Add missing matmul kernel C files

Author: FrancescAlted

src/blosc2/matmul_kernels.c

@@ -0,0 +1,105 @@
+/*
+ * Copyright (c) 2019-present, Blosc Development Team <blosc@blosc.org>
+ * All rights reserved.
+ *
+ * SPDX-License-Identifier: BSD-3-Clause
+ */
+
+/* This module hosts the backend-selection and accelerator-specific GEMM
+ * wrappers used by the matmul fast path. The portable naive kernel stays in
+ * blosc2_ext.pyx, and external BLAS-style integrations live here.
+ */
+
+#include "matmul_kernels.h"
+static int g_b2_matmul_backend = B2_MATMUL_BACKEND_AUTO;
+
+int b2_has_accelerate(void) {
+#if defined(__APPLE__)
+    return 1;
+#else
+    return 0;
+#endif
+}
+
+void b2_set_matmul_backend(int backend) {
+    switch (backend) {
+        case B2_MATMUL_BACKEND_AUTO:
+        case B2_MATMUL_BACKEND_NAIVE:
+        case B2_MATMUL_BACKEND_ACCELERATE:
+            g_b2_matmul_backend = backend;
+            break;
+        default:
+            g_b2_matmul_backend = B2_MATMUL_BACKEND_AUTO;
+            break;
+    }
+}
+
+int b2_get_matmul_backend(void) {
+    return g_b2_matmul_backend;
+}
+
+int b2_get_selected_matmul_backend(void) {
+    if (g_b2_matmul_backend == B2_MATMUL_BACKEND_ACCELERATE && !b2_has_accelerate()) {
+        return B2_MATMUL_BACKEND_NAIVE;
+    }
+    if (g_b2_matmul_backend == B2_MATMUL_BACKEND_AUTO) {
+        return b2_has_accelerate() ? B2_MATMUL_BACKEND_ACCELERATE : B2_MATMUL_BACKEND_NAIVE;
+    }
+    return g_b2_matmul_backend;
+}
+
+const char *b2_get_matmul_backend_name(void) {
+    switch (g_b2_matmul_backend) {
+        case B2_MATMUL_BACKEND_NAIVE:
+            return "naive";
+        case B2_MATMUL_BACKEND_ACCELERATE:
+            return "accelerate";
+        case B2_MATMUL_BACKEND_AUTO:
+        default:
+            return "auto";
+    }
+}
+
+const char *b2_get_selected_matmul_backend_name(void) {
+    switch (b2_get_selected_matmul_backend()) {
+        case B2_MATMUL_BACKEND_ACCELERATE:
+            return "accelerate";
+        case B2_MATMUL_BACKEND_NAIVE:
+        default:
+            return "naive";
+    }
+}
+
+#if defined(__APPLE__)
+#include <Accelerate/Accelerate.h>
+
+int b2_gemm_accelerate_f32(const float *A, const float *B, float *C, int M, int K, int N) {
+    cblas_sgemm(CblasRowMajor, CblasNoTrans, CblasNoTrans, M, N, K, 1.0f, A, K, B, N, 1.0f, C, N);
+    return 0;
+}
+
+int b2_gemm_accelerate_f64(const double *A, const double *B, double *C, int M, int K, int N) {
+    cblas_dgemm(CblasRowMajor, CblasNoTrans, CblasNoTrans, M, N, K, 1.0, A, K, B, N, 1.0, C, N);
+    return 0;
+}
+#else
+int b2_gemm_accelerate_f32(const float *A, const float *B, float *C, int M, int K, int N) {
+    (void)A;
+    (void)B;
+    (void)C;
+    (void)M;
+    (void)K;
+    (void)N;
+    return -1;
+}
+
+int b2_gemm_accelerate_f64(const double *A, const double *B, double *C, int M, int K, int N) {
+    (void)A;
+    (void)B;
+    (void)C;
+    (void)M;
+    (void)K;
+    (void)N;
+    return -1;
+}
+#endif

src/blosc2/matmul_kernels.h

@@ -0,0 +1,39 @@
+/*
+ * Copyright (c) 2019-present, Blosc Development Team <blosc@blosc.org>
+ * All rights reserved.
+ *
+ * SPDX-License-Identifier: BSD-3-Clause
+ *
+ * This header declares the small backend-selection layer for matmul block
+ * acceleration. The portable naive kernel remains in blosc2_ext.pyx, while
+ * platform BLAS backends such as Accelerate are exposed through this module.
+ */
+
+#ifndef PYTHON_BLOSC2_MATMUL_KERNELS_H
+#define PYTHON_BLOSC2_MATMUL_KERNELS_H
+
+#ifdef __cplusplus
+extern "C" {
+#endif
+
+typedef enum b2_matmul_backend {
+    B2_MATMUL_BACKEND_AUTO = 0,
+    B2_MATMUL_BACKEND_NAIVE = 1,
+    B2_MATMUL_BACKEND_ACCELERATE = 2,
+} b2_matmul_backend;
+
+int b2_has_accelerate(void);
+void b2_set_matmul_backend(int backend);
+int b2_get_matmul_backend(void);
+int b2_get_selected_matmul_backend(void);
+const char *b2_get_matmul_backend_name(void);
+const char *b2_get_selected_matmul_backend_name(void);
+
+int b2_gemm_accelerate_f32(const float *A, const float *B, float *C, int M, int K, int N);
+int b2_gemm_accelerate_f64(const double *A, const double *B, double *C, int M, int K, int N);
+
+#ifdef __cplusplus
+}
+#endif
+
+#endif