NVIDIA/open-gpu-kernel-modules
1
0
Fork 0
mirror of https://github.com/NVIDIA/open-gpu-kernel-modules.git synced 2026-02-28 19:03:58 +00:00
Code Issues Packages Projects Releases Wiki Activity

535.43.08

Browse Source
This commit is contained in:
russellcnv
2023-08-17 13:03:06 -07:00
parent eb5c7665a1
commit 18b7303c54
378 changed files with 124043 additions and 161949 deletions
Download Patch File Download Diff File Expand all files Collapse all files

42
src/common/sdk/nvidia/inc/ctrl/ctrl0050.h
View File

@@ -72,6 +72,19 @@
* identity mapped. To use this feature, users need to pass in the
* hVaspace with identity mapped addresses for the entire memory during
* construct.
* PIPELINED
* This flag allows the copy/memset operation to be pipelined with previous dma operations on the same channel
* It means that its reads/writes are allowed happen before writes of preceding operations are tlb-acked
* The flag can be useful when dealing with non-inersecting async operations,
* but it can result in races when 2 async CE operations target the same allocation, and the second operation uses the flag
* Race example:
* 1. async copy A -> B
* 2. pipelined copy B -> C
* Here copy 2 can read B before copy finishes writing it, which will result in C containing invalid data
* Technical details:
* By default, first _LAUNCH_DMA method of a CE operation is marked has _TRANSFER_TYPE_NON_PIPELINED, which the flag overrides
* Subsequent _LAUNCH_DMA methods belonging to the same operation use _TRANSFER_TYPE_PIPELINED, as each of these methods should
* target different addresses
*
* submittedWorkId [OUT]
* The work submission token users can poll on to wait for work
@@ -79,8 +92,9 @@
*/
#define NV0050_CTRL_MEMSET_FLAGS_DEFAULT 0
#define NV0050_CTRL_MEMSET_FLAGS_ASYNC NVBIT(0)
#define NV0050_CTRL_MEMSET_FLAGS_VIRTUAL NVBIT(1)
#define NV0050_CTRL_MEMSET_FLAGS_ASYNC NVBIT(0)
#define NV0050_CTRL_MEMSET_FLAGS_VIRTUAL NVBIT(1)
#define NV0050_CTRL_MEMSET_FLAGS_PIPELINED NVBIT(2)
#define NV0050_CTRL_CMD_MEMSET (0x500101U) /* finn: Evaluated from "(FINN_NV_CE_UTILS_UTILS_INTERFACE_ID << 8) | NV0050_CTRL_MEMSET_PARAMS_MESSAGE_ID" */
@@ -98,7 +112,7 @@ typedef struct NV0050_CTRL_MEMSET_PARAMS {
/*
* NV0050_CTRL_CMD_MEMCOPY
*
* Copies from a source memoryto ssdestination memory and releases a semaphore
* Copies from a source memoryto ssdestination memory and releases a semaphore
* on completion
*
* hDstMemory [IN]
@@ -131,6 +145,19 @@ typedef struct NV0050_CTRL_MEMSET_PARAMS {
* identity mapped. To use this feature, users need to pass in the
* hVaspace with identity mapped addresses for the entire memory during
* construct.
* PIPELINED
* This flag allows the copy/memset operation to be pipelined with previous dma operations on the same channel
* It means that its reads/writes are allowed happen before writes of preceding operations are tlb-acked
* The flag can be useful when dealing with non-inersecting async operations,
* but it can result in races when 2 async CE operations target the same allocation, and the second operation uses the flag
* Race example:
* 1. async copy A -> B
* 2. pipelined copy B -> C
* Here copy 2 can read B before copy finishes writing it, which will result in C containing invalid data
* Technical details:
* By default, first _LAUNCH_DMA method of a CE operation is marked has _TRANSFER_TYPE_NON_PIPELINED, which the flag overrides
* Subsequent _LAUNCH_DMA methods belonging to the same operation use _TRANSFER_TYPE_PIPELINED, as each of these methods should
* target different addresses
*
* submittedWorkId [OUT]
* The work submission token users can poll on to wait for work
@@ -138,8 +165,9 @@ typedef struct NV0050_CTRL_MEMSET_PARAMS {
*/
#define NV0050_CTRL_MEMCOPY_FLAGS_DEFAULT 0
#define NV0050_CTRL_MEMCOPY_FLAGS_ASYNC NVBIT(1)
#define NV0050_CTRL_MEMCOPY_FLAGS_VIRTUAL NVBIT(2)
#define NV0050_CTRL_MEMCOPY_FLAGS_ASYNC NVBIT(0)
#define NV0050_CTRL_MEMCOPY_FLAGS_VIRTUAL NVBIT(1)
#define NV0050_CTRL_MEMCOPY_FLAGS_PIPELINED NVBIT(2)
#define NV0050_CTRL_CMD_MEMCOPY (0x500102U) /* finn: Evaluated from "(FINN_NV_CE_UTILS_UTILS_INTERFACE_ID << 8 | NV0050_CTRL_MEMCOPY_PARAMS_MESSAGE_ID)" */
@@ -157,12 +185,12 @@ typedef struct NV0050_CTRL_MEMCOPY_PARAMS {
/*
* NV0050_CTRL_CMD_CHECK_PROGRESS
*
*
* Check if a previously submitted work item has been completed by HW.
*
* submittedWorkId [IN]
* The work submission token users can poll on to wait for work
* completed by CE.
* completed by CE.
*
*/
#define NV0050_CTRL_CHECK_PROGRESS_RESULT_DEFAULT 0