Initial support for SkylakeX / AVX512

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.

Makefile.system

@@ -62,6 +62,9 @@ ifeq ($(BINARY), 32)
 ifeq ($(TARGET), HASWELL)
 GETARCH_FLAGS := -DFORCE_NEHALEM
 endif
+ifeq ($(TARGET), SKYLAKEX)
+GETARCH_FLAGS := -DFORCE_NEHALEM
+endif
 ifeq ($(TARGET), SANDYBRIDGE)
 GETARCH_FLAGS := -DFORCE_NEHALEM
 endif
@@ -95,6 +98,9 @@ ifeq ($(BINARY), 32)
 ifeq ($(TARGET_CORE), HASWELL)
 GETARCH_FLAGS := -DFORCE_NEHALEM
 endif
+ifeq ($(TARGET_CORE), SKYLAKEX)
+GETARCH_FLAGS := -DFORCE_NEHALEM
+endif
 ifeq ($(TARGET_CORE), SANDYBRIDGE)
 GETARCH_FLAGS := -DFORCE_NEHALEM
 endif
@@ -467,7 +473,7 @@ ifneq ($(NO_AVX), 1)
 DYNAMIC_CORE += SANDYBRIDGE BULLDOZER PILEDRIVER STEAMROLLER EXCAVATOR
 endif
 ifneq ($(NO_AVX2), 1)
-DYNAMIC_CORE += HASWELL ZEN
+DYNAMIC_CORE += HASWELL ZEN SKYLAKEX
 endif
 endif
 

TargetList.txt

@@ -20,6 +20,7 @@ DUNNINGTON
 NEHALEM
 SANDYBRIDGE
 HASWELL
+SKYLAKEX
 ATOM
 
 b)AMD CPU:

cmake/arch.cmake

@@ -56,6 +56,9 @@ if (DYNAMIC_ARCH)
     if (NOT NO_AVX2)
       set(DYNAMIC_CORE ${DYNAMIC_CORE} HASWELL ZEN)
     endif ()
+    if (NOT NO_AVX512)
+      set(DYNAMIC_CORE ${DYNAMIC_CORE} SKYLAKEX)
+    endif ()
   endif ()
 
   if (NOT DYNAMIC_CORE)

cmake/system.cmake

@@ -33,7 +33,7 @@ endif ()
 if (DEFINED BINARY AND DEFINED TARGET AND BINARY EQUAL 32)
   message(STATUS "Compiling a ${BINARY}-bit binary.")
   set(NO_AVX 1)
-  if (${TARGET} STREQUAL "HASWELL" OR ${TARGET} STREQUAL "SANDYBRIDGE")
+  if (${TARGET} STREQUAL "HASWELL" OR ${TARGET} STREQUAL "SANDYBRIDGE" OR ${TARGET} STREQUAL "SKYLAKEX")
     set(TARGET "NEHALEM")
   endif ()
   if (${TARGET} STREQUAL "BULLDOZER" OR ${TARGET} STREQUAL "PILEDRIVER" OR ${TARGET} STREQUAL "ZEN")

cpuid.h

@@ -115,6 +115,7 @@
 #define CORE_STEAMROLLER 25
 #define CORE_EXCAVATOR   26
 #define CORE_ZEN         27
+#define CORE_SKYLAKEX    28
 
 #define HAVE_SSE      (1 <<  0)
 #define HAVE_SSE2     (1 <<  1)
@@ -137,6 +138,7 @@
 #define HAVE_AVX      (1 <<  18)
 #define HAVE_FMA4     (1 <<  19)
 #define HAVE_FMA3     (1 <<  20)
+#define HAVE_AVX512VL (1 <<  21)
 
 #define CACHE_INFO_L1_I     1
 #define CACHE_INFO_L1_D     2
@@ -211,5 +213,6 @@ typedef struct {
 #define CPUTYPE_STEAMROLLER 		49
 #define CPUTYPE_EXCAVATOR 		50
 #define CPUTYPE_ZEN 			51
+#define CPUTYPE_SKYLAKEX		52
 
 #endif

cpuid_x86.c

@@ -50,6 +50,8 @@
 #ifdef NO_AVX
 #define CPUTYPE_HASWELL CPUTYPE_NEHALEM
 #define CORE_HASWELL CORE_NEHALEM
+#define CPUTYPE_SKYLAKEX CPUTYPE_NEHALEM
+#define CORE_SKYLAKEX CORE_NEHALEM
 #define CPUTYPE_SANDYBRIDGE CPUTYPE_NEHALEM
 #define CORE_SANDYBRIDGE CORE_NEHALEM
 #define CPUTYPE_BULLDOZER CPUTYPE_BARCELONA

driver/others/dynamic.c

@@ -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

driver/others/parameter.c

@@ -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();

getarch.c

@@ -326,6 +326,21 @@ USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
 #define CORENAME  "HASWELL"
 #endif
 
