Change dotaxpyf microkernel signature.

The function signature for dotaxpyf has been changed to allow different `alpha` values for the dot and axpy sub-problems. This is needed to support skew-symmetric operations which differ in more than just conjugation of A and A^T.

config/template/kernels/1f/bli_dotxaxpyf_template_noopt_var1.c

@@ -43,7 +43,8 @@ void bli_zdotxaxpyf_template_noopt
        conj_t             conjx,
        dim_t              m,
        dim_t              b_n,
-       dcomplex* restrict alpha,
+       dcomplex* restrict alphaw,
+       dcomplex* restrict alphax,
        dcomplex* restrict a, inc_t inca, inc_t lda,
        dcomplex* restrict w, inc_t incw,
        dcomplex* restrict x, inc_t incx,
@@ -67,8 +68,8 @@ void bli_zdotxaxpyf_template_noopt
 
   This kernel performs the following two gemv-like operations:
 
-    y := beta * y + alpha * conjat( A^T ) * conjw( w )
-    z :=        z + alpha * conja( A )    * conjx( x )
+    y := beta * y + alphaw * conjat( A^T ) * conjw( w )
+    z :=        z + alphax * conja( A )    * conjx( x )
 
   where A is an m x b_n matrix, x and y are vector of length b_n, w and z
   are vectors of length m, and alpha and beta are scalars. The operation
@@ -84,7 +85,8 @@ void bli_zdotxaxpyf_template_noopt
   - m:      The number of rows in matrix A.
   - b_n:    The number of columns in matrix A. Must be equal to or less than
             the fusing factor.
-  - alpha:  The address of the scalar to be applied to A^T*w and A*x.
+  - alphaw: The address of the scalar to be applied to A^T*w.
+  - alphax: The address of the scalar to be applied to A*x.
   - a:      The address of matrix A.
   - inca:   The row stride of A. inca should be unit unless the
             implementation makes special accomodation for non-unit values.
@@ -205,7 +207,8 @@ void bli_zdotxaxpyf_template_noopt
 		  conjx,
 		  m,
 		  b_n,
-		  alpha,
+		  alphaw,
+		  alphax,
 		  a, inca, lda,
 		  w, incw,
 		  x, incx,
@@ -239,15 +242,15 @@ void bli_zdotxaxpyf_template_noopt
 		for ( j = 0; j < b_n; ++j )
 		{
 			bli_zcopys( *xp[ j ], alpha_x[ j ] );
-			bli_zscals( *alpha, alpha_x[ j ] );
+			bli_zscals( *alphax, alpha_x[ j ] );
 		}
 	}
 	else // if ( bli_is_conj( conjx ) )
 	{
 		for ( j = 0; j < b_n; ++j )
 		{
 			bli_zcopyjs( *xp[ j ], alpha_x[ j ] );
-			bli_zscals( *alpha, alpha_x[ j ] );
+			bli_zscals( *alphax, alpha_x[ j ] );
 		}
 	}
 
@@ -468,7 +471,7 @@ void bli_zdotxaxpyf_template_noopt
 	for ( j = 0; j < b_n; ++j )
 	{
 		bli_zscals( *beta, *yp[ j ] );
-		bli_zaxpys( *alpha, At_w[ j ], *yp[ j ] );
+		bli_zaxpys( *alphaw, At_w[ j ], *yp[ j ] );
 	}
 }
 

frame/1f/bli_l1f_check.c

@@ -288,7 +288,8 @@ void bli_dotaxpyv_check
 
 void bli_dotxaxpyf_check
      (
-       const obj_t* alpha,
+       const obj_t* alphaw,
+       const obj_t* alphax,
        const obj_t* at,
        const obj_t* a,
        const obj_t* w,
@@ -302,7 +303,10 @@ void bli_dotxaxpyf_check
 
 	// Check object datatypes.
 
-	e_val = bli_check_noninteger_object( alpha );
+	e_val = bli_check_noninteger_object( alphaw );
+	bli_check_error_code( e_val );
+
+	e_val = bli_check_noninteger_object( alphax );
 	bli_check_error_code( e_val );
 
 	e_val = bli_check_floating_object( at );
@@ -345,7 +349,10 @@ void bli_dotxaxpyf_check
 
 	// Check object dimensions.
 
-	e_val = bli_check_scalar_object( alpha );
+	e_val = bli_check_scalar_object( alphaw );
+	bli_check_error_code( e_val );
+
+	e_val = bli_check_scalar_object( alphax );
 	bli_check_error_code( e_val );
 
 	e_val = bli_check_matrix_object( at );
@@ -397,7 +404,10 @@ void bli_dotxaxpyf_check
 
 	// Check object buffers (for non-NULLness).
 
