maxjhandsome
/
llvm-project
forked from OSchip/llvm-project
1
0
Fork 0
Code Issues Pull Requests Packages Releases Wiki Activity
llvm-project/llvm/test/Transforms/InstCombine/select_meta.ll

315 lines
11 KiB
LLVM
Raw Normal View History

[InstCombine] canonicalize min/max constant to select's false value This is a first step towards canonicalization and improved folding/codegen for integer min/max as discussed here: http://lists.llvm.org/pipermail/llvm-dev/2016-November/106868.html Here, we're just matching the simplest min/max patterns and adjusting the icmp predicate while swapping the select operands. I've included FIXME tests in test/Transforms/InstCombine/select_meta.ll so it's easier to see how this might be extended (corresponds to the TODO comment in the code). That's also why I'm using matchSelectPattern() rather than a simpler check; once the backend is patched, we can just remove some of the restrictions to allow the obfuscated min/max patterns in the FIXME tests to be matched. Differential Revision: https://reviews.llvm.org/D26525 llvm-svn: 287585
2016-11-22 06:04:14 +08:00
; NOTE: Assertions have been autogenerated by utils/update_test_checks.py
; and enhanced to include metadata checking.
[Profile] Propagate branch metadata properly in instcombine Differential Revision: http://reviews.llvm.org/D23590 llvm-svn: 279693
2016-08-25 08:26:32 +08:00
; RUN: opt < %s -instcombine -S | FileCheck %s
define i32 @foo(i32) local_unnamed_addr #0 {
[InstCombine] auto-generate better checks; NFC Note that the existing metadata checking was re-added by hand because the script doesn't currently know how to generate checks for lines outside of functions. llvm-svn: 286460
2016-11-10 22:58:17 +08:00
; CHECK-LABEL: @foo(
; CHECK-NEXT: [[TMP2:%.*]] = icmp sgt i32 %0, 2
; CHECK-NEXT: [[DOTV:%.*]] = select i1 [[TMP2]], i32 20, i32 -20, !prof ![[MD1:[0-9]+]]
; CHECK-NEXT: [[TMP3:%.*]] = add i32 [[DOTV]], %0
; CHECK-NEXT: ret i32 [[TMP3]]
;
[Profile] Propagate branch metadata properly in instcombine Differential Revision: http://reviews.llvm.org/D23590 llvm-svn: 279693
2016-08-25 08:26:32 +08:00
%2 = icmp sgt i32 %0, 2
%3 = add nsw i32 %0, 20
%4 = add nsw i32 %0, -20
select i1 %2, i32 %3, i32 %4, !prof !1
ret i32 %5
}
[InstCombine] don't drop metadata in FoldOpIntoSelect() llvm-svn: 287980
2016-11-26 23:23:20 +08:00
define i8 @shrink_select(i1 %cond, i32 %x) {
; CHECK-LABEL: @shrink_select(
; CHECK-NEXT: [[TMP1:%.*]] = trunc i32 %x to i8
; CHECK-NEXT: [[TRUNC:%.*]] = select i1 %cond, i8 [[TMP1]], i8 42, !prof ![[MD1]]
; CHECK-NEXT: ret i8 [[TRUNC]]
;
%sel = select i1 %cond, i32 %x, i32 42, !prof !1
%trunc = trunc i32 %sel to i8
ret i8 %trunc
}
[Profile] Propagate branch metadata properly in instcombine Differential Revision: http://reviews.llvm.org/D23590 llvm-svn: 279693
2016-08-25 08:26:32 +08:00
define void @min_max_bitcast(<4 x float> %a, <4 x float> %b, <4 x i32>* %ptr1, <4 x i32>* %ptr2) {
[InstCombine] auto-generate better checks; NFC Note that the existing metadata checking was re-added by hand because the script doesn't currently know how to generate checks for lines outside of functions. llvm-svn: 286460
2016-11-10 22:58:17 +08:00
; CHECK-LABEL: @min_max_bitcast(
; CHECK-NEXT: [[CMP:%.*]] = fcmp olt <4 x float> %a, %b
; CHECK-NEXT: [[SEL1_V:%.*]] = select <4 x i1> [[CMP]], <4 x float> %a, <4 x float> %b, !prof ![[MD1]]
; CHECK-NEXT: [[SEL2_V:%.*]] = select <4 x i1> [[CMP]], <4 x float> %b, <4 x float> %a, !prof ![[MD1]]
; CHECK-NEXT: [[TMP1:%.*]] = bitcast <4 x i32>* %ptr1 to <4 x float>*
; CHECK-NEXT: store <4 x float> [[SEL1_V]], <4 x float>* [[TMP1]], align 16
