floraachy/OpenBLAS
1
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity

Initial support for SkylakeX / AVX512

Browse Source 
This patch adds the basic infrastructure for adding the SkylakeX (Intel Skylake server)
target. The SkylakeX target will use the AVX512 (AVX512VL level) instruction set,
which brings 2 basic things:
1) 512 bit wide SIMD (2x width of AVX2)
2) 32 SIMD registers (2x the number on AVX2)

This initial patch only contains a trivial transofrmation of the Haswell SGEMM kernel
to AVX512VL; more will follow later but this patch aims to get the infrastructure
in place for this "later".

Full performance tuning has not been done yet; with more registers and wider SIMD
it's in theory possible to retune the kernels but even without that there's an
interesting enough performance increase (30-40% range) with just this change.
This commit is contained in:
Arjan van de Ven
2018-06-03 07:24:29 +00:00
parent 36c4523d85
commit 99c7bba8e4
57 changed files with 7034 additions and 47 deletions
Download Patch File Download Diff File Expand all files Collapse all files

2
driver/others/dynamic.c
View File

@@ -74,6 +74,7 @@ extern gotoblas_t gotoblas_STEAMROLLER;
extern gotoblas_t gotoblas_EXCAVATOR;
#ifdef NO_AVX2
#define gotoblas_HASWELL gotoblas_SANDYBRIDGE
#define gotoblas_SKYLAKEX gotoblas_SANDYBRIDGE
#define gotoblas_ZEN gotoblas_SANDYBRIDGE
#else
extern gotoblas_t gotoblas_HASWELL;
@@ -83,6 +84,7 @@ extern gotoblas_t gotoblas_ZEN;
//Use NEHALEM kernels for sandy bridge
#define gotoblas_SANDYBRIDGE gotoblas_NEHALEM
#define gotoblas_HASWELL gotoblas_NEHALEM
#define gotoblas_SKYLAKEX gotoblas_NEHALEM
#define gotoblas_BULLDOZER gotoblas_BARCELONA
#define gotoblas_PILEDRIVER gotoblas_BARCELONA
#define gotoblas_STEAMROLLER gotoblas_BARCELONA

4
driver/others/parameter.c
View File

@@ -167,7 +167,7 @@ int get_L2_size(void){
#if defined(ATHLON) || defined(OPTERON) || defined(BARCELONA) || defined(BOBCAT) || defined(BULLDOZER) || \
defined(CORE_PRESCOTT) || defined(CORE_CORE2) || defined(PENRYN) || defined(DUNNINGTON) || \
defined(CORE_NEHALEM) || defined(CORE_SANDYBRIDGE) || defined(ATOM) || defined(GENERIC) || \
defined(PILEDRIVER) || defined(HASWELL) || defined(STEAMROLLER) || defined(EXCAVATOR) || defined(ZEN)
defined(PILEDRIVER) || defined(HASWELL) || defined(STEAMROLLER) || defined(EXCAVATOR) || defined(ZEN) || defined(SKYLAKEX)
cpuid(0x80000006, &eax, &ebx, &ecx, &edx);
@@ -251,7 +251,7 @@ int get_L2_size(void){
void blas_set_parameter(void){
int factor;
#if defined(BULLDOZER) || defined(PILEDRIVER) || defined(SANDYBRIDGE) || defined(NEHALEM) || defined(HASWELL) || defined(STEAMROLLER) || defined(EXCAVATOR) || defined(ZEN)
#if defined(BULLDOZER) || defined(PILEDRIVER) || defined(SANDYBRIDGE) || defined(NEHALEM) || defined(HASWELL) || defined(STEAMROLLER) || defined(EXCAVATOR) || defined(ZEN) || defined(SKYLAKEX)
int size = 16;
#else
int size = get_L2_size();