Skip to content

[SimplifyCFG] Make i1 legal to build lookup tables #87431

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

Open
wants to merge 2 commits into
base: main
Choose a base branch
from
Open

[SimplifyCFG] Make i1 legal to build lookup tables #87431

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
2 commits
Select commit Hold shift + click to select a range
cb6f3c1
Add a test for missing folding
krtab Apr 10, 2024
863cd17
[SimplifyCFG] Make i1 legal to build lookup tables
krtab Apr 2, 2024
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
2 changes: 2 additions & 0 deletions llvm/lib/Transforms/Utils/SimplifyCFG.cpp
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -6448,6 +6448,8 @@ static bool isTypeLegalForLookupTable(Type *Ty, const TargetTransformInfo &TTI,
// on ABI alignment and padding in the table to allow the load to be widened.
// Or we could widen the constants and truncate the load.
unsigned BitWidth = IT->getBitWidth();
if (BitWidth == 1)
return true;
return BitWidth >= 8 && isPowerOf2_32(BitWidth) &&
DL.fitsInLegalInteger(IT->getBitWidth());
}
Expand Down
64 changes: 64 additions & 0 deletions llvm/test/Transforms/SimplifyCFG/fold-i1.ll
View file
Open in desktop
Original file line number Diff line number Diff line change
@@ -0,0 +1,64 @@
; NOTE: Assertions have been autogenerated by utils/update_test_checks.py UTC_ARGS: --version 4
; RUN: opt < %s -S -passes='simplifycfg<switch-to-lookup>' | FileCheck %s

target datalayout = "e-m:e-p270:32:32-p271:32:32-p272:64:64-i64:64-f80:128-n8:16:32:64-S128"

define { i1, i8 } @f(i8 noundef %character) unnamed_addr {
; CHECK-LABEL: define { i1, i8 } @f(
; CHECK-SAME: i8 noundef [[CHARACTER:%.*]]) unnamed_addr {
; CHECK-NEXT: start:
; CHECK-NEXT: [[SWITCH_TABLEIDX:%.*]] = sub i8 [[CHARACTER]], 48
; CHECK-NEXT: [[TMP2:%.*]] = icmp ult i8 [[SWITCH_TABLEIDX]], 9
; CHECK-NEXT: [[_0_SROA_11_0:%.*]] = select i1 [[TMP2]], i8 [[SWITCH_TABLEIDX]], i8 undef
; CHECK-NEXT: [[_0_SROA_0_0:%.*]] = select i1 [[TMP2]], i1 true, i1 false
Copy link
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

It would be good to also test the case where we actually build a lookup table for the booleans. Need some pattern for that which doesn't reduce down to a simple select.

If I'm reading the code right, then what we'd end up doing is generate an array of i1 constants and then generate i1 loads -- the former is okay-ish, but the latter is a big no-no. We'd instead want to generate a table of i8s, generate i8 loads and truncate them to i1.

An alternative would be to not allow table generation, but instead relax the "AllTablesFitInRegister" condition so that we always have either a legal integer width or the lookup mask fitting in a register, without requiring that all of them fit in a register.

; CHECK-NEXT: [[TMP0:%.*]] = insertvalue { i1, i8 } poison, i1 [[_0_SROA_0_0]], 0
; CHECK-NEXT: [[TMP1:%.*]] = insertvalue { i1, i8 } [[TMP0]], i8 [[_0_SROA_11_0]], 1
; CHECK-NEXT: ret { i1, i8 } [[TMP1]]
;
start:
switch i8 %character, label %bb11 [
i8 48, label %bb2
i8 49, label %bb3
i8 50, label %bb4
i8 51, label %bb5
i8 52, label %bb6
i8 53, label %bb7
i8 54, label %bb8
i8 55, label %bb9
i8 56, label %bb10
]

bb2:
br label %bb11

bb3:
br label %bb11

bb4:
br label %bb11

bb5:
br label %bb11

bb6:
br label %bb11

bb7:
br label %bb11

bb8:
br label %bb11

bb9:
br label %bb11

bb10:
br label %bb11

bb11:
%_0.sroa.11.0 = phi i8 [ 8, %bb10 ], [ 7, %bb9 ], [ 6, %bb8 ], [ 5, %bb7 ], [ 4, %bb6 ], [ 3, %bb5 ], [ 2, %bb4 ], [ 1, %bb3 ], [ 0, %bb2 ], [ undef, %start ]
%_0.sroa.0.0 = phi i1 [ true, %bb10 ], [ true, %bb9 ], [ true, %bb8 ], [ true, %bb7 ], [ true, %bb6 ], [ true, %bb5 ], [ true, %bb4 ], [ true, %bb3 ], [ true, %bb2 ], [ false, %start ]
%0 = insertvalue { i1, i8 } poison, i1 %_0.sroa.0.0, 0
%1 = insertvalue { i1, i8 } %0, i8 %_0.sroa.11.0, 1
ret { i1, i8 } %1
}
Loading