* ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR
* OTHER DEALINGS IN THE SOFTWARE.
*
-*/
+ */
/**
* \file amdgpu.h
#include <stdint.h>
#include <stdbool.h>
+#ifdef __cplusplus
+extern "C" {
+#endif
+
struct drm_amdgpu_info_hw_ip;
/*--------------------------------------------------------------------------*/
#define AMDGPU_TIMEOUT_INFINITE 0xffffffffffffffffull
/**
- * Used in amdgpu_cs_query_fence::flags, meaning that the given timeout
+ * Used in amdgpu_cs_query_fence_status(), meaning that the given timeout
* is absolute.
*/
#define AMDGPU_QUERY_FENCE_TIMEOUT_IS_ABSOLUTE (1 << 0)
*/
typedef struct amdgpu_va *amdgpu_va_handle;
+/**
+ * Define handle for semaphore
+ */
+typedef struct amdgpu_semaphore *amdgpu_semaphore_handle;
+
/*--------------------------------------------------------------------------*/
/* -------------------------- Structures ---------------------------------- */
/*--------------------------------------------------------------------------*/
};
/**
- * Structure describing memory allocation request
- *
- * \sa amdgpu_bo_alloc()
-*/
-struct amdgpu_bo_alloc_result {
- /** Assigned virtual MC Base Address */
- uint64_t virtual_mc_base_address;
-
- /** Handle of allocated memory to be used by the given process only. */
- amdgpu_bo_handle buf_handle;
-};
-
-/**
* Special UMD specific information associated with buffer.
*
* It may be need to pass some buffer charactersitic as part
*/
uint64_t phys_alignment;
- /**
- * Assigned virtual MC Base Address.
- * \note This information will be returned only if this buffer was
- * allocated in the same process otherwise 0 will be returned.
- */
- uint64_t virtual_mc_base_address;
-
/** Heap where to allocate memory. */
uint32_t preferred_heap;
/** Buffer size */
uint64_t alloc_size;
-
- /** Assigned virtual MC Base Address */
- uint64_t virtual_mc_base_address;
};
/**
};
/**
- * Structure describing CS dependency
+ * Structure describing CS fence
*
- * \sa amdgpu_cs_request, amdgpu_cs_submit()
+ * \sa amdgpu_cs_query_fence_status(), amdgpu_cs_request, amdgpu_cs_submit()
*
*/
-struct amdgpu_cs_dep_info {
- /** Context to which the fence belongs */
+struct amdgpu_cs_fence {
+
+ /** In which context IB was sent to execution */
amdgpu_context_handle context;
/** To which HW IP type the fence belongs */
};
/**
+ * Structure describing fence information
+ *
+ * \sa amdgpu_cs_request, amdgpu_cs_query_fence,
+ * amdgpu_cs_submit(), amdgpu_cs_query_fence_status()
+*/
+struct amdgpu_cs_fence_info {
+ /** buffer object for the fence */
+ amdgpu_bo_handle handle;
+
+ /** fence offset in the unit of sizeof(uint64_t) */
+ uint64_t offset;
+};
+
+/**
* Structure describing submission request
*
* \note We could have several IBs as packet. e.g. CE, CE, DE case for gfx
* Array of dependencies which need to be met before
* execution can start.
