Add support for new root.expand() calls.

BUG=5444144

Change-Id: I9ab942561a03e9f837a3199457259bb3e07ac76f
diff --git a/driver/rsdBcc.cpp b/driver/rsdBcc.cpp
index 24bb288..bec6fff 100644
--- a/driver/rsdBcc.cpp
+++ b/driver/rsdBcc.cpp
@@ -1,5 +1,5 @@
 /*
- * Copyright (C) 2011 The Android Open Source Project
+ * Copyright (C) 2011-2012 The Android Open Source Project
  *
  * Licensed under the Apache License, Version 2.0 (the "License");
  * you may not use this file except in compliance with the License.
@@ -36,6 +36,7 @@
 
 struct DrvScript {
     int (*mRoot)();
+    int (*mRootExpand)();
     void (*mInit)();
     void (*mFreeChildren)();
 
@@ -52,6 +53,10 @@
     uint32_t mScriptTextLength;
 };
 
+typedef void (*outer_foreach_t)(
+    const android::renderscript::RsForEachStubParamStruct *,
+    uint32_t x1, uint32_t x2,
+    uint32_t instep, uint32_t outstep);
 
 static Script * setTLS(Script *sc) {
     ScriptTLSStruct * tls = (ScriptTLSStruct *)pthread_getspecific(rsdgThreadTLSKey);
@@ -123,6 +128,7 @@
     }
 
     drv->mRoot = reinterpret_cast<int (*)()>(bccGetFuncAddr(drv->mBccScript, "root"));
+    drv->mRootExpand = reinterpret_cast<int (*)()>(bccGetFuncAddr(drv->mBccScript, "root.expand"));
     drv->mInit = reinterpret_cast<void (*)()>(bccGetFuncAddr(drv->mBccScript, "init"));
     drv->mFreeChildren = reinterpret_cast<void (*)()>(bccGetFuncAddr(drv->mBccScript, ".rs.dtor"));
 
@@ -165,7 +171,12 @@
     script->mHal.info.exportedPragmaCount = drv->ME->getPragmaCount();
     script->mHal.info.exportedPragmaKeyList = drv->ME->getPragmaKeyList();
     script->mHal.info.exportedPragmaValueList = drv->ME->getPragmaValueList();
-    script->mHal.info.root = drv->mRoot;
+
+    if (drv->mRootExpand) {
+      script->mHal.info.root = drv->mRootExpand;
+    } else {
+      script->mHal.info.root = drv->mRoot;
+    }
 
     pthread_mutex_unlock(&rsdgInitMutex);
     return true;
@@ -224,7 +235,7 @@
     RsdHal * dc = (RsdHal *)mtls->rsc->mHal.drv;
     uint32_t sig = mtls->sig;
 
-    outer_foreach_t fn = dc->mForEachLaunch[sig];
+    outer_foreach_t fn = (outer_foreach_t) mtls->script->mHal.info.root;
     while (1) {
         uint32_t slice = (uint32_t)android_atomic_inc(&mtls->mSliceNum);
         uint32_t yStart = mtls->yStart + slice * mtls->mSliceSize;
@@ -240,8 +251,7 @@
             uint32_t offset = mtls->dimX * p.y;
             p.out = mtls->ptrOut + (mtls->eStrideOut * offset);
             p.in = mtls->ptrIn + (mtls->eStrideIn * offset);
-            fn(&mtls->script->mHal.info.root, &p, mtls->xStart, mtls->xEnd,
-               mtls->eStrideIn, mtls->eStrideOut);
+            fn(&p, mtls->xStart, mtls->xEnd, mtls->eStrideIn, mtls->eStrideOut);
         }
     }
 }
@@ -255,7 +265,7 @@
     RsdHal * dc = (RsdHal *)mtls->rsc->mHal.drv;
     uint32_t sig = mtls->sig;
 
