[CPU] CACHE_DIR hash optimization (openvinotoolkit#25624)

### Details: - *JIT implementation of the hash function in the ConstantWriter* ### Tickets: - *127331*
a-sidorova · Oct 21, 2024 · 5f4a445 · 5f4a445
1 parent d34cdda
commit 5f4a445
Show file tree

Hide file tree

Showing 10 changed files with 1,410 additions and 89 deletions.
diff --git a/src/core/CMakeLists.txt b/src/core/CMakeLists.txt
@@ -49,6 +49,9 @@ target_include_directories(openvino_core_dev INTERFACE
     $<BUILD_INTERFACE:${OpenVINO_SOURCE_DIR}/src/common/transformations/include>
     $<BUILD_INTERFACE:${OpenVINO_SOURCE_DIR}/src/common/low_precision_transformations/include>)
 
+target_include_directories(openvino_core_dev SYSTEM INTERFACE
+    $<BUILD_INTERFACE:$<$<TARGET_EXISTS:xbyak::xbyak>:$<TARGET_PROPERTY:xbyak::xbyak,INTERFACE_INCLUDE_DIRECTORIES>>>)
+
 target_link_libraries(openvino_core_dev INTERFACE openvino::itt openvino::util)
 
 set_target_properties(openvino_core_dev PROPERTIES EXPORT_NAME core::dev)
@@ -81,7 +84,7 @@ if(ENABLE_SYSTEM_PUGIXML)
     set_target_properties(openvino_core_obj PROPERTIES NO_SYSTEM_FROM_IMPORTED ON)
 endif()
 
-target_compile_definitions(openvino_core_obj PRIVATE IMPLEMENT_OPENVINO_API)
+target_compile_definitions(openvino_core_obj PRIVATE IMPLEMENT_OPENVINO_API XBYAK_NO_OP_NAMES XBYAK64)
 
 ov_build_target_faster(openvino_core_obj
     UNITY

diff --git a/src/core/dev_api/openvino/runtime/compute_hash.hpp b/src/core/dev_api/openvino/runtime/compute_hash.hpp
@@ -0,0 +1,20 @@
+// Copyright (C) 2018-2024 Intel Corporation
+// SPDX-License-Identifier: Apache-2.0
+//
+
+#pragma once
+
+#include <cstddef>
+
+namespace ov {
+namespace runtime {
+
+/**
+ * @brief Computes the hash value for the input data
+ * @param src  A pointer to the input data
+ * @param size The length of the input data in bytes
+ */
+size_t compute_hash(const void* src, size_t size);
+
+}  // namespace runtime
+}  // namespace ov
diff --git a/src/core/reference/CMakeLists.txt b/src/core/reference/CMakeLists.txt
@@ -50,9 +50,6 @@ target_include_directories(${TARGET_NAME} PUBLIC
     $<BUILD_INTERFACE:${OV_CORE_DEV_API_PATH}>
     $<BUILD_INTERFACE:${OV_CORE_INCLUDE_PATH}>)
 
-target_include_directories(${TARGET_NAME} SYSTEM PRIVATE
-    $<BUILD_INTERFACE:$<$<TARGET_EXISTS:xbyak::xbyak>:$<TARGET_PROPERTY:xbyak::xbyak,INTERFACE_INCLUDE_DIRECTORIES>>>)
-
 find_package(Threads REQUIRED)
 target_link_libraries(${TARGET_NAME} PRIVATE Threads::Threads openvino::core::dev)
 

diff --git a/src/core/reference/src/op/jit_generator.hpp → ...penvino/reference/utils/jit_generator.hpp b/src/core/reference/src/op/jit_generator.hpp → ...penvino/reference/utils/jit_generator.hpp
@@ -15,66 +15,69 @@
 namespace ov {
 namespace reference {
 namespace jit {
-#ifdef XBYAK64
 static const Xbyak::Operand::Code abi_save_gpr_regs[] = {
     Xbyak::Operand::RBX,
     Xbyak::Operand::RBP,
     Xbyak::Operand::R12,
     Xbyak::Operand::R13,
     Xbyak::Operand::R14,
     Xbyak::Operand::R15,
-#    ifdef _WIN32
+#ifdef _WIN32
     Xbyak::Operand::RDI,
     Xbyak::Operand::RSI,
-#    endif
+#endif
 };
 
