mirror of
https://github.com/RPCS3/llvm-mirror.git
synced 2024-10-22 20:43:44 +02:00
e1db330ce8
The backend tries to use block operations like MVC, NC, OC and XC for simple scalar operations. For correctness reasons, it rejects any case in which the regions might partially overlap. However, for performance reasons, it should also reject cases where the regions might be equal, since the instruction might then not use the fast path. This fixes a performance regression seen in bzip2. We may want to limit the optimisation even more in future, or even remove it entirely, but I'll try with this for now. llvm-svn: 191525
377 lines
9.1 KiB
LLVM
377 lines
9.1 KiB
LLVM
; Test memory-to-memory ANDs.
|
|
;
|
|
; RUN: llc < %s -mtriple=s390x-linux-gnu | FileCheck %s
|
|
|
|
@g1src = global i8 1
|
|
@g1dst = global i8 1
|
|
@g2src = global i16 2
|
|
@g2dst = global i16 2
|
|
|
|
; Test the simple i8 case.
|
|
define void @f1(i8 *%ptr1) {
|
|
; CHECK-LABEL: f1:
|
|
; CHECK: nc 1(1,%r2), 0(%r2)
|
|
; CHECK: br %r14
|
|
%ptr2 = getelementptr i8 *%ptr1, i64 1
|
|
%val = load i8 *%ptr1
|
|
%old = load i8 *%ptr2
|
|
%and = and i8 %val, %old
|
|
store i8 %and, i8 *%ptr2
|
|
ret void
|
|
}
|
|
|
|
; ...and again in reverse.
|
|
define void @f2(i8 *%ptr1) {
|
|
; CHECK-LABEL: f2:
|
|
; CHECK: nc 1(1,%r2), 0(%r2)
|
|
; CHECK: br %r14
|
|
%ptr2 = getelementptr i8 *%ptr1, i64 1
|
|
%val = load i8 *%ptr1
|
|
%old = load i8 *%ptr2
|
|
%and = and i8 %old, %val
|
|
store i8 %and, i8 *%ptr2
|
|
ret void
|
|
}
|
|
|
|
; Test i8 cases where one value is zero-extended to 32 bits and the other
|
|
; sign-extended.
|
|
define void @f3(i8 *%ptr1) {
|
|
; CHECK-LABEL: f3:
|
|
; CHECK: nc 1(1,%r2), 0(%r2)
|
|
; CHECK: br %r14
|
|
%ptr2 = getelementptr i8 *%ptr1, i64 1
|
|
%val = load i8 *%ptr1
|
|
%extval = zext i8 %val to i32
|
|
%old = load i8 *%ptr2
|
|
%extold = sext i8 %old to i32
|
|
%and = and i32 %extval, %extold
|
|
%trunc = trunc i32 %and to i8
|
|
store i8 %trunc, i8 *%ptr2
|
|
ret void
|
|
}
|
|
|
|
; ...and again with the extension types reversed.
|
|
define void @f4(i8 *%ptr1) {
|
|
; CHECK-LABEL: f4:
|
|
; CHECK: nc 1(1,%r2), 0(%r2)
|
|
; CHECK: br %r14
|
|
%ptr2 = getelementptr i8 *%ptr1, i64 1
|
|
%val = load i8 *%ptr1
|
|
%extval = sext i8 %val to i32
|
|
%old = load i8 *%ptr2
|
|
%extold = zext i8 %old to i32
|
|
%and = and i32 %extval, %extold
|
|
%trunc = trunc i32 %and to i8
|
|
store i8 %trunc, i8 *%ptr2
|
|
ret void
|
|
}
|
|
|
|
; ...and again with two sign extensions.
|
|
define void @f5(i8 *%ptr1) {
|
|
; CHECK-LABEL: f5:
|
|
; CHECK: nc 1(1,%r2), 0(%r2)
|
|
; CHECK: br %r14
|
|
%ptr2 = getelementptr i8 *%ptr1, i64 1
|
|
%val = load i8 *%ptr1
|
|
%extval = sext i8 %val to i32
|
|
%old = load i8 *%ptr2
|
|
%extold = sext i8 %old to i32
|
|
%and = and i32 %extval, %extold
|
|
%trunc = trunc i32 %and to i8
|
|
store i8 %trunc, i8 *%ptr2
|
|
ret void
|
|
}
|
|
|
|
; ...and again with two zero extensions.
