st/nine: Avoid Constant upload when there is no change

It is very common for d3d9 apps to set again the constants
they need before every draw call, even if nothing changed.

Since we are mostly gpu bound, it is better to check
for change, and upload constants again (and thus use
gpu bandwith) only if the constants changed.

Signed-off-by: Axel Davy <axel.davy@ens.fr>

diff --git a/src/gallium/state_trackers/nine/device9.c b/src/gallium/state_trackers/nine/device9.c
index dbe0e16da5f63eb983e012a448bb4b9794724093..7b2a2f14c8fac600eecae30b4dcc5ae8e5815afb 100644 (file)
--- a/src/gallium/state_trackers/nine/device9.c
+++ b/src/gallium/state_trackers/nine/device9.c
@@ -3297,6 +3297,12 @@ NineDevice9_SetVertexShaderConstantF( struct NineDevice9 *This,
        return D3D_OK;
     user_assert(pConstantData, D3DERR_INVALIDCALL);
 
+    if (!This->is_recording) {
+        if (!memcmp(&state->vs_const_f[StartRegister * 4], pConstantData,
+                    Vector4fCount * 4 * sizeof(state->vs_const_f[0])))
+            return D3D_OK;
+    }
+
     memcpy(&state->vs_const_f[StartRegister * 4],
            pConstantData,
            Vector4fCount * 4 * sizeof(state->vs_const_f[0]));
@@ -3346,6 +3352,11 @@ NineDevice9_SetVertexShaderConstantI( struct NineDevice9 *This,
     user_assert(pConstantData, D3DERR_INVALIDCALL);
 
     if (This->driver_caps.vs_integer) {
+        if (!This->is_recording) {
+            if (!memcmp(&state->vs_const_i[StartRegister][0], pConstantData,
+                        Vector4iCount * sizeof(state->vs_const_i[0])))
+                return D3D_OK;
+        }
         memcpy(&state->vs_const_i[StartRegister][0],
                pConstantData,
                Vector4iCount * sizeof(state->vs_const_i[0]));
@@ -3410,6 +3421,16 @@ NineDevice9_SetVertexShaderConstantB( struct NineDevice9 *This,
     user_assert(StartRegister + BoolCount <= NINE_MAX_CONST_B, D3DERR_INVALIDCALL);
     user_assert(pConstantData, D3DERR_INVALIDCALL);
 
+    if (!This->is_recording) {
+        bool noChange = true;
+        for (i = 0; i < BoolCount; i++) {
+            if (!!state->vs_const_b[StartRegister + i] != !!pConstantData[i])
+              noChange = false;
+        }
+        if (noChange)
+            return D3D_OK;
+    }
+
     for (i = 0; i < BoolCount; i++)
         state->vs_const_b[StartRegister + i] = pConstantData[i] ? bool_true : 0;
 
@@ -3635,6 +3656,12 @@ NineDevice9_SetPixelShaderConstantF( struct NineDevice9 *This,
        return D3D_OK;
     user_assert(pConstantData, D3DERR_INVALIDCALL);
 
+    if (!This->is_recording) {
+        if (!memcmp(&state->ps_const_f[StartRegister * 4], pConstantData,
+                    Vector4fCount * 4 * sizeof(state->ps_const_f[0])))
+            return D3D_OK;
+    }
+
     memcpy(&state->ps_const_f[StartRegister * 4],
            pConstantData,
            Vector4fCount * 4 * sizeof(state->ps_const_f[0]));
@@ -3684,6 +3711,11 @@ NineDevice9_SetPixelShaderConstantI( struct NineDevice9 *This,
     user_assert(pConstantData, D3DERR_INVALIDCALL);
 
     if (This->driver_caps.ps_integer) {
+        if (!This->is_recording) {
+            if (!memcmp(&state->ps_const_i[StartRegister][0], pConstantData,
+                        Vector4iCount * sizeof(state->ps_const_i[0])))
+                return D3D_OK;
+        }
         memcpy(&state->ps_const_i[StartRegister][0],
                pConstantData,
                Vector4iCount * sizeof(state->ps_const_i[0]));
@@ -3747,6 +3779,16 @@ NineDevice9_SetPixelShaderConstantB( struct NineDevice9 *This,
     user_assert(StartRegister + BoolCount <= NINE_MAX_CONST_B, D3DERR_INVALIDCALL);
     user_assert(pConstantData, D3DERR_INVALIDCALL);
 
+    if (!This->is_recording) {
+        bool noChange = true;
+        for (i = 0; i < BoolCount; i++) {
+            if (!!state->ps_const_b[StartRegister + i] != !!pConstantData[i])
+              noChange = false;
+        }
+        if (noChange)
+            return D3D_OK;
+    }
+
     for (i = 0; i < BoolCount; i++)
         state->ps_const_b[StartRegister + i] = pConstantData[i] ? bool_true : 0;