90ba33099c
This patch canonicalizes getelementptr instructions with constant indices to use the `i8` source element type. This makes it easier for optimizations to recognize that two GEPs are identical, because they don't need to see past many different ways to express the same offset. This is a first step towards https://discourse.llvm.org/t/rfc-replacing-getelementptr-with-ptradd/68699. This is limited to constant GEPs only for now, as they have a clear canonical form, while we're not yet sure how exactly to deal with variable indices. The test llvm/test/Transforms/PhaseOrdering/switch_with_geps.ll gives two representative examples of the kind of optimization improvement we expect from this change. In the first test SimplifyCFG can now realize that all switch branches are actually the same. In the second test it can convert it into simple arithmetic. These are representative of common optimization failures we see in Rust. Fixes https://github.com/llvm/llvm-project/issues/69841.
103 lines
3.8 KiB
LLVM
103 lines
3.8 KiB
LLVM
; NOTE: Assertions have been autogenerated by utils/update_test_checks.py
|
|
; RUN: opt < %s -passes=instcombine -S | FileCheck %s
|
|
;
|
|
; Exercise folding of memcmp calls with constant arrays including both
|
|
; negative and positive characters and both constant and nonconstant sizes.
|
|
|
|
declare i32 @memcmp(ptr, ptr, i64)
|
|
|
|
@a = constant [7 x i8] c"abcdef\7f"
|
|
@b = constant [7 x i8] c"abcdef\80"
|
|
|
|
|
|
; Exercise memcmp(A + C, B + C, 2) folding of small arrays that differ in
|
|
; a character with the opposite sign and a constant size.
|
|
|
|
define void @fold_memcmp_cst_cst(ptr %pcmp) {
|
|
; CHECK-LABEL: @fold_memcmp_cst_cst(
|
|
; CHECK-NEXT: store i32 -1, ptr [[PCMP:%.*]], align 4
|
|
; CHECK-NEXT: [[SB5_A5:%.*]] = getelementptr i8, ptr [[PCMP]], i64 4
|
|
; CHECK-NEXT: store i32 1, ptr [[SB5_A5]], align 4
|
|
; CHECK-NEXT: [[SA6_B6:%.*]] = getelementptr i8, ptr [[PCMP]], i64 8
|
|
; CHECK-NEXT: store i32 -1, ptr [[SA6_B6]], align 4
|
|
; CHECK-NEXT: [[SB6_A6:%.*]] = getelementptr i8, ptr [[PCMP]], i64 12
|
|
; CHECK-NEXT: store i32 1, ptr [[SB6_A6]], align 4
|
|
; CHECK-NEXT: ret void
|
|
;
|
|
%p5 = getelementptr [7 x i8], ptr @a, i64 0, i64 5
|
|
%p6 = getelementptr [7 x i8], ptr @a, i64 0, i64 6
|
|
|
|
%q5 = getelementptr [7 x i8], ptr @b, i64 0, i64 5
|
|
%q6 = getelementptr [7 x i8], ptr @b, i64 0, i64 6
|
|
|
|
; Fold memcmp(a + 5, b + 5, 2) to -1.
|
|
%ca5_b5 = call i32 @memcmp(ptr %p5, ptr %q5, i64 2)
|
|
store i32 %ca5_b5, ptr %pcmp
|
|
|
|
; Fold memcmp(b + 5, a + 5, 2) to +1.
|
|
%cb5_a5 = call i32 @memcmp(ptr %q5, ptr %p5, i64 2)
|
|
%sb5_a5 = getelementptr i32, ptr %pcmp, i64 1
|
|
store i32 %cb5_a5, ptr %sb5_a5
|
|
|
|
; Fold memcmp(a + 6, b + 6, 1) to -1.
|
|
%ca6_b6 = call i32 @memcmp(ptr %p6, ptr %q6, i64 1)
|
|
%sa6_b6 = getelementptr i32, ptr %pcmp, i64 2
|
|
store i32 %ca6_b6, ptr %sa6_b6
|
|
|
|
; Fold memcmp(b + 6, a + 6, 1) to +1.
