maxjhandsome
/
llvm-project
forked from OSchip/llvm-project
1
0
Fork 0
Code Issues Pull Requests Packages Releases Wiki Activity
llvm-project/llvm/test/Transforms/LoopFusion/four_loops.ll

Ignoring revisions in .git-blame-ignore-revs. Click here to bypass and see the normal blame view.

117 lines
4.5 KiB
LLVM
Raw Normal View History

Add basic loop fusion pass. This patch adds a basic loop fusion pass. It will fuse loops that conform to the following 4 conditions: 1. Adjacent (no code between them) 2. Control flow equivalent (if one loop executes, the other loop executes) 3. Identical bounds (both loops iterate the same number of iterations) 4. No negative distance dependencies between the loop bodies. The pass does not make any changes to the IR to create opportunities for fusion. Instead, it checks if the necessary conditions are met and if so it fuses two loops together. The pass has not been added to the pass pipeline yet, and thus is not enabled by default. It can be run stand alone using the -loop-fusion option. Differential Revision: https://reviews.llvm.org/D55851 llvm-svn: 358607
2019-04-18 02:53:27 +08:00
; RUN: opt -S -loop-fusion < %s | FileCheck %s
@A = common global [1024 x i32] zeroinitializer, align 16
@B = common global [1024 x i32] zeroinitializer, align 16
@C = common global [1024 x i32] zeroinitializer, align 16
@D = common global [1024 x i32] zeroinitializer, align 16
; CHECK: void @dep_free
; CHECK-NEXT: bb:
; CHECK-NEXT: br label %[[LOOP1HEADER:bb[0-9]+]]
; CHECK: [[LOOP1HEADER]]
[LoopFusion] Move instructions from FC0.Latch to FC1.Latch. Summary:This PR move instructions from FC0.Latch bottom up to the beginning of FC1.Latch as long as they are proven safe. To illustrate why this is beneficial, let's consider the following example: Before Fusion: header1: br header2 header2: br header2, latch1 latch1: br header1, preheader3 preheader3: br header3 header3: br header4 header4: br header4, latch3 latch3: br header3, exit3 After Fusion (before this PR): header1: br header2 header2: br header2, latch1 latch1: br header3 header3: br header4 header4: br header4, latch3 latch3: br header1, exit3 Note that preheader3 is removed during fusion before this PR. Notice that we cannot fuse loop2 with loop4 as there exists block latch1 in between. This PR move instructions from latch1 to beginning of latch3, and remove block latch1. LoopFusion is now able to fuse loop nest recursively. After Fusion (after this PR): header1: br header2 header2: br header3 header3: br header4 header4: br header2, latch3 latch3: br header1, exit3 Reviewer: kbarton, jdoerfert, Meinersbur, dmgreen, fhahn, hfinkel, bmahjour, etiotto Reviewed By: kbarton, Meinersbur Subscribers: hiraditya, llvm-commits Tag: LLVM Differential Revision: https://reviews.llvm.org/D71165
2019-12-18 05:44:06 +08:00
; CHECK: br label %[[LOOP2BODY:bb[0-9]+]]
[LoopFusion] Restrict loop fusion to rotated loops. Summary: This patch restricts loop fusion to only consider rotated loops as valid candidates. This simplifies the analysis and transformation and aligns with other loop optimizations. Reviewers: jdoerfert, Meinersbur, dmgreen, etiotto, Whitney, fhahn, hfinkel Reviewed By: Meinersbur Subscribers: ormris, hiraditya, llvm-commits Tags: #llvm Differential Revision: https://reviews.llvm.org/D71025
2019-12-17 00:57:52 +08:00
; CHECK: [[LOOP2BODY]]
[LoopFusion] Move instructions from FC0.Latch to FC1.Latch. Summary:This PR move instructions from FC0.Latch bottom up to the beginning of FC1.Latch as long as they are proven safe. To illustrate why this is beneficial, let's consider the following example: Before Fusion: header1: br header2 header2: br header2, latch1 latch1: br header1, preheader3 preheader3: br header3 header3: br header4 header4: br header4, latch3 latch3: br header3, exit3 After Fusion (before this PR): header1: br header2 header2: br header2, latch1 latch1: br header3 header3: br header4 header4: br header4, latch3 latch3: br header1, exit3 Note that preheader3 is removed during fusion before this PR. Notice that we cannot fuse loop2 with loop4 as there exists block latch1 in between. This PR move instructions from latch1 to beginning of latch3, and remove block latch1. LoopFusion is now able to fuse loop nest recursively. After Fusion (after this PR): header1: br header2 header2: br header3 header3: br header4 header4: br header2, latch3 latch3: br header1, exit3 Reviewer: kbarton, jdoerfert, Meinersbur, dmgreen, fhahn, hfinkel, bmahjour, etiotto Reviewed By: kbarton, Meinersbur Subscribers: hiraditya, llvm-commits Tag: LLVM Differential Revision: https://reviews.llvm.org/D71165
