[llvm] 74f03e4 - [DSE, MSSA] Adjust mda-with-dbg-values.ll to MSSA backed DSE.

[Alina Sbirlea via llvm-commits]

Just a general comment on this one: I'd favor not diverging the test
contents too much, assuming they will eventually get re-merged. So keeping
the same test with different RUN lines and comments (to eventually merge)
could simplify things later.

On Mon, Feb 10, 2020 at 7:24 AM Florian Hahn via llvm-commits <
llvm-commits at lists.llvm.org> wrote:

>
> Author: Florian Hahn
> Date: 2020-02-10T15:24:00Z
> New Revision: 74f03e4ff0c1f11c29102329af95f7d9782426dc
>
> URL:
> https://github.com/llvm/llvm-project/commit/74f03e4ff0c1f11c29102329af95f7d9782426dc
> DIFF:
> https://github.com/llvm/llvm-project/commit/74f03e4ff0c1f11c29102329af95f7d9782426dc.diff
>
> LOG: [DSE,MSSA] Adjust mda-with-dbg-values.ll to MSSA backed DSE.
>
> -memdep-block-scan-limit is not relevant with MSSA.
>
> Added:
>
>
> Modified:
>     llvm/test/Transforms/DeadStoreElimination/MSSA/mda-with-dbg-values.ll
>
> Removed:
>
>
>
>
> ################################################################################
> diff  --git
> a/llvm/test/Transforms/DeadStoreElimination/MSSA/mda-with-dbg-values.ll
> b/llvm/test/Transforms/DeadStoreElimination/MSSA/mda-with-dbg-values.ll
> index 52e2ecdc7290..06bcc346a435 100644
> --- a/llvm/test/Transforms/DeadStoreElimination/MSSA/mda-with-dbg-values.ll
> +++ b/llvm/test/Transforms/DeadStoreElimination/MSSA/mda-with-dbg-values.ll
> @@ -1,13 +1,12 @@
> -; RUN: opt -S -dse -enable-dse-memoryssa -memdep-block-scan-limit=3 < %s
> | FileCheck %s
> -; RUN: opt -S -strip-debug -dse -enable-dse-memoryssa
> -memdep-block-scan-limit=3 < %s | FileCheck %s
> +; RUN: opt -S -dse -enable-dse-memoryssa -dse-memoryssa-scanlimit=2 < %s
> | FileCheck %s
> +; RUN: opt -S -strip-debug -dse -enable-dse-memoryssa
> -dse-memoryssa-scanlimit=2 < %s | FileCheck %s
>
> -; Test case to check that the memory dependency analysis gets the same
> -; result even if we have a dbg value between the memcpy and
> -; store. The memory dependency is then used by DSE to remove the store.
> +; Test case to check that DSE gets the same result even if we have a dbg
> value
> +; between the memcpy.
>
> -; We use -memdep-block-scan-limit=3 to be able to create a small test
> case.
> -; Without it, we would need to squeeze in 100 instructions since the
> default
> -; limit is 100.
> +; This test case is less relevant for the MemorySSA backed version of
> DSE, as
> +; debug values are not modeled in MemorySSA and are skipped regardless of
> the
> +; exploration limit.
>
>  target datalayout = "e-m:e-i64:64-f80:128-n8:16:32:64-S128"
>  target triple = "x86_64-unknown-linux-gnu"
> @@ -20,6 +19,11 @@ entry:
>    %i = alloca i8, align 1
>    store i8 1, i8* %i, align 1, !dbg !19
>    call void @llvm.dbg.value(metadata i32 0, i64 0, metadata !17, metadata
> !DIExpression()), !dbg !18
> +  call void @llvm.dbg.value(metadata i32 0, i64 0, metadata !17, metadata
> !DIExpression()), !dbg !18
> +  call void @llvm.dbg.value(metadata i32 0, i64 0, metadata !17, metadata
> !DIExpression()), !dbg !18
> +  call void @llvm.dbg.value(metadata i32 0, i64 0, metadata !17, metadata
> !DIExpression()), !dbg !18
> +  call void @llvm.dbg.value(metadata i32 0, i64 0, metadata !17, metadata
> !DIExpression()), !dbg !18
> +  call void @llvm.dbg.value(metadata i32 0, i64 0, metadata !17, metadata
> !DIExpression()), !dbg !18
>    %0 = bitcast [1 x i8]* @g to i8*
>    call void @llvm.memcpy.p0i8.p0i8.i64(i8* %i, i8* %0, i64 1, i1 false),
> !dbg !20
>    br label %bb2
>
>
>
> _______________________________________________
> llvm-commits mailing list
> llvm-commits at lists.llvm.org
> https://lists.llvm.org/cgi-bin/mailman/listinfo/llvm-commits
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.llvm.org/pipermail/llvm-commits/attachments/20200210/27392094/attachment.html>