+#ifdef FORCE_SKYLAKEX
+#define FORCE
+#define FORCE_INTEL
+#define ARCHITECTURE    "X86"
+#define SUBARCHITECTURE "SKYLAKEX"
+#define ARCHCONFIG   "-DSKYLAKEX " \
+		     "-DL1_DATA_SIZE=32768 -DL1_DATA_LINESIZE=64 " \
+		     "-DL2_SIZE=262144 -DL2_LINESIZE=64 " \
+		     "-DDTB_DEFAULT_ENTRIES=64 -DDTB_SIZE=4096 " \
+		     "-DHAVE_CMOV -DHAVE_MMX -DHAVE_SSE -DHAVE_SSE2 -DHAVE_SSE3 -DHAVE_SSSE3 -DHAVE_SSE4_1 -DHAVE_SSE4_2 -DHAVE_AVX " \
+                     "-DFMA3 -DHAVE_AVX512VL -march=skylake-avx512"
+#define LIBNAME   "skylakex"
+#define CORENAME  "SKYLAKEX"
+#endif
+
 #ifdef FORCE_ATOM
 #define FORCE
 #define FORCE_INTEL

kernel/CMakeLists.txt

@@ -121,7 +121,7 @@ function (build_core TARGET_CORE KDIR TSUFFIX KERNEL_DEFINITIONS)
     # Makefile.L3
     set(USE_TRMM false)
 
-    if (ARM OR ARM64 OR "${TARGET_CORE}" STREQUAL "LONGSOON3B" OR "${TARGET_CORE}" STREQUAL "GENERIC" OR "${CORE}" STREQUAL "generic" OR "${TARGET_CORE}" STREQUAL "HASWELL" OR "${CORE}" STREQUAL "haswell" OR "${CORE}" STREQUAL "zen")
+    if (ARM OR ARM64 OR "${TARGET_CORE}" STREQUAL "LONGSOON3B" OR "${TARGET_CORE}" STREQUAL "GENERIC" OR "${CORE}" STREQUAL "generic" OR "${TARGET_CORE}" STREQUAL "HASWELL" OR "${CORE}" STREQUAL "haswell" OR "${CORE}" STREQUAL "zen" OR "${TARGET_CORE}" STREQUAL "SKYLAKEX" OR "${CORE}" STREQUAL "skylakex")
       set(USE_TRMM true)
     endif ()
 

kernel/Makefile.L3

@@ -32,6 +32,10 @@ ifeq ($(CORE), HASWELL)
 USE_TRMM = 1
 endif
 
+ifeq ($(CORE), SKYLAKEX)
+USE_TRMM = 1
+endif
+
 ifeq ($(CORE), ZEN)
 USE_TRMM = 1
 endif

kernel/setparam-ref.c

@@ -871,6 +871,22 @@ static void init_parameter(void) {
 #endif
 #endif
 
+#ifdef SKYLAKEX
+
+#ifdef DEBUG
+  fprintf(stderr, "SkylakeX\n");
+#endif
+
+  TABLE_NAME.sgemm_p = SGEMM_DEFAULT_P;
+  TABLE_NAME.dgemm_p = DGEMM_DEFAULT_P;
+  TABLE_NAME.cgemm_p = CGEMM_DEFAULT_P;
+  TABLE_NAME.zgemm_p = ZGEMM_DEFAULT_P;
+#ifdef EXPRECISION
+  TABLE_NAME.qgemm_p = QGEMM_DEFAULT_P;
+  TABLE_NAME.xgemm_p = XGEMM_DEFAULT_P;
+#endif
+#endif
+
 
 #ifdef OPTERON
 

kernel/x86/trsm_kernel_LN_2x4_penryn.S

@@ -62,7 +62,7 @@
 #define PREFETCHSIZE  (8 * 21 + 4)
 #endif
 
-#if defined(NEHALEM) || defined(SANDYBRIDGE) || defined(HASWELL)
+#if defined(NEHALEM) || defined(SANDYBRIDGE) || defined(HASWELL) || defined (SKYLAKEX)
 #define PREFETCH     prefetcht0
 #define PREFETCHSIZE  (8 * 21 + 4)
 #endif

