Message ID | 20230616-sparse-flash-fix-v1-1-6bafeacc567b@baylibre.com |
---|---|
State | Superseded |
Delegated to: | Tom Rini |
Headers | show |
Series | lib: sparse: allocate blkcnt instead of arbitrary small number | expand |
在 2023/6/16 21:26, Mattijs Korpershoek 写道: > Commit 62649165cb02 ("lib: sparse: Make CHUNK_TYPE_RAW buffer aligned") > fixed cache alignment for systems with a D-CACHE. > > However it introduced some performance regressions [1] on system > flashing huge images, such as Android. > > On AM62x SK EVM, we also observe such performance penalty: > Sending sparse 'super' 1/2 (768793 KB) OKAY [ 23.954s] > Writing 'super' OKAY [ 75.926s] > Sending sparse 'super' 2/2 (629819 KB) OKAY [ 19.641s] > Writing 'super' OKAY [ 62.849s] > Finished. Total time: 182.474s > > The reason for this is that we use an arbitrary small buffer > (info->blksz * 100) for transferring. > > Fix it by using a bigger buffer (info->blksz * blkcnt) as suggested in > the original's patch review [2]. > > With this patch, performance impact is mitigated: > Sending sparse 'super' 1/2 (768793 KB) OKAY [ 24.006s] > Writing 'super' OKAY [ 15.920s] > Sending sparse 'super' 2/2 (629819 KB) OKAY [ 19.651s] > Writing 'super' OKAY [ 14.665s] > Finished. Total time: 74.346s > > [1] https://lore.kernel.org/r/20221118121323.4009193-1-gary.bisson@boundarydevices.com > [2] https://lore.kernel.org/r/all/43e4c17c-4483-ec8e-f843-9b4c5569bd18@seco.com/ > > Fixes: 62649165cb02 ("lib: sparse: Make CHUNK_TYPE_RAW buffer aligned") > Signed-off-by: Mattijs Korpershoek <mkorpershoek@baylibre.com> > --- > lib/image-sparse.c | 2 +- > 1 file changed, 1 insertion(+), 1 deletion(-) > > diff --git a/lib/image-sparse.c b/lib/image-sparse.c > index 5ec0f94ab3eb..25aed0604192 100644 > --- a/lib/image-sparse.c > +++ b/lib/image-sparse.c > @@ -55,7 +55,7 @@ static lbaint_t write_sparse_chunk_raw(struct sparse_storage *info, > void *data, > char *response) > { > - lbaint_t n = blkcnt, write_blks, blks = 0, aligned_buf_blks = 100; > + lbaint_t n = blkcnt, write_blks, blks = 0, aligned_buf_blks = blkcnt; Hi: It's a good point that this code report the performance was affected by write large small mmc blks, not memory copy. And I can not make sure whether memalign can always alloc such huge memory when we change the aligned_buf_blks to blkcnt. Could you please set aligned_buf_blks to FASTBOOT_MAX_BLK_WRITE(16384) and test again? > uint32_t *aligned_buf = NULL; > > if (CONFIG_IS_ENABLED(SYS_DCACHE_OFF)) { > > --- > base-commit: 2f4664f5c3edc55b18d8906f256a4c8e303243c0 > change-id: 20230616-sparse-flash-fix-9c2852aa8d16 > > Best regards,
Hi Qianfan, Thank you for your review. On lun., juin 19, 2023 at 14:19, qianfan <qianfanguijin@163.com> wrote: > 在 2023/6/16 21:26, Mattijs Korpershoek 写道: >> Commit 62649165cb02 ("lib: sparse: Make CHUNK_TYPE_RAW buffer aligned") >> fixed cache alignment for systems with a D-CACHE. >> >> However it introduced some performance regressions [1] on system >> flashing huge images, such as Android. >> >> On AM62x SK EVM, we also observe such performance penalty: >> Sending sparse 'super' 1/2 (768793 KB) OKAY [ 23.954s] >> Writing 'super' OKAY [ 75.926s] >> Sending sparse 'super' 2/2 (629819 KB) OKAY [ 19.641s] >> Writing 'super' OKAY [ 62.849s] >> Finished. Total time: 182.474s >> >> The reason for this is that we use an arbitrary small buffer >> (info->blksz * 100) for transferring. >> >> Fix it by using a bigger buffer (info->blksz * blkcnt) as suggested in >> the original's patch review [2]. >> >> With this patch, performance impact is mitigated: >> Sending sparse 'super' 1/2 (768793 KB) OKAY [ 24.006s] >> Writing 'super' OKAY [ 15.920s] >> Sending sparse 'super' 2/2 (629819 KB) OKAY [ 19.651s] >> Writing 