- Mark as New
- Bookmark
- Subscribe
- Mute
- Subscribe to RSS Feed
- Permalink
- Report Inappropriate Content
Hi,
I am currently measuring a database that is being benchmarked, and i use all instruction(x87,64 mmx, 128 mmx, packed single precision,...) events and ratio's.
The weird thing was that i don't get to 100% of instructions if i sum up the ratio's which have a formula like:
'(instructions of the specific kind/instructions retired)*100'.
If i look at the defenition of 'instructions of the specific kind', for instance '128-bit MMX Instructions Retired' it says it doesn not include the .. data movement instructions, in this case the 128-bit SIMD data movement instructions.
Is there an event or ratio which can measure data movement instructions somehow?
Also the formula for 'Branch Mispredict Performance Impact' is:
'((Mispredicted Branches Retired * 20) / Clockticks)*100'.
Is the '20' in the formula the length of the pipeline?
Thx
I am currently measuring a database that is being benchmarked, and i use all instruction(x87,64 mmx, 128 mmx, packed single precision,...) events and ratio's.
The weird thing was that i don't get to 100% of instructions if i sum up the ratio's which have a formula like:
'(instructions of the specific kind/instructions retired)*100'.
If i look at the defenition of 'instructions of the specific kind', for instance '128-bit MMX Instructions Retired' it says it doesn not include the .. data movement instructions, in this case the 128-bit SIMD data movement instructions.
Is there an event or ratio which can measure data movement instructions somehow?
Also the formula for 'Branch Mispredict Performance Impact' is:
'((Mispredicted Branches Retired * 20) / Clockticks)*100'.
Is the '20' in the formula the length of the pipeline?
Thx
Link Copied
0 Replies

Reply
Topic Options
- Subscribe to RSS Feed
- Mark Topic as New
- Mark Topic as Read
- Float this Topic for Current User
- Bookmark
- Subscribe
- Printer Friendly Page