AI/rocm-systems
2
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity

[HIPIFY][#1400] Fix Template Instantiation kernel launch (clang & perl)

Browse Source 
+ Enclose template instantiation kernel calls into round brackets, leave regular kernel names unchanged (hipify-perl doesn't handle cases with macros).
+ Fix corresponding tests.

PS. hipify-perl couldn't handle correctly the following cases due to macros expansion disability, thus hipify-clang should be used instead:

#define KERNEL_NAME_MACRO axpy<float>
#define KERNEL_CALL_MACRO axpy<float><<<1, 2>>>
#define KERNEL_ARG_LIST_MACRO a, x, y

// CUDA:
KERNEL_NAME_MACRO<<<1, 2>>>(KERNEL_ARG_LIST_MACRO);
KERNEL_CALL_MACRO(KERNEL_ARG_LIST_MACRO);

// hipify-perl:
hipLaunchKernelGGL(KERNEL_NAME_MACRO, dim3(1), dim3(2), 0, 0, KERNEL_ARG_LIST_MACRO);
KERNEL_CALL_MACRO(KERNEL_ARG_LIST_MACRO);

// hipify-clang:
hipLaunchKernelGGL((KERNEL_NAME_MACRO), dim3(1), dim3(2), 0, 0, KERNEL_ARG_LIST_MACRO);
hipLaunchKernelGGL((axpy<float>), dim3(1), dim3(2), 0, 0, KERNEL_ARG_LIST_MACRO);
This commit is contained in:
Evgeny Mankov
2019-09-10 15:59:06 +03:00
parent 29d09eb269
commit 56ab105e9d
5 changed files with 65 additions and 35 deletions
Download Patch File Download Diff File Expand all files Collapse all files
bin/hipify-perl
+31 -12
View File
@@ -1662,18 +1662,37 @@ while (@ARGV) {
# match uses ? for <.*> which will be unitialized if this is not present in launch syntax.
no warnings qw/uninitialized/;
my $k = 0;
# Handle the <<numBlocks, blockDim, sharedSize, stream>>> syntax with empty args:
$k += s/(\w+)\s*(<.*>)?\s*<<<\s*(.+)\s*,\s*(.+)\s*,\s*(.+)\s*,\s*(.+)\s*>>>(\s*)\((\s*)\)/hipLaunchKernelGGL($1$2, dim3($3), dim3($4), $5, $6)/g;
# Handle the <<numBlocks, blockDim, sharedSize, stream>>> syntax with non-empty args:
$k += s/(\w+)\s*(<.*>)?\s*<<<\s*(.+)\s*,\s*(.+)\s*,\s*(.+)\s*,\s*(.+)\s*>>>(\s*)\(/hipLaunchKernelGGL($1$2, dim3($3), dim3($4), $5, $6, /g;
# Handle the <<numBlocks, blockDim, sharedSize>>> syntax with empty args:
$k += s/(\w+)\s*(<.*>)?\s*<<<\s*(.+)\s*,\s*(.+)\s*,\s*(.+)\s*>>>(\s*)\((\s*)\)/hipLaunchKernelGGL($1$2, dim3($3), dim3($4), $5, 0)/g;
# Handle the <<numBlocks, blockDim, sharedSize>>> syntax with non-empty args:
$k += s/(\w+)\s*(<.*>)?\s*<<<\s*(.+)\s*,\s*(.+)\s*,\s*(.+)\s*>>>(\s*)\(/hipLaunchKernelGGL($1$2, dim3($3), dim3($4), $5, 0, /g;
# Handle the <<numBlocks, blockDim>>> syntax with empty args:
$k += s/(\w+)\s*(<.*>)?\s*<<<\s*(.+)\s*,\s*(.+)\s*>>>(\s*)\((\s*)\)/hipLaunchKernelGGL($1$2, dim3($3), dim3($4), 0, 0)/g;
# Handle the <<numBlocks, blockDim>>> syntax with non-empty args:
$k += s/(\w+)\s*(<.*>)?\s*<<<\s*(.+)\s*,\s*(.+)\s*>>>(\s*)\(/hipLaunchKernelGGL($1$2, dim3($3), dim3($4), 0, 0, /g;
