c58d4c4bd3
Currently, the IROutliner uses a simple metric to outline the largest amount of IR possible to outline first if it fits the cost model. This is model loses out on smaller blocks of code that have higher reductions in cost that are contained within larger blocks of IR. This reverses the order, where we calculate all of the costs first, and then reorder and extract items based on the calculated results. Reviewers: paquette Differential Revision: https://reviews.llvm.org/D106440
133 lines
5.0 KiB
LLVM
133 lines
5.0 KiB
LLVM
; NOTE: Assertions have been autogenerated by utils/update_test_checks.py UTC_ARGS: --include-generated-funcs
|
|
; RUN: opt -S -verify -iroutliner -ir-outlining-no-cost < %s | FileCheck %s
|
|
|
|
; This test ensures that an extra output is not added when there is a bitcast
|
|
; that is relocated to outside of the extraction due to a starting lifetime
|
|
; instruction outside of the extracted region.
|
|
|
|
; Additionally, we check that the newly added bitcast instruction is excluded in
|
|
; further extractions.
|
|
|
|
declare void @llvm.lifetime.start.p0i8(i64, i8* nocapture)
|
|
declare void @llvm.lifetime.end.p0i8(i64, i8* nocapture)
|
|
|
|
define void @outline_bitcast_base() {
|
|
entry:
|
|
%a = alloca i32, align 4
|
|
%b = alloca i32, align 4
|
|
%c = alloca i32, align 4
|
|
%d = alloca i32, align 4
|
|
store i32 2, i32* %a, align 4
|
|
store i32 3, i32* %b, align 4
|
|
store i32 4, i32* %c, align 4
|
|
%X = bitcast i32* %d to i8*
|
|
%al = load i32, i32* %a
|
|
%bl = load i32, i32* %b
|
|
%cl = load i32, i32* %c
|
|
ret void
|
|
}
|
|
|
|
define void @outline_bitcast_removed() {
|
|
entry:
|
|
%a = alloca i32, align 4
|
|
%b = alloca i32, align 4
|
|
%c = alloca i32, align 4
|
|
%d = alloca i32, align 4
|
|
store i32 2, i32* %a, align 4
|
|
store i32 3, i32* %b, align 4
|
|
store i32 4, i32* %c, align 4
|
|
%X = bitcast i32* %d to i8*
|
|
%al = load i32, i32* %a
|
|
%bl = load i32, i32* %b
|
|
%cl = load i32, i32* %c
|
|
call void @llvm.lifetime.start.p0i8(i64 -1, i8* %X)
|
|
%am = load i32, i32* %b
|
|
%bm = load i32, i32* %a
|
|
%cm = load i32, i32* %c
|
|
call void @llvm.lifetime.end.p0i8(i64 -1, i8* %X)
|
|
ret void
|
|
}
|
|
|
|
; The first bitcast is moved down to lifetime start, and, since the original
|
|
; endpoint does not match the new endpoint, we cannot extract and outline the
|
|
; second bitcast and set of adds. Outlining only occurs in this case due to
|
|
; the lack of a cost model, as denoted by the debug command line argument.
|
|
|
|
define void @outline_bitcast_base2(i32 %a, i32 %b, i32 %c) {
|
|
entry:
|
|
%d = alloca i32, align 4
|
|
%X = bitcast i32* %d to i8*
|
|
%al = add i32 %a, %b
|
|
%bl = add i32 %b, %a
|
|
%cl = add i32 %b, %c
|
|
%buffer = mul i32 %a, %b
|
|
%Y = bitcast i32* %d to i8*
|
|
%am = add i32 %a, %b
|
|
%bm = add i32 %b, %a
|
|
%cm = add i32 %b, %c
|
|
call void @llvm.lifetime.start.p0i8(i64 -1, i8* %X)
|
|
