// // Copyright (c) 2017 The Khronos Group Inc. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. // You may obtain a copy of the License at // // http://www.apache.org/licenses/LICENSE-2.0 // // Unless required by applicable law or agreed to in writing, software // distributed under the License is distributed on an "AS IS" BASIS, // WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. // See the License for the specific language governing permissions and // limitations under the License. // #include "testBase.h" #include "harness/conversions.h" #define TEST_SIZE 512 #ifndef MIN #define MIN( _a, _b ) ((_a) < (_b) ? (_a) : (_b)) #endif #ifndef MAX #define MAX( _a, _b ) ((_a) > (_b) ? (_a) : (_b)) #endif const char *singleParamIntegerKernelSourcePattern = "__kernel void sample_test(__global %s *sourceA, __global %s *destValues)\n" "{\n" " int tid = get_global_id(0);\n" " %s%s tmp = vload%s( tid, destValues );\n" " tmp %s= %s( vload%s( tid, sourceA ) );\n" " vstore%s( tmp, tid, destValues );\n" "\n" "}\n"; const char *singleParamSingleSizeIntegerKernelSourcePattern = "__kernel void sample_test(__global %s *sourceA, __global %s *destValues)\n" "{\n" " int tid = get_global_id(0);\n" " destValues[tid] %s= %s( sourceA[tid] );\n" "}\n"; typedef bool (*singleParamIntegerVerifyFn)( void *source, void *destination, ExplicitType vecType ); static void patchup_divide_results( void *outData, const void *inDataA, const void *inDataB, size_t count, ExplicitType vecType ); bool verify_integer_divideAssign( void *source, void *destination, ExplicitType vecType ); bool verify_integer_moduloAssign( void *source, void *destination, ExplicitType vecType ); int test_single_param_integer_kernel(cl_command_queue queue, cl_context context, const char *fnName, ExplicitType vecType, size_t vecSize, singleParamIntegerVerifyFn verifyFn, MTdata d, bool useOpKernel = false ) { clProgramWrapper program; clKernelWrapper kernel; clMemWrapper streams[2]; cl_long inDataA[TEST_SIZE * 16], outData[TEST_SIZE * 16], inDataB[TEST_SIZE * 16], expected; int error, i; size_t threads[1], localThreads[1]; char kernelSource[10240]; char *programPtr; char sizeName[4]; if (! gHasLong && strstr(get_explicit_type_name(vecType),"long")) { log_info( "WARNING: 64 bit integers are not supported on this device. Skipping %s\n", get_explicit_type_name(vecType) ); return CL_SUCCESS; } /* Create the source */ if( vecSize == 1 ) sizeName[ 0 ] = 0; else sprintf( sizeName, "%d", (int)vecSize ); if( vecSize == 1 ) sprintf( kernelSource, singleParamSingleSizeIntegerKernelSourcePattern, get_explicit_type_name( vecType ), get_explicit_type_name( vecType ), useOpKernel ? fnName : "", useOpKernel ? "" : fnName ); else sprintf( kernelSource, singleParamIntegerKernelSourcePattern, get_explicit_type_name( vecType ), get_explicit_type_name( vecType ), get_explicit_type_name( vecType ), sizeName, sizeName, useOpKernel ? fnName : "", useOpKernel ? "" : fnName, sizeName, sizeName ); /* Create kernels */ programPtr = kernelSource; if (create_single_kernel_helper(context, &program, &kernel, 1, (const char **)&programPtr, "sample_test")) { log_error("The program we attempted to compile was: \n%s\n", kernelSource); return -1; } /* Generate some streams */ generate_random_data( vecType, vecSize * TEST_SIZE, d, inDataA ); streams[0] = clCreateBuffer( context, CL_MEM_COPY_HOST_PTR, get_explicit_type_size(vecType) * vecSize * TEST_SIZE, inDataA, NULL); if( streams[0] == NULL ) { log_error("ERROR: Creating input array A failed!\n"); return -1; } if( useOpKernel ) { // Op kernels use an r/w buffer for the second param, so we need to init it with data generate_random_data( vecType, vecSize * TEST_SIZE, d, inDataB ); } streams[1] = clCreateBuffer( context, (CL_MEM_READ_WRITE | (useOpKernel ? CL_MEM_COPY_HOST_PTR : 0)), get_explicit_type_size(vecType) * vecSize * TEST_SIZE, (useOpKernel) ? &inDataB : NULL, NULL); if( streams[1] == NULL ) { log_error("ERROR: Creating output array failed!\n"); return -1; } /* Assign streams and execute */ error = clSetKernelArg( kernel, 0, sizeof( streams[0] ), &streams[0] ); test_error( error, "Unable to set indexed kernel arguments" ); error = clSetKernelArg( kernel, 1, sizeof( streams[1] ), &streams[1] ); test_error( error, "Unable to set indexed kernel arguments" ); /* Run the kernel */ threads[0] = TEST_SIZE; error = get_max_common_work_group_size( context, kernel, threads[0], &localThreads[0] ); test_error( error, "Unable to get work group size to use" ); error = clEnqueueNDRangeKernel( queue, kernel, 1, NULL, threads, localThreads, 0, NULL, NULL ); test_error( error, "Unable to execute test kernel" ); memset(outData, 0xFF, get_explicit_type_size( vecType ) * TEST_SIZE * vecSize ); /* Now get the results */ error = clEnqueueReadBuffer( queue, streams[1], CL_TRUE, 0, get_explicit_type_size( vecType ) * TEST_SIZE * vecSize, outData, 0, NULL, NULL ); test_error( error, "Unable to read output array!" ); // deal with division by 0 -- any answer is allowed here if( verifyFn == verify_integer_divideAssign || verifyFn == verify_integer_moduloAssign ) patchup_divide_results( outData, inDataA, inDataB, TEST_SIZE * vecSize, vecType ); /* And verify! */ char *p = (char *)outData; char *in = (char *)inDataA; char *in2 = (char *)inDataB; for( i = 0; i < (int)TEST_SIZE; i++ ) { for( size_t j = 0; j < vecSize; j++ ) { if( useOpKernel ) memcpy( &expected, in2, get_explicit_type_size( vecType ) ); verifyFn( in, &expected, vecType ); if( memcmp( &expected, p, get_explicit_type_size( vecType ) ) != 0 ) { switch( get_explicit_type_size( vecType )) { case 1: if( useOpKernel ) log_error( "ERROR: Data sample %d:%d does not validate! Expected (0x%2.2x), got (0x%2.2x), sources (0x%2.2x, 0x%2.2x)\n", (int)i, (int)j, ((cl_uchar*)&expected)[0], *( (cl_uchar *)p ), *( (cl_uchar *)in ), *( (cl_uchar *)in2 ) ); else log_error( "ERROR: Data sample %d:%d does not validate! Expected (0x%2.2x), got (0x%2.2x), sources (0x%2.2x)\n", (int)i, (int)j, ((cl_uchar*)&expected)[0], *( (cl_uchar *)p ), *( (cl_uchar *)in ) ); break; case 2: if( useOpKernel ) log_error( "ERROR: Data sample %d:%d does not validate! Expected (0x%4.4x), got (0x%4.4x), sources (0x%4.4x, 0x%4.4x)\n", (int)i, (int)j, ((cl_ushort*)&expected)[0], *( (cl_ushort *)p ), *( (cl_ushort *)in ), *( (cl_ushort *)in2 ) ); else log_error( "ERROR: Data sample %d:%d does not validate! Expected (0x%4.4x), got (0x%4.4x), sources (0x%4.4x)\n", (int)i, (int)j, ((cl_ushort*)&expected)[0], *( (cl_ushort *)p ), *( (cl_ushort *)in ) ); break; case 4: if( useOpKernel ) log_error( "ERROR: Data sample %d:%d does not validate! Expected (0x%8.8x), got (0x%8.8x), sources (0x%8.8x, 0x%8.8x)\n", (int)i, (int)j, ((cl_uint*)&expected)[0], *( (cl_uint *)p ), *( (cl_uint *)in ), *( (cl_uint *)in2 ) ); else log_error( "ERROR: Data sample %d:%d does