Hi Julian! On Tue, 28 Oct 2014 11:16:19 +0000, Julian Brown <jul...@codesourcery.com> wrote: > This patch rationalises TLS support by moving all thread-local > variables into a single structure. Because this meant interfering with > how per-thread/per-device initialisation was done, I took the > opportunity to tidy up a couple of other bits along the way. Highlights > are: [...]
Thanks for all that! > OK to apply? The rationale given certainly sounds plausible, and considering that the bulk of the code that we're adding to libgomp has not yet been reviewed, please just check this in on top. As far as I'm concerned, that also applies to future such patches, unless you're explicitly asking for review now. (Please do continue to post patches, though.) > libgomp/ > * env.c (oacc-int.h): Include. > * libgomp-plugin.h (GOMP_PLUGIN_async_unmap_vars): Don't declare > here. > * libgomp.map (PLUGIN_1.0): Add GOMP_PLUGIN_acc_thread symbol. > * oacc-async.c (oacc-int.h): Include. > (acc_async_test, acc_async_test_all, acc_wait, acc_wait_async) > (acc_wait_all, acc_wait_all_async): Use base_dev not ACC_dev. > * oacc-cuda.c (oacc-int.h): Include. > (acc_get_current_cuda_device, acc_get_current_cuda_context) > (acc_get_cuda_stream): Use base_dev not ACC_dev. > (acc_set_cuda_stream): Call ACC_lazy_initialize, use base_dev not > ACC_dev. > * oacc-host.c (oacc-plugin.h, oacc-int.h): Include one or the > other, depending on HOST_NONSHM_PLUGIN setting. > (openacc_create_thread_data, openacc_destroy_thread_data): New. > (host_dispatch): Initialize create_thread_data_func, > destroy_thread_data_func hooks. > * oacc-init.c (oacc-int.h): Include. > (acc_device_lock): Make static. > (ACC_dev, ACC_handle, handle_num, struct ACC_context, ACC_contexts) > (_ACC_contexts): Remove. > (base_dev, goacc_threads, goacc_thread_lock, goacc_cleanup_key): > Add. > (goacc_tls_data, goacc_tls_key): Define one or the other, depending > on HAVE_TLS setting. > (dispatchers): Add comment. > (ACC_register): Only register 0th device. > (close_handle): Remove. > (_acc_init): Add comment. Don't use ACC_dev. > (goacc_new_thread, goacc_destroy_thread): New. > (lazy_open): Update comment. Remove code relating to contexts and > memory maps. Create thread-specific data if necessary. > (acc_init): Use base_dev not ACC_dev. > (_acc_shutdown): Don't call close_handle. Use base_dev not ACC_dev. > Free target-specific TLS data and close devices that are open at > the time of shutdown. > (lazy_init): Add comment. Use base_dev not ACC_dev. > (lazy_init_and_open): Add comment. Use base_dev not ACC_dev. > (acc_get_num_devices, acc_get_device_type, acc_get_device_num): Use > base_dev not ACC_dev. > (acc_set_device_num): Likewise. If switching device number, > de-associate previous device number from the current thread, but do > not close the device. > (acc_on_device): Check type for device for current thread, not > ACC_dev. > (ACC_runtime_initialize): Initialize goacc_tls_key, > goacc_cleanup_key, base_dev, goacc_threads, goacc_thread_lock. Don't > initialize ACC_contexts. > (saved_bound_dev): Remove. > (ACC_save_and_set_bind, ACC_restore_bind): Use goacc_thread. > (ACC_lazy_initialize): Use base_dev, goacc_thread instead of > ACC_dev, ACC_memmap. > * oacc-int.h (struct ACC_dispatch_t): Don't declare here. > (struct goacc_thread): New. > (goacc_tls_data): Define for HAVE_TLS. > (goacc_tls_key): Define for !HAVE_TLS. > (goacc_thread): New (versions with and without HAVE_TLS). > (ACC_mem_open, ACC_mem_close, ACC_resolve_device, ACC_dev) > (ACC_handle, struct memmap_t, ACC_memmap): Remove. > (base_dev): Add extern