maxjhandsome
/
llvm-project
forked from OSchip/llvm-project
1
0
Fork 0
Code Issues Pull Requests Packages Releases Wiki Activity
llvm-project/llvm/test/CodeGen/PowerPC/swaps-le-2.ll

99 lines
3.1 KiB
LLVM
Raw Normal View History

Adding -verify-machineinstrs option to PowerPC tests Currently we have a number of tests that fail with -verify-machineinstrs. To detect this cases earlier we add the option to the testcases with the exception of tests that will currently fail with this option. PR 27456 keeps track of this failures. No code review, as discussed with Hal Finkel. llvm-svn: 277624
2016-08-04 02:17:35 +08:00
; RUN: llc -verify-machineinstrs -O3 -mcpu=pwr8 -mtriple=powerpc64le-unknown-linux-gnu < %s | FileCheck %s
[PPC64LE] Adjust vector splats during VSX swap optimization The initial code drop for VSX swap optimization permitted the optimization only when all operations in a web of related computation are lane-insensitive. For some lane-sensitive operations, we can still permit the optimization provided that we make adjustments to those operations. This patch adds special handling for vector splats so that their presence doesn't kill the optimization. Vector splats are lane-sensitive since they identify by number a vector element to be used as the source of a splat. When swap optimizations take place, the desired vector element will move to the opposite doubleword of the quadword vector. We thus replace the index I by (I + N/2) % N, where N is the number of elements in the vector. A new test case is added to test that swap optimization succeeds when vector splats are present, and that the proper input element is used as the source of the splat. An ancillary change removes SH_BUILDVEC as one of the kinds of special handling that may be required by VSX swap optimization. From experience with GCC, I had expected to need some modifications for vector build operations, but I did not find that to be the case. llvm-svn: 236606
2015-05-06 23:40:46 +08:00
; Test swap removal when a vector splat must be adjusted to make it legal.
;
[PowerPC] Emit VMX loads/stores for aligned ops to avoid adding swaps on LE Fixes PR30730. This is a re-commit of a pulled commit. The commit was pulled because some software projects contained uses of Altivec vectors that violated alignment requirements. Known issues have now been fixed. Committing on behalf of Lei Huang. Differential Revision: https://reviews.llvm.org/D26861 llvm-svn: 301892
2017-05-02 09:47:34 +08:00
; LH: 2016-11-17
; Updated align attritue from 16 to 8 to keep swap instructions tests.
; Changes have been made on little-endian to use lvx and stvx
; instructions instead of lxvd2x/xxswapd and xxswapd/stxvd2x for
; aligned vectors with elements up to 4 bytes
[PPC64LE] Adjust vector splats during VSX swap optimization The initial code drop for VSX swap optimization permitted the optimization only when all operations in a web of related computation are lane-insensitive. For some lane-sensitive operations, we can still permit the optimization provided that we make adjustments to those operations. This patch adds special handling for vector splats so that their presence doesn't kill the optimization. Vector splats are lane-sensitive since they identify by number a vector element to be used as the source of a splat. When swap optimizations take place, the desired vector element will move to the opposite doubleword of the quadword vector. We thus replace the index I by (I + N/2) % N, where N is the number of elements in the vector. A new test case is added to test that swap optimization succeeds when vector splats are present, and that the proper input element is used as the source of the splat. An ancillary change removes SH_BUILDVEC as one of the kinds of special handling that may be required by VSX swap optimization. From experience with GCC, I had expected to need some modifications for vector build operations, but I did not find that to be the case. llvm-svn: 236606
2015-05-06 23:40:46 +08:00
; Test generated from following C code:
;
; vector char vc = {0, 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15};
; vector char vcr;
; vector short vs = {0, 1, 2, 3, 4, 5, 6, 7};
; vector short vsr;
; vector int vi = {0, 1, 2, 3};
; vector int vir;
;
; void cfoo ()
; {
; vcr = (vector char){vc[5], vc[5], vc[5], vc[5], vc[5], vc[5], vc[5], vc[5],
; vc[5], vc[5], vc[5], vc[5], vc[5], vc[5], vc[5], vc[5]};
; }
;
; void sfoo ()
; {
; vsr = (vector short){vs[6], vs[6], vs[6], vs[6],
; vs[6], vs[6], vs[6], vs[6]};
; }
;
; void ifoo ()
; {
; vir = (vector int){vi[1], vi[1], vi[1], vi[1]};
; }
[PowerPC] Emit VMX loads/stores for aligned ops to avoid adding swaps on LE Fixes PR30730. This is a re-commit of a pulled commit. The commit was pulled because some software projects contained uses of Altivec vectors that violated alignment requirements. Known issues have now been fixed. Committing on behalf of Lei Huang. Differential Revision: https://reviews.llvm.org/D26861 llvm-svn: 301892