not validate! Expected (0x%8.8x), got (0x%8.8x), sources (0x%8.8x)\n", (int)i, (int)j, ((cl_uint*)&expected)[0], *( (cl_uint *)p ), *( (cl_uint *)in ) ); break; case 8: if( useOpKernel ) log_error( "ERROR: Data sample %d:%d does not validate! Expected (0x%16.16llx), got (0x%16.16llx), sources (0x%16.16llx, 0x%16.16llx)\n", (int)i, (int)j, ((cl_ulong*)&expected)[0], *( (cl_ulong *)p ), *( (cl_ulong *)in ), *( (cl_ulong *)in2 ) ); else log_error( "ERROR: Data sample %d:%d does not validate! Expected (0x%16.16llx), got (0x%16.16llx), sources (0x%16.16llx)\n", (int)i, (int)j, ((cl_ulong*)&expected)[0], *( (cl_ulong *)p ), *( (cl_ulong *)in ) ); break; } return -1; } p += get_explicit_type_size( vecType ); in += get_explicit_type_size( vecType ); in2 += get_explicit_type_size( vecType ); } } return 0; } int test_single_param_integer_fn( cl_command_queue queue, cl_context context, const char *fnName, singleParamIntegerVerifyFn verifyFn, bool useOpKernel = false ) { ExplicitType types[] = { kChar, kUChar, kShort, kUShort, kInt, kUInt, kLong, kULong, kNumExplicitTypes }; unsigned int vecSizes[] = { 1, 2, 3, 4, 8, 16, 0 }; // TODO 3 not tested unsigned int index, typeIndex; int retVal = 0; RandomSeed seed(gRandomSeed ); for( typeIndex = 0; types[ typeIndex ] != kNumExplicitTypes; typeIndex++ ) { if ((types[ typeIndex ] == kLong || types[ typeIndex ] == kULong) && !gHasLong) continue; for( index = 0; vecSizes[ index ] != 0; index++ ) { if( test_single_param_integer_kernel(queue, context, fnName, types[ typeIndex ], vecSizes[ index ], verifyFn, seed, useOpKernel ) != 0 ) { log_error( " Vector %s%d FAILED\n", get_explicit_type_name( types[ typeIndex ] ), vecSizes[ index ] ); retVal = -1; } } } return retVal; } bool verify_integer_clz( void *source, void *destination, ExplicitType vecType ) { cl_long testValue; int count; int typeBits; switch( vecType ) { case kChar: testValue = *( (cl_char *)source ); typeBits = 8 * sizeof( cl_char ); break; case kUChar: testValue = *( (cl_uchar *)source ); typeBits = 8 * sizeof( cl_uchar ); break; case kShort: testValue = *( (cl_short *)source ); typeBits = 8 * sizeof( cl_short ); break; case kUShort: testValue = *( (cl_ushort *)source ); typeBits = 8 * sizeof( cl_ushort ); break; case kInt: testValue = *( (cl_int *)source ); typeBits = 8 * sizeof( cl_int ); break; case kUInt: testValue = *( (cl_uint *)source ); typeBits = 8 * sizeof( cl_uint ); break; case kLong: testValue = *( (cl_long *)source ); typeBits = 8 * sizeof( cl_long ); break; case kULong: // Hack for now: just treat it as a signed cl_long, since it won't matter for bitcounting testValue = *( (cl_ulong *)source ); typeBits = 8 * sizeof( cl_ulong ); break; default: // Should never happen return false; } count = typeBits; if( testValue ) { testValue <<= 8 * sizeof( testValue ) - typeBits; for( count = 0; 0 == (testValue & CL_LONG_MIN); count++ ) testValue <<= 1; } switch( vecType ) { case kChar: *( (cl_char *)destination ) = count; break; case kUChar: *( (cl_uchar *)destination ) = count; break; case kShort: *( (cl_short *)destination ) = count; break; case kUShort: *( (cl_ushort *)destination ) = count; break; case kInt: *( (cl_int *)destination ) = count; break; case kUInt: *( (cl_uint *)destination ) = count; break; case kLong: *( (cl_long *)destination ) = count; break; case kULong: *( (cl_ulong *)destination ) = count; break; default: // Should never happen return false; } return true; } int test_integer_clz(cl_device_id deviceID, cl_context context, cl_command_queue queue, int num_elements) { return test_single_param_integer_fn( queue, context, "clz", verify_integer_clz ); } bool verify_integer_ctz( void *source, void *destination, ExplicitType vecType ) { cl_long testValue; int count; int typeBits; switch( vecType ) { case kChar: testValue = *( (cl_char *)source ); typeBits = 8 * sizeof( cl_char ); break; case kUChar: testValue = *( (cl_uchar *)source ); typeBits = 8 * sizeof( cl_uchar ); break; case kShort: testValue = *( (cl_short *)source ); typeBits = 8 * sizeof( cl_short ); break; case kUShort: testValue = *( (cl_ushort *)source ); typeBits = 8 * sizeof( cl_ushort ); break; case kInt: testValue = *( (cl_int *)source ); typeBits = 8 * sizeof( cl_int ); break; case kUInt: testValue = *( (cl_uint *)source ); typeBits = 8 * sizeof( cl_uint ); break; case kLong: testValue = *( (cl_long *)source ); typeBits = 8 * sizeof( cl_long ); break; case kULong: // Hack for now: just treat it as a signed cl_long, since it won't matter for bitcounting testValue = *( (cl_ulong *)source ); typeBits = 8 * sizeof( cl_ulong ); break; default: // Should never happen return false; } if ( testValue == 0 ) count = typeBits; else { for( count = 0; (0 == (testValue & 0x1)); count++ ) testValue >>= 1; } switch( vecType ) { case kChar: *( (cl_char *)destination ) = count; break; case kUChar: *( (cl_uchar *)destination ) = count; break; case kShort: *( (cl_short *)destination ) = count; break; case kUShort: *( (cl_ushort *)destination ) = count; break; case kInt: *( (cl_int *)destination ) = count; break; case kUInt: *( (cl_uint *)destination ) = count; break; case kLong: *( (cl_long *)destination ) = count; break; case kULong: *( (cl_ulong *)destination ) = count; break; default: // Should never happen return false; } return true; } int test_integer_ctz(cl_device_id deviceID, cl_context context, cl_command_queue queue, int num_elements) { return test_single_param_integer_fn( queue, context, "ctz", verify_integer_ctz ); } #define OP_CASE( op, sizeName, size ) \ case sizeName: \ { \ cl_##size *d = (cl_##size *)destination; \ *d op##= *( (cl_##size *)source ); \ break; \ } #define OP_CASES( op ) \ switch( vecType ) \ { \ OP_CASE( op, kChar, char ) \ OP_CASE( op, kUChar, uchar ) \ OP_CASE( op, kShort, short ) \ OP_CASE( op, kUShort, ushort ) \ OP_CASE( op, kInt, int ) \ OP_CASE( op, kUInt, uint ) \ OP_CASE( op, kLong, long ) \ OP_CASE( op, kULong, ulong ) \ default: \ break; \ } #define OP_TEST( op, opName ) \ bool verify_integer_##opName##Assign( void *source, void *destination, ExplicitType vecType ) \ { \ OP_CASES( op ) \ return true; \ } \ int test_integer_##opName##Assign(cl_device_id deviceID, cl_context context, cl_command_queue queue, int num_elements) \ { \ return test_single_param_integer_fn( queue, context, #op, verify_integer_##opName##Assign, true ); \ } OP_TEST( +, add ) OP_TEST( -, subtract ) OP_TEST( *, multiply ) OP_TEST( ^, exclusiveOr ) OP_TEST( |, or ) OP_TEST( &, and ) #define OP_CASE_GUARD( op, sizeName, size ) \ case sizeName: \ { \ cl_##size *d = (cl_##size *)destination; \ cl_##size *s = (cl_##size *)source; \ if( *s == 0 ) \ *d = -1; \ else \ *d op##= *s; \ break; \ } #define OP_CASE_GUARD_SIGNED( op, sizeName, size, MIN_VAL ) \ case sizeName: \ { \ cl_##size *d = (cl_##size *)destination; \ cl_##size *s = (cl_##size *)source; \ if( *s == 0 || (*d == MIN_VAL && *s == -1)) \ *d = -1 - MIN_VAL; \ else \ *d op##= *s; \ break; \ } #define OP_CASES_GUARD( op ) \ switch( vecType ) \ { \ OP_CASE_GUARD_SIGNED( op, kChar, char, CL_CHAR_MIN ) \ OP_CASE_GUARD( op, kUChar, uchar ) \ OP_CASE_GUARD_SIGNED( op, kShort, short, CL_SHRT_MIN ) \ OP_CASE_GUARD( op, kUShort, ushort ) \ OP_CASE_GUARD_SIGNED( op, kInt, int, CL_INT_MIN ) \ OP_CASE_GUARD( op, kUInt, uint ) \ OP_CASE_GUARD_SIGNED( op, kLong, long, CL_LONG_MIN ) \ OP_CASE_GUARD( op, kULong, ulong ) \ default: \ break; \ } #define OP_TEST_GUARD( op, opName ) \ bool verify_integer_##opName##Assign( void *source, void *destination, ExplicitType vecType ) \ { \ OP_CASES_GUARD( op ) \ return true; \ } \ int test_integer_##opName##Assign(cl_device_id deviceID, cl_context context, cl_command_queue queue, int num_elements) \ { \ return test_single_param_integer_fn( queue, context, #op, verify_integer_##opName##Assign, true ); \ } OP_TEST_GUARD( /, divide ) OP_TEST_GUARD( %, modulo ) #define PATCH_CASE( _out, _src, _dest, _count, _cl_type ) \ { \ const _cl_type *denom = (const _cl_type* ) _src; \ _cl_type *result = (_cl_type* ) _out; \ for( size_t i = 0; i < _count; i++ ) \ if( denom[i] == 0 ) \ result[i] = (_cl_type) -1; \ } #define PATCH_CASE_SIGNED( _out, _src, _dest, _count, _cl_type, _MIN_VAL ) \ { \ const _cl_type *num = (const _cl_type* ) _dest; \ const _cl_type *denom = (const _cl_type* ) _src; \ _cl_type *result = (_cl_type* ) _out; \ for( size_t i = 0; i < _count; i++ ) \ if( denom[i] == 0 || ( num[i] == _MIN_VAL && denom[i] == -1)) \ result[i] = -1 - _MIN_VAL; \ } static void patchup_divide_results( void *outData, const void *inDataA, const void *inDataB, size_t count, ExplicitType vecType ) { switch( vecType ) { case kChar: PATCH_CASE_SIGNED( outData, inDataA, inDataB, count, cl_char, CL_CHAR_MIN ) break; case kUChar: PATCH_CASE( outData, inDataA, inDataB, count, cl_uchar ) break; case kShort: PATCH_CASE_SIGNED( outData, inDataA, inDataB, count, cl_short, CL_SHRT_MIN ) break; case kUShort: PATCH_CASE( outData, inDataA, inDataB, count, cl_ushort ) break; case kInt: PATCH_CASE_SIGNED( outData, inDataA, inDataB, count, cl_int, CL_INT_MIN ) break; case kUInt: PATCH_CASE( outData, inDataA, inDataB, count, cl_uint ) break; case kLong: PATCH_CASE_SIGNED( outData, inDataA, inDataB, count, cl_long, CL_LONG_MIN ) break; case kULong: PATCH_CASE( outData, inDataA, inDataB, count, cl_ulong ) break; default: log_error( "ERROR: internal test error -- unknown data type %d\n", vecType ); break; } } const char *twoParamIntegerKernelSourcePattern = "__kernel void sample_test(__global %s%s *sourceA, __global %s%s *sourceB, __global %s%s *destValues)\n" "{\n" " int tid = get_global_id(0);\n" " %s%s sA = %s;\n" " %s%s sB = %s;\n" " %s%s dst = %s( sA, sB );\n" " %s;\n" "\n" "}\n"; typedef bool (*twoParamIntegerVerifyFn)( void *sourceA, void *sourceB, void *destination, ExplicitType vecType ); static char * build_load_statement( char *outString, size_t vecSize, const char *name ) { if( vecSize != 3 ) sprintf( outString, "%s[ tid ]", name ); else sprintf( outString, "vload3( tid, %s )", name ); return outString; } static char * build_store_statement( char *outString, size_t vecSize, const char *name, const char *srcName ) { if( vecSize != 3 ) sprintf( outString, "%s[ tid ] = %s", name, srcName ); else sprintf( outString, "vstore3( %s, tid, %s )", srcName, name ); return outString; } int test_two_param_integer_kernel(cl_command_queue queue, cl_context context, const char *fnName, ExplicitType vecAType, ExplicitType vecBType, unsigned int vecSize, twoParamIntegerVerifyFn verifyFn, MTdata d ) { clProgramWrapper program; clKernelWrapper kernel; clMemWrapper streams[3]; cl_long inDataA[TEST_SIZE * 16], inDataB[TEST_SIZE * 16], outData[TEST_SIZE * 16], expected; int error, i; size_t threads[1], localThreads[1]; char kernelSource[10240]; char *programPtr; char sizeName[4], paramSizeName[4]; // embedded profiles don't support long/ulong datatypes if (! gHasLong && strstr(get_explicit_type_name(vecAType),"long")) { log_info( "WARNING: 64 bit integers are not supported on this device. Skipping %s\n", get_explicit_type_name(vecAType) ); return CL_SUCCESS; } /* Create the source */ if( vecSize == 1 ) sizeName[ 0 ] = 0; else sprintf( sizeName, "%d", vecSize ); if( ( vecSize == 1 ) || ( vecSize == 3 ) ) paramSizeName[ 0 ] = 0; else sprintf( paramSizeName, "%d", vecSize ); char sourceALoad[ 128 ], sourceBLoad[ 128 ], destStore[ 128 ]; sprintf( kernelSource, twoParamIntegerKernelSourcePattern, get_explicit_type_name( vecAType ), paramSizeName, get_explicit_type_name( vecBType ), paramSizeName, get_explicit_type_name( vecAType ), paramSizeName, get_explicit_type_name( vecAType ), sizeName, build_load_statement( sourceALoad, (size_t)vecSize, "sourceA" ), get_explicit_type_name( vecBType ), sizeName, build_load_statement( sourceBLoad, (size_t)vecSize, "sourceB" ), get_explicit_type_name( vecAType ), sizeName, fnName, build_store_statement( destStore, (size_t)vecSize, "destValues", "dst" ) ); /* Create kernels */ programPtr = kernelSource; if( create_single_kernel_helper( context, &program, &kernel, 1, (const char **)&programPtr, "sample_test" ) ) { log_error("The program we attempted to compile was: \n%s\n", kernelSource); return -1; } /* Generate some streams */ generate_random_data( vecAType, vecSize * TEST_SIZE, d, inDataA ); generate_random_data( vecBType, vecSize * TEST_SIZE, d, inDataB ); streams[0] = clCreateBuffer( context, CL_MEM_COPY_HOST_PTR, get_explicit_type_size(vecAType) * vecSize * TEST_SIZE, &inDataA, NULL); if( streams[0] == NULL ) { log_error("ERROR: Creating input array A failed!\n"); return -1; } streams[1] = clCreateBuffer( context, CL_MEM_COPY_HOST_PTR, get_explicit_type_size(vecBType) * vecSize * TEST_SIZE, &inDataB, NULL); if( streams[1] == NULL ) { log_error("ERROR: Creating input array B failed!\n"); return -1; } streams[2] = clCreateBuffer( context, CL_MEM_READ_WRITE, get_explicit_type_size(vecAType) * vecSize * TEST_SIZE, NULL, NULL); if( streams[2] == NULL ) { log_error("ERROR: Creating output array failed!