-	e_val = bli_check_object_buffer( alpha );
+	e_val = bli_check_object_buffer( alphaw );
+	bli_check_error_code( e_val );
+
+	e_val = bli_check_object_buffer( alphax );
 	bli_check_error_code( e_val );
 
 	e_val = bli_check_object_buffer( at );

frame/1f/bli_l1f_check.h

@@ -87,7 +87,8 @@ GENTPROT( dotaxpyv )
 \
 void PASTEMAC(opname,_check) \
      ( \
-       const obj_t* alpha, \
+       const obj_t* alphaw, \
+       const obj_t* alphax, \
        const obj_t* at, \
        const obj_t* a, \
        const obj_t* w, \

frame/1f/bli_l1f_ft.h

@@ -132,7 +132,8 @@ typedef void (*PASTECH(ch,opname,EX_SUF,tsuf)) \
              conj_t conjx, \
              dim_t  m, \
              dim_t  b_n, \
-       const ctype* alpha, \
+       const ctype* alphaw, \
+       const ctype* alphax, \
        const ctype* a, inc_t inca, inc_t lda, \
        const ctype* w, inc_t incw, \
        const ctype* x, inc_t incx, \

frame/1f/bli_l1f_ker_params.h

@@ -78,7 +78,8 @@
              conj_t  conjx, \
              dim_t   m, \
              dim_t   b_n, \
-       const void*   alpha, \
+       const void*   alphaw, \
+       const void*   alphax, \
        const void*   a, inc_t inca, inc_t lda, \
        const void*   w, inc_t incw, \
        const void*   x, inc_t incx, \

frame/1f/bli_l1f_oapi.c

@@ -253,7 +253,8 @@ GENFRONT( dotaxpyv )
 \
 void PASTEMAC(opname,EX_SUF) \
      ( \
-       const obj_t* alpha, \
+       const obj_t* alphaw, \
+       const obj_t* alphax, \
        const obj_t* at, \
        const obj_t* a, \
        const obj_t* w, \
@@ -288,23 +289,28 @@ void PASTEMAC(opname,EX_SUF) \
 	void*     buf_z     = bli_obj_buffer_at_off( z ); \
 	inc_t     inc_z     = bli_obj_vector_inc( z ); \
 \
-	void*     buf_alpha; \
+	void*     buf_alphaw; \
+	void*     buf_alphax; \
 	void*     buf_beta; \
 \
-	obj_t     alpha_local; \
+	obj_t     alphaw_local; \
+	obj_t     alphax_local; \
 	obj_t     beta_local; \
 \
 	if ( bli_error_checking_is_enabled() ) \
-		PASTEMAC(opname,_check)( alpha, at, a, w, x, beta, y, z ); \
+		PASTEMAC(opname,_check)( alphaw, alphax, at, a, w, x, beta, y, z ); \
 \
 	/* Create local copy-casts of scalars (and apply internal conjugation
 	   as needed). */ \
 	bli_obj_scalar_init_detached_copy_of( dt, BLIS_NO_CONJUGATE, \
-	                                      alpha, &alpha_local ); \
+	                                      alphaw, &alphaw_local ); \
+	bli_obj_scalar_init_detached_copy_of( dt, BLIS_NO_CONJUGATE, \
+	                                      alphax, &alphax_local ); \
 	bli_obj_scalar_init_detached_copy_of( dt, BLIS_NO_CONJUGATE, \
 	                                      beta, &beta_local ); \
-	buf_alpha = bli_obj_buffer_for_1x1( dt, &alpha_local ); \
-	buf_beta  = bli_obj_buffer_for_1x1( dt, &beta_local ); \
+	buf_alphaw = bli_obj_buffer_for_1x1( dt, &alphaw_local ); \
+	buf_alphax = bli_obj_buffer_for_1x1( dt, &alphax_local ); \
+	buf_beta   = bli_obj_buffer_for_1x1( dt, &beta_local ); \
 \
 	/* Support cases where matrix A requires a transposition. */ \
     if ( bli_obj_has_trans( a ) ) { bli_swap_incs( &rs_a, &cs_a ); } \
@@ -322,7 +328,8 @@ void PASTEMAC(opname,EX_SUF) \
 	   conjx, \
 	   m, \
 	   b_n, \
-	   buf_alpha, \
+	   buf_alphaw, \
+	   buf_alphax, \
 	   buf_a, rs_a, cs_a, \
 	   buf_w, inc_w, \
 	   buf_x, inc_x, \

frame/1f/bli_l1f_oapi.h

@@ -90,7 +90,8 @@ GENTPROT( dotaxpyv )
 \
 BLIS_EXPORT_BLIS void PASTEMAC(opname,EX_SUF) \
      ( \
-       const obj_t* alpha, \
+       const obj_t* alphaw, \
+       const obj_t* alphax, \
        const obj_t* at, \
        const obj_t* a, \
        const obj_t* w, \

