#0: Revert recent fabric changes because it breaks TG unit tests (#18702

)

### Ticket

TG unit tests broken. Revert everything relevant

### Problem description
Provide context for the problem.

### What's changed
Describe the approach used to solve the problem.
Summarize the changes made and its impact.

### Checklist
- [ ] [All post
commit](https://github.com/tenstorrent/tt-metal/actions/workflows/all-post-commit-workflows.yaml)
CI passes
- [ ] [Blackhole Post
commit](https://github.com/tenstorrent/tt-metal/actions/workflows/blackhole-post-commit.yaml)
CI passes (if applicable)
- [ ] [Model
regression](https://github.com/tenstorrent/tt-metal/actions/workflows/perf-models.yaml)
CI passes (if applicable)
- [ ] [Device performance
regression](https://github.com/tenstorrent/tt-metal/actions/workflows/perf-device-models.yaml)
CI passes (if applicable)
- [ ] **(For models and ops writers)** Full [new models
tests](https://github.com/tenstorrent/tt-metal/actions/workflows/full-new-models-suite.yaml)
CI passes (if applicable)
- [ ] New/Existing tests provide coverage for changes

---------

Co-authored-by: Raymond Kim <raymond.kim243@outlook.com>

tests/tt_metal/tt_metal/perf_microbenchmark/dispatch/test_dispatcher.cpp

@@ -565,12 +565,6 @@ int main(int argc, char** argv) {
             host_completion_queue_wr_ptr,
             dev_completion_queue_wr_ptr,
             dev_completion_queue_rd_ptr,
-            0,
-            0,
-            0,
-            0,
-            0,
-            0,
             true,  // is_dram_variant
             true,  // is_host_variant
         };

tests/tt_metal/tt_metal/perf_microbenchmark/dispatch/test_prefetcher.cpp

@@ -66,10 +66,6 @@ constexpr CoreType DISPATCH_CORE_TYPE = CoreType::WORKER;
 //////////////////////////////////////////////////////////////////////////////////////////
 using std::vector;
 using namespace tt;
-using tt::packet_queue::dispatch_packet_header_t;
-using tt::packet_queue::DispatchRemoteNetworkType;
-using tt::packet_queue::packet_switch_4B_pack;
-using tt::packet_queue::packet_switch_dest_pack;
 
 uint32_t iterations_g = DEFAULT_ITERATIONS;
 
@@ -1843,11 +1839,11 @@ void configure_for_single_chip(
     const uint32_t dispatch_h_cb_sem = dispatch_h_core_sem_0_id;
 
     std::vector<uint32_t> prefetch_compile_args = {
-        dispatch_buffer_base,                             // overridden below for prefetch_h
+        dispatch_buffer_base,                              // overridden below for prefetch_h
         DispatchSettings::DISPATCH_BUFFER_LOG_PAGE_SIZE,  // overridden below for prefetch_h
-        dispatch_buffer_pages,                            // overridden below for prefetch_h
-        prefetch_downstream_cb_sem,                       // overridden below for prefetch_d
-        dispatch_cb_sem,                                  // overridden below for prefetch_h
+        dispatch_buffer_pages,                             // overridden below for prefetch_h
+        prefetch_downstream_cb_sem,                        // overridden below for prefetch_d
+        dispatch_cb_sem,                                   // overridden below for prefetch_h
         dev_hugepage_base_g,
         hugepage_issue_buffer_size_g,
         prefetch_q_base,
@@ -1864,17 +1860,11 @@ void configure_for_single_chip(
         prefetch_downstream_cb_sem,  // prefetch_d only
         DispatchSettings::PREFETCH_D_BUFFER_LOG_PAGE_SIZE,
         DispatchSettings::PREFETCH_D_BUFFER_BLOCKS,  // prefetch_d only
-        0,                                           // unused: for prefetch_hd <--> dispatch_hd
-        0,                                           // unused: for prefetch_hd <--> dispatch_hd
-        0,                                           // unused: for prefetch_hd <--> dispatch_hd
-        0,                                           // unused: for prefetch_hd <--> dispatch_hd
-        0,                                           // unused: for prefetch_hd <--> dispatch_hd
-        0,
-        0,
-        0,
-        0,
-        0,
-        0,
+        0,                                            // unused: for prefetch_hd <--> dispatch_hd
+        0,                                            // unused: for prefetch_hd <--> dispatch_hd
+        0,                                            // unused: for prefetch_hd <--> dispatch_hd
+        0,                                            // unused: for prefetch_hd <--> dispatch_hd
+        0,                                            // unused: for prefetch_hd <--> dispatch_hd
     };
 
