multiple changes update

[ROCm/roctracer commit: c1cca98a71]
This commit is contained in:
Evgeny
2019-06-24 21:05:12 -05:00
parent 775e64f392
commit 7a00dcb37e
19 changed files with 1420 additions and 308 deletions
@@ -145,7 +145,7 @@ HsaRsrcFactory::HsaRsrcFactory(bool initialize_hsa) : initialize_hsa_(initialize
CHECK_STATUS("loader API table query failed", status);
// Instantiate HSA timer
timer_ = new HsaTimer;
timer_ = new HsaTimer(&hsa_api_);
CHECK_STATUS("HSA timer allocation failed",
(timer_ == NULL) ? HSA_STATUS_ERROR : HSA_STATUS_SUCCESS);
@@ -172,7 +172,6 @@ void HsaRsrcFactory::InitHsaApiTable(HsaApiTable* table) {
hsa_api_.hsa_init = table->core_->hsa_init_fn;
hsa_api_.hsa_shut_down = table->core_->hsa_shut_down_fn;
hsa_api_.hsa_agent_get_info = table->core_->hsa_agent_get_info_fn;
hsa_api_.hsa_iterate_agents = table->core_->hsa_iterate_agents_fn;
hsa_api_.hsa_queue_create = table->core_->hsa_queue_create_fn;
@@ -180,14 +179,13 @@ void HsaRsrcFactory::InitHsaApiTable(HsaApiTable* table) {
hsa_api_.hsa_queue_load_write_index_relaxed = table->core_->hsa_queue_load_write_index_relaxed_fn;
hsa_api_.hsa_queue_store_write_index_relaxed = table->core_->hsa_queue_store_write_index_relaxed_fn;
hsa_api_.hsa_queue_load_read_index_relaxed = table->core_->hsa_queue_load_read_index_relaxed_fn;
hsa_api_.hsa_signal_create = table->core_->hsa_signal_create_fn;
hsa_api_.hsa_signal_destroy = table->core_->hsa_signal_destroy_fn;
hsa_api_.hsa_signal_load_relaxed = table->core_->hsa_signal_load_relaxed_fn;
hsa_api_.hsa_signal_store_relaxed = table->core_->hsa_signal_store_relaxed_fn;
hsa_api_.hsa_signal_store_screlease = table->core_->hsa_signal_store_screlease_fn;
hsa_api_.hsa_signal_wait_scacquire = table->core_->hsa_signal_wait_scacquire_fn;
hsa_api_.hsa_system_get_major_extension_table = table->core_->hsa_system_get_major_extension_table_fn;
hsa_api_.hsa_signal_store_screlease = table->core_->hsa_signal_store_screlease_fn;
hsa_api_.hsa_code_object_reader_create_from_file = table->core_->hsa_code_object_reader_create_from_file_fn;
hsa_api_.hsa_executable_create_alt = table->core_->hsa_executable_create_alt_fn;
@@ -195,21 +193,23 @@ void HsaRsrcFactory::InitHsaApiTable(HsaApiTable* table) {
hsa_api_.hsa_executable_freeze = table->core_->hsa_executable_freeze_fn;
hsa_api_.hsa_executable_get_symbol = table->core_->hsa_executable_get_symbol_fn;
hsa_api_.hsa_system_get_info = table->core_->hsa_system_get_info_fn;
hsa_api_.hsa_system_get_major_extension_table = table->core_->hsa_system_get_major_extension_table_fn;
hsa_api_.hsa_amd_agent_iterate_memory_pools = table->amd_ext_->hsa_amd_agent_iterate_memory_pools_fn;
hsa_api_.hsa_amd_memory_pool_get_info = table->amd_ext_->hsa_amd_memory_pool_get_info_fn;
hsa_api_.hsa_amd_memory_pool_allocate = table->amd_ext_->hsa_amd_memory_pool_allocate_fn;
