shylie/llvm-project
1
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity
llvm-project/llvm/test/Transforms/SeparateConstOffsetFromGEP/RISCV/split-gep.ll
Fabian Ritter 332f060363
[SeparateConstOffsetFromGEP] Don't set unsound inbounds flag (#130616) 
The language reference says about inbounds geps that "if the
getelementptr has any non-zero indices[...] [t]he base pointer has an in
bounds address of the allocated object that it is based on [and]
[d]uring the successive addition of offsets to the address, the
resulting pointer must remain in bounds of the allocated object at each
step."

If (gep inbounds p, (a + 5)) is translated to (gep [inbounds] (gep p,
a), 5) with p pointing to the beginning of an object and a=-4, as the
example in the comments suggests, that's the case for neither of the
resulting geps. Therefore, we need to clear the inbounds flag for both
geps.

We might want to use ValueTracking to check if a is known to be
non-negative to preserve the inbounds flags.

For the AMDGPU tests with scratch instructions, removing the unsound
inbounds flag means that AMDGPUDAGToDAGISel::isFlatScratchBaseLegal sees
no NUW flag at the pointer add, which prevents generation of scratch
instructions with immediate offsets.

For SWDEV-516125.
2025-03-18 12:30:20 +01:00

292 lines
12 KiB
LLVM
Raw Blame History

; NOTE: Assertions have been autogenerated by utils/update_test_checks.py
; RUN: opt < %s -mtriple=riscv64-unknown-elf -passes='separate-const-offset-from-gep,early-cse' \
; RUN: -S | FileCheck %s
; Several tests for separate-const-offset-from-gep. The transformation
; heavily relies on TargetTransformInfo, so we put these tests under
; target-specific folders.
; Simple case when GEPs should be optimized.
define i64 @test1(ptr %array, i64 %i, i64 %j) {
; CHECK-LABEL: @test1(
; CHECK-NEXT: entry:
; CHECK-NEXT: [[ADD:%.*]] = add nsw i64 [[I:%.*]], 5
; CHECK-NEXT: [[TMP0:%.*]] = getelementptr i64, ptr [[ARRAY:%.*]], i64 [[I]]
; CHECK-NEXT: [[GEP4:%.*]] = getelementptr i8, ptr [[TMP0]], i64 40
; CHECK-NEXT: store i64 [[J:%.*]], ptr [[GEP4]], align 8
; CHECK-NEXT: [[GEP26:%.*]] = getelementptr i8, ptr [[TMP0]], i64 48
; CHECK-NEXT: store i64 [[J]], ptr [[GEP26]], align 8
; CHECK-NEXT: [[GEP38:%.*]] = getelementptr i8, ptr [[TMP0]], i64 280
; CHECK-NEXT: store i64 [[ADD]], ptr [[GEP38]], align 8
; CHECK-NEXT: ret i64 undef
;
entry:
%add = add nsw i64 %i, 5
%gep = getelementptr inbounds i64, ptr %array, i64 %add
store i64 %j, ptr %gep
%add2 = add nsw i64 %i, 6
%gep2 = getelementptr inbounds i64, ptr %array, i64 %add2
store i64 %j, ptr %gep2
%add3 = add nsw i64 %i, 35
%gep3 = getelementptr inbounds i64, ptr %array, i64 %add3
store i64 %add, ptr %gep3
ret i64 undef
}
; Optimize GEPs when there sext instructions are needed to cast index value to expected type.