     constexpr NOC my_noc_index = NOC::NOC_0;
@@ -2156,14 +2146,7 @@ void configure_for_single_chip(
         0,
         host_completion_queue_wr_ptr,
         dev_completion_queue_wr_ptr,
-        dev_completion_queue_rd_ptr,
-        0,
-        0,
-        0,
-        0,
-        0,
-        0,
-    };
+        dev_completion_queue_rd_ptr};
 
     CoreCoord phys_upstream_from_dispatch_core = split_prefetcher_g ? phys_prefetch_d_core : phys_prefetch_core_g;
     if (split_dispatcher_g) {
@@ -3522,10 +3505,6 @@ int main(int argc, char** argv) {
         }
 
         if (packetized_path_en_g) {
-            using tt::packet_queue::PACKET_QUEUE_TEST_PASS;
-            using tt::packet_queue::packet_queue_test_status_to_string;
-            using tt::packet_queue::PQ_TEST_STATUS_INDEX;
-
             vector<uint32_t> prefetch_relay_mux_results = tt::llrt::read_hex_vec_from_core(
                 device_r->id(),
                 phys_prefetch_relay_mux_core,

tests/tt_metal/tt_metal/perf_microbenchmark/routing/kernels/traffic_gen.hpp

@@ -59,7 +59,7 @@ struct input_queue_raw_state_t {
         uint32_t max_packet_size_words,
         uint64_t total_data_words) {
         this->curr_packet_dest = this->num_dests_sent_last_packet + dest_endpoint_start_id;
-        this->curr_packet_flags = tt::packet_queue::DispatchPacketFlag::PACKET_TEST_LAST;
+        this->curr_packet_flags = DispatchPacketFlag::PACKET_TEST_LAST;
         this->curr_packet_size_words = 2;
         this->curr_packet_words_remaining = this->curr_packet_size_words;
         this->data_words_input += 2;
@@ -174,10 +174,10 @@ constexpr auto select_input_queue() {
 }
 
 inline void fill_packet_data(tt_l1_ptr uint32_t* start_addr, uint32_t num_words, uint32_t start_val) {
-    tt_l1_ptr uint32_t* addr = start_addr + (tt::packet_queue::PACKET_WORD_SIZE_BYTES / 4 - 1);
+    tt_l1_ptr uint32_t* addr = start_addr + (PACKET_WORD_SIZE_BYTES / 4 - 1);
     for (uint32_t i = 0; i < num_words; i++) {
         *addr = start_val++;
-        addr += (tt::packet_queue::PACKET_WORD_SIZE_BYTES / 4);
+        addr += (PACKET_WORD_SIZE_BYTES / 4);
     }
 }
 
@@ -188,7 +188,7 @@ inline bool check_packet_data(
     uint32_t& mismatch_addr,
     uint32_t& mismatch_val,
     uint32_t& expected_val) {
-    tt_l1_ptr uint32_t* addr = start_addr + (tt::packet_queue::PACKET_WORD_SIZE_BYTES / 4 - 1);
+    tt_l1_ptr uint32_t* addr = start_addr + (PACKET_WORD_SIZE_BYTES / 4 - 1);
     for (uint32_t i = 0; i < num_words; i++) {
         if (*addr != start_val) {
             mismatch_addr = reinterpret_cast<uint32_t>(addr);
@@ -197,7 +197,7 @@ inline bool check_packet_data(
             return false;
         }
         start_val++;
-        addr += (tt::packet_queue::PACKET_WORD_SIZE_BYTES / 4);
+        addr += (PACKET_WORD_SIZE_BYTES / 4);
     }
     return true;
 }

tests/tt_metal/tt_metal/perf_microbenchmark/routing/kernels/traffic_gen_rx.cpp

@@ -7,9 +7,6 @@
 #include "tt_metal/impl/dispatch/kernels/packet_queue.hpp"
 #include "tests/tt_metal/tt_metal/perf_microbenchmark/routing/kernels/traffic_gen.hpp"
 