frame/1f/bli_l1f_tapi.c

@@ -185,7 +185,8 @@ void PASTEMAC(ch,opname,EX_SUF) \
              conj_t conjx, \
              dim_t  m, \
              dim_t  b_n, \
-       const ctype* alpha, \
+       const ctype* alphaw, \
+       const ctype* alphax, \
        const ctype* a, inc_t inca, inc_t lda, \
        const ctype* w, inc_t incw, \
        const ctype* x, inc_t incx, \
@@ -214,7 +215,8 @@ void PASTEMAC(ch,opname,EX_SUF) \
 	  conjx, \
 	  m, \
 	  b_n, \
-	  ( ctype* )alpha, \
+	  ( ctype* )alphaw, \
+	  ( ctype* )alphax, \
 	  ( ctype* )a, inca, lda, \
 	  ( ctype* )w, incw, \
 	  ( ctype* )x, incx, \

frame/1f/bli_l1f_tapi.h

@@ -106,7 +106,8 @@ BLIS_EXPORT_BLIS void PASTEMAC(ch,opname,EX_SUF) \
              conj_t conjx, \
              dim_t  m, \
              dim_t  b_n, \
-       const ctype* alpha, \
+       const ctype* alphaw, \
+       const ctype* alphax, \
        const ctype* a, inc_t inca, inc_t lda, \
        const ctype* w, inc_t incw, \
        const ctype* x, inc_t incx, \

frame/2/hemv/bli_hemv_unf_var1.c

@@ -153,6 +153,7 @@ void PASTEMAC(ch,varname) \
 		  n_behind, \
 		  f, \
 		  alpha, \
+		  alpha, \
 		  A10, cs_at, rs_at, \
 		  x0,  incx, \
 		  x1,  incx, \

frame/2/hemv/bli_hemv_unf_var3.c

@@ -202,6 +202,7 @@ void PASTEMAC(ch,varname) \
 		  n_ahead, \
 		  f, \
 		  alpha, \
+		  alpha, \
 		  A21, rs_at, cs_at, \
 		  x2,  incx, \
 		  x1,  incx, \

kernels/penryn/1f/bli_dotxaxpyf_penryn_int.c

@@ -51,7 +51,8 @@ void bli_ddotxaxpyf_penryn_int
              conj_t  conjx,
              dim_t   m,
              dim_t   b_n,
-       const void*   alpha,
+       const void*   alphaw,
+       const void*   alphax,
        const void*   a, inc_t inca, inc_t lda,
        const void*   w, inc_t incw,
        const void*   x, inc_t incx,
@@ -61,13 +62,14 @@ void bli_ddotxaxpyf_penryn_int
        const cntx_t* cntx
      )
 {
-	const double*  restrict alpha_cast = alpha;
-	const double*  restrict a_cast     = a;
-	const double*  restrict w_cast     = w;
-	const double*  restrict x_cast     = x;
-	const double*  restrict beta_cast  = beta;
-	      double*  restrict y_cast     = y;
-	      double*  restrict z_cast     = z;
+	const double*  restrict alphaw_cast = alphaw;
+	const double*  restrict alphax_cast = alphax;
+	const double*  restrict a_cast      = a;
+	const double*  restrict w_cast      = w;
+	const double*  restrict x_cast      = x;
+	const double*  restrict beta_cast   = beta;
+	      double*  restrict y_cast      = y;
+	      double*  restrict z_cast      = z;
 
 	const dim_t             n_elem_per_reg = 2;
 	const dim_t             n_iter_unroll  = 2;
@@ -152,7 +154,8 @@ void bli_ddotxaxpyf_penryn_int
 		  conjx,
 		  m,
 		  b_n,
-		  alpha_cast,
+		  alphaw_cast,
+		  alphax_cast,
 		  a_cast, inca, lda,
 		  w_cast, incw,
 		  x_cast, incx,
@@ -182,10 +185,10 @@ void bli_ddotxaxpyf_penryn_int
 	chi2 = *(x_cast + 2*incx);
 	chi3 = *(x_cast + 3*incx);
 
-	PASTEMAC(d,d,scals)( *alpha_cast, chi0 );
-	PASTEMAC(d,d,scals)( *alpha_cast, chi1 );
-	PASTEMAC(d,d,scals)( *alpha_cast, chi2 );
-	PASTEMAC(d,d,scals)( *alpha_cast, chi3 );
+	PASTEMAC(d,d,scals)( *alphax_cast, chi0 );
+	PASTEMAC(d,d,scals)( *alphax_cast, chi1 );
+	PASTEMAC(d,d,scals)( *alphax_cast, chi2 );
+	PASTEMAC(d,d,scals)( *alphax_cast, chi3 );
 
 	PASTEMAC(d,set0s)( rho0 );
 	PASTEMAC(d,set0s)( rho1 );
@@ -341,7 +344,7 @@ void bli_ddotxaxpyf_penryn_int
 	rho1v.d[1] = rho3;
 
 	betav.v  = _mm_loaddup_pd( ( double* ) beta_cast );
-	alphav.v = _mm_loaddup_pd( ( double* ) alpha_cast );
+	alphav.v = _mm_loaddup_pd( ( double* ) alphaw_cast );
 
 	psi0v.v = _mm_load_pd( ( double* )(y_cast + 0*n_elem_per_reg ) );
 	psi1v.v = _mm_load_pd( ( double* )(y_cast + 1*n_elem_per_reg ) );