\n"); return -1; } /* Assign streams and execute */ error = clSetKernelArg( kernel, 0, sizeof( streams[0] ), &streams[0] ); test_error( error, "Unable to set indexed kernel arguments" ); error = clSetKernelArg( kernel, 1, sizeof( streams[1] ), &streams[1] ); test_error( error, "Unable to set indexed kernel arguments" ); error = clSetKernelArg( kernel, 2, sizeof( streams[2] ), &streams[2] ); test_error( error, "Unable to set indexed kernel arguments" ); /* Run the kernel */ threads[0] = TEST_SIZE; error = get_max_common_work_group_size( context, kernel, threads[0], &localThreads[0] ); test_error( error, "Unable to get work group size to use" ); error = clEnqueueNDRangeKernel( queue, kernel, 1, NULL, threads, localThreads, 0, NULL, NULL ); test_error( error, "Unable to execute test kernel" ); memset(outData, 0xFF, get_explicit_type_size( vecAType ) * TEST_SIZE * vecSize); /* Now get the results */ error = clEnqueueReadBuffer( queue, streams[2], CL_TRUE, 0, get_explicit_type_size( vecAType ) * TEST_SIZE * vecSize, outData, 0, NULL, NULL ); test_error( error, "Unable to read output array!" ); /* And verify! */ char *inA = (char *)inDataA; char *inB = (char *)inDataB; char *out = (char *)outData; for( i = 0; i < (int)TEST_SIZE; i++ ) { for( size_t j = 0; j < vecSize; j++ ) { bool test = verifyFn( inA, inB, &expected, vecAType ); if( test && ( memcmp( &expected, out, get_explicit_type_size( vecAType ) ) != 0 ) ) { switch( get_explicit_type_size( vecAType )) { case 1: log_error( "ERROR: Data sample %d:%d does not validate! Expected (0x%2.2x), got (0x%2.2x), sources (0x%2.2x, 0x%2.2x), TEST_SIZE %d\n", (int)i, (int)j, ((cl_uchar*)&expected)[ 0 ], *( (cl_uchar *)out ), *( (cl_uchar *)inA ), *( (cl_uchar *)inB ) , TEST_SIZE); break; case 2: log_error( "ERROR: Data sample %d:%d does not validate! Expected (0x%4.4x), got (0x%4.4x), sources (0x%4.4x, 0x%4.4x), TEST_SIZE %d\n", (int)i, (int)j, ((cl_ushort*)&expected)[ 0 ], *( (cl_ushort *)out ), *( (cl_ushort *)inA ), *( (cl_ushort *)inB ), TEST_SIZE); break; case 4: log_error( "ERROR: Data sample %d:%d does not validate! Expected (0x%8.8x), got (0x%8.8x), sources (0x%8.8x, 0x%8.8x)\n", (int)i, (int)j, ((cl_uint*)&expected)[ 0 ], *( (cl_uint *)out ), *( (cl_uint *)inA ), *( (cl_uint *)inB ) ); break; case 8: log_error( "ERROR: Data sample %d:%d does not validate! Expected (0x%16.16llx), got (0x%16.16llx), sources (0x%16.16llx, 0x%16.16llx)\n", (int)i, (int)j, ((cl_ulong*)&expected)[ 0 ], *( (cl_ulong *)out ), *( (cl_ulong *)inA ), *( (cl_ulong *)inB ) ); break; } return -1; } inA += get_explicit_type_size( vecAType ); inB += get_explicit_type_size( vecBType ); out += get_explicit_type_size( vecAType ); } } return 0; } int test_two_param_integer_fn(cl_command_queue queue, cl_context context, const char *fnName, twoParamIntegerVerifyFn verifyFn) { ExplicitType types[] = { kChar, kUChar, kShort, kUShort, kInt, kUInt, kLong, kULong, kNumExplicitTypes }; unsigned int vecSizes[] = { 1, 2, 3, 4, 8, 16, 0 }; // TODO : 3 not tested unsigned int index, typeIndex; int retVal = 0; RandomSeed seed(gRandomSeed ); for( typeIndex = 0; types[ typeIndex ] != kNumExplicitTypes; typeIndex++ ) { if (( types[ typeIndex ] == kLong || types[ typeIndex ] == kULong) && !gHasLong) continue; for( index = 0; vecSizes[ index ] != 0; index++ ) { if( test_two_param_integer_kernel(queue, context, fnName, types[ typeIndex ], types[ typeIndex ], vecSizes[ index ], verifyFn, seed ) != 0 ) { log_error( " Vector %s%d FAILED\n", get_explicit_type_name( types[ typeIndex ] ), vecSizes[ index ] ); retVal = -1; } } } return retVal; } int test_two_param_unmatched_integer_fn(cl_command_queue queue, cl_context context, const char *fnName, twoParamIntegerVerifyFn verifyFn) { ExplicitType types[] = { kChar, kUChar, kShort, kUShort, kInt, kUInt, kLong, kULong, kNumExplicitTypes }; unsigned int vecSizes[] = { 1, 2, 4, 8, 16, 0 }; unsigned int index, typeAIndex, typeBIndex; int retVal = 0; RandomSeed seed( gRandomSeed ); for( typeAIndex = 0; types[ typeAIndex ] != kNumExplicitTypes; typeAIndex++ ) { if (( types[ typeAIndex ] == kLong || types[ typeAIndex ] == kULong) && !gHasLong) continue; for( typeBIndex = 0; types[ typeBIndex ] != kNumExplicitTypes; typeBIndex++ ) { if (( types[ typeBIndex ] == kLong || types[ typeBIndex ] == kULong) && !gHasLong) continue; for( index = 0; vecSizes[ index ] != 0; index++ ) { if( test_two_param_integer_kernel( queue, context, fnName, types[ typeAIndex ], types[ typeBIndex ], vecSizes[ index ], verifyFn, seed ) != 0 ) { log_error( " Vector %s%d / %s%d FAILED\n", get_explicit_type_name( types[ typeAIndex ] ), vecSizes[ index ], get_explicit_type_name( types[ typeBIndex ] ), vecSizes[ index ] ); retVal = -1; } } } } return retVal; } bool verify_integer_hadd( void *sourceA, void *sourceB, void *destination, ExplicitType vecType ) { cl_long testValueA, testValueB, overflow; cl_ulong uValueA, uValueB, uOverflow; switch( vecType ) { case kChar: testValueA = *( (cl_char *)sourceA ); testValueB = *( (cl_char *)sourceB ); *( (cl_char *)destination ) = (cl_char)( ( testValueA + testValueB ) >> 1 ); break; case kUChar: testValueA = *( (cl_uchar *)sourceA ); testValueB = *( (cl_uchar *)sourceB ); *( (cl_uchar *)destination ) = (cl_uchar)( ( testValueA + testValueB ) >> 1 ); break; case kShort: testValueA = *( (cl_short *)sourceA ); testValueB = *( (cl_short *)sourceB ); *( (cl_short *)destination ) = (cl_short)( ( testValueA + testValueB ) >> 1 ); break; case kUShort: testValueA = *( (cl_ushort *)sourceA ); testValueB = *( (cl_ushort *)sourceB ); *( (cl_ushort *)destination ) = (cl_ushort)( ( testValueA + testValueB ) >> 1 ); break; case kInt: testValueA = *( (cl_int *)sourceA ); testValueB = *( (cl_int *)sourceB ); *( (cl_int *)destination ) = (cl_int)( ( testValueA + testValueB ) >> 1 ); break; case kUInt: testValueA = *( (cl_uint *)sourceA ); testValueB = *( (cl_uint *)sourceB ); *( (cl_uint *)destination ) = (cl_uint)( ( testValueA + testValueB ) >> 1 ); break; case kLong: // The long way to avoid dropping bits testValueA = *( (cl_long *)sourceA ); testValueB = *( (cl_long *)sourceB ); overflow = ( testValueA & 0x1 ) + ( testValueB & 0x1 ); *( (cl_long *)destination ) = ( ( testValueA >> 1 ) + ( testValueB >> 1 ) ) + ( overflow >> 1 ); break; case kULong: // The long way to avoid dropping bits uValueA = *( (cl_ulong *)sourceA ); uValueB = *( (cl_ulong *)sourceB ); uOverflow = ( uValueA & 0x1 ) + ( uValueB & 0x1 ); *( (cl_ulong *)destination ) = ( ( uValueA >> 1 ) + ( uValueB >> 1 ) ) + ( uOverflow >> 1 ); break; default: // Should never happen return false; } return true; } int test_integer_hadd(cl_device_id deviceID, cl_context context, cl_command_queue queue, int num_elements) { return test_two_param_integer_fn( queue, context, "hadd", verify_integer_hadd ); } bool verify_integer_rhadd( void *sourceA, void *sourceB, void *destination, ExplicitType vecType ) { cl_long testValueA, testValueB, overflow; cl_ulong uValueA, uValueB, uOverflow; switch( vecType ) { case kChar: testValueA = *( (cl_char *)sourceA ); testValueB = *( (cl_char *)sourceB ); *( (cl_char *)destination ) = (cl_char)( ( testValueA + testValueB + 1 ) >> 1 ); break; case kUChar: testValueA = *( (cl_uchar *)sourceA ); testValueB = *( (cl_uchar *)sourceB ); *( (cl_uchar *)destination ) = (cl_uchar)( ( testValueA + testValueB + 1 ) >> 1 ); break; case kShort: testValueA = *( (cl_short *)sourceA ); testValueB = *( (cl_short *)sourceB ); *( (cl_short *)destination ) = (cl_short)( ( testValueA + testValueB + 