|
|
define void @f6(i8 *%ptr1) {
|
|
; CHECK-LABEL: f6:
|
|
; CHECK: nc 1(1,%r2), 0(%r2)
|
|
; CHECK: br %r14
|
|
%ptr2 = getelementptr i8 *%ptr1, i64 1
|
|
%val = load i8 *%ptr1
|
|
%extval = zext i8 %val to i32
|
|
%old = load i8 *%ptr2
|
|
%extold = zext i8 %old to i32
|
|
%and = and i32 %extval, %extold
|
|
%trunc = trunc i32 %and to i8
|
|
store i8 %trunc, i8 *%ptr2
|
|
ret void
|
|
}
|
|
|
|
; Test i8 cases where the value is extended to 64 bits (just one case
|
|
; this time).
|
|
define void @f7(i8 *%ptr1) {
|
|
; CHECK-LABEL: f7:
|
|
; CHECK: nc 1(1,%r2), 0(%r2)
|
|
; CHECK: br %r14
|
|
%ptr2 = getelementptr i8 *%ptr1, i64 1
|
|
%val = load i8 *%ptr1
|
|
%extval = sext i8 %val to i64
|
|
%old = load i8 *%ptr2
|
|
%extold = zext i8 %old to i64
|
|
%and = and i64 %extval, %extold
|
|
%trunc = trunc i64 %and to i8
|
|
store i8 %trunc, i8 *%ptr2
|
|
ret void
|
|
}
|
|
|
|
; Test the simple i16 case.
|
|
define void @f8(i16 *%ptr1) {
|
|
; CHECK-LABEL: f8:
|
|
; CHECK: nc 2(2,%r2), 0(%r2)
|
|
; CHECK: br %r14
|
|
%ptr2 = getelementptr i16 *%ptr1, i64 1
|
|
%val = load i16 *%ptr1
|
|
%old = load i16 *%ptr2
|
|
%and = and i16 %val, %old
|
|
store i16 %and, i16 *%ptr2
|
|
ret void
|
|
}
|
|
|
|
; Test i16 cases where the value is extended to 32 bits.
|
|
define void @f9(i16 *%ptr1) {
|
|
; CHECK-LABEL: f9:
|
|
; CHECK: nc 2(2,%r2), 0(%r2)
|
|
; CHECK: br %r14
|
|
%ptr2 = getelementptr i16 *%ptr1, i64 1
|
|
%val = load i16 *%ptr1
|
|
%extval = zext i16 %val to i32
|
|
%old = load i16 *%ptr2
|
|
%extold = sext i16 %old to i32
|
|
%and = and i32 %extval, %extold
|
|
%trunc = trunc i32 %and to i16
|
|
store i16 %trunc, i16 *%ptr2
|
|
ret void
|
|
}
|
|
|
|
; Test i16 cases where the value is extended to 64 bits.
|
|
define void @f10(i16 *%ptr1) {
|
|
; CHECK-LABEL: f10:
|
|
; CHECK: nc 2(2,%r2), 0(%r2)
|
|
; CHECK: br %r14
|
|
%ptr2 = getelementptr i16 *%ptr1, i64 1
|
|
%val = load i16 *%ptr1
|
|
%extval = sext i16 %val to i64
|
|
%old = load i16 *%ptr2
|
|
%extold = zext i16 %old to i64
|
|
%and = and i64 %extval, %extold
|
|
%trunc = trunc i64 %and to i16
|
|
store i16 %trunc, i16 *%ptr2
|
|
ret void
|
|
}
|
|
|
|
; Test the simple i32 case.
|
|
define void @f11(i32 *%ptr1) {
|
|
; CHECK-LABEL: f11:
|
|
; CHECK: nc 4(4,%r2), 0(%r2)
|
|
; CHECK: br %r14
|
|
%ptr2 = getelementptr i32 *%ptr1, i64 1
|
|
%val = load i32 *%ptr1
|
|
%old = load i32 *%ptr2
|
|
%and = and i32 %old, %val
|
|
store i32 %and, i32 *%ptr2
|
|
ret void
|
|
}
|
|
|
|
; Test i32 cases where the value is extended to 64 bits.
