diff --git a/README.md b/README.md
index 0504391b..6d4c477f 100644
--- a/README.md
+++ b/README.md
@@ -48,11 +48,14 @@ and Intel(R) graphics processors.
       to make the generated code easier for humans to read.
       You'll get a warning when running `make` if one of these doesn't exist.
       Everything will still work, but the generated code will be difficult to read.
-      Reading the generated code is only necessary for debug or curiosity.
+      Reading the generated code is only necessary for debug, performance analysis, etc.
     * SWIG (4.0.0 or later):
       http://www.swig.org, for creating the Python interface.
     * Python 3 (3.6.1 or later):
       https://www.python.org/downloads, for creating and using the Python interface.
+      Included with Intel(R) oneAPI HPC Toolkit.
+    * Python `numpy` package for running Python interface tests.
+      Included with Intel(R) oneAPI HPC Toolkit.
     * Doxygen (1.9.0 or later):
       https://www.doxygen.nl, for creating updated API documentation.
       If you're not changing the API documentation, you can view the existing documentation
diff --git a/docs/YASK-tutorial.pdf b/docs/YASK-tutorial.pdf
index d4b45851..dd4bcdbb 100644
Binary files a/docs/YASK-tutorial.pdf and b/docs/YASK-tutorial.pdf differ
diff --git a/src/common/common_utils.cpp b/src/common/common_utils.cpp
index caa77582..94dab876 100644
--- a/src/common/common_utils.cpp
+++ b/src/common/common_utils.cpp
@@ -44,7 +44,7 @@ namespace yask {
     // for numbers above 9 (at least up to 99).
 
     // Format: "major.minor.patch[-alpha|-beta]".
-    const string version = "4.04.04";
+    const string version = "4.04.05";
 
     string yask_get_version_string() {
         return version;
@@ -57,10 +57,12 @@ namespace yask {
     
     // Return num with SI multiplier and "iB" suffix,
     // e.g., 412KiB.
+    // Use only for storage bytes, e.g., not for
+    // rates like bytes/sec.
     string make_byte_str(size_t nbytes)
     {
         if (!is_suffix_print_enabled)
-            return to_string(nbytes);
+            return to_string(nbytes) + " Bytes";
         
         ostringstream os;
         double num = double(nbytes);
@@ -89,7 +91,8 @@ namespace yask {
     }
 
     // Return num with SI multiplier, e.g. "3.14M".
-    // Use this one for rates, etc.
+    // Use this one for printing any number that is
+    // not number of storage bytes.
     string make_num_str(idx_t num) {
         if (!is_suffix_print_enabled || (num > -1000 && num < 1000))
             return to_string(num);
diff --git a/src/common/tuple.cpp b/src/common/tuple.cpp
index 9427ee1b..e964b353 100644
--- a/src/common/tuple.cpp
+++ b/src/common/tuple.cpp
@@ -202,8 +202,8 @@ namespace yask {
         size_t prev_size = 1;
 
         // Loop thru dims.
-        int start_dim = _first_inner ? 0 : size()-1;
-        int end_dim = _first_inner ? size() : -1;
+        int start_dim = _first_inner ? 0 : get_num_dims()-1;
+        int end_dim = _first_inner ? get_num_dims() : -1;
         int step_dim = _first_inner ? 1 : -1;
         for (int di = start_dim; di != end_dim; di += step_dim) {
             auto& i = _q.at(di);
@@ -242,8 +242,8 @@ namespace yask {
         size_t prev_size = 1;
 
         // Loop thru dims.
-        int start_dim = _first_inner ? 0 : size()-1;
-        int stop_dim = _first_inner ? size() : -1;
+        int start_dim = _first_inner ? 0 : get_num_dims()-1;
+        int stop_dim = _first_inner ? get_num_dims() : -1;
         int step_dim = _first_inner ? 1 : -1;
         for (int di = start_dim; di != stop_dim; di += step_dim) {
             auto& i = _q.at(di);
diff --git a/src/compiler/Makefile b/src/compiler/Makefile
index f4346609..bebe05bc 100644
--- a/src/compiler/Makefile
+++ b/src/compiler/Makefile
@@ -62,8 +62,8 @@ YC_CXX    	:=	$(CXX)
 YC_CXXOPT	?= 	-O2
 YC_CXXDBG	?=	-g
 YC_CXXFLAGS 	:=
-YC_CXXFLAGS_API	:=	-std=c++17
-YC_CXXFLAGS_API	+=	-Wall -Wno-unknown-pragmas -Wno-unused-variable
+YC_CXXWARN	:=	-Wall -Wno-unknown-pragmas -Wno-unused-variable
+YC_CXXFLAGS_API	:=	-std=c++17 $(YC_CXXWARN)
 YC_CXX_INCFLAGS 	:=	$(addprefix -I,$(YC_INC_DIRS))
 YC_CXX_INCFLAGS_API	:=	$(addprefix -I,$(INC_DIR))
 
