[Qemu-devel] [PATCH 6/7] migraion: optimiztion xbzrle by reducing data copy

[Qemu-devel] [PATCH 6/7] migraion: optimiztion xbzrle by reducing data copy

[Gonglei (Arei)]

Reducing data copy can reduce cpu overheah.

Signed-off-by: ChenLiang <chenliang88@huawei.com>
Signed-off-by: Gonglei <arei.gonglei@huawei.com>
---
 arch_init.c | 8 +++-----
 1 file changed, 3 insertions(+), 5 deletions(-)

diff --git a/arch_init.c b/arch_init.c
index 2211e0b..cc88875 100644
--- a/arch_init.c
+++ b/arch_init.c
@@ -344,11 +344,8 @@ static int save_xbzrle_page(QEMUFile *f, uint8_t **current_data,
 
     prev_cached_page = get_cached_data(XBZRLE.cache, current_addr);
 
-    /* save current buffer into memory */
-    memcpy(XBZRLE.current_buf, *current_data, TARGET_PAGE_SIZE);
-
     /* XBZRLE encoding (if there is no overflow) */
-    encoded_len = xbzrle_encode_buffer(prev_cached_page, XBZRLE.current_buf,
+    encoded_len = xbzrle_encode_buffer(prev_cached_page, *current_data,
                                        TARGET_PAGE_SIZE, XBZRLE.encoded_buf,
                                        TARGET_PAGE_SIZE);
     if (encoded_len == 0) {
@@ -367,7 +364,8 @@ static int save_xbzrle_page(QEMUFile *f, uint8_t **current_data,
 
     /* we need to update the data in the cache, in order to get the same data */
     if (!last_stage) {
-        memcpy(prev_cached_page, XBZRLE.current_buf, TARGET_PAGE_SIZE);
+        xbzrle_decode_buffer(XBZRLE.encoded_buf, encoded_len, prev_cached_page,
+                                                              TARGET_PAGE_SIZE);
     }
 
     /* Send XBZRLE based compressed page */
-- 
1.7.12.4


Best regards,
-Gonglei

Re: [Qemu-devel] [PATCH 6/7] migraion: optimiztion xbzrle by reducing data copy

[Dr. David Alan Gilbert]

* Gonglei (Arei) (arei.gonglei@huawei.com) wrote:
> Reducing data copy can reduce cpu overheah.

(Note a few typos in subject/title)

> Signed-off-by: ChenLiang <chenliang88@huawei.com>
> Signed-off-by: Gonglei <arei.gonglei@huawei.com>
> ---
>  arch_init.c | 8 +++-----
>  1 file changed, 3 insertions(+), 5 deletions(-)
> 
> diff --git a/arch_init.c b/arch_init.c
> index 2211e0b..cc88875 100644
> --- a/arch_init.c
> +++ b/arch_init.c
> @@ -344,11 +344,8 @@ static int save_xbzrle_page(QEMUFile *f, uint8_t **current_data,
>  
>      prev_cached_page = get_cached_data(XBZRLE.cache, current_addr);
>  
> -    /* save current buffer into memory */
> -    memcpy(XBZRLE.current_buf, *current_data, TARGET_PAGE_SIZE);
> -
>      /* XBZRLE encoding (if there is no overflow) */
> -    encoded_len = xbzrle_encode_buffer(prev_cached_page, XBZRLE.current_buf,
> +    encoded_len = xbzrle_encode_buffer(prev_cached_page, *current_data,
>                                         TARGET_PAGE_SIZE, XBZRLE.encoded_buf,
>                                         TARGET_PAGE_SIZE);

Is xbzrle_encode_buffer safe if the main memory is still being changed while
it is run? Even with multiple CPUs changing it? Even on CPUs with looser
memory ordering semantics?

>      if (encoded_len == 0) {
> @@ -367,7 +364,8 @@ static int save_xbzrle_page(QEMUFile *f, uint8_t **current_data,
>  
>      /* we need to update the data in the cache, in order to get the same data */
>      if (!last_stage) {
> -        memcpy(prev_cached_page, XBZRLE.current_buf, TARGET_PAGE_SIZE);
> +        xbzrle_decode_buffer(XBZRLE.encoded_buf, encoded_len, prev_cached_page,
> +                                                              TARGET_PAGE_SIZE);
>      }
>  
>      /* Send XBZRLE based compressed page */
> -- 
> 1.7.12.4
> 
> 
> Best regards,
> -Gonglei
> 
> 
--
Dr. David Alan Gilbert / dgilbert@redhat.com / Manchester, UK

Re: [Qemu-devel] [PATCH 6/7] migraion: optimiztion xbzrle by reducing data copy

[Gonglei]

On 2014/2/28 18:30, Dr. David Alan Gilbert wrote:

> * Gonglei (Arei) (arei.gonglei@huawei.com) wrote:
>> Reducing data copy can reduce cpu overheah.
> 
> (Note a few typos in subject/title)
> 
>> Signed-off-by: ChenLiang <chenliang88@huawei.com>
>> Signed-off-by: Gonglei <arei.gonglei@huawei.com>
>> ---
>>  arch_init.c | 8 +++-----
>>  1 file changed, 3 insertions(+), 5 deletions(-)
>>
>> diff --git a/arch_init.c b/arch_init.c
>> index 2211e0b..cc88875 100644
>> --- a/arch_init.c
>> +++ b/arch_init.c
>> @@ -344,11 +344,8 @@ static int save_xbzrle_page(QEMUFile *f, uint8_t **current_data,
>>  
>>      prev_cached_page = get_cached_data(XBZRLE.cache, current_addr);
>>  
>> -    /* save current buffer into memory */
>> -    memcpy(XBZRLE.current_buf, *current_data, TARGET_PAGE_SIZE);
>> -
>>      /* XBZRLE encoding (if there is no overflow) */
>> -    encoded_len = xbzrle_encode_buffer(prev_cached_page, XBZRLE.current_buf,
>> +    encoded_len = xbzrle_encode_buffer(prev_cached_page, *current_data,
>>                                         TARGET_PAGE_SIZE, XBZRLE.encoded_buf,
>>                                         TARGET_PAGE_SIZE);
> 
> Is xbzrle_encode_buffer safe if the main memory is still being changed while
> it is run? Even with multiple CPUs changing it? Even on CPUs with looser
> memory ordering semantics?
> 

It's OK. We just need to assure the pages of cache and the pages which be sent coherence.

>>      if (encoded_len == 0) {
>> @@ -367,7 +364,8 @@ static int save_xbzrle_page(QEMUFile *f, uint8_t **current_data,
>>  
>>      /* we need to update the data in the cache, in order to get the same data */
>>      if (!last_stage) {
>> -        memcpy(prev_cached_page, XBZRLE.current_buf, TARGET_PAGE_SIZE);
>> +        xbzrle_decode_buffer(XBZRLE.encoded_buf, encoded_len, prev_cached_page,
>> +                                                              TARGET_PAGE_SIZE);
>>      }
>>  
>>      /* Send XBZRLE based compressed page */
>> -- 
>> 1.7.12.4
>>
>>
>> Best regards,
>> -Gonglei
>>
>>
> --
> Dr. David Alan Gilbert / dgilbert@redhat.com / Manchester, UK


Best regards,
-Gonglei