fix_pals

utils/yaml_ast.rb

@@ -442,6 +442,7 @@ def find_types_map(types, cast_type, map)
   "size_t" => [false, 8, "ffi_type_pointer"],
   "intptr_t" => [true, 8, "ffi_type_pointer"],
   "uintptr_t" => [false, 8, "ffi_type_pointer"],
+  "_Bool" => [false, 4, "ffi_type_uint32"],
 }
 
 INT_SIGN_MAP = INT_TYPE_MAP.map { |k, v| [k, v[0]] }.to_h

ze/btx_zeinterval_callbacks.hpp

@@ -2,6 +2,7 @@
 
 #include "xprof_utils.hpp"
 #include <cstddef> // Bytes
+typedef bool _Bool;
 #include <metababel/metababel.h>
 #include <optional>
 #include <stdexcept>

ze/gen_ze.rb

@@ -67,6 +67,8 @@ def gen_struct_printer(namespace, types)
     puts <<EOF
   case ZES_STRUCTURE_TYPE_BASE_STATE:
     break;
+  case ZES_STRUCTURE_TYPE_DEVICE_UUID:
+    break;
 EOF
   end
   puts <<EOF
@@ -219,6 +221,11 @@ def gen_struct_printer(namespace, types)
     puts <<EOF
   _init_tracer();
 EOF
+    if c.name == 'zeInit'
+      puts <<EOF
+  _init_tracer_dump();
+EOF
+    end
   end
   common_block.call(c, provider, types)
   if c.has_return_type?

ze/headers.patch

@@ -42,7 +42,7 @@ diff -u4 -r --new-file include/layers/zel_tracing_register_cb.h modified_include
 diff -u4 -r --new-file include/loader/ze_loader_api.h modified_include/loader/ze_loader_api.h
 --- include/loader/ze_loader_api.h	2024-04-02 13:42:16.693493309 -0500
 +++ modified_include/loader/ze_loader_api.h	2024-04-02 13:11:54.060768340 -0500
-@@ -7,10 +7,12 @@
+@@ -7,10 +7,13 @@
   * @file ze_loader_api.cpp
   *
   */
@@ -52,12 +52,13 @@ diff -u4 -r --new-file include/loader/ze_loader_api.h modified_include/loader/ze
 +//#include "ze_loader_internal.h"
 +//#include "loader/ze_loader.h"
 +#include "../ze_api.h"
++#include <stdbool.h>
 +#define HMODULE void *
 
  #define TRACING_COMP_NAME "loader"
 
  #if defined(__cplusplus)
-@@ -54,27 +56,10 @@
+@@ -54,27 +57,10 @@
 
  ///////////////////////////////////////////////////////////////////////////////
  /// @brief Exported function for getting version

ze/include/layers/zel_tracing_register_cb.h

@@ -3251,4 +3251,4 @@ zelTracerResetAllCallbacks(zel_tracer_handle_t hTracer);
 } // extern "C"
 #endif
 
-#endif // zel_tracing_register_cb_H
+#endif // zel_tracing_register_cb_H

ze/include/loader/ze_loader_api.h

@@ -33,13 +33,14 @@ zeLoaderInit();
 ///     - ::ZE_RESULT_SUCCESS
 ///     - ::ZE_RESULT_ERROR_UNINITIALIZED
 ZE_DLLEXPORT ze_result_t ZE_APICALL
-zelLoaderDriverCheck(ze_init_flags_t flags);
-
+//zelLoaderDriverCheck(ze_init_flags_t flags);
+zelLoaderDriverCheck(ze_init_flags_t flags, void *globalInitStored, void *sysmanGlobalInitStored, bool *requireDdiReinit, bool sysmanOnly);
 
 ///////////////////////////////////////////////////////////////////////////////
 /// @brief Internal function for Setting the ddi tables for the Tracing Layer.
 ///
 ZE_DLLEXPORT ze_result_t ZE_APICALL
+//zelLoaderTracingLayerInit(std::atomic<ze_dditable_t *> &zeDdiTable, std::atomic<zet_dditable_t *> &zetDdiTable, std::atomic<zes_dditable_t *> &zesDdiTable);
 zelLoaderTracingLayerInit(void *zeDdiTable, void *zetDdiTable, void *zesDdiTable);
 
