v4.17.0 (#15)

hailo-ai · Apr 1, 2024 · a984e26 · a984e26
1 parent f2aad50
commit a984e26
Show file tree

Hide file tree

Showing 361 changed files with 19,220 additions and 19,592 deletions.
diff --git a/.hailort.jpg b/.hailort.jpg
diff --git a/CMakeLists.txt b/CMakeLists.txt
@@ -1,8 +1,12 @@
 cmake_minimum_required(VERSION 3.0.0)
 
 find_program(CCACHE_PROGRAM ccache)
+find_program(CLACHE_PROGRAM clcache)
+
 if(CCACHE_PROGRAM)
     set_property(GLOBAL PROPERTY RULE_LAUNCH_COMPILE "${CCACHE_PROGRAM}")
+elseif(CLCACHE_PROGRAM)
+    set_property(GLOBAL PROPERTY RULE_LAUNCH_COMPILE "${CLCACHE_PROGRAM}")
 endif()
 
 project(HailoRT)

diff --git a/common/include/context_switch_defs.h b/common/include/context_switch_defs.h
@@ -56,6 +56,13 @@ extern "C" {
 
 #define CONTEXT_SWITCH_DEFS__WRITE_ACTION_BY_TYPE_MAX_SIZE (4)
 
+// TODO HRT-12512: Update variable when / If DDR has it's own CMA region
+#define CONTEXT_SWITCH_DEFS__START_M4_MAPPED_DDR_ADDRESS (0x80000000)
+#define CONTEXT_SWITCH_DEFS__END_M4_MAPPED_DDR_ADDRESS (0x90000000)
+#define CONTEXT_SWITCH_DEFS__START_M4_MAPPED_DDR_ADDRESS_AFTER_LUT (0x50000000)
+#define CONTEXT_SWITCH_DEFS__DDR_ADDRESS_MASK (0x0FFFFFFF)
+#define CONTEXT_SWITCH_DEFS__INVALID_DDR_CONTEXTS_BUFFER_ADDRESS (0)
+
 
 #pragma pack(push, 1)
 typedef struct {
@@ -207,7 +214,7 @@ typedef struct {
     uint32_t kernel_done_count;
 } CONTEXT_SWITCH_DEFS__enable_lcu_action_non_default_data_t;
 
-/* Default action - kernel_done_address and kernel_done_count has default values */
+/* Default action - kernel_done_address, kernel_done_count have default values */
 typedef struct {
     uint8_t packed_lcu_id;
     uint8_t network_index;

diff --git a/common/include/control_protocol.h b/common/include/control_protocol.h
@@ -81,6 +81,7 @@ extern "C" {
 /* Value to represent an operation should be performed on all streams. */
 #define CONTROL_PROTOCOL__ALL_DATAFLOW_MANAGERS (0xFF)
 
+#define CONTROL_PROTOCOL__MAX_CONTEXT_SIZE (3072)
 
 #define CONTROL_PROTOCOL__OPCODES_VARIABLES \
     CONTROL_PROTOCOL__OPCODE_X(HAILO_CONTROL_OPCODE_IDENTIFY,                                  true,  CPU_ID_APP_CPU)\
@@ -868,15 +869,18 @@ typedef struct {
 
 typedef struct {
     bool preliminary_run_asap;
+    bool batch_register_config;
+    bool can_fast_batch_switch;
 } CONTROL_PROTOCOL__INFER_FEATURE_LIST_t;
 
 typedef struct {
-    uint8_t dynamic_contexts_count;
+    uint16_t dynamic_contexts_count;
     CONTROL_PROTOCOL__INFER_FEATURE_LIST_t infer_features;
     CONTROL_PROTOCOL__VALIDATION_FEATURE_LIST_t validation_features;
     uint8_t networks_count;
     uint16_t csm_buffer_size;
     uint16_t batch_size[CONTROL_PROTOCOL__MAX_NETWORKS_PER_NETWORK_GROUP];
+    uint32_t external_action_list_address;
     uint32_t boundary_channels_bitmap[CONTROL_PROTOCOL__MAX_VDMA_ENGINES_COUNT];
 } CONTROL_PROTOCOL__application_header_t;
 