-    outer_foreach_t fn = dc->mForEachLaunch[sig];
+    outer_foreach_t fn = (outer_foreach_t) mtls->script->mHal.info.root;
     while (1) {
         uint32_t slice = (uint32_t)android_atomic_inc(&mtls->mSliceNum);
         uint32_t xStart = mtls->xStart + slice * mtls->mSliceSize;
@@ -270,7 +280,7 @@
 
         p.out = mtls->ptrOut + (mtls->eStrideOut * xStart);
         p.in = mtls->ptrIn + (mtls->eStrideIn * xStart);
-        fn(&mtls->script->mHal.info.root, &p, xStart, xEnd, mtls->eStrideIn, mtls->eStrideOut);
+        fn(&p, xStart, xEnd, mtls->eStrideIn, mtls->eStrideOut);
     }
 }
 
@@ -381,7 +391,7 @@
         uint32_t sig = mtls.sig;
 
         //ALOGE("launch 3");
-        outer_foreach_t fn = dc->mForEachLaunch[sig];
+        outer_foreach_t fn = (outer_foreach_t) mtls.script->mHal.info.root;
         for (p.ar[0] = mtls.arrayStart; p.ar[0] < mtls.arrayEnd; p.ar[0]++) {
             for (p.z = mtls.zStart; p.z < mtls.zEnd; p.z++) {
                 for (p.y = mtls.yStart; p.y < mtls.yEnd; p.y++) {
@@ -390,8 +400,8 @@
                                       mtls.dimX * p.y;
                     p.out = mtls.ptrOut + (mtls.eStrideOut * offset);
                     p.in = mtls.ptrIn + (mtls.eStrideIn * offset);
-                    fn(&mtls.script->mHal.info.root, &p, mtls.xStart, mtls.xEnd,
-                       mtls.eStrideIn, mtls.eStrideOut);
+                    fn(&p, mtls.xStart, mtls.xEnd, mtls.eStrideIn,
+                       mtls.eStrideOut);
                 }
             }
         }
diff --git a/driver/rsdCore.cpp b/driver/rsdCore.cpp
index b514e21..1a535d0 100644
--- a/driver/rsdCore.cpp
+++ b/driver/rsdCore.cpp
@@ -1,5 +1,5 @@
 /*
- * Copyright (C) 2011 The Android Open Source Project
+ * Copyright (C) 2011-2012 The Android Open Source Project
  *
  * Licensed under the Apache License, Version 2.0 (the "License");
  * you may not use this file except in compliance with the License.
@@ -43,7 +43,6 @@
 
 static void Shutdown(Context *rsc);
 static void SetPriority(const Context *rsc, int32_t priority);
-static void initForEach(outer_foreach_t* forEachLaunch);
 
 static RsdHalFunctions FunctionTable = {
     rsdGLInit,
@@ -208,8 +207,6 @@
     rsdgThreadTLSKeyCount++;
     pthread_mutex_unlock(&rsdgInitMutex);
 
-    initForEach(dc->mForEachLaunch);
-
     dc->mTlsStruct.mContext = rsc;
     dc->mTlsStruct.mScript = NULL;
     int status = pthread_setspecific(rsdgThreadTLSKey, &dc->mTlsStruct);
@@ -291,173 +288,3 @@
 
