[PATCH] D71739: [WIP] Use operand bundles to encode alignment assumptions

[Tyker via Phabricator via cfe-commits]

Tyker added a comment.

In D71739#1961508 , @jdoerfert wrote:

> @lebedev.ri  We'd need to identify other uses of the alignment encoding 
> in-tree so we can replace them as well. Also, this patch uses not only the 
> alignment but also the offset in the operand bundle. We can either allow that 
> or encode the offset via a gep in the IR. I guess the latter is easier to 
> implement until we have more reasons to allow more complex operand bundles 
> (which we will need to have eventually).


for now i think we will stay with the current "simple" alignment assumptions in 
operand bundles. but we can improve it later.

> @Tyker Do you want to take this?

i am fine with taking this. but there is a few thing to do before this.

i think that this patch depends on a few things:

- add an API to build assumes from provided knowledge.
- update users of alignment assumptions.

and a few things that i would like to do before:

- finish patches currently in review.
- improve generation of assume with operand bundles to minimize duplicates and 
extra instructions.


Repository:
  rG LLVM Github Monorepo

CHANGES SINCE LAST ACTION
  https://reviews.llvm.org/D71739/new/

https://reviews.llvm.org/D71739



___
cfe-commits mailing list
cfe-commits@lists.llvm.org
https://lists.llvm.org/cgi-bin/mailman/listinfo/cfe-commits

[PATCH] D71739: [WIP] Use operand bundles to encode alignment assumptions

[Johannes Doerfert via Phabricator via cfe-commits]

jdoerfert added a comment.

@lebedev.ri  We'd need to identify other uses of the alignment encoding in-tree 
so we can replace them as well. Also, this patch uses not only the alignment 
but also the offset in the operand bundle. We can either allow that or encode 
the offset via a gep in the IR. I guess the latter is easier to implement until 
we have more reasons to allow more complex operand bundles (which we will need 
to have eventually).

@Tyker Do you want to take this?


Repository:
  rG LLVM Github Monorepo

CHANGES SINCE LAST ACTION
  https://reviews.llvm.org/D71739/new/

https://reviews.llvm.org/D71739



___
cfe-commits mailing list
cfe-commits@lists.llvm.org
https://lists.llvm.org/cgi-bin/mailman/listinfo/cfe-commits

[PATCH] D71739: [WIP] Use operand bundles to encode alignment assumptions

[Roman Lebedev via Phabricator via cfe-commits]

lebedev.ri added a comment.

What's the status here?


Repository:
  rG LLVM Github Monorepo

CHANGES SINCE LAST ACTION
  https://reviews.llvm.org/D71739/new/

https://reviews.llvm.org/D71739



___
cfe-commits mailing list
cfe-commits@lists.llvm.org
https://lists.llvm.org/cgi-bin/mailman/listinfo/cfe-commits

[PATCH] D71739: [WIP] Use operand bundles to encode alignment assumptions

[pre-merge checks [bot] via Phabricator via cfe-commits]

merge_guards_bot added a comment.

{icon times-circle color=red} Unit tests: fail. 60947 tests passed, 20 failed 
and 726 were skipped.

  failed: Clang.CodeGen/align_value.cpp
  failed: Clang.CodeGen/builtin-assume-aligned.c
  failed: Clang.CodeGen/builtin-movdir.c
  failed: 
Clang.CodeGen/catch-alignment-assumption-attribute-align_value-on-lvalue.cpp
  failed: 
Clang.CodeGen/catch-alignment-assumption-attribute-align_value-on-paramvar.cpp
  failed: 
Clang.CodeGen/catch-alignment-assumption-attribute-alloc_align-on-function-variable.cpp
  failed: 
Clang.CodeGen/catch-alignment-assumption-attribute-alloc_align-on-function.cpp
  failed: 
Clang.CodeGen/catch-alignment-assumption-attribute-assume_aligned-on-function-two-params.cpp
  failed: 
Clang.CodeGen/catch-alignment-assumption-attribute-assume_aligned-on-function.cpp
  failed: Clang.CodeGen/catch-alignment-assumption-blacklist.c
  failed: 
Clang.CodeGen/catch-alignment-assumption-builtin_assume_aligned-three-params-variable.cpp
  failed: 
Clang.CodeGen/catch-alignment-assumption-builtin_assume_aligned-three-params.cpp
  failed: 
Clang.CodeGen/catch-alignment-assumption-builtin_assume_aligned-two-params.cpp
  failed: Clang.CodeGen/catch-alignment-assumption-openmp.cpp
  failed: Clang.OpenMP/simd_codegen.cpp
  failed: Clang.OpenMP/simd_metadata.c
  failed: Clang.OpenMP/target_teams_distribute_parallel_for_simd_codegen.cpp
  failed: LLVM.Transforms/AlignmentFromAssumptions/simple32.ll
  failed: LLVM.Transforms/Inline/align.ll
  failed: LLVM.Transforms/InstCombine/assume.ll

