clean up some comments and log messages

tenstorrent · Dec 6, 2024 · c03bac4 · c03bac4
1 parent 1003a47
commit c03bac4
Show file tree

Hide file tree

Showing 2 changed files with 20 additions and 23 deletions.
diff --git a/device/cluster.cpp b/device/cluster.cpp
@@ -268,18 +268,24 @@ void Cluster::create_device(
                 !(arch_name == tt::ARCH::BLACKHOLE && num_host_mem_channels > 1),
                 "More channels are not yet supported for Blackhole");
             // Same number of host channels per device for now
-            bool hugepages_initialized = m_pci_device_map.at(logical_device_id)->init_hugepage(num_host_mem_channels);
-            // Large writes to remote chips require hugepages to be initialized.
-            // Conservative assert - end workload if remote chips present but hugepages not initialized (failures caused
-            // if using remote only for small transactions)
-            if (target_remote_chips.size()) {
+            bool hugepages_initialized = dev->init_hugepage(num_host_mem_channels);
+
+            if (!hugepages_initialized) {
+                log_warning(
+                    LogSiliconDriver,
+                    "Hugepages not initialized for device {} (logical_device_id: {} pci_interface_id: {})",
+                    dev->get_device_num(),
+                    logical_device_id,
+                    pci_interface_id);
+            }
+
+            // Large writes to remote chips require at least one hugepage.
+            bool no_hugepages = (dev->get_hugepage_mapping(0).mapping == nullptr);
+            if (target_remote_chips.size() && no_hugepages) {
                 log_assert(
                     hugepages_initialized,
                     "Hugepages must be successfully initialized if workload contains remote chips!");
             }
-            if (not m_pci_device_map.at(logical_device_id)->get_hugepage_mapping(0).mapping) {
-                log_warning(LogSiliconDriver, "No hugepage mapping at device {}.", logical_device_id);
-            }
         }
         // translation layer for harvested coords. Default is identity map
         harvested_coord_translation.insert({logical_device_id, create_harvested_coord_translation(arch_name, true)});

diff --git a/device/pcie/pci_device.cpp b/device/pcie/pci_device.cpp
@@ -723,10 +723,7 @@ bool PCIDevice::init_hugepage(uint32_t num_host_mem_channels) {
     auto physical_device_id = get_device_num();
     std::string hugepage_dir = find_hugepage_dir(hugepage_size);
     if (hugepage_dir.empty()) {
-        log_warning(
-            LogSiliconDriver,
-            "ttSiliconDevice::init_hugepage: no huge page mount found for hugepage_size: {}.",
-            hugepage_size);
+        log_warning(LogSiliconDriver, "init_hugepage: no huge page mount found for hugepage_size: {}.", hugepage_size);
         return false;
     }
 
@@ -741,7 +738,7 @@ bool PCIDevice::init_hugepage(uint32_t num_host_mem_channels) {
             // Probably a permissions problem.
             log_warning(
                 LogSiliconDriver,
-                "ttSiliconDevice::init_hugepage: physical_device_id: {} ch: {} creating hugepage mapping file failed.",
+                "init_hugepage: physical_device_id: {} ch: {} creating hugepage mapping file failed.",
                 physical_device_id,
                 ch);
             success = false;
@@ -762,10 +759,10 @@ bool PCIDevice::init_hugepage(uint32_t num_host_mem_channels) {
         if (mapping == MAP_FAILED) {
             log_warning(
                 LogSiliconDriver,
-                "UMD: Mapping a hugepage failed. (device: {}, {}/{} errno: {}).",
+                "Mapping a hugepage failed. (device: {}, channel {}/{} errno: {}).",
                 physical_device_id,
                 ch,
-                num_host_mem_channels,
+                num_host_mem_channels - 1,
                 strerror(errno));
             if (hugepage_st.st_size == 0) {
                 log_warning(
@@ -790,13 +787,7 @@ bool PCIDevice::init_hugepage(uint32_t num_host_mem_channels) {
         auto fd = get_fd();
 
         if (ioctl(fd, TENSTORRENT_IOCTL_PIN_PAGES, &pin_pages) == -1) {
-            log_warning(
-                LogSiliconDriver,
-                "---- ttSiliconDevice::init_hugepage: physical_device_id: {} ch: {} TENSTORRENT_IOCTL_PIN_PAGES failed "
-                "(errno: {}). Common Issue: Requires TTMKD >= 1.11, see following file contents...",
-                physical_device_id,
-                ch,
-                strerror(errno));
+            log_warning(LogSiliconDriver, "Failed to pin pages (errno: {}).", strerror(errno));
             munmap(mapping, hugepage_size);
             print_file_contents("/sys/module/tenstorrent/version", "(TTKMD version)");
             print_file_contents("/proc/meminfo");
@@ -809,7 +800,7 @@ bool PCIDevice::init_hugepage(uint32_t num_host_mem_channels) {
 
         log_debug(
             LogSiliconDriver,
-            "ttSiliconDevice::init_hugepage: physical_device_id: {} ch: {} mapping_size: {} physical address 0x{:x}",
+            "init_hugepage: physical_device_id: {} ch: {} mapping_size: {} physical address 0x{:x}",
             physical_device_id,
             ch,
             hugepage_size,