-#    ifdef _WIN32
-#        define abi_param1 Xbyak::Reg64(Xbyak::Operand::RCX)  // RCX
-#    else
-#        define abi_param1 Xbyak::Reg64(Xbyak::Operand::RDI)  // RDI
-#    endif
-#endif  // XBYAK64
+#ifdef _WIN32
+#    define abi_param1 Xbyak::Reg64(Xbyak::Operand::RCX)  // RCX
+#else
+#    define abi_param1 Xbyak::Reg64(Xbyak::Operand::RDI)  // RDI
+#endif
 
-class Generator : public Xbyak::CodeGenerator {
-    static constexpr size_t xmm_len = 16;
+typedef enum {
+    isa_any,
+    sse42,
+    avx,
+    avx2,
+    avx512_common,
+    avx512_core,
+    avx512_core_vnni,
+    avx512_mic,
+    avx512_mic_4ops,
+    avx512_core_bf16,
+    avx512_vpopcnt,
+    fp16,
+    pclmulqdq,
+    vpclmulqdq
+} cpu_isa_t;
 
+class Generator : public Xbyak::CodeGenerator {
 #ifdef _WIN32
-    static constexpr size_t xmm_to_preserve_start = 6;
-    static constexpr size_t xmm_to_preserve = 10;
+    static constexpr size_t xmm_to_preserve_start = 6llu;
+    static constexpr size_t xmm_to_preserve = 10llu;
 #else
-    static constexpr size_t xmm_to_preserve_start = 0;
-    static constexpr size_t xmm_to_preserve = 0;
+    static constexpr size_t xmm_to_preserve_start = 0lu;
+    static constexpr size_t xmm_to_preserve = 0lu;
 #endif
 
     static const size_t num_abi_save_gpr_regs = sizeof(abi_save_gpr_regs) / sizeof(abi_save_gpr_regs[0]);
     const size_t size_of_abi_save_regs;
 
     const Xbyak::Reg64 reg_EVEX_max_8b_offt;
     static constexpr int EVEX_max_8b_offt = 0x200;
+    size_t m_vlen = ymm_len;
 
 public:
-    const Xbyak::Reg64 param = abi_param1;
+    static constexpr size_t xmm_len = 16lu;
+    static constexpr size_t ymm_len = 32lu;
+    static constexpr size_t zmm_len = 64lu;
 
-    typedef enum {
-        isa_any,
-        sse42,
-        avx,
-        avx2,
-        avx512_common,
-        avx512_core,
-        avx512_core_vnni,
-        avx512_mic,
-        avx512_mic_4ops,
-        avx512_core_bf16,
-        avx512_vpopcnt,
-        fp16
-    } cpu_isa_t;
+    const Xbyak::Reg64 param = abi_param1;
 
     static bool mayiuse(const cpu_isa_t cpu_isa);
     static bool is_x64();
 
-    Generator(void* code_ptr = nullptr, size_t code_size = 16 * 1024);
+    Generator(cpu_isa_t isa = avx2, void* code_ptr = nullptr, size_t code_size = 16lu * 1024lu);
     void preamble();
     void postamble();
 
@@ -85,7 +88,12 @@ class Generator : public Xbyak::CodeGenerator {
 
     template <typename T>
     void copy(const Xbyak::Reg64& dst, const Xbyak::Reg64& src, const Xbyak::Reg64& size);
+
+    size_t get_vlen() {
+        return m_vlen;
+    }
 };
+
 }  // namespace jit
 }  // namespace reference
 }  // namespace ov