2017-05-02 09:47:34 +08:00
@vc = global <16 x i8> <i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 8, i8 9, i8 10, i8 11, i8 12, i8 13, i8 14, i8 15>, align 8
@vs = global <8 x i16> <i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7>, align 8
@vi = global <4 x i32> <i32 0, i32 1, i32 2, i32 3>, align 8
@vcr = common global <16 x i8> zeroinitializer, align 8
@vsr = common global <8 x i16> zeroinitializer, align 8
@vir = common global <4 x i32> zeroinitializer, align 8
[PPC64LE] Adjust vector splats during VSX swap optimization The initial code drop for VSX swap optimization permitted the optimization only when all operations in a web of related computation are lane-insensitive. For some lane-sensitive operations, we can still permit the optimization provided that we make adjustments to those operations. This patch adds special handling for vector splats so that their presence doesn't kill the optimization. Vector splats are lane-sensitive since they identify by number a vector element to be used as the source of a splat. When swap optimizations take place, the desired vector element will move to the opposite doubleword of the quadword vector. We thus replace the index I by (I + N/2) % N, where N is the number of elements in the vector. A new test case is added to test that swap optimization succeeds when vector splats are present, and that the proper input element is used as the source of the splat. An ancillary change removes SH_BUILDVEC as one of the kinds of special handling that may be required by VSX swap optimization. From experience with GCC, I had expected to need some modifications for vector build operations, but I did not find that to be the case. llvm-svn: 236606
2015-05-06 23:40:46 +08:00
; Function Attrs: nounwind
define void @cfoo() {
entry:
[PowerPC] Emit VMX loads/stores for aligned ops to avoid adding swaps on LE Fixes PR30730. This is a re-commit of a pulled commit. The commit was pulled because some software projects contained uses of Altivec vectors that violated alignment requirements. Known issues have now been fixed. Committing on behalf of Lei Huang. Differential Revision: https://reviews.llvm.org/D26861 llvm-svn: 301892
2017-05-02 09:47:34 +08:00
%0 = load <16 x i8>, <16 x i8>* @vc, align 8
[PPC64LE] Adjust vector splats during VSX swap optimization The initial code drop for VSX swap optimization permitted the optimization only when all operations in a web of related computation are lane-insensitive. For some lane-sensitive operations, we can still permit the optimization provided that we make adjustments to those operations. This patch adds special handling for vector splats so that their presence doesn't kill the optimization. Vector splats are lane-sensitive since they identify by number a vector element to be used as the source of a splat. When swap optimizations take place, the desired vector element will move to the opposite doubleword of the quadword vector. We thus replace the index I by (I + N/2) % N, where N is the number of elements in the vector. A new test case is added to test that swap optimization succeeds when vector splats are present, and that the proper input element is used as the source of the splat. An ancillary change removes SH_BUILDVEC as one of the kinds of special handling that may be required by VSX swap optimization. From experience with GCC, I had expected to need some modifications for vector build operations, but I did not find that to be the case. llvm-svn: 236606
2015-05-06 23:40:46 +08:00
%vecinit30 = shufflevector <16 x i8> %0, <16 x i8> undef, <16 x i32> <i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5>
[PowerPC] Emit VMX loads/stores for aligned ops to avoid adding swaps on LE Fixes PR30730. This is a re-commit of a pulled commit. The commit was pulled because some software projects contained uses of Altivec vectors that violated alignment requirements. Known issues have now been fixed. Committing on behalf of Lei Huang. Differential Revision: https://reviews.llvm.org/D26861 llvm-svn: 301892
2017-05-02 09:47:34 +08:00
store <16 x i8> %vecinit30, <16 x i8>* @vcr, align 8
[PPC64LE] Adjust vector splats during VSX swap optimization The initial code drop for VSX swap optimization permitted the optimization only when all operations in a web of related computation are lane-insensitive. For some lane-sensitive operations, we can still permit the optimization provided that we make adjustments to those operations. This patch adds special handling for vector splats so that their presence doesn't kill the optimization. Vector splats are lane-sensitive since they identify by number a vector element to be used as the source of a splat. When swap optimizations take place, the desired vector element will move to the opposite doubleword of the quadword vector. We thus replace the index I by (I + N/2) % N, where N is the number of elements in the vector. A new test case is added to test that swap optimization succeeds when vector splats are present, and that the proper input element is used as the source of the splat. An ancillary change removes SH_BUILDVEC as one of the kinds of special handling that may be required by VSX swap optimization. From experience with GCC, I had expected to need some modifications for vector build operations, but I did not find that to be the case. llvm-svn: 236606
