
+          ctest -C ${{ matrix.config.build_type }} --output-junit report.xml
 
       - name: Archive unit test results
         uses: actions/upload-artifact@v4
         if: always()
         with:
-          name: unit-test-results-${{ matrix.config.os }}
+          name: unit-test-results-${{ matrix.config.os }}-${{ matrix.config.shared_libs }}
           path: build/report.xml
 
       - name: Install and strip

diff --git a/src/arithmetic_ansi.h b/src/arithmetic_ansi.h
@@ -23,7 +23,7 @@
  * THE SOFTWARE.
  */
 
-/* $Id: arithmetic_ansi.h 65 2010-01-29 12:19:16Z naoaki $ */
+/* $Id$ */
 
 #include <stdlib.h>
 #include <memory.h>
@@ -51,7 +51,7 @@ inline static void vecfree(void *memblock)
 inline static void vecset(lbfgsfloatval_t *x, const lbfgsfloatval_t c, const int n)
 {
     int i;
-
+    
     for (i = 0;i < n;++i) {
         x[i] = c;
     }

diff --git a/src/arithmetic_sse_double.h b/src/arithmetic_sse_double.h
@@ -23,14 +23,12 @@
  * THE SOFTWARE.
  */
 
-/* $Id: arithmetic_sse_double.h 65 2010-01-29 12:19:16Z naoaki $ */
+/* $Id$ */
 
 #include <stdlib.h>
-
-#if !defined(__APPLE__)
+#ifndef __APPLE__
 #include <malloc.h>
 #endif
-
 #include <memory.h>
 
 #if     1400 <= _MSC_VER
@@ -43,13 +41,15 @@
 
 inline static void* vecalloc(size_t size)
 {
-#ifdef	_WIN32
+#if     defined(_WIN32)
     void *memblock = _aligned_malloc(size, 16);
-#elif defined(__APPLE__)
-	/* Memory on Mac OS X is already aligned to 16 bytes */
-	void *memblock = malloc(size);
+#elif   defined(__APPLE__)  /* OS X always aligns on 16-byte boundaries */
+    void *memblock = malloc(size);
 #else
-    void *memblock = memalign(16, size);
+    void *memblock = NULL, *p = NULL;
+    if (posix_memalign(&p, 16, size) == 0) {
+        memblock = p;
+    }
 #endif
     if (memblock != NULL) {
         memset(memblock, 0, size);
@@ -59,7 +59,7 @@ inline static void* vecalloc(size_t size)
 
 inline static void vecfree(void *memblock)
 {
-#ifdef	_WIN32
+#ifdef	_MSC_VER
     _aligned_free(memblock);
 #else
     free(memblock);
@@ -199,7 +199,7 @@ inline static void vecfree(void *memblock)
 
 
 
-#if     3 <= __SSE__
+#if     3 <= __SSE__ || defined(__SSE3__)
 /*
     Horizontal add with haddps SSE3 instruction. The work register (rw)
     is unused.

diff --git a/src/arithmetic_sse_float.h b/src/arithmetic_sse_float.h
@@ -23,14 +23,12 @@
  * THE SOFTWARE.
  */
 
-/* $Id: arithmetic_sse_float.h 65 2010-01-29 12:19:16Z naoaki $ */
+/* $Id$ */
 
 #include <stdlib.h>
-
-#if !defined(__APPLE__)
+#ifndef __APPLE__
 #include <malloc.h>
 #endif
-
 #include <memory.h>
 
 #if     1400 <= _MSC_VER
@@ -49,7 +47,16 @@
 
 inline static void* vecalloc(size_t size)
 {
+#if     defined(_MSC_VER)
     void *memblock = _aligned_malloc(size, 16);
+#elif   defined(__APPLE__)  /* OS X always aligns on 16-byte boundaries */
+    void *memblock = malloc(size);
+#else
+    void *memblock = NULL, *p = NULL;
+    if (posix_memalign(&p, 16, size) == 0) {
+        memblock = p;
+    }
+#endif
     if (memblock != NULL) {
         memset(memblock, 0, size);
     }
@@ -58,7 +65,11 @@ inline static void* vecalloc(size_t size)
 
 inline static void vecfree(void *memblock)
 {
+#ifdef	_MSC_VER
     _aligned_free(memblock);
+#else
+    free(memblock);
+#endif
 }
 
 #define vecset(x, c, n) \
@@ -189,7 +200,7 @@ inline static void vecfree(void *memblock)
 
 
 
-#if     3 <= __SSE__
+#if     3 <= __SSE__ || defined(__SSE3__)
 /*
     Horizontal add with haddps SSE3 instruction. The work register (rw)
     is unused.