-using tt::packet_queue::PACKET_WORD_SIZE_BYTES;
-using tt::packet_queue::dispatch_packet_header_t;
-using tt::packet_queue::DispatchRemoteNetworkType;
 
 constexpr uint32_t endpoint_id = get_compile_time_arg_val(0);
 
@@ -57,12 +54,11 @@ constexpr uint32_t disable_header_check = get_compile_time_arg_val(18);
 // the same random seed as the corresponding traffic_gen_tx
 input_queue_rnd_state_t src_rnd_state[num_src_endpoints];
 
-tt::packet_queue::packet_input_queue_state_t input_queues[tt::packet_queue::MAX_SWITCH_FAN_IN];
-using input_queue_network_sequence = tt::packet_queue::NetworkTypeSequence<rx_rptr_update_network_type>;
-using input_queue_cb_mode_sequence = tt::packet_queue::CBModeTypeSequence<false>;
+packet_input_queue_state_t input_queues[MAX_SWITCH_FAN_IN];
+using input_queue_network_sequence = NetworkTypeSequence<rx_rptr_update_network_type>;
+using input_queue_cb_mode_sequence = CBModeTypeSequence<false>;
 
 void kernel_main() {
-    using namespace tt::packet_queue;
 
     zero_l1_buf(test_results, test_results_size_bytes);
     test_results[PQ_TEST_STATUS_INDEX] = PACKET_QUEUE_TEST_STARTED;

tests/tt_metal/tt_metal/perf_microbenchmark/routing/kernels/traffic_gen_tx.cpp

@@ -8,10 +8,6 @@
 #include "tt_metal/fabric/hw/inc/tt_fabric_status.h"
 #include "tests/tt_metal/tt_metal/perf_microbenchmark/routing/kernels/traffic_gen.hpp"
 
-using tt::packet_queue::PACKET_WORD_SIZE_BYTES;
-using tt::packet_queue::dispatch_packet_header_t;
-using tt::packet_queue::DispatchRemoteNetworkType;
-
 constexpr uint32_t src_endpoint_id = get_compile_time_arg_val(0);
 constexpr uint32_t num_dest_endpoints = get_compile_time_arg_val(1);
 
@@ -65,16 +61,16 @@ constexpr uint32_t data_sent_per_iter_high = get_compile_time_arg_val(21);
 constexpr uint32_t input_queue_id = 0;
 constexpr uint32_t output_queue_id = 1;
 
-tt::packet_queue::packet_input_queue_state_t input_queue;
-using input_queue_network_sequence = tt::packet_queue::NetworkTypeSequence<DispatchRemoteNetworkType::NONE>;
-using input_queue_cb_mode_sequence = tt::packet_queue::CBModeTypeSequence<false>;
+packet_input_queue_state_t input_queue;
+using input_queue_network_sequence = NetworkTypeSequence<DispatchRemoteNetworkType::NONE>;
+using input_queue_cb_mode_sequence = CBModeTypeSequence<false>;
 
-tt::packet_queue::packet_output_queue_state_t output_queue;
-using output_queue_network_sequence = tt::packet_queue::NetworkTypeSequence<tx_network_type>;
-using output_queue_cb_mode_sequence = tt::packet_queue::CBModeTypeSequence<false>;
+packet_output_queue_state_t output_queue;
+using output_queue_network_sequence = NetworkTypeSequence<tx_network_type>;
+using output_queue_cb_mode_sequence = CBModeTypeSequence<false>;
 
-constexpr tt::packet_queue::packet_input_queue_state_t* input_queue_ptr = &input_queue;
-constexpr tt::packet_queue::packet_output_queue_state_t* output_queue_ptr = &output_queue;
+constexpr packet_input_queue_state_t* input_queue_ptr = &input_queue;
+constexpr packet_output_queue_state_t* output_queue_ptr = &output_queue;
 
 // input_queue_rnd_state_t input_queue_state;
 auto input_queue_state = select_input_queue<pkt_dest_size_choice>();
@@ -135,8 +131,6 @@ inline bool input_queue_handler() {
 }
 
 void kernel_main() {
-    using namespace tt::packet_queue;
-
     zero_l1_buf(test_results, test_results_size_bytes);
     test_results[PQ_TEST_STATUS_INDEX] = PACKET_QUEUE_TEST_STARTED;
     test_results[PQ_TEST_MISC_INDEX] = 0xff000000;

tests/tt_metal/tt_metal/perf_microbenchmark/routing/test_mux_demux.cpp

@@ -12,11 +12,11 @@
 #include "routing_test_common.hpp"
 #include "llrt.hpp"
 