{icon times-circle color=red} clang-tidy: fail. Please fix clang-tidy findings 
.

{icon times-circle color=red} clang-format: fail. Please format your changes 
with clang-format by running `git-clang-format HEAD^` or applying this patch 
.

Build artifacts 
: 
diff.json 
,
 clang-tidy.txt 
,
 clang-format.patch 
,
 CMakeCache.txt 
,
 console-log.txt 
,
 test-results.xml 



Repository:
  rG LLVM Github Monorepo

CHANGES SINCE LAST ACTION
  https://reviews.llvm.org/D71739/new/

https://reviews.llvm.org/D71739



___
cfe-commits mailing list
cfe-commits@lists.llvm.org
https://lists.llvm.org/cgi-bin/mailman/listinfo/cfe-commits

[PATCH] D71739: [WIP] Use operand bundles to encode alignment assumptions

[Johannes Doerfert via Phabricator via cfe-commits]

jdoerfert updated this revision to Diff 234817.
jdoerfert added a comment.

Rebase


Repository:
  rG LLVM Github Monorepo

CHANGES SINCE LAST ACTION
  https://reviews.llvm.org/D71739/new/

https://reviews.llvm.org/D71739

Files:
  clang/lib/CodeGen/CodeGenFunction.cpp
  clang/test/CodeGen/alloc-align-attr.c
  llvm/include/llvm/IR/IRBuilder.h
  llvm/lib/IR/IRBuilder.cpp
  llvm/lib/Transforms/InstCombine/InstCombineCalls.cpp
  llvm/lib/Transforms/Scalar/AlignmentFromAssumptions.cpp
  llvm/test/Transforms/AlignmentFromAssumptions/simple.ll

Index: llvm/test/Transforms/AlignmentFromAssumptions/simple.ll
===
--- llvm/test/Transforms/AlignmentFromAssumptions/simple.ll
+++ llvm/test/Transforms/AlignmentFromAssumptions/simple.ll
@@ -4,10 +4,7 @@
 
 define i32 @foo(i32* nocapture %a) nounwind uwtable readonly {
 entry:
-  %ptrint = ptrtoint i32* %a to i64
-  %maskedptr = and i64 %ptrint, 31
-  %maskcond = icmp eq i64 %maskedptr, 0
-  tail call void @llvm.assume(i1 %maskcond)
+  tail call void @llvm.assume(i1 true) ["align"(i32* %a, i32 32)]
   %0 = load i32, i32* %a, align 4
   ret i32 %0
 
@@ -18,11 +15,7 @@
 
 define i32 @foo2(i32* nocapture %a) nounwind uwtable readonly {
 entry:
-  %ptrint = ptrtoint i32* %a to i64
-  %offsetptr = add i64 %ptrint, 24
-  %maskedptr = and i64 %offsetptr, 31
-  %maskcond = icmp eq i64 %maskedptr, 0
-  tail call void @llvm.assume(i1 %maskcond)
+  tail call void @llvm.assume(i1 true) ["align"(i32* %a, i32 32, i32 24)]
   %arrayidx = getelementptr inbounds i32, i32* %a, i64 2
   %0 = load i32, i32* %arrayidx, align 4
   ret i32 %0
@@ -34,11 +27,7 @@
 