*/
- struct amdgpu_cs_dep_info *dependencies;
+ struct amdgpu_cs_fence *dependencies;
/** Number of IBs to submit in the field ibs. */
uint32_t number_of_ibs;
* IBs to submit. Those IBs will be submit together as single entity
*/
struct amdgpu_cs_ib_info *ibs;
-};
-
-/**
- * Structure describing request to check submission state using fence
- *
- * \sa amdgpu_cs_query_fence_status()
- *
-*/
-struct amdgpu_cs_query_fence {
- /** In which context IB was sent to execution */
- amdgpu_context_handle context;
-
- /** Timeout in nanoseconds. */
- uint64_t timeout_ns;
-
- /** To which HW IP type the fence belongs */
- unsigned ip_type;
-
- /** IP instance index if there are several IPs of the same type. */
- unsigned ip_instance;
-
- /** Ring index of the HW IP */
- uint32_t ring;
-
- /** Flags */
- uint64_t flags;
+ /**
+ * The returned sequence number for the command submission
+ */
+ uint64_t seq_no;
- /** Specify fence for which we need to check submission status.*/
- uint64_t fence;
+ /**
+ * The fence information
+ */
+ struct amdgpu_cs_fence_info fence_info;
};
/**
uint32_t vram_bit_width;
/** constant engine ram size*/
uint32_t ce_ram_size;
+ /* vce harvesting instance */
+ uint32_t vce_harvest_config;
+ /* PCI revision ID */
+ uint32_t pci_rev_id;
};
* See #amdgpu_device_initialize()
* \param alloc_buffer - \c [in] Pointer to the structure describing an
* allocation request
- * \param info - \c [out] Pointer to structure which return
- * information about allocated memory
+ * \param buf_handle - \c [out] Allocated buffer handle
*
* \return 0 on success\n
* <0 - Negative POSIX Error code
*/
int amdgpu_bo_alloc(amdgpu_device_handle dev,
struct amdgpu_bo_alloc_request *alloc_buffer,
- struct amdgpu_bo_alloc_result *info);
+ amdgpu_bo_handle *buf_handle);
/**
* Associate opaque data with buffer to be queried by another UMD
* want to map to GPU address space (make GPU accessible)
* (This address must be correctly aligned).
* \param size - [in] Size of allocation (must be correctly aligned)
- * \param amdgpu_bo_alloc_result - [out] Handle of allocation to be passed as
+ * \param buf_handle - [out] Buffer handle for the userptr memory
* resource on submission and be used in other operations.
*
*
*/
int amdgpu_create_bo_from_user_mem(amdgpu_device_handle dev,
void *cpu, uint64_t size,
- struct amdgpu_bo_alloc_result *info);
+ amdgpu_bo_handle *buf_handle);
/**
* Free previosuly allocated memory
int amdgpu_bo_free(amdgpu_bo_handle buf_handle);
/**
- * Request CPU access to GPU accessable memory
+ * Request CPU access to GPU accessible memory
*
* \param buf_handle - \c [in] Buffer handle
* \param cpu - \c [out] CPU address to be used for access
/**
* Wait until a buffer is not used by the device.
*
- * \param dev - \c [in] Device handle. See #amdgpu_lib_initialize()
+ * \param dev - \c [in] Device handle. See #amdgpu_device_initialize()
* \param buf_handle - \c [in] Buffer handle.
* \param timeout_ns - Timeout in nanoseconds.
* \param buffer_busy - 0 if buffer is idle, all GPU access was completed
* from the same GPU context to the same ip:ip_instance:ring will be executed in
* order.
*
+ * The caller can specify the user fence buffer/location with the fence_info in the
+ * cs_request.The sequence number is returned via the 'seq_no' parameter
+ * in ibs_request structure.
+ *
*
* \param dev - \c [in] Device handle.
* See #amdgpu_device_initialize()
* \param context - \c [in] GPU Context
* \param flags - \c [in] Global submission flags
- * \param ibs_request - \c [in] Pointer to submission requests.
+ * \param ibs_request - \c [in/out] Pointer to submission requests.