kernel/x86/trsm_kernel_LN_4x4_penryn.S

@@ -62,7 +62,7 @@
 #define PREFETCHSIZE  (8 * 21 + 4)
 #endif
 
-#if defined(NEHALEM) || defined(SANDYBRIDGE) || defined(HASWELL)
+#if defined(NEHALEM) || defined(SANDYBRIDGE) || defined(HASWELL) || defined (SKYLAKEX)
 #define PREFETCH     prefetcht0
 #define PREFETCHSIZE  (8 * 21 + 4)
 #endif

kernel/x86/trsm_kernel_LT_2x4_penryn.S

@@ -62,7 +62,7 @@
 #define PREFETCHSIZE  (8 * 21 + 4)
 #endif
 
-#if defined(NEHALEM) || defined(SANDYBRIDGE) || defined(HASWELL)
+#if defined(NEHALEM) || defined(SANDYBRIDGE) || defined(HASWELL) || defined (SKYLAKEX)
 #define PREFETCH     prefetcht0
 #define PREFETCHSIZE  (8 * 21 + 4)
 #endif

kernel/x86/trsm_kernel_LT_4x4_penryn.S

@@ -62,7 +62,7 @@
 #define PREFETCHSIZE  (8 * 21 + 4)
 #endif
 
-#if defined(NEHALEM) || defined(SANDYBRIDGE) || defined(HASWELL)
+#if defined(NEHALEM) || defined(SANDYBRIDGE) || defined(HASWELL || defined (SKYLAKEX))
 #define PREFETCH     prefetcht0
 #define PREFETCHSIZE  (8 * 21 + 4)
 #endif

kernel/x86/trsm_kernel_RT_2x4_penryn.S

@@ -62,7 +62,7 @@
 #define PREFETCHSIZE  (8 * 21 + 4)
 #endif
 
-#if defined(NEHALEM) || defined(SANDYBRIDGE) || defined(HASWELL)
+#if defined(NEHALEM) || defined(SANDYBRIDGE) || defined(HASWELL) || defined (SKYLAKEX)
 #define PREFETCH     prefetcht0
 #define PREFETCHSIZE  (8 * 21 + 4)
 #endif

kernel/x86/trsm_kernel_RT_4x4_penryn.S

@@ -62,7 +62,7 @@
 #define PREFETCHSIZE  (8 * 21 + 4)
 #endif
 
-#if defined(NEHALEM) || defined(SANDYBRIDGE) || defined(HASWELL)
+#if defined(NEHALEM) || defined(SANDYBRIDGE) || defined(HASWELL) || defined (SKYLAKEX)
 #define PREFETCH     prefetcht0
 #define PREFETCHSIZE  (8 * 21 + 4)
 #endif

kernel/x86/ztrsm_kernel_LN_2x2_penryn.S

@@ -61,7 +61,7 @@
 #define PREFETCHSIZE 84
 #endif
 
-#if defined(NEHALEM) || defined(SANDYBRIDGE) || defined(HASWELL)
+#if defined(NEHALEM) || defined(SANDYBRIDGE) || defined(HASWELL) || defined (SKYLAKEX)
 #define PREFETCH	prefetcht1
 #define PREFETCHSIZE 84
 #endif

kernel/x86/ztrsm_kernel_LT_1x2_penryn.S

@@ -63,7 +63,7 @@
 #define PREFETCHSIZE 84
 #endif
 
-#if defined(NEHALEM) || defined(SANDYBRIDGE) || defined(HASWELL)
+#if defined(NEHALEM) || defined(SANDYBRIDGE) || defined(HASWELL) || defined (SKYLAKEX)
 #define PREFETCH	prefetcht1
 #define PREFETCHSIZE 84
 #endif

kernel/x86/ztrsm_kernel_LT_2x2_penryn.S

@@ -61,7 +61,7 @@
 #define PREFETCHSIZE 84
 #endif
 
-#if defined(NEHALEM) || defined(SANDYBRIDGE) || defined(HASWELL)
+#if defined(NEHALEM) || defined(SANDYBRIDGE) || defined(HASWELL) || defined (SKYLAKEX)
 #define PREFETCH	prefetcht1
 #define PREFETCHSIZE 84
 #endif

kernel/x86/ztrsm_kernel_RT_1x2_penryn.S

@@ -63,7 +63,7 @@
 #define PREFETCHSIZE 84
 #endif
 
-#if defined(NEHALEM) || defined(SANDYBRIDGE) || defined(HASWELL)
+#if defined(NEHALEM) || defined(SANDYBRIDGE) || defined(HASWELL) || defined (SKYLAKEX)
 #define PREFETCH	prefetcht1
 #define PREFETCHSIZE 84
 #endif