; CHECK-NEXT: [[TMP2:%.*]] = bitcast <4 x i32>* %ptr2 to <4 x float>*
; CHECK-NEXT: store <4 x float> [[SEL2_V]], <4 x float>* [[TMP2]], align 16
; CHECK-NEXT: ret void
;
[Profile] Propagate branch metadata properly in instcombine Differential Revision: http://reviews.llvm.org/D23590 llvm-svn: 279693
2016-08-25 08:26:32 +08:00
%cmp = fcmp olt <4 x float> %a, %b
%bc1 = bitcast <4 x float> %a to <4 x i32>
%bc2 = bitcast <4 x float> %b to <4 x i32>
%sel1 = select <4 x i1> %cmp, <4 x i32> %bc1, <4 x i32> %bc2, !prof !1
%sel2 = select <4 x i1> %cmp, <4 x i32> %bc2, <4 x i32> %bc1, !prof !1
store <4 x i32> %sel1, <4 x i32>* %ptr1
store <4 x i32> %sel2, <4 x i32>* %ptr2
ret void
}
define i32 @foo2(i32, i32) local_unnamed_addr #0 {
[InstCombine] auto-generate better checks; NFC Note that the existing metadata checking was re-added by hand because the script doesn't currently know how to generate checks for lines outside of functions. llvm-svn: 286460
2016-11-10 22:58:17 +08:00
; CHECK-LABEL: @foo2(
; CHECK-NEXT: [[TMP3:%.*]] = icmp sgt i32 %0, 2
; CHECK-NEXT: [[TMP4:%.*]] = sub i32 0, %1
; CHECK-NEXT: [[DOTP:%.*]] = select i1 [[TMP3]], i32 %1, i32 [[TMP4]], !prof ![[MD1]]
; CHECK-NEXT: [[TMP5:%.*]] = add i32 [[DOTP]], %0
; CHECK-NEXT: ret i32 [[TMP5]]
;
[Profile] Propagate branch metadata properly in instcombine Differential Revision: http://reviews.llvm.org/D23590 llvm-svn: 279693
2016-08-25 08:26:32 +08:00
%3 = icmp sgt i32 %0, 2
%4 = add nsw i32 %0, %1
%5 = sub nsw i32 %0, %1
select i1 %3, i32 %4, i32 %5, !prof !1
ret i32 %6
}
define i64 @test43(i32 %a) nounwind {
[InstCombine] auto-generate better checks; NFC Note that the existing metadata checking was re-added by hand because the script doesn't currently know how to generate checks for lines outside of functions. llvm-svn: 286460
2016-11-10 22:58:17 +08:00
; CHECK-LABEL: @test43(
; CHECK-NEXT: [[A_EXT:%.*]] = sext i32 %a to i64
[InstCombine] canonicalize min/max constant to select's false value This is a first step towards canonicalization and improved folding/codegen for integer min/max as discussed here: http://lists.llvm.org/pipermail/llvm-dev/2016-November/106868.html Here, we're just matching the simplest min/max patterns and adjusting the icmp predicate while swapping the select operands. I've included FIXME tests in test/Transforms/InstCombine/select_meta.ll so it's easier to see how this might be extended (corresponds to the TODO comment in the code). That's also why I'm using matchSelectPattern() rather than a simpler check; once the backend is patched, we can just remove some of the restrictions to allow the obfuscated min/max patterns in the FIXME tests to be matched. Differential Revision: https://reviews.llvm.org/D26525 llvm-svn: 287585
2016-11-22 06:04:14 +08:00
; CHECK-NEXT: [[TMP1:%.*]] = icmp sgt i64 [[A_EXT]], 0
; CHECK-NEXT: [[MAX:%.*]] = select i1 [[TMP1]], i64 [[A_EXT]], i64 0, !prof ![[MD1]]
[InstCombine] auto-generate better checks; NFC Note that the existing metadata checking was re-added by hand because the script doesn't currently know how to generate checks for lines outside of functions. llvm-svn: 286460
2016-11-10 22:58:17 +08:00
; CHECK-NEXT: ret i64 [[MAX]]
;
[Profile] Propagate branch metadata properly in instcombine Differential Revision: http://reviews.llvm.org/D23590 llvm-svn: 279693
2016-08-25 08:26:32 +08:00
%a_ext = sext i32 %a to i64
%is_a_nonnegative = icmp sgt i32 %a, -1
%max = select i1 %is_a_nonnegative, i64 %a_ext, i64 0, !prof !1
ret i64 %max
}
define <2 x i32> @scalar_select_of_vectors_sext(<2 x i1> %cca, i1 %ccb) {
