Errata fixes (Apress#105)

* Don't use octal initializer

* Use C++-style max limit

* Add comments clarifying what T is within the snip blocks

* Move some comments closer to initial use site

samples/Ch09_work_item_communication/fig_9_11_matmul_broadcast.cpp

@@ -9,6 +9,7 @@ using namespace sycl;
 extern const int matrixSize = 128;
 static const int iterations = 16;
 
+// T is the type of data stored in the matrix
 template <typename T>
 double run_sycl(const std::vector<T>& vecA,
                 const std::vector<T>& vecB,
@@ -56,6 +57,8 @@ double run_sycl(const std::vector<T>& vecA,
             // Index in the local index space:
             int i = item.get_local_id()[1];
 
+            // Template type T is the type of data stored in
+            // the matrix
             T sum = 0;
             for (int kk = 0; kk < K; kk += tile_size) {
               // Load the matrix tile from matrix A, and

samples/Ch09_work_item_communication/fig_9_12_ndrange_sub_group_matmul.cpp

@@ -9,6 +9,7 @@ using namespace sycl;
 extern const int matrixSize = 128;
 static const int iterations = 16;
 
+// T is the type of data stored in the matrix
 template <typename T>
 double run_sycl(const std::vector<T>& vecA,
                 const std::vector<T>& vecB,
@@ -55,6 +56,8 @@ double run_sycl(const std::vector<T>& vecA,
             // Index in the local index space:
             int i = item.get_local_id()[1];
 
+            // Template type T is the type of data stored
+            // in the matrix
             T sum = 0;
             for (int kk = 0; kk < K; kk += tile_size) {
               // Load the matrix tile from matrix A.

samples/Ch09_work_item_communication/fig_9_4_naive_matmul.cpp

@@ -44,6 +44,8 @@ double run_sycl(const std::vector<T>& vecA,
         int m = id[0];
         int n = id[1];
 
+        // Template type T is the type of data stored
+        // in the matrix
         T sum = 0;
         for (int k = 0; k < K; k++) {
           sum += matrixA[m][k] * matrixB[k][n];

samples/Ch09_work_item_communication/fig_9_8_ndrange_tiled_matmul.cpp

@@ -44,6 +44,8 @@ double run_sycl(const std::vector<T>& vecA,
 
       // Local accessor, for one matrix tile:
       constexpr int tile_size = 16;
+
+      // Template type T is the type of data stored in the matrix
       auto tileA = local_accessor<T, 1>(tile_size, h);
 
       h.parallel_for(

samples/Ch16_cpus/fig_16_18_vector_swizzle.cpp

@@ -16,11 +16,11 @@ int main() {
 
   q.single_task([=]() {
      sycl::vec<int, 4> old_v =
-         sycl::vec<int, 4>(000, 100, 200, 300);
+         sycl::vec<int, 4>(0, 100, 200, 300);
      sycl::vec<int, 4> new_v = sycl::vec<int, 4>();
 
      new_v.rgba() = old_v.abgr();
-     int vals[] = {300, 200, 100, 000};
+     int vals[] = {300, 200, 100, 0};
 
      if (new_v.r() != vals[0] || new_v.g() != vals[1] ||
          new_v.b() != vals[2] || new_v.a() != vals[3]) {

samples/Ch16_cpus/fig_16_6_stream_triad.cpp

@@ -6,6 +6,7 @@
 #include <cassert>
 #include <cfloat>
 #include <iostream>
+#include <limits>
 #include <string>
 #include <sycl/sycl.hpp>
 using namespace sycl;
@@ -20,7 +21,7 @@ double triad(const std::vector<float>& vecA,
   assert(vecA.size() == vecB.size() &&
          vecB.size() == vecC.size());
   const size_t array_size = vecA.size();
-  double min_time_ns = DBL_MAX;
+  double min_time_ns = std::numeric_limits<double>::max();
 
   queue q{property::queue::enable_profiling{}};
   std::cout << "Running on device: "

samples/Ch16_cpus/figure_16_6.hpp

@@ -13,7 +13,7 @@ double triad(const std::vector<double>& vecA,
              std::vector<double>& vecC) {
   assert(vecA.size() == vecB.size() == vecC.size());
   const size_t array_size = vecA.size();
-  double min_time_ns = DBL_MAX;
+  double min_time_ns = std::numeric_limits<double>::max();
 
   queue Q{property::queue::enable_profiling{}};
   std::cout << "Running on device: "