kernel/x86/ztrsm_kernel_RT_2x2_penryn.S

@@ -61,7 +61,7 @@
 #define PREFETCHSIZE 84
 #endif
 
-#if defined(NEHALEM) || defined(SANDYBRIDGE) || defined(HASWELL)
+#if defined(NEHALEM) || defined(SANDYBRIDGE) || defined(HASWELL) || defined (SKYLAKEX)
 #define PREFETCH	prefetcht1
 #define PREFETCHSIZE 84
 #endif

kernel/x86_64/KERNEL.SKYLAKEX

@@ -0,0 +1,4 @@
+include $(KERNELDIR)/KERNEL.HASWELL
+
+SGEMMKERNEL    =  sgemm_kernel_16x4_skylakex.S
+

kernel/x86_64/caxpy.c

@@ -33,7 +33,7 @@ USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
 #include "caxpy_microk_steamroller-2.c"
 #elif defined(BULLDOZER)
 #include "caxpy_microk_bulldozer-2.c"
-#elif defined(HASWELL) || defined(ZEN)
+#elif defined(HASWELL) || defined(ZEN) || defined(SKYLAKEX)
 #include "caxpy_microk_haswell-2.c"
 #elif defined(SANDYBRIDGE)
 #include "caxpy_microk_sandy-2.c"

kernel/x86_64/cdot.c

@@ -34,7 +34,7 @@ USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
 #include "cdot_microk_bulldozer-2.c"
 #elif defined(STEAMROLLER) || defined(PILEDRIVER)  || defined(EXCAVATOR)
 #include "cdot_microk_steamroller-2.c"
-#elif defined(HASWELL) || defined(ZEN)
+#elif defined(HASWELL) || defined(ZEN) || defined (SKYLAKEX)
 #include "cdot_microk_haswell-2.c"
 #elif defined(SANDYBRIDGE)
 #include "cdot_microk_sandy-2.c"

kernel/x86_64/cgemv_n_4.c

@@ -29,7 +29,7 @@ USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
 #include <stdio.h>
 #include "common.h"
 
-#if defined(HASWELL) || defined(ZEN)
+#if defined(HASWELL) || defined(ZEN) || defined (SKYLAKEX)
 #include "cgemv_n_microk_haswell-4.c"
 #elif defined(BULLDOZER) || defined(PILEDRIVER) || defined(STEAMROLLER) || defined(EXCAVATOR)
 #include "cgemv_n_microk_bulldozer-4.c"

kernel/x86_64/cgemv_t_4.c

@@ -28,7 +28,7 @@ USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
 
 #include "common.h"
 
-#if defined(HASWELL) || defined(ZEN)
+#if defined(HASWELL) || defined(ZEN) || defined (SKYLAKEX)
 #include "cgemv_t_microk_haswell-4.c"
 #elif defined(BULLDOZER) || defined(PILEDRIVER) || defined(STEAMROLLER)  || defined(EXCAVATOR)
 #include "cgemv_t_microk_bulldozer-4.c"

kernel/x86_64/cscal.c

@@ -28,7 +28,7 @@ USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
 #include "common.h"
 
 
-#if defined(HASWELL) || defined(ZEN)
+#if defined(HASWELL) || defined(ZEN) || defined (SKYLAKEX)
 #include "cscal_microk_haswell-2.c"
 #elif defined(BULLDOZER)  || defined(PILEDRIVER)
 #include "cscal_microk_bulldozer-2.c"

kernel/x86_64/daxpy.c

@@ -37,7 +37,7 @@ USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
 #include "daxpy_microk_steamroller-2.c"
 #elif defined(PILEDRIVER)
 #include "daxpy_microk_piledriver-2.c"
-#elif defined(HASWELL) || defined(ZEN)
+#elif defined(HASWELL) || defined(ZEN) || defined (SKYLAKEX)
 #include "daxpy_microk_haswell-2.c"
 #elif defined(SANDYBRIDGE)
 #include "daxpy_microk_sandy-2.c"

kernel/x86_64/ddot.c

@@ -37,7 +37,7 @@ USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
 #include "ddot_microk_piledriver-2.c"
 #elif defined(NEHALEM) 
 #include "ddot_microk_nehalem-2.c"
-#elif defined(HASWELL) || defined(ZEN)
+#elif defined(HASWELL) || defined(ZEN) || defined (SKYLAKEX)
 #include "ddot_microk_haswell-2.c"
 #elif defined(SANDYBRIDGE)
 #include "ddot_microk_sandy-2.c"

kernel/x86_64/dgemv_n_4.c

@@ -31,7 +31,7 @@ USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
 
 #if defined(NEHALEM)
 #include "dgemv_n_microk_nehalem-4.c"
-#elif defined(HASWELL) || defined(ZEN) || defined(STEAMROLLER) || defined(EXCAVATOR)
+#elif defined(HASWELL) || defined(ZEN) || defined(STEAMROLLER) || defined(EXCAVATOR) || defined (SKYLAKEX)
 #include "dgemv_n_microk_haswell-4.c"
 #endif
 

kernel/x86_64/dgemv_t_4.c

@@ -28,7 +28,7 @@ USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
 
 #include "common.h"
 
