@@ -71,18 +71,14 @@ fn main(@builtin(global_invocation_id) gid: vec3<u32>) {
7171void testNumTypeSizes () {
7272 LOG (kDefLog , kInfo , " Running testNumTypeSizes..." );
7373
74- // kf16 and kf32 expected sizes
75- // Adjust these values if your implementation differs.
74+
7675 assert (sizeBytes (kf16) == 2 );
7776 assert (sizeBytes (kf32) == 4 );
78-
79- // For the integer types, we compare against the sizeof the respective type.
8077 assert (sizeBytes (ki8) == sizeof (uint8_t )); // typically 1
8178 assert (sizeBytes (ki16) == sizeof (uint16_t )); // typically 2
8279 assert (sizeBytes (ki32) == sizeof (int32_t )); // typically 4
8380 assert (sizeBytes (ku8) == sizeof (uint8_t )); // typically 1
8481 assert (sizeBytes (ku16) == sizeof (uint16_t )); // typically 2
85- // Assuming ku32 should be sizeof(uint32_t)
8682 assert (sizeBytes (ku32) == sizeof (uint32_t )); // typically 4
8783
8884 LOG (kDefLog , kInfo , " testNumTypeSizes passed." );
@@ -105,7 +101,6 @@ void testToCPUWithHalf() {
105101 inputData[i] = half (static_cast <float >(i));
106102 }
107103
108- // Create a tensor for half data using the kf16 type.
109104 Tensor inputTensor = createTensor (ctx, Shape{N}, kf16, inputData.data ());
110105
111106 // Copy GPU output to CPU.
@@ -138,7 +133,6 @@ void testToCPUWithFloat() {
138133 outputData[i] = 0 .0f ;
139134 }
140135
141- // Create a tensor for float data using the kf32 type.
142136 Tensor inputTensor = createTensor (ctx, Shape{N}, kf32, inputData.data ());
143137
144138 // Copy GPU output to CPU.
@@ -299,7 +293,6 @@ void testToCPUWithint64() {
299293 outputData[i] = 0 ;
300294 }
301295
302- // Assuming a new NumType 'ki64' for 64-bit integers.
303296 Tensor inputTensor = createTensor (ctx, Shape{N}, ki64, inputData.data ());
304297
305298 // Copy GPU output to CPU.
@@ -331,7 +324,6 @@ void testToCPUWithUint8() {
331324 Tensor inputTensor = createTensor (
332325 ctx, Shape{N}, ku8, reinterpret_cast <const uint8_t *>(inputData.data ()));
333326
334- // Synchronously copy GPU output to CPU using the tensor overload.
335327 toCPU (ctx, inputTensor, outputData.data (), sizeof (outputData));
336328
337329 // Verify the output matches the input.
0 commit comments