ahmed/OpenCL-CTS
1
0
Fork 0
mirror of https://github.com/KhronosGroup/OpenCL-CTS.git synced 2026-03-20 14:39:01 +00:00
Code Packages Projects Releases Wiki Activity

Factor out GetTernaryKernel (#1511)

Browse Source 
Use a common function to create the kernel source code for testing
3-argument math builtins.  This reduces code duplication.  1-argument
and 2-argument math kernel construction will be factored out in future
work.

Change the kernels to use preprocessor defines for argument types and
undef values, to make the CTS code easier to read.

Co-authored-by: Marco Antognini <marco.antognini@arm.com>
Signed-off-by: Marco Antognini <marco.antognini@arm.com>
Signed-off-by: Sven van Haastregt <sven.vanhaastregt@arm.com>

Signed-off-by: Marco Antognini <marco.antognini@arm.com>
Signed-off-by: Sven van Haastregt <sven.vanhaastregt@arm.com>
Co-authored-by: Marco Antognini <marco.antognini@arm.com>
This commit is contained in:
Sven van Haastregt
2022-10-04 17:28:29 +01:00
committed by GitHub
parent 07b055cd68
commit d9a938b698
7 changed files with 216 additions and 348 deletions
Download Patch File Download Diff File Expand all files Collapse all files

93
test_conformance/math_brute_force/mad_float.cpp
View File

@@ -26,92 +26,13 @@ namespace {
int BuildKernel(const char *name, int vectorSize, cl_kernel *k, cl_program *p,
bool relaxedMode)
{
const char *c[] = { "__kernel void math_kernel",
sizeNames[vectorSize],
"( __global float",
sizeNames[vectorSize],
"* out, __global float",
sizeNames[vectorSize],
"* in1, __global float",
sizeNames[vectorSize],
"* in2, __global float",
sizeNames[vectorSize],
"* in3 )\n"
"{\n"
" size_t i = get_global_id(0);\n"
" out[i] = ",
name,
"( in1[i], in2[i], in3[i] );\n"
"}\n" };
const char *c3[] = {
"__kernel void math_kernel",
sizeNames[vectorSize],
"( __global float* out, __global float* in, __global float* in2, "
"__global float* in3)\n"
"{\n"
" size_t i = get_global_id(0);\n"
" if( i + 1 < get_global_size(0) )\n"
" {\n"
" float3 f0 = vload3( 0, in + 3 * i );\n"
" float3 f1 = vload3( 0, in2 + 3 * i );\n"
" float3 f2 = vload3( 0, in3 + 3 * i );\n"
" f0 = ",
name,
"( f0, f1, f2 );\n"
" vstore3( f0, 0, out + 3*i );\n"
" }\n"
" else\n"
" {\n"
" size_t parity = i & 1; // Figure out how many elements are "
"left over after BUFFER_SIZE % (3*sizeof(float)). Assume power of two "
"buffer size \n"
" float3 f0;\n"
" float3 f1;\n"
" float3 f2;\n"
" switch( parity )\n"
" {\n"
" case 1:\n"
" f0 = (float3)( in[3*i], NAN, NAN ); \n"
" f1 = (float3)( in2[3*i], NAN, NAN ); \n"
" f2 = (float3)( in3[3*i], NAN, NAN ); \n"
" break;\n"
" case 0:\n"
" f0 = (float3)( in[3*i], in[3*i+1], NAN ); \n"
" f1 = (float3)( in2[3*i], in2[3*i+1], NAN ); \n"
" f2 = (float3)( in3[3*i], in3[3*i+1], NAN ); \n"
" break;\n"
" }\n"
" f0 = ",
name,
"( f0, f1, f2 );\n"
" switch( parity )\n"
" {\n"
" case 0:\n"
" out[3*i+1] = f0.y; \n"
" // fall through\n"
" case 1:\n"
" out[3*i] = f0.x; \n"
" break;\n"
" }\n"
" }\n"
"}\n"
};
const char **kern = c;
size_t kernSize = sizeof(c) / sizeof(c[0]);
if (sizeValues[vectorSize] == 3)
{
kern = c3;
kernSize = sizeof(c3) / sizeof(c3[0]);
}
char testName[32];
snprintf(testName, sizeof(testName) - 1, "math_kernel%s",
sizeNames[vectorSize]);
return MakeKernel(kern, (cl_uint)kernSize, testName, k, p, relaxedMode);
auto kernel_name = GetKernelName(vectorSize);
auto source = GetTernaryKernel(kernel_name, name, ParameterType::Float,
ParameterType::Float, ParameterType::Float,
ParameterType::Float, vectorSize);
std::array<const char *, 1> sources{ source.c_str() };
return MakeKernel(sources.data(), sources.size(), kernel_name.c_str(), k, p,
relaxedMode);
}
struct BuildKernelInfo2