+using std::vector;
+using namespace tt;
+using json = nlohmann::json;
+
 int main(int argc, char **argv) {
-    using std::vector;
-    using namespace tt;
-    using namespace tt::packet_queue;
-    using json = nlohmann::json;
 
     constexpr uint32_t default_tx_x = 0;
     constexpr uint32_t default_tx_y = 0;

tests/tt_metal/tt_metal/perf_microbenchmark/routing/test_mux_demux_2level.cpp

@@ -11,10 +11,10 @@
 #include "routing_test_common.hpp"
 #include "llrt.hpp"
 
+using std::vector;
+using namespace tt;
+
 int main(int argc, char **argv) {
-    using std::vector;
-    using namespace tt;
-    using namespace tt::packet_queue;
 
     constexpr uint32_t default_prng_seed = 0x100;
     constexpr uint32_t default_data_kb_per_tx = 64*1024;

tests/tt_metal/tt_metal/perf_microbenchmark/routing/test_tx_rx.cpp

@@ -12,10 +12,10 @@
 #include "utils.hpp"
 #include "llrt.hpp"
 
+using std::vector;
+using namespace tt;
+
 int main(int argc, char **argv) {
-    using std::vector;
-    using namespace tt;
-    using namespace tt::packet_queue;
 
     bool pass = true;
     try {

tests/tt_metal/tt_metal/perf_microbenchmark/routing/test_vc_bi_tunnel_2ep.cpp

@@ -11,12 +11,11 @@
 #include "test_common.hpp"
 #include "routing_test_common.hpp"
 
-int main(int argc, char** argv) {
-    using std::vector;
-    using namespace tt;
-    using namespace tt::packet_queue;
+using std::vector;
+using namespace tt;
+using json = nlohmann::json;
 
-    using json = nlohmann::json;
+int main(int argc, char** argv) {
     constexpr uint32_t default_tx_x = 0;
     constexpr uint32_t default_tx_y = 0;
     constexpr uint32_t default_rx_x = 0;

tests/tt_metal/tt_metal/perf_microbenchmark/routing/test_vc_bi_tunnel_4ep.cpp

@@ -11,12 +11,11 @@
 #include "test_common.hpp"
 #include "routing_test_common.hpp"
 
-int main(int argc, char** argv) {
-    using std::vector;
-    using namespace tt;
-    using namespace tt::packet_queue;
-    using json = nlohmann::json;
+using std::vector;
+using namespace tt;
+using json = nlohmann::json;
 
+int main(int argc, char** argv) {
     constexpr uint32_t default_tx_x = 0;
     constexpr uint32_t default_tx_y = 0;
     constexpr uint32_t default_rx_x = 0;

tests/tt_metal/tt_metal/perf_microbenchmark/routing/test_vc_loopback_tunnel.cpp

@@ -11,11 +11,11 @@
 #include "test_common.hpp"
 #include "routing_test_common.hpp"
 
+using std::vector;
+using namespace tt;
+using json = nlohmann::json;
+
 int main(int argc, char **argv) {
-    using std::vector;
-    using namespace tt;
-    using namespace tt::packet_queue;
-    using json = nlohmann::json;
 
     constexpr uint32_t default_tx_x = 0;
     constexpr uint32_t default_tx_y = 0;

tests/tt_metal/tt_metal/perf_microbenchmark/routing/test_vc_mux_demux.cpp

@@ -11,11 +11,11 @@
 #include "test_common.hpp"
 #include "routing_test_common.hpp"
 
+using std::vector;
+using namespace tt;
+using json = nlohmann::json;
+
 int main(int argc, char **argv) {
-    using std::vector;
-    using namespace tt;
-    using namespace tt::packet_queue;
-    using json = nlohmann::json;
 
     constexpr uint32_t default_tx_x = 0;
     constexpr uint32_t default_tx_y = 0;

tests/tt_metal/tt_metal/perf_microbenchmark/routing/test_vc_uni_tunnel.cpp

@@ -11,11 +11,12 @@
 #include "test_common.hpp"
 #include "routing_test_common.hpp"
 