# Handle the kern<...><<<numBlocks, blockDim, sharedSize, stream>>> syntax with empty args:
$k += s/(\w+)\s*<(.+)>\s*<<<\s*(.+)\s*,\s*(.+)\s*,\s*(.+)\s*,\s*(.+)\s*>>>(\s*)\((\s*)\)/hipLaunchKernelGGL(($1<$2>), dim3($3), dim3($4), $5, $6)/g;
# Handle the kern<<<numBlocks, blockDim, sharedSize, stream>>> syntax with empty args:
$k += s/(\w+)\s*<<<\s*(.+)\s*,\s*(.+)\s*,\s*(.+)\s*,\s*(.+)\s*>>>(\s*)\((\s*)\)/hipLaunchKernelGGL($1, dim3($2), dim3($3), $4, $5)/g;
# Handle the kern<...><<<numBlocks, blockDim, sharedSize, stream>>> syntax with non-empty args:
$k += s/(\w+)\s*<(.+)>\s*<<<\s*(.+)\s*,\s*(.+)\s*,\s*(.+)\s*,\s*(.+)\s*>>>(\s*)\(/hipLaunchKernelGGL(($1<$2>), dim3($3), dim3($4), $5, $6, /g;
# Handle the kern<<<numBlocks, blockDim, sharedSize, stream>>> syntax with non-empty args:
$k += s/(\w+)\s*<<<\s*(.+)\s*,\s*(.+)\s*,\s*(.+)\s*,\s*(.+)\s*>>>(\s*)\(/hipLaunchKernelGGL($1, dim3($2), dim3($3), $4, $5, /g;
# Handle the kern<...><<<numBlocks, blockDim, sharedSize>>> syntax with empty args:
$k += s/(\w+)\s*<(.+)>\s*<<<\s*(.+)\s*,\s*(.+)\s*,\s*(.+)\s*>>>(\s*)\((\s*)\)/hipLaunchKernelGGL(($1<$2>), dim3($3), dim3($4), $5, 0)/g;
# Handle the kern<<<numBlocks, blockDim, sharedSize>>> syntax with empty args:
$k += s/(\w+)\s*<<<\s*(.+)\s*,\s*(.+)\s*,\s*(.+)\s*>>>(\s*)\((\s*)\)/hipLaunchKernelGGL($1, dim3($2), dim3($3), $4, 0)/g;
# Handle the kern<...><<<numBlocks, blockDim, sharedSize>>> syntax with non-empty args:
$k += s/(\w+)\s*<(.+)>\s*<<<\s*(.+)\s*,\s*(.+)\s*,\s*(.+)\s*>>>(\s*)\(/hipLaunchKernelGGL(($1<$2>), dim3($3), dim3($4), $5, 0, /g;
# Handle the kern<<<numBlocks, blockDim, sharedSize>>> syntax with non-empty args:
$k += s/(\w+)\s*<<<\s*(.+)\s*,\s*(.+)\s*,\s*(.+)\s*>>>(\s*)\(/hipLaunchKernelGGL($1, dim3($2), dim3($3), $4, 0, /g;
# Handle the kern<...><<<numBlocks, blockDim>>> syntax with empty args:
$k += s/(\w+)\s*<(.+)>\s*<<<\s*(.+)\s*,\s*(.+)\s*>>>(\s*)\((\s*)\)/hipLaunchKernelGGL(($1<$2>), dim3($3), dim3($4), 0, 0)/g;
# Handle the kern<<<numBlocks, blockDim>>> syntax with empty args:
$k += s/(\w+)\s*<<<\s*(.+)\s*,\s*(.+)\s*>>>(\s*)\((\s*)\)/hipLaunchKernelGGL($1, dim3($2), dim3($3), 0, 0)/g;
# Handle the kern<...><<<numBlocks, blockDim>>> syntax with non-empty args:
$k += s/(\w+)\s*<(.+)>\s*<<<\s*(.+)\s*,\s*(.+)\s*>>>(\s*)\(/hipLaunchKernelGGL(($1<$2>), dim3($3), dim3($4), 0, 0, /g;
# Handle the kern<<<numBlocks, blockDim>>> syntax with non-empty args:
$k += s/(\w+)\s*<<<\s*(.+)\s*,\s*(.+)\s*>>>(\s*)\(/hipLaunchKernelGGL($1, dim3($2), dim3($3), 0, 0, /g;
if ($k) {
$ft{'kern_launch'} += $k;
$Tkernels{$1} ++;