'super' OKAY [ 14.665s] >> Finished. Total time: 74.346s >> >> [1] https://lore.kernel.org/r/20221118121323.4009193-1-gary.bisson@boundarydevices.com >> [2] https://lore.kernel.org/r/all/43e4c17c-4483-ec8e-f843-9b4c5569bd18@seco.com/ >> >> Fixes: 62649165cb02 ("lib: sparse: Make CHUNK_TYPE_RAW buffer aligned") >> Signed-off-by: Mattijs Korpershoek <mkorpershoek@baylibre.com> >> --- >> lib/image-sparse.c | 2 +- >> 1 file changed, 1 insertion(+), 1 deletion(-) >> >> diff --git a/lib/image-sparse.c b/lib/image-sparse.c >> index 5ec0f94ab3eb..25aed0604192 100644 >> --- a/lib/image-sparse.c >> +++ b/lib/image-sparse.c >> @@ -55,7 +55,7 @@ static lbaint_t write_sparse_chunk_raw(struct sparse_storage *info, >> void *data, >> char *response) >> { >> - lbaint_t n = blkcnt, write_blks, blks = 0, aligned_buf_blks = 100; >> + lbaint_t n = blkcnt, write_blks, blks = 0, aligned_buf_blks = blkcnt; > Hi: > > It's a good point that this code report the performance was affected by > write large small > mmc blks, not memory copy. I believe memory copy also affects performance, but in my case, it has less impact than small mmc blks. With 62649165cb02 reverted: Sending sparse 'super' 1/2 (768793 KB) OKAY [ 23.947s] Writing 'super' OKAY [ 12.983s] Sending sparse 'super' 2/2 (629819 KB) OKAY [ 19.600s] Writing 'super' OKAY [ 12.796s] Finished. Total time: 69.430s With aligned_buf_blks = blkcnt: Sending sparse 'super' 1/2 (768793 KB) OKAY [ 24.072s] Writing 'super' OKAY [ 16.177s] Sending sparse 'super' 2/2 (629819 KB) OKAY [ 19.681s] Writing 'super' OKAY [ 14.845s] Finished. Total time: 74.919s > > And I can not make sure whether memalign can always alloc such huge > memory when we change the > aligned_buf_blks to blkcnt. Could you clarify the concern here? I've dumped blkcnt for my board (AM62x SK EVK) and the biggest blkcnt I found was: 131072 With info->blksz = 512, this gives me: 512 * 131072 = 67108864 Which is a memalign (memory alloc) of 64MB. Is 64MB really that big? (I don't realize it's that much) > > Could you please set aligned_buf_blks to FASTBOOT_MAX_BLK_WRITE(16384) > and test again? With aligned_buf_blks = FASTBOOT_MAX_BLK_WRITE(16384): Sending sparse 'super' 1/2 (768793 KB) OKAY [ 23.912s] Writing 'super' OKAY [ 15.780s] Sending sparse 'super' 2/2 (629819 KB) OKAY [ 19.581s] Writing 'super' OKAY [ 17.192s] Finished. Total time: 76.569s So using FASTBOOT_MAX_BLK_WRITE is slightly worse than using blkcnt. But allocations (for blksz = 512) are smaller: 8MB instead of 64MB in my example. I can spin up a v2 with FASTBOOT_MAX_BLK_WRITE but i'm waiting a little more feedback before doing so. >> uint32_t *aligned_buf = NULL; >> >> if (CONFIG_IS_ENABLED(SYS_DCACHE_OFF)) { >> >> --- >> base-commit: 2f4664f5c3edc55b18d8906f256a4c8e303243c0 >> change-id: 20230616-sparse-flash-fix-9c2852aa8d16 >> >> Best regards,
On lun., juin 19, 2023 at 10:21, Mattijs Korpershoek <mkorpershoek@baylibre.com> wrote: > Hi Qianfan, > > Thank you for your review. > > On lun., juin 19, 2023 at 14:19, qianfan <qianfanguijin@163.com> wrote: > >> 在 2023/6/16 21:26, Mattijs Korpershoek 写道: >>> Commit 62649165cb02 ("lib: sparse: Make CHUNK_TYPE_RAW buffer aligned") >>> fixed cache alignment for systems with a D-CACHE. >>> >>> However it introduced some performance regressions [1] on system >>> flashing huge images, such as Android. >>> >>> On AM62x SK EVM, we also observe such performance penalty: >>> Sending sparse 'super' 1/2 (768793 KB) OKAY [ 23.954s] >>> Writing 'super' OKAY [ 75.926s] >>> Sending sparse 'super' 2/2 (629819 KB) OKAY [ 19.641s] >>> Writing 'super' OKAY [ 62.849s] >>> Finished. Total time: 182.474s >>> >>> The reason for this is that we use an arbitrary small buffer >>> (info->blksz * 100) for