diff --git a/Makefile b/Makefile
index 6542b03acd..33b43f53e9 100644
--- a/Makefile
+++ b/Makefile
@@ -13,7 +13,7 @@ $(HIP_OBJECTS): HIPCC_FLAGS += -I$(HSA_PATH)/include
$(HIP_OBJECTS):
%.o:: %.cpp
- $(HIPCC) $(HIPCC_FLAGS) $< -c -o $@
+ $(HIPCC) $(HIPCC_FLAGS) $< -c -O3 -o $@
clean:
diff --git a/bin/hipconfig b/bin/hipconfig
index 5d982111ed..2af9ad0937 100755
--- a/bin/hipconfig
+++ b/bin/hipconfig
@@ -14,6 +14,7 @@ GetOptions(
,"compiler|c" => \$p_compiler
,"platform|P" => \$p_platform
,"cpp_config|cxx_config|C" => \$p_cpp_config
+ ,"full|f" => \$p_full,
,"newline|n" => \$p_newline
);
@@ -23,6 +24,7 @@ if ($p_help) {
print " --cpp_config, -C : print C++ compiler options\n";
print " --compiler, -c : print compiler (hcc or nvcc)\n";
print " --platform, -P : print platform (hcc or nvcc)\n";
+ print " --full, -f : print full config\n";
print " --newline, -n : print newline\n";
print " --help, -h : print help message\n";
exit();
@@ -47,8 +49,13 @@ $HIP_PATH=$ENV{'HIP_PATH'};
$HIP_PATH=Cwd::realpath (dirname (dirname $0)) unless defined $HIP_PATH; # use parent directory of this tool
+if ($HIP_PLATFORM eq "hcc") {
+ $CPP_CONFIG= " -D__HIP_PLATFORM_HCC__= -I$HIP_PATH/include -I$HCC_HOME/include";
+}
+if ($HIP_PLATFORM eq "nvcc") {
+ $CPP_CONFIG = " -D__HIP_PLATFORM_NVCC__= -I$HIP_PATH/include -I$CUDA_PATH/include";
+};
-$printed = 0;
if ($p_path) {
print "$HIP_PATH";
$printed = 1;
@@ -56,15 +63,39 @@ if ($p_path) {
if ($p_cpp_config) {
- if ($HIP_PLATFORM eq "hcc") {
- print " -D__HIP_PLATFORM_HCC__= -I$HIP_PATH/include -I$HCC_HOME/include";
- }
- if ($HIP_PLATFORM eq "nvcc") {
- print " -D__HIP_PLATFORM_NVCC__= -I$HIP_PATH/include -I$CUDA_PATH/include";
- };
+ print $CPP_CONFIG;
$printed = 1;
}
+if ($p_full) {
+ print "== hipconfig\n";
+ print "HIP_PATH : ", $HIP_PATH, "\n";
+ print "HIP_PLATFORM : ", $HIP_PLATFORM, "\n";
+ print "CPP_CONFIG : ", $CPP_CONFIG, "\n";
+ if ($HIP_PLATFORM eq "hcc")
+ {
+ print "\n" ;
+ print "== hcc\n";
+ print ("HCC_HOME : $HCC_HOME\n");
+ system("$HCC_HOME/bin/hcc --version");
+ }
+ if ($HIP_PLATFORM eq "nvcc") {
+ print "\n" ;
+ print "== nvcc\n";
+ #print "CUDA_PATH :", $CUDA_PATH";
+ system("nvcc --version");
+
+ }
+ print "\n" ;
+
+ print "=== Environment Variables\n";
+ system("env | egrep '^HIP|^HSA|^HCC|^CUDA'");
+
+ print "\n" ;
+ print "== Linux Kernel\n";
+ system ("uname -a");
+ $printed = 1;
+}
if (!$printed or $p_compiler or $p_platform) {
diff --git a/bin/hipconvertinplace.sh b/bin/hipconvertinplace.sh
index 8381687853..a8c8d6d9e8 100755
--- a/bin/hipconvertinplace.sh
+++ b/bin/hipconvertinplace.sh
@@ -4,10 +4,10 @@
#hipify "inplace" all code files in specified directory.
# This can be quite handy when dealing with an existing CUDA code base since the script
-# preseeves the existing directory structure.
+# preserves the existing directory structure.
# For each code file, this script will:
-# - If ".prehip file does not exist, copy the original code to a new file withextension ".prehip". Then Hipify the code file.
+# - If ".prehip file does not exist, copy the original code to a new file with extension ".prehip". Then Hipify the code file.
# - If ".prehip" file exists, this is used as input to hipify.
# (this is useful for testing improvements to the hipify toolset).
diff --git a/docs/markdown/hip_terms.md b/docs/markdown/hip_terms.md
index d55d5da7f5..4ab04f459a 100644
--- a/docs/markdown/hip_terms.md
+++ b/docs/markdown/hip_terms.md
@@ -34,9 +34,9 @@
|Vector|`float4`|`float4`|`hc::`
`short_vector::float4`|`concurrency::`
`graphics::float_4`|`float4`
###Notes
-1. For HC and C++AMP, assume captured _tiled_ext_ t_ext and captured _extent_ ext. These languages use captured variables to pass information to the kernel rather than special built-in functions so variable name may vary.
-2. The indexig functions (starting with thread-index) show the terminology for a 1D grid. Some APIs use reverse order of xyz / 012 indexing for 3D grids.
-3. HC allos tile dimensions to be specified at runtime while C++AMP requires that tile dimensions are specified at compile-time. Thus hc syntax for tile dims is `t_ext.tile_dim[0]` while C++AMP is t_ext.tile_dim0.
+1. For HC and C++AMP, assume a captured _tiled_ext_ named "t_ext" and captured _extent_ named "ext". These languages use captured variables to pass information to the kernel rather than using special built-in functions so the exact variable name may vary.
+2. The indexing functions (starting with `thread-index`) show the terminology for a 1D grid. Some APIs use reverse order of xyz / 012 indexing for 3D grids.
+3. HC allows tile dimensions to be specified at runtime while C++AMP requires that tile dimensions be specified at compile-time. Thus hc syntax for tile dims is `t_ext.tile_dim[0]` while C++AMP is t_ext.tile_dim0.