1 ) >> 1 ); break; case kUShort: testValueA = *( (cl_ushort *)sourceA ); testValueB = *( (cl_ushort *)sourceB ); *( (cl_ushort *)destination ) = (cl_ushort)( ( testValueA + testValueB + 1 ) >> 1 ); break; case kInt: testValueA = *( (cl_int *)sourceA ); testValueB = *( (cl_int *)sourceB ); *( (cl_int *)destination ) = (cl_int)( ( testValueA + testValueB + 1 ) >> 1 ); break; case kUInt: testValueA = *( (cl_uint *)sourceA ); testValueB = *( (cl_uint *)sourceB ); *( (cl_uint *)destination ) = (cl_uint)( ( testValueA + testValueB + 1 ) >> 1 ); break; case kLong: // The long way to avoid dropping bits testValueA = *( (cl_long *)sourceA ); testValueB = *( (cl_long *)sourceB ); overflow = ( testValueA | testValueB ) & 0x1; *( (cl_long *)destination ) = ( ( testValueA >> 1 ) + ( testValueB >> 1 ) ) + overflow; break; case kULong: // The long way to avoid dropping bits uValueA = *( (cl_ulong *)sourceA ); uValueB = *( (cl_ulong *)sourceB ); uOverflow = ( uValueA | uValueB ) & 0x1; *( (cl_ulong *)destination ) = ( ( uValueA >> 1 ) + ( uValueB >> 1 ) ) + uOverflow; break; default: // Should never happen return false; } return true; } int test_integer_rhadd(cl_device_id deviceID, cl_context context, cl_command_queue queue, int num_elements) { return test_two_param_integer_fn( queue, context, "rhadd", verify_integer_rhadd ); } #define MIN_CASE( type, const ) \ case const : \ { \ cl_##type valueA = *( (cl_##type *)sourceA ); \ cl_##type valueB = *( (cl_##type *)sourceB ); \ *( (cl_##type *)destination ) = (cl_##type)( valueB < valueA ? valueB : valueA ); \ break; \ } bool verify_integer_min( void *sourceA, void *sourceB, void *destination, ExplicitType vecType ) { switch( vecType ) { MIN_CASE( char, kChar ) MIN_CASE( uchar, kUChar ) MIN_CASE( short, kShort ) MIN_CASE( ushort, kUShort ) MIN_CASE( int, kInt ) MIN_CASE( uint, kUInt ) MIN_CASE( long, kLong ) MIN_CASE( ulong, kULong ) default: // Should never happen return false; } return true; } int test_integer_min(cl_device_id deviceID, cl_context context, cl_command_queue queue, int num_elements) { return test_two_param_integer_fn( queue, context, "min", verify_integer_min); } #define MAX_CASE( type, const ) \ case const : \ { \ cl_##type valueA = *( (cl_##type *)sourceA ); \ cl_##type valueB = *( (cl_##type *)sourceB ); \ *( (cl_##type *)destination ) = (cl_##type)( valueA < valueB ? valueB : valueA ); \ break; \ } bool verify_integer_max( void *sourceA, void *sourceB, void *destination, ExplicitType vecType ) { switch( vecType ) { MAX_CASE( char, kChar ) MAX_CASE( uchar, kUChar ) MAX_CASE( short, kShort ) MAX_CASE( ushort, kUShort ) MAX_CASE( int, kInt ) MAX_CASE( uint, kUInt ) MAX_CASE( long, kLong ) MAX_CASE( ulong, kULong ) default: // Should never happen return false; } return true; } int test_integer_max(cl_device_id deviceID, cl_context context, cl_command_queue queue, int num_elements) { return test_two_param_integer_fn( queue, context, "max", verify_integer_max ); } void multiply_unsigned_64_by_64( cl_ulong sourceA, cl_ulong sourceB, cl_ulong &destLow, cl_ulong &destHi ) { cl_ulong lowA, lowB; cl_ulong highA, highB; // Split up the values lowA = sourceA & 0xffffffff; highA = sourceA >> 32; lowB = sourceB & 0xffffffff; highB = sourceB >> 32; // Note that, with this split, our multiplication becomes: // ( a * b ) // = ( ( aHI << 32 + aLO ) * ( bHI << 32 + bLO ) ) >> 64 // = ( ( aHI << 32 * bHI << 32 ) + ( aHI << 32 * bLO ) + ( aLO * bHI << 32 ) + ( aLO * bLO ) ) >> 64 // = ( ( aHI * bHI << 64 ) + ( aHI * bLO << 32 ) + ( aLO * bHI << 32 ) + ( aLO * bLO ) ) >> 64 // = ( aHI * bHI ) + ( aHI * bLO >> 32 ) + ( aLO * bHI >> 32 ) + ( aLO * bLO >> 64 ) // Now, since each value is 32 bits, the max size of any multiplication is: // ( 2 ^ 32 - 1 ) * ( 2 ^ 32 - 1 ) = 2^64 - 4^32 + 1 = 2^64 - 2^33 + 1, which fits within 64 bits // Which means we can do each component within a 64-bit integer as necessary (each component above marked as AB1 - AB4) cl_ulong aHibHi = highA * highB; cl_ulong aHibLo = highA * lowB; cl_ulong aLobHi = lowA * highB; cl_ulong aLobLo = lowA * lowB; // Assemble terms. // We note that in certain cases, sums of products cannot overflow: // // The maximum product of two N-bit unsigned numbers is // // (2**N-1)^2 = 2**2N - 2**(N+1) + 1 // // We note that we can add the maximum N-bit number to the 2N-bit product twice without overflow: // // (2**N-1)^2 + 2*(2**N-1) = 2**2N - 2**(N+1) + 1 + 2**(N+1) - 2 = 2**2N - 1 // // If we breakdown the product of two numbers a,b into high and low halves of partial products as follows: // // a.hi a.lo // x b.hi b.lo //=============================================================================== // (b.hi*a.hi).hi (b.hi*a.hi).lo // (b.lo*a.hi).hi (b.lo*a.hi).lo // (b.hi*a.lo).hi (b.hi*a.lo).lo // + (b.lo*a.lo).hi (b.lo*a.lo).lo //=============================================================================== // // The (b.lo*a.lo).lo term cannot cause a carry, so we can ignore them for now. We also know from above, that we can add (b.lo*a.lo).hi // and (b.hi*a.lo).lo to the 2N bit term [(b.lo*a.hi).hi + (b.lo*a.hi).lo] without overflow. That takes care of all of the terms // on the right half that might carry. Do that now. // cl_ulong aLobLoHi = aLobLo >> 32; cl_ulong aLobHiLo = aLobHi & 0xFFFFFFFFULL; aHibLo += aLobLoHi + aLobHiLo; // That leaves us with these terms: // // a.hi a.lo // x b.hi b.lo //=============================================================================== // (b.hi*a.hi).hi (b.hi*a.hi).lo // (b.hi*a.lo).hi // [ (b.lo*a.hi).hi + (b.lo*a.hi).lo + other ] // + (b.lo*a.lo).lo //=============================================================================== // All of the overflow potential from the right half has now been accumulated into the [ (b.lo*a.hi).hi + (b.lo*a.hi).lo ] 2N bit term. // We can safely separate into high and low parts. Per our rule above, we know we can accumulate the high part of that and (b.hi*a.lo).hi // into the 2N bit term (b.lo*a.hi) without carry. The low part can be pieced together with (b.lo*a.lo).lo, to give the final low result destHi = aHibHi + (aHibLo >> 32 ) + (aLobHi >> 32); // Cant overflow destLow = (aHibLo << 32) | ( aLobLo & 0xFFFFFFFFULL ); } void multiply_signed_64_by_64( cl_long sourceA, cl_long sourceB, cl_ulong &destLow, cl_long &destHi ) { // Find sign of result cl_long aSign = sourceA >> 63; cl_long bSign = sourceB >> 63; cl_long resultSign = aSign ^ bSign; // take absolute values of the argument sourceA = (sourceA ^ aSign) - aSign; sourceB = (sourceB ^ bSign) - bSign; cl_ulong hi; multiply_unsigned_64_by_64( (cl_ulong) sourceA, (cl_ulong) sourceB, destLow, hi ); // Fix the sign if( resultSign ) { destLow ^= resultSign; hi ^= resultSign; destLow -= resultSign; //carry if necessary if( 0 == destLow ) hi -= resultSign; } destHi = (cl_long) hi; } bool