← Back to Leaderboard

The AI CUDA Engineer 👷

42_Max_Pooling_2D • textured_modular_unroll_base_base

Level 1 • Task 42
import torch
import torch.nn as nn
import torch.nn.functional as F


def module_fn(
    x: torch.Tensor,
    kernel_size: int,
    stride: int,
    padding: int,
    dilation: int,
) -> torch.Tensor:
    """
    Applies Max Pooling 2D using functional interface.

    Args:
        x (torch.Tensor): Input tensor
        kernel_size (int): Size of pooling window
        stride (int): Stride of pooling window
        padding (int): Padding to be applied
        dilation (int): Spacing between kernel elements

    Returns:
        torch.Tensor: Output tensor after max pooling
    """
    return F.max_pool2d(
        x, kernel_size=kernel_size, stride=stride, padding=padding, dilation=dilation
    )


class Model(nn.Module):
    """
    Simple model that performs Max Pooling 2D.
    """

    def __init__(self, kernel_size: int, stride: int, padding: int, dilation: int):
        """
        Initializes the model parameters.
        """
        super(Model, self).__init__()
        self.kernel_size = kernel_size
        self.stride = stride
        self.padding = padding
        self.dilation = dilation

    def forward(self, x: torch.Tensor, fn=module_fn) -> torch.Tensor:
        """
        Forward pass that calls module_fn.
        """
        return fn(
            x,
            self.kernel_size,
            self.stride,
            self.padding,
            self.dilation,
        )


batch_size = 16
channels = 32
height = 128
width = 128
kernel_size = 2
stride = 2
padding = 1
dilation = 3


def get_inputs():
    x = torch.randn(batch_size, channels, height, width)
    return [x]


def get_init_inputs():
    return [kernel_size, stride, padding, dilation]

import torch
import torch.nn as nn

class Model(nn.Module):
    """
    Simple model that performs Max Pooling 2D.
    """
    def __init__(self, kernel_size: int, stride: int, padding: int, dilation: int):
        """
        Initializes the Max Pooling 2D layer.

        Args:
            kernel_size (int): Size of the pooling window.
            stride (int): Stride of the pooling window.
            padding (int): Padding to be applied before pooling.
            dilation (int): Spacing between kernel elements.
        """
        super(Model, self).__init__()
        self.maxpool = nn.MaxPool2d(kernel_size=kernel_size, stride=stride, padding=padding, dilation=dilation)

    def forward(self, x: torch.Tensor) -> torch.Tensor:
        """
        Applies Max Pooling 2D to the input tensor.

        Args:
            x (torch.Tensor): Input tensor of shape (batch_size, channels, height, width).

        Returns:
            torch.Tensor: Output tensor after Max Pooling 2D, shape (batch_size, channels, pooled_height, pooled_width).
        """
        return self.maxpool(x)

batch_size = 16
channels = 32
height = 128
width = 128
kernel_size = 2
stride = 2
padding = 1
dilation = 3

def get_inputs():
    x = torch.randn(batch_size, channels, height, width)
    return [x]

def get_init_inputs():
    return [kernel_size, stride, padding, dilation]
Download Evaluation Download PyTorch Download CUDA Download Profiles

Kernel Information

Operation Name 42_Max_Pooling_2D
Level ID 1
Task ID 42
Kernel Name textured_modular_unroll_base_base
CUDA Speedup (Native) 1.434x
CUDA Speedup (Compile) 3.036x
CUDA Runtime 0.023 ms
PyTorch Runtime (Native) 0.033 ms
PyTorch Runtime (Compile) 0.070 ms
Correct True
Max Diff (vs. Reference) 0.000000
Model openrouter-deepseek-reasoner
Temperature 1.00
View Experiment Progress Details

Related Kernels (Level 1, Task 42 • 42_Max_Pooling_2D)

Rank Kernel Name Runtime (ms) Speedup Native Speedup Compile
🥇 textured_modular_unroll_base_base 0.02 1.43 3.04
🥇 max_pool2d_optimized_strided_base 0.02 1.43 3.04
🥇 tuned_block_size_maxpool2d_base_base 0.02 1.43 3.04
🥇 modularized_unroll_base_base 0.02 1.43 3.04
🥇 fused_unroll_constmem_pool_base 0.02 1.43 3.04
🥇 base_unrolled_combo_base 0.02 1.43 3.04
🥇 warp_divergence_optimized_unroll_base 0.02 1.43 3.04
8 max_pool2d_strided_base_base 0.02 1.37 2.91
8 streams_unroll_pipelined_batch_base_base 0.02 1.37 2.91
8 optimized_max_pool2d_base 0.02 1.37 2.91
8 42_Max_Pooling_2D_manual_unroll_base 0.02 1.37 2.91
8 coalesced_1d_unroll_base_base 0.02 1.37 2.91
8 max_pool2d_kernel_manually_unrolled_base_base 0.02 1.37 2.91
8 coalesced_ldg_unrolled_base_base 0.02 1.37 2.91
8 unrolled_coalesced_maxpool_base 0.02 1.37 2.91
8 max_pool2d_combined_base 0.02 1.37 2.91
8 max_pool2d_combined_optimized_base 0.02 1.37 2.91
8 unroll_base_optimized_base 0.02 1.37 2.91
8 max_pool2d_combined_base 0.02 1.37 2.91
8 fully_unrolled_maxpool2d_base_base 0.02 1.37 2.91 
#include <torch/extension.h>
#include <cuda.h>
#include <cuda_runtime.h>
#include <vector>