[InstCombine] auto-generate better checks; NFC Note that the existing metadata checking was re-added by hand because the script doesn't currently know how to generate checks for lines outside of functions. llvm-svn: 286460
2016-11-10 22:58:17 +08:00
; CHECK-LABEL: @scalar_select_of_vectors_sext(
; CHECK-NEXT: [[NARROW:%.*]] = select i1 %ccb, <2 x i1> %cca, <2 x i1> zeroinitializer, !prof ![[MD1]]
; CHECK-NEXT: [[R:%.*]] = sext <2 x i1> [[NARROW]] to <2 x i32>
; CHECK-NEXT: ret <2 x i32> [[R]]
;
[Profile] Propagate branch metadata properly in instcombine Differential Revision: http://reviews.llvm.org/D23590 llvm-svn: 279693
2016-08-25 08:26:32 +08:00
%ccax = sext <2 x i1> %cca to <2 x i32>
%r = select i1 %ccb, <2 x i32> %ccax, <2 x i32> <i32 0, i32 0>, !prof !1
ret <2 x i32> %r
}
define i16 @t7(i32 %a) {
[InstCombine] auto-generate better checks; NFC Note that the existing metadata checking was re-added by hand because the script doesn't currently know how to generate checks for lines outside of functions. llvm-svn: 286460
2016-11-10 22:58:17 +08:00
; CHECK-LABEL: @t7(
; CHECK-NEXT: [[TMP1:%.*]] = icmp slt i32 %a, -32768
; CHECK-NEXT: [[TMP2:%.*]] = select i1 [[TMP1]], i32 %a, i32 -32768, !prof ![[MD1]]
; CHECK-NEXT: [[TMP3:%.*]] = trunc i32 [[TMP2]] to i16
; CHECK-NEXT: ret i16 [[TMP3]]
;
[Profile] Propagate branch metadata properly in instcombine Differential Revision: http://reviews.llvm.org/D23590 llvm-svn: 279693
2016-08-25 08:26:32 +08:00
%1 = icmp slt i32 %a, -32768
%2 = trunc i32 %a to i16
%3 = select i1 %1, i16 %2, i16 -32768, !prof !1
ret i16 %3
}
define i32 @abs_nabs_x01(i32 %x) {
[InstCombine] auto-generate better checks; NFC Note that the existing metadata checking was re-added by hand because the script doesn't currently know how to generate checks for lines outside of functions. llvm-svn: 286460
2016-11-10 22:58:17 +08:00
; CHECK-LABEL: @abs_nabs_x01(
; CHECK-NEXT: [[CMP:%.*]] = icmp sgt i32 %x, -1
; CHECK-NEXT: [[SUB:%.*]] = sub nsw i32 0, %x
; CHECK-NEXT: [[COND1:%.*]] = select i1 [[CMP]], i32 %x, i32 [[SUB]], !prof ![[MD1]]
; CHECK-NEXT: ret i32 [[COND1]]
;
[Profile] Propagate branch metadata properly in instcombine Differential Revision: http://reviews.llvm.org/D23590 llvm-svn: 279693
2016-08-25 08:26:32 +08:00
%cmp = icmp sgt i32 %x, -1
%sub = sub nsw i32 0, %x
%cond = select i1 %cmp, i32 %sub, i32 %x, !prof !1
%cmp1 = icmp sgt i32 %cond, -1
%sub16 = sub nsw i32 0, %cond
%cond18 = select i1 %cmp1, i32 %cond, i32 %sub16, !prof !2
ret i32 %cond18
}
[InstCombine] add tests for missing folds of vector abs/nabs/min/max llvm-svn: 285299
2016-10-27 23:02:45 +08:00
define <2 x i32> @abs_nabs_x01_vec(<2 x i32> %x) {
; CHECK-LABEL: @abs_nabs_x01_vec(
; CHECK-NEXT: [[CMP:%.*]] = icmp sgt <2 x i32> %x, <i32 -1, i32 -1>
; CHECK-NEXT: [[SUB:%.*]] = sub nsw <2 x i32> zeroinitializer, %x
[ValueTracking] fix matchSelectPattern to allow vector splat folds of min/max/abs/nabs llvm-svn: 285303
2016-10-27 23:26:10 +08:00
; CHECK-NEXT: [[COND1:%.*]] = select <2 x i1> [[CMP]], <2 x i32> %x, <2 x i32> [[SUB]], !prof ![[MD1]]
; CHECK-NEXT: ret <2 x i32> [[COND1]]
[InstCombine] add tests for missing folds of vector abs/nabs/min/max llvm-svn: 285299
2016-10-27 23:02:45 +08:00
;
%cmp = icmp sgt <2 x i32> %x, <i32 -1, i32 -1>
%sub = sub nsw <2 x i32> zeroinitializer, %x
%cond = select <2 x i1> %cmp, <2 x i32> %sub, <2 x i32> %x, !prof !1
%cmp1 = icmp sgt <2 x i32> %cond, <i32 -1, i32 -1>
%sub16 = sub nsw <2 x i32> zeroinitializer, %cond
%cond18 = select <2 x i1> %cmp1, <2 x i32> %cond, <2 x i32> %sub16, !prof !2
ret <2 x i32> %cond18
}
[Profile] Propagate branch metadata properly in instcombine Differential Revision: http://reviews.llvm.org/D23590 llvm-svn: 279693