 }
 
-static void rsdForEach17(const void *vRoot,
-        const android::renderscript::RsForEachStubParamStruct *p,
-                                uint32_t x1, uint32_t x2,
-                                uint32_t instep, uint32_t outstep) {
-    typedef void (*fe)(const void *, uint32_t);
-    (*(fe*)vRoot)(p->in, p->y);
-}
-
-static void rsdForEach18(const void *vRoot,
-        const android::renderscript::RsForEachStubParamStruct *p,
-                                uint32_t x1, uint32_t x2,
-                                uint32_t instep, uint32_t outstep) {
-    typedef void (*fe)(void *, uint32_t);
-    (*(fe*)vRoot)(p->out, p->y);
-}
-
-static void rsdForEach19(const void *vRoot,
-        const android::renderscript::RsForEachStubParamStruct *p,
-                                uint32_t x1, uint32_t x2,
-                                uint32_t instep, uint32_t outstep) {
-    typedef void (*fe)(const void *, void *, uint32_t);
-    (*(fe*)vRoot)(p->in, p->out, p->y);
-}
-
-static void rsdForEach21(const void *vRoot,
-        const android::renderscript::RsForEachStubParamStruct *p,
-                                uint32_t x1, uint32_t x2,
-                                uint32_t instep, uint32_t outstep) {
-    typedef void (*fe)(const void *, const void *, uint32_t);
-    (*(fe*)vRoot)(p->in, p->usr, p->y);
-}
-
-static void rsdForEach22(const void *vRoot,
-        const android::renderscript::RsForEachStubParamStruct *p,
-                                uint32_t x1, uint32_t x2,
-                                uint32_t instep, uint32_t outstep) {
-    typedef void (*fe)(void *, const void *, uint32_t);
-    (*(fe*)vRoot)(p->out, p->usr, p->y);
-}
-
-static void rsdForEach23(const void *vRoot,
-        const android::renderscript::RsForEachStubParamStruct *p,
-                                uint32_t x1, uint32_t x2,
-                                uint32_t instep, uint32_t outstep) {
-    typedef void (*fe)(const void *, void *, const void *, uint32_t);
-    (*(fe*)vRoot)(p->in, p->out, p->usr, p->y);
-}
-
-static void rsdForEach25(const void *vRoot,
-        const android::renderscript::RsForEachStubParamStruct *p,
-                                uint32_t x1, uint32_t x2,
-                                uint32_t instep, uint32_t outstep) {
-    typedef void (*fe)(const void *, uint32_t, uint32_t);
-    const uint8_t *pin = (const uint8_t *)p->in;
-    uint32_t y = p->y;
-    for (uint32_t x = x1; x < x2; x++) {
-        (*(fe*)vRoot)(pin, x, y);
-        pin += instep;
-    }
-}
-
-static void rsdForEach26(const void *vRoot,
-        const android::renderscript::RsForEachStubParamStruct *p,
-                                uint32_t x1, uint32_t x2,
-                                uint32_t instep, uint32_t outstep) {
-    typedef void (*fe)(void *, uint32_t, uint32_t);
-    uint8_t *pout = (uint8_t *)p->out;
-    uint32_t y = p->y;
-    for (uint32_t x = x1; x < x2; x++) {
-        (*(fe*)vRoot)(pout, x, y);
-        pout += outstep;
-    }
-}
-
-static void rsdForEach27(const void *vRoot,
-        const android::renderscript::RsForEachStubParamStruct *p,
-                                uint32_t x1, uint32_t x2,
-                                uint32_t instep, uint32_t outstep) {
-    typedef void (*fe)(const void *, void *, uint32_t, uint32_t);
-    uint8_t *pout = (uint8_t *)p->out;
-    const uint8_t *pin = (const uint8_t *)p->in;
-    uint32_t y = p->y;
-    for (uint32_t x = x1; x < x2; x++) {
-        (*(fe*)vRoot)(pin, pout, x, y);
-        pin += instep;
-        pout += outstep;
-    }
-}
-
-static void rsdForEach29(const void *vRoot,
-        const android::renderscript::RsForEachStubParamStruct *p,
-                                uint32_t x1, uint32_t x2,
-                                uint32_t instep, uint32_t outstep) {
-    typedef void (*fe)(const void *, const void *, uint32_t, uint32_t);
-    const uint8_t *pin = (const uint8_t *)p->in;
-    const void *usr = p->usr;
-    const uint32_t y = p->y;
-    for (uint32_t x = x1; x < x2; x++) {
