remove reference to global datastructure on inner loop

1 files changed, 25 insertions, 14 deletions

diff --git a/vowpalwabbit/sparse_dense.h b/vowpalwabbit/sparse_dense.h
index ac0b7037..16e67553 100644
--- a/vowpalwabbit/sparse_dense.h
+++ b/vowpalwabbit/sparse_dense.h
@@ -10,22 +10,28 @@ license as described in the file LICENSE.
 #include "parse_example.h"
 #include "constant.h"
 
+template<class R> 
+struct predict_data {
+  float prediction;
+  R extra;
+};
+  
 inline float sign(float w){ if (w < 0.) return -1.; else  return 1.;}
 
 inline float trunc_weight(float w, float gravity){
   return (gravity < fabsf(w)) ? w - sign(w) * gravity : 0.f;
 }
 
-inline void vec_add(vw& all, float& p, float fx, uint32_t fi) {
-  p += all.reg.weight_vector[fi & all.reg.weight_mask] * fx;
+inline void vec_add(vw& all, predict_data<float>& p, float fx, uint32_t fi) {
+  p.prediction += all.reg.weight_vector[fi & all.reg.weight_mask] * fx;
 }
 
-inline void vec_add_trunc(vw& all, float& p, float fx, uint32_t fi) {
-  p += trunc_weight(all.reg.weight_vector[fi & all.reg.weight_mask], (float)all.sd->gravity) * fx;
+inline void vec_add_trunc(vw& all, predict_data<float>& p, float fx, uint32_t fi) {
+  p.prediction += trunc_weight(all.reg.weight_vector[fi & all.reg.weight_mask], p.extra) * fx;
 }
 
 template <bool adaptive, size_t normalized_idx>
-inline void vec_add_rescale(vw& all, float& p, float fx, uint32_t fi) {
+inline void vec_add_rescale(vw& all, predict_data<float>& p, float fx, uint32_t fi) {
   weight* w = &all.reg.weight_vector[fi & all.reg.weight_mask];
   float x_abs = fabs(fx);
   if( x_abs > w[normalized_idx] ) {// new scale discovered
@@ -35,11 +41,11 @@ inline void vec_add_rescale(vw& all, float& p, float fx, uint32_t fi) {
     }
     w[normalized_idx] = x_abs;
   }
-  p += w[0] * fx;
+  p.prediction += w[0] * fx;
 }
 
 template <bool adaptive, size_t normalized_idx>
-inline void vec_add_trunc_rescale(vw& all, float& p, float fx, uint32_t fi) {
+inline void vec_add_trunc_rescale(vw& all, predict_data<float>& p, float fx, uint32_t fi) {
   weight* w = &all.reg.weight_vector[fi & all.reg.weight_mask];
   float x_abs = fabs(fx);
   if( x_abs > w[normalized_idx] ) {
@@ -49,14 +55,14 @@ inline void vec_add_trunc_rescale(vw& all, float& p, float fx, uint32_t fi) {
     }
     w[normalized_idx] = x_abs;
   }
-  p += trunc_weight(w[0], (float)all.sd->gravity) * fx;
+  p.prediction += trunc_weight(w[0], p.extra) * fx;
 }
 
 template <bool adaptive, size_t normalized_idx>
-inline void vec_add_rescale_general(vw& all, float& p, float fx, uint32_t fi) {
+inline void vec_add_rescale_general(vw& all, predict_data<float>& p, float fx, uint32_t fi) {
   weight* w = &all.reg.weight_vector[fi & all.reg.weight_mask];
   float x_abs = fabs(fx);
-  float power_t_norm = 1.f - (adaptive ? all.power_t : 0.f);
+  float power_t_norm = 1.f - (adaptive ? p.extra : 0.f);
   if( x_abs > w[normalized_idx] ) {
     if( w[normalized_idx] > 0. ) {
       float rescale = (w[normalized_idx]/x_abs);
@@ -64,14 +70,19 @@ inline void vec_add_rescale_general(vw& all, float& p, float fx, uint32_t fi) {
     }
     w[normalized_idx] = x_abs;
   }
-  p += w[0] * fx;
+  p.prediction += w[0] * fx;
 }
 
+struct gnp {
+  float gravity;
+  float power_t;
+};
+
 template <bool adaptive, size_t normalized_idx>
-inline void vec_add_trunc_rescale_general(vw& all, float& p, float fx, uint32_t fi) {
+inline void vec_add_trunc_rescale_general(vw& all, predict_data<gnp>& p, float fx, uint32_t fi) {
   weight* w = &all.reg.weight_vector[fi & all.reg.weight_mask];
   float x_abs = fabs(fx);
-  float power_t_norm = 1.f - (adaptive ? all.power_t : 0.f);
+  float power_t_norm = 1.f - (adaptive ? p.extra.power_t : 0.f);
   if( x_abs > w[normalized_idx] ) {
     if( w[normalized_idx] > 0. ) {
       float rescale = (w[normalized_idx]/x_abs);
@@ -79,7 +90,7 @@ inline void vec_add_trunc_rescale_general(vw& all, float& p, float fx, uint32_t
     }
     w[normalized_idx] = x_abs;
   }
-  p += trunc_weight(w[0], (float)all.sd->gravity) * fx;
+  p.prediction += trunc_weight(w[0], p.extra.gravity) * fx;
 }
 
 void sd_offset_update(weight* weights, size_t mask, feature* begin, feature* end, size_t offset, float update, float regularization);