 
@@ -77,4 +78,4 @@ zelLoaderTranslateHandleInternal(
 
 #if defined(__cplusplus)
 }
-#endif
+#endif

ze/include/ze_api.h

@@ -5,7 +5,7 @@
  * SPDX-License-Identifier: MIT
  *
  * @file ze_api.h
- * @version v1.9-r1.9.1
+ * @version v1.9-r1.9.3
  *
  */
 #ifndef _ZE_API_H
@@ -1443,7 +1443,7 @@ typedef struct _ze_device_properties_t
                                                                             ///< structure (i.e. contains stype and pNext).
     ze_device_type_t type;                                                  ///< [out] generic device type
     uint32_t vendorId;                                                      ///< [out] vendor id from PCI configuration
-    uint32_t deviceId;                                                      ///< [out] device id from PCI configuration
+    uint32_t deviceId;                                                      ///< [out] device id from PCI configuration.
                                                                             ///< Note, the device id uses little-endian format.
     ze_device_property_flags_t flags;                                       ///< [out] 0 (none) or a valid combination of ::ze_device_property_flag_t
     uint32_t subdeviceId;                                                   ///< [out] sub-device id. Only valid if ::ZE_DEVICE_PROPERTY_FLAG_SUBDEVICE
@@ -4697,6 +4697,9 @@ typedef enum _ze_image_format_layout_t
     ZE_IMAGE_FORMAT_LAYOUT_444P = 40,                                       ///< Media Format: 444P. Format type and swizzle is ignored for this.
     ZE_IMAGE_FORMAT_LAYOUT_RGBP = 41,                                       ///< Media Format: RGBP. Format type and swizzle is ignored for this.
     ZE_IMAGE_FORMAT_LAYOUT_BRGP = 42,                                       ///< Media Format: BRGP. Format type and swizzle is ignored for this.
+    ZE_IMAGE_FORMAT_LAYOUT_8_8_8 = 43,                                      ///< 3-component 8-bit layout
+    ZE_IMAGE_FORMAT_LAYOUT_16_16_16 = 44,                                   ///< 3-component 16-bit layout
+    ZE_IMAGE_FORMAT_LAYOUT_32_32_32 = 45,                                   ///< 3-component 32-bit layout
     ZE_IMAGE_FORMAT_LAYOUT_FORCE_UINT32 = 0x7fffffff
 
 } ze_image_format_layout_t;
@@ -10545,6 +10548,9 @@ typedef enum _ze_image_bindless_exp_flag_t
 {
     ZE_IMAGE_BINDLESS_EXP_FLAG_BINDLESS = ZE_BIT(0),                        ///< Bindless images are created with ::zeImageCreate. The image handle
                                                                             ///< created with this flag is valid on both host and device.
+    ZE_IMAGE_BINDLESS_EXP_FLAG_SAMPLED_IMAGE = ZE_BIT(1),                   ///< Bindless sampled images are created with ::zeImageCreate by combining
+                                                                            ///< BINDLESS and SAMPLED_IMAGE.
+                                                                            ///< Create sampled image view from bindless unsampled image using SAMPLED_IMAGE.
     ZE_IMAGE_BINDLESS_EXP_FLAG_FORCE_UINT32 = 0x7fffffff
 
 } ze_image_bindless_exp_flag_t;
@@ -10564,6 +10570,8 @@ typedef struct _ze_image_bindless_exp_desc_t
                                                                             ///< When the flag is passed to ::zeImageCreate, then only the memory for
                                                                             ///< the image is allocated.
                                                                             ///< Additional image handles can be created with ::zeImageViewCreateExt.
+                                                                            ///< When ::ZE_IMAGE_BINDLESS_EXP_FLAG_SAMPLED_IMAGE flag is passed,
+                                                                            ///< ::ze_sampler_desc_t must be attached via pNext member of ::ze_image_bindless_exp_desc_t.
 
 } ze_image_bindless_exp_desc_t;
 

ze/include/ze_ddi.h

@@ -5,7 +5,7 @@
  * SPDX-License-Identifier: MIT
  *
  * @file ze_ddi.h
- * @version v1.9-r1.9.1
+ * @version v1.9-r1.9.3
  *
  */
 #ifndef _ZE_DDI_H

ze/include/zes_api.h

@@ -5,7 +5,7 @@
  * SPDX-License-Identifier: MIT
  *
  * @file zes_api.h
- * @version v1.9-r1.9.1
+ * @version v1.9-r1.9.3
  *
  */
 #ifndef _ZES_API_H
@@ -150,6 +150,7 @@ typedef enum _zes_structure_type_t
     ZES_STRUCTURE_TYPE_ENGINE_EXT_PROPERTIES = 0x2b,                        ///< ::zes_engine_ext_properties_t
     ZES_STRUCTURE_TYPE_RESET_PROPERTIES = 0x2c,                             ///< ::zes_reset_properties_t
     ZES_STRUCTURE_TYPE_DEVICE_EXT_PROPERTIES = 0x2d,                        ///< ::zes_device_ext_properties_t
+    ZES_STRUCTURE_TYPE_DEVICE_UUID = 0x2e,                                  ///< ::zes_uuid_t
     ZES_STRUCTURE_TYPE_POWER_DOMAIN_EXP_PROPERTIES = 0x00020001,            ///< ::zes_power_domain_exp_properties_t
     ZES_STRUCTURE_TYPE_MEM_BANDWIDTH_COUNTER_BITS_EXP_PROPERTIES = 0x00020002,  ///< ::zes_mem_bandwidth_counter_bits_exp_properties_t
     ZES_STRUCTURE_TYPE_MEMORY_PAGE_OFFLINE_STATE_EXP = 0x00020003,          ///< ::zes_mem_page_offline_state_exp_t