* We could submit to the several
* engines/rings simulteniously as
* 'atomic' operation
* \param number_of_requests - \c [in] Number of submission requests
- * \param fences - \c [out] Pointer to array of data to get
- * fences to identify submission
- * requests. Timestamps are valid
- * in this GPU context and could be used
- * to identify/detect completion of
- * submission request
*
* \return 0 on success\n
* <0 - Negative POSIX Error code
int amdgpu_cs_submit(amdgpu_context_handle context,
uint64_t flags,
struct amdgpu_cs_request *ibs_request,
- uint32_t number_of_requests,
- uint64_t *fences);
+ uint32_t number_of_requests);
/**
* Query status of Command Buffer Submission
*
- * \param dev - \c [in] Device handle. See #amdgpu_device_initialize()
* \param fence - \c [in] Structure describing fence to query
+ * \param timeout_ns - \c [in] Timeout value to wait
+ * \param flags - \c [in] Flags for the query
* \param expired - \c [out] If fence expired or not.\n
* 0 – if fence is not expired\n
* !0 - otherwise
*
* \sa amdgpu_cs_submit()
*/
-int amdgpu_cs_query_fence_status(struct amdgpu_cs_query_fence *fence,
+int amdgpu_cs_query_fence_status(struct amdgpu_cs_fence *fence,
+ uint64_t timeout_ns,
+ uint64_t flags,
uint32_t *expired);
+/**
+ * Wait for multiple fences
+ *
+ * \param fences - \c [in] The fence array to wait
+ * \param fence_count - \c [in] The fence count
+ * \param wait_all - \c [in] If true, wait all fences to be signaled,
+ * otherwise, wait at least one fence
+ * \param timeout_ns - \c [in] The timeout to wait, in nanoseconds
+ * \param status - \c [out] '1' for signaled, '0' for timeout
+ * \param first - \c [out] the index of the first signaled fence from @fences
+ *
+ * \return 0 on success
+ * <0 - Negative POSIX Error code
+ *
+ * \note Currently it supports only one amdgpu_device. All fences come from
+ * the same amdgpu_device with the same fd.
+*/
+int amdgpu_cs_wait_fences(struct amdgpu_cs_fence *fences,
+ uint32_t fence_count,
+ bool wait_all,
+ uint64_t timeout_ns,
+ uint32_t *status, uint32_t *first);
+
/*
* Query / Info API
*
struct amdgpu_gds_resource_info *gds_info);
/**
+ * Query information about sensor.
+ *
+ * The return size is query-specific and depends on the "sensor_type"
+ * parameter. No more than "size" bytes is returned.
+ *
+ * \param dev - \c [in] Device handle. See #amdgpu_device_initialize()
+ * \param sensor_type - \c [in] AMDGPU_INFO_SENSOR_*
+ * \param size - \c [in] Size of the returned value.
+ * \param value - \c [out] Pointer to the return value.
+ *
+ * \return 0 on success\n
+ * <0 - Negative POSIX Error code
+ *
+*/
+int amdgpu_query_sensor_info(amdgpu_device_handle dev, unsigned sensor_type,
+ unsigned size, void *value);
+
+/**
* Read a set of consecutive memory-mapped registers.
* Not all registers are allowed to be read by userspace.
*
uint32_t *values);
/**
+ * Flag to request VA address range in the 32bit address space
+*/
+#define AMDGPU_VA_RANGE_32_BIT 0x1
+
+/**
* Allocate virtual address range
*
* \param dev - [in] Device handle. See #amdgpu_device_initialize()
* \param va_base_allocated - \c [out] On return: Allocated VA base to be used
* by client.
* \param va_range_handle - \c [out] On return: Handle assigned to allocation
+ * \param flags - \c [in] flags for special VA range
*
* \return 0 on success\n
* >0 - AMD specific error code\n
uint64_t va_base_alignment,
uint64_t va_base_required,
uint64_t *va_base_allocated,
- amdgpu_va_handle *va_range_handle);
+ amdgpu_va_handle *va_range_handle,
+ uint64_t flags);
/**
* Free previously allocated virtual address range
*/
int amdgpu_va_range_free(amdgpu_va_handle va_range_handle);
+/**
+* Query virtual address range
+*
+* UMD can query GPU VM range supported by each device
+* to initialize its own VAM accordingly.
+*
+* \param dev - [in] Device handle. See #amdgpu_device_initialize()
+* \param type - \c [in] Type of virtual address range
+* \param offset - \c [out] Start offset of virtual address range
+* \param size - \c [out] Size of virtual address range
+*
+* \return 0 on success\n
+* <0 - Negative POSIX Error code
+*
+*/
+
+int amdgpu_va_range_query(amdgpu_device_handle dev,
+ enum amdgpu_gpu_va_range type,
+ uint64_t *start,
+ uint64_t *end);
+
+/**
+ * VA mapping/unmapping for the buffer object
+ *
+ * \param bo - \c [in] BO handle
+ * \param offset - \c [in] Start offset to map
+ * \param size - \c [in] Size to map
+ * \param addr - \c [in] Start virtual address.
