Skip to contentSkip to conversation

Subject: [PATCH] pci/msi: remove redundant calculation in msi_setup_msi_desc

PATCH 黄少波 3 participants 6 messages Aug 18, 2023 – Aug 24, 2023 PCI

lore
Share & tools
lore

黄少波

author
#0 lore raw

Whether to support 64-bit address status has been calculated before, and the calculation result can be used directly afterwards, so use msi_attrib.is_64 to avoid double calculation.

---
 drivers/pci/msi/msi.c | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)
drivers/pci/msi/msi.c +1 -1
index ef1d885..304e889 100644
@@ -303,7 +303,7 @@ static int msi_setup_msi_desc(struct pci_dev *dev, int nvec,
        desc.pci.msi_attrib.multiple    = ilog2(__roundup_pow_of_two(nvec));
        desc.affinity                   = masks;
 
-       if (control & PCI_MSI_FLAGS_64BIT)
+       if (desc.pci.msi_attrib.is_64)
                desc.pci.mask_pos = dev->msi_cap + PCI_MSI_MASK_64;
        else
                desc.pci.mask_pos = dev->msi_cap + PCI_MSI_MASK_32;

--
2.7.4

#/******本邮件及其附件含有小米公司的保密信息,仅限于发送给上面地址中列出的个人或群组。禁止任何其他人以任何形式使用(包括但不限于全部或部分地泄露、复制、或散发)本邮件中的信息。如果您错收了本邮件,请您立即电话或邮件通知发件人并删除本邮件! This e-mail and its attachments contain confidential information from XIAOMI, which is intended only for the person or entity whose address is listed above. Any use of the information contained herein in any way (including, but not limited to, total or partial disclosure, reproduction, or dissemination) by persons other than the intended recipient(s) is prohibited. If you receive this e-mail in error, please notify the sender by phone or email immediately and delete it!******/#

Reviewed-by
#1 lore raw

On Fri, 18 Aug 2023, 黄少波 wrote:

You should add () to the function name in the subject and use capital PCI/MSI.

Quoted 24 lines

Whether to support 64-bit address status has been calculated before, and the calculation result can be used directly afterwards, so use msi_attrib.is_64 to avoid double calculation.

---
 drivers/pci/msi/msi.c | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)
drivers/pci/msi/msi.c +1 -2
index ef1d885..304e889 100644
@@ -303,7 +303,7 @@ static int msi_setup_msi_desc(struct pci_dev *dev, int nvec,
        desc.pci.msi_attrib.multiple    = ilog2(__roundup_pow_of_two(nvec));
        desc.affinity                   = masks;
 
-       if (control & PCI_MSI_FLAGS_64BIT)
+       if (desc.pci.msi_attrib.is_64)
                desc.pci.mask_pos = dev->msi_cap + PCI_MSI_MASK_64;
        else
                desc.pci.mask_pos = dev->msi_cap + PCI_MSI_MASK_32;
--
Quoted 1 line

2.7.4

Signature · 1 line
 i.
#2 lore raw
In reply to 黄少波

On Fri, Aug 18 2023 at 07:26, 黄少波 wrote:

Your patch is corrupt:

Applying: Subject: [PATCH] pci/msi: remove redundant calculation in msi_setup_msi_desc
error: corrupt patch at line 12

It's DOS mangled and whitespace damaged.

Quoted 3 lines

Whether to support 64-bit address status has been calculated before, and the calculation result can be used directly afterwards, so use msi_attrib.is_64 to avoid double calculation.

I'm not seeing what this solves:

drivers/pci/msi/msi.c static int msi_setup_msi_desc(struct pci_dev *dev, int nvec, · line 306

Quoted 2 lines
drivers/pci/msi/msi.c +1 -1
-       if (control & PCI_MSI_FLAGS_64BIT)
+       if (desc.pci.msi_attrib.is_64)

Both variants resolve to a test of a bit and a conditional instruction on the result. It's exactly zero difference in terms of "calculation".

So all this does is change the memory location to test. Not more not less. It does not generate better code and does not save anything.

Thanks,

tglx

#3 lore raw
Quoted 8 lines · Thomas Gleixner

On Fri, Aug 18 2023 at 07:26, 黄少波 wrote:

Your patch is corrupt:

Applying: Subject: [PATCH] pci/msi: remove redundant calculation in msi_setup_msi_desc
error: corrupt patch at line 12

It's DOS mangled and whitespace damaged.