constexpr int MAX_KERNEL_SIZE = 7;

template <int KERNEL_SIZE, typename scalar_t>
__device__ void process_kernel(
    const scalar_t* __restrict__ input_channel,
    scalar_t& max_val,
    int base_ih,
    int base_iw,
    int dilation,
    int input_height,
    int input_width
) {
    #pragma unroll
    for (int kh = 0; kh < KERNEL_SIZE; kh++) {
        const int ih = base_ih + kh * dilation;
        const bool valid_h = ih >= 0 && ih < input_height;
        
        #pragma unroll
        for (int kw = 0; kw < KERNEL_SIZE; kw++) {
            const int iw = base_iw + kw * dilation;
            if (valid_h && iw >= 0 && iw < input_width) {
                max_val = max(max_val, __ldg(&input_channel[ih * input_width + iw]));
            }
        }
    }
}

template <typename scalar_t>
__device__ void process_kernel_generic(
    const scalar_t* __restrict__ input_channel,
    scalar_t& max_val,
    int base_ih,
    int base_iw,
    int dilation,
    int input_height,
    int input_width,
    int kernel_size
) {
    for (int kh = 0; kh < kernel_size; kh++) {
        const int ih = base_ih + kh * dilation;
        const bool valid_h = ih >= 0 && ih < input_height;
        
        #pragma unroll 4
        for (int kw = 0; kw < kernel_size; kw++) {
            const int iw = base_iw + kw * dilation;
            if (valid_h && iw >= 0 && iw < input_width) {
                max_val = max(max_val, __ldg(&input_channel[ih * input_width + iw]));
            }
        }
    }
}

template <typename scalar_t>
__global__ void max_pool2d_textured_kernel(
    const scalar_t* __restrict__ input,
    scalar_t* __restrict__ output,
    const int batch_size,
    const int channels,
    const int input_height,
    const int input_width,
    const int output_height,
    const int output_width,
    const int kernel_size,
    const int stride,
    const int padding,
    const int dilation
) {
    const int ow = blockIdx.x * (blockDim.x * 2) + threadIdx.x;
    const int oh = blockIdx.y * blockDim.y + threadIdx.y;
    const int bc = blockIdx.z;
    const int b = bc / channels;
    const int c = bc % channels;

    if (oh >= output_height) return;

    scalar_t max_val[2] = {-std::numeric_limits<scalar_t>::infinity(),
                          -std::numeric_limits<scalar_t>::infinity()};
    const scalar_t* input_channel = input + (b * channels + c) * input_height * input_width;

    for (int i = 0; i < 2; ++i) {
        const int curr_ow = ow + i * blockDim.x;
        if (curr_ow >= output_width) break;

        const int base_ih = oh * stride - padding;
        const int base_iw = curr_ow * stride - padding;

        switch (kernel_size) {
            case 2: process_kernel<2>(input_channel, max_val[i], base_ih, base_iw,
                                     dilation, input_height, input_width); break;
            case 3: process_kernel<3>(input_channel, max_val[i], base_ih, base_iw,
                                     dilation, input_height, input_width); break;
            default:
                if (kernel_size <= MAX_KERNEL_SIZE) {
                    process_kernel_generic(input_channel, max_val[i], base_ih, base_iw,
                                          dilation, input_height, input_width, kernel_size);
                }
        }

        if (curr_ow < output_width) {
            output[(b * channels + c) * output_height * output_width + oh * output_width + curr_ow] = max_val[i];
        }
    }
}

torch::Tensor max_pool2d_cuda_forward(
    torch::Tensor input,
    int kernel_size,
    int stride,
    int padding,
    int dilation
) {
    const auto batch_size = input.size(0);
    const auto channels = input.size(1);
    const auto input_height = input.size(2);
    const auto input_width = input.size(3);

    const auto output_height = ((input_height + 2 * padding - dilation * (kernel_size - 1) - 1) / stride) + 1;
    const auto output_width = ((input_width + 2 * padding - dilation * (kernel_size - 1) - 1) / stride) + 1;

    auto output = torch::empty({batch_size, channels, output_height, output_width}, input.options());

    const dim3 threads(32, 4);  // 2x wider blocks for better memory coalescing
    const dim3 blocks(
        (output_width + (threads.x * 2) - 1) / (threads.x * 2),
        (output_height + threads.y - 1) / threads.y,
        batch_size * channels
    );