2016-08-25 08:26:32 +08:00
; SMAX(SMAX(x, y), x) -> SMAX(x, y)
define i32 @test30(i32 %x, i32 %y) {
[InstCombine] auto-generate better checks; NFC Note that the existing metadata checking was re-added by hand because the script doesn't currently know how to generate checks for lines outside of functions. llvm-svn: 286460
2016-11-10 22:58:17 +08:00
; CHECK-LABEL: @test30(
; CHECK-NEXT: [[CMP:%.*]] = icmp sgt i32 %x, %y
; CHECK-NEXT: [[COND:%.*]] = select i1 [[CMP]], i32 %x, i32 %y, !prof ![[MD1]]
; CHECK-NEXT: ret i32 [[COND]]
;
[Profile] Propagate branch metadata properly in instcombine Differential Revision: http://reviews.llvm.org/D23590 llvm-svn: 279693
2016-08-25 08:26:32 +08:00
%cmp = icmp sgt i32 %x, %y
%cond = select i1 %cmp, i32 %x, i32 %y, !prof !1
%cmp5 = icmp sgt i32 %cond, %x
%retval = select i1 %cmp5, i32 %cond, i32 %x, !prof !2
ret i32 %retval
}
[InstCombine] canonicalize min/max constant to select's false value This is a first step towards canonicalization and improved folding/codegen for integer min/max as discussed here: http://lists.llvm.org/pipermail/llvm-dev/2016-November/106868.html Here, we're just matching the simplest min/max patterns and adjusting the icmp predicate while swapping the select operands. I've included FIXME tests in test/Transforms/InstCombine/select_meta.ll so it's easier to see how this might be extended (corresponds to the TODO comment in the code). That's also why I'm using matchSelectPattern() rather than a simpler check; once the backend is patched, we can just remove some of the restrictions to allow the obfuscated min/max patterns in the FIXME tests to be matched. Differential Revision: https://reviews.llvm.org/D26525 llvm-svn: 287585
2016-11-22 06:04:14 +08:00
; Swap predicate / metadata order
; SMAX(SMAX(X, 75), 36) -> SMAX(X, 75)
[Profile] Propagate branch metadata properly in instcombine Differential Revision: http://reviews.llvm.org/D23590 llvm-svn: 279693
2016-08-25 08:26:32 +08:00
define i32 @test70(i32 %x) {
[InstCombine] auto-generate better checks; NFC Note that the existing metadata checking was re-added by hand because the script doesn't currently know how to generate checks for lines outside of functions. llvm-svn: 286460
2016-11-10 22:58:17 +08:00
; CHECK-LABEL: @test70(
[InstCombine] canonicalize min/max constant to select's false value This is a first step towards canonicalization and improved folding/codegen for integer min/max as discussed here: http://lists.llvm.org/pipermail/llvm-dev/2016-November/106868.html Here, we're just matching the simplest min/max patterns and adjusting the icmp predicate while swapping the select operands. I've included FIXME tests in test/Transforms/InstCombine/select_meta.ll so it's easier to see how this might be extended (corresponds to the TODO comment in the code). That's also why I'm using matchSelectPattern() rather than a simpler check; once the backend is patched, we can just remove some of the restrictions to allow the obfuscated min/max patterns in the FIXME tests to be matched. Differential Revision: https://reviews.llvm.org/D26525 llvm-svn: 287585
2016-11-22 06:04:14 +08:00
; CHECK-NEXT: [[TMP1:%.*]] = icmp sgt i32 %x, 75
; CHECK-NEXT: [[COND:%.*]] = select i1 [[TMP1]], i32 %x, i32 75, !prof ![[MD3:[0-9]+]]
[InstCombine] auto-generate better checks; NFC Note that the existing metadata checking was re-added by hand because the script doesn't currently know how to generate checks for lines outside of functions. llvm-svn: 286460
2016-11-10 22:58:17 +08:00
; CHECK-NEXT: ret i32 [[COND]]
;
[Profile] Propagate branch metadata properly in instcombine Differential Revision: http://reviews.llvm.org/D23590 llvm-svn: 279693