2015-05-06 23:40:46 +08:00
ret void
}
; Function Attrs: nounwind
define void @sfoo() {
entry:
[PowerPC] Emit VMX loads/stores for aligned ops to avoid adding swaps on LE Fixes PR30730. This is a re-commit of a pulled commit. The commit was pulled because some software projects contained uses of Altivec vectors that violated alignment requirements. Known issues have now been fixed. Committing on behalf of Lei Huang. Differential Revision: https://reviews.llvm.org/D26861 llvm-svn: 301892
2017-05-02 09:47:34 +08:00
%0 = load <8 x i16>, <8 x i16>* @vs, align 8
[PPC64LE] Adjust vector splats during VSX swap optimization The initial code drop for VSX swap optimization permitted the optimization only when all operations in a web of related computation are lane-insensitive. For some lane-sensitive operations, we can still permit the optimization provided that we make adjustments to those operations. This patch adds special handling for vector splats so that their presence doesn't kill the optimization. Vector splats are lane-sensitive since they identify by number a vector element to be used as the source of a splat. When swap optimizations take place, the desired vector element will move to the opposite doubleword of the quadword vector. We thus replace the index I by (I + N/2) % N, where N is the number of elements in the vector. A new test case is added to test that swap optimization succeeds when vector splats are present, and that the proper input element is used as the source of the splat. An ancillary change removes SH_BUILDVEC as one of the kinds of special handling that may be required by VSX swap optimization. From experience with GCC, I had expected to need some modifications for vector build operations, but I did not find that to be the case. llvm-svn: 236606
2015-05-06 23:40:46 +08:00
%vecinit14 = shufflevector <8 x i16> %0, <8 x i16> undef, <8 x i32> <i32 6, i32 6, i32 6, i32 6, i32 6, i32 6, i32 6, i32 6>
[PowerPC] Emit VMX loads/stores for aligned ops to avoid adding swaps on LE Fixes PR30730. This is a re-commit of a pulled commit. The commit was pulled because some software projects contained uses of Altivec vectors that violated alignment requirements. Known issues have now been fixed. Committing on behalf of Lei Huang. Differential Revision: https://reviews.llvm.org/D26861 llvm-svn: 301892
2017-05-02 09:47:34 +08:00
store <8 x i16> %vecinit14, <8 x i16>* @vsr, align 8
[PPC64LE] Adjust vector splats during VSX swap optimization The initial code drop for VSX swap optimization permitted the optimization only when all operations in a web of related computation are lane-insensitive. For some lane-sensitive operations, we can still permit the optimization provided that we make adjustments to those operations. This patch adds special handling for vector splats so that their presence doesn't kill the optimization. Vector splats are lane-sensitive since they identify by number a vector element to be used as the source of a splat. When swap optimizations take place, the desired vector element will move to the opposite doubleword of the quadword vector. We thus replace the index I by (I + N/2) % N, where N is the number of elements in the vector. A new test case is added to test that swap optimization succeeds when vector splats are present, and that the proper input element is used as the source of the splat. An ancillary change removes SH_BUILDVEC as one of the kinds of special handling that may be required by VSX swap optimization. From experience with GCC, I had expected to need some modifications for vector build operations, but I did not find that to be the case. llvm-svn: 236606
2015-05-06 23:40:46 +08:00
ret void
}
; Function Attrs: nounwind
define void @ifoo() {
entry:
[PowerPC] Emit VMX loads/stores for aligned ops to avoid adding swaps on LE Fixes PR30730. This is a re-commit of a pulled commit. The commit was pulled because some software projects contained uses of Altivec vectors that violated alignment requirements. Known issues have now been fixed. Committing on behalf of Lei Huang. Differential Revision: https://reviews.llvm.org/D26861 llvm-svn: 301892
2017-05-02 09:47:34 +08:00
%0 = load <4 x i32>, <4 x i32>* @vi, align 8