transferring. >>> >>> Fix it by using a bigger buffer (info->blksz * blkcnt) as suggested in >>> the original's patch review [2]. >>> >>> With this patch, performance impact is mitigated: >>> Sending sparse 'super' 1/2 (768793 KB) OKAY [ 24.006s] >>> Writing 'super' OKAY [ 15.920s] >>> Sending sparse 'super' 2/2 (629819 KB) OKAY [ 19.651s] >>> Writing 'super' OKAY [ 14.665s] >>> Finished. Total time: 74.346s >>> >>> [1] https://lore.kernel.org/r/20221118121323.4009193-1-gary.bisson@boundarydevices.com >>> [2] https://lore.kernel.org/r/all/43e4c17c-4483-ec8e-f843-9b4c5569bd18@seco.com/ >>> >>> Fixes: 62649165cb02 ("lib: sparse: Make CHUNK_TYPE_RAW buffer aligned") >>> Signed-off-by: Mattijs Korpershoek <mkorpershoek@baylibre.com> >>> --- >>> lib/image-sparse.c | 2 +- >>> 1 file changed, 1 insertion(+), 1 deletion(-) >>> >>> diff --git a/lib/image-sparse.c b/lib/image-sparse.c >>> index 5ec0f94ab3eb..25aed0604192 100644 >>> --- a/lib/image-sparse.c >>> +++ b/lib/image-sparse.c >>> @@ -55,7 +55,7 @@ static lbaint_t write_sparse_chunk_raw(struct sparse_storage *info, >>> void *data, >>> char *response) >>> { >>> - lbaint_t n = blkcnt, write_blks, blks = 0, aligned_buf_blks = 100; >>> + lbaint_t n = blkcnt, write_blks, blks = 0, aligned_buf_blks = blkcnt; >> Hi: >> >> It's a good point that this code report the performance was affected by >> write large small >> mmc blks, not memory copy. > > I believe memory copy also affects performance, but in my case, > it has less impact than small mmc blks. > > With 62649165cb02 reverted: > Sending sparse 'super' 1/2 (768793 KB) OKAY [ 23.947s] > Writing 'super' OKAY [ 12.983s] > Sending sparse 'super' 2/2 (629819 KB) OKAY [ 19.600s] > Writing 'super' OKAY [ 12.796s] > Finished. Total time: 69.430s > > With aligned_buf_blks = blkcnt: > Sending sparse 'super' 1/2 (768793 KB) OKAY [ 24.072s] > Writing 'super' OKAY [ 16.177s] > Sending sparse 'super' 2/2 (629819 KB) OKAY [ 19.681s] > Writing 'super' OKAY [ 14.845s] > Finished. Total time: 74.919s > >> >> And I can not make sure whether memalign can always alloc such huge >> memory when we change the >> aligned_buf_blks to blkcnt. > > Could you clarify the concern here? I've dumped blkcnt for my board > (AM62x SK EVK) and the biggest blkcnt I found was: 131072 > > With info->blksz = 512, this gives me: 512 * 131072 = 67108864 > > Which is a memalign (memory alloc) of 64MB. Is 64MB really that big? (I > don't realize it's that much) > >> >> Could you please set aligned_buf_blks to FASTBOOT_MAX_BLK_WRITE(16384) >> and test again? > > With aligned_buf_blks = FASTBOOT_MAX_BLK_WRITE(16384): > Sending sparse 'super' 1/2 (768793 KB) OKAY [ 23.912s] > Writing 'super' OKAY [ 15.780s] > Sending sparse 'super' 2/2 (629819 KB) OKAY [ 19.581s] > Writing 'super' OKAY [ 17.192s] > Finished. Total time: 76.569s > > So using FASTBOOT_MAX_BLK_WRITE is slightly worse than using blkcnt. > But allocations (for blksz = 512) are smaller: 8MB instead of 64MB in my example. > > I can spin up a v2 with FASTBOOT_MAX_BLK_WRITE but i'm waiting a little > more feedback before doing so. Hi Marek, Tom, What's your take on this ? Can we keep blkcnt or should I respin using FASTBOOT_MAX_BLK_WRITE ? I have also tested this on VIM3, on U-Boot 2023.07-rc6-00003-g923de765ee1a: Sending sparse 'super' 1/13 (114684 KB) OKAY [ 5.442s] Writing 'super' OKAY [ 5.791s] Sending sparse 'super' 2/13 (114684 KB) OKAY [ 5.706s] Writing 'super' OKAY [ 5.607s] Sending sparse 'super' 3/13 (114684 KB) OKAY [ 5.468s] Writing 'super' OKAY [ 5.835s] Sending sparse 'super' 4/13 (114684 KB) OKAY [ 5.703s] Writing 'super' OKAY [ 5.618s] Sending sparse 'super' 5/13 (114684 KB) OKAY [ 6.176s] Writing 'super' OKAY [ 5.421s] Sending sparse 'super' 6/13 (104176 