call void @llvm.lifetime.end.p0i8(i64 -1, i8* %X)
|
|
ret void
|
|
}
|
|
|
|
; CHECK-LABEL: @outline_bitcast_base(
|
|
; CHECK-NEXT: entry:
|
|
; CHECK-NEXT: [[A:%.*]] = alloca i32, align 4
|
|
; CHECK-NEXT: [[B:%.*]] = alloca i32, align 4
|
|
; CHECK-NEXT: [[C:%.*]] = alloca i32, align 4
|
|
; CHECK-NEXT: [[D:%.*]] = alloca i32, align 4
|
|
; CHECK-NEXT: call void @outlined_ir_func_0(i32* [[A]], i32* [[B]], i32* [[C]], i32* [[D]])
|
|
; CHECK-NEXT: ret void
|
|
;
|
|
;
|
|
; CHECK-LABEL: @outline_bitcast_removed(
|
|
; CHECK-NEXT: entry:
|
|
; CHECK-NEXT: [[A:%.*]] = alloca i32, align 4
|
|
; CHECK-NEXT: [[B:%.*]] = alloca i32, align 4
|
|
; CHECK-NEXT: [[C:%.*]] = alloca i32, align 4
|
|
; CHECK-NEXT: [[D:%.*]] = alloca i32, align 4
|
|
; CHECK-NEXT: call void @outlined_ir_func_0(i32* [[A]], i32* [[B]], i32* [[C]], i32* [[D]])
|
|
; CHECK-NEXT: [[LT_CAST1:%.*]] = bitcast i32* [[D]] to i8*
|
|
; CHECK-NEXT: call void @llvm.lifetime.start.p0i8(i64 -1, i8* [[LT_CAST1]])
|
|
; CHECK-NEXT: [[AM:%.*]] = load i32, i32* [[B]], align 4
|
|
; CHECK-NEXT: [[BM:%.*]] = load i32, i32* [[A]], align 4
|
|
; CHECK-NEXT: [[CM:%.*]] = load i32, i32* [[C]], align 4
|
|
; CHECK-NEXT: [[LT_CAST:%.*]] = bitcast i32* [[D]] to i8*
|
|
; CHECK-NEXT: call void @llvm.lifetime.end.p0i8(i64 -1, i8* [[LT_CAST]])
|
|
; CHECK-NEXT: ret void
|
|
;
|
|
;
|
|
; CHECK-LABEL: @outline_bitcast_base2(
|
|
; CHECK-NEXT: entry:
|
|
; CHECK-NEXT: [[D:%.*]] = alloca i32, align 4
|
|
; CHECK-NEXT: [[X:%.*]] = bitcast i32* [[D]] to i8*
|
|
; CHECK-NEXT: [[AL:%.*]] = add i32 [[A:%.*]], [[B:%.*]]
|
|
; CHECK-NEXT: [[BL:%.*]] = add i32 [[B]], [[A]]
|
|
; CHECK-NEXT: [[CL:%.*]] = add i32 [[B]], [[C:%.*]]
|
|
; CHECK-NEXT: [[BUFFER:%.*]] = mul i32 [[A]], [[B]]
|
|
; CHECK-NEXT: [[Y:%.*]] = bitcast i32* [[D]] to i8*
|
|
; CHECK-NEXT: [[AM:%.*]] = add i32 [[A]], [[B]]
|
|
; CHECK-NEXT: [[BM:%.*]] = add i32 [[B]], [[A]]
|
|
; CHECK-NEXT: [[CM:%.*]] = add i32 [[B]], [[C]]
|
|
; CHECK-NEXT: [[LT_CAST1:%.*]] = bitcast i32* [[D]] to i8*
|
|
; CHECK-NEXT: call void @llvm.lifetime.start.p0i8(i64 -1, i8* [[LT_CAST1]])
|
|
; CHECK-NEXT: [[LT_CAST:%.*]] = bitcast i32* [[D]] to i8*
|
|
; CHECK-NEXT: call void @llvm.lifetime.end.p0i8(i64 -1, i8* [[LT_CAST]])
|
|
; CHECK-NEXT: ret void
|
|
;
|
|
;
|
|
; CHECK-LABEL: @outlined_ir_func_0(
|
|
; CHECK-NEXT: newFuncRoot:
|
|
; CHECK-NEXT: br label [[ENTRY_TO_OUTLINE:%.*]]
|
|
; CHECK: entry_to_outline:
|
|
; CHECK-NEXT: store i32 2, i32* [[TMP0:%.*]], align 4
|
|
; CHECK-NEXT: store i32 3, i32* [[TMP1:%.*]], align 4
|
|
; CHECK-NEXT: store i32 4, i32* [[TMP2:%.*]], align 4
|
|
; CHECK-NEXT: [[X:%.*]] = bitcast i32* [[TMP3:%.*]] to i8*
|
|
; CHECK-NEXT: [[AL:%.*]] = load i32, i32* [[TMP0]], align 4
|
|
; CHECK-NEXT: [[BL:%.*]] = load i32, i32* [[TMP1]], align 4
|
|
; CHECK-NEXT: [[CL:%.*]] = load i32, i32* [[TMP2]], align 4
|
|
; CHECK-NEXT: br label [[ENTRY_AFTER_OUTLINE_EXITSTUB:%.*]]
|
|
; CHECK: entry_after_outline.exitStub:
|
|
; CHECK-NEXT: ret void
|