add nosva support #682
Open
add nosva support #682
Conversation
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Remove nosva limitation to permit nosva run Signed-off-by: Zhangfei Gao <zhangfei.gao@linaro.org>
Signed-off-by: Zhangfei Gao <zhangfei.gao@linaro.org>
gaozhangfei
force-pushed
the
nosva-patch-7.24
branch
4 times, most recently
from
8e83665 to
589e0af
Compare
Add new apis: wd_blkpool_new; wd_blkpool_delete; wd_blkpool_phy; wd_blkpool_alloc; wd_blkpool_free; wd_blkpool_setup; wd_blkpool_destroy_mem; wd_blkpool_create_sglpool; wd_blkpool_destroy_sglpool; App only use two apis after setup blkpool wd_blkpool_alloc; wd_blkpool_free; Signed-off-by: Zhangfei Gao <zhangfei.gao@linaro.org>
When init, all ctxs will call wd_blkpool_new, Only nosva case will get a pointer, while sva case get NULL. When uninit, delete blkpool and related resources. Signed-off-by: Zhangfei Gao <zhangfei.gao@linaro.org>
Add api wd_comp_setup_blkpool. Other alg.c will need wd_xxx_setup_blkpool as well. The reason is app does not know ctx. It will setup blkpool for ctx[0] and sglpool for sgl mode. The blkpool will be used by app and driver. App need call wd_xxx_setup_blkpool for user pointer mode and sgl mode. The returned blkpool will be used for wd_blkpool_alloc/free. Alloc_sess will call wd_xxx_setup_blkpool if it is not called by app. Then uadk library will alloc blkpool and memcpy to user memory, with poorer performance. The driver will translate va to pa when configure register. Signed-off-by: Zhangfei Gao <zhangfei.gao@linaro.org>
Support sync: mempcy, user pointer and sgl case Support async: memcpy, user pointer For flat memory: If user call wd_xxx_setup_blkpool, will use user pointer mode. uadk directly use pointer for app, assume it is continuous memory and translate va to pa when configure register Otherwise, alloc_sess will setup blkpool and use memcpy mode. wd_comp alloc continuous memory for hardware and memcpy from src pointer and memcpy results to dst pointer For sgl memory: App has to call wd_xxx_setup_blkpool. The wd_datalist.data has to use continuous memory Signed-off-by: Zhangfei Gao <zhangfei.gao@linaro.org>
sync: mempcy mode ./uadk_tool/uadk_tool benchmark --alg zlib --mode sva \ --opt 0 --sync --pktlen 1024 user pointer mode, --user ./uadk_tool/uadk_tool benchmark --alg zlib --mode sva \ --user --opt 0 --sync --pktlen 1024 sgl mode, --sgl ./uadk_tool/uadk_tool benchmark --alg zlib --mode sva \ --sgl --opt 0 --sync --pktlen 1024 async: memcpy mode ./uadk_tool/uadk_tool benchmark --alg zlib --mode sva \ --opt 0 --async --pktlen 1024 user pointer mode, --user ./uadk_tool/uadk_tool benchmark --alg zlib --mode sva \ --user --opt 0 --async --pktlen 1024 sgl mode, --sgl ./uadk_tool/uadk_tool benchmark --alg zlib --mode sva \ --sgl --opt 0 --async --pktlen 1024 Signed-off-by: Zhangfei Gao <zhangfei.gao@linaro.org>
This patch addresses the SGL (Scatter-Gather List) next pointer handling issue in both SVA and non-SVA scenarios. Background: - A single SGL can typically cover 255 SGEs (Scatter-Gather Elements) * 8MB - Multi-SGL cases require proper next pointer handling Implementation details: 1. In SVA (Shared Virtual Addressing) case: - hisi_sgl.next_dma serves both hardware and CPU access - Address translation is handled automatically by the IOMMU 2. In non-SVA case: - hisi_sgl.next_dma contains DMA address for hardware use only - CPU cannot directly use next_dma to access SGL members - Solution: Reuse one pad1 field as hisi_sgl.next for CPU access The modification maintains the original 64B hardware SGL header size while adding proper CPU-accessible pointer support. This ensures correct operation in both single and multi-SGL scenarios. Signed-off-by: Zhangfei Gao <zhangfei.gao@linaro.org>
test: digest: uadk_tool benchmark --alg aes-128-ecb --mode sva --opt 0 \ --sync --pktlen 1024 --seconds 1 --multi 1 --thread 1 uadk_tool benchmark --alg aes-128-ecb --mode sva --opt 0 \ --async --pktlen 1024 --seconds 1 --multi 1 --thread 1 cipher: uadk_tool benchmark --alg sm4-128-cbc --mode sva --opt 0 \ --sync --pktlen 1024 --seconds 1 --multi 1 --thread 1 uadk_tool benchmark --alg sm4-128-cbc --mode sva --opt 0 \ --async --pktlen 1024 --seconds 1 --multi 1 --thread 1 Signed-off-by: Zhangfei Gao <zhangfei.gao@linaro.org>
Signed-off-by: Zhangfei Gao <zhangfei.gao@linaro.org>
gaozhangfei
force-pushed
the
nosva-patch-7.24
branch
from
e0b81ee to
208c637
Compare
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
2 participants
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
Add nosva support
design doc
【腾讯文档】nosva
https://docs.qq.com/doc/DRVBtbVh6TkdaTFVB
++ echo 'test zip, sync'
++ ./uadk_tool/uadk_tool benchmark --alg zlib --mode sva --opt 0 --sync --pktlen 1024
algname: length: perf: iops: CPU_rate:
zlib 1024Bytes 316882.67KiB/s 316.9Kops 290.67%
++ ./uadk_tool/uadk_tool benchmark --alg zlib --mode sva --user --opt 0 --sync --pktlen 1024
algname: length: perf: iops: CPU_rate:
zlib 1024Bytes 498163.67KiB/s 498.2Kops 299.67%
++ ./uadk_tool/uadk_tool benchmark --alg zlib --mode sva --sgl --opt 0 --sync --pktlen 1024
algname: length: perf: iops: CPU_rate:
zlib 1024Bytes 417162.33KiB/s 417.2Kops 299.67%
++ echo 'test zip, async'
++ ./uadk_tool/uadk_tool benchmark --alg zlib --mode sva --opt 0 --async --pktlen 1024
algname: length: perf: iops: CPU_rate:
zlib 1024Bytes 415351.67KiB/s 415.4Kops 312.33%
++ ./uadk_tool/uadk_tool benchmark --alg zlib --mode sva --user --opt 0 --async --pktlen 1024
algname: length: perf: iops: CPU_rate:
zlib 1024Bytes 681221.00KiB/s 681.2Kops 376.00%
++ ./uadk_tool/uadk_tool benchmark --alg zlib --mode sva --sgl --opt 0 --async --pktlen 1024
algname: length: perf: iops: CPU_rate:
zlib 1024Bytes 491375.00KiB/s 491.4Kops 320.33%
digest:
uadk_tool benchmark --alg aes-128-ecb --mode sva --opt 0
--sync --pktlen 1024 --seconds 1 --multi 1 --thread 1
uadk_tool benchmark --alg aes-128-ecb --mode sva --opt 0
--async --pktlen 1024 --seconds 1 --multi 1 --thread 1