KB) OKAY [ 5.204s] Writing 'super' OKAY [ 5.199s] Sending sparse 'super' 7/13 (108856 KB) OKAY [ 5.456s] Writing 'super' OKAY [ 5.290s] Sending sparse 'super' 8/13 (114684 KB) OKAY [ 6.122s] Writing 'super' OKAY [ 5.838s] Sending sparse 'super' 9/13 (114684 KB) OKAY [ 5.951s] Writing 'super' OKAY [ 5.857s] Sending sparse 'super' 10/13 (100980 KB) OKAY [ 4.902s] Writing 'super' OKAY [ 4.749s] Sending sparse 'super' 11/13 (114681 KB) OKAY [ 6.041s] Writing 'super' OKAY [ 5.779s] Sending sparse 'super' 12/13 (107212 KB) OKAY [ 5.174s] Writing 'super' OKAY [ 6.587s] Sending sparse 'super' 13/13 (71496 KB) OKAY [ 3.717s] Writing 'super' OKAY [ 3.744s] Finished. Total time: 142.578s With this patch: Sending sparse 'super' 1/13 (114684 KB) OKAY [ 7.149s] Writing 'super' OKAY [ 1.639s] Sending sparse 'super' 2/13 (114684 KB) OKAY [ 6.993s] Writing 'super' OKAY [ 1.713s] Sending sparse 'super' 3/13 (114684 KB) OKAY [ 7.029s] Writing 'super' OKAY [ 1.107s] Sending sparse 'super' 4/13 (114684 KB) OKAY [ 7.027s] Writing 'super' OKAY [ 0.162s] Sending sparse 'super' 5/13 (114684 KB) OKAY [ 6.930s] Writing 'super' OKAY [ 1.643s] Sending sparse 'super' 6/13 (104176 KB) OKAY [ 6.253s] Writing 'super' OKAY [ 2.348s] Sending sparse 'super' 7/13 (108856 KB) OKAY [ 6.346s] Writing 'super' OKAY [ 0.723s] Sending sparse 'super' 8/13 (114684 KB) OKAY [ 6.715s] Writing 'super' OKAY [ 2.848s] Sending sparse 'super' 9/13 (114684 KB) OKAY [ 6.888s] Writing 'super' OKAY [ 1.928s] Sending sparse 'super' 10/13 (100980 KB) OKAY [ 5.979s] Writing 'super' OKAY [ 1.178s] Sending sparse 'super' 11/13 (114681 KB) OKAY [ 6.822s] Writing 'super' OKAY [ 2.652s] Sending sparse 'super' 12/13 (107212 KB) OKAY [ 6.414s] Writing 'super' OKAY [ 5.109s] Sending sparse 'super' 13/13 (71496 KB) OKAY [ 4.238s] Writing 'super' OKAY [ 0.252s] Finished. Total time: 108.151s It's probably too late for v2023.07 to pick this up but can we consider taking it for next? Thanks a lot Mattijs > >>> uint32_t *aligned_buf = NULL; >>> >>> if (CONFIG_IS_ENABLED(SYS_DCACHE_OFF)) { >>> >>> --- >>> base-commit: 2f4664f5c3edc55b18d8906f256a4c8e303243c0 >>> change-id: 20230616-sparse-flash-fix-9c2852aa8d16 >>> >>> Best regards,
On Thu, Jul 06, 2023 at 11:43:13AM +0200, Mattijs Korpershoek wrote: > On lun., juin 19, 2023 at 10:21, Mattijs Korpershoek <mkorpershoek@baylibre.com> wrote: > > > Hi Qianfan, > > > > Thank you for your review. > > > > On lun., juin 19, 2023 at 14:19, qianfan <qianfanguijin@163.com> wrote: > > > >> 在 2023/6/16 21:26, Mattijs Korpershoek 写道: > >>> Commit 62649165cb02 ("lib: sparse: Make CHUNK_TYPE_RAW buffer aligned") > >>> fixed cache alignment for systems with a D-CACHE. > >>> > >>> However it introduced some performance regressions [1] on system > >>> flashing huge images, such as Android. > >>> > >>> On AM62x SK EVM, we also observe such performance penalty: > >>> Sending sparse 'super' 1/2 (768793 KB) OKAY [ 23.954s] > >>> Writing 'super' OKAY [ 75.926s] > >>> Sending sparse 'super' 2/2 (629819 KB) OKAY [ 19.641s] > >>> Writing 'super' OKAY [ 62.849s] > >>> Finished. Total time: 182.474s > >>> > >>> The reason for this is that we use an arbitrary small buffer > >>> (info->blksz * 100) for transferring. > >>> > >>> Fix it by using a bigger buffer (info->blksz * blkcnt) as suggested in > >>> the original's patch review [2]. > >>> > >>> With this patch, performance impact is mitigated: > >>> Sending sparse 'super' 1/2 (768793 KB) OKAY [ 24.006s] > >>> Writing 'super' OKAY [ 15.920s] > >>> Sending sparse 'super' 2/2 (629819 KB) OKAY [ 19.651s] > >>> Writing 'super' OKAY [ 14.665s] > >>> Finished. Total time: 74.346s > >>> > >>> [1] https://lore.kernel.org/r/20221118121323.4009193-1-gary.bisson@boundarydevices.com > >>> [2] https://lore.kernel.org/r/all/43e4c17c-4483-ec8e-f843-9b4c5569bd18@seco.com/ > >>> > >>> Fixes: 62649165cb02 ("lib: sparse: Make CHUNK_TYPE_RAW buffer aligned") > >>> Signed-off-by: Mattijs Korpershoek <mkorpershoek@baylibre.com> > >>> --- > >>> lib/image-sparse.c | 2 +- > >>> 1 file changed, 1 insertion(+), 1 deletion(-) > >>> > >>> diff --git a/lib/image-sparse.c b/lib/image-sparse.c > >>> index 5ec0f94ab3eb..25aed0604192 100644 > >>> --- a/lib/image-sparse.c > >>> +++ b/lib/image-sparse.c > >>> @@ -55,7 +55,7 @@ static lbaint_t write_sparse_chunk_raw(struct sparse_storage *info, > >>> void *data, > >>> char *response) > >>> { > >>> - lbaint_t n = blkcnt, write_blks, blks = 0, aligned_buf_blks = 100; > >>> + lbaint_t n = blkcnt, write_blks, blks = 0, aligned_buf_blks = blkcnt; > >> Hi: > >> > >> It's a good point that this code report the performance was affected by > >> write large small > >> mmc blks, not memory copy. > > > > I believe memory copy also affects performance, but in my case, > > it has less impact than small mmc blks. > > > > With 62649165cb02 reverted: > > Sending sparse 'super' 1/2 (768793 KB) OKAY [ 23.947s] > > Writing 'super' OKAY [ 12.983s] > > Sending sparse 'super' 2/2 (629819 KB) OKAY [ 19.600s] > > Writing 'super' OKAY [ 12.796s] > > Finished. Total time: 69.430s > > > > With aligned_buf_blks = blkcnt: > > Sending sparse 'super' 1/2 (768793 KB) OKAY [ 24.072s] > > Writing 'super' OKAY [ 16.177s] > > Sending sparse 'super' 2/2 (629819 KB) OKAY [ 19.681s] > > Writing 'super' OKAY [ 14.845s] > > Finished. Total time: 74.919s > > > >> > >> And I can not make sure whether memalign can always alloc such huge > >> memory when we change the > >> aligned_buf_blks to blkcnt. > > > > Could you clarify the concern here? I've dumped blkcnt for my board > > (AM62x SK EVK) and the biggest blkcnt I found was: 131072 > > > > With info->blksz = 512, this gives me: 512 * 131072 = 67108864 > > > > Which is a memalign (memory alloc) of 64MB. Is 64MB really that big? (I > > don't realize it's that much) > > > >> > >> Could you please set aligned_buf_blks to FASTBOOT_MAX_BLK_WRITE(16384) > >> and test again? > > > > With aligned_buf_blks = FASTBOOT_MAX_BLK_WRITE(16384): > > Sending sparse 'super' 1/2 (768793 KB) OKAY [ 23.912s] > > Writing 'super' OKAY [ 15.780s] > > Sending sparse 'super' 2/2 (629819 KB) OKAY [ 19.581s] > > Writing 'super' OKAY [ 17.192s] > > Finished. Total time: 76.569s > > > > So using FASTBOOT_MAX_BLK_WRITE is slightly worse than using blkcnt. > > But allocations (for blksz = 512) are smaller: 8MB instead of 64MB in my example. > > > > I can spin up a v2 with FASTBOOT_MAX_BLK_WRITE but i'm waiting a little > > more feedback before doing so. > > Hi Marek, Tom, > > What's your take on this ? Can we keep blkcnt or should I respin using > FASTBOOT_MAX_BLK_WRITE ? > > I have also tested this on VIM3, on > U-Boot 2023.07-rc6-00003-g923de765ee1a: > > Sending sparse 'super' 1/13 (114684 KB) OKAY [ 5.442s] > Writing 'super' OKAY [ 5.791s] > Sending sparse 'super' 2/13 (114684 KB) OKAY [ 5.706s] > Writing 'super' OKAY [ 5.607s] > Sending sparse 'super' 3/13 (114684 KB) OKAY [ 5.468s] > Writing 'super' OKAY [ 5.835s] > Sending sparse 'super' 4/13 (114684 KB) OKAY [ 5.703s] > Writing 'super' OKAY [ 5.618s] > Sending sparse 'super' 5/13 (114684 KB) OKAY [ 6.176s] > Writing 'super' OKAY [ 5.421s] > Sending sparse 'super' 6/13 (104176 KB) OKAY [ 5.204s] > Writing 'super' OKAY [ 5.199s] > Sending sparse 'super' 7/13 (108856 KB) OKAY [ 5.456s] > Writing 'super' OKAY [ 5.290s] > Sending sparse 'super' 8/13 (114684 KB) OKAY [ 6.122s] > Writing 'super' OKAY [ 5.838s] > Sending sparse 'super' 9/13 (114684 KB) OKAY [ 5.951s] > Writing 'super' OKAY [ 5.857s] > Sending