[1/1] venus : Fix for H265 decoding failure.

Message ID 20221115121004.28197-2-quic_vboma@quicinc.com (mailing list archive)
State Accepted
Delegated to: Stanimir Varbanov
Headers
Series Fix for H265 decoding failure . |

Commit Message

Viswanath Boma Nov. 15, 2022, 12:10 p.m. UTC
  From: Viswanath Boma <quic_vboma@quicinc.com>

Aligned the mismatch of persist1 and scratch1 buffer calculation,
as per the firmware requirements .

Signed-off-by: Vikash Garodia <vgarodia@qti.qualcomm.com>
Signed-off-by: Viswanath Boma <quic_vboma@quicinc.com>
---
 drivers/media/platform/qcom/venus/hfi_plat_bufs_v6.c | 4 ++--
 1 file changed, 2 insertions(+), 2 deletions(-)
  

Comments

Nathan Hebert Dec. 2, 2022, 12:12 a.m. UTC | #1
On Tue, Nov 15, 2022 at 4:10 AM <quic_vboma@quicinc.com> wrote:
>
> From: Viswanath Boma <quic_vboma@quicinc.com>
>
> Aligned the mismatch of persist1 and scratch1 buffer calculation,
> as per the firmware requirements .
>
> Signed-off-by: Vikash Garodia <vgarodia@qti.qualcomm.com>
> Signed-off-by: Viswanath Boma <quic_vboma@quicinc.com>
> ---
>  drivers/media/platform/qcom/venus/hfi_plat_bufs_v6.c | 4 ++--
>  1 file changed, 2 insertions(+), 2 deletions(-)
>
> diff --git a/drivers/media/platform/qcom/venus/hfi_plat_bufs_v6.c b/drivers/media/platform/qcom/venus/hfi_plat_bufs_v6.c
> index ea25c451222b..a9be31ec6927 100644
> --- a/drivers/media/platform/qcom/venus/hfi_plat_bufs_v6.c
> +++ b/drivers/media/platform/qcom/venus/hfi_plat_bufs_v6.c
> @@ -93,7 +93,7 @@
>  #define LCU_MIN_SIZE_PELS              16
>  #define SIZE_SEI_USERDATA              4096
>
> -#define H265D_MAX_SLICE                        600
> +#define H265D_MAX_SLICE                        3600
>  #define SIZE_H265D_HW_PIC_T            SIZE_H264D_HW_PIC_T
>  #define SIZE_H265D_BSE_CMD_PER_BUF     (16 * sizeof(u32))
>  #define SIZE_H265D_VPP_CMD_PER_BUF     256
> @@ -1021,7 +1021,7 @@ static u32 h264d_persist1_size(void)
>  static u32 h265d_persist1_size(void)
>  {
>         return ALIGN((SIZE_SLIST_BUF_H265 * NUM_SLIST_BUF_H265 + H265_NUM_TILE
> -                       * sizeof(u32)), HFI_DMA_ALIGNMENT);
> +                       * sizeof(u32) + NUM_HW_PIC_BUF * SIZE_SEI_USERDATA), HFI_DMA_ALIGNMENT);
>  }
>
>  static u32 vp8d_persist1_size(void)
> --
> 2.17.1
>
Hi Viswanath. I tested this patch on ChromiumOS's downstream
5.15-based fork. Using ITU-T HEVC test vectors [0] I am seeing
firmware errors "qcom-venus-decoder aa00000.video-codec:video-decoder:
VenusMed : event not sufficient resources". Does this change fix HEVC
decoding for you?

