I think it's actually the other way around. Development is driven to the product that produces the largest net revenue stream. Windows costs the same whether it's installed on a Celeron or P4, but there should have been more Celerons sold than P4, hence optimization is for Celeron, not P4.
Additionally, it's unlikely that Microsoft has had to worry about throughput, since that's just a clarion call for a bigger processor with another copy of Windows installed. Compiler switches are notorious for not being set the way you intend; yet I've seen little evidence that there have been significant bugs caused by improper targeting of instruction sets between Celeron and Pentium.
Based on the the Intel documentation, they make zero distinction between Celeron and Pentium software architecture; there is only one IA-32 software and P-6 hardware architecture manual, and zero mention of Celeron as a special case. TTFN
Additionally, it's unlikely that Microsoft has had to worry about throughput, since that's just a clarion call for a bigger processor with another copy of Windows installed. Compiler switches are notorious for not being set the way you intend; yet I've seen little evidence that there have been significant bugs caused by improper targeting of instruction sets between Celeron and Pentium.
Based on the the Intel documentation, they make zero distinction between Celeron and Pentium software architecture; there is only one IA-32 software and P-6 hardware architecture manual, and zero mention of Celeron as a special case. TTFN