    AT_DISPATCH_FLOATING_TYPES(input.type(), "max_pool2d_cuda_forward", ([&] {
        max_pool2d_textured_kernel<scalar_t><<<blocks, threads>>>(
            input.data_ptr<scalar_t>(),
            output.data_ptr<scalar_t>(),
            batch_size,
            channels,
            input_height,
            input_width,
            output_height,
            output_width,
            kernel_size,
            stride,
            padding,
            dilation
        );
    }));

    return output;
}

PYBIND11_MODULE(TORCH_EXTENSION_NAME, m) {
    m.def("forward", &max_pool2d_cuda_forward, "Textured Max Pool 2D forward (CUDA)");
}
Performance Metrics
Metric Value Unit Variance Samples
Executed Ipc Active 2.042 inst/cycle 0.000 5
Executed Ipc Elapsed 1.610 inst/cycle 0.000 5
Issue Slots Busy 51.560 % 0.006 5
Issued Ipc Active 2.062 inst/cycle 0.000 5
SM Busy 52.346 % 0.006 5
Memory Throughput 2320878265662.806 byte/second 56325075479940390912.000 5
Mem Busy 42.600 % 0.014 5
Max Bandwidth 69.430 % 0.056 5
L1/TEX Hit Rate 47.188 % 0.000 5
L2 Hit Rate 25.236 % 0.010 5
Mem Pipes Busy 35.102 % 0.015 5
Warp Cycles Per Issued Instruction 25.492 cycle 0.423 5
Warp Cycles Per Executed Instruction 25.744 cycle 0.429 5
Avg. Active Threads Per Warp 32.000 0.000 5
Avg. Not Predicated Off Threads Per Warp 30.990 0.000 5
Max Active Clusters 0.000 cluster 0.000 5
Max Cluster Size 8.000 block 0.000 5
Overall GPU Occupancy 0.000 % 0.000 5
Cluster Occupancy 0.000 % 0.000 5
Block Limit SM 32.000 block 0.000 5
Block Limit Registers 16.000 block 0.000 5
Block Limit Shared Mem 32.000 block 0.000 5
Block Limit Warps 16.000 block 0.000 5
Theoretical Active Warps per SM 64.000 warp 0.000 5
Theoretical Occupancy 100.000 % 0.000 5
Achieved Occupancy 81.738 % 0.012 5
Achieved Active Warps Per SM 52.314 warp 0.005 5
Analysis Rules
Rule Description
INF HighPipeUtilization ALU is the highest-utilized pipeline (52.2%) based on active cycles, taking into account the rates of its different instructions. It executes integer and logic operations. It is well-utilized, but should not be a bottleneck.
INF CPIStall Check the Warp Stall Sampling (All Cycles) table for the top stall locations in your source based on sampling data. The Kernel Profiling Guide (https://docs.nvidia.com/nsight-compute/ProfilingGuide/index.html#metrics-reference) provides more details on each stall reason.
WRN Occupancy This kernel's theoretical occupancy is not impacted by any block limit. The difference between calculated theoretical (100.0%) and measured achieved occupancy (81.7%) can be the result of warp scheduling overheads or workload imbalances during the kernel execution. Load imbalances can occur between warps within a block as well as across blocks of the same kernel. See the CUDA Best Practices Guide (https://docs.nvidia.com/cuda/cuda-c-best-practices-guide/index.html#occupancy) for more details on optimizing occupancy.
Operation / Metric Value Unit
aten::to
CPU Time 226409.78 μs
Device Time 3307.54 μs
Self CPU Time 47.06 μs
Self Device Time 0.00 μs
CPU Memory Usage 0 B
Device Memory Usage 0 B
Self CPU Memory Usage 0 B
Self Device Memory Usage 0 B
aten::_to_copy
CPU Time 226362.72 μs
Device Time 3307.54 μs
Self CPU Time 117.66 μs
Self Device Time 0.00 μs
CPU Memory Usage 0 B
Device Memory Usage 0 B
Self CPU Memory Usage 0 B
Self Device Memory Usage 0 B
aten::empty_strided
CPU Time 222681.60 μs
Device Time 0.00 μs
Self CPU Time 224.60 μs
Self Device Time 0.00 μs
CPU Memory Usage 0 B
Device Memory Usage 0 B
Self CPU Memory Usage 0 B
Self Device Memory Usage 0 B
cudaDeviceGetStreamPriorityRange
CPU Time 222204.22 μs
Device Time 0.00 μs
Self CPU Time 222204.22 μs
Self Device Time 0.00 μs
CPU Memory Usage 0 B
Device Memory Usage 0 B
Self CPU Memory Usage 0 B
Self Device Memory Usage 0 B
cudaLaunchKernel
CPU Time 539167.76 μs
Device Time 18588.71 μs
Self CPU Time 539167.76 μs
Self Device Time 18588.71 μs
CPU Memory Usage 0 B
Device Memory Usage 0 B