2019-12-18 05:44:06 +08:00
; CHECK: br label %[[LOOP3BODY:bb[0-9]+]]
[LoopFusion] Restrict loop fusion to rotated loops. Summary: This patch restricts loop fusion to only consider rotated loops as valid candidates. This simplifies the analysis and transformation and aligns with other loop optimizations. Reviewers: jdoerfert, Meinersbur, dmgreen, etiotto, Whitney, fhahn, hfinkel Reviewed By: Meinersbur Subscribers: ormris, hiraditya, llvm-commits Tags: #llvm Differential Revision: https://reviews.llvm.org/D71025
2019-12-17 00:57:52 +08:00
; CHECK: [[LOOP3BODY]]
[LoopFusion] Move instructions from FC0.Latch to FC1.Latch. Summary:This PR move instructions from FC0.Latch bottom up to the beginning of FC1.Latch as long as they are proven safe. To illustrate why this is beneficial, let's consider the following example: Before Fusion: header1: br header2 header2: br header2, latch1 latch1: br header1, preheader3 preheader3: br header3 header3: br header4 header4: br header4, latch3 latch3: br header3, exit3 After Fusion (before this PR): header1: br header2 header2: br header2, latch1 latch1: br header3 header3: br header4 header4: br header4, latch3 latch3: br header1, exit3 Note that preheader3 is removed during fusion before this PR. Notice that we cannot fuse loop2 with loop4 as there exists block latch1 in between. This PR move instructions from latch1 to beginning of latch3, and remove block latch1. LoopFusion is now able to fuse loop nest recursively. After Fusion (after this PR): header1: br header2 header2: br header3 header3: br header4 header4: br header2, latch3 latch3: br header1, exit3 Reviewer: kbarton, jdoerfert, Meinersbur, dmgreen, fhahn, hfinkel, bmahjour, etiotto Reviewed By: kbarton, Meinersbur Subscribers: hiraditya, llvm-commits Tag: LLVM Differential Revision: https://reviews.llvm.org/D71165
2019-12-18 05:44:06 +08:00
; CHECK: br label %[[LOOP4BODY:bb[0-9]+]]
[LoopFusion] Restrict loop fusion to rotated loops. Summary: This patch restricts loop fusion to only consider rotated loops as valid candidates. This simplifies the analysis and transformation and aligns with other loop optimizations. Reviewers: jdoerfert, Meinersbur, dmgreen, etiotto, Whitney, fhahn, hfinkel Reviewed By: Meinersbur Subscribers: ormris, hiraditya, llvm-commits Tags: #llvm Differential Revision: https://reviews.llvm.org/D71025
2019-12-17 00:57:52 +08:00
; CHECK: [[LOOP4BODY]]
[LoopFusion] Move instructions from FC0.Latch to FC1.Latch. Summary:This PR move instructions from FC0.Latch bottom up to the beginning of FC1.Latch as long as they are proven safe. To illustrate why this is beneficial, let's consider the following example: Before Fusion: header1: br header2 header2: br header2, latch1 latch1: br header1, preheader3 preheader3: br header3 header3: br header4 header4: br header4, latch3 latch3: br header3, exit3 After Fusion (before this PR): header1: br header2 header2: br header2, latch1 latch1: br header3 header3: br header4 header4: br header4, latch3 latch3: br header1, exit3 Note that preheader3 is removed during fusion before this PR. Notice that we cannot fuse loop2 with loop4 as there exists block latch1 in between. This PR move instructions from latch1 to beginning of latch3, and remove block latch1. LoopFusion is now able to fuse loop nest recursively. After Fusion (after this PR): header1: br header2 header2: br header3 header3: br header4 header4: br header2, latch3 latch3: br header1, exit3 Reviewer: kbarton, jdoerfert, Meinersbur, dmgreen, fhahn, hfinkel, bmahjour, etiotto Reviewed By: kbarton, Meinersbur Subscribers: hiraditya, llvm-commits Tag: LLVM Differential Revision: https://reviews.llvm.org/D71165
2019-12-18 05:44:06 +08:00
; CHECK: br label %[[LOOP1LATCH:bb[0-9]+]]
; CHECK: [[LOOP1LATCH]]
[LoopFusion] Restrict loop fusion to rotated loops. Summary: This patch restricts loop fusion to only consider rotated loops as valid candidates. This simplifies the analysis and transformation and aligns with other loop optimizations. Reviewers: jdoerfert, Meinersbur, dmgreen, etiotto, Whitney, fhahn, hfinkel Reviewed By: Meinersbur Subscribers: ormris, hiraditya, llvm-commits Tags: #llvm Differential Revision: https://reviews.llvm.org/D71025