verify_integer_mul_hi( void *sourceA, void *sourceB, void *destination, ExplicitType vecType ) { cl_long testValueA, testValueB, highSigned; cl_ulong highUnsigned, lowHalf; switch( vecType ) { case kChar: testValueA = *( (cl_char *)sourceA ); testValueB = *( (cl_char *)sourceB ); *( (cl_char *)destination ) = (cl_char)( ( testValueA * testValueB ) >> 8 ); break; case kUChar: testValueA = *( (cl_uchar *)sourceA ); testValueB = *( (cl_uchar *)sourceB ); *( (cl_uchar *)destination ) = (cl_uchar)( ( testValueA * testValueB ) >> 8 ); break; case kShort: testValueA = *( (cl_short *)sourceA ); testValueB = *( (cl_short *)sourceB ); *( (cl_short *)destination ) = (cl_short)( ( testValueA * testValueB ) >> 16 ); break; case kUShort: testValueA = *( (cl_ushort *)sourceA ); testValueB = *( (cl_ushort *)sourceB ); *( (cl_ushort *)destination ) = (cl_ushort)( ( testValueA * testValueB ) >> 16 ); break; case kInt: testValueA = *( (cl_int *)sourceA ); testValueB = *( (cl_int *)sourceB ); *( (cl_int *)destination ) = (cl_int)( ( testValueA * testValueB ) >> 32 ); break; case kUInt: testValueA = *( (cl_uint *)sourceA ); testValueB = *( (cl_uint *)sourceB ); *( (cl_uint *)destination ) = (cl_uint)( ( testValueA * testValueB ) >> 32 ); break; case kLong: testValueA = *( (cl_long *)sourceA ); testValueB = *( (cl_long *)sourceB ); multiply_signed_64_by_64( testValueA, testValueB, lowHalf, highSigned ); *( (cl_long *)destination ) = highSigned; break; case kULong: testValueA = *( (cl_ulong *)sourceA ); testValueB = *( (cl_ulong *)sourceB ); multiply_unsigned_64_by_64( testValueA, testValueB, lowHalf, highUnsigned ); *( (cl_ulong *)destination ) = highUnsigned; break; default: // Should never happen return false; } return true; } int test_integer_mul_hi(cl_device_id deviceID, cl_context context, cl_command_queue queue, int num_elements) { return test_two_param_integer_fn( queue, context, "mul_hi", verify_integer_mul_hi ); } bool verify_integer_rotate( void *sourceA, void *sourceB, void *destination, ExplicitType vecType ) { cl_ulong testValueA; char numBits; switch( vecType ) { case kChar: case kUChar: testValueA = *( (cl_uchar *)sourceA ); numBits = *( (cl_uchar *)sourceB ); numBits &= 7; if ( numBits == 0 ) *( (cl_uchar *)destination ) = (cl_uchar)testValueA; else *( (cl_uchar *)destination ) = (cl_uchar)( ( testValueA << numBits ) | ( testValueA >> ( 8 - numBits ) ) ); break; case kShort: case kUShort: testValueA = *( (cl_ushort *)sourceA ); numBits = *( (cl_ushort *)sourceB ); numBits &= 15; if ( numBits == 0 ) *( (cl_ushort *)destination ) = (cl_ushort)testValueA; else *( (cl_ushort *)destination ) = (cl_ushort)( ( testValueA << numBits ) | ( testValueA >> ( 16 - numBits ) ) ); break; case kInt: case kUInt: testValueA = *( (cl_uint *)sourceA ); numBits = *( (cl_uint *)sourceB ); numBits &= 31; if ( numBits == 0 ) *( (cl_uint *)destination ) = (cl_uint) testValueA; else *( (cl_uint *)destination ) = (cl_uint)( ( testValueA << numBits ) | ( testValueA >> ( 32 - numBits ) ) ); break; case kLong: case kULong: testValueA = *( (cl_ulong *)sourceA ); numBits = *( (cl_ulong *)sourceB ); numBits &= 63; if ( numBits == 0 ) *( (cl_ulong *)destination ) = (cl_ulong)testValueA; else *( (cl_ulong *)destination ) = (cl_ulong)( ( testValueA << numBits ) | ( testValueA >> ( 64 - numBits ) ) ); break; default: // Should never happen log_error( "Unknown type encountered in verify_integer_rotate. Test failed. Aborting...\n" ); abort(); return false; } return true; } int test_integer_rotate(cl_device_id deviceID, cl_context context, cl_command_queue queue, int num_elements) { return test_two_param_integer_fn( queue, context, "rotate", verify_integer_rotate ); } const char *threeParamIntegerKernelSourcePattern = "__kernel void sample_test(__global %s%s *sourceA, __global %s%s *sourceB, __global %s%s *sourceC, __global %s%s *destValues)\n" "{\n" " int tid = get_global_id(0);\n" " %s%s sA = %s;\n" " %s%s sB = %s;\n" " %s%s sC = %s;\n" " %s%s dst = %s( sA, sB, sC );\n" " %s;\n" "\n" "}\n"; typedef bool (*threeParamIntegerVerifyFn)( void *sourceA, void *sourceB, void *sourceC, void *destination, ExplicitType vecAType, ExplicitType vecBType, ExplicitType vecCType, ExplicitType destType ); int test_three_param_integer_kernel(cl_command_queue queue, cl_context context, const char *fnName, ExplicitType vecAType, ExplicitType vecBType, ExplicitType vecCType, ExplicitType destType, unsigned int vecSize, threeParamIntegerVerifyFn verifyFn, MTdata d ) { clProgramWrapper program; clKernelWrapper kernel; clMemWrapper streams[4]; cl_long inDataA[TEST_SIZE * 16], inDataB[TEST_SIZE * 16], inDataC[TEST_SIZE * 16], outData[TEST_SIZE * 16], expected; int error, i; size_t threads[1], localThreads[1]; char kernelSource[10240]; char *programPtr; char sizeName[4], paramSizeName[4]; if (! gHasLong && strstr(get_explicit_type_name(vecAType),"long")) { log_info( "WARNING: 64 bit integers are not supported on this device. Skipping %s\n", get_explicit_type_name(vecAType) ); return CL_SUCCESS; } /* Create the source */ if( vecSize == 1 ) sizeName[ 0 ] = 0; else sprintf( sizeName, "%d", vecSize ); if( ( vecSize == 1 ) || ( vecSize == 3 ) ) paramSizeName[ 0 ] = 0; else sprintf( paramSizeName, "%d", vecSize ); char sourceALoad[ 128 ], sourceBLoad[ 128 ], sourceCLoad[ 128 ], destStore[ 128 ]; sprintf( kernelSource, threeParamIntegerKernelSourcePattern, get_explicit_type_name( vecAType ), paramSizeName, get_explicit_type_name( vecBType ), paramSizeName, get_explicit_type_name( vecCType ), paramSizeName, get_explicit_type_name( destType ), paramSizeName, get_explicit_type_name( vecAType ), sizeName, build_load_statement( sourceALoad, (size_t)vecSize, "sourceA" ), get_explicit_type_name( vecBType ), sizeName, build_load_statement( sourceBLoad, (size_t)vecSize, "sourceB" ), get_explicit_type_name( vecCType ), sizeName, build_load_statement( sourceCLoad, (size_t)vecSize, "sourceC" ), get_explicit_type_name( destType ), sizeName, fnName, build_store_statement( destStore, (size_t)vecSize, "destValues", "dst" ) ); /* Create kernels */ programPtr = kernelSource; if( create_single_kernel_helper( context, &program, &kernel, 1, (const char **)&programPtr, "sample_test" ) ) { log_error("The program we attempted to compile was: \n%s\n", kernelSource); return -1; } /* Generate some streams */ generate_random_data( vecAType, vecSize * TEST_SIZE, d, inDataA ); generate_random_data( vecBType, vecSize * TEST_SIZE, d, inDataB ); generate_random_data( vecCType, vecSize * TEST_SIZE, d, inDataC ); streams[0] = clCreateBuffer( context, CL_MEM_COPY_HOST_PTR, get_explicit_type_size(vecAType) * vecSize * TEST_SIZE, &inDataA, NULL); if( streams[0] == NULL ) { log_error("ERROR: Creating input array A failed!