@@ -954,10 +958,10 @@ typedef struct {
 #pragma warning(disable: 4200)
 #endif
 typedef struct {
-    uint32_t is_first_control_per_context_length;
-    uint8_t is_first_control_per_context;
-    uint32_t is_last_control_per_context_length;
-    uint8_t is_last_control_per_context;
+    uint32_t is_first_chunk_per_context_length;
+    uint8_t is_first_chunk_per_context;
+    uint32_t is_last_chunk_per_context_length;
+    uint8_t is_last_chunk_per_context;
     uint32_t context_type_length;
     uint8_t context_type; // CONTROL_PROTOCOL__context_switch_context_type_t
     uint32_t context_network_data_length;
@@ -988,7 +992,7 @@ typedef struct {
     uint32_t context_type_length;
     uint8_t context_type; // CONTROL_PROTOCOL__context_switch_context_type_t
     uint32_t context_index_length;
-    uint8_t context_index;
+    uint16_t context_index;
     uint32_t action_list_offset_length;
     uint16_t action_list_offset;
 } CONTROL_PROTOCOL__download_context_action_list_request_t;
@@ -1160,7 +1164,7 @@ typedef struct {
     bool break_at_any_batch_index;
     uint16_t batch_index;
     bool break_at_any_context_index;
-    uint8_t context_index;
+    uint16_t context_index;
     bool break_at_any_action_index;
     uint16_t action_index;
 } CONTROL_PROTOCOL__context_switch_breakpoint_data_t;
@@ -1470,15 +1474,21 @@ typedef enum {
     CONTROL_PROTOCOL__CONTEXT_SWITCH_INDEX_COUNT,
 } CONTROL_PROTOCOL__context_switch_context_index_t;
 
-#define CONTROL_PROTOCOL__MAX_CONTEXTS_PER_NETWORK_GROUP (64)
+#define CONTROL_PROTOCOL__MAX_CONTEXTS_PER_NETWORK_GROUP (1024)
 
+// This struct will be used for both ControlActionList and DDRActionlist (in order to keep flow in FW as similar as possible)
+// The context_network_data array will never have more data than CONTROL_PROTOCOL__CONTEXT_NETWORK_DATA_SINGLE_CONTROL_MAX_SIZE
+// In case of ControlActionList - this is verified when sending and receiving control. We make it larger here to be
+// able to hold DDRActionList Contexts without needing to copy or do more processing in fw.
+// In both cases this struct holds a chunk of the context - in ControlActionList - it will be as much of the context a
+// Single control message is able to carry and in DDRActionlist will be the whole context
 typedef struct {
-    bool is_first_control_per_context;
-    bool is_last_control_per_context;
+    bool is_first_chunk_per_context;
+    bool is_last_chunk_per_context;
     uint8_t context_type; // CONTROL_PROTOCOL__context_switch_context_type_t
     uint32_t context_network_data_length;
-    uint8_t context_network_data[CONTROL_PROTOCOL__CONTEXT_NETWORK_DATA_SINGLE_CONTROL_MAX_SIZE];
-} CONTROL_PROTOCOL__context_switch_context_info_single_control_t;
+    uint8_t context_network_data[CONTROL_PROTOCOL__MAX_CONTEXT_SIZE];
+} CONTROL_PROTOCOL__context_switch_context_info_chunk_t;
 
 CASSERT(sizeof(CONTROL_PROTOCOL__context_switch_context_index_t)<=UINT8_MAX, control_protocol_h);
 CASSERT(sizeof(CONTROL_PROTOCOL__context_switch_context_type_t)<=UINT8_MAX, control_protocol_h);

diff --git a/common/include/d2h_events.h b/common/include/d2h_events.h
@@ -123,14 +123,14 @@ typedef struct {
 
 #define D2H_EVENT_HEALTH_MONITOR_CPU_ECC_EVENT_PARAMETER_COUNT  (1)
 