2016-08-25 08:26:32 +08:00
%cmp = icmp slt i32 %x, 75
%cond = select i1 %cmp, i32 75, i32 %x, !prof !1
%cmp3 = icmp slt i32 %cond, 36
%retval = select i1 %cmp3, i32 36, i32 %cond, !prof !2
ret i32 %retval
}
[InstCombine] canonicalize min/max constant to select's false value This is a first step towards canonicalization and improved folding/codegen for integer min/max as discussed here: http://lists.llvm.org/pipermail/llvm-dev/2016-November/106868.html Here, we're just matching the simplest min/max patterns and adjusting the icmp predicate while swapping the select operands. I've included FIXME tests in test/Transforms/InstCombine/select_meta.ll so it's easier to see how this might be extended (corresponds to the TODO comment in the code). That's also why I'm using matchSelectPattern() rather than a simpler check; once the backend is patched, we can just remove some of the restrictions to allow the obfuscated min/max patterns in the FIXME tests to be matched. Differential Revision: https://reviews.llvm.org/D26525 llvm-svn: 287585
2016-11-22 06:04:14 +08:00
; Swap predicate / metadata order
[Profile] Propagate branch metadata properly in instcombine Differential Revision: http://reviews.llvm.org/D23590 llvm-svn: 279693
2016-08-25 08:26:32 +08:00
; SMIN(SMIN(X, 92), 11) -> SMIN(X, 11)
define i32 @test72(i32 %x) {
[InstCombine] auto-generate better checks; NFC Note that the existing metadata checking was re-added by hand because the script doesn't currently know how to generate checks for lines outside of functions. llvm-svn: 286460
2016-11-10 22:58:17 +08:00
; CHECK-LABEL: @test72(
[InstCombine] canonicalize min/max constant to select's false value This is a first step towards canonicalization and improved folding/codegen for integer min/max as discussed here: http://lists.llvm.org/pipermail/llvm-dev/2016-November/106868.html Here, we're just matching the simplest min/max patterns and adjusting the icmp predicate while swapping the select operands. I've included FIXME tests in test/Transforms/InstCombine/select_meta.ll so it's easier to see how this might be extended (corresponds to the TODO comment in the code). That's also why I'm using matchSelectPattern() rather than a simpler check; once the backend is patched, we can just remove some of the restrictions to allow the obfuscated min/max patterns in the FIXME tests to be matched. Differential Revision: https://reviews.llvm.org/D26525 llvm-svn: 287585
2016-11-22 06:04:14 +08:00
; CHECK-NEXT: [[TMP1:%.*]] = icmp slt i32 %x, 11
; CHECK-NEXT: [[RETVAL:%.*]] = select i1 [[TMP1]], i32 %x, i32 11, !prof ![[MD4:[0-9]+]]
[InstCombine] auto-generate better checks; NFC Note that the existing metadata checking was re-added by hand because the script doesn't currently know how to generate checks for lines outside of functions. llvm-svn: 286460
2016-11-10 22:58:17 +08:00
; CHECK-NEXT: ret i32 [[RETVAL]]
;
[Profile] Propagate branch metadata properly in instcombine Differential Revision: http://reviews.llvm.org/D23590 llvm-svn: 279693
2016-08-25 08:26:32 +08:00
%cmp = icmp sgt i32 %x, 92
%cond = select i1 %cmp, i32 92, i32 %x, !prof !1
%cmp3 = icmp sgt i32 %cond, 11
%retval = select i1 %cmp3, i32 11, i32 %cond, !prof !2
ret i32 %retval
}