2019-12-17 00:57:52 +08:00
; CHECK: br i1 %{{.*}}, label %[[LOOP1HEADER]], label %[[LOOPEXIT:bb[0-9]+]]
Add basic loop fusion pass. This patch adds a basic loop fusion pass. It will fuse loops that conform to the following 4 conditions: 1. Adjacent (no code between them) 2. Control flow equivalent (if one loop executes, the other loop executes) 3. Identical bounds (both loops iterate the same number of iterations) 4. No negative distance dependencies between the loop bodies. The pass does not make any changes to the IR to create opportunities for fusion. Instead, it checks if the necessary conditions are met and if so it fuses two loops together. The pass has not been added to the pass pipeline yet, and thus is not enabled by default. It can be run stand alone using the -loop-fusion option. Differential Revision: https://reviews.llvm.org/D55851 llvm-svn: 358607
2019-04-18 02:53:27 +08:00
; CHECK: ret void
define void @dep_free() {
bb:
[LoopFusion] Restrict loop fusion to rotated loops. Summary: This patch restricts loop fusion to only consider rotated loops as valid candidates. This simplifies the analysis and transformation and aligns with other loop optimizations. Reviewers: jdoerfert, Meinersbur, dmgreen, etiotto, Whitney, fhahn, hfinkel Reviewed By: Meinersbur Subscribers: ormris, hiraditya, llvm-commits Tags: #llvm Differential Revision: https://reviews.llvm.org/D71025
2019-12-17 00:57:52 +08:00
br label %bb15
Add basic loop fusion pass. This patch adds a basic loop fusion pass. It will fuse loops that conform to the following 4 conditions: 1. Adjacent (no code between them) 2. Control flow equivalent (if one loop executes, the other loop executes) 3. Identical bounds (both loops iterate the same number of iterations) 4. No negative distance dependencies between the loop bodies. The pass does not make any changes to the IR to create opportunities for fusion. Instead, it checks if the necessary conditions are met and if so it fuses two loops together. The pass has not been added to the pass pipeline yet, and thus is not enabled by default. It can be run stand alone using the -loop-fusion option. Differential Revision: https://reviews.llvm.org/D55851 llvm-svn: 358607
2019-04-18 02:53:27 +08:00
[LoopFusion] Restrict loop fusion to rotated loops. Summary: This patch restricts loop fusion to only consider rotated loops as valid candidates. This simplifies the analysis and transformation and aligns with other loop optimizations. Reviewers: jdoerfert, Meinersbur, dmgreen, etiotto, Whitney, fhahn, hfinkel Reviewed By: Meinersbur Subscribers: ormris, hiraditya, llvm-commits Tags: #llvm Differential Revision: https://reviews.llvm.org/D71025
2019-12-17 00:57:52 +08:00
bb25.preheader: ; preds = %bb22
br label %bb27
Add basic loop fusion pass. This patch adds a basic loop fusion pass. It will fuse loops that conform to the following 4 conditions: 1. Adjacent (no code between them) 2. Control flow equivalent (if one loop executes, the other loop executes) 3. Identical bounds (both loops iterate the same number of iterations) 4. No negative distance dependencies between the loop bodies. The pass does not make any changes to the IR to create opportunities for fusion. Instead, it checks if the necessary conditions are met and if so it fuses two loops together. The pass has not been added to the pass pipeline yet, and thus is not enabled by default. It can be run stand alone using the -loop-fusion option. Differential Revision: https://reviews.llvm.org/D55851 llvm-svn: 358607
2019-04-18 02:53:27 +08:00
[LoopFusion] Restrict loop fusion to rotated loops. Summary: This patch restricts loop fusion to only consider rotated loops as valid candidates. This simplifies the analysis and transformation and aligns with other loop optimizations. Reviewers: jdoerfert, Meinersbur, dmgreen, etiotto, Whitney, fhahn, hfinkel Reviewed By: Meinersbur Subscribers: ormris, hiraditya, llvm-commits Tags: #llvm Differential Revision: https://reviews.llvm.org/D71025
2019-12-17 00:57:52 +08:00
bb15: ; preds = %bb, %bb22
%.08 = phi i32 [ 0, %bb ], [ %tmp23, %bb22 ]
