rocr: Add WaitMultiple to core Signal

Replaces WaitAny with WaitMultiple to more closely align with the
underlying driver API for waiting on multiple events.

WaitMultiple adds a single parameter, wait_on_all, to the WaitAny
interface providing a single function for waiting on multiple
events when we only need AND and OR semantics for the signal
checking logic.

Change-Id: I68a4a45d48151d9d69aef02fd8f7263b9e6c0e75

runtime/hsa-runtime/inc/hsa_api_trace.h

@@ -3,7 +3,7 @@
 // The University of Illinois/NCSA
 // Open Source License (NCSA)
 //
-// Copyright (c) 2014-2020, Advanced Micro Devices, Inc. All rights reserved.
+// Copyright (c) 2014-2024, Advanced Micro Devices, Inc. All rights reserved.
 //
 // Developed by:
 //
@@ -268,6 +268,7 @@ struct AmdExtTable {
   decltype(hsa_amd_queue_get_info)* hsa_amd_queue_get_info_fn;
   decltype(hsa_amd_vmem_address_reserve_align)* hsa_amd_vmem_address_reserve_align_fn;
   decltype(hsa_amd_enable_logging)* hsa_amd_enable_logging_fn;
+  decltype(hsa_amd_signal_wait_all)* hsa_amd_signal_wait_all_fn;
 };
 
 // Table to export HSA Core Runtime Apis

runtime/hsa-runtime/inc/hsa_api_trace_version.h

@@ -58,7 +58,7 @@
 // Step Ids of the Api tables exported by Hsa Core Runtime
 #define HSA_API_TABLE_STEP_VERSION                  0x01
 #define HSA_CORE_API_TABLE_STEP_VERSION             0x00
-#define HSA_AMD_EXT_API_TABLE_STEP_VERSION          0x04
+#define HSA_AMD_EXT_API_TABLE_STEP_VERSION          0x05
 #define HSA_FINALIZER_API_TABLE_STEP_VERSION        0x00
 #define HSA_IMAGE_API_TABLE_STEP_VERSION            0x01
 // Rocprofiler just checks HSA_MAGE_EXT_API_TABLE_STEP_VERSION

runtime/hsa-runtime/inc/hsa_ext_amd.h

@@ -57,9 +57,10 @@
  * - 1.4 - Virtual Memory API
  * - 1.5 - hsa_amd_agent_info: HSA_AMD_AGENT_INFO_MEMORY_PROPERTIES
  * - 1.6 - Virtual Memory API: hsa_amd_vmem_address_reserve_align
+ * - 1.7 - hsa_amd_signal_wait_all
  */
 #define HSA_AMD_INTERFACE_VERSION_MAJOR 1
-#define HSA_AMD_INTERFACE_VERSION_MINOR 6
+#define HSA_AMD_INTERFACE_VERSION_MINOR 7
 
 #ifdef __cplusplus
 extern "C" {
@@ -1179,6 +1180,20 @@ hsa_status_t HSA_API
                                  hsa_signal_value_t value,
                                  hsa_amd_signal_handler handler, void* arg);
 
+/**
+ * @brief Wait for all signal-condition pairs to be satisfied.
+ *
+ * @details Allows waiting for all of several signal and condition pairs to be
+ * satisfied. The function returns 0 if all signals met their conditions and -1
+ * on a timeout. The value of each signal's satisfying value is returned in
+ * satisfying_value unless satisfying_value is nullptr. This function provides
+ * only relaxed memory semantics.
+ */
+uint32_t HSA_API hsa_amd_signal_wait_all(uint32_t signal_count, hsa_signal_t* signals,
+                                         hsa_signal_condition_t* conds, hsa_signal_value_t* values,
+                                         uint64_t timeout_hint, hsa_wait_state_t wait_hint,
+                                         hsa_signal_value_t* satisfying_values);
+
 /**
  * @brief Wait for any signal-condition pair to be satisfied.
  *
@@ -1429,7 +1444,7 @@ typedef enum {
  * following its memory access model. The actual placement may vary or migrate
  * due to the system's NUMA policy and state, which is beyond the scope of
  * HSA APIs.
- */ 
+ */
 typedef struct hsa_amd_memory_pool_s {
   /**
    * Opaque handle.
@@ -2972,7 +2987,7 @@ typedef enum hsa_amd_svm_attribute_s {
   HSA_AMD_SVM_ATTRIB_ACCESS_QUERY = 0x203,
 } hsa_amd_svm_attribute_t;
 
-// List type for hsa_amd_svm_attributes_set/get. 
+// List type for hsa_amd_svm_attributes_set/get.
 typedef struct hsa_amd_svm_attribute_pair_s {
   // hsa_amd_svm_attribute_t value.
   uint64_t attribute;