-#if defined(HASWELL) || defined(ZEN) || defined(STEAMROLLER)  || defined(EXCAVATOR)
+#if defined(HASWELL) || defined(ZEN) || defined(STEAMROLLER)  || defined(EXCAVATOR) || defined (SKYLAKEX)
 #include "dgemv_t_microk_haswell-4.c"
 #endif
 

kernel/x86_64/dscal.c

@@ -31,7 +31,7 @@ USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
 #include "dscal_microk_bulldozer-2.c"
 #elif defined(SANDYBRIDGE)
 #include "dscal_microk_sandy-2.c"
-#elif defined(HASWELL) || defined(ZEN)
+#elif defined(HASWELL) || defined(ZEN) || defined (SKYLAKEX)
 #include "dscal_microk_haswell-2.c"
 #endif
 

kernel/x86_64/dsymv_L.c

@@ -30,7 +30,7 @@ USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
 
 #if defined(BULLDOZER) || defined(PILEDRIVER) || defined(STEAMROLLER) || defined(EXCAVATOR)
 #include "dsymv_L_microk_bulldozer-2.c"
-#elif defined(HASWELL) || defined(ZEN)
+#elif defined(HASWELL) || defined(ZEN) || defined (SKYLAKEX)
 #include "dsymv_L_microk_haswell-2.c"
 #elif defined(SANDYBRIDGE)
 #include "dsymv_L_microk_sandy-2.c"

kernel/x86_64/dsymv_U.c

@@ -31,7 +31,7 @@ USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
 
 #if defined(BULLDOZER) || defined(PILEDRIVER) || defined(STEAMROLLER)  || defined(EXCAVATOR)
 #include "dsymv_U_microk_bulldozer-2.c"
-#elif defined(HASWELL) || defined(ZEN)
+#elif defined(HASWELL) || defined(ZEN) || defined (SKYLAKEX)
 #include "dsymv_U_microk_haswell-2.c"
 #elif defined(SANDYBRIDGE)
 #include "dsymv_U_microk_sandy-2.c"

kernel/x86_64/saxpy.c

@@ -31,7 +31,7 @@ USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
 
 #if defined(NEHALEM)
 #include "saxpy_microk_nehalem-2.c"
-#elif defined(HASWELL) || defined(ZEN)
+#elif defined(HASWELL) || defined(ZEN) || defined (SKYLAKEX)
 #include "saxpy_microk_haswell-2.c"
 #elif defined(SANDYBRIDGE)
 #include "saxpy_microk_sandy-2.c"

kernel/x86_64/sdot.c

@@ -34,7 +34,7 @@ USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
 #include "sdot_microk_steamroller-2.c"
 #elif defined(NEHALEM)
 #include "sdot_microk_nehalem-2.c"
-#elif defined(HASWELL) || defined(ZEN)
+#elif defined(HASWELL) || defined(ZEN) || defined (SKYLAKEX)
 #include "sdot_microk_haswell-2.c"
 #elif defined(SANDYBRIDGE)
 #include "sdot_microk_sandy-2.c"

kernel/x86_64/sgemv_n_4.c

@@ -35,7 +35,7 @@ USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
 #include "sgemv_n_microk_nehalem-4.c"
 #elif defined(SANDYBRIDGE)
 #include "sgemv_n_microk_sandy-4.c"
-#elif defined(HASWELL) || defined(ZEN)
+#elif defined(HASWELL) || defined(ZEN) || defined (SKYLAKEX)
 #include "sgemv_n_microk_haswell-4.c"
 #endif
 

kernel/x86_64/sgemv_t_4.c

@@ -34,7 +34,7 @@ USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
 #include "sgemv_t_microk_bulldozer-4.c"
 #elif defined(SANDYBRIDGE)
 #include "sgemv_t_microk_sandy-4.c"
-#elif defined(HASWELL) || defined(ZEN)
+#elif defined(HASWELL) || defined(ZEN) || defined (SKYLAKEX)
 #include "sgemv_t_microk_haswell-4.c"
 #endif
 