hsa_api_.hsa_amd_agents_allow_access = table->amd_ext_->hsa_amd_agents_allow_access_fn;
hsa_api_.hsa_amd_memory_async_copy = table->amd_ext_->hsa_amd_memory_async_copy_fn;
hsa_api_.hsa_amd_signal_async_handler = table->amd_ext_->hsa_amd_signal_async_handler_fn;
hsa_api_.hsa_amd_profiling_set_profiler_enabled = table->amd_ext_->hsa_amd_profiling_set_profiler_enabled_fn;
hsa_api_.hsa_amd_profiling_get_async_copy_time = table->amd_ext_->hsa_amd_profiling_get_async_copy_time_fn;
hsa_api_.hsa_amd_profiling_get_dispatch_time = table->amd_ext_->hsa_amd_profiling_get_dispatch_time_fn;
} else {
hsa_api_.hsa_init = hsa_init;
hsa_api_.hsa_shut_down = hsa_shut_down;
hsa_api_.hsa_agent_get_info = hsa_agent_get_info;
hsa_api_.hsa_iterate_agents = hsa_iterate_agents;
hsa_api_.hsa_queue_create = hsa_queue_create;
@@ -217,19 +217,13 @@ void HsaRsrcFactory::InitHsaApiTable(HsaApiTable* table) {
hsa_api_.hsa_queue_load_write_index_relaxed = hsa_queue_load_write_index_relaxed;
hsa_api_.hsa_queue_store_write_index_relaxed = hsa_queue_store_write_index_relaxed;
hsa_api_.hsa_queue_load_read_index_relaxed = hsa_queue_load_read_index_relaxed;
hsa_api_.hsa_signal_create = hsa_signal_create;
hsa_api_.hsa_signal_destroy = hsa_signal_destroy;
hsa_api_.hsa_signal_load_relaxed = hsa_signal_load_relaxed;
hsa_api_.hsa_signal_store_relaxed = hsa_signal_store_relaxed;
hsa_api_.hsa_signal_wait_scacquire = hsa_signal_wait_scacquire;
hsa_api_.hsa_amd_agent_iterate_memory_pools = hsa_amd_agent_iterate_memory_pools;
hsa_api_.hsa_amd_memory_pool_get_info = hsa_amd_memory_pool_get_info;
hsa_api_.hsa_amd_memory_pool_allocate = hsa_amd_memory_pool_allocate;
hsa_api_.hsa_amd_agents_allow_access = hsa_amd_agents_allow_access;
hsa_api_.hsa_amd_memory_async_copy = hsa_amd_memory_async_copy;
hsa_api_.hsa_system_get_major_extension_table = hsa_system_get_major_extension_table;
hsa_api_.hsa_signal_store_screlease = hsa_signal_store_screlease;
hsa_api_.hsa_code_object_reader_create_from_file = hsa_code_object_reader_create_from_file;
hsa_api_.hsa_executable_create_alt = hsa_executable_create_alt;
@@ -237,11 +231,19 @@ void HsaRsrcFactory::InitHsaApiTable(HsaApiTable* table) {
hsa_api_.hsa_executable_freeze = hsa_executable_freeze;
hsa_api_.hsa_executable_get_symbol = hsa_executable_get_symbol;
hsa_api_.hsa_system_get_info = hsa_system_get_info;
hsa_api_.hsa_system_get_major_extension_table = hsa_system_get_major_extension_table;
hsa_api_.hsa_amd_agent_iterate_memory_pools = hsa_amd_agent_iterate_memory_pools;
hsa_api_.hsa_amd_memory_pool_get_info = hsa_amd_memory_pool_get_info;
hsa_api_.hsa_amd_memory_pool_allocate = hsa_amd_memory_pool_allocate;
hsa_api_.hsa_amd_agents_allow_access = hsa_amd_agents_allow_access;
hsa_api_.hsa_amd_memory_async_copy = hsa_amd_memory_async_copy;
hsa_api_.hsa_amd_signal_async_handler = hsa_amd_signal_async_handler;
hsa_api_.hsa_amd_profiling_set_profiler_enabled = hsa_amd_profiling_set_profiler_enabled;
