Fix SERIAL -> Serial.

Fix a few broken Serial linAlg kernels.

okl/linAlg/linAlgAXMY.c

@@ -21,14 +21,14 @@ SOFTWARE.
 
 extern "C" 
 void axmy(const dlong & N, 
+          const dfloat& alpha,
                  const dfloat * __restrict__ cpu_w,
-                 const dfloat * __restrict__ cpu_a,
-                 dfloat * __restrict__ cpu_wa){
+                 dfloat * __restrict__ cpu_a){
 
   for(int i=0;i<N;++i){
     const dfloat ai = cpu_a[i];
     const dfloat wi = cpu_w[i];
-    cpu_wa[i] = ai*wi;
+    cpu_a[i] = alpha*ai*wi;
   }
 
 }
@@ -38,35 +38,34 @@ void axmyMany(const dlong & N,
                       const dlong & Nfields,
                       const dlong & offset,
                       const dlong & mode,
+                      const dfloat & alpha,
                       const dfloat * __restrict__ cpu_w,
-                      const dfloat * __restrict__ cpu_a,
-                      dfloat * __restrict__ cpu_wa){
+                      dfloat * __restrict__ cpu_a){
 
   for(int fld=0;fld<Nfields;fld++) {
     for(int i=0;i<N;++i){
       const dlong id = i + fld*offset;
       const dfloat ai = cpu_a[id];
       const dfloat wi = cpu_w[i + mode * fld * offset];
-      cpu_wa[id] = ai*wi;
+      cpu_a[id] = alpha*ai*wi;
     }
   }
 
 }
 extern "C" 
 void axmyVector(const dlong & N, 
-                      const dlong & Nfields,
                       const dlong & offset,
                       const dlong & mode,
+                      const dfloat & alpha,
                       const dfloat * __restrict__ cpu_w,
-                      const dfloat * __restrict__ cpu_a,
-                      dfloat * __restrict__ cpu_wa){
+                      dfloat * __restrict__ cpu_a){
 
   for(int fld=0;fld<p_NVec;fld++) {
     for(int i=0;i<N;++i){
       const dlong id = i + fld*offset;
       const dfloat ai = cpu_a[id];
       const dfloat wi = cpu_w[i + mode * fld * offset];
-      cpu_wa[id] = ai*wi;
+      cpu_a[id] = alpha*ai*wi;
     }
   }
 

okl/linAlg/linAlgAXPBY.c

@@ -31,7 +31,7 @@ void axpby(const dlong & N, const dlong & xOffset, const dlong& yOffset, const d
 }
 
 extern "C"
-void axpbyMany(const dlong & N, const dlong Nfields, const dlong & offset, const dfloat & alpha, const dfloat * __restrict__ cpu_a, 
+void axpbyMany(const dlong & N, const dlong & Nfields, const dlong & offset, const dfloat & alpha, const dfloat * __restrict__ cpu_a, 
                     const dfloat & beta, dfloat * __restrict__ cpu_b){
 
   for(int fld=0;fld<Nfields;fld++) { 

okl/linAlg/linAlgWeightedInnerProd.c

@@ -19,7 +19,7 @@ SOFTWARE.
 */
 
 extern "C" 
-void weightedInnerProduct2(
+void weightedInnerProd(
             const dlong & Nblocks,
             const dlong & N,
             const dfloat * __restrict__ cpu_w,
@@ -41,7 +41,7 @@ void weightedInnerProduct2(
 }
 
 extern "C" 
-void weightedInnerProduct2Many(
+void weightedInnerProdMany(
             const dlong & Nblocks,
             const dlong & N,
             const dlong & Nfields,
@@ -58,7 +58,7 @@ void weightedInnerProduct2Many(
       const dlong id = i + fld*offset;
       const dfloat ai = cpu_a[id];
       const dfloat bi = cpu_b[id];
-      const dfloat wi = cpu_w[id];
+      const dfloat wi = cpu_w[i];
       wab += ai*bi*wi;
     }
   }

src/linAlg/linAlg.cpp

@@ -77,7 +77,7 @@ void linAlg_t::setup() {
 
   occa::properties kernelInfoNoOkl = platform->kernelInfo;
   kernelInfoNoOkl["okl/enabled"] = false;
-  const bool serial = platform->device.mode() == "SERIAL";
+  const bool serial = platform->device.mode() == "Serial";
 