|
|
define void @f12(i32 *%ptr1) {
|
|
; CHECK-LABEL: f12:
|
|
; CHECK: nc 4(4,%r2), 0(%r2)
|
|
; CHECK: br %r14
|
|
%ptr2 = getelementptr i32 *%ptr1, i64 1
|
|
%val = load i32 *%ptr1
|
|
%extval = sext i32 %val to i64
|
|
%old = load i32 *%ptr2
|
|
%extold = zext i32 %old to i64
|
|
%and = and i64 %extval, %extold
|
|
%trunc = trunc i64 %and to i32
|
|
store i32 %trunc, i32 *%ptr2
|
|
ret void
|
|
}
|
|
|
|
; Test the i64 case.
|
|
define void @f13(i64 *%ptr1) {
|
|
; CHECK-LABEL: f13:
|
|
; CHECK: nc 8(8,%r2), 0(%r2)
|
|
; CHECK: br %r14
|
|
%ptr2 = getelementptr i64 *%ptr1, i64 1
|
|
%val = load i64 *%ptr1
|
|
%old = load i64 *%ptr2
|
|
%and = and i64 %old, %val
|
|
store i64 %and, i64 *%ptr2
|
|
ret void
|
|
}
|
|
|
|
; Make sure that we don't use NC if the first load is volatile.
|
|
define void @f14(i64 *%ptr1) {
|
|
; CHECK-LABEL: f14:
|
|
; CHECK-NOT: nc
|
|
; CHECK: br %r14
|
|
%ptr2 = getelementptr i64 *%ptr1, i64 1
|
|
%val = load volatile i64 *%ptr1
|
|
%old = load i64 *%ptr2
|
|
%and = and i64 %old, %val
|
|
store i64 %and, i64 *%ptr2
|
|
ret void
|
|
}
|
|
|
|
; ...likewise the second.
|
|
define void @f15(i64 *%ptr1) {
|
|
; CHECK-LABEL: f15:
|
|
; CHECK-NOT: nc
|
|
; CHECK: br %r14
|
|
%ptr2 = getelementptr i64 *%ptr1, i64 1
|
|
%val = load i64 *%ptr1
|
|
%old = load volatile i64 *%ptr2
|
|
%and = and i64 %old, %val
|
|
store i64 %and, i64 *%ptr2
|
|
ret void
|
|
}
|
|
|
|
; ...likewise the store.
|
|
define void @f16(i64 *%ptr1) {
|
|
; CHECK-LABEL: f16:
|
|
; CHECK-NOT: nc
|
|
; CHECK: br %r14
|
|
%ptr2 = getelementptr i64 *%ptr1, i64 1
|
|
%val = load i64 *%ptr1
|
|
%old = load i64 *%ptr2
|
|
%and = and i64 %old, %val
|
|
store volatile i64 %and, i64 *%ptr2
|
|
ret void
|
|
}
|
|
|
|
; Test that NC is not used for aligned loads and stores if there is
|
|
; no way of telling whether they alias. We don't want to use NC in
|
|
; cases where the addresses could be equal.
|
|
define void @f17(i64 *%ptr1, i64 *%ptr2) {
|
|
; CHECK-LABEL: f17:
|
|
; CHECK-NOT: nc
|
|
; CHECK: br %r14
|
|
%val = load i64 *%ptr1
|
|
%old = load i64 *%ptr2
|
|
%and = and i64 %old, %val
|
|
store i64 %and, i64 *%ptr2
|
|
ret void
|
|
}
|
|
|
|
; ...but if one of the loads isn't aligned, we can't be sure.
|
|
define void @f18(i64 *%ptr1, i64 *%ptr2) {
|
|
; CHECK-LABEL: f18:
|
|
; CHECK-NOT: nc
|
|
; CHECK: br %r14
|
|
%val = load i64 *%ptr1, align 2
|
|
%old = load i64 *%ptr2
|
|
%and = and i64 %old, %val
|
|
store i64 %and, i64 *%ptr2
|
|
ret void
|
|
}
|
|
|
|
; Repeat the previous test with the operands in the opposite order.
|
|
define void @f19(i64 *%ptr1, i64 *%ptr2) {
|
|
; CHECK-LABEL: f19:
|
|
; CHECK-NOT: nc
|
|
; CHECK: br %r14
|
|
%val = load i64 *%ptr1, align 2
|
|
%old = load i64 *%ptr2
|
|
%and = and i64 %val, %old
|
|
store i64 %and, i64 *%ptr2
|
|
ret void
|
|
}
|
|
|
|
; ...and again with the other operand being unaligned.