%indvars.iv107 = phi i64 [ 0, %bb ], [ %indvars.iv.next11, %bb22 ]
%tmp = add nsw i32 %.08, -3
%tmp16 = add nuw nsw i64 %indvars.iv107, 3
Add basic loop fusion pass. This patch adds a basic loop fusion pass. It will fuse loops that conform to the following 4 conditions: 1. Adjacent (no code between them) 2. Control flow equivalent (if one loop executes, the other loop executes) 3. Identical bounds (both loops iterate the same number of iterations) 4. No negative distance dependencies between the loop bodies. The pass does not make any changes to the IR to create opportunities for fusion. Instead, it checks if the necessary conditions are met and if so it fuses two loops together. The pass has not been added to the pass pipeline yet, and thus is not enabled by default. It can be run stand alone using the -loop-fusion option. Differential Revision: https://reviews.llvm.org/D55851 llvm-svn: 358607
2019-04-18 02:53:27 +08:00
%tmp17 = trunc i64 %tmp16 to i32
%tmp18 = mul nsw i32 %tmp, %tmp17
[LoopFusion] Restrict loop fusion to rotated loops. Summary: This patch restricts loop fusion to only consider rotated loops as valid candidates. This simplifies the analysis and transformation and aligns with other loop optimizations. Reviewers: jdoerfert, Meinersbur, dmgreen, etiotto, Whitney, fhahn, hfinkel Reviewed By: Meinersbur Subscribers: ormris, hiraditya, llvm-commits Tags: #llvm Differential Revision: https://reviews.llvm.org/D71025
2019-12-17 00:57:52 +08:00
%tmp19 = trunc i64 %indvars.iv107 to i32
Add basic loop fusion pass. This patch adds a basic loop fusion pass. It will fuse loops that conform to the following 4 conditions: 1. Adjacent (no code between them) 2. Control flow equivalent (if one loop executes, the other loop executes) 3. Identical bounds (both loops iterate the same number of iterations) 4. No negative distance dependencies between the loop bodies. The pass does not make any changes to the IR to create opportunities for fusion. Instead, it checks if the necessary conditions are met and if so it fuses two loops together. The pass has not been added to the pass pipeline yet, and thus is not enabled by default. It can be run stand alone using the -loop-fusion option. Differential Revision: https://reviews.llvm.org/D55851 llvm-svn: 358607
2019-04-18 02:53:27 +08:00
%tmp20 = srem i32 %tmp18, %tmp19
[LoopFusion] Restrict loop fusion to rotated loops. Summary: This patch restricts loop fusion to only consider rotated loops as valid candidates. This simplifies the analysis and transformation and aligns with other loop optimizations. Reviewers: jdoerfert, Meinersbur, dmgreen, etiotto, Whitney, fhahn, hfinkel Reviewed By: Meinersbur Subscribers: ormris, hiraditya, llvm-commits Tags: #llvm Differential Revision: https://reviews.llvm.org/D71025
2019-12-17 00:57:52 +08:00
%tmp21 = getelementptr inbounds [1024 x i32], [1024 x i32]* @A, i64 0, i64 %indvars.iv107
Add basic loop fusion pass. This patch adds a basic loop fusion pass. It will fuse loops that conform to the following 4 conditions: 1. Adjacent (no code between them) 2. Control flow equivalent (if one loop executes, the other loop executes) 3. Identical bounds (both loops iterate the same number of iterations) 4. No negative distance dependencies between the loop bodies. The pass does not make any changes to the IR to create opportunities for fusion. Instead, it checks if the necessary conditions are met and if so it fuses two loops together. The pass has not been added to the pass pipeline yet, and thus is not enabled by default. It can be run stand alone using the -loop-fusion option. Differential Revision: https://reviews.llvm.org/D55851 llvm-svn: 358607
2019-04-18 02:53:27 +08:00
store i32 %tmp20, i32* %tmp21, align 4
br label %bb22
bb22: ; preds = %bb15
[LoopFusion] Restrict loop fusion to rotated loops. Summary: This patch restricts loop fusion to only consider rotated loops as valid candidates. This simplifies the analysis and transformation and aligns with other loop optimizations. Reviewers: jdoerfert, Meinersbur, dmgreen, etiotto, Whitney, fhahn, hfinkel Reviewed By: Meinersbur Subscribers: ormris, hiraditya, llvm-commits Tags: #llvm Differential Revision: https://reviews.llvm.org/D71025
2019-12-17 00:57:52 +08:00
%indvars.iv.next11 = add nuw nsw i64 %indvars.iv107, 1