[PPC64LE] Adjust vector splats during VSX swap optimization The initial code drop for VSX swap optimization permitted the optimization only when all operations in a web of related computation are lane-insensitive. For some lane-sensitive operations, we can still permit the optimization provided that we make adjustments to those operations. This patch adds special handling for vector splats so that their presence doesn't kill the optimization. Vector splats are lane-sensitive since they identify by number a vector element to be used as the source of a splat. When swap optimizations take place, the desired vector element will move to the opposite doubleword of the quadword vector. We thus replace the index I by (I + N/2) % N, where N is the number of elements in the vector. A new test case is added to test that swap optimization succeeds when vector splats are present, and that the proper input element is used as the source of the splat. An ancillary change removes SH_BUILDVEC as one of the kinds of special handling that may be required by VSX swap optimization. From experience with GCC, I had expected to need some modifications for vector build operations, but I did not find that to be the case. llvm-svn: 236606
2015-05-06 23:40:46 +08:00
%vecinit6 = shufflevector <4 x i32> %0, <4 x i32> undef, <4 x i32> <i32 1, i32 1, i32 1, i32 1>
[PowerPC] Emit VMX loads/stores for aligned ops to avoid adding swaps on LE Fixes PR30730. This is a re-commit of a pulled commit. The commit was pulled because some software projects contained uses of Altivec vectors that violated alignment requirements. Known issues have now been fixed. Committing on behalf of Lei Huang. Differential Revision: https://reviews.llvm.org/D26861 llvm-svn: 301892
2017-05-02 09:47:34 +08:00
store <4 x i32> %vecinit6, <4 x i32>* @vir, align 8
[PPC64LE] Adjust vector splats during VSX swap optimization The initial code drop for VSX swap optimization permitted the optimization only when all operations in a web of related computation are lane-insensitive. For some lane-sensitive operations, we can still permit the optimization provided that we make adjustments to those operations. This patch adds special handling for vector splats so that their presence doesn't kill the optimization. Vector splats are lane-sensitive since they identify by number a vector element to be used as the source of a splat. When swap optimizations take place, the desired vector element will move to the opposite doubleword of the quadword vector. We thus replace the index I by (I + N/2) % N, where N is the number of elements in the vector. A new test case is added to test that swap optimization succeeds when vector splats are present, and that the proper input element is used as the source of the splat. An ancillary change removes SH_BUILDVEC as one of the kinds of special handling that may be required by VSX swap optimization. From experience with GCC, I had expected to need some modifications for vector build operations, but I did not find that to be the case. llvm-svn: 236606
2015-05-06 23:40:46 +08:00
ret void
}
; Justification:
; Byte splat of element 5 (BE) becomes element 15-5 = 10 (LE)
; which becomes (10+8)%16 = 2 (LE swapped).
;
; Halfword splat of element 6 (BE) becomes element 7-6 = 1 (LE)
; which becomes (1+4)%8 = 5 (LE swapped).
;
; Word splat of element 1 (BE) becomes element 3-1 = 2 (LE)
; which becomes (2+2)%4 = 0 (LE swapped).
; CHECK-NOT: xxpermdi
; CHECK-NOT: xxswapd
; CHECK-LABEL: @cfoo
; CHECK: lxvd2x
; CHECK: vspltb {{[0-9]+}}, {{[0-9]+}}, 2
; CHECK: stxvd2x
; CHECK-LABEL: @sfoo
; CHECK: lxvd2x
; CHECK: vsplth {{[0-9]+}}, {{[0-9]+}}, 5
; CHECK: stxvd2x
; CHECK-LABEL: @ifoo
; CHECK: lxvd2x
[PowerPC] Generate VSX version of splat word This patch corresponds to review: http://reviews.llvm.org/D18592 It allows the PPC back end to generate the xxspltw instruction where we previously only emitted vspltw. llvm-svn: 268516
2016-05-05 00:04:02 +08:00
; CHECK: xxspltw {{[0-9]+}}, {{[0-9]+}}, 0
[PPC64LE] Adjust vector splats during VSX swap optimization The initial code drop for VSX swap optimization permitted the optimization only when all operations in a web of related computation are lane-insensitive. For some lane-sensitive operations, we can still permit the optimization provided that we make adjustments to those operations. This patch adds special handling for vector splats so that their presence doesn't kill the optimization. Vector splats are lane-sensitive since they identify by number a vector element to be used as the source of a splat. When swap optimizations take place, the desired vector element will move to the opposite doubleword of the quadword vector. We thus replace the index I by (I + N/2) % N, where N is the number of elements in the vector. A new test case is added to test that swap optimization succeeds when vector splats are present, and that the proper input element is used as the source of the splat. An ancillary change removes SH_BUILDVEC as one of the kinds of special handling that may be required by VSX swap optimization. From experience with GCC, I had expected to need some modifications for vector build operations, but I did not find that to be the case. llvm-svn: 236606
2015-05-06 23:40:46 +08:00
; CHECK: stxvd2x