|
|
define void @f20(i64 *%ptr1, i64 *%ptr2) {
|
|
; CHECK-LABEL: f20:
|
|
; CHECK-NOT: nc
|
|
; CHECK: br %r14
|
|
%val = load i64 *%ptr1
|
|
%old = load i64 *%ptr2, align 2
|
|
%and = and i64 %val, %old
|
|
store i64 %and, i64 *%ptr2, align 2
|
|
ret void
|
|
}
|
|
|
|
; Test a case where there is definite overlap.
|
|
define void @f21(i64 %base) {
|
|
; CHECK-LABEL: f21:
|
|
; CHECK-NOT: nc
|
|
; CHECK: br %r14
|
|
%add = add i64 %base, 1
|
|
%ptr1 = inttoptr i64 %base to i64 *
|
|
%ptr2 = inttoptr i64 %add to i64 *
|
|
%val = load i64 *%ptr1
|
|
%old = load i64 *%ptr2, align 1
|
|
%and = and i64 %old, %val
|
|
store i64 %and, i64 *%ptr2, align 1
|
|
ret void
|
|
}
|
|
|
|
; Test that we can use NC for global addresses for i8.
|
|
define void @f22(i8 *%ptr) {
|
|
; CHECK-LABEL: f22:
|
|
; CHECK-DAG: larl [[SRC:%r[0-5]]], g1src
|
|
; CHECK-DAG: larl [[DST:%r[0-5]]], g1dst
|
|
; CHECK: nc 0(1,[[DST]]), 0([[SRC]])
|
|
; CHECK: br %r14
|
|
%val = load i8 *@g1src
|
|
%old = load i8 *@g1dst
|
|
%and = and i8 %val, %old
|
|
store i8 %and, i8 *@g1dst
|
|
ret void
|
|
}
|
|
|
|
; Test that we use NC even where LHRL and STHRL are available.
|
|
define void @f23(i16 *%ptr) {
|
|
; CHECK-LABEL: f23:
|
|
; CHECK-DAG: larl [[SRC:%r[0-5]]], g2src
|
|
; CHECK-DAG: larl [[DST:%r[0-5]]], g2dst
|
|
; CHECK: nc 0(2,[[DST]]), 0([[SRC]])
|
|
; CHECK: br %r14
|
|
%val = load i16 *@g2src
|
|
%old = load i16 *@g2dst
|
|
%and = and i16 %val, %old
|
|
store i16 %and, i16 *@g2dst
|
|
ret void
|
|
}
|
|
|
|
; Test a case where offset disambiguation is enough.
|
|
define void @f24(i64 *%ptr1) {
|
|
; CHECK-LABEL: f24:
|
|
; CHECK: nc 8(8,%r2), 0(%r2)
|
|
; CHECK: br %r14
|
|
%ptr2 = getelementptr i64 *%ptr1, i64 1
|
|
%val = load i64 *%ptr1, align 1
|
|
%old = load i64 *%ptr2, align 1
|
|
%and = and i64 %old, %val
|
|
store i64 %and, i64 *%ptr2, align 1
|
|
ret void
|
|
}
|
|
|
|
; Test a case where TBAA tells us there is no alias.
|
|
define void @f25(i64 *%ptr1, i64 *%ptr2) {
|
|
; CHECK-LABEL: f25:
|
|
; CHECK: nc 0(8,%r3), 0(%r2)
|
|
; CHECK: br %r14
|
|
%val = load i64 *%ptr1, align 2, !tbaa !1
|
|
%old = load i64 *%ptr2, align 2, !tbaa !2
|
|
%and = and i64 %old, %val
|
|
store i64 %and, i64 *%ptr2, align 2, !tbaa !2
|
|
ret void
|
|
}
|
|
|
|
; Test a case where TBAA information is present but doesn't help.
|
|
define void @f26(i64 *%ptr1, i64 *%ptr2) {
|
|
; CHECK-LABEL: f26:
|
|
; CHECK-NOT: nc
|
|
; CHECK: br %r14
|
|
%val = load i64 *%ptr1, align 2, !tbaa !1
|
|
%old = load i64 *%ptr2, align 2, !tbaa !1
|
|
%and = and i64 %old, %val
|
|
store i64 %and, i64 *%ptr2, align 2, !tbaa !1
|
|
ret void
|
|
}
|
|
|
|
!0 = metadata !{ metadata !"root" }
|
|
!1 = metadata !{ metadata !"set1", metadata !0 }
|
|
!2 = metadata !{ metadata !"set2", metadata !0 }
|