diff --git a/src/kernel/Makefile b/src/kernel/Makefile
index 570a8ecd..165d8387 100644
--- a/src/kernel/Makefile
+++ b/src/kernel/Makefile
@@ -31,13 +31,8 @@
 YASK_BASE	?=	$(abspath ../..)
 include $(YASK_BASE)/src/common/common.mk
 
-# Initial default settings for the YASK kernel library.
+# Default settings for invoking the YASK compiler.
 # These can be overridden on the 'make' command-line.
-# See src/common/common.mk for more setting vars.
-numa			?=	1
-allow_new_var_types	?=	1
-streaming_stores	?=	0
-use_rcp			?=	0
 use_ptrs		?=	1
 use_safe_ptrs		?=	0
 outer_domain_layout	?=	0
@@ -45,6 +40,14 @@ inner_misc_layout 	?=	1
 first_inner		?=	1
 early_loads		?=	0
 min_buffer_len		?=	1
+
+# Default settings for the YASK kernel library.
+# These can be overridden on the 'make' command-line.
+# See src/common/common.mk for more setting vars.
+numa			?=	1
+allow_new_var_types	?=	1
+streaming_stores	?=	0
+use_rcp			?=	0
 trace			?=	0
 trace_mem		?=	0
 check			?=	0
@@ -56,9 +59,7 @@ check			?=	0
 
 # YASK compiler settings for offload.
 ifeq ($(offload),1)
- pfd_l1		:=	0
- pfd_l2		:=	0
- inner_loop_dim :=	1
+ inner_loop_dim 	:=	1
  outer_domain_layout	:=	1
 endif
 
@@ -87,7 +88,7 @@ ifeq ($(TARGET),knl)
   VEC_MACROS	+=	USE_RCP28
  endif
  MACROS		+=	NUMA_PREF=1
- pfd_l1		?=	1
+ pfd_l1		:=	1
 
 else ifeq ($(TARGET),avx512)
 
@@ -156,18 +157,18 @@ endif
 ifneq ($(step_dim),)
  YC_FLAGS	+=	-step-dim $(step_dim)
 endif
-ifneq ($(pfd_l1),)
- YC_FLAGS	+=	-l1-prefetch-dist $(pfd_l1)
-endif
-ifneq ($(pfd_l2),)
- YC_FLAGS	+=	-l2-prefetch-dist $(pfd_l2)
-endif
 ifneq ($(inner_loop_dim),)
  YC_FLAGS	+=	-inner-loop-dim $(inner_loop_dim)
 endif
 ifneq ($(min_buffer_len),)
  YC_FLAGS	+=	-min-buffer-len $(min_buffer_len)
 endif
+ifneq ($(pfd_l1),)
+ YC_FLAGS	+=	-l1-prefetch-dist $(pfd_l1)
+endif
+ifneq ($(pfd_l2),)
+ YC_FLAGS	+=	-l2-prefetch-dist $(pfd_l2)
+endif
 
 # Stencil compiler flags that are boolean.
 ifeq ($(use_ptrs),1)
@@ -351,7 +352,7 @@ endif
 # Compiler-specific settings.
 