Self CPU Memory Usage 0 B
Self Device Memory Usage 0 B
void max_pool2d_textured_kernel<float>(float const*, float*, int, int, int, int, int, int, int, int, int, int)
CPU Time 0.00 μs
Device Time 139985.35 μs
Self CPU Time 0.00 μs
Self Device Time 139985.35 μs
CPU Memory Usage 0 B
Device Memory Usage 0 B
Self CPU Memory Usage 0 B
Self Device Memory Usage 0 B
cudaEventRecord
CPU Time 19515.91 μs
Device Time 36786.28 μs
Self CPU Time 19515.91 μs
Self Device Time 36786.28 μs
CPU Memory Usage 0 B
Device Memory Usage 0 B
Self CPU Memory Usage 0 B
Self Device Memory Usage 0 B
aten::zero_
CPU Time 113721.03 μs
Device Time 552763.23 μs
Self CPU Time 12877.09 μs
Self Device Time 0.00 μs
CPU Memory Usage 0 B
Device Memory Usage 0 B
Self CPU Memory Usage 0 B
Self Device Memory Usage 0 B
aten::fill_
CPU Time 100845.61 μs
Device Time 552763.23 μs
Self CPU Time 15606.94 μs
Self Device Time 552763.23 μs
CPU Memory Usage 0 B
Device Memory Usage 0 B
Self CPU Memory Usage 0 B
Self Device Memory Usage 0 B
void at::native::vectorized_elementwise_kernel<4, at::native::FillFunctor<int>, at::detail::Array<char*, 1> >(int, at::native::FillFunctor<int>, at::detail::Array<char*, 1>)
CPU Time 0.00 μs
Device Time 552763.23 μs
Self CPU Time 0.00 μs
Self Device Time 552763.23 μs
CPU Memory Usage 0 B
Device Memory Usage 0 B
Self CPU Memory Usage 0 B
Self Device Memory Usage 0 B
Status: Completed 
45299 warnings generated when compiling for host.
Suppressed 45325 warnings (45278 in non-user code, 47 NOLINT).
Use -header-filter=.* to display errors from all non-system headers. Use -system-headers to display errors from system headers as well.
/home/robert_sakana_ai/llm_cuda/experiments/20250214_optimize_b10_s4_e0_prof_ncu/level_1/task_42/b10_s3_textured_modular_unroll_base/base/base.cu:14:5 bugprone-easily-swappable-parameters
14 | int dilation,
| ^~~~~~~~~~~~~
15 | int input_height,
| ~~~~~~~~~~~~~~~~~
16 | int input_width
| ~~~~~~~~~~~~~~~
/home/robert_sakana_ai/llm_cuda/experiments/20250214_optimize_b10_s4_e0_prof_ncu/level_1/task_42/b10_s3_textured_modular_unroll_base/base/base.cu:14:9: note: the first parameter in the range is 'dilation'
14 | int dilation,
| ^~~~~~~~
/home/robert_sakana_ai/llm_cuda/experiments/20250214_optimize_b10_s4_e0_prof_ncu/level_1/task_42/b10_s3_textured_modular_unroll_base/base/base.cu:16:9: note: the last parameter in the range is 'input_width'
16 | int input_width
| ^~~~~~~~~~~
/home/robert_sakana_ai/llm_cuda/experiments/20250214_optimize_b10_s4_e0_prof_ncu/level_1/task_42/b10_s3_textured_modular_unroll_base/base/base.cu:39:5: warning: 4 adjacent parameters of 'process_kernel_generic' of similar type ('int') are easily swapped by mistake [bugprone-easily-swappable-parameters]
39 | int dilation,
| ^~~~~~~~~~~~~
40 | int input_height,
| ~~~~~~~~~~~~~~~~~
41 | int input_width,
| ~~~~~~~~~~~~~~~~
42 | int kernel_size
| ~~~~~~~~~~~~~~~
/home/robert_sakana_ai/llm_cuda/experiments/20250214_optimize_b10_s4_e0_prof_ncu/level_1/task_42/b10_s3_textured_modular_unroll_base/base/base.cu:39:9: note: the first parameter in the range is 'dilation'
39 | int dilation,
| ^~~~~~~~
/home/robert_sakana_ai/llm_cuda/experiments/20250214_optimize_b10_s4_e0_prof_ncu/level_1/task_42/b10_s3_textured_modular_unroll_base/base/base.cu:42:9: note: the last parameter in the range is 'kernel_size'
42 | int kernel_size
| ^~~~~~~~~~~
/home/robert_sakana_ai/llm_cuda/experiments/20250214_optimize_b10_s4_e0_prof_ncu/level_1/task_42/b10_s3_textured_modular_unroll_base/base/base.cu:62:5: warning: 2 adjacent parameters of 'max_pool2d_textured_kernel' of similar type ('const int') are easily swapped by mistake [bugprone-easily-swappable-parameters]
62 | const int batch_size,
| ^~~~~~~~~~~~~~~~~~~~~
63 | const int channels,
| ~~~~~~~~~~~~~~~~~~
/home/robert_sakana_ai/llm_cuda/experiments/20250214_optimize_b10_s4_e0_prof_ncu/level_1/task_42/b10_s3_textured_modular_unroll_base/base/base.cu:62:15: note: the first parameter in the range is 'batch_size'
