upsize EDM fabric channel buffer slots to be able to fit 4 bfp8 tiles…

… per packet (#18000)

The current default EDM buffer slot size is 4096 which can only store 3
bfp8 tiles. There is enough space in erisc L1 unreserved space such that
all channels can have a power of 2 buffer slot count and also have a
slot size of 4 bfp8 tiles. There is inefficient space for 5 bfp8 tiles
per slot.

This commit bumps up the buffer slot size to fit 4 bfp8 tiles per packet,
which is preferable for workloads with bfp8 tiles sent over fabric.

tests/ttnn/unit_tests/gtests/ccl/test_fabric_erisc_data_mover_loopback_with_workers.cpp

@@ -427,7 +427,8 @@ bool RunLoopbackTest(
     // EDM Builder Setup
     ////////////////////////////////////////////////////////////////////////////
 
-    static constexpr std::size_t edm_buffer_size = 4096 + PACKET_HEADER_SIZE_BYTES;
+    static constexpr std::size_t edm_buffer_size =
+        ttnn::ccl::FabricEriscDatamoverBuilder::default_packet_payload_size_bytes + PACKET_HEADER_SIZE_BYTES;
 
     auto chip0_worker_fabric_connection = chip_0_edm_builder.build_connection_to_worker_channel();
     ////////////////////////////////////////////////////////////////////////////
@@ -910,7 +911,8 @@ bool RunLineFabricTest(
     std::size_t page_plus_header_size = page_size + sizeof(tt::fabric::PacketHeader);
     std::size_t tensor_size_bytes = num_pages_total * page_size;
 
-    static constexpr std::size_t edm_buffer_size = 4096 + PACKET_HEADER_SIZE_BYTES;
+    static constexpr std::size_t edm_buffer_size =
+        ttnn::ccl::FabricEriscDatamoverBuilder::default_packet_payload_size_bytes + PACKET_HEADER_SIZE_BYTES;
     const size_t local_chip_id = 0;
     const size_t remote_chip_id = 1;
     auto program_ptrs = std::vector<Program*>(devices.size());
@@ -1237,7 +1239,8 @@ int TestLoopbackEntrypoint(
     IDevice* sender_device = device_0;
     IDevice* receiver_device = device_1;
 
-    static constexpr std::size_t edm_buffer_size = 4096 + PACKET_HEADER_SIZE_BYTES;
+    static constexpr std::size_t edm_buffer_size =
+        ttnn::ccl::FabricEriscDatamoverBuilder::default_packet_payload_size_bytes + PACKET_HEADER_SIZE_BYTES;
     const chip_id_t local_chip_id = 0;
     const chip_id_t remote_chip_id = 1;
     auto const& edm_config = ttnn::ccl::FabricEriscDatamoverConfig(edm_buffer_size, 1, 2);
@@ -2988,7 +2991,8 @@ void RunWriteThroughputStabilityTestWithPersistentFabric(
     static constexpr uint32_t source_payload_cb_index = tt::CB::c_in1;
     static constexpr size_t packet_header_cb_size_in_headers = 4;
     static constexpr bool enable_persistent_fabric_mode = true;
-    static constexpr size_t packet_payload_size_bytes = 4096;
+    static constexpr size_t packet_payload_size_bytes =
+        ttnn::ccl::FabricEriscDatamoverBuilder::default_packet_payload_size_bytes;
     static constexpr size_t dest_buffer_size = packet_payload_size_bytes * 4;
     static constexpr tt::DataFormat cb_df = tt::DataFormat::Bfp8;
 

ttnn/cpp/ttnn/operations/ccl/erisc_datamover_builder.hpp

@@ -183,7 +183,7 @@ class FabricEriscDatamoverBuilder {
    public:
        static constexpr size_t default_firmware_context_switch_interval = 200000;
        // payload only, no header
-       static constexpr size_t default_packet_payload_size_bytes = 4096;
+       static constexpr size_t default_packet_payload_size_bytes = tt::tile_size(tt::DataFormat::Bfp8_b) * 4;
 
        FabricEriscDatamoverBuilder(
            const CoreCoord& my_eth_core_logical,