kernel/x86_64/ssymv_L.c

@@ -32,7 +32,7 @@ USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
 #include "ssymv_L_microk_bulldozer-2.c"
 #elif defined(NEHALEM)
 #include "ssymv_L_microk_nehalem-2.c"
-#elif defined(HASWELL) || defined(ZEN)
+#elif defined(HASWELL) || defined(ZEN) || defined (SKYLAKEX)
 #include "ssymv_L_microk_haswell-2.c"
 #elif defined(SANDYBRIDGE)
 #include "ssymv_L_microk_sandy-2.c"

kernel/x86_64/ssymv_U.c

@@ -33,7 +33,7 @@ USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
 #include "ssymv_U_microk_bulldozer-2.c"
 #elif defined(NEHALEM)
 #include "ssymv_U_microk_nehalem-2.c"
-#elif defined(HASWELL) || defined(ZEN)
+#elif defined(HASWELL) || defined(ZEN) || defined (SKYLAKEX)
 #include "ssymv_U_microk_haswell-2.c"
 #elif defined(SANDYBRIDGE)
 #include "ssymv_U_microk_sandy-2.c"

kernel/x86_64/symv_L_sse.S

@@ -57,7 +57,7 @@
 #define PREFETCHSIZE	(16 * 12)
 #endif
 
-#if defined(NEHALEM) || defined(SANDYBRIDGE) || defined(HASWELL) || defined(ZEN)
+#if defined(NEHALEM) || defined(SANDYBRIDGE) || defined(HASWELL) || defined(ZEN) || defined (SKYLAKEX)
 #define PREFETCH	prefetcht0
 #define PREFETCHW	prefetcht0
 #define PREFETCHSIZE	(16 * 12)

kernel/x86_64/symv_L_sse2.S

@@ -57,7 +57,7 @@
 #define PREFETCHSIZE	(16 * 12)
 #endif
 
-#if defined(NEHALEM) || defined(SANDYBRIDGE) || defined(HASWELL) || defined(ZEN)
+#if defined(NEHALEM) || defined(SANDYBRIDGE) || defined(HASWELL) || defined(ZEN) || defined (SKYLAKEX)
 #define PREFETCH	prefetcht0
 #define PREFETCHW	prefetcht0
 #define PREFETCHSIZE	(16 * 12)

kernel/x86_64/symv_U_sse.S

@@ -57,7 +57,7 @@
 #define PREFETCHSIZE	(16 * 12)
 #endif
 
-#if defined(NEHALEM) || defined(SANDYBRIDGE)  || defined(HASWELL) || defined(ZEN)
+#if defined(NEHALEM) || defined(SANDYBRIDGE)  || defined(HASWELL) || defined(ZEN) || defined (SKYLAKEX)
 #define PREFETCH	prefetcht0
 #define PREFETCHW	prefetcht0
 #define PREFETCHSIZE	(16 * 12)

kernel/x86_64/symv_U_sse2.S

@@ -57,7 +57,7 @@
 #define PREFETCHSIZE	(16 * 12)
 #endif
 
-#if defined(NEHALEM) || defined(SANDYBRIDGE)  || defined(HASWELL) || defined(ZEN)
+#if defined(NEHALEM) || defined(SANDYBRIDGE)  || defined(HASWELL) || defined(ZEN) || defined (SKYLAKEX)
 #define PREFETCH	prefetcht0
 #define PREFETCHW	prefetcht0
 #define PREFETCHSIZE	(16 * 24)

kernel/x86_64/zaxpy.c

@@ -33,7 +33,7 @@ USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
 #include "zaxpy_microk_bulldozer-2.c"
 #elif defined(PILEDRIVER) || defined(STEAMROLLER) || defined(EXCAVATOR)
 #include "zaxpy_microk_steamroller-2.c"
-#elif defined(HASWELL) || defined(ZEN)
+#elif defined(HASWELL) || defined(ZEN) || defined (SKYLAKEX)
 #include "zaxpy_microk_haswell-2.c"
 #elif defined(SANDYBRIDGE)
 #include "zaxpy_microk_sandy-2.c"

kernel/x86_64/zdot.c

@@ -33,7 +33,7 @@ USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
 #include "zdot_microk_bulldozer-2.c"
 #elif defined(STEAMROLLER) || defined(PILEDRIVER) || defined(EXCAVATOR)
 #include "zdot_microk_steamroller-2.c"
-#elif defined(HASWELL) || defined(ZEN)
+#elif defined(HASWELL) || defined(ZEN) || defined (SKYLAKEX)
 #include "zdot_microk_haswell-2.c"
 #elif defined(SANDYBRIDGE)
 #include "zdot_microk_sandy-2.c"

kernel/x86_64/zgemv_n_4.c

@@ -30,7 +30,7 @@ USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
 #include "common.h"
 
