SWDEV-299773 - Enable performance tests on NV (#2337)

1. Simply enable test on NV Some need minor fix performance/compute/hipPerfDotProduct.cpp performance/dispatch/hipPerfDispatchSpeed.cpp performance/memory/hipPerfBufferCopyRectSpeed.cpp performance/memory/hipPerfBufferCopySpeed.cpp performance/memory/hipPerfDevMemReadSpeed.cpp performance/memory/hipPerfDevMemWriteSpeed.cpp performance/memory/hipPerfMemcpy.cpp performance/memory/hipPerfMemset.cpp performance/memory/hipPerfSharedMemReadSpeed.cpp performance/stream/hipPerfDeviceConcurrency.cpp performance/stream/hipPerfStreamCreateCopyDestroy.cpp 2. Enable and fix on NV performance/compute/hipPerfMandelbrot.cpp Root cause: coordIdx is random Solution: Initialize coordIdx correctly performance/memory/hipPerfMemFill.cpp Root cause: Hip ext Apis called. Solution: Exclude case with Hip ext Apis involved performance/memory/hipPerfMemMallocCpyFree.cpp Root cause: Test allocates device memory more than GPU has. Solution: Allocate device memory in terms of GPU capacity. tests/performance/memory/hipPerfSampleRate.cpp Root cause: Cuda has no operators += for float2 and float4. Solution: Provide the operators. performance/stream/hipPerfStreamConcurrency.cpp Root cause:float4 format doesn't match cude. operators are missing in cuda lib. Solution: Use (x, y, z, w) format. Add necessary float4 operatoris for cuda. Change-Id: I5add29ebabcfb21fb3ef89d09004c5d13423a291
2021-09-14 04:07:13 -04:00
@@ -18,7 +18,7 @@ THE SOFTWARE.
 */

 /* HIT_START
- * BUILD: %t %s ../../src/test_common.cpp ../../src/timer.cpp EXCLUDE_HIP_PLATFORM nvidia
+ * BUILD: %t %s ../../src/test_common.cpp ../../src/timer.cpp
 * TEST: %t
 * HIT_END
 */
@@ -18,7 +18,7 @@ THE SOFTWARE.
 */

 /* HIT_START
- * BUILD: %t %s ../../src/test_common.cpp ../../src/timer.cpp EXCLUDE_HIP_PLATFORM nvidia
+ * BUILD: %t %s ../../src/test_common.cpp ../../src/timer.cpp
 * TEST: %t
 * HIT_END
 */
@@ -18,7 +18,7 @@ THE SOFTWARE.
 */

 /* HIT_START
- * BUILD: %t %s ../../src/test_common.cpp EXCLUDE_HIP_PLATFORM nvidia
+ * BUILD: %t %s ../../src/test_common.cpp
 * TEST: %t
 * HIT_END
 */
@@ -18,7 +18,7 @@ THE SOFTWARE.
 */

 /* HIT_START
- * BUILD: %t %s ../../src/test_common.cpp EXCLUDE_HIP_PLATFORM nvidia
+ * BUILD: %t %s ../../src/test_common.cpp
 * TEST: %t
 * HIT_END
 */
@@ -18,7 +18,7 @@
 */

 /* HIT_START
- * BUILD: %t %s ../../src/test_common.cpp EXCLUDE_HIP_PLATFORM nvidia
+ * BUILD: %t %s ../../src/test_common.cpp
 * TEST: %t
 * HIT_END
 */
@@ -136,6 +136,7 @@ class hipPerfMemFill {
    }

    HIPCHECK(hipSetDevice(deviceId));
+    memset(&props_, 0, sizeof(props_));
    HIPCHECK(hipGetDeviceProperties(&props_, deviceId));
    std::cout << "Info: running on device: id: " << deviceId << ", bus: 0x"
        << props_.pciBusID << " " << props_.name << " with "
@@ -397,8 +398,9 @@ class hipPerfMemFill {
    return true;
  }

-  /* This fuction should be via device attribute query*/
+  /* This function should be via device attribute query*/
  bool supportDeviceMallocFinegrained() {
+#ifdef __HIP_PLATFORM_AMD__
    T *A = nullptr;
    hipExtMallocWithFlags((void **)&A, sizeof(T), hipDeviceMallocFinegrained);
    if (!A) {
@@ -406,6 +408,9 @@ class hipPerfMemFill {
    }
    HIPCHECK(hipFree(A));
    return true;
+#else
+    return false;
+#endif
  }

  unsigned int setNumBlocks(size_t size) {
@@ -419,6 +424,7 @@ class hipPerfMemFill {
 #endif
  }

+#ifdef __HIP_PLATFORM_AMD__
  bool testExtDeviceMemoryHostFill(size_t size, unsigned int flags) {
    double GBytes = (double) size / (1024.0 * 1024.0 * 1024.0);

@@ -481,6 +487,7 @@ class hipPerfMemFill {

    return true;
  }
+#endif

  bool run() {
    if (supportLargeBar()) {
@@ -499,11 +506,13 @@ class hipPerfMemFill {
      return false;
    }

+#ifdef __HIP_PLATFORM_AMD__
    if (supportDeviceMallocFinegrained()) {
      if (!testExtDeviceMemory()) {
        return false;
      }
    }
+#endif
    return true;
  }