%tmp23 = add nuw nsw i32 %.08, 1
%exitcond12 = icmp ne i64 %indvars.iv.next11, 100
br i1 %exitcond12, label %bb15, label %bb25.preheader
bb38.preheader: ; preds = %bb35
br label %bb40
bb27: ; preds = %bb25.preheader, %bb35
%.016 = phi i32 [ 0, %bb25.preheader ], [ %tmp36, %bb35 ]
%indvars.iv75 = phi i64 [ 0, %bb25.preheader ], [ %indvars.iv.next8, %bb35 ]
%tmp28 = add nsw i32 %.016, -3
%tmp29 = add nuw nsw i64 %indvars.iv75, 3
Add basic loop fusion pass. This patch adds a basic loop fusion pass. It will fuse loops that conform to the following 4 conditions: 1. Adjacent (no code between them) 2. Control flow equivalent (if one loop executes, the other loop executes) 3. Identical bounds (both loops iterate the same number of iterations) 4. No negative distance dependencies between the loop bodies. The pass does not make any changes to the IR to create opportunities for fusion. Instead, it checks if the necessary conditions are met and if so it fuses two loops together. The pass has not been added to the pass pipeline yet, and thus is not enabled by default. It can be run stand alone using the -loop-fusion option. Differential Revision: https://reviews.llvm.org/D55851 llvm-svn: 358607
2019-04-18 02:53:27 +08:00
%tmp30 = trunc i64 %tmp29 to i32
%tmp31 = mul nsw i32 %tmp28, %tmp30
[LoopFusion] Restrict loop fusion to rotated loops. Summary: This patch restricts loop fusion to only consider rotated loops as valid candidates. This simplifies the analysis and transformation and aligns with other loop optimizations. Reviewers: jdoerfert, Meinersbur, dmgreen, etiotto, Whitney, fhahn, hfinkel Reviewed By: Meinersbur Subscribers: ormris, hiraditya, llvm-commits Tags: #llvm Differential Revision: https://reviews.llvm.org/D71025
2019-12-17 00:57:52 +08:00
%tmp32 = trunc i64 %indvars.iv75 to i32
Add basic loop fusion pass. This patch adds a basic loop fusion pass. It will fuse loops that conform to the following 4 conditions: 1. Adjacent (no code between them) 2. Control flow equivalent (if one loop executes, the other loop executes) 3. Identical bounds (both loops iterate the same number of iterations) 4. No negative distance dependencies between the loop bodies. The pass does not make any changes to the IR to create opportunities for fusion. Instead, it checks if the necessary conditions are met and if so it fuses two loops together. The pass has not been added to the pass pipeline yet, and thus is not enabled by default. It can be run stand alone using the -loop-fusion option. Differential Revision: https://reviews.llvm.org/D55851 llvm-svn: 358607
2019-04-18 02:53:27 +08:00
%tmp33 = srem i32 %tmp31, %tmp32
[LoopFusion] Restrict loop fusion to rotated loops. Summary: This patch restricts loop fusion to only consider rotated loops as valid candidates. This simplifies the analysis and transformation and aligns with other loop optimizations. Reviewers: jdoerfert, Meinersbur, dmgreen, etiotto, Whitney, fhahn, hfinkel Reviewed By: Meinersbur Subscribers: ormris, hiraditya, llvm-commits Tags: #llvm Differential Revision: https://reviews.llvm.org/D71025
2019-12-17 00:57:52 +08:00
%tmp34 = getelementptr inbounds [1024 x i32], [1024 x i32]* @B, i64 0, i64 %indvars.iv75
Add basic loop fusion pass. This patch adds a basic loop fusion pass. It will fuse loops that conform to the following 4 conditions: 1. Adjacent (no code between them) 2. Control flow equivalent (if one loop executes, the other loop executes) 3. Identical bounds (both loops iterate the same number of iterations) 4. No negative distance dependencies between the loop bodies. The pass does not make any changes to the IR to create opportunities for fusion. Instead, it checks if the necessary conditions are met and if so it fuses two loops together. The pass has not been added to the pass pipeline yet, and thus is not enabled by default. It can be run stand alone using the -loop-fusion option. Differential Revision: https://reviews.llvm.org/D55851 llvm-svn: 358607
2019-04-18 02:53:27 +08:00
store i32 %tmp33, i32* %tmp34, align 4
br label %bb35
bb35: ; preds = %bb27
[LoopFusion] Restrict loop fusion to rotated loops. Summary: This patch restricts loop fusion to only consider rotated loops as valid candidates. This simplifies the analysis and transformation and aligns with other loop optimizations. Reviewers: jdoerfert, Meinersbur, dmgreen, etiotto, Whitney, fhahn, hfinkel Reviewed By: Meinersbur Subscribers: ormris, hiraditya, llvm-commits Tags: #llvm Differential Revision: https://reviews.llvm.org/D71025
