mirror of
https://github.com/RPCS3/llvm-mirror.git
synced 2024-11-22 18:54:02 +01:00
4fd7f95ec0
Currently the SCEVExpander tries to re-use existing casts, even if they are not exactly at the insertion point it was asked to create the cast. To do so in some case, it creates a new cast at the insertion point and updates all users to use the new cast. This behavior is problematic, because it changes the IR outside of the instructions created during the expansion. Therefore we cannot completely undo all changes made during expansion. This re-use should be only an extra optimization, so only using the new cast in the expanded instructions should not be a correctness issue. There are many cases equivalent instructions are created during expansion. This patch also adjusts findInsertPointAfter to skip instructions inserted during expansion. This enables re-using existing casts without the renaming any uses, by picking a better insertion point. Reviewed By: efriedma, lebedev.ri Differential Revision: https://reviews.llvm.org/D84399
133 lines
6.1 KiB
LLVM
133 lines
6.1 KiB
LLVM
; NOTE: Assertions have been autogenerated by utils/update_test_checks.py
|
|
; RUN: opt -loop-idiom -S %s | FileCheck %s
|
|
|
|
define void @reuse_cast_1(float** %ptr, i1 %c) {
|
|
; CHECK-LABEL: @reuse_cast_1(
|
|
; CHECK-NEXT: entry:
|
|
; CHECK-NEXT: br label [[LOOP_0:%.*]]
|
|
; CHECK: loop.0:
|
|
; CHECK-NEXT: [[TMP:%.*]] = load float*, float** [[PTR:%.*]], align 8
|
|
; CHECK-NEXT: [[TMP2:%.*]] = bitcast float* [[TMP]] to i8*
|
|
; CHECK-NEXT: br i1 [[C:%.*]], label [[LOOP_2_PREHEADER:%.*]], label [[LOOP_1_PREHEADER:%.*]]
|
|
; CHECK: loop.1.preheader:
|
|
; CHECK-NEXT: call void @llvm.memset.p0i8.i64(i8* align 4 [[TMP2]], i8 0, i64 400, i1 false)
|
|
; CHECK-NEXT: br label [[LOOP_1:%.*]]
|
|
; CHECK: loop.2.preheader:
|
|
; CHECK-NEXT: br label [[LOOP_2:%.*]]
|
|
; CHECK: loop.1:
|
|
; CHECK-NEXT: [[IV_1:%.*]] = phi i64 [ [[IV_1_NEXT:%.*]], [[LOOP_1]] ], [ 0, [[LOOP_1_PREHEADER]] ]
|
|
; CHECK-NEXT: [[TMP4:%.*]] = getelementptr inbounds float, float* [[TMP]], i64 [[IV_1]]
|
|
; CHECK-NEXT: [[IV_1_NEXT]] = add nuw nsw i64 [[IV_1]], 1
|
|
; CHECK-NEXT: [[C_1:%.*]] = icmp ne i64 [[IV_1_NEXT]], 100
|
|
; CHECK-NEXT: br i1 [[C_1]], label [[LOOP_1]], label [[LOOP_0_LATCH_LOOPEXIT1:%.*]]
|
|
; CHECK: loop.2:
|
|
; CHECK-NEXT: [[IV_2:%.*]] = phi i64 [ [[IV_2_NEXT:%.*]], [[LOOP_2]] ], [ 0, [[LOOP_2_PREHEADER]] ]
|
|