ze/include/zes_ddi.h

@@ -5,7 +5,7 @@
  * SPDX-License-Identifier: MIT
  *
  * @file zes_ddi.h
- * @version v1.9-r1.9.1
+ * @version v1.9-r1.9.3
  *
  */
 #ifndef _ZES_DDI_H

ze/include/zet_api.h

@@ -5,7 +5,7 @@
  * SPDX-License-Identifier: MIT
  *
  * @file zet_api.h
- * @version v1.9-r1.9.1
+ * @version v1.9-r1.9.3
  *
  */
 #ifndef _ZET_API_H

ze/include/zet_ddi.h

@@ -5,7 +5,7 @@
  * SPDX-License-Identifier: MIT
  *
  * @file zet_ddi.h
- * @version v1.9-r1.9.1
+ * @version v1.9-r1.9.3
  *
  */
 #ifndef _ZET_DDI_H

ze/tracer_ze_helpers.include.c

@@ -38,6 +38,7 @@ static int _do_cleanup = 0;
 static int _do_chained_structs = 0;
 static int _do_paranoid_drift = 0;
 static int _do_paranoid_memory_location = 0;
+static int _do_ddi_table_forward = 0;
 
 pthread_mutex_t ze_closures_mutex = PTHREAD_MUTEX_INITIALIZER;
 
@@ -627,6 +628,10 @@ static pthread_once_t _init = PTHREAD_ONCE_INIT;
 static __thread volatile int in_init = 0;
 static volatile unsigned int _initialized = 0;
 
+static pthread_once_t _init_dump = PTHREAD_ONCE_INIT;
+static __thread volatile int in_init_dump = 0;
+static volatile unsigned int _initialized_dump = 0;
+
 static inline int _do_state() {
   return _do_profile ||
          tracepoint_enabled(lttng_ust_ze_properties, memory_info_properties) ||
@@ -814,13 +819,11 @@ static void _load_tracer(void) {
   if (s)
     _do_chained_structs = 1;
 
-  find_ze_symbols(handle, verbose);
-
-  //FIX for intel tracing layer that needs to register its callbacks first...
-  ZE_INIT_PTR(0);
+  s = getenv("LTTNG_UST_ZE_DDI_TABLE_FORWARD");
+  if (s)
+    _do_ddi_table_forward = 1;
 
-  if (tracepoint_enabled(lttng_ust_ze_properties, driver) || tracepoint_enabled(lttng_ust_ze_properties, device) || tracepoint_enabled(lttng_ust_ze_properties, subdevice))
-    _dump_properties();
+  find_ze_symbols(handle, verbose);
 
   s = getenv("LTTNG_UST_ZE_PROFILE");
   if (s)
@@ -845,6 +848,13 @@ static void _load_tracer(void) {
 #endif
 }
 
+static void _load_tracer_dump(void) {
+  //FIX for intel tracing layer that needs to register its callbacks first...
+  ZE_INIT_PTR(0);
+  if (tracepoint_enabled(lttng_ust_ze_properties, driver) || tracepoint_enabled(lttng_ust_ze_properties, device) || tracepoint_enabled(lttng_ust_ze_properties, subdevice))
+    _dump_properties();
+}
+
 static inline void _init_tracer(void) {
   if( __builtin_expect (_initialized, 1) )
     return;
@@ -858,3 +868,17 @@ static inline void _init_tracer(void) {
   }
   _initialized = 1;
 }
+
+static inline void _init_tracer_dump(void) {
+  if( __builtin_expect (_initialized_dump, 1) )
+    return;
+  /* Avoid reentrancy */
+  if (!in_init_dump) {
+    in_init_dump=1;
+    __sync_synchronize();
+    pthread_once(&_init_dump, _load_tracer_dump);
+    __sync_synchronize();
+    in_init_dump=0;
+  }
+  _initialized_dump = 1;
+}

ze/ze_model.rb

@@ -387,7 +387,7 @@ def upper_snake_case(str)
   child_types = STRUCT_MAP.select { |k, _| k.match(parent_type) }
   str = <<EOF
   #{c.type} _retval;
-  if (pDdiTable && version <= ZE_API_VERSION_CURRENT) {
+  if (!_do_ddi_table_forward && pDdiTable && version <= ZE_API_VERSION_CURRENT) {
 EOF
   str << "   "
   str << child_types.reverse_each.collect { |k, v|