+ * \param flags - \c [in] Supported flags for mapping/unmapping
+ * \param ops - \c [in] AMDGPU_VA_OP_MAP or AMDGPU_VA_OP_UNMAP
+ *
+ * \return 0 on success\n
+ * <0 - Negative POSIX Error code
+ *
+*/
+
+int amdgpu_bo_va_op(amdgpu_bo_handle bo,
+ uint64_t offset,
+ uint64_t size,
+ uint64_t addr,
+ uint64_t flags,
+ uint32_t ops);
+
+/**
+ * VA mapping/unmapping for a buffer object or PRT region.
+ *
+ * This is not a simple drop-in extension for amdgpu_bo_va_op; instead, all
+ * parameters are treated "raw", i.e. size is not automatically aligned, and
+ * all flags must be specified explicitly.
+ *
+ * \param dev - \c [in] device handle
+ * \param bo - \c [in] BO handle (may be NULL)
+ * \param offset - \c [in] Start offset to map
+ * \param size - \c [in] Size to map
+ * \param addr - \c [in] Start virtual address.
+ * \param flags - \c [in] Supported flags for mapping/unmapping
+ * \param ops - \c [in] AMDGPU_VA_OP_MAP or AMDGPU_VA_OP_UNMAP
+ *
+ * \return 0 on success\n
+ * <0 - Negative POSIX Error code
+ *
+*/
+
+int amdgpu_bo_va_op_raw(amdgpu_device_handle dev,
+ amdgpu_bo_handle bo,
+ uint64_t offset,
+ uint64_t size,
+ uint64_t addr,
+ uint64_t flags,
+ uint32_t ops);
+
+/**
+ * create semaphore
+ *
+ * \param sem - \c [out] semaphore handle
+ *
+ * \return 0 on success\n
+ * <0 - Negative POSIX Error code
+ *
+*/
+int amdgpu_cs_create_semaphore(amdgpu_semaphore_handle *sem);
+
+/**
+ * signal semaphore
+ *
+ * \param context - \c [in] GPU Context
+ * \param ip_type - \c [in] Hardware IP block type = AMDGPU_HW_IP_*
+ * \param ip_instance - \c [in] Index of the IP block of the same type
+ * \param ring - \c [in] Specify ring index of the IP
+ * \param sem - \c [in] semaphore handle
+ *
+ * \return 0 on success\n
+ * <0 - Negative POSIX Error code
+ *
+*/
+int amdgpu_cs_signal_semaphore(amdgpu_context_handle ctx,
+ uint32_t ip_type,
+ uint32_t ip_instance,
+ uint32_t ring,
+ amdgpu_semaphore_handle sem);
+
+/**
+ * wait semaphore
+ *
+ * \param context - \c [in] GPU Context
+ * \param ip_type - \c [in] Hardware IP block type = AMDGPU_HW_IP_*
+ * \param ip_instance - \c [in] Index of the IP block of the same type
+ * \param ring - \c [in] Specify ring index of the IP
+ * \param sem - \c [in] semaphore handle
+ *
+ * \return 0 on success\n
+ * <0 - Negative POSIX Error code
+ *
+*/
+int amdgpu_cs_wait_semaphore(amdgpu_context_handle ctx,
+ uint32_t ip_type,
+ uint32_t ip_instance,
+ uint32_t ring,
+ amdgpu_semaphore_handle sem);
+
+/**
+ * destroy semaphore
+ *
+ * \param sem - \c [in] semaphore handle
+ *
+ * \return 0 on success\n
+ * <0 - Negative POSIX Error code
+ *
+*/
+int amdgpu_cs_destroy_semaphore(amdgpu_semaphore_handle sem);
+
+/**
+ * Get the ASIC marketing name
+ *
+ * \param dev - \c [in] Device handle. See #amdgpu_device_initialize()
+ *
+ * \return the constant string of the marketing name
+ * "NULL" means the ASIC is not found
+*/
+const char *amdgpu_get_marketing_name(amdgpu_device_handle dev);
+
+#ifdef __cplusplus
+}
+#endif
+
#endif /* #ifdef _AMDGPU_H_ */