 
-#if defined(HASWELL) || defined(ZEN)
+#if defined(HASWELL) || defined(ZEN) || defined (SKYLAKEX)
 #include "zgemv_n_microk_haswell-4.c"
 #elif defined(SANDYBRIDGE)
 #include "zgemv_n_microk_sandy-4.c"

kernel/x86_64/zgemv_t_4.c

@@ -31,7 +31,7 @@ USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
 
 #if defined(BULLDOZER) || defined(PILEDRIVER) || defined(STEAMROLLER)  || defined(EXCAVATOR)
 #include "zgemv_t_microk_bulldozer-4.c"
-#elif defined(HASWELL) || defined(ZEN)
+#elif defined(HASWELL) || defined(ZEN) || defined (SKYLAKEX)
 #include "zgemv_t_microk_haswell-4.c"
 #endif
 

kernel/x86_64/zscal.c

@@ -28,7 +28,7 @@ USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
 #include "common.h"
 
 
-#if defined(HASWELL) || defined(ZEN)
+#if defined(HASWELL) || defined(ZEN) || defined (SKYLAKEX)
 #include "zscal_microk_haswell-2.c"
 #elif defined(BULLDOZER)  || defined(PILEDRIVER)
 #include "zscal_microk_bulldozer-2.c"

kernel/x86_64/zsymv_L_sse.S

@@ -57,7 +57,7 @@
 #define PREFETCHSIZE	(16 * 24)
 #endif
 
-#if defined(NEHALEM) || defined(SANDYBRIDGE) || defined(HASWELL) || defined(ZEN)
+#if defined(NEHALEM) || defined(SANDYBRIDGE) || defined(HASWELL) || defined(ZEN) || defined (SKYLAKEX)
 #define PREFETCH	prefetcht0
 #define PREFETCHW	prefetcht0
 #define PREFETCHSIZE	(16 * 24)

kernel/x86_64/zsymv_L_sse2.S

@@ -57,7 +57,7 @@
 #define PREFETCHSIZE	(16 * 24)
 #endif
 
-#if defined(NEHALEM) || defined(SANDYBRIDGE) || defined(HASWELL) || defined(ZEN)
+#if defined(NEHALEM) || defined(SANDYBRIDGE) || defined(HASWELL) || defined(ZEN) || defined (SKYLAKEX)
 #define PREFETCH	prefetcht0
 #define PREFETCHW	prefetcht0
 #define PREFETCHSIZE	(16 * 24)

kernel/x86_64/zsymv_U_sse.S

@@ -57,7 +57,7 @@
 #define PREFETCHSIZE	(16 * 24)
 #endif
 
-#if defined(NEHALEM) || defined(SANDYBRIDGE)  || defined(HASWELL) || defined(ZEN)
+#if defined(NEHALEM) || defined(SANDYBRIDGE)  || defined(HASWELL) || defined(ZEN) || defined (SKYLAKEX)
 #define PREFETCH	prefetcht0
 #define PREFETCHW	prefetcht0
 #define PREFETCHSIZE	(16 * 24)

kernel/x86_64/zsymv_U_sse2.S

@@ -57,7 +57,7 @@
 #define PREFETCHSIZE	(16 * 24)
 #endif
 
-#if defined(NEHALEM) || defined(SANDYBRIDGE)  || defined(HASWELL) || defined(ZEN)
+#if defined(NEHALEM) || defined(SANDYBRIDGE)  || defined(HASWELL) || defined(ZEN) || defined (SKYLAKEX)
 #define PREFETCH	prefetcht0
 #define PREFETCHW	prefetcht0
 #define PREFETCHSIZE	(16 * 24)

param.h

@@ -1613,6 +1613,125 @@ USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
 #endif
 