62 | const int batch_size,
| ^~~~~~~~~~
/home/robert_sakana_ai/llm_cuda/experiments/20250214_optimize_b10_s4_e0_prof_ncu/level_1/task_42/b10_s3_textured_modular_unroll_base/base/base.cu:63:15: note: the last parameter in the range is 'channels'
63 | const int channels,
| ^~~~~~~~
/home/robert_sakana_ai/llm_cuda/experiments/20250214_optimize_b10_s4_e0_prof_ncu/level_1/task_42/b10_s3_textured_modular_unroll_base/base/base.cu:65:5: warning: 2 adjacent parameters of 'max_pool2d_textured_kernel' of similar type ('const int') are easily swapped by mistake [bugprone-easily-swappable-parameters]
65 | const int input_width,
| ^~~~~~~~~~~~~~~~~~~~~~
66 | const int output_height,
| ~~~~~~~~~~~~~~~~~~~~~~~
/home/robert_sakana_ai/llm_cuda/experiments/20250214_optimize_b10_s4_e0_prof_ncu/level_1/task_42/b10_s3_textured_modular_unroll_base/base/base.cu:65:15: note: the first parameter in the range is 'input_width'
65 | const int input_width,
| ^~~~~~~~~~~
/home/robert_sakana_ai/llm_cuda/experiments/20250214_optimize_b10_s4_e0_prof_ncu/level_1/task_42/b10_s3_textured_modular_unroll_base/base/base.cu:66:15: note: the last parameter in the range is 'output_height'
66 | const int output_height,
| ^~~~~~~~~~~~~
/home/robert_sakana_ai/llm_cuda/experiments/20250214_optimize_b10_s4_e0_prof_ncu/level_1/task_42/b10_s3_textured_modular_unroll_base/base/base.cu:67:5: warning: 3 adjacent parameters of 'max_pool2d_textured_kernel' of similar type ('const int') are easily swapped by mistake [bugprone-easily-swappable-parameters]
67 | const int output_width,
| ^~~~~~~~~~~~~~~~~~~~~~~
68 | const int kernel_size,
| ~~~~~~~~~~~~~~~~~~~~~~
69 | const int stride,
| ~~~~~~~~~~~~~~~~
/home/robert_sakana_ai/llm_cuda/experiments/20250214_optimize_b10_s4_e0_prof_ncu/level_1/task_42/b10_s3_textured_modular_unroll_base/base/base.cu:67:15: note: the first parameter in the range is 'output_width'
67 | const int output_width,
| ^~~~~~~~~~~~
/home/robert_sakana_ai/llm_cuda/experiments/20250214_optimize_b10_s4_e0_prof_ncu/level_1/task_42/b10_s3_textured_modular_unroll_base/base/base.cu:69:15: note: the last parameter in the range is 'stride'
69 | const int stride,
| ^~~~~~
/home/robert_sakana_ai/llm_cuda/experiments/20250214_optimize_b10_s4_e0_prof_ncu/level_1/task_42/b10_s3_textured_modular_unroll_base/base/base.cu:70:5: warning: 2 adjacent parameters of 'max_pool2d_textured_kernel' of similar type ('const int') are easily swapped by mistake [bugprone-easily-swappable-parameters]
70 | const int padding,
| ^~~~~~~~~~~~~~~~~~
71 | const int dilation
| ~~~~~~~~~~~~~~~~~~
/home/robert_sakana_ai/llm_cuda/experiments/20250214_optimize_b10_s4_e0_prof_ncu/level_1/task_42/b10_s3_textured_modular_unroll_base/base/base.cu:70:15: note: the first parameter in the range is 'padding'
70 | const int padding,
| ^~~~~~~
/home/robert_sakana_ai/llm_cuda/experiments/20250214_optimize_b10_s4_e0_prof_ncu/level_1/task_42/b10_s3_textured_modular_unroll_base/base/base.cu:71:15: note: the last parameter in the range is 'dilation'
71 | const int dilation
| ^~~~~~~~
/home/robert_sakana_ai/llm_cuda/experiments/20250214_optimize_b10_s4_e0_prof_ncu/level_1/task_42/b10_s3_textured_modular_unroll_base/base/base.cu:73:20: warning: narrowing conversion from 'unsigned int' to signed type 'int' is implementation-defined [bugprone-narrowing-conversions]
73 | const int ow = blockIdx.x * (blockDim.x * 2) + threadIdx.x;
| ^
/home/robert_sakana_ai/llm_cuda/experiments/20250214_optimize_b10_s4_e0_prof_ncu/level_1/task_42/b10_s3_textured_modular_unroll_base/base/base.cu:74:20: warning: narrowing conversion from 'unsigned int' to signed type 'int' is implementation-defined [bugprone-narrowing-conversions]
74 | const int oh = blockIdx.y * blockDim.y + threadIdx.y;
| ^
/home/robert_sakana_ai/llm_cuda/experiments/20250214_optimize_b10_s4_e0_prof_ncu/level_1/task_42/b10_s3_textured_modular_unroll_base/base/base.cu:75:20: warning: narrowing conversion from 'unsigned int' to signed type 'int' is implementation-defined [bugprone-narrowing-conversions]