   {
       if (fillKernel.isInitialized()==false)
@@ -549,7 +549,7 @@ dfloat linAlg_t::innerProd(const dlong N, occa::memory& o_x, occa::memory& o_y,
 dfloat linAlg_t::weightedInnerProd(const dlong N, occa::memory& o_w,
                                    occa::memory& o_x, occa::memory& o_y,
                                    MPI_Comm _comm) {
-  const bool serial = platform->device.mode() == "SERIAL";
+  const bool serial = platform->device.mode() == "Serial";
 #ifdef ENABLE_TIMER
   platform->timer.tic("dotp",1);
 #endif
@@ -559,13 +559,14 @@ dfloat linAlg_t::weightedInnerProd(const dlong N, occa::memory& o_w,
 
   weightedInnerProdKernel(Nblock, N, o_w, o_x, o_y, o_scratch);
 
-  o_scratch.copyTo(scratch, Nbytes);
 
   dfloat dot = 0;
 
   if(serial){
+    o_scratch.copyTo(scratch, sizeof(dfloat));
     dot = scratch[0];
   } else {
+    o_scratch.copyTo(scratch, Nbytes);
     for(dlong n=0;n<Nblock;++n){
       dot += scratch[n];
     }
@@ -617,7 +618,7 @@ dfloat linAlg_t::weightedInnerProdMany(const dlong N,
                                    occa::memory& o_w,
                                    occa::memory& o_x, occa::memory& o_y,
                                    MPI_Comm _comm) {
-  const bool serial = platform->device.mode() == "SERIAL";
+  const bool serial = platform->device.mode() == "Serial";
 #ifdef ENABLE_TIMER
   platform->timer.tic("dotp",1);
 #endif
@@ -627,12 +628,13 @@ dfloat linAlg_t::weightedInnerProdMany(const dlong N,
 
   weightedInnerProdManyKernel(Nblock, N, Nfields, fieldOffset, o_w, o_x, o_y, o_scratch);
 
-  o_scratch.copyTo(scratch, Nbytes);
   dfloat dot = 0;
 
   if(serial){
+    o_scratch.copyTo(scratch, sizeof(dfloat));
     dot = scratch[0];
   } else {
+    o_scratch.copyTo(scratch, Nbytes);
     for(dlong n=0;n<Nblock;++n){
       dot += scratch[n];
     }
@@ -650,7 +652,7 @@ dfloat linAlg_t::weightedInnerProdMany(const dlong N,
 // ||o_a||_w2
 dfloat linAlg_t::weightedNorm2(const dlong N, occa::memory& o_w,
                                occa::memory& o_a, MPI_Comm _comm) {
-  const bool serial = platform->device.mode() == "SERIAL";
+  const bool serial = platform->device.mode() == "Serial";
 #ifdef ENABLE_TIMER
   platform->timer.tic("dotp",1);
 #endif
@@ -660,13 +662,14 @@ dfloat linAlg_t::weightedNorm2(const dlong N, occa::memory& o_w,
 
   weightedNorm2Kernel(Nblock, N, o_w, o_a, o_scratch);
 
-  o_scratch.copyTo(scratch, Nbytes);
 
 
   dfloat norm = 0;
   if(serial){
+    o_scratch.copyTo(scratch, sizeof(dfloat));
     norm = scratch[0];
   } else {
+    o_scratch.copyTo(scratch, Nbytes);
     for(dlong n=0;n<Nblock;++n){
       norm += scratch[n];
     }
@@ -685,7 +688,7 @@ dfloat linAlg_t::weightedNorm2Many(const dlong N,
                                    const dlong fieldOffset,
                                    occa::memory& o_w,
                                occa::memory& o_a, MPI_Comm _comm) {
-  const bool serial = platform->device.mode() == "SERIAL";
+  const bool serial = platform->device.mode() == "Serial";
 #ifdef ENABLE_TIMER
   platform->timer.tic("dotp",1);
 #endif
@@ -695,12 +698,13 @@ dfloat linAlg_t::weightedNorm2Many(const dlong N,
 
   weightedNorm2ManyKernel(Nblock, N, Nfields, fieldOffset, o_w, o_a, o_scratch);
 
-  o_scratch.copyTo(scratch, Nbytes);
 
   dfloat norm = 0;
   if(serial){
+    o_scratch.copyTo(scratch, sizeof(dfloat));
     norm = scratch[0];
   } else {
+    o_scratch.copyTo(scratch, Nbytes);
     for(dlong n=0;n<Nblock;++n){
       norm += scratch[n];
     }