sparse 'super' 10/13 (100980 KB) OKAY [ 4.902s] > Writing 'super' OKAY [ 4.749s] > Sending sparse 'super' 11/13 (114681 KB) OKAY [ 6.041s] > Writing 'super' OKAY [ 5.779s] > Sending sparse 'super' 12/13 (107212 KB) OKAY [ 5.174s] > Writing 'super' OKAY [ 6.587s] > Sending sparse 'super' 13/13 (71496 KB) OKAY [ 3.717s] > Writing 'super' OKAY [ 3.744s] > Finished. Total time: 142.578s > > With this patch: > Sending sparse 'super' 1/13 (114684 KB) OKAY [ 7.149s] > Writing 'super' OKAY [ 1.639s] > Sending sparse 'super' 2/13 (114684 KB) OKAY [ 6.993s] > Writing 'super' OKAY [ 1.713s] > Sending sparse 'super' 3/13 (114684 KB) OKAY [ 7.029s] > Writing 'super' OKAY [ 1.107s] > Sending sparse 'super' 4/13 (114684 KB) OKAY [ 7.027s] > Writing 'super' OKAY [ 0.162s] > Sending sparse 'super' 5/13 (114684 KB) OKAY [ 6.930s] > Writing 'super' OKAY [ 1.643s] > Sending sparse 'super' 6/13 (104176 KB) OKAY [ 6.253s] > Writing 'super' OKAY [ 2.348s] > Sending sparse 'super' 7/13 (108856 KB) OKAY [ 6.346s] > Writing 'super' OKAY [ 0.723s] > Sending sparse 'super' 8/13 (114684 KB) OKAY [ 6.715s] > Writing 'super' OKAY [ 2.848s] > Sending sparse 'super' 9/13 (114684 KB) OKAY [ 6.888s] > Writing 'super' OKAY [ 1.928s] > Sending sparse 'super' 10/13 (100980 KB) OKAY [ 5.979s] > Writing 'super' OKAY [ 1.178s] > Sending sparse 'super' 11/13 (114681 KB) OKAY [ 6.822s] > Writing 'super' OKAY [ 2.652s] > Sending sparse 'super' 12/13 (107212 KB) OKAY [ 6.414s] > Writing 'super' OKAY [ 5.109s] > Sending sparse 'super' 13/13 (71496 KB) OKAY [ 4.238s] > Writing 'super' OKAY [ 0.252s] > Finished. Total time: 108.151s > > It's probably too late for v2023.07 to pick this up but can we consider > taking it for next? I was waiting for a v2, and yes, it's too late for v2023.07. Sorry for not being clear enough.
On jeu., juil. 06, 2023 at 13:00, Tom Rini <trini@konsulko.com> wrote: > On Thu, Jul 06, 2023 at 11:43:13AM +0200, Mattijs Korpershoek wrote: >> On lun., juin 19, 2023 at 10:21, Mattijs Korpershoek <mkorpershoek@baylibre.com> wrote: >> >> > Hi Qianfan, >> > >> > Thank you for your review. >> > >> > On lun., juin 19, 2023 at 14:19, qianfan <qianfanguijin@163.com> wrote: >> > >> >> 在 2023/6/16 21:26, Mattijs Korpershoek 写道: >> >>> Commit 62649165cb02 ("lib: sparse: Make CHUNK_TYPE_RAW buffer aligned") >> >>> fixed cache alignment for systems with a D-CACHE. >> >>> >> >>> However it introduced some performance regressions [1] on system >> >>> flashing huge images, such as Android. >> >>> >> >>> On AM62x SK EVM, we also observe such performance penalty: >> >>> Sending sparse 'super' 1/2 (768793 KB) OKAY [ 23.954s] >> >>> Writing 'super' OKAY [ 75.926s] >> >>> Sending sparse 'super' 2/2 (629819 KB) OKAY [ 19.641s] >> >>> Writing 'super' OKAY [ 62.849s] >> >>> Finished. Total time: 182.474s >> >>> >> >>> The reason for this is that we use an arbitrary small buffer >> >>> (info->blksz * 100) for transferring. >> >>> >> >>> Fix it by using a bigger buffer (info->blksz * blkcnt) as suggested in >> >>> the original's patch review [2]. >> >>> >> >>> With this patch, performance impact is mitigated: >> >>> Sending sparse 'super' 1/2 (768793 KB) OKAY [ 24.006s] >> >>> Writing 'super' OKAY [ 15.920s] >> >>> Sending sparse 'super' 2/2 (629819 KB) OKAY [ 19.651s] >> >>> Writing 'super' OKAY [ 14.665s] >> >>> Finished. Total time: 74.346s >> >>> >> >>> [1] https://lore.kernel.org/r/20221118121323.4009193-1-gary.bisson@boundarydevices.com >> >>> [2] https://lore.kernel.org/r/all/43e4c17c-4483-ec8e-f843-9b4c5569bd18@seco.com/ >> >>> >> >>> Fixes: 62649165cb02 ("lib: sparse: Make CHUNK_TYPE_RAW buffer aligned") >> >>> Signed-off-by: Mattijs Korpershoek <mkorpershoek@baylibre.com> >> >>> --- >> >>> lib/image-sparse.c | 2 +- >> >>> 