2019-12-17 00:57:52 +08:00
%indvars.iv.next8 = add nuw nsw i64 %indvars.iv75, 1
%tmp36 = add nuw nsw i32 %.016, 1
%exitcond9 = icmp ne i64 %indvars.iv.next8, 100
br i1 %exitcond9, label %bb27, label %bb38.preheader
bb51.preheader: ; preds = %bb48
br label %bb53
bb40: ; preds = %bb38.preheader, %bb48
%.024 = phi i32 [ 0, %bb38.preheader ], [ %tmp49, %bb48 ]
%indvars.iv43 = phi i64 [ 0, %bb38.preheader ], [ %indvars.iv.next5, %bb48 ]
%tmp41 = add nsw i32 %.024, -3
%tmp42 = add nuw nsw i64 %indvars.iv43, 3
Add basic loop fusion pass. This patch adds a basic loop fusion pass. It will fuse loops that conform to the following 4 conditions: 1. Adjacent (no code between them) 2. Control flow equivalent (if one loop executes, the other loop executes) 3. Identical bounds (both loops iterate the same number of iterations) 4. No negative distance dependencies between the loop bodies. The pass does not make any changes to the IR to create opportunities for fusion. Instead, it checks if the necessary conditions are met and if so it fuses two loops together. The pass has not been added to the pass pipeline yet, and thus is not enabled by default. It can be run stand alone using the -loop-fusion option. Differential Revision: https://reviews.llvm.org/D55851 llvm-svn: 358607
2019-04-18 02:53:27 +08:00
%tmp43 = trunc i64 %tmp42 to i32
%tmp44 = mul nsw i32 %tmp41, %tmp43
[LoopFusion] Restrict loop fusion to rotated loops. Summary: This patch restricts loop fusion to only consider rotated loops as valid candidates. This simplifies the analysis and transformation and aligns with other loop optimizations. Reviewers: jdoerfert, Meinersbur, dmgreen, etiotto, Whitney, fhahn, hfinkel Reviewed By: Meinersbur Subscribers: ormris, hiraditya, llvm-commits Tags: #llvm Differential Revision: https://reviews.llvm.org/D71025
2019-12-17 00:57:52 +08:00
%tmp45 = trunc i64 %indvars.iv43 to i32
Add basic loop fusion pass. This patch adds a basic loop fusion pass. It will fuse loops that conform to the following 4 conditions: 1. Adjacent (no code between them) 2. Control flow equivalent (if one loop executes, the other loop executes) 3. Identical bounds (both loops iterate the same number of iterations) 4. No negative distance dependencies between the loop bodies. The pass does not make any changes to the IR to create opportunities for fusion. Instead, it checks if the necessary conditions are met and if so it fuses two loops together. The pass has not been added to the pass pipeline yet, and thus is not enabled by default. It can be run stand alone using the -loop-fusion option. Differential Revision: https://reviews.llvm.org/D55851 llvm-svn: 358607
2019-04-18 02:53:27 +08:00
%tmp46 = srem i32 %tmp44, %tmp45
[LoopFusion] Restrict loop fusion to rotated loops. Summary: This patch restricts loop fusion to only consider rotated loops as valid candidates. This simplifies the analysis and transformation and aligns with other loop optimizations. Reviewers: jdoerfert, Meinersbur, dmgreen, etiotto, Whitney, fhahn, hfinkel Reviewed By: Meinersbur Subscribers: ormris, hiraditya, llvm-commits Tags: #llvm Differential Revision: https://reviews.llvm.org/D71025
2019-12-17 00:57:52 +08:00
%tmp47 = getelementptr inbounds [1024 x i32], [1024 x i32]* @C, i64 0, i64 %indvars.iv43
Add basic loop fusion pass. This patch adds a basic loop fusion pass. It will fuse loops that conform to the following 4 conditions: 1. Adjacent (no code between them) 2. Control flow equivalent (if one loop executes, the other loop executes) 3. Identical bounds (both loops iterate the same number of iterations) 4. No negative distance dependencies between the loop bodies. The pass does not make any changes to the IR to create opportunities for fusion. Instead, it checks if the necessary conditions are met and if so it fuses two loops together. The pass has not been added to the pass pipeline yet, and thus is not enabled by default. It can be run stand alone using the -loop-fusion option. Differential Revision: https://reviews.llvm.org/D55851 llvm-svn: 358607
2019-04-18 02:53:27 +08:00
store i32 %tmp46, i32* %tmp47, align 4