|
|
%cb6_a6 = call i32 @memcmp(ptr %q6, ptr %p6, i64 1)
|
|
%sb6_a6 = getelementptr i32, ptr %pcmp, i64 3
|
|
store i32 %cb6_a6, ptr %sb6_a6
|
|
|
|
ret void
|
|
}
|
|
|
|
|
|
; Exercise memcmp(A, B, N) folding of arrays that differ in a character
|
|
; with the opposite sign and a variable size
|
|
|
|
define void @fold_memcmp_cst_var(ptr %pcmp, i64 %n) {
|
|
; CHECK-LABEL: @fold_memcmp_cst_var(
|
|
; CHECK-NEXT: [[TMP1:%.*]] = icmp ugt i64 [[N:%.*]], 6
|
|
; CHECK-NEXT: [[CA0_B0:%.*]] = sext i1 [[TMP1]] to i32
|
|
; CHECK-NEXT: store i32 [[CA0_B0]], ptr [[PCMP:%.*]], align 4
|
|
; CHECK-NEXT: [[TMP2:%.*]] = icmp ugt i64 [[N]], 6
|
|
; CHECK-NEXT: [[CB0_A0:%.*]] = zext i1 [[TMP2]] to i32
|
|
; CHECK-NEXT: [[SB0_A0:%.*]] = getelementptr i8, ptr [[PCMP]], i64 4
|
|
; CHECK-NEXT: store i32 [[CB0_A0]], ptr [[SB0_A0]], align 4
|
|
; CHECK-NEXT: [[TMP3:%.*]] = icmp ne i64 [[N]], 0
|
|
; CHECK-NEXT: [[CA6_B6:%.*]] = sext i1 [[TMP3]] to i32
|
|
; CHECK-NEXT: [[SA6_B6:%.*]] = getelementptr i8, ptr [[PCMP]], i64 8
|
|
; CHECK-NEXT: store i32 [[CA6_B6]], ptr [[SA6_B6]], align 4
|
|
; CHECK-NEXT: [[TMP4:%.*]] = icmp ne i64 [[N]], 0
|
|
; CHECK-NEXT: [[CB6_A6:%.*]] = zext i1 [[TMP4]] to i32
|
|
; CHECK-NEXT: [[SB6_A6:%.*]] = getelementptr i8, ptr [[PCMP]], i64 12
|
|
; CHECK-NEXT: store i32 [[CB6_A6]], ptr [[SB6_A6]], align 4
|
|
; CHECK-NEXT: ret void
|
|
;
|
|
%p6 = getelementptr [7 x i8], ptr @a, i64 0, i64 6
|
|
|
|
%q6 = getelementptr [7 x i8], ptr @b, i64 0, i64 6
|
|
|
|
; Fold memcmp(a, b, n) to -1.
|
|
%ca0_b0 = call i32 @memcmp(ptr @a, ptr @b, i64 %n)
|
|
store i32 %ca0_b0, ptr %pcmp
|
|
|
|
; Fold memcmp(b, a, n) to +1.
|
|
%cb0_a0 = call i32 @memcmp(ptr @b, ptr @a, i64 %n)
|
|
%sb0_a0 = getelementptr i32, ptr %pcmp, i64 1
|
|
store i32 %cb0_a0, ptr %sb0_a0
|
|
|
|
; Fold memcmp(a + 6, b + 6, n) to -1.
|
|
%ca6_b6 = call i32 @memcmp(ptr %p6, ptr %q6, i64 %n)
|
|
%sa6_b6 = getelementptr i32, ptr %pcmp, i64 2
|
|
store i32 %ca6_b6, ptr %sa6_b6
|
|
|
|
; Fold memcmp(b + 6, a + 6, n) to +1.
|
|
%cb6_a6 = call i32 @memcmp(ptr %q6, ptr %p6, i64 %n)
|
|
%sb6_a6 = getelementptr i32, ptr %pcmp, i64 3
|
|
store i32 %cb6_a6, ptr %sb6_a6
|
|
|
|
ret void
|
|
}
|