1 file changed, 1 insertion(+), 1 deletion(-) >> >>> >> >>> diff --git a/lib/image-sparse.c b/lib/image-sparse.c >> >>> index 5ec0f94ab3eb..25aed0604192 100644 >> >>> --- a/lib/image-sparse.c >> >>> +++ b/lib/image-sparse.c >> >>> @@ -55,7 +55,7 @@ static lbaint_t write_sparse_chunk_raw(struct sparse_storage *info, >> >>> void *data, >> >>> char *response) >> >>> { >> >>> - lbaint_t n = blkcnt, write_blks, blks = 0, aligned_buf_blks = 100; >> >>> + lbaint_t n = blkcnt, write_blks, blks = 0, aligned_buf_blks = blkcnt; >> >> Hi: >> >> >> >> It's a good point that this code report the performance was affected by >> >> write large small >> >> mmc blks, not memory copy. >> > >> > I believe memory copy also affects performance, but in my case, >> > it has less impact than small mmc blks. >> > >> > With 62649165cb02 reverted: >> > Sending sparse 'super' 1/2 (768793 KB) OKAY [ 23.947s] >> > Writing 'super' OKAY [ 12.983s] >> > Sending sparse 'super' 2/2 (629819 KB) OKAY [ 19.600s] >> > Writing 'super' OKAY [ 12.796s] >> > Finished. Total time: 69.430s >> > >> > With aligned_buf_blks = blkcnt: >> > Sending sparse 'super' 1/2 (768793 KB) OKAY [ 24.072s] >> > Writing 'super' OKAY [ 16.177s] >> > Sending sparse 'super' 2/2 (629819 KB) OKAY [ 19.681s] >> > Writing 'super' OKAY [ 14.845s] >> > Finished. Total time: 74.919s >> > >> >> >> >> And I can not make sure whether memalign can always alloc such huge >> >> memory when we change the >> >> aligned_buf_blks to blkcnt. >> > >> > Could you clarify the concern here? I've dumped blkcnt for my board >> > (AM62x SK EVK) and the biggest blkcnt I found was: 131072 >> > >> > With info->blksz = 512, this gives me: 512 * 131072 = 67108864 >> > >> > Which is a memalign (memory alloc) of 64MB. Is 64MB really that big? (I >> > don't realize it's that much) >> > >> >> >> >> Could you please set aligned_buf_blks to FASTBOOT_MAX_BLK_WRITE(16384) >> >> and test again? >> > >> > With aligned_buf_blks = FASTBOOT_MAX_BLK_WRITE(16384): >> > Sending sparse 'super' 1/2 (768793 KB) OKAY [ 23.912s] >> > Writing 'super' OKAY [ 15.780s] >> > Sending sparse 'super' 2/2 (629819 KB) OKAY [ 19.581s] >> > Writing 'super' OKAY [ 17.192s] >> > Finished. Total time: 76.569s >> > >> > So using FASTBOOT_MAX_BLK_WRITE is slightly worse than using blkcnt. >> > But allocations (for blksz = 512) are smaller: 8MB instead of 64MB in my example. >> > >> > I can spin up a v2 with FASTBOOT_MAX_BLK_WRITE but i'm waiting a little >> > more feedback before doing so. >> >> Hi Marek, Tom, >> >> What's your take on this ? Can we keep blkcnt or should I respin using >> FASTBOOT_MAX_BLK_WRITE ? >> >> I have also tested this on VIM3, on >> U-Boot 2023.07-rc6-00003-g923de765ee1a: >> >> Sending sparse 'super' 1/13 (114684 KB) OKAY [ 5.442s] >> Writing 'super' OKAY [ 5.791s] >> Sending sparse 'super' 2/13 (114684 KB) OKAY [ 5.706s] >> Writing 'super' OKAY [ 5.607s] >> Sending sparse 'super' 3/13 (114684 KB) OKAY [ 5.468s] >> Writing 'super' OKAY [ 5.835s] >> Sending sparse 'super' 4/13 (114684 KB) OKAY [ 5.703s] >> Writing 'super' OKAY [ 5.618s] >> Sending sparse 'super' 5/13 (114684 KB) OKAY [ 6.176s] >> Writing 'super' OKAY [ 5.421s] >> Sending sparse 'super' 6/13 (104176 KB) OKAY [ 5.204s] >> Writing 'super' OKAY [ 5.199s] >> Sending sparse 'super' 7/13 (108856 KB) OKAY [ 5.456s] >> Writing 'super' OKAY [ 5.290s] >> Sending sparse 'super' 8/13 (114684 KB) OKAY [ 6.122s] >> Writing 'super' OKAY [ 5.838s] >> Sending sparse 'super' 9/13 (114684 KB) OKAY [ 5.951s] >> Writing 'super' OKAY [ 5.857s] >> Sending sparse 'super' 10/13 (100980 KB) OKAY [ 4.902s] >> Writing 'super' OKAY [ 4.749s] >> Sending sparse 'super' 11/13 (114681 KB) OKAY [ 6.041s] >> Writing 'super' OKAY [ 5.779s] >> Sending