; CHECK-NEXT: [[TMP10:%.*]] = getelementptr inbounds float, float* [[TMP]], i64 [[IV_2]]
|
|
; CHECK-NEXT: store float 0.000000e+00, float* [[TMP10]], align 4
|
|
; CHECK-NEXT: [[TMP11:%.*]] = load float*, float** [[PTR]], align 8
|
|
; CHECK-NEXT: [[TMP12:%.*]] = getelementptr inbounds float, float* [[TMP11]], i64 [[IV_2]]
|
|
; CHECK-NEXT: [[TMP13:%.*]] = load float, float* [[TMP12]], align 4
|
|
; CHECK-NEXT: [[IV_2_NEXT]] = add nuw nsw i64 [[IV_2]], 1
|
|
; CHECK-NEXT: [[C_2:%.*]] = icmp ne i64 [[IV_2_NEXT]], 100
|
|
; CHECK-NEXT: br i1 [[C_2]], label [[LOOP_2]], label [[LOOP_0_LATCH_LOOPEXIT:%.*]]
|
|
; CHECK: loop.0.latch.loopexit:
|
|
; CHECK-NEXT: br label [[LOOP_0_LATCH:%.*]]
|
|
; CHECK: loop.0.latch.loopexit1:
|
|
; CHECK-NEXT: br label [[LOOP_0_LATCH]]
|
|
; CHECK: loop.0.latch:
|
|
; CHECK-NEXT: [[C_0:%.*]] = call i1 @cond()
|
|
; CHECK-NEXT: br i1 [[C_0]], label [[LOOP_0]], label [[END:%.*]]
|
|
; CHECK: end:
|
|
; CHECK-NEXT: ret void
|
|
;
|
|
entry:
|
|
br label %loop.0
|
|
|
|
loop.0: ; preds = %loop.0.latch, %entry
|
|
%tmp = load float*, float** %ptr, align 8
|
|
br i1 %c, label %loop.2, label %loop.1
|
|
|
|
loop.1: ; preds = %loop.1, %loop.0
|
|
%iv.1 = phi i64 [ %iv.1.next, %loop.1 ], [ 0, %loop.0 ]
|
|
%tmp4 = getelementptr inbounds float, float* %tmp, i64 %iv.1
|
|
store float 0.000000e+00, float* %tmp4, align 4
|
|
%iv.1.next = add nuw nsw i64 %iv.1, 1
|
|
%c.1 = icmp ne i64 %iv.1.next, 100
|
|
br i1 %c.1, label %loop.1, label %loop.0.latch
|
|
|
|
loop.2: ; preds = %loop.2, %loop.0
|
|
%iv.2 = phi i64 [ %iv.2.next, %loop.2 ], [ 0, %loop.0 ]
|
|
%tmp10 = getelementptr inbounds float, float* %tmp, i64 %iv.2
|
|
store float 0.000000e+00, float* %tmp10, align 4
|
|
%tmp11 = load float*, float** %ptr, align 8
|
|
%tmp12 = getelementptr inbounds float, float* %tmp11, i64 %iv.2
|
|
%tmp13 = load float, float* %tmp12, align 4
|
|
%iv.2.next = add nuw nsw i64 %iv.2, 1
|
|
%c.2 = icmp ne i64 %iv.2.next, 100
|
|
br i1 %c.2, label %loop.2, label %loop.0.latch
|
|
|
|
loop.0.latch: ; preds = %loop.2, %loop.1
|
|
%c.0 = call i1 @cond()
|
|
br i1 %c.0, label %loop.0, label %end
|
|
|
|
end: ; preds = %loop.0.latch
|
|
ret void
|
|
}
|
|
|
|
declare i1 @cond()
|
|
|
|
declare void @use.i8(i8*)
|
|
|
|
declare void @use.i1(i1)
|
|
|
|
define void @reuse_cast_2(i32 %x, i32* %ptr.1.start) {
|
|
; CHECK-LABEL: @reuse_cast_2(
|
|
; CHECK-NEXT: entry:
|
|