 define i32 @foo2a(i32* nocapture %a) nounwind uwtable readonly {
 entry:
-  %ptrint = ptrtoint i32* %a to i64
-  %offsetptr = add i64 %ptrint, 28
-  %maskedptr = and i64 %offsetptr, 31
-  %maskcond = icmp eq i64 %maskedptr, 0
-  tail call void @llvm.assume(i1 %maskcond)
+  tail call void @llvm.assume(i1 true) ["align"(i32* %a, i32 32, i32 28)]
   %arrayidx = getelementptr inbounds i32, i32* %a, i64 -1
   %0 = load i32, i32* %arrayidx, align 4
   ret i32 %0
@@ -50,10 +39,7 @@
 
 define i32 @goo(i32* nocapture %a) nounwind uwtable readonly {
 entry:
-  %ptrint = ptrtoint i32* %a to i64
-  %maskedptr = and i64 %ptrint, 31
-  %maskcond = icmp eq i64 %maskedptr, 0
-  tail call void @llvm.assume(i1 %maskcond)
+  tail call void @llvm.assume(i1 true) ["align"(i32* %a, i32 32, i32 0)]
   %0 = load i32, i32* %a, align 4
   ret i32 %0
 
@@ -64,10 +50,7 @@
 
 define i32 @hoo(i32* nocapture %a) nounwind uwtable readonly {
 entry:
-  %ptrint = ptrtoint i32* %a to i64
-  %maskedptr = and i64 %ptrint, 31
-  %maskcond = icmp eq i64 %maskedptr, 0
-  tail call void @llvm.assume(i1 %maskcond)
+  tail call void @llvm.assume(i1 true) ["align"(i32* %a, i64 32, i32 0)]
   br label %for.body
 
 for.body: ; preds = %entry, %for.body
@@ -98,10 +81,7 @@
 ; load(a, i0+i1+i2+32)
 define void @hoo2(i32* nocapture %a, i64 %id, i64 %num) nounwind uwtable readonly {
 entry:
-  %ptrint = ptrtoint i32* %a to i64
-  %maskedptr = and i64 %ptrint, 31
-  %maskcond = icmp eq i64 %maskedptr, 0
-  tail call void @llvm.assume(i1 %maskcond)
+  tail call void @llvm.assume(i1 true) ["align"(i32* %a, i8 32, i64 0)]
   %id.mul = shl nsw i64 %id, 6
   %num.mul = shl nsw i64 %num, 6
   br label %for0.body
@@ -147,10 +127,7 @@
 
 define i32 @joo(i32* nocapture %a) nounwind uwtable readonly {
 entry:
-  %ptrint = ptrtoint i32* %a to i64
-  %maskedptr = and i64 %ptrint, 31
-  %maskcond = icmp eq i64 %maskedptr, 0
-  tail call void @llvm.assume(i1 %maskcond)
+  tail call void @llvm.assume(i1 true) ["align"(i32* %a, i8 32, i8 0)]
   br label %for.body
 
 for.body: ; preds = %entry, %for.body
@@ -175,16 +152,13 @@
 
 define i32 @koo(i32* nocapture %a) nounwind uwtable readonly {
 entry:
-  %ptrint = ptrtoint i32* %a to i64
-  %maskedptr = and i64 %ptrint, 31
-  %maskcond = icmp eq i64 %maskedptr, 0
-  tail call void @llvm.assume(i1 %maskcond)
   br label %for.body
 
 for.body: ; preds = %entry, %for.body
   %indvars.iv = phi i64 [ 0, %entry ], [ %indvars.iv.next, %for.body ]
   %r.06 = phi i32 [ 0, %entry ], [ %add, %for.body ]
   %arrayidx = getelementptr inbounds i32, i32* %a, i64 %indvars.iv
+  tail call void @llvm.assume(i1 true) ["align"(i32* %a, i8 32, i8 0)]
   %0 = load i32, i32* %arrayidx, align 4
   %add = add nsw i32 %0, %r.06
   %indvars.iv.next = add i64 %indvars.iv, 4
@@ -203,10 +177,7 @@
 