-/* D2H_EVENT_context_switch_breakpoint_reached_event_massage_t should be the same as 
+/* D2H_EVENT_context_switch_breakpoint_reached_event_message_t should be the same as 
  * CONTROL_PROTOCOL__context_switch_breakpoint_data_t and hailo_context_switch_breakpoint_reached_notification_message_t */
 typedef struct {
     uint8_t application_index;
     uint16_t batch_index;
-    uint8_t context_index;
+    uint16_t context_index;
     uint16_t action_index;
-} D2H_EVENT_context_switch_breakpoint_reached_event_massage_t;
+} D2H_EVENT_context_switch_breakpoint_reached_event_message_t;
 
 #define D2H_EVENT_CONTEXT_SWITCH_BREAKPOINT_REACHED_EVENT_PARAMETER_COUNT  (4)
 
@@ -151,7 +151,7 @@ typedef struct {
     uint32_t exit_status;
     uint8_t application_index;
     uint16_t batch_index;
-    uint8_t context_index;
+    uint16_t context_index;
     uint16_t action_index;
 } D2H_EVENT_context_switch_run_time_error_event_message_t;
 
@@ -166,7 +166,7 @@ typedef union {
    D2H_EVENT_health_monitor_overcurrent_alert_event_message_t health_monitor_overcurrent_alert_event;
    D2H_EVENT_health_monitor_lcu_ecc_error_event_message_t health_monitor_lcu_ecc_error_event;
    D2H_EVENT_health_monitor_cpu_ecc_event_message_t health_monitor_cpu_ecc_event;
-   D2H_EVENT_context_switch_breakpoint_reached_event_massage_t context_switch_breakpoint_reached_event;
+   D2H_EVENT_context_switch_breakpoint_reached_event_message_t context_switch_breakpoint_reached_event;
    D2H_EVENT_health_monitor_clock_changed_event_message_t health_monitor_clock_changed_event;
    D2H_EVENT_hw_infer_mamager_infer_done_message_t hw_infer_manager_infer_done_event;
    D2H_EVENT_context_switch_run_time_error_event_message_t context_switch_run_time_error_event;

