3 ms·
The author seems to have ended up in a silly debate when defending the use of SSE3 (and not SSE2). "You are absolutely not going to 100% SSE2 out in the real w
by kyberias 6y ago
The author seems to have ended up in a silly debate when defending the use of SSE3 (and not SSE2).
"You are absolutely not going to 100% SSE2 out in the real world. In fact, you have many 16-bit x86 systems out there. There are people running companies on Apple II computers."
Right, "companies running business on Apple II computers" is a relevant argument for... absolutely nothing.
- cesarb 6y agoNote that the author is not defending the use of SSE3 (https://en.wikipedia.org/wiki/SSE3 https://en.wikipedia.org/wiki/SSE3), the author is defending the use of SSSE3 (https://en.wikipedia.org/wiki/SSSE3 https://en.wikipedia.org/wiki/SSSE3), which is something completely different. In fact, the whole point of this post is a novel use of the PSHUFB instruction, which comes from SSSE3. I happen to have a somewhat old laptop right next to me (it originally came with Windows Vista). According to /proc/cpuinfo, it does have SSE, SSE2, and SSSE3 (but not SSE3), so even that old laptop is already new enough to run the proposed algorithm.
- account42 6y agoAMD added SSSE3 with the FX processors, Phenom didn't have it. Not unreasonable to still expect there to be users without SSSE3 - e.g. across all Steam users over 1% don't have SSSE3. [0]. [0] https://store.steampowered.com/hwsurvey/ https://store.steampowered.com/hwsurvey/