 define i32 @koo2(i32* nocapture %a) nounwind uwtable readonly {
 entry:
-  %ptrint = ptrtoint i32* %a to i64
-  %maskedptr = and i64 %ptrint, 31
-  %maskcond = icmp eq i64 %maskedptr, 0
-  tail call void @llvm.assume(i1 %maskcond)
+  tail call void @llvm.assume(i1 true) ["align"(i32* %a, i128 32, i128 0)]
   br label %for.body
 
 for.body: ; preds = %entry, %for

[PATCH] D71739: [WIP] Use operand bundles to encode alignment assumptions

[pre-merge checks [bot] via Phabricator via cfe-commits]

merge_guards_bot added a comment.

{icon question-circle color=gray} Unit tests: unknown.

{icon question-circle color=gray} clang-tidy: unknown.

{icon question-circle color=gray} clang-format: unknown.

Build artifacts 
: 
diff.json 
,
 console-log.txt 



Repository:
  rG LLVM Github Monorepo

CHANGES SINCE LAST ACTION
  https://reviews.llvm.org/D71739/new/

https://reviews.llvm.org/D71739



___
cfe-commits mailing list
cfe-commits@lists.llvm.org
https://lists.llvm.org/cgi-bin/mailman/listinfo/cfe-commits

[PATCH] D71739: [WIP] Use operand bundles to encode alignment assumptions

[Johannes Doerfert via Phabricator via cfe-commits]

jdoerfert created this revision.
jdoerfert added reviewers: hfinkel, xbolva00, lebedev.ri, nikic, Tyker, 
rjmccall, spatel.
Herald added subscribers: bollu, hiraditya.
Herald added projects: clang, LLVM.

NOTE: This is a prototype not a finished patch!
NOTE: There is a mailing list discussion on this: 
http://lists.llvm.org/pipermail/llvm-dev/2019-December/137632.html

Complemantary to the assumption outliner prototype in D71692 
, this patch
shows how we could simplify the code emitted for an alignemnt
assumption. The generated code is smaller, less fragile, and it makes it
easier to recognize the additional use as a "assumption use".

As mentioned in D71692  and on the mailing 
list, we could adopt this
scheme, and similar schemes for other patterns, without adopting the
assumption outlining.


Repository:
  rG LLVM Github Monorepo

https://reviews.llvm.org/D71739

Files:
  clang/lib/CodeGen/CodeGenFunction.cpp
  clang/test/CodeGen/alloc-align-attr.c
  llvm/include/llvm/IR/IRBuilder.h
  llvm/lib/IR/IRBuilder.cpp
  llvm/lib/Transforms/InstCombine/InstCombineCalls.cpp
  llvm/lib/Transforms/Scalar/AlignmentFromAssumptions.cpp
  llvm/test/Transforms/AlignmentFromAssumptions/simple.ll

Index: llvm/test/Transforms/AlignmentFromAssumptions/simple.ll
===
--- llvm/test/Transforms/AlignmentFromAssumptions/simple.ll
+++ llvm/test/Transforms/AlignmentFromAssumptions/simple.ll
@@ -6,10 +6,7 @@
 
 define i32 @foo(i32* nocapture %a) nounwind uwtable readonly {
 entry:
-  %ptrint = ptrtoint i32* %a to i64
-  %maskedptr = and i64 %ptrint, 31
-  %maskcond = icmp eq i64 %maskedptr, 0
-  tail call void @llvm.assume(i1 %maskcond)
+  tail call void @llvm.assume(i1 true) ["align"(i32* %a, i32 32)]
   %0 = load i32, i32* %a, align 4
   ret i32 %0
 