hsa_api_.hsa_amd_profiling_get_async_copy_time = hsa_amd_profiling_get_async_copy_time;
hsa_api_.hsa_amd_profiling_get_dispatch_time = hsa_amd_profiling_get_dispatch_time;
hsa_api_.hsa_signal_load_relaxed = hsa_signal_load_relaxed;
hsa_api_.hsa_signal_store_screlease = hsa_signal_store_screlease;
}
}
}
+36 -32
View File
@@ -71,41 +71,43 @@ static const size_t MEM_PAGE_MASK = MEM_PAGE_BYTES - 1;
typedef decltype(hsa_agent_t::handle) hsa_agent_handle_t;
struct hsa_pfn_t {
decltype(::hsa_init)* hsa_init;
decltype(::hsa_shut_down)* hsa_shut_down;
decltype(::hsa_agent_get_info)* hsa_agent_get_info;
decltype(hsa_init)* hsa_init;
decltype(hsa_shut_down)* hsa_shut_down;
decltype(hsa_agent_get_info)* hsa_agent_get_info;
decltype(hsa_iterate_agents)* hsa_iterate_agents;
decltype(::hsa_iterate_agents)* hsa_iterate_agents;
decltype(hsa_queue_create)* hsa_queue_create;
decltype(hsa_queue_destroy)* hsa_queue_destroy;
decltype(hsa_queue_load_write_index_relaxed)* hsa_queue_load_write_index_relaxed;
decltype(hsa_queue_store_write_index_relaxed)* hsa_queue_store_write_index_relaxed;
decltype(hsa_queue_load_read_index_relaxed)* hsa_queue_load_read_index_relaxed;
decltype(::hsa_queue_create)* hsa_queue_create;
decltype(::hsa_queue_destroy)* hsa_queue_destroy;
decltype(::hsa_queue_load_write_index_relaxed)* hsa_queue_load_write_index_relaxed;
decltype(::hsa_queue_store_write_index_relaxed)* hsa_queue_store_write_index_relaxed;
decltype(::hsa_queue_load_read_index_relaxed)* hsa_queue_load_read_index_relaxed;
decltype(::hsa_signal_create)* hsa_signal_create;
decltype(::hsa_signal_destroy)* hsa_signal_destroy;
decltype(::hsa_signal_store_relaxed)* hsa_signal_store_relaxed;
decltype(::hsa_signal_wait_scacquire)* hsa_signal_wait_scacquire;
decltype(hsa_signal_create)* hsa_signal_create;
decltype(hsa_signal_destroy)* hsa_signal_destroy;
decltype(hsa_signal_load_relaxed)* hsa_signal_load_relaxed;
decltype(hsa_signal_store_relaxed)* hsa_signal_store_relaxed;
decltype(hsa_signal_wait_scacquire)* hsa_signal_wait_scacquire;
decltype(hsa_signal_store_screlease)* hsa_signal_store_screlease;
decltype(::hsa_amd_agent_iterate_memory_pools)* hsa_amd_agent_iterate_memory_pools;
decltype(::hsa_amd_memory_pool_get_info)* hsa_amd_memory_pool_get_info;
decltype(::hsa_amd_memory_pool_allocate)* hsa_amd_memory_pool_allocate;
decltype(::hsa_amd_agents_allow_access)* hsa_amd_agents_allow_access;
decltype(::hsa_amd_memory_async_copy)* hsa_amd_memory_async_copy;
decltype(hsa_code_object_reader_create_from_file)* hsa_code_object_reader_create_from_file;
decltype(hsa_executable_create_alt)* hsa_executable_create_alt;
decltype(hsa_executable_load_agent_code_object)* hsa_executable_load_agent_code_object;
decltype(hsa_executable_freeze)* hsa_executable_freeze;
decltype(hsa_executable_get_symbol)* hsa_executable_get_symbol;