br label %bb48
bb48: ; preds = %bb40
[LoopFusion] Restrict loop fusion to rotated loops. Summary: This patch restricts loop fusion to only consider rotated loops as valid candidates. This simplifies the analysis and transformation and aligns with other loop optimizations. Reviewers: jdoerfert, Meinersbur, dmgreen, etiotto, Whitney, fhahn, hfinkel Reviewed By: Meinersbur Subscribers: ormris, hiraditya, llvm-commits Tags: #llvm Differential Revision: https://reviews.llvm.org/D71025
2019-12-17 00:57:52 +08:00
%indvars.iv.next5 = add nuw nsw i64 %indvars.iv43, 1
%tmp49 = add nuw nsw i32 %.024, 1
%exitcond6 = icmp ne i64 %indvars.iv.next5, 100
br i1 %exitcond6, label %bb40, label %bb51.preheader
Add basic loop fusion pass. This patch adds a basic loop fusion pass. It will fuse loops that conform to the following 4 conditions: 1. Adjacent (no code between them) 2. Control flow equivalent (if one loop executes, the other loop executes) 3. Identical bounds (both loops iterate the same number of iterations) 4. No negative distance dependencies between the loop bodies. The pass does not make any changes to the IR to create opportunities for fusion. Instead, it checks if the necessary conditions are met and if so it fuses two loops together. The pass has not been added to the pass pipeline yet, and thus is not enabled by default. It can be run stand alone using the -loop-fusion option. Differential Revision: https://reviews.llvm.org/D55851 llvm-svn: 358607
2019-04-18 02:53:27 +08:00
[LoopFusion] Restrict loop fusion to rotated loops. Summary: This patch restricts loop fusion to only consider rotated loops as valid candidates. This simplifies the analysis and transformation and aligns with other loop optimizations. Reviewers: jdoerfert, Meinersbur, dmgreen, etiotto, Whitney, fhahn, hfinkel Reviewed By: Meinersbur Subscribers: ormris, hiraditya, llvm-commits Tags: #llvm Differential Revision: https://reviews.llvm.org/D71025
2019-12-17 00:57:52 +08:00
bb52: ; preds = %bb61
Add basic loop fusion pass. This patch adds a basic loop fusion pass. It will fuse loops that conform to the following 4 conditions: 1. Adjacent (no code between them) 2. Control flow equivalent (if one loop executes, the other loop executes) 3. Identical bounds (both loops iterate the same number of iterations) 4. No negative distance dependencies between the loop bodies. The pass does not make any changes to the IR to create opportunities for fusion. Instead, it checks if the necessary conditions are met and if so it fuses two loops together. The pass has not been added to the pass pipeline yet, and thus is not enabled by default. It can be run stand alone using the -loop-fusion option. Differential Revision: https://reviews.llvm.org/D55851 llvm-svn: 358607
2019-04-18 02:53:27 +08:00
br label %bb63
[LoopFusion] Restrict loop fusion to rotated loops. Summary: This patch restricts loop fusion to only consider rotated loops as valid candidates. This simplifies the analysis and transformation and aligns with other loop optimizations. Reviewers: jdoerfert, Meinersbur, dmgreen, etiotto, Whitney, fhahn, hfinkel Reviewed By: Meinersbur Subscribers: ormris, hiraditya, llvm-commits Tags: #llvm Differential Revision: https://reviews.llvm.org/D71025
2019-12-17 00:57:52 +08:00
bb53: ; preds = %bb51.preheader, %bb61
%.032 = phi i32 [ 0, %bb51.preheader ], [ %tmp62, %bb61 ]
%indvars.iv1 = phi i64 [ 0, %bb51.preheader ], [ %indvars.iv.next, %bb61 ]
%tmp54 = add nsw i32 %.032, -3
%tmp55 = add nuw nsw i64 %indvars.iv1, 3
Add basic loop fusion pass. This patch adds a basic loop fusion pass. It will fuse loops that conform to the following 4 conditions: 1. Adjacent (no code between them) 2. Control flow equivalent (if one loop executes, the other loop executes) 3. Identical bounds (both loops iterate the same number of iterations) 4. No negative distance dependencies between the loop bodies. The pass does not make any changes to the IR to create opportunities for fusion. Instead, it checks if the necessary conditions are met and if so it fuses two loops together. The pass has not been added to the pass pipeline yet, and thus is not enabled by default. It can be run stand alone using the -loop-fusion option. Differential Revision: https://reviews.llvm.org/D55851 llvm-svn: 358607
2019-04-18 02:53:27 +08:00