@@ -20,11 +17,7 @@
 
 define i32 @foo2(i32* nocapture %a) nounwind uwtable readonly {
 entry:
-  %ptrint = ptrtoint i32* %a to i64
-  %offsetptr = add i64 %ptrint, 24
-  %maskedptr = and i64 %offsetptr, 31
-  %maskcond = icmp eq i64 %maskedptr, 0
-  tail call void @llvm.assume(i1 %maskcond)
+  tail call void @llvm.assume(i1 true) ["align"(i32* %a, i32 32, i32 24)]
   %arrayidx = getelementptr inbounds i32, i32* %a, i64 2
   %0 = load i32, i32* %arrayidx, align 4
   ret i32 %0
@@ -36,11 +29,7 @@
 
 define i32 @foo2a(i32* nocapture %a) nounwind uwtable readonly {
 entry:
-  %ptrint = ptrtoint i32* %a to i64
-  %offsetptr = add i64 %ptrint, 28
-  %maskedptr = and i64 %offsetptr, 31
-  %maskcond = icmp eq i64 %maskedptr, 0
-  tail call void @llvm.assume(i1 %maskcond)
+  tail call void @llvm.assume(i1 true) ["align"(i32* %a, i32 32, i32 28)]
   %arrayidx = getelementptr inbounds i32, i32* %a, i64 -1
   %0 = load i32, i32* %arrayidx, align 4
   ret i32 %0
@@ -52,10 +41,7 @@
 
 define i32 @goo(i32* nocapture %a) nounwind uwtable readonly {
 entry:
-  %ptrint = ptrtoint i32* %a to i64
-  %maskedptr = and i64 %ptrint, 31
-  %maskcond = icmp eq i64 %maskedptr, 0
-  tail call void @llvm.assume(i1 %maskcond)
+  tail call void @llvm.assume(i1 true) ["align"(i32* %a, i32 32, i32 0)]
   %0 = load i32, i32* %a, align 4
   ret i32 %0
 
@@ -66,10 +52,7 @@
 
 define i32 @hoo(i32* nocapture %a) nounwind uwtable readonly {
 entry:
-  %ptrint = ptrtoint i32* %a to i64
-  %maskedptr = and i64 %ptrint, 31
-  %maskcond = icmp eq i64 %maskedptr, 0
-  tail call void @llvm.assume(i1 %maskcond)
+  tail call void @llvm.assume(i1 true) ["align"(i32* %a, i64 32, i32 0)]
   br label %for.body
 
 for.body: ; preds = %entry, %for.body
@@ -100,10 +83,7 @@
 ; load(a, i0+i1+i2+32)
 define void @hoo2(i32* nocapture %a, i64 %id, i64 %num) nounwind uwtable readonly {
 entry:
-  %ptrint = ptrtoint i32* %a to i64
-  %maskedptr = and i64 %ptrint, 31
-  %maskcond = icmp eq i64 %maskedptr, 0
-  tail call void @llvm.assume(i1 %maskcond)
+  tail call void @llvm.assume(i1 true) ["align"(i32* %a, i8 32, i64 0)]
   %id.mul = shl nsw i64 %id, 6
   %num.mul = shl nsw i64 %num, 6
   br label %for0.body
@@ -149,10 +129,7 @@
 
 define i32 @joo(i32* nocapture %a) nounwind uwtable readonly {
 entry:
-  %ptrint = ptrtoint i32* %a to i64
-  %maskedptr = and i64 %ptrint, 31
-  %maskcond = icmp eq i64 %maskedptr, 0
-  tail call void @llvm.assume(i1 %maskcond)
+  tail call void @llvm.assume(i1 true) ["align"(i32* %a, i8 32, i8 0)]
   br label %for.body
 
 for.body: ; preds = %entry, %for.body
@@ -177,16 +154,13 @@
 
 define i32 @koo(i32* nocapture %a) nounwind uwtable readonly {
 entry:
-  %ptrint = ptrtoint i32* %a to i64
-  %maskedptr = and i64 %ptrint, 31
-  %maskcond = icmp eq i64 %maskedptr, 0
-  tail call void @llvm.assume(i1 %maskcond)
   br label %for.body
 
 for.body: ; preds = %entry, %for.body
   %indvars.iv = phi i64 [ 0, %entry ], [ %indvars.iv.next, %for.body ]
   %r.06 = phi i3