decltype(::hsa_system_get_major_extension_table)* hsa_system_get_major_extension_table;
decltype(hsa_system_get_info)* hsa_system_get_info;
decltype(hsa_system_get_major_extension_table)* hsa_system_get_major_extension_table;
decltype(::hsa_code_object_reader_create_from_file)* hsa_code_object_reader_create_from_file;
decltype(::hsa_executable_create_alt)* hsa_executable_create_alt;
decltype(::hsa_executable_load_agent_code_object)* hsa_executable_load_agent_code_object;
decltype(::hsa_executable_freeze)* hsa_executable_freeze;
decltype(::hsa_executable_get_symbol)* hsa_executable_get_symbol;
decltype(hsa_amd_agent_iterate_memory_pools)* hsa_amd_agent_iterate_memory_pools;
decltype(hsa_amd_memory_pool_get_info)* hsa_amd_memory_pool_get_info;
decltype(hsa_amd_memory_pool_allocate)* hsa_amd_memory_pool_allocate;
decltype(hsa_amd_agents_allow_access)* hsa_amd_agents_allow_access;
decltype(hsa_amd_memory_async_copy)* hsa_amd_memory_async_copy;
decltype(::hsa_amd_signal_async_handler)* hsa_amd_signal_async_handler;
decltype(::hsa_amd_profiling_get_async_copy_time)* hsa_amd_profiling_get_async_copy_time;
decltype(::hsa_amd_profiling_get_dispatch_time)* hsa_amd_profiling_get_dispatch_time;
decltype(::hsa_signal_load_relaxed)* hsa_signal_load_relaxed;
decltype(::hsa_signal_store_screlease)* hsa_signal_store_screlease;
decltype(hsa_amd_signal_async_handler)* hsa_amd_signal_async_handler;
decltype(hsa_amd_profiling_set_profiler_enabled)* hsa_amd_profiling_set_profiler_enabled;
decltype(hsa_amd_profiling_get_async_copy_time)* hsa_amd_profiling_get_async_copy_time;
decltype(hsa_amd_profiling_get_dispatch_time)* hsa_amd_profiling_get_dispatch_time;
};
// Encapsulates information about a Hsa Agent such as its
@@ -167,9 +169,9 @@ class HsaTimer {
static const timestamp_t TIMESTAMP_MAX = UINT64_MAX;
typedef long double freq_t;
HsaTimer() {
HsaTimer(const hsa_pfn_t* hsa_api) : hsa_api_(hsa_api) {
timestamp_t sysclock_hz = 0;
hsa_status_t status = hsa_system_get_info(HSA_SYSTEM_INFO_TIMESTAMP_FREQUENCY, &sysclock_hz);
hsa_status_t status = hsa_api_->hsa_system_get_info(HSA_SYSTEM_INFO_TIMESTAMP_FREQUENCY, &sysclock_hz);
CHECK_STATUS("hsa_system_get_info(HSA_SYSTEM_INFO_TIMESTAMP_FREQUENCY)", status);
sysclock_factor_ = (freq_t)1000000000 / (freq_t)sysclock_hz;
}
@@ -185,7 +187,7 @@ class HsaTimer {
// Return timestamp in 'ns'
timestamp_t timestamp_ns() const {
timestamp_t sysclock;
hsa_status_t status = hsa_system_get_info(HSA_SYSTEM_INFO_TIMESTAMP, &sysclock);
hsa_status_t status = hsa_api_->hsa_system_get_info(HSA_SYSTEM_INFO_TIMESTAMP, &sysclock);
CHECK_STATUS("hsa_system_get_info(HSA_SYSTEM_INFO_TIMESTAMP)", status);
return sysclock_to_ns(sysclock);
}
@@ -193,6 +195,8 @@ class HsaTimer {
private:
// Timestamp frequency factor
freq_t sysclock_factor_;
// HSA API table
const hsa_pfn_t* const hsa_api_;
};
class HsaRsrcFactory {