[InstCombine] canonicalize min/max constant to select's false value This is a first step towards canonicalization and improved folding/codegen for integer min/max as discussed here: http://lists.llvm.org/pipermail/llvm-dev/2016-November/106868.html Here, we're just matching the simplest min/max patterns and adjusting the icmp predicate while swapping the select operands. I've included FIXME tests in test/Transforms/InstCombine/select_meta.ll so it's easier to see how this might be extended (corresponds to the TODO comment in the code). That's also why I'm using matchSelectPattern() rather than a simpler check; once the backend is patched, we can just remove some of the restrictions to allow the obfuscated min/max patterns in the FIXME tests to be matched. Differential Revision: https://reviews.llvm.org/D26525 llvm-svn: 287585
2016-11-22 06:04:14 +08:00
; Swap predicate / metadata order
[Profile] Propagate branch metadata properly in instcombine Differential Revision: http://reviews.llvm.org/D23590 llvm-svn: 279693
2016-08-25 08:26:32 +08:00
; SMAX(SMAX(X, 36), 75) -> SMAX(X, 75)
define i32 @test74(i32 %x) {
[InstCombine] auto-generate better checks; NFC Note that the existing metadata checking was re-added by hand because the script doesn't currently know how to generate checks for lines outside of functions. llvm-svn: 286460
2016-11-10 22:58:17 +08:00
; CHECK-LABEL: @test74(
[InstCombine] canonicalize min/max constant to select's false value This is a first step towards canonicalization and improved folding/codegen for integer min/max as discussed here: http://lists.llvm.org/pipermail/llvm-dev/2016-November/106868.html Here, we're just matching the simplest min/max patterns and adjusting the icmp predicate while swapping the select operands. I've included FIXME tests in test/Transforms/InstCombine/select_meta.ll so it's easier to see how this might be extended (corresponds to the TODO comment in the code). That's also why I'm using matchSelectPattern() rather than a simpler check; once the backend is patched, we can just remove some of the restrictions to allow the obfuscated min/max patterns in the FIXME tests to be matched. Differential Revision: https://reviews.llvm.org/D26525 llvm-svn: 287585
2016-11-22 06:04:14 +08:00
; CHECK-NEXT: [[TMP1:%.*]] = icmp sgt i32 %x, 75
; CHECK-NEXT: [[RETVAL:%.*]] = select i1 [[TMP1]], i32 %x, i32 75, !prof ![[MD4]]
[InstCombine] auto-generate better checks; NFC Note that the existing metadata checking was re-added by hand because the script doesn't currently know how to generate checks for lines outside of functions. llvm-svn: 286460
2016-11-10 22:58:17 +08:00
; CHECK-NEXT: ret i32 [[RETVAL]]
;
[Profile] Propagate branch metadata properly in instcombine Differential Revision: http://reviews.llvm.org/D23590 llvm-svn: 279693
2016-08-25 08:26:32 +08:00
%cmp = icmp slt i32 %x, 36
%cond = select i1 %cmp, i32 36, i32 %x, !prof !1
%cmp3 = icmp slt i32 %cond, 75
%retval = select i1 %cmp3, i32 75, i32 %cond, !prof !2
ret i32 %retval
}
[InstCombine] canonicalize min/max constant to select's false value This is a first step towards canonicalization and improved folding/codegen for integer min/max as discussed here: http://lists.llvm.org/pipermail/llvm-dev/2016-November/106868.html Here, we're just matching the simplest min/max patterns and adjusting the icmp predicate while swapping the select operands. I've included FIXME tests in test/Transforms/InstCombine/select_meta.ll so it's easier to see how this might be extended (corresponds to the TODO comment in the code). That's also why I'm using matchSelectPattern() rather than a simpler check; once the backend is patched, we can just remove some of the restrictions to allow the obfuscated min/max patterns in the FIXME tests to be matched. Differential Revision: https://reviews.llvm.org/D26525 llvm-svn: 287585
2016-11-22 06:04:14 +08:00
; FIXME:
; The compare should change, but the metadata remains the same because the select operands are not swapped.
