[patch #8727] AX_EXT(ax_ext.m4) support for AVX2/AVX512 and other CPU extensions

[Michael Petch]

URL:
  <http://savannah.gnu.org/patch/?8727>

                 Summary: AX_EXT(ax_ext.m4) support for AVX2/AVX512 and other
CPU extensions
                 Project: GNU Autoconf Archive
            Submitted by: mdpetch
            Submitted on: Tue 25 Aug 2015 23:56:36 MDT
                Priority: 5 - Normal
                  Status: None
             Assigned to: None
        Originator Email: 
             Open/Closed: Open
         Discussion Lock: Any

    _______________________________________________________

Details:

Attached is a patch for AX_EXT(ax_ext.m4) that should remain backwards
compatible with the current version (serial 13) in git with support for
AVX/AVX2/AVX512 as well as existing SIMD extensions. There is also a new
environment variable returned - AC_SUBST(CPUEXT_FLAGS) . CPUEXT_FLAGS works in
the same fashion as SIMD_FLAGS except it contains a list of GCC compiler
options to enable the non-SIMD extensions that a processor may support.

Proper OS and CPU checking is performed as well as a compiler test.

This patch really needs to be tested on newer architectures as I didn't have
access to AVX2 and AVX512 systems. The checks are based upon information
gathered across Intel specifications. This type of functionality has been
requested a lot over the past year so anyone who stumbles on this patch
submission can try it out.

The defines current supported are: 

AH_TEMPLATE([HAVE_MMX],[Define to 1 to support Multimedia Extensions])
  AH_TEMPLATE([HAVE_RDRND],[Define to 1 to support Digital Random Number
Generator])
  AH_TEMPLATE([HAVE_BMI1],[Define to 1 to support Bit Manipulation
Instruction Set 1])
  AH_TEMPLATE([HAVE_BMI2],[Define to 1 to support Bit Manipulation
Instruction Set 2])
  AH_TEMPLATE([HAVE_ADX],[Define to 1 to support Multi-Precision
Add-Carry Instruction Extensions])
  AH_TEMPLATE([HAVE_MPX],[Define to 1 to support Memory Protection
Extensions])
  AH_TEMPLATE([HAVE_PREFETCHWT1],[Define to 1 to support Prefetch Vector
Data Into Caches WT1])
  AH_TEMPLATE([HAVE_ABM],[Define to 1 to support Advanced Bit Manipulation])
  AH_TEMPLATE([HAVE_SSE],[Define to 1 to support Streaming SIMD Extensions])
  AH_TEMPLATE([HAVE_SSE2],[Define to 1 to support Streaming SIMD
Extensions])
  AH_TEMPLATE([HAVE_SSE3],[Define to 1 to support Streaming SIMD
Extensions 3])
  AH_TEMPLATE([HAVE_SSSE3],[Define to 1 to support Supplemental
Streaming SIMD Extensions 3])
  AH_TEMPLATE([HAVE_SSE4_1],[Define to 1 to support Streaming SIMD
Extensions 4.1])
  AH_TEMPLATE([HAVE_SSE4_2],[Define to 1 to support Streaming SIMD
Extensions 4.2])
  AH_TEMPLATE([HAVE_SSE4a],[Define to 1 to support AMD Streaming SIMD
Extensions 4a])
  AH_TEMPLATE([HAVE_SHA],[Define to 1 to support Secure Hash Algorithm
Extension])
  AH_TEMPLATE([HAVE_AES],[Define to 1 to support Advanced Encryption
Standard New Instruction Set (AES-NI)])
  AH_TEMPLATE([HAVE_AVX],[Define to 1 to support Advanced Vector
Extensions])
  AH_TEMPLATE([HAVE_FMA3],[Define to 1 to support  Fused Multiply-Add
Extensions 3])
  AH_TEMPLATE([HAVE_FMA4],[Define to 1 to support Fused Multiply-Add
Extensions 3])
  AH_TEMPLATE([HAVE_XOP],[Define to 1 to support eXtended Operations
Extensions])
  AH_TEMPLATE([HAVE_AVX2],[Define to 1 to support Advanced Vector
Extensions 2])
  AH_TEMPLATE([HAVE_AVX512_F],[Define to 1 to support AVX-512 Foundation
Extensions])
  AH_TEMPLATE([HAVE_AVX512_CD],[Define to 1 to support AVX-512 Conflict
Detection Instructions])
  AH_TEMPLATE([HAVE_AVX512_PF],[Define to 1 to support AVX-512 Conflict
Prefetch Instructions])
  AH_TEMPLATE([HAVE_AVX512_ER],[Define to 1 to support AVX-512
Exponential & Reciprocal Instructions])
  AH_TEMPLATE([HAVE_AVX512_VL],[Define to 1 to support AVX-512 Vector
Length Extensions])
  AH_TEMPLATE([HAVE_AVX512_BW],[Define to 1 to support AVX-512 Byte and
Word Instructions])
  AH_TEMPLATE([HAVE_AVX512_DQ],[Define to 1 to support AVX-512
Doubleword and Quadword Instructions])
  AH_TEMPLATE([HAVE_AVX512_IFMA],[Define to 1 to support AVX-512 Integer
Fused Multiply Add Instructions])
  AH_TEMPLATE([HAVE_AVX512_VBMI],[Define to 1 to support AVX-512 Vector
Byte Manipulation Instructions])



    _______________________________________________________

File Attachments:


-------------------------------------------------------
Date: Tue 25 Aug 2015 23:56:36 MDT  Name: ax_ext.patch  Size: 20kB   By:
mdpetch
ax_ext.m4 patch to be applied to Serial 13 to support AVX/AVX/AVX512
<http://savannah.gnu.org/patch/download.php?file_id=34731>

    _______________________________________________________

Reply to this item at:

  <http://savannah.gnu.org/patch/?8727>

_______________________________________________
  Message sent via/by Savannah
  http://savannah.gnu.org/