A few points, if I may:

• On modern linux configured with transparent hugepages, you (likely) get hugepages for big segments without any special code or handling.
• Your 1st conservative unrolling is more accurately called loop peeling. The second form is complete loop peeling. To unroll a loop you merge N*loop bodies and iterate for (i=0; i<len; i += N) { BODY(i); BODY(i+1); ... }
• I expect very little benefits from unrolling the complex inner loop. (But certainly I'd test a 2x unroll. Sometimes weird effects may be at work.)
• Random memory access with hugepages is still on the order of 200 clocks or so. This time would be well spent on calculations. I'd therefore start with a few iterations of type a (=calculated), then switch over to type b (=lookup) once the prefetch has completed. Either loop might be unrolled.
• Better still: prefetch the data needed for next loop iteration, that is vecM[m7next]. It will then be ready at the right time. (*)
• Unrolling benefits may not carry over when hyperthreading is accounted for.
• The hashing step+modulus involves a chain of 3 muls, meaning latency > 3*4+some. Loop iterations may overlap, except quarter of the tests vs 500+-127 will mispredict. This suggests another option of calculating a temporary q8_to_d9[128] in a tight loop (that allows good overlap), followed by a loop with the tests.
• The q8_to_d9[] is a good candidate for vectorization. So is q9_to_c9[]. Inbetween, grep from d9 to q9. I think vectorized code will beat the lookup variant. Or not. Lookup+vectorized q9_to_c9 might be better. (*)
• Test with different compilers. gcc, clang.

All in all, this problem seems like a nice little assignment for a competent cryptanalyst.

Modulo cmps. Since (2**31)%1000003 == 477207, we know we may safely add small values to the hashval without fear of sign change or overflow. This allows one to rewrite tests such as
    v %= 1001; if (v < 0) v += 1001; if (v != 100) continue;
in a simpler, speedier form:
    v += 1001 - 100; if (v % 1001) continue;

(*) Updated.

---

---

Posts are HTML formatted. Put <p> </p> tags around your paragraphs. Put <code> </code> tags around your code and data!

Titles consisting of a single word are discouraged, and in most cases are disallowed outright.

Read Where should I post X? if you're not absolutely sure you're posting in the right place.

Please read these before you post! —

• How do I compose an effective node title?
• How do I post a question effectively?
• Markup in the Monastery

Posts may use any of the Perl Monks Approved HTML tags:

a, abbr, b, big, blockquote, br, caption, center, col, colgroup, dd, del, details, div, dl, dt, em, font, h1, h2, h3, h4, h5, h6, hr, i, ins, li, ol, p, pre, readmore, small, span, spoiler, strike, strong, sub, summary, sup, table, tbody, td, tfoot, th, thead, tr, tt, u, ul, wbr

You may need to use entities for some characters, as follows. (Exception: Within code tags, you can put the characters literally.)

	For:		Use:
	&		&
	<		&lt;
	>		&gt;
	[		&#91;
	]		&#93;

Link using PerlMonks shortcuts! What shortcuts can I use for linking?

See Writeup Formatting Tips and other pages linked from there for more info.