- Mark as New
- Bookmark
- Subscribe
- Mute
- Subscribe to RSS Feed
- Permalink
- Report Inappropriate Content
The VTune docs describes this metric as "how the application code vectorization relates to the floating point computations. A value of 100% means that all floating point instructions are vectorized with the full vector capacity."
The last phrase "full vector capacity" is bugging me. What exactly does it mean? Full capacity of what - Vector hardware or vector instructions?
Harsh
- Mark as New
- Bookmark
- Subscribe
- Mute
- Subscribe to RSS Feed
- Permalink
- Report Inappropriate Content
Hello,
"Vector Capacity Usage" metric was an attempt to show both vectorizaton efficiency and SIMD utilization efficiency in one metric. For example, if you have 100% of FP instructions vectorized, but the vector instructions used in the application are 128-bit with the processor capable to execute 256-bit instructions - Vector Capacity metric will show 50%. If 50% of FP instructions are vectorized and vector instructions used in the application are 128-bit with the processor capable to execute 256-bit instructions - Vector Capacity metric will show 25%. In the latest product versions this metric was replaced with more straightforward metrics.
Thanks & Regards, Dmitry
Link Copied
- Mark as New
- Bookmark
- Subscribe
- Mute
- Subscribe to RSS Feed
- Permalink
- Report Inappropriate Content
Hello,
"Vector Capacity Usage" metric was an attempt to show both vectorizaton efficiency and SIMD utilization efficiency in one metric. For example, if you have 100% of FP instructions vectorized, but the vector instructions used in the application are 128-bit with the processor capable to execute 256-bit instructions - Vector Capacity metric will show 50%. If 50% of FP instructions are vectorized and vector instructions used in the application are 128-bit with the processor capable to execute 256-bit instructions - Vector Capacity metric will show 25%. In the latest product versions this metric was replaced with more straightforward metrics.
Thanks & Regards, Dmitry
- Mark as New
- Bookmark
- Subscribe
- Mute
- Subscribe to RSS Feed
- Permalink
- Report Inappropriate Content
Also, for in-depth SIMD / Vectorization analysis you can use Intel (Vectorization) Advisor, available for individual download or as a part of oneAPI base toolkit/Parallel Studio
See:
https://software.intel.com/content/www/us/en/develop/tools/oneapi/components/advisor.html
- Mark as New
- Bookmark
- Subscribe
- Mute
- Subscribe to RSS Feed
- Permalink
- Report Inappropriate Content
Hi HarshVardhanKumar,
We have not heard from you. Has the solution provided helped? Could you please give us an update?
Thanks
- Mark as New
- Bookmark
- Subscribe
- Mute
- Subscribe to RSS Feed
- Permalink
- Report Inappropriate Content
The provided solution has helped. I've selected the answer which helped me as the accepted solution. Thanks.
- Mark as New
- Bookmark
- Subscribe
- Mute
- Subscribe to RSS Feed
- Permalink
- Report Inappropriate Content
Hi,
Thanks for accepting our solution. If you need any additional information, please post a new question as this thread will no longer be monitored by Intel.
Thanks
- Subscribe to RSS Feed
- Mark Topic as New
- Mark Topic as Read
- Float this Topic for Current User
- Bookmark
- Subscribe
- Printer Friendly Page