-        (*(fe*)vRoot)(pin, usr, x, y);
-        pin += instep;
-    }
-}
-
-static void rsdForEach30(const void *vRoot,
-        const android::renderscript::RsForEachStubParamStruct *p,
-                                uint32_t x1, uint32_t x2,
-                                uint32_t instep, uint32_t outstep) {
-    typedef void (*fe)(void *, const void *, uint32_t, uint32_t);
-    uint8_t *pout = (uint8_t *)p->out;
-    const void *usr = p->usr;
-    const uint32_t y = p->y;
-    for (uint32_t x = x1; x < x2; x++) {
-        (*(fe*)vRoot)(pout, usr, x, y);
-        pout += outstep;
-    }
-}
-
-static void rsdForEach31(const void *vRoot,
-        const android::renderscript::RsForEachStubParamStruct *p,
-                                uint32_t x1, uint32_t x2,
-                                uint32_t instep, uint32_t outstep) {
-    typedef void (*fe)(const void *, void *, const void *, uint32_t, uint32_t);
-    uint8_t *pout = (uint8_t *)p->out;
-    const uint8_t *pin = (const uint8_t *)p->in;
-    const void *usr = p->usr;
-    const uint32_t y = p->y;
-    for (uint32_t x = x1; x < x2; x++) {
-        (*(fe*)vRoot)(pin, pout, usr, x, y);
-        pin += instep;
-        pout += outstep;
-    }
-}
-
-
-static void initForEach(outer_foreach_t* forEachLaunch) {
-    rsAssert(forEachLaunch);
-    forEachLaunch[0x00] = NULL;
-    forEachLaunch[0x01] = rsdForEach31; // in
-    forEachLaunch[0x02] = rsdForEach30; //     out
-    forEachLaunch[0x03] = rsdForEach31; // in, out
-    forEachLaunch[0x04] = NULL;
-    forEachLaunch[0x05] = rsdForEach29;  // in,      usr
-    forEachLaunch[0x06] = rsdForEach30; //     out, usr
-    forEachLaunch[0x07] = rsdForEach31; // in, out, usr
-    forEachLaunch[0x08] = NULL;
-    forEachLaunch[0x09] = rsdForEach25; // in,           x
-    forEachLaunch[0x0a] = rsdForEach26; //     out,      x
-    forEachLaunch[0x0b] = rsdForEach27; // in, out,      x
-    forEachLaunch[0x0c] = NULL;
-    forEachLaunch[0x0d] = rsdForEach29; // in,      usr, x
-    forEachLaunch[0x0e] = rsdForEach30; //     out, usr, x
-    forEachLaunch[0x0f] = rsdForEach31; // in, out, usr, x
-    forEachLaunch[0x10] = NULL;
-    forEachLaunch[0x11] = rsdForEach17; // in               y
-    forEachLaunch[0x12] = rsdForEach18; //     out,         y
-    forEachLaunch[0x13] = rsdForEach19; // in, out,         y
-    forEachLaunch[0x14] = NULL;
-    forEachLaunch[0x15] = rsdForEach21; // in,      usr,    y
-    forEachLaunch[0x16] = rsdForEach22; //     out, usr,    y
-    forEachLaunch[0x17] = rsdForEach23; // in, out, usr,    y
-    forEachLaunch[0x18] = NULL;
-    forEachLaunch[0x19] = rsdForEach25; // in,           x, y
-    forEachLaunch[0x1a] = rsdForEach26; //     out,      x, y
-    forEachLaunch[0x1b] = rsdForEach27; // in, out,      x, y
-    forEachLaunch[0x1c] = NULL;
-    forEachLaunch[0x1d] = rsdForEach29; // in,      usr, x, y
-    forEachLaunch[0x1e] = rsdForEach30; //     out, usr, x, y
-    forEachLaunch[0x1f] = rsdForEach31; // in, out, usr, x, y
-}
-
diff --git a/driver/rsdCore.h b/driver/rsdCore.h
index ce86d11..126c87a 100644
--- a/driver/rsdCore.h
+++ b/driver/rsdCore.h
@@ -1,5 +1,5 @@
 /*
- * Copyright (C) 2011 The Android Open Source Project
+ * Copyright (C) 2011-2012 The Android Open Source Project
  *
  * Licensed under the Apache License, Version 2.0 (the "License");
  * you may not use this file except in compliance with the License.
@@ -27,11 +27,6 @@
 typedef void (* InvokeFunc_t)(void);
 typedef void (*WorkerCallback_t)(void *usr, uint32_t idx);
 
-typedef void (*outer_foreach_t)(const void *,
-    const android::renderscript::RsForEachStubParamStruct *,
-                                uint32_t x1, uint32_t x2,
-                                uint32_t instep, uint32_t outstep);
-
 typedef struct RsdSymbolTableRec {
     const char * mName;
     void * mPtr;
@@ -62,8 +57,6 @@
     Workers mWorkers;
     bool mExit;
 
-    outer_foreach_t mForEachLaunch[32];
-
     ScriptTLSStruct mTlsStruct;
 
     RsdGL gl;