declaration. > * oacc-mem.c (oacc-int.h, assert.h): Include. > (ACC_memmap, ACC_mem_open, ACC_mem_close): Remove. > (lookup_host): Change memmap_t argument to struct > gomp_memory_mapping. > (lookup_dev): Change memmap_t argument to struct target_mem_desc. > Use list_count not refcount for iterating over mapped elements. > (acc_malloc): Use base_dev not ACC_dev. > (acc_free): Update call to lookup_dev. Use base_dev not ACC_dev. > (acc_memcpy_to_device, acc_memcpy_from_device): Use base_dev not > ACC_dev. > (acc_deviceptr, acc_is_present): Update call to lookup_host. > (acc_hostptr): Update call to lookup_dev. > (acc_map_data): Look up thread device instead of using ACC_dev, > update calls to lookup_host, lookup_dev. Use data environment in > device descriptor. > (acc_unmap_data): Update call to lookup_host. Remove mapped data > from data environment not ACC_memmap. > (present_create_copy): Update call to lookup_host. Use data > environment instead of list in ACC_memmap. > (delete_copyout): Update call to lookup_host. Look up device in > current thread info instead of using ACC_dev. > (update_dev_host): Look up device in current thread info instead of > using ACC_dev. > * oacc-parallel.c (oacc-int.h): Include. > (struct devgeom, devgeom, dump_devaddrs): Remove. > (select_acc_device): Call ACC_lazy_initialize earlier. > (GOACC_parallel): Use device for current thread instead of ACC_dev. > Use memory map from current device. > (GOACC_data_start): Likewise. Use thread info block for mapped data. > (GOACC_data_end): Use thread info block for mapped data. > (goacc_wait): Use device for current thread instead of ACC_dev. > (GOACC_update): Likewise. Formatting fixes. > * oacc-plugin.c (ACC_plugin_register): Remove. > (oacc-int.h): Include. > (GOMP_PLUGIN_acc_thread): New. > * oacc-plugin.h (target.h): Don't include. > (ACC_plugin_register): Remove. > (GOMP_PLUGIN_async_unmap_vars, GOMP_PLUGIN_acc_thread): Add extern > declarations. > * plugin-nvptx.c (oacc-plugin.h): Include. > (current_stream, PTX_dev, PTX_devices): Remove. > (struct nvptx_thread): New. > (nvptx_thread): New function. > (select_stream_for_async): Locate ptx_dev in device-specific TLS > data instead of using TLS PTX_dev variable. > (PTX_init): Don't initialize PTX_devices. > (PTX_open_device): Remove PTX_devices list handling. Tweak context > initialization. > (PTX_close_device): Remove PTX_devices list handling. Find PTX > device info via function argument instead of global TLS variable. > (PTX_get_num_devices): Make callable when backend has not been > initialized. > (event_gc): Find PTX device info, current stream via nvptx_thread. > (event_add, PTX_exec, PTX_host2dev, PTX_dev2host) > (PTX_async_test_all, PTX_wait_all, PTX_wait_all_async) > (PTX_get_current_cuda_device, PTX_get_current_cuda_context) > (PTX_get_cuda_stream, PTX_set_cuda_stream, openacc_close_device) > (openacc_set_device_num, openacc_register_async_cleanup) > (openacc_async_set_async): Likewise. > (openacc_create_thread_data, openacc_destroy_thread_data): New. > * target.c (oacc-int.h): Include. > (gomp_fini_device): Split out memory-map freeing into... > (gomp_free_memmap): ...this new function. > (gomp_load_plugin_for_device): Initialize > openacc.create_thread_data_func, openacc.destroy_thread_data_func > hooks. > (gomp_find_available_plugins): Initialize one target_device_descr > per physical device. > * target.h (oacc-int.h): Don't include. > (ACC_dispatch_t): Declare here. Add data_environ, ord fields. > Update comment for mem_map field. > (gomp_free_memmap): Add prototype. Grüße, Thomas
pgpLcbrF7b6Xx.pgp
Description: PGP signature