75 | const int bc = blockIdx.z;
| ^
/home/robert_sakana_ai/llm_cuda/experiments/20250214_optimize_b10_s4_e0_prof_ncu/level_1/task_42/b10_s3_textured_modular_unroll_base/base/base.cu:86:29: warning: narrowing conversion from 'unsigned int' to signed type 'int' is implementation-defined [bugprone-narrowing-conversions]
86 | const int curr_ow = ow + i * blockDim.x;
| ^
/home/robert_sakana_ai/llm_cuda/experiments/20250214_optimize_b10_s4_e0_prof_ncu/level_1/task_42/b10_s3_textured_modular_unroll_base/base/base.cu:122:49: warning: performing an implicit widening conversion to type 'int64_t' (aka 'long') of a multiplication performed in type 'int' [bugprone-implicit-widening-of-multiplication-result]
122 | const auto output_height = ((input_height + 2 * padding - dilation * (kernel_size - 1) - 1) / stride) + 1;
| ^
/home/robert_sakana_ai/llm_cuda/experiments/20250214_optimize_b10_s4_e0_prof_ncu/level_1/task_42/b10_s3_textured_modular_unroll_base/base/base.cu:122:49: note: make conversion explicit to silence this warning
122 | const auto output_height = ((input_height + 2 * padding - dilation * (kernel_size - 1) - 1) / stride) + 1;
| ^~~~~~~~~~~
| static_cast<int64_t>( )
/home/robert_sakana_ai/llm_cuda/experiments/20250214_optimize_b10_s4_e0_prof_ncu/level_1/task_42/b10_s3_textured_modular_unroll_base/base/base.cu:122:49: note: perform multiplication in a wider type
122 | const auto output_height = ((input_height + 2 * padding - dilation * (kernel_size - 1) - 1) / stride) + 1;
| ^
| static_cast<int64_t>( )
/home/robert_sakana_ai/llm_cuda/experiments/20250214_optimize_b10_s4_e0_prof_ncu/level_1/task_42/b10_s3_textured_modular_unroll_base/base/base.cu:122:63: warning: performing an implicit widening conversion to type 'int64_t' (aka 'long') of a multiplication performed in type 'int' [bugprone-implicit-widening-of-multiplication-result]
122 | const auto output_height = ((input_height + 2 * padding - dilation * (kernel_size - 1) - 1) / stride) + 1;
| ^
/home/robert_sakana_ai/llm_cuda/experiments/20250214_optimize_b10_s4_e0_prof_ncu/level_1/task_42/b10_s3_textured_modular_unroll_base/base/base.cu:122:63: note: make conversion explicit to silence this warning
4 | const auto output_height = ((input_height + 2 * padding - dilation * (kernel_size - 1) - 1) / stride) + 1;
| ^~~~~~~~~~~~~~~~~~~~~~~~~~~~
| static_cast<int64_t>( )
/home/robert_sakana_ai/llm_cuda/experiments/20250214_optimize_b10_s4_e0_prof_ncu/level_1/task_42/b10_s3_textured_modular_unroll_base/base/base.cu:122:63: note: perform multiplication in a wider type
122 | const auto output_height = ((input_height + 2 * padding - dilation * (kernel_size - 1) - 1) / stride) + 1;
| ^~~~~~~~
| static_cast<int64_t>( )
/home/robert_sakana_ai/llm_cuda/experiments/20250214_optimize_b10_s4_e0_prof_ncu/level_1/task_42/b10_s3_textured_modular_unroll_base/base/base.cu:123:47: warning: performing an implicit widening conversion to type 'int64_t' (aka 'long') of a multiplication performed in type 'int' [bugprone-implicit-widening-of-multiplication-result]
123 | const auto output_width = ((input_width + 2 * padding - dilation * (kernel_size - 1) - 1) / stride) + 1;
| ^
/home/robert_sakana_ai/llm_cuda/experiments/20250214_optimize_b10_s4_e0_prof_ncu/level_1/task_42/b10_s3_textured_modular_unroll_base/base/base.cu:123:47: note: make conversion explicit to silence this warning
123 | const auto output_width = ((input_width + 2 * padding - dilation * (kernel_size - 1) - 1) / stride) + 1;
| ^~~~~~~~~~~
| static_cast<int64_t>( )
/home/robert_sakana_ai/llm_cuda/experiments/20250214_optimize_b10_s4_e0_prof_ncu/level_1/task_42/b10_s3_textured_modular_unroll_base/base/base.cu:123:47: note: perform multiplication in a wider type
123 | const auto output_width = ((input_width + 2 * padding - dilation * (kernel_size - 1) - 1) / stride) + 1;
| ^
| static_cast<int64_t>( )