%tmp56 = trunc i64 %tmp55 to i32
%tmp57 = mul nsw i32 %tmp54, %tmp56
[LoopFusion] Restrict loop fusion to rotated loops. Summary: This patch restricts loop fusion to only consider rotated loops as valid candidates. This simplifies the analysis and transformation and aligns with other loop optimizations. Reviewers: jdoerfert, Meinersbur, dmgreen, etiotto, Whitney, fhahn, hfinkel Reviewed By: Meinersbur Subscribers: ormris, hiraditya, llvm-commits Tags: #llvm Differential Revision: https://reviews.llvm.org/D71025
2019-12-17 00:57:52 +08:00
%tmp58 = trunc i64 %indvars.iv1 to i32
Add basic loop fusion pass. This patch adds a basic loop fusion pass. It will fuse loops that conform to the following 4 conditions: 1. Adjacent (no code between them) 2. Control flow equivalent (if one loop executes, the other loop executes) 3. Identical bounds (both loops iterate the same number of iterations) 4. No negative distance dependencies between the loop bodies. The pass does not make any changes to the IR to create opportunities for fusion. Instead, it checks if the necessary conditions are met and if so it fuses two loops together. The pass has not been added to the pass pipeline yet, and thus is not enabled by default. It can be run stand alone using the -loop-fusion option. Differential Revision: https://reviews.llvm.org/D55851 llvm-svn: 358607
2019-04-18 02:53:27 +08:00
%tmp59 = srem i32 %tmp57, %tmp58
[LoopFusion] Restrict loop fusion to rotated loops. Summary: This patch restricts loop fusion to only consider rotated loops as valid candidates. This simplifies the analysis and transformation and aligns with other loop optimizations. Reviewers: jdoerfert, Meinersbur, dmgreen, etiotto, Whitney, fhahn, hfinkel Reviewed By: Meinersbur Subscribers: ormris, hiraditya, llvm-commits Tags: #llvm Differential Revision: https://reviews.llvm.org/D71025
2019-12-17 00:57:52 +08:00
%tmp60 = getelementptr inbounds [1024 x i32], [1024 x i32]* @D, i64 0, i64 %indvars.iv1
Add basic loop fusion pass. This patch adds a basic loop fusion pass. It will fuse loops that conform to the following 4 conditions: 1. Adjacent (no code between them) 2. Control flow equivalent (if one loop executes, the other loop executes) 3. Identical bounds (both loops iterate the same number of iterations) 4. No negative distance dependencies between the loop bodies. The pass does not make any changes to the IR to create opportunities for fusion. Instead, it checks if the necessary conditions are met and if so it fuses two loops together. The pass has not been added to the pass pipeline yet, and thus is not enabled by default. It can be run stand alone using the -loop-fusion option. Differential Revision: https://reviews.llvm.org/D55851 llvm-svn: 358607
2019-04-18 02:53:27 +08:00
store i32 %tmp59, i32* %tmp60, align 4
br label %bb61
bb61: ; preds = %bb53
[LoopFusion] Restrict loop fusion to rotated loops. Summary: This patch restricts loop fusion to only consider rotated loops as valid candidates. This simplifies the analysis and transformation and aligns with other loop optimizations. Reviewers: jdoerfert, Meinersbur, dmgreen, etiotto, Whitney, fhahn, hfinkel Reviewed By: Meinersbur Subscribers: ormris, hiraditya, llvm-commits Tags: #llvm Differential Revision: https://reviews.llvm.org/D71025
2019-12-17 00:57:52 +08:00
%indvars.iv.next = add nuw nsw i64 %indvars.iv1, 1
%tmp62 = add nuw nsw i32 %.032, 1
%exitcond = icmp ne i64 %indvars.iv.next, 100
br i1 %exitcond, label %bb53, label %bb52
Add basic loop fusion pass. This patch adds a basic loop fusion pass. It will fuse loops that conform to the following 4 conditions: 1. Adjacent (no code between them) 2. Control flow equivalent (if one loop executes, the other loop executes) 3. Identical bounds (both loops iterate the same number of iterations) 4. No negative distance dependencies between the loop bodies. The pass does not make any changes to the IR to create opportunities for fusion. Instead, it checks if the necessary conditions are met and if so it fuses two loops together. The pass has not been added to the pass pipeline yet, and thus is not enabled by default. It can be run stand alone using the -loop-fusion option. Differential Revision: https://reviews.llvm.org/D55851 llvm-svn: 358607
2019-04-18 02:53:27 +08:00
bb63: ; preds = %bb52
ret void
}