x86: lpc: simd av_update_lls

4x-6x faster on sandybridge

Signed-off-by: Luca Barbato <lu_zero@gentoo.org>

1 files changed, 9 insertions, 3 deletions

diff --git a/libavutil/lls.h b/libavutil/lls.h
index 81834402f7..27c0d5e3fe 100644
--- a/libavutil/lls.h
+++ b/libavutil/lls.h
@@ -23,9 +23,12 @@
 #ifndef AVUTIL_LLS_H
 #define AVUTIL_LLS_H
 
+#include "common.h"
+#include "mem.h"
 #include "version.h"
 
 #define MAX_VARS 32
+#define MAX_VARS_ALIGN FFALIGN(MAX_VARS+1,4)
 
 //FIXME avoid direct access to LLSModel from outside
 
@@ -33,26 +36,29 @@
  * Linear least squares model.
  */
 typedef struct LLSModel {
-    double covariance[MAX_VARS + 1][MAX_VARS + 1];
-    double coeff[MAX_VARS][MAX_VARS];
+    DECLARE_ALIGNED(32, double, covariance[MAX_VARS_ALIGN][MAX_VARS_ALIGN]);
+    DECLARE_ALIGNED(32, double, coeff[MAX_VARS][MAX_VARS]);
     double variance[MAX_VARS];
     int indep_count;
     /**
      * Take the outer-product of var[] with itself, and add to the covariance matrix.
      * @param m this context
      * @param var training samples, starting with the value to be predicted
+     *            32-byte aligned, and any padding elements must be initialized
+     *            (i.e not denormal/nan).
      */
     void (*update_lls)(struct LLSModel *m, double *var);
     /**
      * Inner product of var[] and the LPC coefs.
      * @param m this context
-     * @param var training samples, excluding the value to be predicted
+     * @param var training samples, excluding the value to be predicted. unaligned.
      * @param order lpc order
      */
     double (*evaluate_lls)(struct LLSModel *m, double *var, int order);
 } LLSModel;
 
 void avpriv_init_lls(LLSModel *m, int indep_count);
+void ff_init_lls_x86(LLSModel *m);
 void avpriv_solve_lls(LLSModel *m, double threshold, unsigned short min_order);
 
 #if FF_API_LLS_PRIVATE