@@ -22,7 +22,7 @@ THE SOFTWARE.
 #include <time.h>

 /* HIT_START
- * BUILD: %t %s ../../src/test_common.cpp EXCLUDE_HIP_PLATFORM nvidia
+ * BUILD: %t %s ../../src/test_common.cpp
 * TEST: %t
 * HIT_END
 */
@@ -37,10 +37,15 @@ void valSet(int* A, int val, size_t size) {
    }
 }

-void setup(size_t *size, const int num, int **pA) {
+void setup(size_t *size, int &num, int **pA, const size_t totalGlobalMem) {
+
    std::cout << "size: ";
    for (int i = 0; i < num; i++) {
        size[i] = 1 << (i + 6);
+        if((NUM_ITER + 1) * size[i] > totalGlobalMem) {
+          num = i;
+          break;
+        }
        std::cout << size[i] << " ";
    }
    std::cout << std::endl;
@@ -77,11 +82,16 @@ int main() {
    size_t size[NUM_SIZE] = { 0 };
    int *Ad[NUM_ITER] = { nullptr };
    int *A;
+    hipDeviceProp_t props;
+    memset(&props, 0, sizeof(props));
+    HIPCHECK(hipGetDeviceProperties(&props, 0));
+    std::cout << "totalGlobalMem: " << props.totalGlobalMem << std::endl;

-    setup(size, NUM_SIZE, &A);
+    int num = NUM_SIZE;
+    setup(size, num, &A, props.totalGlobalMem);
    testInit(size[0], A);

-    for (int i = 0; i < NUM_SIZE; i++) {
+    for (int i = 0; i < num; i++) {
        std::cout << size[i] << std::endl;
        start = clock();
        for (int j = 0; j < NUM_ITER; j++) {
@@ -18,7 +18,7 @@
 */

 /* HIT_START
- * BUILD: %t %s ../../src/test_common.cpp EXCLUDE_HIP_PLATFORM nvidia
+ * BUILD: %t %s ../../src/test_common.cpp
 * TEST: %t
 * HIT_END
 */
@@ -18,7 +18,7 @@
 */

 /* HIT_START
- * BUILD: %t %s ../../src/test_common.cpp EXCLUDE_HIP_PLATFORM nvidia
+ * BUILD: %t %s ../../src/test_common.cpp
 * TEST: %t
 * HIT_END
 */
@@ -140,13 +140,13 @@ void hipPerfMemset::run1D(unsigned int test, T memsetval, enum MemsetType type,
  HIPCHECK(hipStreamCreate(&stream));

  // Warm-up
-  HIPCHECK(hipMemset((hipDeviceptr_t)A_d, memsetval, bufSize_));
+  HIPCHECK(hipMemset((void *)A_d, memsetval, bufSize_));

  auto start = chrono::steady_clock::now();

  for (uint i = 0; i < NUM_ITER; i++) {
    if (type == hipMemsetTypeDefault && !async) {
-      HIPCHECK(hipMemset((hipDeviceptr_t)A_d, memsetval, bufSize_));
+      HIPCHECK(hipMemset((void *)A_d, memsetval, bufSize_));
    }
    else if (type == hipMemsetTypeDefault && async) {
      HIPCHECK(hipMemsetAsync(A_d, memsetval, bufSize_, stream));
@@ -18,7 +18,7 @@
 */

 /* HIT_START
- * BUILD: %t %s ../../src/test_common.cpp EXCLUDE_HIP_PLATFORM nvidia
+ * BUILD: %t %s ../../src/test_common.cpp
 * TEST: %t
 * HIT_END
 */
@@ -42,6 +42,18 @@ vector<unsigned int> sizes = {1,  2,   4,   8,   16,   32,
 #define NUM_BUFS 6
 #define MAX_BUFS (1 << (NUM_BUFS - 1))

+#ifdef __HIP_PLATFORM_NVIDIA__
+inline __host__ __device__ void operator+=(float2 &a, float2 b)
+{
+  a.x += b.x; a.y += b.y;
+}
+
+inline __host__ __device__ void operator+=(float4 &a, float4 b)
+{
+  a.x += b.x; a.y += b.y; a.z += b.z; a.w += b.w;
+}
+#endif
+
 template <typename T>
 __global__ void sampleRate(T * outBuffer, unsigned int inBufSize, unsigned int writeIt,
                           T **inBuffer, int numBufs) {
@@ -49,7 +61,8 @@ __global__ void sampleRate(T * outBuffer, unsigned int inBufSize, unsigned int w
  uint gid = (blockIdx.x * blockDim.x + threadIdx.x);
  uint inputIdx = gid % inBufSize;

-  T tmp = (T)0.0f;
+  T tmp;
+  memset(&tmp, 0, sizeof(T));
  for(int i = 0; i < numBufs; i++) {
    tmp += *(*(inBuffer+i)+inputIdx);
  }
@@ -264,11 +277,11 @@ void hipPerfSampleRate::run(unsigned int test) {

   // Free host and device memory
   for (uint i = 0; i < numBufs_; i++) {
-    HIPCHECK(hipFree(hInPtr[i]));
+    HIPCHECK(hipHostFree(hInPtr[i]));
    HIPCHECK(hipFree(dInPtr[i]));
    }

-   HIPCHECK(hipFree(hOutPtr));
+   HIPCHECK(hipHostFree(hOutPtr));
   HIPCHECK(hipFree(dPtr));
 }

@@ -18,7 +18,7 @@
 */

 /* HIT_START
- * BUILD: %t %s ../../src/test_common.cpp EXCLUDE_HIP_PLATFORM nvidia
+ * BUILD: %t %s ../../src/test_common.cpp
 * TEST: %t
 * HIT_END
 */