Add support for cmpeq and cmpgt MMX intrinsics

Add primop support for the following MMX intrinsics:

__m64 _mm_cmpeq_pi16 (__m64 a, __m64 b)
__m64 _mm_cmpeq_pi32 (__m64 a, __m64 b)
__m64 _mm_cmpeq_pi8 (__m64 a, __m64 b)
__m64 _mm_cmpeq_pi8 (__m64 a, __m64 b)
__m64 _mm_cmpgt_pi16 (__m64 a, __m64 b)
__m64 _mm_cmpgt_pi32 (__m64 a, __m64 b)
__m64 _mm_cmpgt_pi8 (__m64 a, __m64 b)

Trac metadata

Trac field	Value
Version	8.4.3
Type	Task
TypeOfFailure	OtherFailure
Priority	normal
Resolution	Unresolved
Component	Compiler
Test case
Differential revisions
BlockedBy
Related
Blocking
CC
Operating system
Architecture

changed milestone to %8.6.1

changed weight to 5

added Ttask Trac import labels

More information about these intrinsics here:

https://software.intel.com/sites/landingpage/IntrinsicsGuide/#text=cmp&expand=765,767,802&cats=Compare&techs=MMX

I need this instruction to optimise my succinct data structure based CSV parser. If this instruction was available to me, I would be able to replace code that generates six instructions with code that generates one.

https://github.com/haskell-works/hw-dsv/pull/12/files

assigned to @newhoggy

changed milestone to %8.8.1

These won't be addressed in GHC 8.6.

For motivation, SIMD instructions are really good for writing very high performance parsers. For example, the following C example shows a short C program using AVX2 intrinsics to produce two Rank-Select-Bit-Strings from a CSV file that can be used to index into the CSV file:

https://github.com/haskell-works/hw-dsv/tree/simd

The program is able to index a 7G CSV file in under 4s:

$ time ./simd
./simd  1.10s user 2.34s system 89% cpu 3.840 total

To implement this in pure haskell, I would need equivalent primops for the following CPU intrinsics:

_mm256_set1_epi8
_mm256_cmpeq_epi8
_mm256_movemask_epi8
_mm256_cmpeq_epi8
_mm256_movemask_epi8
_mm256_maskload_epi64
_mm256_maskstore_epi64

See: https://software.intel.com/sites/landingpage/IntrinsicsGuide/#text=vpmaskmovq&expand=765,767,802,5241,3295,3296,3297

changed milestone to %8.10.1

Bumping milestones of low-priority tickets.

added Pnormal label

added SIMD Tbug Tfeature request labels

removed milestone

removed Tbug label

mentioned in issue #25643