Very sorry, may have accidentally broken the patch format, and will make sure to use plain text next time.

Quoted 6 lines
Quoted 3 lines

Whether to support 64-bit address status has been calculated before, and the calculation result can be used directly afterwards, so use msi_attrib.is_64 to avoid double calculation.

I'm not seeing what this solves:

Quoted 8 lines

drivers/pci/msi/msi.c static int msi_setup_msi_desc(struct pci_dev *dev, int nvec, · line 306

Quoted 2 lines
drivers/pci/msi/msi.c +1 -1
-       if (control & PCI_MSI_FLAGS_64BIT)
+       if (desc.pci.msi_attrib.is_64)

Both variants resolve to a test of a bit and a conditional instruction on the result. It's exactly zero difference in terms of "calculation".

So all this does is change the memory location to test. Not more not less. It does not generate better code and does not save anything.

It may not be appropriate to write to eliminate duplicate calculations, can it be proposed again with clean code?

Quoted 3 lines

Thanks,

tglx

thanks,
sparkhuang
#/******本邮件及其附件含有小米公司的保密信息,仅限于发送给上面地址中列出的个人或群组。禁止任何其他人以任何形式使用(包括但不限于全部或部分地泄露、复制、或散发)本邮件中的信息。如果您错收了本邮件,请您立即电话或邮件通知发件人并删除本邮件! This e-mail and its attachments contain confidential information from XIAOMI, which is intended only for the person or entity whose address is listed above. Any use of the information contained herein in any way (including, but not limited to, total or partial disclosure, reproduction, or dissemination) by persons other than the intended recipient(s) is prohibited. If you receive this e-mail in error, please notify the sender by phone or email immediately and delete it!******/#

#4 lore raw
Quoted 6 lines · On Thu, Aug 24 2023 at 00:27, [email protected] wrote:
Quoted 4 lines · Thomas Gleixner

I'm not seeing what this solves:

Quoted 2 lines
drivers/pci/msi/msi.c +1 -1
-       if (control & PCI_MSI_FLAGS_64BIT)
+       if (desc.pci.msi_attrib.is_64)
Quoted 8 lines
Quoted 5 lines

Both variants resolve to a test of a bit and a conditional instruction on the result. It's exactly zero difference in terms of "calculation".

So all this does is change the memory location to test. Not more not less. It does not generate better code and does not save anything.

It may not be appropriate to write to eliminate duplicate calculations, can it be proposed again with clean code?

What's wrong with the existing code?

author
#5 lore raw
Quoted 10 lines · Thomas Gleixner
Quoted 9 lines · On Thu, Aug 24 2023 at 00:27, [email protected] wrote:
Quoted 7 lines · Thomas Gleixner

I'm not seeing what this solves:

drivers/pci/msi/msi.c static int msi_setup_msi_desc(struct pci_dev *dev, int nvec, · line 306

Quoted 2 lines
drivers/pci/msi/msi.c +1 -1
-       if (control & PCI_MSI_FLAGS_64BIT)
+       if (desc.pci.msi_attrib.is_64)

Both variants resolve to a test of a bit and a conditional instruction on the result. It's exactly zero difference in terms of "calculation".

Quoted 7 lines
Quoted 5 lines
Quoted 2 lines

So all this does is change the memory location to test. Not more not less. It does not generate better code and does not save anything.

It may not be appropriate to write to eliminate duplicate calculations, can it be proposed again with clean code?

What's wrong with the existing code?

There is nothing wrong with the current code, using desc.pci.msi_attrib.is_64 does not solve any problems, thank you very much for reviewing comments.

Thanks,
sparkhuang
#/******���ʼ����丽������С�׹�˾�ı�����Ϣ�������ڷ��͸������ַ���г��ĸ��˻�Ⱥ�顣��ֹ�κ����������κ���ʽʹ�ã�������������ȫ���򲿷ֵ�й¶�����ơ���ɢ�������ʼ��е���Ϣ������������˱��ʼ������������绰���ʼ�֪ͨ�����˲�ɾ�����ʼ��� This e-mail and its attachments contain confidential information from XIAOMI, which is intended only for the person or entity whose address is listed above. Any use of the information contained herein in any way (including, but not limited to, total or partial disclosure, reproduction, or dissemination) by persons other than the intended recipient(s) is prohibited. If you receive this e-mail in error, please notify the sender by phone or email immediately and delete it!******/#