define i32 @test2(ptr %array, i32 %i, i32 %j) {
; CHECK-LABEL: @test2(
; CHECK-NEXT: entry:
; CHECK-NEXT: [[ADD:%.*]] = add nsw i32 [[I:%.*]], 5
; CHECK-NEXT: [[TMP0:%.*]] = sext i32 [[I]] to i64
; CHECK-NEXT: [[TMP1:%.*]] = getelementptr i32, ptr [[ARRAY:%.*]], i64 [[TMP0]]
; CHECK-NEXT: [[GEP2:%.*]] = getelementptr i8, ptr [[TMP1]], i64 20
; CHECK-NEXT: store i32 [[J:%.*]], ptr [[GEP2]], align 4
; CHECK-NEXT: [[GEP54:%.*]] = getelementptr i8, ptr [[TMP1]], i64 24
; CHECK-NEXT: store i32 [[J]], ptr [[GEP54]], align 4
; CHECK-NEXT: [[GEP86:%.*]] = getelementptr i8, ptr [[TMP1]], i64 140
; CHECK-NEXT: store i32 [[ADD]], ptr [[GEP86]], align 4
; CHECK-NEXT: ret i32 undef
;
entry:
%add = add nsw i32 %i, 5
%sext = sext i32 %add to i64
%gep = getelementptr inbounds i32, ptr %array, i64 %sext
store i32 %j, ptr %gep
%add3 = add nsw i32 %i, 6
%sext4 = sext i32 %add3 to i64
%gep5 = getelementptr inbounds i32, ptr %array, i64 %sext4
store i32 %j, ptr %gep5
%add6 = add nsw i32 %i, 35
%sext7 = sext i32 %add6 to i64
%gep8 = getelementptr inbounds i32, ptr %array, i64 %sext7
store i32 %add, ptr %gep8
ret i32 undef
}
; No need to modify because all values are also used in other expressions.
; Modification doesn't decrease register pressure.
define i32 @test3(ptr %array, i32 %i) {
; CHECK-LABEL: @test3(
; CHECK-NEXT: entry:
; CHECK-NEXT: [[ADD:%.*]] = add nsw i32 [[I:%.*]], 5
; CHECK-NEXT: [[TMP0:%.*]] = sext i32 [[I]] to i64
; CHECK-NEXT: [[TMP1:%.*]] = getelementptr i32, ptr [[ARRAY:%.*]], i64 [[TMP0]]
; CHECK-NEXT: [[GEP2:%.*]] = getelementptr i8, ptr [[TMP1]], i64 20
; CHECK-NEXT: store i32 [[ADD]], ptr [[GEP2]], align 4
; CHECK-NEXT: [[ADD3:%.*]] = add nsw i32 [[I]], 6
; CHECK-NEXT: [[GEP54:%.*]] = getelementptr i8, ptr [[TMP1]], i64 24
; CHECK-NEXT: store i32 [[ADD3]], ptr [[GEP54]], align 4
; CHECK-NEXT: [[ADD6:%.*]] = add nsw i32 [[I]], 35
; CHECK-NEXT: [[GEP86:%.*]] = getelementptr i8, ptr [[TMP1]], i64 140
; CHECK-NEXT: store i32 [[ADD6]], ptr [[GEP86]], align 4
; CHECK-NEXT: ret i32 undef
;
entry:
%add = add nsw i32 %i, 5
%sext = sext i32 %add to i64
%gep = getelementptr inbounds i32, ptr %array, i64 %sext
store i32 %add, ptr %gep
%add3 = add nsw i32 %i, 6
%sext4 = sext i32 %add3 to i64
%gep5 = getelementptr inbounds i32, ptr %array, i64 %sext4
store i32 %add3, ptr %gep5
%add6 = add nsw i32 %i, 35
%sext7 = sext i32 %add6 to i64
%gep8 = getelementptr inbounds i32, ptr %array, i64 %sext7
store i32 %add6, ptr %gep8
ret i32 undef
}
; Optimized GEPs for multidimensional array with same base
define i32 @test4(ptr %array2, i32 %i) {
; CHECK-LABEL: @test4(
; CHECK-NEXT: entry:
; CHECK-NEXT: [[ADD:%.*]] = add nsw i32 [[I:%.*]], 5