[0]: https://www.itu.int/wftp3/av-arch/jctvc-site/bitstream_exchange/draft_conformance/HEVC_v1/

Best regards,
Nathan Hebert
  
Bryan O'Donoghue Dec. 2, 2022, 12:51 a.m. UTC | #2
On 15/11/2022 12:10, quic_vboma@quicinc.com wrote:
> From: Viswanath Boma <quic_vboma@quicinc.com>
> 
> Aligned the mismatch of persist1 and scratch1 buffer calculation,
> as per the firmware requirements .
> 
> Signed-off-by: Vikash Garodia <vgarodia@qti.qualcomm.com>
> Signed-off-by: Viswanath Boma <quic_vboma@quicinc.com>
> ---
>   drivers/media/platform/qcom/venus/hfi_plat_bufs_v6.c | 4 ++--
>   1 file changed, 2 insertions(+), 2 deletions(-)
> 
> diff --git a/drivers/media/platform/qcom/venus/hfi_plat_bufs_v6.c b/drivers/media/platform/qcom/venus/hfi_plat_bufs_v6.c
> index ea25c451222b..a9be31ec6927 100644
> --- a/drivers/media/platform/qcom/venus/hfi_plat_bufs_v6.c
> +++ b/drivers/media/platform/qcom/venus/hfi_plat_bufs_v6.c
> @@ -93,7 +93,7 @@
>   #define LCU_MIN_SIZE_PELS		16
>   #define SIZE_SEI_USERDATA		4096
>   
> -#define H265D_MAX_SLICE			600
> +#define H265D_MAX_SLICE			3600
>   #define SIZE_H265D_HW_PIC_T		SIZE_H264D_HW_PIC_T
>   #define SIZE_H265D_BSE_CMD_PER_BUF	(16 * sizeof(u32))
>   #define SIZE_H265D_VPP_CMD_PER_BUF	256
> @@ -1021,7 +1021,7 @@ static u32 h264d_persist1_size(void)
>   static u32 h265d_persist1_size(void)
>   {
>   	return ALIGN((SIZE_SLIST_BUF_H265 * NUM_SLIST_BUF_H265 + H265_NUM_TILE
> -			* sizeof(u32)), HFI_DMA_ALIGNMENT);
> +			* sizeof(u32) + NUM_HW_PIC_BUF * SIZE_SEI_USERDATA), HFI_DMA_ALIGNMENT);
>   }
>   
>   static u32 vp8d_persist1_size(void)

If this is a real fix it need to be applied elsewhere

Please add a Fixes tag

Fixes: 3a75bf4e7925 ("media: venus: Add platform buffers for v6")

and

Cc: <stable@vger.kernel.org> # 5.14.x

---
bod
  
Nathan Hebert Dec. 2, 2022, 7:48 p.m. UTC | #3
On Thu, Dec 1, 2022 at 4:12 PM Nathan Hebert <nhebert@chromium.org> wrote:
>
> On Tue, Nov 15, 2022 at 4:10 AM <quic_vboma@quicinc.com> wrote:
> >
> > From: Viswanath Boma <quic_vboma@quicinc.com>
> >
> > Aligned the mismatch of persist1 and scratch1 buffer calculation,
> > as per the firmware requirements .
> >
> > Signed-off-by: Vikash Garodia <vgarodia@qti.qualcomm.com>
> > Signed-off-by: Viswanath Boma <quic_vboma@quicinc.com>
> > ---
> >  drivers/media/platform/qcom/venus/hfi_plat_bufs_v6.c | 4 ++--
> >  1 file changed, 2 insertions(+), 2 deletions(-)
> >
> > diff --git a/drivers/media/platform/qcom/venus/hfi_plat_bufs_v6.c b/drivers/media/platform/qcom/venus/hfi_plat_bufs_v6.c
> > index ea25c451222b..a9be31ec6927 100644
> > --- a/drivers/media/platform/qcom/venus/hfi_plat_bufs_v6.c
> > +++ b/drivers/media/platform/qcom/venus/hfi_plat_bufs_v6.c
> > @@ -93,7 +93,7 @@
> >  #define LCU_MIN_SIZE_PELS              16
> >  #define SIZE_SEI_USERDATA              4096
> >
> > -#define H265D_MAX_SLICE                        600
> > +#define H265D_MAX_SLICE                        3600
> >  #define SIZE_H265D_HW_PIC_T            SIZE_H264D_HW_PIC_T
> >  #define SIZE_H265D_BSE_CMD_PER_BUF     (16 * sizeof(u32))
> >  #define SIZE_H265D_VPP_CMD_PER_BUF     256
> > @@ -1021,7 +1021,7 @@ static u32 h264d_persist1_size(void)
> >  static u32 h265d_persist1_size(void)
> >  {
> >         return ALIGN((SIZE_SLIST_BUF_H265 * NUM_SLIST_BUF_H265 + H265_NUM_TILE
> > -                       * sizeof(u32)), HFI_DMA_ALIGNMENT);
> > +                       * sizeof(u32) + NUM_HW_PIC_BUF * SIZE_SEI_USERDATA), HFI_DMA_ALIGNMENT);
> >  }
> >
> >  static u32 vp8d_persist1_size(void)
> > --
> > 2.17.1
> >
> Hi Viswanath. I tested this patch on ChromiumOS's downstream
> 5.15-based fork. Using ITU-T HEVC test vectors [0] I am seeing
> firmware errors "qcom-venus-decoder aa00000.video-codec:video-decoder:
> VenusMed : event not sufficient resources". Does this change fix HEVC
> decoding for you?
>