diff --git a/common/include/firmware_status.h b/common/include/firmware_status.h
@@ -527,6 +527,7 @@ Updating rules:
    FIRMWARE_STATUS__X(QSPI_STATUS_MISALIGNED_ADDRESS)\
    FIRMWARE_STATUS__X(QSPI_STATUS_BLOCK_ERASE_FAILED)\
    FIRMWARE_STATUS__X(QSPI_STATUS_CLEAR_AHB_REMAP_FAILED)\
+   FIRMWARE_STATUS__X(QSPI_STATUS_NOT_SUPPORTED)\
    \
    FIRMWARE_MODULE__X(FIRMWARE_MODULE__PCIE_SERVICE)\
    FIRMWARE_STATUS__X(PCIE_SERVICE_STATUS_INVALID_PARAMETERS)\
@@ -763,6 +764,7 @@ Updating rules:
    FIRMWARE_STATUS__X(CONTEXT_SWITCH_STATUS_WRITE_DATA_BY_TYPE_ACTION_INVALID_TYPE)\
    FIRMWARE_STATUS__X(CONTEXT_SWITCH_STATUS_WRITE_DATA_BY_TYPE_ACTION_INVALID_MEMORY_SPACE)\
    FIRMWARE_STATUS__X(CONTEXT_SWITCH_STATUS_REACHED_TIMEOUT_WHILE_WAITING_FOR_BATCH_SWITCH_CONTEXT_TO_END)\
+   FIRMWARE_STATUS__X(CONTEXT_SWITCH_STATUS_INVALID_EXTERNAL_ACTION_LIST_ADDRESS)\
    \
    FIRMWARE_MODULE__X(FIRMWARE_MODULE__D2H_EVENT_MANAGER)\
    FIRMWARE_STATUS__X(HAILO_D2H_EVENT_MANAGER_STATUS_MESSAGE_HIGH_PRIORITY_QUEUE_CREATE_FAILED)\
@@ -1080,6 +1082,7 @@ Updating rules:
    FIRMWARE_STATUS__X(BURST_CREDITS_TASK_STATUS_FAILED_TO_FIND_STREAM_INDEX)\
    FIRMWARE_STATUS__X(BURST_CREDITS_TASK_STATUS_TASK_NO_CONFIGURED_ACTIONS)\
    FIRMWARE_STATUS__X(BURST_CREDITS_TASK_STATUS_TASK_EXPECTED_HIGHER_BATCH)\
+   FIRMWARE_STATUS__X(BURST_CREDITS_TASK_STATUS_TASK_REACHED_TIMEOUT_WAITING_FOR_DEACTIVATION)\
    \
    FIRMWARE_MODULE__X(FIRMWARE_MODULE__TASK_SYNC_EVENTS)\
    FIRMWARE_STATUS__X(TASK_SYNC_EVENTS_STATUS_START_TASK_WHILE_IT_IS_RUNNING)\
@@ -1111,13 +1114,19 @@ Updating rules:
    FIRMWARE_STATUS__X(CLUSTER_MANAGER_STATUS_RECEIVED_UNEXPECTED_INTERRUPT)\
    FIRMWARE_STATUS__X(CLUSTER_MANAGER_STATUS_INVALID_NETWORK_INDEX)\
    FIRMWARE_STATUS__X(CLUSTER_MANAGER_STATUS_INVALID_KERNEL_DONE_COUNT)\
+   FIRMWARE_STATUS__X(CLUSTER_MANAGER_STATUS_INVALID_EXTENSION)\
    \
    FIRMWARE_MODULE__X(FIRMWARE_MODULE__HW_INFER_MANAGER)\
    FIRMWARE_STATUS__X(HW_INFER_MANAGER_STATUS_NETWORK_GROUP_NOT_CONFIGURED_BEFORE_INFER_START)\
    FIRMWARE_STATUS__X(HW_INFER_MANAGER_STATUS_NETWORK_GROUP_ALREADY_ACTIVATED)\
    FIRMWARE_STATUS__X(HW_INFER_MANAGER_STATUS_STATE_MACHINE_NOT_IN_RESET_STATE_BEFORE_DEACTIVATE)\
    FIRMWARE_STATUS__X(HW_INFER_MANAGER_STATUS_INVALID_STATE)\
-
+   \
+   FIRMWARE_MODULE__X(FIRMWARE_MODULE__INFINITE_CONTEXT_LOADER)\
+   FIRMWARE_STATUS__X(INFINITE_CONTEXT_LOADER_STATUS_EVENT_BITS_NOT_CLEARED_BEFORE_COPY_CALL)\
+   FIRMWARE_STATUS__X(INFINITE_CONTEXT_LOADER_STATUS_TIMEOUT_OCCURED_WAITING_FOR_COPY)\
+   FIRMWARE_STATUS__X(INFINITE_CONTEXT_LOADER_STATUS_NOT_SUPPORTED)\
+   FIRMWARE_STATUS__X(INFINITE_CONTEXT_LOADER_STATUS_NOT_MODULE_NOT_INITIALIZED)\
 
 typedef enum {
 #define FIRMWARE_MODULE__X(module) module,

diff --git a/common/include/utils.h b/common/include/utils.h
@@ -54,6 +54,10 @@
 #define DIV_ROUND_UP(n,d) (((n) + (d) - 1) / (d))
 #endif
 
+#ifndef DIV_ROUND_DOWN
+#define DIV_ROUND_DOWN(n,d) ((n) / (d))
+#endif
+
 #ifndef ROUND_UNSIGNED_FLOAT
 #define ROUND_UNSIGNED_FLOAT(n) ((n - (uint32_t)(n)) > 0.5) ? (uint32_t)(n + 1) : (uint32_t)(n)
 #endif

diff --git a/hailort/.gitignore b/hailort/.gitignore
@@ -1,3 +1,4 @@
+build/
+dist/
 /external/
 cmake/external/*/
-prepare_externals/build/
diff --git a/hailort/CMakeLists.txt b/hailort/CMakeLists.txt
@@ -3,6 +3,7 @@ cmake_minimum_required(VERSION 3.0.0)
 option(HAILO_BUILD_PYBIND "Build Python binding" OFF)
 option(HAILO_BUILD_EMULATOR "Build hailort for emulator" OFF)
 option(HAILO_BUILD_UT "Build Unit Tests" OFF)
+option(HAILO_BUILD_DMABUF_TESTS "Build DMA buffer tests. Relevant only if HAILO_BUILD_UT is ON" OFF)
 option(HAILO_BUILD_HW_DEBUG_TOOL "Build hw debug tool" OFF)
 option(HAILO_BUILD_GSTREAMER "Compile gstreamer plugins" OFF)
 option(HAILO_BUILD_EXAMPLES "Build examples" OFF)
@@ -30,8 +31,8 @@ endif()
 