; CHECK-NEXT: [[TMP0:%.*]] = sext i32 [[I]] to i64
; CHECK-NEXT: [[TMP1:%.*]] = getelementptr [50 x i32], ptr [[ARRAY2:%.*]], i64 [[TMP0]], i64 [[TMP0]]
; CHECK-NEXT: [[GEP3:%.*]] = getelementptr i8, ptr [[TMP1]], i64 1020
; CHECK-NEXT: store i32 [[I]], ptr [[GEP3]], align 4
; CHECK-NEXT: [[GEP56:%.*]] = getelementptr i8, ptr [[TMP1]], i64 1024
; CHECK-NEXT: store i32 [[ADD]], ptr [[GEP56]], align 4
; CHECK-NEXT: [[GEP89:%.*]] = getelementptr i8, ptr [[TMP1]], i64 1140
; CHECK-NEXT: store i32 [[I]], ptr [[GEP89]], align 4
; CHECK-NEXT: ret i32 undef
;
entry:
%add = add nsw i32 %i, 5
%sext = sext i32 %add to i64
%gep = getelementptr inbounds [50 x i32], ptr %array2, i64 %sext, i64 %sext
store i32 %i, ptr %gep
%add3 = add nsw i32 %i, 6
%sext4 = sext i32 %add3 to i64
%gep5 = getelementptr inbounds [50 x i32], ptr %array2, i64 %sext, i64 %sext4
store i32 %add, ptr %gep5
%add6 = add nsw i32 %i, 35
%sext7 = sext i32 %add6 to i64
%gep8 = getelementptr inbounds [50 x i32], ptr %array2, i64 %sext, i64 %sext7
store i32 %i, ptr %gep8
ret i32 undef
}
; Don't optimize GEPs for multidimensional array with same base because RISC-V doesn't support the addressing mode
define i32 @test5(ptr %array2, i32 %i, i64 %j) {
; CHECK-LABEL: @test5(
; CHECK-NEXT: entry:
; CHECK-NEXT: [[ADD:%.*]] = add nsw i32 [[I:%.*]], 5
; CHECK-NEXT: [[TMP0:%.*]] = sext i32 [[I]] to i64
; CHECK-NEXT: [[TMP1:%.*]] = getelementptr [50 x i32], ptr [[ARRAY2:%.*]], i64 [[TMP0]], i64 [[TMP0]]
; CHECK-NEXT: [[GEP3:%.*]] = getelementptr i8, ptr [[TMP1]], i64 1020
; CHECK-NEXT: store i32 [[ADD]], ptr [[GEP3]], align 4
; CHECK-NEXT: [[TMP2:%.*]] = getelementptr [50 x i32], ptr [[ARRAY2]], i64 [[TMP0]], i64 [[J:%.*]]
; CHECK-NEXT: [[GEP55:%.*]] = getelementptr i8, ptr [[TMP2]], i64 1200
; CHECK-NEXT: store i32 [[I]], ptr [[GEP55]], align 4
; CHECK-NEXT: [[ADD6:%.*]] = add nsw i32 [[I]], 35
; CHECK-NEXT: [[SEXT7:%.*]] = sext i32 [[ADD6]] to i64
; CHECK-NEXT: [[GEP8:%.*]] = getelementptr inbounds [50 x i32], ptr [[ARRAY2]], i64 [[SEXT7]], i64 [[J]]
; CHECK-NEXT: store i32 [[I]], ptr [[GEP8]], align 4
; CHECK-NEXT: ret i32 undef
;
entry:
%add = add nsw i32 %i, 5
%sext = sext i32 %add to i64
%gep = getelementptr inbounds [50 x i32], ptr %array2, i64 %sext, i64 %sext
store i32 %add, ptr %gep
%add3 = add nsw i32 %i, 6
%sext4 = sext i32 %add3 to i64
%gep5 = getelementptr inbounds [50 x i32], ptr %array2, i64 %sext4, i64 %j
store i32 %i, ptr %gep5
%add6 = add nsw i32 %i, 35
%sext7 = sext i32 %add6 to i64
%gep8 = getelementptr inbounds [50 x i32], ptr %array2, i64 %sext7, i64 %j
store i32 %i, ptr %gep8
ret i32 undef
}
; No need to optimize GEPs, because there is critical amount with non-constant offsets.