Hi Viswanath. Thanks for looking at the logs offline. I re-tested
after adding Q08C (UBWC) support to my decoder client, and it fixed
the error that I was seeing. My mistake. Each of the main profile test
vectors now decodes correctly after applying your patch!

Tested-by: Nathan Hebert <nhebert@chromium.org>

> [0]: https://www.itu.int/wftp3/av-arch/jctvc-site/bitstream_exchange/draft_conformance/HEVC_v1/
>
> Best regards,
> Nathan Hebert
  
Viswanath Boma (Temp) Dec. 5, 2022, 4:48 a.m. UTC | #4
-----Original Message-----
From: Nathan Hebert <nhebert@chromium.org> 
Sent: Saturday, December 3, 2022 1:19 AM
To: Viswanath Boma (Temp) (QUIC) <quic_vboma@quicinc.com>
Cc: stanimir.varbanov@linaro.org; Andy Gross <agross@kernel.org>; bjorn.andersson@linaro.org; Mauro Carvalho Chehab <mchehab@kernel.org>; linux-media@vger.kernel.org; linux-arm-msm@vger.kernel.org; linux-kernel@vger.kernel.org; Vikash Garodia <vgarodia@qti.qualcomm.com>
Subject: Re: [PATCH 1/1] venus : Fix for H265 decoding failure.

WARNING: This email originated from outside of Qualcomm. Please be wary of any links or attachments, and do not enable macros.

On Thu, Dec 1, 2022 at 4:12 PM Nathan Hebert <nhebert@chromium.org> wrote:
>
> On Tue, Nov 15, 2022 at 4:10 AM <quic_vboma@quicinc.com> wrote:
> >
> > From: Viswanath Boma <quic_vboma@quicinc.com>
> >
> > Aligned the mismatch of persist1 and scratch1 buffer calculation, as 
> > per the firmware requirements .
> >
> > Signed-off-by: Vikash Garodia <vgarodia@qti.qualcomm.com>
> > Signed-off-by: Viswanath Boma <quic_vboma@quicinc.com>
> > ---
> >  drivers/media/platform/qcom/venus/hfi_plat_bufs_v6.c | 4 ++--
> >  1 file changed, 2 insertions(+), 2 deletions(-)
> >
> > diff --git a/drivers/media/platform/qcom/venus/hfi_plat_bufs_v6.c 
> > b/drivers/media/platform/qcom/venus/hfi_plat_bufs_v6.c
> > index ea25c451222b..a9be31ec6927 100644
> > --- a/drivers/media/platform/qcom/venus/hfi_plat_bufs_v6.c
> > +++ b/drivers/media/platform/qcom/venus/hfi_plat_bufs_v6.c
> > @@ -93,7 +93,7 @@
> >  #define LCU_MIN_SIZE_PELS              16
> >  #define SIZE_SEI_USERDATA              4096
> >
> > -#define H265D_MAX_SLICE                        600
> > +#define H265D_MAX_SLICE                        3600
> >  #define SIZE_H265D_HW_PIC_T            SIZE_H264D_HW_PIC_T
> >  #define SIZE_H265D_BSE_CMD_PER_BUF     (16 * sizeof(u32))
> >  #define SIZE_H265D_VPP_CMD_PER_BUF     256
> > @@ -1021,7 +1021,7 @@ static u32 h264d_persist1_size(void)  static 
> > u32 h265d_persist1_size(void)  {
> >         return ALIGN((SIZE_SLIST_BUF_H265 * NUM_SLIST_BUF_H265 + H265_NUM_TILE
> > -                       * sizeof(u32)), HFI_DMA_ALIGNMENT);
> > +                       * sizeof(u32) + NUM_HW_PIC_BUF * 
> > + SIZE_SEI_USERDATA), HFI_DMA_ALIGNMENT);
> >  }
> >
> >  static u32 vp8d_persist1_size(void)
> > --
> > 2.17.1
> >
> Hi Viswanath. I tested this patch on ChromiumOS's downstream 
> 5.15-based fork. Using ITU-T HEVC test vectors [0] I am seeing 
> firmware errors "qcom-venus-decoder aa00000.video-codec:video-decoder:
> VenusMed : event not sufficient resources". Does this change fix HEVC 
> decoding for you?
>