 # Create a compiler invocation to test for macro settings.
-YK_CXX_TEST 	:= 	$(YK_CXX)
+YK_CXX_TEST 	:= 	$(YK_CXXCMD)
 cxx_is_llvm_intel :=	$(call MACRO_DEF,$(YK_CXX_TEST),__INTEL_LLVM_COMPILER)
 cxx_is_clang	:=	$(call MACRO_DEF,$(YK_CXX_TEST),__clang__)
 cxx_is_intel	:=	$(call MACRO_DEF,$(YK_CXX_TEST),__INTEL_COMPILER)
@@ -512,7 +513,8 @@ NANO_BLOCK_LOOP_ORDER	?=	DOMAIN_LOOP_DIMS
 ifeq ($(offload),1)
 NANO_BLOCK_LOOP_OMP	?=	omp target teams distribute thread_limit(thread_limit) device(KernelEnv::_omp_devn)
 NANO_BLOCK_LOOP_FLAGS	+=	-omp '$(NANO_BLOCK_LOOP_OMP)'
-NANO_BLOCK_LOOP_CODE	:=	$(NANO_BLOCK_LOOP_MODS) omp loop($(NANO_BLOCK_LOOP_ORDER)) { }
+NANO_BLOCK_LOOP_MODS	+=	omp
+NANO_BLOCK_LOOP_CODE	:=	$(NANO_BLOCK_LOOP_MODS) loop($(NANO_BLOCK_LOOP_ORDER)) { }
 else
 NANO_BLOCK_LOOP_CODE	:=	$(NANO_BLOCK_LOOP_MODS) loop($(NANO_BLOCK_LOOP_ORDER)) { }
 endif
diff --git a/src/kernel/lib/alloc.cpp b/src/kernel/lib/alloc.cpp
index ec09062b..136f5313 100644
--- a/src/kernel/lib/alloc.cpp
+++ b/src/kernel/lib/alloc.cpp
@@ -103,10 +103,14 @@ namespace yask {
         else if (numa_available() != -1) {
             numa_set_bind_policy(0);
             if (numa_pref >= 0 && numa_pref <= numa_max_node())
-                numa_alloc_onnode(nbytes, numa_pref);
+                p = numa_alloc_onnode(nbytes, numa_pref);
             else
-                numa_alloc_local(nbytes);
+                p = numa_alloc_local(nbytes);
             // Interleaved not available.
+
+            if (!p)
+                THROW_YASK_EXCEPTION("cannot allocate " + make_byte_str(nbytes) +
+                                     " using numa-node (or policy) " + to_string(numa_pref));
         }
         else
             THROW_YASK_EXCEPTION("explicit NUMA policy allocation is not available");
@@ -158,15 +162,6 @@ namespace yask {
 
         #endif // not USE_NUMA_POLICY_LIB.
 
-        #else
-        THROW_YASK_EXCEPTION("NUMA allocation is not enabled; build with numa=1");
-        #endif // USE_NUMA.
-
-        // Should not get here w/null p; throw exception.
-        if (!p)
-            THROW_YASK_EXCEPTION("cannot allocate " + make_byte_str(nbytes) +
-                                 " using numa-node (or policy) " + to_string(numa_pref));
-
         // Check alignment.
         if ((size_t(p) & (CACHELINE_BYTES - 1)) != 0)
             FORMAT_AND_THROW_YASK_EXCEPTION("NUMA-allocated " << p << " is not " <<
@@ -177,6 +172,10 @@ namespace yask {
 
         // Return as a char* as required for shared_ptr ctor.
         return static_cast<char*>(p);
+        
+        #else
+        THROW_YASK_EXCEPTION("NUMA allocation is not enabled; build with numa=1");
+        #endif // USE_NUMA.
     }
 
     // Reverse numa_alloc().
@@ -217,38 +216,36 @@ namespace yask {
 
         void *p = 0;
 
+        #ifdef USE_OFFLOAD
+        THROW_YASK_EXCEPTION("mapping offload device memory to shm not yet supported; "
+                             "use '-no-use_shm' option");
+
         // Allocate using MPI shm.
-        #ifdef USE_MPI
+        #elif defined(USE_MPI)
         assert(shm_comm);
         assert(shm_win);
         MPI_Info win_info;
         MPI_Info_create(&win_info);
         MPI_Info_set(win_info, "alloc_shared_noncontig", "true");
         MPI_Win_allocate_shared(nbytes, 1, win_info, *shm_comm, &p, shm_win);
-        MPI_Info_free(&win_info);
-        MPI_Win_lock_all(0, *shm_win);
-        #else
-        THROW_YASK_EXCEPTION("MPI shm allocation is not enabled; build with mpi=1");
-        #endif
-
         if (!p)
             THROW_YASK_EXCEPTION("cannot allocate " + make_byte_str(nbytes) +
                                  " using MPI shm");
+        MPI_Info_free(&win_info);
+        MPI_Win_lock_all(0, *shm_win);
 
         // Check alignment.
         if ((size_t(p) & (CACHELINE_BYTES - 1)) != 0)
             FORMAT_AND_THROW_YASK_EXCEPTION("MPI shm-allocated " << p << " is not " <<
                                             CACHELINE_BYTES << "-byte aligned");
 
-        #ifdef USE_OFFLOAD
-        THROW_YASK_EXCEPTION("mapping offload device memory to shm not yet supported; "
-                             "use '-no-use_shm'");
-        #endif
-
         // Cannot typically use huge pages for shm, so not calling set_huge().
         