\n"); return -1; } streams[1] = clCreateBuffer( context, CL_MEM_COPY_HOST_PTR, get_explicit_type_size(vecBType) * vecSize * TEST_SIZE, &inDataB, NULL); if( streams[1] == NULL ) { log_error("ERROR: Creating input array B failed!\n"); return -1; } streams[2] = clCreateBuffer( context, CL_MEM_COPY_HOST_PTR, get_explicit_type_size(vecCType) * vecSize * TEST_SIZE, &inDataC, NULL); if( streams[2] == NULL ) { log_error("ERROR: Creating input array C failed!\n"); return -1; } streams[3] = clCreateBuffer( context, CL_MEM_READ_WRITE, get_explicit_type_size(destType) * vecSize * TEST_SIZE, NULL, NULL); if( streams[3] == NULL ) { log_error("ERROR: Creating output array failed!\n"); return -1; } /* Assign streams and execute */ error = clSetKernelArg( kernel, 0, sizeof( streams[0] ), &streams[0] ); test_error( error, "Unable to set indexed kernel arguments" ); error = clSetKernelArg( kernel, 1, sizeof( streams[1] ), &streams[1] ); test_error( error, "Unable to set indexed kernel arguments" ); error = clSetKernelArg( kernel, 2, sizeof( streams[2] ), &streams[2] ); test_error( error, "Unable to set indexed kernel arguments" ); error = clSetKernelArg( kernel, 3, sizeof( streams[3] ), &streams[3] ); test_error( error, "Unable to set indexed kernel arguments" ); /* Run the kernel */ threads[0] = TEST_SIZE; error = get_max_common_work_group_size( context, kernel, threads[0], &localThreads[0] ); test_error( error, "Unable to get work group size to use" ); error = clEnqueueNDRangeKernel( queue, kernel, 1, NULL, threads, localThreads, 0, NULL, NULL ); test_error( error, "Unable to execute test kernel" ); memset(outData, 0xFF, get_explicit_type_size( destType ) * TEST_SIZE * vecSize); /* Now get the results */ error = clEnqueueReadBuffer( queue, streams[3], CL_TRUE, 0, get_explicit_type_size( destType ) * TEST_SIZE * vecSize, outData, 0, NULL, NULL ); test_error( error, "Unable to read output array!" ); /* And verify! */ char *inA = (char *)inDataA; char *inB = (char *)inDataB; char *inC = (char *)inDataC; char *out = (char *)outData; for( i = 0; i < (int)TEST_SIZE; i++ ) { for( size_t j = 0; j < vecSize; j++ ) { bool test = verifyFn( inA, inB, inC, &expected, vecAType, vecBType, vecCType, destType ); if( test && ( memcmp( &expected, out, get_explicit_type_size( destType ) ) != 0 ) ) { switch( get_explicit_type_size( vecAType )) { case 1: log_error( "ERROR: Data sample %d:%d does not validate! Expected (0x%2.2x), got (0x%2.2x), sources (0x%2.2x, 0x%2.2x, 0x%2.2x)\n", (int)i, (int)j, ((cl_uchar*)&expected)[ 0 ], *( (cl_uchar *)out ), *( (cl_uchar *)inA ), *( (cl_uchar *)inB ), *( (cl_uchar *)inC ) ); break; case 2: log_error( "ERROR: Data sample %d:%d does not validate! Expected (0x%4.4x), got (0x%4.4x), sources (0x%4.4x, 0x%4.4x, 0x%4.4x)\n", (int)i, (int)j, ((cl_ushort*)&expected)[ 0 ], *( (cl_ushort *)out ), *( (cl_ushort *)inA ), *( (cl_ushort *)inB ), *( (cl_ushort *)inC ) ); break; case 4: log_error( "ERROR: Data sample %d:%d does not validate! Expected (0x%8.8x), got (0x%8.8x), sources (0x%8.8x, 0x%8.8x, 0x%8.8x)\n", (int)i, (int)j, ((cl_uint*)&expected)[ 0 ], *( (cl_uint *)out ), *( (cl_uint *)inA ), *( (cl_uint *)inB ), *( (cl_uint *)inC ) ); break; case 8: log_error( "ERROR: Data sample %d:%d does not validate! Expected (0x%16.16llx), got (0x%16.16llx), sources (0x%16.16llx, 0x%16.16llx, 0x%16.16llx)\n", (int)i, (int)j, ((cl_ulong*)&expected)[ 0 ], *( (cl_ulong *)out ), *( (cl_ulong *)inA ), *( (cl_ulong *)inB ), *( (cl_ulong *)inC ) ); break; } return -1; } inA += get_explicit_type_size( vecAType ); inB += get_explicit_type_size( vecBType ); inC += get_explicit_type_size( vecCType ); out += get_explicit_type_size( destType ); } } return 0; } int test_three_param_integer_fn(cl_command_queue queue, cl_context context, const char *fnName, threeParamIntegerVerifyFn verifyFn) { ExplicitType types[] = { kChar, kUChar, kShort, kUShort, kInt, kUInt, kLong, kULong, kNumExplicitTypes }; unsigned int vecSizes[] = { 1, 2, 3, 4, 8, 16, 0 }; unsigned int index, typeAIndex; int retVal = 0; RandomSeed seed(gRandomSeed); for( typeAIndex = 0; types[ typeAIndex ] != kNumExplicitTypes; typeAIndex++ ) { if ((types[ typeAIndex ] == kLong || types[ typeAIndex] == kULong) && !gHasLong) continue; for( index = 0; vecSizes[ index ] != 0; index++ ) { if( test_three_param_integer_kernel(queue, context, fnName, types[ typeAIndex ], types[ typeAIndex ], types[ typeAIndex ], types[ typeAIndex ], vecSizes[ index ], verifyFn, seed ) != 0 ) { log_error( " Vector %s%d,%s%d,%s%d FAILED\n", get_explicit_type_name( types[ typeAIndex ] ), vecSizes[ index ], get_explicit_type_name( types[ typeAIndex ] ), vecSizes[ index ] , get_explicit_type_name( types[ typeAIndex ] ), vecSizes[ index ] ); retVal = -1; } } } return retVal; } bool verify_integer_clamp( void *sourceA, void *sourceB, void *sourceC, void *destination, ExplicitType vecAType, ExplicitType vecBType, ExplicitType vecCType, ExplicitType destType ) { if( vecAType == kULong || vecAType == kUInt || vecAType == kUShort || vecAType == kUChar ) { cl_ulong valueA, valueB, valueC; switch( vecAType ) { case kULong: valueA = ((cl_ulong*) sourceA)[0]; valueB = ((cl_ulong*) sourceB)[0]; valueC = ((cl_ulong*) sourceC)[0]; break; case kUInt: valueA = ((cl_uint*) sourceA)[0]; valueB = ((cl_uint*) sourceB)[0]; valueC = ((cl_uint*) sourceC)[0]; break; case kUShort: valueA = ((cl_ushort*) sourceA)[0]; valueB = ((cl_ushort*) sourceB)[0]; valueC = ((cl_ushort*) sourceC)[0]; break; case kUChar: valueA = ((cl_uchar*) sourceA)[0]; valueB = ((cl_uchar*) sourceB)[0]; valueC = ((cl_uchar*) sourceC)[0]; break; default: //error -- should never get here abort(); break; } if(valueB > valueC) { return false; // results are undefined : let expected alone. } switch( vecAType ) { case kULong: ((cl_ulong*) destination)[0] = MAX(MIN(valueA, valueC), valueB); break; case kUInt: ((cl_uint*) destination)[0] = (cl_uint) (MAX(MIN(valueA, valueC), valueB)); break; case kUShort: ((cl_ushort*) destination)[0] = (cl_ushort) (MAX(MIN(valueA, valueC), valueB)); break; case kUChar: ((cl_uchar*) destination)[0] = (cl_uchar) (MAX(MIN(valueA, valueC), valueB)); break; default: //error -- should never get here abort(); break; } } else { cl_long valueA, valueB, valueC; switch( vecAType ) { case kLong: valueA = ((cl_long*) sourceA)[0]; valueB = ((cl_long*) sourceB)[0]; valueC = ((cl_long*) sourceC)[0]; break; case kInt: valueA = ((cl_int*) sourceA)[0]; valueB = ((cl_int*) sourceB)[0]; valueC = ((cl_int*) sourceC)[0]; break; case kShort: valueA = ((cl_short*) sourceA)[0]; valueB = ((cl_short*) sourceB)[0]; valueC = ((cl_short*) sourceC)[0]; break; case kChar: valueA = ((cl_char*) sourceA)[0]; valueB = ((cl_char*) sourceB)[0]; valueC = ((cl_char*) sourceC)[0]; break; default: //error -- should never get here abort(); break; } if(valueB > valueC) { return false; // undefined behavior : leave "expected" alone } switch( vecAType ) { case kLong: ((cl_long*) destination)[0] = MAX(MIN(valueA, valueC), valueB); break; case kInt: ((cl_int*) destination)[0] = (cl_int) (MAX(MIN(valueA, valueC), valueB)); break; case kShort: ((cl_short*) destination)[0] = (cl_short) (MAX(MIN(valueA, valueC), valueB)); break; case kChar: ((cl_char*) destination)[0] = (cl_char) (MAX(MIN(valueA, valueC), valueB)); break; default: //error -- should never get here abort(); break; } } return true; } int test_integer_clamp(cl_device_id deviceID, cl_context context, cl_command_queue queue, int num_elements) { return test_three_param_integer_fn( queue, context, "clamp", verify_integer_clamp ); } bool verify_integer_mad_sat( void *sourceA, void *sourceB, void *sourceC, void *destination, ExplicitType vecAType, ExplicitType vecBType, ExplicitType vecCType, ExplicitType destType ) { if( vecAType == kULong || vecAType == kUInt || vecAType == kUShort || vecAType == kUChar ) { cl_ulong valueA, valueB, valueC; switch( vecAType ) { case kULong: valueA = ((cl_ulong*) sourceA)[0]; valueB = ((cl_ulong*) sourceB)[0]; valueC = ((cl_ulong*) sourceC)[0]; break; case kUInt: valueA = ((cl_uint*) sourceA)[0]; valueB = ((cl_uint*) sourceB)[0]; valueC = ((cl_uint*) sourceC)[0]; break; case kUShort: valueA = ((cl_ushort*) sourceA)[0]; valueB = ((cl_ushort*) sourceB)[0]; valueC = ((cl_ushort*) sourceC)[0]; break; case kUChar: valueA = ((cl_uchar*) sourceA)[0]; valueB = ((cl_uchar*) sourceB)[0]; valueC = ((cl_uchar*) sourceC)[0]; break; default: //error -- should never get here abort(); break; } cl_ulong multHi, multLo; multiply_unsigned_64_by_64( valueA, valueB, multLo, multHi ); multLo += valueC; multHi += multLo < valueC; // carry if overflow if( multHi ) multLo = 0xFFFFFFFFFFFFFFFFULL; switch( vecAType ) { case kULong: ((cl_ulong*) destination)[0] = multLo; break; case kUInt: ((cl_uint*) destination)[0] = (cl_uint) MIN( multLo, (cl_ulong) CL_UINT_MAX ); break; case kUShort: ((cl_ushort*) destination)[0] = (cl_ushort) MIN( multLo, (cl_ulong) CL_USHRT_MAX ); break; case kUChar: ((cl_uchar*) destination)[0] = (cl_uchar) MIN( multLo, (cl_ulong) CL_UCHAR_MAX ); break; default: //error -- should never get here abort(); break; } } else { cl_long valueA, valueB, valueC; switch( vecAType ) { case kLong: valueA = ((cl_long*) sourceA)[0]; valueB = ((cl_long*) sourceB)[0]; valueC = ((cl_long*) sourceC)[0]; break; case kInt: valueA = ((cl_int*) sourceA)[0]; valueB = ((cl_int*) sourceB)[0]; valueC = ((cl_int*) sourceC)[0]; break; case kShort: valueA = ((cl_short*) sourceA)[0]; valueB = ((cl_short*) sourceB)[0]; valueC = ((cl_short*) sourceC)[0]; break; case kChar: valueA = ((cl_char*) sourceA)[0]; valueB = ((cl_char*) sourceB)[0]; valueC = ((cl_char*) sourceC)[0]; break; default: //error -- should never get here abort(); break; } cl_long multHi; cl_ulong multLo; multiply_signed_64_by_64( valueA, valueB, multLo, multHi ); cl_ulong sum = multLo + valueC; // carry if overflow if( valueC >= 0 ) { if( multLo > sum ) { multHi++; if( CL_LONG_MIN == multHi ) { multHi = CL_LONG_MAX; sum = CL_ULONG_MAX; } } } else { if( multLo < sum ) { multHi--; if( CL_LONG_MAX == multHi ) { multHi = CL_LONG_MIN; sum = 0; } } } // saturate if( multHi > 0 ) sum = CL_LONG_MAX; else if( multHi < -1 ) sum = CL_LONG_MIN; cl_long result = (cl_long) sum; switch( vecAType ) { case kLong: ((cl_long*) destination)[0] = result; break; case kInt: result = MIN( result, (cl_long) CL_INT_MAX ); result = MAX( result, (cl_long) CL_INT_MIN ); ((cl_int*) destination)[0] = (cl_int) result; break; case kShort: result = MIN( result, (cl_long) CL_SHRT_MAX ); result = MAX( result, (cl_long) CL_SHRT_MIN ); ((cl_short*) destination)[0] = (cl_short) result; break; case kChar: result = MIN( result, (cl_long) CL_CHAR_MAX ); result = MAX( result, (cl_long) CL_CHAR_MIN ); ((cl_char*) destination)[0] = (cl_char) result; break; default: //error -- should never get here abort(); break; } } return true; } int test_integer_mad_sat(cl_device_id deviceID, cl_context context, cl_command_queue queue, int num_elements) { return test_three_param_integer_fn( queue, context, "mad_sat", verify_integer_mad_sat ); } bool verify_integer_mad_hi( void *sourceA, void *sourceB, void *sourceC, void *destination, ExplicitType vecAType, ExplicitType vecBType, ExplicitType vecCType, ExplicitType destType ) { if( vecAType == kULong || vecAType == kUInt || vecAType == kUShort || vecAType == kUChar ) { cl_ulong valueA, valueB, valueC; switch( vecAType ) { case kULong: valueA = ((cl_ulong*) sourceA)[0]; valueB = ((cl_ulong*) sourceB)[0]; valueC = ((cl_ulong*) sourceC)[0]; break; case kUInt: valueA = ((cl_uint*) sourceA)[0]; valueB = ((cl_uint*) sourceB)[0]; valueC = ((cl_uint*) sourceC)[0]; break; case kUShort: valueA = ((cl_ushort*) sourceA)[0]; valueB = ((cl_ushort*) sourceB)[0]; valueC = ((cl_ushort*) sourceC)[0]; break; case kUChar: valueA = ((cl_uchar*) sourceA)[0]; valueB = ((cl_uchar*) sourceB)[0]; valueC = ((cl_uchar*) sourceC)[0]; break; default: //error -- should never get here abort(); break; } cl_ulong multHi, multLo; multiply_unsigned_64_by_64( valueA, valueB, multLo, multHi ); switch( vecAType ) { case kULong: ((cl_ulong*) destination)[0] = multHi + valueC; break; case kUInt: ((cl_uint*) destination)[0] = (cl_uint) (( multLo >> 32) + valueC ); break; case kUShort: ((cl_ushort*) destination)[0] = (cl_ushort) (( multLo >> 16) + valueC ); break; case kUChar: ((cl_uchar*) destination)[0] = (cl_uchar) (( multLo >> 8) + valueC ); break; default: //error -- should never get here abort(); break; } } else { cl_long valueA, valueB, valueC; switch( vecAType ) { case kLong: valueA = ((cl_long*) sourceA)[0]; valueB = ((cl_long*) sourceB)[0]; valueC = ((cl_long*) sourceC)[0]; break; case kInt: valueA = ((cl_int*) sourceA)[0]; valueB = ((cl_int*) sourceB)[0]; valueC = ((cl_int*) sourceC)[0]; break; case kShort: valueA = ((cl_short*) sourceA)[0]; valueB = ((cl_short*) sourceB)[0]; valueC = ((cl_short*) sourceC)[0]; break; case kChar: valueA = ((cl_char*) sourceA)[0]; valueB = ((cl_char*) sourceB)[0]; valueC = ((cl_char*) sourceC)[0]; break; default: //error -- should never get here abort(); break; } cl_long multHi; cl_ulong multLo; multiply_signed_64_by_64( valueA, valueB, multLo, multHi ); switch( vecAType ) { case kLong: ((cl_long*) destination)[0] = multHi + valueC; break; case kInt: ((cl_int*) destination)[0] = (cl_int) ((multLo >> 32) + valueC); break; case kShort: ((cl_short*) destination)[0] = (cl_int) ((multLo >> 16) + valueC); break; case kChar: ((cl_char*) destination)[0] = (cl_char) (cl_int) ((multLo >> 8) + valueC); break; default: //error -- should never get here abort(); break; } } return true; } int test_integer_mad_hi( cl_device_id deviceID, cl_context context, cl_command_queue queue, int num_elements) { return test_three_param_integer_fn( queue, context, "mad_hi", verify_integer_mad_hi ); }