Created attachment 188520 [details]
patch

Support blendpd (Blend Packed Double Precision Floating-Point
Values (XMM)) instruction in guest_x86_toIR.c. To be able to
use amd64 math_BLENDPD_128 function for x86 implementation, add
a new VEX/priv/guest_generic_sse.h header and move math_BLENDPD_128
there.

Add test function to sse4-common.h and update none/tests/x86/sse4-x86.c
to test the instruction.

Created attachment 188530 [details]
patch

I see you added priv/guest_generic_sse.h to Makefile.vex.am, good (although odd indentation).
It is missing a NEWS (bug) entry. 
Why are the guest_generic_sse.h includes in the middle of the files?
Can they be moved up to where we include other generic headers like guest_generic_x87.h?

(In reply to Mark Wielaard from comment #3)
> Why are the guest_generic_sse.h includes in the middle of the files?
> Can they be moved up to where we include other generic headers like
> guest_generic_x87.h?

OK, I see because you need newTemp, assign, mkV128, binop, unop, mkexpr, etc.
hmmm, maybe it isn't worth it to create a guest_generic_x87.h then, just keep the
math_BLENDPD_128 in guest_amd64_toIR.c and repeat it in guest_x86_toIR.c?

I don't really mind it being included at that odd place in the middle if you think
it will help reduce redundancy in the future for other shared sse shared functions.
But in that case please do document it clearly to list all requirements before it
can be included.

Created attachment 188756 [details]
patch