; CHECK-NEXT: [[PTR_1_START2:%.*]] = bitcast i32* [[PTR_1_START:%.*]] to i8*
|
|
; CHECK-NEXT: [[STACK:%.*]] = alloca [2 x i32], align 4
|
|
; CHECK-NEXT: [[STACK1:%.*]] = bitcast [2 x i32]* [[STACK]] to i8*
|
|
; CHECK-NEXT: [[C_0:%.*]] = icmp sgt i32 [[X:%.*]], 0
|
|
; CHECK-NEXT: [[CAST_TO_REUSE:%.*]] = bitcast [2 x i32]* [[STACK]] to i8*
|
|
; CHECK-NEXT: [[PTR_2_START:%.*]] = getelementptr inbounds [2 x i32], [2 x i32]* [[STACK]], i64 0, i64 0
|
|
; CHECK-NEXT: call void @use.i8(i8* [[CAST_TO_REUSE]])
|
|
; CHECK-NEXT: call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 4 [[STACK1]], i8* align 4 [[PTR_1_START2]], i64 8, i1 false)
|
|
; CHECK-NEXT: br label [[LOOP:%.*]]
|
|
; CHECK: loop:
|
|
; CHECK-NEXT: [[IV:%.*]] = phi i32 [ 0, [[ENTRY:%.*]] ], [ [[IV_NEXT:%.*]], [[LOOP]] ]
|
|
; CHECK-NEXT: [[PTR_1:%.*]] = phi i32* [ [[PTR_1_START]], [[ENTRY]] ], [ [[PTR_1_NEXT:%.*]], [[LOOP]] ]
|
|
; CHECK-NEXT: [[PTR_2:%.*]] = phi i32* [ [[PTR_2_START]], [[ENTRY]] ], [ [[PTR_2_NEXT:%.*]], [[LOOP]] ]
|
|
; CHECK-NEXT: [[PTR_1_NEXT]] = getelementptr inbounds i32, i32* [[PTR_1]], i64 1
|
|
; CHECK-NEXT: [[LV:%.*]] = load i32, i32* [[PTR_1]], align 4
|
|
; CHECK-NEXT: [[PTR_2_NEXT]] = getelementptr inbounds i32, i32* [[PTR_2]], i64 1
|
|
; CHECK-NEXT: [[IV_NEXT]] = add nuw nsw i32 [[IV]], 1
|
|
; CHECK-NEXT: [[C_1:%.*]] = icmp eq i32 [[IV]], 0
|
|
; CHECK-NEXT: br i1 [[C_1]], label [[LOOP]], label [[EXIT:%.*]]
|
|
; CHECK: exit:
|
|
; CHECK-NEXT: call void @use.i1(i1 [[C_0]])
|
|
; CHECK-NEXT: ret void
|
|
;
|
|
entry:
|
|
%stack = alloca [2 x i32], align 4
|
|
%c.0 = icmp sgt i32 %x, 0
|
|
%cast.to.reuse = bitcast [2 x i32]* %stack to i8*
|
|
%ptr.2.start = getelementptr inbounds [2 x i32], [2 x i32]* %stack, i64 0, i64 0
|
|
call void @use.i8(i8* %cast.to.reuse)
|
|
br label %loop
|
|
|
|
loop: ; preds = %loop, %entry
|
|
%iv = phi i32 [ 0, %entry ], [ %iv.next, %loop ]
|
|
%ptr.1 = phi i32* [ %ptr.1.start, %entry ], [ %ptr.1.next, %loop ]
|
|
%ptr.2 = phi i32* [ %ptr.2.start, %entry ], [ %ptr.2.next, %loop ]
|
|
%ptr.1.next = getelementptr inbounds i32, i32* %ptr.1, i64 1
|
|
%lv = load i32, i32* %ptr.1, align 4
|
|
%ptr.2.next = getelementptr inbounds i32, i32* %ptr.2, i64 1
|
|
store i32 %lv, i32* %ptr.2, align 4
|
|
%iv.next = add nuw nsw i32 %iv, 1
|
|
%c.1 = icmp eq i32 %iv, 0
|
|
br i1 %c.1, label %loop, label %exit
|
|
|
|
exit: ; preds = %loop
|
|
call void @use.i1(i1 %c.0)
|
|
ret void
|
|
}
|