 
+#endif
+
+#ifdef SKYLAKEX
+
+#define SNUMOPT         16
+#define DNUMOPT         8
+
+#define GEMM_DEFAULT_OFFSET_A     0
+#define GEMM_DEFAULT_OFFSET_B     0
+#define GEMM_DEFAULT_ALIGN 0x03fffUL
+
+#define SYMV_P  8
+
+#define SWITCH_RATIO	4
+
+#ifdef ARCH_X86
+
+#define SGEMM_DEFAULT_UNROLL_M 4
+#define DGEMM_DEFAULT_UNROLL_M 2
+#define QGEMM_DEFAULT_UNROLL_M 2
+#define CGEMM_DEFAULT_UNROLL_M 2
+#define ZGEMM_DEFAULT_UNROLL_M 1
+#define XGEMM_DEFAULT_UNROLL_M 1
+
+#define SGEMM_DEFAULT_UNROLL_N 4
+#define DGEMM_DEFAULT_UNROLL_N 4
+#define QGEMM_DEFAULT_UNROLL_N 2
+#define CGEMM_DEFAULT_UNROLL_N 2
+#define ZGEMM_DEFAULT_UNROLL_N 2
+#define XGEMM_DEFAULT_UNROLL_N 1
+
+#else
+
+#define SGEMM_DEFAULT_UNROLL_M 16
+#define DGEMM_DEFAULT_UNROLL_M 4
+#define QGEMM_DEFAULT_UNROLL_M 2
+#define CGEMM_DEFAULT_UNROLL_M 8
+#define ZGEMM_DEFAULT_UNROLL_M 4
+#define XGEMM_DEFAULT_UNROLL_M 1
+
+#define SGEMM_DEFAULT_UNROLL_N 4
+#define DGEMM_DEFAULT_UNROLL_N 8
+#define QGEMM_DEFAULT_UNROLL_N 2
+#define CGEMM_DEFAULT_UNROLL_N 2
+#define ZGEMM_DEFAULT_UNROLL_N 2
+#define XGEMM_DEFAULT_UNROLL_N 1
+
+#define SGEMM_DEFAULT_UNROLL_MN 32
+#define DGEMM_DEFAULT_UNROLL_MN 32
+#endif
+
+#ifdef ARCH_X86
+
+#define SGEMM_DEFAULT_P 512
+#define SGEMM_DEFAULT_R sgemm_r
+#define DGEMM_DEFAULT_P 512
+#define DGEMM_DEFAULT_R dgemm_r
+#define QGEMM_DEFAULT_P 504
+#define QGEMM_DEFAULT_R qgemm_r
+#define CGEMM_DEFAULT_P 128
+#define CGEMM_DEFAULT_R 1024
+#define ZGEMM_DEFAULT_P 512
+#define ZGEMM_DEFAULT_R zgemm_r
+#define XGEMM_DEFAULT_P 252
+#define XGEMM_DEFAULT_R xgemm_r
+#define SGEMM_DEFAULT_Q 256
+#define DGEMM_DEFAULT_Q 256
+#define QGEMM_DEFAULT_Q 128
+#define CGEMM_DEFAULT_Q 256
+#define ZGEMM_DEFAULT_Q 192
+#define XGEMM_DEFAULT_Q 128
+
+#else
+
+#define SGEMM_DEFAULT_P 768
+#define DGEMM_DEFAULT_P 512
+#define CGEMM_DEFAULT_P 384
+#define ZGEMM_DEFAULT_P 256
+
+#ifdef WINDOWS_ABI
+#define SGEMM_DEFAULT_Q 320
+#define DGEMM_DEFAULT_Q 128
+#else
+#define SGEMM_DEFAULT_Q 384
+#define DGEMM_DEFAULT_Q 256
+#endif
+#define CGEMM_DEFAULT_Q 192
+#define ZGEMM_DEFAULT_Q 128
+
+#define SGEMM_DEFAULT_R sgemm_r
+#define DGEMM_DEFAULT_R 13824
+#define CGEMM_DEFAULT_R cgemm_r
+#define ZGEMM_DEFAULT_R zgemm_r
+
+#define QGEMM_DEFAULT_Q 128
+#define QGEMM_DEFAULT_P 504
+#define QGEMM_DEFAULT_R qgemm_r
+#define XGEMM_DEFAULT_P 252
+#define XGEMM_DEFAULT_R xgemm_r
+#define XGEMM_DEFAULT_Q 128
+
+#define CGEMM3M_DEFAULT_UNROLL_N 8
+#define CGEMM3M_DEFAULT_UNROLL_M 4
+#define ZGEMM3M_DEFAULT_UNROLL_N 8
+#define ZGEMM3M_DEFAULT_UNROLL_M 2
+
+#define CGEMM3M_DEFAULT_P 448
+#define ZGEMM3M_DEFAULT_P 224
+#define XGEMM3M_DEFAULT_P 112
+#define CGEMM3M_DEFAULT_Q 224
+#define ZGEMM3M_DEFAULT_Q 224
+#define XGEMM3M_DEFAULT_Q 224
+#define CGEMM3M_DEFAULT_R 12288
+#define ZGEMM3M_DEFAULT_R 12288
+#define XGEMM3M_DEFAULT_R 12288
+
+#endif
+
+
 #endif