define i32 @smin1(i32 %x) {
; CHECK-LABEL: @smin1(
; CHECK-NEXT: [[NOT_X:%.*]] = xor i32 %x, -1
; CHECK-NEXT: [[CMP:%.*]] = icmp sgt i32 %x, 0
; CHECK-NEXT: [[SEL:%.*]] = select i1 [[CMP]], i32 [[NOT_X]], i32 -1, !prof ![[MD1]]
; CHECK-NEXT: ret i32 [[SEL]]
;
%not_x = xor i32 %x, -1
%cmp = icmp sgt i32 %x, 0
%sel = select i1 %cmp, i32 %not_x, i32 -1, !prof !1
ret i32 %sel
}
; FIXME:
; The compare should change, and the metadata is swapped because the select operands are swapped.
define i32 @smin2(i32 %x) {
; CHECK-LABEL: @smin2(
; CHECK-NEXT: [[NOT_X:%.*]] = xor i32 %x, -1
; CHECK-NEXT: [[CMP:%.*]] = icmp slt i32 %x, 0
; CHECK-NEXT: [[SEL:%.*]] = select i1 [[CMP]], i32 -1, i32 [[NOT_X]], !prof ![[MD1]]
; CHECK-NEXT: ret i32 [[SEL]]
;
%not_x = xor i32 %x, -1
%cmp = icmp slt i32 %x, 0
%sel = select i1 %cmp, i32 -1, i32 %not_x, !prof !1
ret i32 %sel
}
; FIXME:
; The compare should change, but the metadata remains the same because the select operands are not swapped.
define i32 @smax1(i32 %x) {
; CHECK-LABEL: @smax1(
; CHECK-NEXT: [[NOT_X:%.*]] = xor i32 %x, -1
; CHECK-NEXT: [[CMP:%.*]] = icmp slt i32 %x, 0
; CHECK-NEXT: [[SEL:%.*]] = select i1 [[CMP]], i32 [[NOT_X]], i32 -1, !prof ![[MD1]]
; CHECK-NEXT: ret i32 [[SEL]]
;
%not_x = xor i32 %x, -1
%cmp = icmp slt i32 %x, 0
%sel = select i1 %cmp, i32 %not_x, i32 -1, !prof !1
ret i32 %sel
}
; FIXME:
; The compare should change, and the metadata is swapped because the select operands are swapped.
define i32 @smax2(i32 %x) {
; CHECK-LABEL: @smax2(
; CHECK-NEXT: [[NOT_X:%.*]] = xor i32 %x, -1
; CHECK-NEXT: [[CMP:%.*]] = icmp sgt i32 %x, 0
; CHECK-NEXT: [[SEL:%.*]] = select i1 [[CMP]], i32 -1, i32 [[NOT_X]], !prof ![[MD1]]
; CHECK-NEXT: ret i32 [[SEL]]
;
%not_x = xor i32 %x, -1
%cmp = icmp sgt i32 %x, 0
%sel = select i1 %cmp, i32 -1, i32 %not_x, !prof !1
ret i32 %sel
}
; FIXME:
; The compare should change, but the metadata remains the same because the select operands are not swapped.