define i64 @test6(ptr %array, i64 %i, i64 %j) {
; CHECK-LABEL: @test6(
; CHECK-NEXT: entry:
; CHECK-NEXT: [[ADD:%.*]] = add nsw i64 [[I:%.*]], 5
; CHECK-NEXT: [[GEP:%.*]] = getelementptr inbounds i64, ptr [[ARRAY:%.*]], i64 [[J:%.*]]
; CHECK-NEXT: store i64 [[ADD]], ptr [[GEP]], align 8
; CHECK-NEXT: [[TMP0:%.*]] = getelementptr i64, ptr [[ARRAY]], i64 [[I]]
; CHECK-NEXT: [[GEP52:%.*]] = getelementptr i8, ptr [[TMP0]], i64 48
; CHECK-NEXT: store i64 [[I]], ptr [[GEP52]], align 8
; CHECK-NEXT: store i64 [[I]], ptr [[TMP0]], align 8
; CHECK-NEXT: ret i64 undef
;
entry:
%add = add nsw i64 %i, 5
%gep = getelementptr inbounds i64, ptr %array, i64 %j
store i64 %add, ptr %gep
%add3 = add nsw i64 %i, 6
%gep5 = getelementptr inbounds i64, ptr %array, i64 %add3
store i64 %i, ptr %gep5
%add6 = add nsw i64 %i, 35
%gep8 = getelementptr inbounds i64, ptr %array, i64 %i
store i64 %i, ptr %gep8
ret i64 undef
}
; No need to optimize GEPs, because the base variable is different.
define i32 @test7(ptr %array, i32 %i, i32 %j, i32 %k) {
; CHECK-LABEL: @test7(
; CHECK-NEXT: entry:
; CHECK-NEXT: [[ADD:%.*]] = add nsw i32 [[I:%.*]], 5
; CHECK-NEXT: [[TMP0:%.*]] = sext i32 [[I]] to i64
; CHECK-NEXT: [[TMP1:%.*]] = getelementptr i32, ptr [[ARRAY:%.*]], i64 [[TMP0]]
; CHECK-NEXT: [[GEP2:%.*]] = getelementptr i8, ptr [[TMP1]], i64 20
; CHECK-NEXT: store i32 [[ADD]], ptr [[GEP2]], align 4
; CHECK-NEXT: [[TMP2:%.*]] = sext i32 [[K:%.*]] to i64
; CHECK-NEXT: [[TMP3:%.*]] = getelementptr i32, ptr [[ARRAY]], i64 [[TMP2]]
; CHECK-NEXT: [[GEP54:%.*]] = getelementptr i8, ptr [[TMP3]], i64 24
; CHECK-NEXT: store i32 [[I]], ptr [[GEP54]], align 4
; CHECK-NEXT: [[TMP4:%.*]] = sext i32 [[J:%.*]] to i64
; CHECK-NEXT: [[TMP5:%.*]] = getelementptr i32, ptr [[ARRAY]], i64 [[TMP4]]
; CHECK-NEXT: [[GEP86:%.*]] = getelementptr i8, ptr [[TMP5]], i64 140
; CHECK-NEXT: store i32 [[I]], ptr [[GEP86]], align 4
; CHECK-NEXT: ret i32 undef
;
entry:
%add = add nsw i32 %i, 5
%sext = sext i32 %add to i64
%gep = getelementptr inbounds i32, ptr %array, i64 %sext
store i32 %add, ptr %gep
%add3 = add nsw i32 %k, 6
%sext4 = sext i32 %add3 to i64
%gep5 = getelementptr inbounds i32, ptr %array, i64 %sext4
store i32 %i, ptr %gep5
%add6 = add nsw i32 %j, 35
%sext7 = sext i32 %add6 to i64
%gep8 = getelementptr inbounds i32, ptr %array, i64 %sext7
store i32 %i, ptr %gep8
ret i32 undef
}
; No need to optimize GEPs, because the base of GEP instructions is different.
