AI/rocm-systems
2
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity

Extend ExecutePM4() to accept completion signal and fences

Browse Source 
ExecutePM4() function can optionally accept extra arguments for
acquire fence scope, release fence scope andcompletion signal. When
a completion signal is provided, ExecutePM4() does not wait for the
commands to complete.

Change-Id: Ib2a433b7bce1cb6260be8b76fe902335bd5dfada
This commit is contained in:
David Yat Sin
2023-08-30 17:31:13 +00:00
parent d7adc94e3f
commit 721e56ef5c
5 changed files with 59 additions and 22 deletions
Download Patch File Download Diff File Expand all files Collapse all files
runtime/hsa-runtime/core/inc/queue.h
+18 -2
View File
@@ -346,8 +346,24 @@ class Queue : public Checked<0xFA3906A679F9DB49>, private LocalQueue {
/// @return hsa_status_t
virtual hsa_status_t GetCUMasking(uint32_t num_cu_mask_count, uint32_t* cu_mask) = 0;
// @brief Submits a block of PM4 and waits until it has been executed.
virtual void ExecutePM4(uint32_t* cmd_data, size_t cmd_size_b) = 0;
/// @brief Submits a block of PM4.
///
/// @param cmd_data pointer to command buffer
///
/// @param cmd_size_b command buffer size in bytes
///
/// @param acquireFence acquire-fence type
///
/// @param releaseFence acquire-fence type
///
/// @param signal optional wait signal
///
/// if @p signal is provided, function will return without waiting for commands to be executed
/// if @p signal is NULL, waits until commands have been executed.
virtual void ExecutePM4(uint32_t* cmd_data, size_t cmd_size_b,
hsa_fence_scope_t acquireFence = HSA_FENCE_SCOPE_NONE,
hsa_fence_scope_t releaseFence = HSA_FENCE_SCOPE_NONE,
hsa_signal_t* signal = NULL) = 0;
virtual void SetProfiling(bool enabled) {
AMD_HSA_BITS_SET(amd_queue_.queue_properties, AMD_QUEUE_PROPERTIES_ENABLE_PROFILING,