diff --git a/features/feature_case/math/math-emu-half2.cu b/features/feature_case/math/math-emu-half2.cu index 747673949..30341561b 100644 --- a/features/feature_case/math/math-emu-half2.cu +++ b/features/feature_case/math/math-emu-half2.cu @@ -562,7 +562,7 @@ void testHle2(float *const Result, __half2 Input1, __half2 Input2) { void testHle2Cases(const vector> &TestCases) { float *Result; - cudaMallocManaged(&Result, sizeof(*Result)); + cudaMallocManaged(&Result, 2 * sizeof(*Result)); for (const auto &TestCase : TestCases) { testHle2(Result, TestCase.first.first, TestCase.first.second); checkResult("__hle2", {TestCase.first.first, TestCase.first.second}, @@ -584,7 +584,7 @@ void testHleu2(float *const Result, __half2 Input1, __half2 Input2) { void testHleu2Cases(const vector> &TestCases) { float *Result; - cudaMallocManaged(&Result, sizeof(*Result)); + cudaMallocManaged(&Result, 2 * sizeof(*Result)); for (const auto &TestCase : TestCases) { testHleu2(Result, TestCase.first.first, TestCase.first.second); checkResult("__hleu2", {TestCase.first.first, TestCase.first.second}, @@ -606,7 +606,7 @@ void testHlt2(float *const Result, __half2 Input1, __half2 Input2) { void testHlt2Cases(const vector> &TestCases) { float *Result; - cudaMallocManaged(&Result, sizeof(*Result)); + cudaMallocManaged(&Result, 2 * sizeof(*Result)); for (const auto &TestCase : TestCases) { testHlt2(Result, TestCase.first.first, TestCase.first.second); checkResult("__hlt2", {TestCase.first.first, TestCase.first.second}, @@ -628,7 +628,7 @@ void testHltu2(float *const Result, __half2 Input1, __half2 Input2) { void testHltu2Cases(const vector> &TestCases) { float *Result; - cudaMallocManaged(&Result, sizeof(*Result)); + cudaMallocManaged(&Result, 2 * sizeof(*Result)); for (const auto &TestCase : TestCases) { testHltu2(Result, TestCase.first.first, TestCase.first.second); checkResult("__hltu2", {TestCase.first.first, TestCase.first.second}, @@ -650,7 +650,7 @@ void testHne2(float *const Result, __half2 Input1, __half2 Input2) { void testHne2Cases(const vector> &TestCases) { float *Result; - cudaMallocManaged(&Result, sizeof(*Result)); + cudaMallocManaged(&Result, 2 * sizeof(*Result)); for (const auto &TestCase : TestCases) { testHne2(Result, TestCase.first.first, TestCase.first.second); checkResult("__hne2", {TestCase.first.first, TestCase.first.second}, @@ -672,7 +672,7 @@ void testHneu2(float *const Result, __half2 Input1, __half2 Input2) { void testHneu2Cases(const vector> &TestCases) { float *Result; - cudaMallocManaged(&Result, sizeof(*Result)); + cudaMallocManaged(&Result, 2 * sizeof(*Result)); for (const auto &TestCase : TestCases) { testHneu2(Result, TestCase.first.first, TestCase.first.second); checkResult("__hneu2", {TestCase.first.first, TestCase.first.second}, diff --git a/features/feature_case/math/math-ext-half2.cu b/features/feature_case/math/math-ext-half2.cu index f4753c86d..f98602409 100644 --- a/features/feature_case/math/math-ext-half2.cu +++ b/features/feature_case/math/math-ext-half2.cu @@ -561,7 +561,7 @@ void testHle2(float *const Result, __half2 Input1, __half2 Input2) { void testHle2Cases(const vector> &TestCases) { float *Result; - cudaMallocManaged(&Result, sizeof(*Result)); + cudaMallocManaged(&Result, 2 * sizeof(*Result)); for (const auto &TestCase : TestCases) { testHle2(Result, TestCase.first.first, TestCase.first.second); checkResult("__hle2", {TestCase.first.first, TestCase.first.second}, @@ -583,7 +583,7 @@ void testHleu2(float *const Result, __half2 Input1, __half2 Input2) { void testHleu2Cases(const vector> &TestCases) { float *Result; - cudaMallocManaged(&Result, sizeof(*Result)); + cudaMallocManaged(&Result, 2 * sizeof(*Result)); for (const auto &TestCase : TestCases) { testHleu2(Result, TestCase.first.first, TestCase.first.second); checkResult("__hleu2", {TestCase.first.first, TestCase.first.second}, @@ -605,7 +605,7 @@ void testHlt2(float *const Result, __half2 Input1, __half2 Input2) { void testHlt2Cases(const vector> &TestCases) { float *Result; - cudaMallocManaged(&Result, sizeof(*Result)); + cudaMallocManaged(&Result, 2 * sizeof(*Result)); for (const auto &TestCase : TestCases) { testHlt2(Result, TestCase.first.first, TestCase.first.second); checkResult("__hlt2", {TestCase.first.first, TestCase.first.second}, @@ -627,7 +627,7 @@ void testHltu2(float *const Result, __half2 Input1, __half2 Input2) { void testHltu2Cases(const vector> &TestCases) { float *Result; - cudaMallocManaged(&Result, sizeof(*Result)); + cudaMallocManaged(&Result, 2 * sizeof(*Result)); for (const auto &TestCase : TestCases) { testHltu2(Result, TestCase.first.first, TestCase.first.second); checkResult("__hltu2", {TestCase.first.first, TestCase.first.second}, @@ -649,7 +649,7 @@ void testHne2(float *const Result, __half2 Input1, __half2 Input2) { void testHne2Cases(const vector> &TestCases) { float *Result; - cudaMallocManaged(&Result, sizeof(*Result)); + cudaMallocManaged(&Result, 2 * sizeof(*Result)); for (const auto &TestCase : TestCases) { testHne2(Result, TestCase.first.first, TestCase.first.second); checkResult("__hne2", {TestCase.first.first, TestCase.first.second}, @@ -671,7 +671,7 @@ void testHneu2(float *const Result, __half2 Input1, __half2 Input2) { void testHneu2Cases(const vector> &TestCases) { float *Result; - cudaMallocManaged(&Result, sizeof(*Result)); + cudaMallocManaged(&Result, 2 * sizeof(*Result)); for (const auto &TestCase : TestCases) { testHneu2(Result, TestCase.first.first, TestCase.first.second); checkResult("__hneu2", {TestCase.first.first, TestCase.first.second},