Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Jump to bottom

Enhanced Loop Unrolling #766

Merged
merged 5 commits into from
Mar 29, 2022
Merged

Enhanced Loop Unrolling #766

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
5 commits
Select commit Hold shift + click to select a range
be91950
Polished Utilities.Swap implementation to use tuple-based operations.
m4rs-mt Mar 28, 2022
f04ab51
Improved LoopInfo to determine the trip count of more complex loops.
m4rs-mt Mar 28, 2022
53f821f
Improved LoopUnrolling to unroll more complex loops.
m4rs-mt Mar 28, 2022
4e822d1
Used expensive loop optimizations in favor of LICM in the scope of th…
m4rs-mt Mar 28, 2022
db61215
Extended BasicLoop tests to cover loop-unrolling edge cases.
m4rs-mt Mar 28, 2022
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
128 changes: 127 additions & 1 deletion Src/ILGPU.Tests/BasicLoops.cs
View file
Open in desktop
Original file line number Diff line number Diff line change
@@ -1,6 +1,6 @@
// ---------------------------------------------------------------------------------------
// ILGPU
// Copyright (c) 2021 ILGPU Project
// Copyright (c) 2021-2022 ILGPU Project
// www.ilgpu.net
//
// File: BasicLoops.cs
Expand Down Expand Up @@ -919,6 +919,132 @@ public void DoLoopWithoutEntryBlock(int length)
var expected = Enumerable.Repeat(0, length).ToArray();
Verify(target.View, expected);
}

static void UnrollDivKernel(
Index1D index,
ArrayView<int> data1,
ArrayView<int> data2,
ArrayView<int> data3,
ArrayView<int> data4,
ArrayView<int> data5,
int value)
{
for (int laneOffset = Warp.WarpSize / 2; laneOffset > 0; laneOffset >>= 1)
value += Warp.ShuffleDown(value, laneOffset);
data1[index] = value;

// Simple skip loop
for (int i = 0; i > 0; i >>= 1)
value += 1;
data2[index] = value;

// Negative start value
for (int i = -1; i >= 0; i >>= 2)
value += i;
data3[index] = value;

// Trip count not a multiple of the divisor
for (int i = 17; i > 0; i /= 3)
value += i;
data4[index] = value;

// Loop down to 1
for (int i = 21; i >= 1; i /= 3)
value += 13;
data5[index] = value;
}

[Fact]
[KernelMethod(nameof(UnrollDivKernel))]
public void UnrollDiv()
{
int length = Accelerator.WarpSize * 2;
using var data1 = Accelerator.Allocate1D<int>(length);
using var data2 = Accelerator.Allocate1D<int>(length);
using var data3 = Accelerator.Allocate1D<int>(length);
using var data4 = Accelerator.Allocate1D<int>(length);
using var data5 = Accelerator.Allocate1D<int>(length);
Accelerator.Synchronize();

Execute(
length,
data1.View.AsContiguous(),
data2.View.AsContiguous(),
data3.View.AsContiguous(),
data4.View.AsContiguous(),
data5.View.AsContiguous(),
1);

var expected = Enumerable.Repeat(length / 2, length).ToArray();
Verify(data1.View, expected);
Verify(data2.View, expected);
Verify(data3.View, expected);

var expected4 = Enumerable.Repeat(expected[0] + 23, length).ToArray();
Verify(data4.View, expected4);

var expected5 = Enumerable.Repeat(expected4[0] + 39, length).ToArray();
Verify(data5.View, expected5);
}

static void UnrollMulKernel(
Index1D index,
ArrayView<int> data1,
ArrayView<int> data2,
ArrayView<int> data3,
ArrayView<int> data4,
int value)
{
for (int laneOffset = 1; laneOffset < Warp.WarpSize; laneOffset <<= 1)
value += Warp.ShuffleUp(value, laneOffset);
data1[index] = value;

// Trip count not a multiple of the multiplier
for (int i = 1; i < 13; i *= 3)
value += 1;
data2[index] = value;

// Negative start value
for (int i = -1; i > -17; i *= 4)
value += 2;
data3[index] = value;

for (int i = 17; i < 51; i *= 2)
value += i;
data4[index] = value;
}

[Fact]
[KernelMethod(nameof(UnrollMulKernel))]
public void UnrollMul()
{
int length = Accelerator.WarpSize * 2;
using var data1 = Accelerator.Allocate1D<int>(length);
using var data2 = Accelerator.Allocate1D<int>(length);
using var data3 = Accelerator.Allocate1D<int>(length);
using var data4 = Accelerator.Allocate1D<int>(length);
Accelerator.Synchronize();

Execute(
length,
data1.View.AsContiguous(),
data2.View.AsContiguous(),
data3.View.AsContiguous(),
data4.View.AsContiguous(),
1);

var expected = Enumerable.Repeat(length / 2, length).ToArray();
Verify(data1.View, expected);

var expected2 = Enumerable.Repeat(expected[0] + 3, length).ToArray();
Verify(data2.View, expected2);

var expected3 = Enumerable.Repeat(expected2[0] + 6, length).ToArray();
Verify(data3.View, expected3);

var expected4 = Enumerable.Repeat(expected3[0] + 17 * 3, length).ToArray();
Verify(data4.View, expected4);
}
}
}

Expand Down
Loading