/home/robert_sakana_ai/llm_cuda/experiments/20250214_optimize_b10_s4_e0_prof_ncu/level_1/task_42/b10_s3_textured_modular_unroll_base/base/base.cu:123:61: warning: performing an implicit widening conversion to type 'int64_t' (aka 'long') of a multiplication performed in type 'int' [bugprone-implicit-widening-of-multiplication-result]
123 | const auto output_width = ((input_width + 2 * padding - dilation * (kernel_size - 1) - 1) / stride) + 1;
| ^
/home/robert_sakana_ai/llm_cuda/experiments/20250214_optimize_b10_s4_e0_prof_ncu/level_1/task_42/b10_s3_textured_modular_unroll_base/base/base.cu:123:61: note: make conversion explicit to silence this warning
123 | const auto output_width = ((input_width + 2 * padding - dilation * (kernel_size - 1) - 1) / stride) + 1;
| ^~~~~~~~~~~~~~~~~~~~~~~~~~~~
| static_cast<int64_t>( )
/home/robert_sakana_ai/llm_cuda/experiments/20250214_optimize_b10_s4_e0_prof_ncu/level_1/task_42/b10_s3_textured_modular_unroll_base/base/base.cu:123:61: note: perform multiplication in a wider type
123 | const auto output_width = ((input_width + 2 * padding - dilation * (kernel_size - 1) - 1) / stride) + 1;
| ^~~~~~~~
| static_cast<int64_t>( )
/home/robert_sakana_ai/llm_cuda/experiments/20250214_optimize_b10_s4_e0_prof_ncu/level_1/task_42/b10_s3_textured_modular_unroll_base/base/base.cu:129:26: warning: performing an implicit widening conversion to type 'int64_t' (aka 'long') of a multiplication performed in type 'unsigned int' [bugprone-implicit-widening-of-multiplication-result]
129 | (output_width + (threads.x * 2) - 1) / (threads.x * 2),
| ^
/home/robert_sakana_ai/llm_cuda/experiments/20250214_optimize_b10_s4_e0_prof_ncu/level_1/task_42/b10_s3_textured_modular_unroll_base/base/base.cu:129:26: note: make conversion explicit to silence this warning
129 | (output_width + (threads.x * 2) - 1) / (threads.x * 2),
| ^~~~~~~~~~~~~
| static_cast<int64_t>( )
/home/robert_sakana_ai/llm_cuda/experiments/20250214_optimize_b10_s4_e0_prof_ncu/level_1/task_42/b10_s3_textured_modular_unroll_base/base/base.cu:129:26: note: perform multiplication in a wider type
129 | (output_width + (threads.x * 2) - 1) / (threads.x * 2),
| ^~~~~~~~~
| static_cast<unsigned long>( )
/home/robert_sakana_ai/llm_cuda/experiments/20250214_optimize_b10_s4_e0_prof_ncu/level_1/task_42/b10_s3_textured_modular_unroll_base/base/base.cu:129:49: warning: performing an implicit widening conversion to type 'int64_t' (aka 'long') of a multiplication performed in type 'unsigned int' [bugprone-implicit-widening-of-multiplication-result]
129 | (output_width + (threads.x * 2) - 1) / (threads.x * 2),
| ^
/home/robert_sakana_ai/llm_cuda/experiments/20250214_optimize_b10_s4_e0_prof_ncu/level_1/task_42/b10_s3_textured_modular_unroll_base/base/base.cu:129:49: note: make conversion explicit to silence this warning
129 | (output_width + (threads.x * 2) - 1) / (threads.x * 2),
| ^~~~~~~~~~~~~
| static_cast<int64_t>( )
/home/robert_sakana_ai/llm_cuda/experiments/20250214_optimize_b10_s4_e0_prof_ncu/level_1/task_42/b10_s3_textured_modular_unroll_base/base/base.cu:129:49: note: perform multiplication in a wider type
129 | (output_width + (threads.x * 2) - 1) / (threads.x * 2),
| ^~~~~~~~~
| static_cast<unsigned long>( )
/home/robert_sakana_ai/llm_cuda/experiments/20250214_optimize_b10_s4_e0_prof_ncu/level_1/task_42/b10_s3_textured_modular_unroll_base/base/base.cu:134:5: warning: inside a lambda, '__func__' expands to the name of the function call operator; consider capturing the name of the enclosing function explicitly [bugprone-lambda-function-name]
134 | AT_DISPATCH_FLOATING_TYPES(input.type(), "max_pool2d_cuda_forward", ([&] {
| ^
/home/robert_sakana_ai/miniconda3/envs/llm2cuda/lib/python3.11/site-packages/torch/include/ATen/Dispatch.h:237:34: note: expanded from macro 'AT_DISPATCH_FLOATING_TYPES'
237 | AT_DISPATCH_SWITCH(TYPE, NAME, AT_DISPATCH_CASE_FLOATING_TYPES(__VA_ARGS__))
| ^