+using std::vector;
+using namespace tt;
+using json = nlohmann::json;
+
+
 int main(int argc, char **argv) {
-    using std::vector;
-    using namespace tt;
-    using namespace tt::packet_queue;
-    using json = nlohmann::json;
 
     constexpr uint32_t default_tx_x = 0;
     constexpr uint32_t default_tx_y = 0;

tt_metal/CMakeLists.txt

@@ -28,7 +28,6 @@ target_sources(
             api/tt-metalium/cq_commands.hpp
             api/tt-metalium/dev_msgs.h
             api/tt-metalium/tt_log.h
-            api/tt-metalium/fabric_host_interface.h
             core_descriptors/blackhole_140_arch.yaml
             core_descriptors/grayskull_120_arch.yaml
             core_descriptors/wormhole_b0_80_arch.yaml

tt_metal/api/tt-metalium/command_queue_interface.hpp

@@ -7,7 +7,6 @@
 #include <magic_enum/magic_enum.hpp>
 #include <mutex>
 #include <tt-metalium/tt_align.hpp>
-#include <tt-metalium/fabric_host_interface.h>
 #include <unordered_map>
 
 #include "cq_commands.hpp"
@@ -18,7 +17,6 @@
 #include "dispatch_settings.hpp"
 #include "helpers.hpp"
 #include "buffer.hpp"
-#include "rtoptions.hpp"
 #include "umd/device/tt_core_coordinates.h"
 
 namespace tt::tt_metal {
@@ -35,9 +33,7 @@ enum class CommandQueueDeviceAddrType : uint8_t {
     COMPLETION_Q1_LAST_EVENT = 5,
     DISPATCH_S_SYNC_SEM = 6,
     DISPATCH_MESSAGE = 7,
-    // Enable for FD on Fabric
-    FABRIC_INTERFACE = 8,
-    UNRESERVED = 9,
+    UNRESERVED = 8
 };
 
 enum class CommandQueueHostAddrType : uint8_t {
@@ -175,12 +171,6 @@ class DispatchMemMap {
                 device_cq_addr_sizes_[dev_addr_idx] = settings.dispatch_s_sync_sem_;
             } else if (dev_addr_type == CommandQueueDeviceAddrType::DISPATCH_MESSAGE) {
                 device_cq_addr_sizes_[dev_addr_idx] = settings.dispatch_message_;
-            } else if (dev_addr_type == CommandQueueDeviceAddrType::FABRIC_INTERFACE) {
-                if (llrt::RunTimeOptions::get_instance().get_fd_fabric()) {
-                    device_cq_addr_sizes_[dev_addr_idx] = tt_fabric::PACKET_HEADER_SIZE_BYTES;
-                } else {
-                    device_cq_addr_sizes_[dev_addr_idx] = 0;
-                }
             } else {
                 device_cq_addr_sizes_[dev_addr_idx] = settings.other_ptrs_size;
             }
@@ -194,8 +184,6 @@ class DispatchMemMap {
             CommandQueueDeviceAddrType dev_addr_type = magic_enum::enum_value<CommandQueueDeviceAddrType>(dev_addr_idx);
             if (dev_addr_type == CommandQueueDeviceAddrType::UNRESERVED) {
                 device_cq_addrs_[dev_addr_idx] = align(device_cq_addrs_[dev_addr_idx], pcie_alignment);
-            } else if (dev_addr_type == CommandQueueDeviceAddrType::FABRIC_INTERFACE) {
-                device_cq_addrs_[dev_addr_idx] = align(device_cq_addrs_[dev_addr_idx], l1_alignment);
             }
         }
 

tt_metal/api/tt-metalium/control_plane.hpp

@@ -43,7 +43,7 @@ class ControlPlane {
            chip_id_t dst_chip_id,
            chan_id_t src_chan_id) const;
 
-       // Return routers to get to the destination chip, avoid local eth to eth routing. CoreCoord is a virtual coord.
+       // Return routers to get to the destination chip, avoid local eth to eth routing
        std::vector<std::pair<routing_plane_id_t, CoreCoord>> get_routers_to_chip(
            mesh_id_t src_mesh_id, chip_id_t src_chip_id, mesh_id_t dst_mesh_id, chip_id_t dst_chip_id) const;