         // Return as a char* as required for shared_ptr ctor.
         return static_cast<char*>(p);
+        #else
+        THROW_YASK_EXCEPTION("MPI shm allocation is not enabled; build with mpi=1");
+        #endif
     }
 
     // Reverse shm_alloc().
diff --git a/src/kernel/lib/auto_tuner.cpp b/src/kernel/lib/auto_tuner.cpp
index 27beadeb..7bbaf6fc 100644
--- a/src/kernel/lib/auto_tuner.cpp
+++ b/src/kernel/lib/auto_tuner.cpp
@@ -431,8 +431,6 @@ namespace yask {
                 }
             } // beyond next neighbor of center.
         } // while(true) search for new setting to try.
-
-        THROW_YASK_EXCEPTION("(internal fault) exited from infinite loop");
     } // eval.
 
     // Apply best settings if avail, and adjust other settings.
diff --git a/src/kernel/lib/indices.hpp b/src/kernel/lib/indices.hpp
index 2cde897a..9ff4e0c2 100644
--- a/src/kernel/lib/indices.hpp
+++ b/src/kernel/lib/indices.hpp
@@ -544,6 +544,7 @@ namespace yask {
     static_assert(std::is_trivially_copyable<Indices>::value,
                   "Needed for OpenMP offload");
 
+    #if 0
     // Define OMP reductions on Indices.
     #pragma omp declare reduction(min_idxs : Indices :                  \
                                   omp_out = omp_out.min_elements(omp_in) ) \
@@ -551,6 +552,7 @@ namespace yask {
     #pragma omp declare reduction(max_idxs : Indices :                  \
                                   omp_out = omp_out.max_elements(omp_in) ) \
         initializer (omp_priv = omp_orig)
+    #endif
 
     // Layout base class.
     // This class hierarchy is NOT virtual.
diff --git a/src/kernel/yask.sh b/src/kernel/yask.sh
index 24b3f3a8..fab7dac0 100755
--- a/src/kernel/yask.sh
+++ b/src/kernel/yask.sh
@@ -181,6 +181,7 @@ while true; do
         echo "  -v"
         echo "     Shortcut for the following options:"
         echo "       $val"
+        echo "     Adds '/tests' to path of log_dir."
         echo "     If you want to override any of these values, place them after '-v'."
         echo "  -show_arch"
         echo "     Print the default architecture string and exit."
@@ -357,6 +358,9 @@ dump="head -v -n -0"
 # Init log file.
 : ${logfile:=yask.$stencil.$arch.$exe_host.n$nnodes.r$nranks.`date +%Y-%m-%d_%H-%M-%S`_p$$.log}
 if [[ -n "$logdir" ]]; then
+    if [[ $doval == 1 ]]; then
+        logdir="$logdir/tests"
+    fi
     logfile="$logdir/$logfile"
 fi
 echo "Writing log to '$logfile'."
diff --git a/utils/bin/gen_loops.pl b/utils/bin/gen_loops.pl
index 6b70227f..30075878 100755
--- a/utils/bin/gen_loops.pl
+++ b/utils/bin/gen_loops.pl
@@ -961,8 +961,14 @@ ($)
         # use OpenMP on next loop.
         elsif (lc $tok eq 'omp') {
 
-            $features |= $bOmpPar;
-            print "info: using OpenMP on following loop(s).\n";
+	    if ($OPT{omp} !~ /\w/) {
+		warn "info: ignoring OpenMP loop modifier because '-omp' argument is empty.\n";
+	    } 
+
+	    else {
+		$features |= $bOmpPar;
+		print "info: using OpenMP on following loop(s).\n";
+	    }
         }
 
         # generate manual-scheduling optimizations in next loop.
@@ -1238,7 +1244,7 @@ ()
             "  tiled:           generate tiled scan within a >1D loop.\n",
             "  serpentine:      generate reverse scan when enclosing loop index is odd.*\n",
             "  square_wave:     generate 2D square-wave scan for two innermost dims of >1D loop.*\n",
-            "      * Do not use these modifiers for YASK rank or block loops because they must\n",
+            "      * Do not use these modifiers for YASK block or Mega-block loops because they must\n",
             "        execute with strictly-increasing indices when using temporal tiling.\n",
             "        Also, do not combile these modifiers with 'tiled' or 'manual'.\n",
             "A 'ScanIndices' type must be defined in C++ code prior to including the generated code.\n",