 # Set firmware version
 add_definitions( -DFIRMWARE_VERSION_MAJOR=4 )
-add_definitions( -DFIRMWARE_VERSION_MINOR=16 )
-add_definitions( -DFIRMWARE_VERSION_REVISION=2 )
+add_definitions( -DFIRMWARE_VERSION_MINOR=17 )
+add_definitions( -DFIRMWARE_VERSION_REVISION=0 )
 if(HAILO_BUILD_SERVICE)
     add_definitions( -DHAILO_SUPPORT_MULTI_PROCESS )
 endif()

diff --git a/hailort/LICENSE-3RD-PARTY.md b/hailort/LICENSE-3RD-PARTY.md
@@ -1,16 +1,17 @@
-| Package                          | Copyright (c)                     | License            | Version        | Notes                                         | References                                                                    |
-|:---------------------------------|:----------------------------------|:-------------------|:---------------|:----------------------------------------------|:------------------------------------------------------------------------------|
-| CLI11                            | University of Cincinnati          | 3-Clause BSD       | 2.2.0          | Fork                                          | https://github.com/hailo-ai/CLI11                                             |
-| Catch2                           | Catch2 Authors                    | BSL-1.0            | 2.13.7         | Cloned entire package                         | https://github.com/catchorg/Catch2                                            |
-| protobuf                         | Google Inc.                       | BSD                | 21.12          | Cloned entire package                         | https://github.com/protocolbuffers/protobuf                                   |
-| pybind11                         | Wenzel Jakob                      | BSD                | 2.10.1         | Cloned entire package                         | https://github.com/pybind/pybind11                                            |
-| spdlog                           | Gabi Melman                       | MIT                | 1.6.1          | Cloned entire package                         | https://github.com/gabime/spdlog                                              |
-| folly                            | Facebook, Inc. and its affiliates | Apache License 2.0 | v2020.08.17.00 | Copied only the file `folly/TokenBucket.h`    | https://github.com/facebook/folly                                             |
-| nlohmann_json_cmake_fetchcontent | ArthurSonzogni                    | MIT License        | v3.9.1         | Cloned entire package                         | https://github.com/ArthurSonzogni/nlohmann_json_cmake_fetchcontent            |
-| readerwriterqueue                | Cameron Desrochers                | Simplified BSD     | 1.0.3          | Cloned entire package                         | https://github.com/cameron314/readerwriterqueue                               |
-| DotWriter                        | John Vilk                         | MIT License        | master         | Fork                                          | https://github.com/hailo-ai/DotWriter                                         |
-| benchmark                        | Google Inc.                       | Apache License 2.0 | 1.6.0          | Cloned entire package                         | https://github.com/google/benchmark.git                                       |
-| md5                              | Alexander Peslyak                 | cut-down BSD       | -              | Copied code from website                      | http://openwall.info/wiki/people/solar/software/public-domain-source-code/md5 |
-| pevents                          | Mahmoud Al-Qudsi                  | MIT License        | master         | Cloned entire package                         | https://github.com/neosmart/pevents.git                                       |
-| grpc                             | Google Inc.                       | Apache License 2.0 | 1.46.3         | Cloned entire package                         | https://github.com/grpc/grpc                                                  |
-| stb                              | Sean Barrett                      | MIT License        | 0.97           | Copied only the file `stb/stb_image_resize.h` | https://github.com/nothings/stb                                               |
+| Package                          | Copyright (c)                     | License                    | Version        | Notes                                         | References                                                                    |
+|:---------------------------------|:----------------------------------|:---------------------------|:---------------|:----------------------------------------------|:------------------------------------------------------------------------------|