define i32 @test8(ptr %array, ptr %array2, ptr %array3, i32 %i) {
; CHECK-LABEL: @test8(
; CHECK-NEXT: entry:
; CHECK-NEXT: [[ADD:%.*]] = add nsw i32 [[I:%.*]], 5
; CHECK-NEXT: [[TMP0:%.*]] = sext i32 [[I]] to i64
; CHECK-NEXT: [[TMP1:%.*]] = getelementptr i32, ptr [[ARRAY:%.*]], i64 [[TMP0]]
; CHECK-NEXT: [[GEP2:%.*]] = getelementptr i8, ptr [[TMP1]], i64 20
; CHECK-NEXT: store i32 [[ADD]], ptr [[GEP2]], align 4
; CHECK-NEXT: [[TMP2:%.*]] = getelementptr i32, ptr [[ARRAY2:%.*]], i64 [[TMP0]]
; CHECK-NEXT: [[GEP54:%.*]] = getelementptr i8, ptr [[TMP2]], i64 24
; CHECK-NEXT: store i32 [[I]], ptr [[GEP54]], align 4
; CHECK-NEXT: [[TMP3:%.*]] = getelementptr i32, ptr [[ARRAY3:%.*]], i64 [[TMP0]]
; CHECK-NEXT: [[GEP86:%.*]] = getelementptr i8, ptr [[TMP3]], i64 140
; CHECK-NEXT: store i32 [[I]], ptr [[GEP86]], align 4
; CHECK-NEXT: ret i32 undef
;
entry:
%add = add nsw i32 %i, 5
%sext = sext i32 %add to i64
%gep = getelementptr inbounds i32, ptr %array, i64 %sext
store i32 %add, ptr %gep
%add3 = add nsw i32 %i, 6
%sext4 = sext i32 %add3 to i64
%gep5 = getelementptr inbounds i32, ptr %array2, i64 %sext4
store i32 %i, ptr %gep5
%add6 = add nsw i32 %i, 35
%sext7 = sext i32 %add6 to i64
%gep8 = getelementptr inbounds i32, ptr %array3, i64 %sext7
store i32 %i, ptr %gep8
ret i32 undef
}
; No need to optimize GEPs of multidimensional array, because the base of GEP instructions is different.
define i32 @test9(ptr %array, i32 %i) {
; CHECK-LABEL: @test9(
; CHECK-NEXT: entry:
; CHECK-NEXT: [[ADD:%.*]] = add nsw i32 [[I:%.*]], 5
; CHECK-NEXT: [[TMP0:%.*]] = sext i32 [[I]] to i64
; CHECK-NEXT: [[TMP1:%.*]] = getelementptr [50 x i32], ptr [[ARRAY:%.*]], i64 0, i64 [[TMP0]]
; CHECK-NEXT: [[GEP2:%.*]] = getelementptr i8, ptr [[TMP1]], i64 20
; CHECK-NEXT: store i32 [[ADD]], ptr [[GEP2]], align 4
; CHECK-NEXT: [[TMP2:%.*]] = getelementptr [50 x i32], ptr [[ARRAY]], i64 [[TMP0]], i64 [[TMP0]]
; CHECK-NEXT: [[GEP54:%.*]] = getelementptr i8, ptr [[TMP2]], i64 24
; CHECK-NEXT: store i32 [[I]], ptr [[GEP54]], align 4
; CHECK-NEXT: [[GEP87:%.*]] = getelementptr i8, ptr [[TMP2]], i64 1340
; CHECK-NEXT: store i32 [[I]], ptr [[GEP87]], align 4
; CHECK-NEXT: ret i32 undef
;
entry:
%add = add nsw i32 %i, 5
%sext = sext i32 %add to i64
%gep = getelementptr inbounds [50 x i32], ptr %array, i64 0, i64 %sext
store i32 %add, ptr %gep
%add3 = add nsw i32 %i, 6
%sext4 = sext i32 %add3 to i64
%int = sext i32 %i to i64
%gep5 = getelementptr inbounds [50 x i32], ptr %array, i64 %int, i64 %sext4
store i32 %i, ptr %gep5
%add6 = add nsw i32 %i, 35
%sext7 = sext i32 %add6 to i64
%gep8 = getelementptr inbounds [50 x i32], ptr %array, i64 %sext4, i64 %sext7
store i32 %i, ptr %gep8
ret i32 undef
}
Reference in New Issue View Git Blame Copy Permalink