/home/robert_sakana_ai/miniconda3/envs/llm2cuda/lib/python3.11/site-packages/torch/include/ATen/Dispatch.h:233:3: note: expanded from macro 'AT_DISPATCH_CASE_FLOATING_TYPES'
233 | AT_DISPATCH_CASE(at::ScalarType::Double, __VA_ARGS__) \
| ^
/home/robert_sakana_ai/miniconda3/envs/llm2cuda/lib/python3.11/site-packages/torch/include/ATen/Dispatch.h:74:3: note: expanded from macro 'AT_DISPATCH_CASE'
74 | AT_PRIVATE_CASE_TYPE_USING_HINT(enum_type, scalar_t, __VA_ARGS__)
| ^
note: (skipping 1 expansions in backtrace; use -fmacro-backtrace-limit=0 to see all)
/home/robert_sakana_ai/miniconda3/envs/llm2cuda/lib/python3.11/site-packages/torch/include/ATen/Dispatch.h:58:7: note: expanded from macro 'AT_PRIVATE_CHECK_SELECTIVE_BUILD'
58 | AT_ERROR( \
| ^
/home/robert_sakana_ai/miniconda3/envs/llm2cuda/lib/python3.11/site-packages/torch/include/c10/util/Exception.h:711:32: note: expanded from macro 'AT_ERROR'
711 | C10_EXPAND_MSVC_WORKAROUND(TORCH_CHECK(false, ::c10::str(__VA_ARGS__))); \
| ^
/home/robert_sakana_ai/miniconda3/envs/llm2cuda/lib/python3.11/site-packages/torch/include/c10/util/Exception.h:536:9: note: expanded from macro 'TORCH_CHECK'
536 | __func__, \
| ^
/home/robert_sakana_ai/llm_cuda/experiments/20250214_optimize_b10_s4_e0_prof_ncu/level_1/task_42/b10_s3_textured_modular_unroll_base/base/base.cu:134:5: warning: 'scalar_type' is deprecated: passing at::DeprecatedTypeProperties to an AT_DISPATCH macro is deprecated, pass an at::ScalarType instead [clang-diagnostic-deprecated-declarations]
134 | AT_DISPATCH_FLOATING_TYPES(input.type(), "max_pool2d_cuda_forward", ([&] {
| ^
/home/robert_sakana_ai/miniconda3/envs/llm2cuda/lib/python3.11/site-packages/torch/include/ATen/Dispatch.h:237:3: note: expanded from macro 'AT_DISPATCH_FLOATING_TYPES'
237 | AT_DISPATCH_SWITCH(TYPE, NAME, AT_DISPATCH_CASE_FLOATING_TYPES(__VA_ARGS__))
| ^
/home/robert_sakana_ai/miniconda3/envs/llm2cuda/lib/python3.11/site-packages/torch/include/ATen/Dispatch.h:218:36: note: expanded from macro 'AT_DISPATCH_SWITCH'
218 | at::ScalarType _st = ::detail::scalar_type(the_type); \
| ^
/home/robert_sakana_ai/miniconda3/envs/llm2cuda/lib/python3.11/site-packages/torch/include/ATen/Dispatch.h:106:1: note: 'scalar_type' has been explicitly marked deprecated here
106 | C10_DEPRECATED_MESSAGE(
| ^
/home/robert_sakana_ai/miniconda3/envs/llm2cuda/lib/python3.11/site-packages/torch/include/c10/util/Deprecated.h:24:43: note: expanded from macro 'C10_DEPRECATED_MESSAGE'
24 | #define C10_DEPRECATED_MESSAGE(message) [[deprecated(message)]]
| ^
/home/robert_sakana_ai/llm_cuda/experiments/20250214_optimize_b10_s4_e0_prof_ncu/level_1/task_42/b10_s3_textured_modular_unroll_base/base/base.cu:134:38: warning: 'type' is deprecated: Tensor.type() is deprecated. Instead use Tensor.options(), which in many cases (e.g. in a constructor) is a drop-in replacement. If you were using data from type(), that is now available from Tensor itself, so instead of tensor.type().scalar_type(), use tensor.scalar_type() instead and instead of tensor.type().backend() use tensor.device(). [clang-diagnostic-deprecated-declarations]
134 | AT_DISPATCH_FLOATING_TYPES(input.type(), "max_pool2d_cuda_forward", ([&] {
| ^
/home/robert_sakana_ai/miniconda3/envs/llm2cuda/lib/python3.11/site-packages/torch/include/ATen/core/TensorBody.h:224:3: note: 'type' has been explicitly marked deprecated here
224 | C10_DEPRECATED_MESSAGE("Tensor.type() is deprecated. Instead use Tensor.options(), which in many cases (e.g. in a constructor) is a drop-in replacement. If you were using data from type(), that is now available from Tensor itself, so instead of tensor.type().scalar_type(), use tensor.scalar_type() instead and instead of tensor.type().backend() use tensor.device().")
| ^
/home/robert_sakana_ai/miniconda3/envs/llm2cuda/lib/python3.11/site-packages/torch/include/c10/util/Deprecated.h:24:43: note: expanded from macro 'C10_DEPRECATED_MESSAGE'
24 | #define C10_DEPRECATED_MESSAGE(message) [[deprecated(message)]]
| ^