+| CLI11                            | University of Cincinnati          | 3-Clause BSD               | 2.2.0          | Fork                                          | https://github.com/hailo-ai/CLI11                                             |
+| Catch2                           | Catch2 Authors                    | BSL-1.0                    | 2.13.7         | Cloned entire package                         | https://github.com/catchorg/Catch2                                            |
+| protobuf                         | Google Inc.                       | BSD                        | 21.12          | Cloned entire package                         | https://github.com/protocolbuffers/protobuf                                   |
+| pybind11                         | Wenzel Jakob                      | BSD                        | 2.10.1         | Cloned entire package                         | https://github.com/pybind/pybind11                                            |
+| spdlog                           | Gabi Melman                       | MIT                        | 1.6.1          | Cloned entire package                         | https://github.com/gabime/spdlog                                              |
+| folly                            | Facebook, Inc. and its affiliates | Apache License 2.0         | v2020.08.17.00 | Copied only the file `folly/TokenBucket.h`    | https://github.com/facebook/folly                                             |
+| nlohmann_json_cmake_fetchcontent | ArthurSonzogni                    | MIT License                | v3.9.1         | Cloned entire package                         | https://github.com/ArthurSonzogni/nlohmann_json_cmake_fetchcontent            |
+| readerwriterqueue                | Cameron Desrochers                | Simplified BSD             | 1.0.3          | Cloned entire package                         | https://github.com/cameron314/readerwriterqueue                               |
+| DotWriter                        | John Vilk                         | MIT License                | master         | Fork                                          | https://github.com/hailo-ai/DotWriter                                         |
+| benchmark                        | Google Inc.                       | Apache License 2.0         | 1.6.0          | Cloned entire package                         | https://github.com/google/benchmark.git                                       |
+| md5                              | Alexander Peslyak                 | cut-down BSD               | -              | Copied code from website                      | http://openwall.info/wiki/people/solar/software/public-domain-source-code/md5 |
+| pevents                          | Mahmoud Al-Qudsi                  | MIT License                | master         | Cloned entire package                         | https://github.com/neosmart/pevents.git                                       |
+| grpc                             | Google Inc.                       | Apache License 2.0         | 1.46.3         | Cloned entire package                         | https://github.com/grpc/grpc                                                  |
+| stb                              | Sean Barrett                      | MIT License                | 0.97           | Copied only the file `stb/stb_image_resize.h` | https://github.com/nothings/stb                                               |
+| eigen                            |                                   | Mozilla Public License 2.0 | 3.4.0          | Cloned entire package                         | https://gitlab.com/libeigen/eigen                                             |
diff --git a/hailort/cmake/external/eigen.cmake b/hailort/cmake/external/eigen.cmake
@@ -0,0 +1,28 @@
+cmake_minimum_required(VERSION 3.11.0)
+
+include(FetchContent)
+
+FetchContent_Declare(
+    eigen
+    GIT_REPOSITORY https://gitlab.com/libeigen/eigen
+    GIT_TAG 3147391d946bb4b6c68edd901f2add6ac1f31f8c # Version 3.4.0
+    GIT_SHALLOW TRUE
+    SOURCE_DIR ${HAILO_EXTERNAL_DIR}/eigen-src
+    SUBBUILD_DIR ${HAILO_EXTERNAL_DIR}/eigen-subbuild
+)
+
+
+# https://stackoverflow.com/questions/65527126/disable-install-for-fetchcontent
+FetchContent_GetProperties(eigen)
+if(NOT eigen_POPULATED)
+    FetchContent_Populate(eigen)
+    option(EIGEN_BUILD_DOC OFF)
+    option(BUILD_TESTING OFF)
+    option(EIGEN_LEAVE_TEST_IN_ALL_TARGET OFF)
+    option(EIGEN_BUILD_PKGCONFIG OFF)
+    option(CMAKE_Fortran_COMPILER OFF)
+
+    if (NOT HAILO_EXTERNALS_EXCLUDE_TARGETS)
+        add_subdirectory(${eigen_SOURCE_DIR} ${eigen_BINARY_DIR} EXCLUDE_FROM_ALL)
+    endif()
+endif()