sparse 'super' 12/13 (107212 KB) OKAY [ 5.174s] >> Writing 'super' OKAY [ 6.587s] >> Sending sparse 'super' 13/13 (71496 KB) OKAY [ 3.717s] >> Writing 'super' OKAY [ 3.744s] >> Finished. Total time: 142.578s >> >> With this patch: >> Sending sparse 'super' 1/13 (114684 KB) OKAY [ 7.149s] >> Writing 'super' OKAY [ 1.639s] >> Sending sparse 'super' 2/13 (114684 KB) OKAY [ 6.993s] >> Writing 'super' OKAY [ 1.713s] >> Sending sparse 'super' 3/13 (114684 KB) OKAY [ 7.029s] >> Writing 'super' OKAY [ 1.107s] >> Sending sparse 'super' 4/13 (114684 KB) OKAY [ 7.027s] >> Writing 'super' OKAY [ 0.162s] >> Sending sparse 'super' 5/13 (114684 KB) OKAY [ 6.930s] >> Writing 'super' OKAY [ 1.643s] >> Sending sparse 'super' 6/13 (104176 KB) OKAY [ 6.253s] >> Writing 'super' OKAY [ 2.348s] >> Sending sparse 'super' 7/13 (108856 KB) OKAY [ 6.346s] >> Writing 'super' OKAY [ 0.723s] >> Sending sparse 'super' 8/13 (114684 KB) OKAY [ 6.715s] >> Writing 'super' OKAY [ 2.848s] >> Sending sparse 'super' 9/13 (114684 KB) OKAY [ 6.888s] >> Writing 'super' OKAY [ 1.928s] >> Sending sparse 'super' 10/13 (100980 KB) OKAY [ 5.979s] >> Writing 'super' OKAY [ 1.178s] >> Sending sparse 'super' 11/13 (114681 KB) OKAY [ 6.822s] >> Writing 'super' OKAY [ 2.652s] >> Sending sparse 'super' 12/13 (107212 KB) OKAY [ 6.414s] >> Writing 'super' OKAY [ 5.109s] >> Sending sparse 'super' 13/13 (71496 KB) OKAY [ 4.238s] >> Writing 'super' OKAY [ 0.252s] >> Finished. Total time: 108.151s >> >> It's probably too late for v2023.07 to pick this up but can we consider >> taking it for next? > > I was waiting for a v2, and yes, it's too late for v2023.07. Sorry for > not being clear enough. Oh, sorry I did not understand that. I understand for v2023.07. Thank you for the quick answer. Will send a v2 shortly using FASTBOOT_MAX_BLK_WRITE. > > -- > Tom
diff --git a/lib/image-sparse.c b/lib/image-sparse.c index 5ec0f94ab3eb..25aed0604192 100644 --- a/lib/image-sparse.c +++ b/lib/image-sparse.c @@ -55,7 +55,7 @@ static lbaint_t write_sparse_chunk_raw(struct sparse_storage *info, void *data, char *response) { - lbaint_t n = blkcnt, write_blks, blks = 0, aligned_buf_blks = 100; + lbaint_t n = blkcnt, write_blks, blks = 0, aligned_buf_blks = blkcnt; uint32_t *aligned_buf = NULL; if (CONFIG_IS_ENABLED(SYS_DCACHE_OFF)) {
Commit 62649165cb02 ("lib: sparse: Make CHUNK_TYPE_RAW buffer aligned") fixed cache alignment for systems with a D-CACHE. However it introduced some performance regressions [1] on system flashing huge images, such as Android. On AM62x SK EVM, we also observe such performance penalty: Sending sparse 'super' 1/2 (768793 KB) OKAY [ 23.954s] Writing 'super' OKAY [ 75.926s] Sending sparse 'super' 2/2 (629819 KB) OKAY [ 19.641s] Writing 'super' OKAY [ 62.849s] Finished. Total time: 182.474s The reason for this is that we use an arbitrary small buffer (info->blksz * 100) for transferring. Fix it by using a bigger buffer (info->blksz * blkcnt) as suggested in the original's patch review [2]. With this patch, performance impact is mitigated: Sending sparse 'super' 1/2 (768793 KB) OKAY [ 24.006s] Writing 'super' OKAY [ 15.920s] Sending sparse 'super' 2/2 (629819 KB) OKAY [ 19.651s] Writing 'super' OKAY [ 14.665s] Finished. Total time: 74.346s [1] https://lore.kernel.org/r/20221118121323.4009193-1-gary.bisson@boundarydevices.com [2] https://lore.kernel.org/r/all/43e4c17c-4483-ec8e-f843-9b4c5569bd18@seco.com/ Fixes: 62649165cb02 ("lib: sparse: Make CHUNK_TYPE_RAW buffer aligned") Signed-off-by: Mattijs Korpershoek <mkorpershoek@baylibre.com> --- lib/image-sparse.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) --- base-commit: 2f4664f5c3edc55b18d8906f256a4c8e303243c0 change-id: 20230616-sparse-flash-fix-9c2852aa8d16 Best regards,