added dgemm_kernel for Piledriver

common.h

@@ -310,6 +310,13 @@ typedef int blasint;
 #define YIELDING	SwitchToThread()
 #endif
 
+/***************************************************
+Some no-oprations are enough
+***************************************************/
+#ifdef PILEDRIVER
+#define YIELDING	__asm__ __volatile__ ("nop;nop;nop;nop;nop;nop;nop;nop;\n");
+#endif
+
 #ifndef YIELDING
 #define YIELDING	sched_yield()
 #endif

driver/level3/level3.c

@@ -333,7 +333,7 @@ int CNAME(blas_arg_t *args, BLASLONG *range_m, BLASLONG *range_n,
       for(jjs = js; jjs < js + min_j; jjs += min_jj){
 	min_jj = min_j + js - jjs;
 
-#if defined(BULLDOZER) && defined(ARCH_X86_64) && !defined(XDOUBLE) && !defined(COMPLEX)
+#if ( defined(BULLDOZER) || defined(PILEDRIVER) ) && defined(ARCH_X86_64) && !defined(XDOUBLE) && !defined(COMPLEX)
         if (min_jj >= 12*GEMM_UNROLL_N) min_jj = 12*GEMM_UNROLL_N;
         else
                 if (min_jj >= 6*GEMM_UNROLL_N) min_jj = 6*GEMM_UNROLL_N;

driver/level3/level3_thread.c

@@ -367,7 +367,7 @@ static int inner_thread(blas_arg_t *args, BLASLONG *range_m, BLASLONG *range_n,
       for(jjs = xxx; jjs < MIN(n_to, xxx + div_n); jjs += min_jj){
 	min_jj = MIN(n_to, xxx + div_n) - jjs;
 
-#if defined(BULLDOZER) && defined(ARCH_X86_64) && !defined(XDOUBLE) && !defined(COMPLEX)
+#if ( defined(BULLDOZER) || defined(PILEDRIVER) ) && defined(ARCH_X86_64) && !defined(XDOUBLE) && !defined(COMPLEX)
 	if (min_jj >= 12*GEMM_UNROLL_N) min_jj = 12*GEMM_UNROLL_N;
 	else
 		if (min_jj >= 6*GEMM_UNROLL_N) min_jj = 6*GEMM_UNROLL_N;

kernel/x86_64/KERNEL.PILEDRIVER

@@ -17,11 +17,11 @@ SGEMMITCOPYOBJ =  sgemm_itcopy$(TSUFFIX).$(SUFFIX)
 SGEMMONCOPYOBJ =  sgemm_oncopy$(TSUFFIX).$(SUFFIX)
 SGEMMOTCOPYOBJ =  sgemm_otcopy$(TSUFFIX).$(SUFFIX)
 
-DGEMMKERNEL    =  dgemm_kernel_6x4_piledriver.S
-DGEMMINCOPY    = ../generic/gemm_ncopy_6.c
-DGEMMITCOPY    = ../generic/gemm_tcopy_6.c
-DGEMMONCOPY    = ../generic/gemm_ncopy_4.c
-DGEMMOTCOPY    = ../generic/gemm_tcopy_4.c
+DGEMMKERNEL    =  dgemm_kernel_8x2_piledriver.S
+DGEMMINCOPY    =  dgemm_ncopy_8_bulldozer.S
+DGEMMITCOPY    =  dgemm_tcopy_8_bulldozer.S
+DGEMMONCOPY    =  gemm_ncopy_2_bulldozer.S
+DGEMMOTCOPY    =  gemm_tcopy_2_bulldozer.S
 DGEMMINCOPYOBJ =  dgemm_incopy$(TSUFFIX).$(SUFFIX)
 DGEMMITCOPYOBJ =  dgemm_itcopy$(TSUFFIX).$(SUFFIX)
 DGEMMONCOPYOBJ =  dgemm_oncopy$(TSUFFIX).$(SUFFIX)

param.h

@@ -330,9 +330,9 @@ USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
 #define XGEMM_DEFAULT_UNROLL_M 1
 #else
 #define SGEMM_DEFAULT_UNROLL_N 2
-#define DGEMM_DEFAULT_UNROLL_N 4
+#define DGEMM_DEFAULT_UNROLL_N 2
 #define SGEMM_DEFAULT_UNROLL_M 16
-#define DGEMM_DEFAULT_UNROLL_M 6
+#define DGEMM_DEFAULT_UNROLL_M 8
 #define QGEMM_DEFAULT_UNROLL_M 2
 #define CGEMM_DEFAULT_UNROLL_M 4
 #define ZGEMM_DEFAULT_UNROLL_M 2
@@ -347,7 +347,7 @@ USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
 
 #if defined(ARCH_X86_64)
 #define SGEMM_DEFAULT_P 768
-#define DGEMM_DEFAULT_P 480
+#define DGEMM_DEFAULT_P 384
 #else
 #define SGEMM_DEFAULT_P 448
 #define DGEMM_DEFAULT_P 480
@@ -359,7 +359,7 @@ USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
 
 #if defined(ARCH_X86_64)
 #define SGEMM_DEFAULT_Q 168
-#define DGEMM_DEFAULT_Q 128
+#define DGEMM_DEFAULT_Q 168
 #else
 #define SGEMM_DEFAULT_Q 224
 #define DGEMM_DEFAULT_Q 224
@@ -371,7 +371,7 @@ USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
 
 #define SGEMM_DEFAULT_R sgemm_r
 #define QGEMM_DEFAULT_R qgemm_r
-#define DGEMM_DEFAULT_R dgemm_r
+#define DGEMM_DEFAULT_R 12288
 #define CGEMM_DEFAULT_R cgemm_r
 #define ZGEMM_DEFAULT_R zgemm_r
 #define XGEMM_DEFAULT_R xgemm_r