define i32 @umin1(i32 %x) {
; CHECK-LABEL: @umin1(
; CHECK-NEXT: [[CMP:%.*]] = icmp sgt i32 %x, -1
; CHECK-NEXT: [[SEL:%.*]] = select i1 [[CMP]], i32 %x, i32 -2147483648, !prof ![[MD1]]
; CHECK-NEXT: ret i32 [[SEL]]
;
%cmp = icmp sgt i32 %x, -1
%sel = select i1 %cmp, i32 %x, i32 -2147483648, !prof !1
ret i32 %sel
}
; FIXME:
; The compare should change, and the metadata is swapped because the select operands are swapped.
define i32 @umin2(i32 %x) {
; CHECK-LABEL: @umin2(
; CHECK-NEXT: [[CMP:%.*]] = icmp slt i32 %x, 0
; CHECK-NEXT: [[SEL:%.*]] = select i1 [[CMP]], i32 2147483647, i32 %x, !prof ![[MD1]]
; CHECK-NEXT: ret i32 [[SEL]]
;
%cmp = icmp slt i32 %x, 0
%sel = select i1 %cmp, i32 2147483647, i32 %x, !prof !1
ret i32 %sel
}
; FIXME:
; The compare should change, but the metadata remains the same because the select operands are not swapped.
define i32 @umax1(i32 %x) {
; CHECK-LABEL: @umax1(
; CHECK-NEXT: [[CMP:%.*]] = icmp slt i32 %x, 0
; CHECK-NEXT: [[SEL:%.*]] = select i1 [[CMP]], i32 %x, i32 2147483647, !prof ![[MD1]]
; CHECK-NEXT: ret i32 [[SEL]]
;
%cmp = icmp slt i32 %x, 0
%sel = select i1 %cmp, i32 %x, i32 2147483647, !prof !1
ret i32 %sel
}
; FIXME:
; The compare should change, and the metadata is swapped because the select operands are swapped.
define i32 @umax2(i32 %x) {
; CHECK-LABEL: @umax2(
; CHECK-NEXT: [[CMP:%.*]] = icmp sgt i32 %x, -1
; CHECK-NEXT: [[SEL:%.*]] = select i1 [[CMP]], i32 -2147483648, i32 %x, !prof ![[MD1]]
; CHECK-NEXT: ret i32 [[SEL]]
;
%cmp = icmp sgt i32 %x, -1
%sel = select i1 %cmp, i32 -2147483648, i32 %x, !prof !1
ret i32 %sel
}
[Profile] Propagate branch metadata properly in instcombine Differential Revision: http://reviews.llvm.org/D23590 llvm-svn: 279693
2016-08-25 08:26:32 +08:00
!1 = !{!"branch_weights", i32 2, i32 10}
!2 = !{!"branch_weights", i32 3, i32 10}
; CHECK-DAG: ![[MD1]] = !{!"branch_weights", i32 2, i32 10}
; CHECK-DAG: ![[MD3]] = !{!"branch_weights", i32 10, i32 2}
[InstCombine] canonicalize min/max constant to select's false value This is a first step towards canonicalization and improved folding/codegen for integer min/max as discussed here: http://lists.llvm.org/pipermail/llvm-dev/2016-November/106868.html Here, we're just matching the simplest min/max patterns and adjusting the icmp predicate while swapping the select operands. I've included FIXME tests in test/Transforms/InstCombine/select_meta.ll so it's easier to see how this might be extended (corresponds to the TODO comment in the code). That's also why I'm using matchSelectPattern() rather than a simpler check; once the backend is patched, we can just remove some of the restrictions to allow the obfuscated min/max patterns in the FIXME tests to be matched. Differential Revision: https://reviews.llvm.org/D26525 llvm-svn: 287585
2016-11-22 06:04:14 +08:00
; CHECK-DAG: ![[MD4]] = !{!"branch_weights", i32 10, i32 3}
[Profile] Propagate branch metadata properly in instcombine Differential Revision: http://reviews.llvm.org/D23590 llvm-svn: 279693
2016-08-25 08:26:32 +08:00