Hi Viswanath. Thanks for looking at the logs offline. I re-tested after adding Q08C (UBWC) support to my decoder client, and it fixed the error that I was seeing. My mistake. Each of the main profile test vectors now decodes correctly after applying your patch!
[vboma] Thank you Nathan for the update .

Tested-by: Nathan Hebert <nhebert@chromium.org>

> [0]: 
> https://www.itu.int/wftp3/av-arch/jctvc-site/bitstream_exchange/draft_
> conformance/HEVC_v1/
>
> Best regards,
> Nathan Hebert
  
Viswanath Boma March 16, 2023, 7:57 a.m. UTC | #5
From: Viswanath Boma <quic_vboma@quicinc.com>

Fixed strict patch warnings and ensured rebase.

Fixed the buffer size calculation mismatch with firmware requirements.
Tested  on v5.15  kernel.
For testing utilised the chrome utilities .


Viswanath Boma (1):
  venus : Fix for H265 decoding failure.

 drivers/media/platform/qcom/venus/hfi_plat_bufs_v6.c | 4 ++--
 1 file changed, 2 insertions(+), 2 deletions(-)
  

Patch

diff --git a/drivers/media/platform/qcom/venus/hfi_plat_bufs_v6.c b/drivers/media/platform/qcom/venus/hfi_plat_bufs_v6.c
index ea25c451222b..a9be31ec6927 100644
--- a/drivers/media/platform/qcom/venus/hfi_plat_bufs_v6.c
+++ b/drivers/media/platform/qcom/venus/hfi_plat_bufs_v6.c
@@ -93,7 +93,7 @@ 
 #define LCU_MIN_SIZE_PELS		16
 #define SIZE_SEI_USERDATA		4096
 
-#define H265D_MAX_SLICE			600
+#define H265D_MAX_SLICE			3600
 #define SIZE_H265D_HW_PIC_T		SIZE_H264D_HW_PIC_T
 #define SIZE_H265D_BSE_CMD_PER_BUF	(16 * sizeof(u32))
 #define SIZE_H265D_VPP_CMD_PER_BUF	256
@@ -1021,7 +1021,7 @@  static u32 h264d_persist1_size(void)
 static u32 h265d_persist1_size(void)
 {
 	return ALIGN((SIZE_SLIST_BUF_H265 * NUM_SLIST_BUF_H265 + H265_NUM_TILE
-			* sizeof(u32)), HFI_DMA_ALIGNMENT);
+			* sizeof(u32) + NUM_HW_PIC_BUF * SIZE_SEI_USERDATA), HFI_DMA_ALIGNMENT);
 }
 
 static u32 vp8d_persist1_size(void)