Timeline for Converting Array of Floats to UINT8 (`char`) or UINT16 (`unsigned short`) Using SSE4
Current License: CC BY-SA 3.0
21 events
| when toggle format | what | by | license | comment | |
|---|---|---|---|---|---|
| Apr 26, 2019 at 7:11 | comment | added | Royi |
Similar question with AVX2 - codereview.stackexchange.com/questions/219158.
|
|
| Oct 22, 2017 at 20:12 | answer | added | Royi | timeline score: 1 | |
| Oct 22, 2017 at 19:47 | vote | accept | Royi | ||
| Oct 22, 2017 at 15:49 | comment | added | Peter Cordes | See my answer on the original question you already linked (stackoverflow.com/a/34076778/224132) for the full version of that comment. And also for 4->1 vector packing exactly the same way Harold's answer is doing it, and yes I still think that's optimal. | |
| Oct 22, 2017 at 15:42 | comment | added | Royi | @PeterCordes, Any remarks on harold's solution? Could you elaborate on your remark? Thank You. | |
| Oct 22, 2017 at 9:17 | history | tweeted | twitter.com/StackCodeReview/status/922029171580301314 | ||
| Oct 22, 2017 at 6:36 | comment | added | Royi | @MartinR, Got it. I can see the logic in that. Thank You. | |
| Oct 22, 2017 at 6:16 | comment | added | Martin R | That issue was addressed in an answer as "Get Rid of Unused Arguments". This site is quite strict about not editing a question in a way that might invalidate any part of an existing answer. | |
| Oct 22, 2017 at 6:11 | comment | added | Royi |
@MartinR, nothing changed in the code. I made mistake adding something not used mF and mB. It doesn't change the validity of any answer. It was done because it made user1118321 confused.
|
|
| Oct 22, 2017 at 5:53 | history | rollback | Martin R |
Rollback to Revision 4
|
|
| Oct 22, 2017 at 5:53 | comment | added | Martin R | I have rolled back the last edit. Please do not add, remove, or edit code in a question after you've received an answer. The site policy is explained in What to do when someone answers. | |
| Oct 22, 2017 at 5:48 | history | edited | Royi | CC BY-SA 3.0 |
deleted 53 characters in body
|
| Oct 22, 2017 at 5:23 | answer | added | user1118321 | timeline score: 2 | |
| Oct 22, 2017 at 2:10 | comment | added | Peter Cordes |
_mm_packus_epi16 still treats its input as signed, so the first step in a 2-step pack should be packss to create input for packus. And yes, of course you should be using each pack with 2 different inputs.
|
|
| Oct 21, 2017 at 21:57 | answer | added | user555045 | timeline score: 4 | |
| Oct 21, 2017 at 21:12 | history | edited | Royi | CC BY-SA 3.0 |
added 120 characters in body
|
| Oct 21, 2017 at 21:10 | comment | added | Royi | @harold It should be limited in the range [0, 255] (Updated assumptions accordingly). Feel free to offer a different code. I'm pretty struggling with it. Thank You. | |
| Oct 21, 2017 at 20:32 | comment | added | user555045 |
What should happen with the saturation? I shouldn't have put _mm_packus_epi32 first probably.
|
|
| Oct 21, 2017 at 20:26 | history | edited | Royi | CC BY-SA 3.0 |
edited title
|
| Oct 21, 2017 at 20:22 | history | edited | 200_success | CC BY-SA 3.0 |
deleted 46 characters in body; edited tags; edited title
|
| Oct 21, 2017 at 19:36 | history | asked | Royi | CC BY-SA 3.0 |