6502/llvm-6502
1
0
Fork 0
mirror of https://github.com/c64scene-ar/llvm-6502.git synced 2024-12-21 00:32:23 +00:00
Code Issues Projects Releases Wiki Activity
b0b708854e
llvm-6502/test/CodeGen/Mips/msa/bitwise.ll

1641 lines
48 KiB
LLVM
Raw Normal View History

[mips][msa] MSA requires FR=1 mode (64-bit FPU register file). Report fatal error when using it in FR=0 mode. git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@191498 91177308-0d34-0410-b5e6-96231b3b80d8
2013-09-27 10:08:31 +00:00
; RUN: llc -march=mips -mattr=+msa,+fp64 < %s | FileCheck %s
[mips][msa] Build all the tests in little and big endian modes and correct an incorrect test. Summary: This patch (correctly) breaks some MSA tests by exposing the cases when SelectionDAG::getConstant() produces illegal types. These have been temporarily marked XFAIL and the XFAIL flag will be removed when SelectionDAG::getConstant() is fixed. There are three categories of failure: * Immediate instructions are not selected in one endian mode. * Immediates used in ldi.[bhwd] must be different according to endianness. (this only affects cases where the 'wrong' ldi is used to load the correct bitpattern. E.g. (bitcast:v2i64 (build_vector:v4i32 ...))) * Non-immediate instructions that rely on immediates affected by the previous two categories as part of their match pattern. For example, the bset match pattern is the vector equivalent of 'ws | (1 << wt)'. One test needed correcting to expect different output depending on whether big or little endian was in use. This test was test/CodeGen/Mips/msa/basic_operations.ll and experiences the second category of failure shown above. The little endian version of this test is named basic_operations_little.ll and will be merged back into basic_operations.ll in a follow up commit now that FileCheck supports multiple check prefixes. Reviewers: bkramer, jacksprat, dsanders Reviewed By: dsanders CC: llvm-commits Differential Revision: http://llvm-reviews.chandlerc.com/D1972 git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@194806 91177308-0d34-0410-b5e6-96231b3b80d8
2013-11-15 11:04:16 +00:00
; RUN: llc -march=mipsel -mattr=+msa,+fp64 < %s | FileCheck %s
[mips][msa] Added test cases that were supposed to be part of r190507, r190509, r190512, and r190518. git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@190522 91177308-0d34-0410-b5e6-96231b3b80d8
2013-09-11 12:39:25 +00:00
[mips][msa] Added support for matching and, or, and xor from normal IR (i.e. not intrinsics) git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@191194 91177308-0d34-0410-b5e6-96231b3b80d8
2013-09-23 12:57:42 +00:00
define void @and_v16i8(<16 x i8>* %c, <16 x i8>* %a, <16 x i8>* %b) nounwind {
; CHECK: and_v16i8:
%1 = load <16 x i8>* %a
; CHECK-DAG: ld.b [[R1:\$w[0-9]+]], 0($5)
%2 = load <16 x i8>* %b
; CHECK-DAG: ld.b [[R2:\$w[0-9]+]], 0($6)
%3 = and <16 x i8> %1, %2
; CHECK-DAG: and.v [[R3:\$w[0-9]+]], [[R1]], [[R2]]
store <16 x i8> %3, <16 x i8>* %c
; CHECK-DAG: st.b [[R3]], 0($4)
ret void
; CHECK: .size and_v16i8
}
define void @and_v8i16(<8 x i16>* %c, <8 x i16>* %a, <8 x i16>* %b) nounwind {
; CHECK: and_v8i16:
%1 = load <8 x i16>* %a
; CHECK-DAG: ld.h [[R1:\$w[0-9]+]], 0($5)
%2 = load <8 x i16>* %b
; CHECK-DAG: ld.h [[R2:\$w[0-9]+]], 0($6)
%3 = and <8 x i16> %1, %2
; CHECK-DAG: and.v [[R3:\$w[0-9]+]], [[R1]], [[R2]]
store <8 x i16> %3, <8 x i16>* %c
; CHECK-DAG: st.h [[R3]], 0($4)
ret void
; CHECK: .size and_v8i16
}
define void @and_v4i32(<4 x i32>* %c, <4 x i32>* %a, <4 x i32>* %b) nounwind {
; CHECK: and_v4i32:
%1 = load <4 x i32>* %a
; CHECK-DAG: ld.w [[R1:\$w[0-9]+]], 0($5)
%2 = load <4 x i32>* %b
; CHECK-DAG: ld.w [[R2:\$w[0-9]+]], 0($6)
%3 = and <4 x i32> %1, %2
; CHECK-DAG: and.v [[R3:\$w[0-9]+]], [[R1]], [[R2]]
store <4 x i32> %3, <4 x i32>* %c
; CHECK-DAG: st.w [[R3]], 0($4)
ret void
; CHECK: .size and_v4i32
}
define void @and_v2i64(<2 x i64>* %c, <2 x i64>* %a, <2 x i64>* %b) nounwind {
; CHECK: and_v2i64:
%1 = load <2 x i64>* %a
; CHECK-DAG: ld.d [[R1:\$w[0-9]+]], 0($5)
%2 = load <2 x i64>* %b
; CHECK-DAG: ld.d [[R2:\$w[0-9]+]], 0($6)
%3 = and <2 x i64> %1, %2
; CHECK-DAG: and.v [[R3:\$w[0-9]+]], [[R1]], [[R2]]
store <2 x i64> %3, <2 x i64>* %c
; CHECK-DAG: st.d [[R3]], 0($4)
ret void
; CHECK: .size and_v2i64
}
[mips][msa] Added support for matching andi, ori, nori, and xori from normal IR (i.e. not intrinsics) git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@191293 91177308-0d34-0410-b5e6-96231b3b80d8
2013-09-24 12:32:47 +00:00
define void @and_v16i8_i(<16 x i8>* %c, <16 x i8>* %a) nounwind {
; CHECK: and_v16i8_i:
%1 = load <16 x i8>* %a
; CHECK-DAG: ld.b [[R1:\$w[0-9]+]], 0($5)
%2 = and <16 x i8> %1, <i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1>
; CHECK-DAG: andi.b [[R4:\$w[0-9]+]], [[R1]], 1
store <16 x i8> %2, <16 x i8>* %c
; CHECK-DAG: st.b [[R4]], 0($4)
ret void
; CHECK: .size and_v16i8_i
}
define void @and_v8i16_i(<8 x i16>* %c, <8 x i16>* %a) nounwind {
; CHECK: and_v8i16_i:
%1 = load <8 x i16>* %a
; CHECK-DAG: ld.h [[R1:\$w[0-9]+]], 0($5)
%2 = and <8 x i16> %1, <i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1>
; CHECK-DAG: ldi.h [[R3:\$w[0-9]+]], 1
; CHECK-DAG: and.v [[R4:\$w[0-9]+]], [[R1]], [[R3]]
store <8 x i16> %2, <8 x i16>* %c
; CHECK-DAG: st.h [[R4]], 0($4)
ret void
; CHECK: .size and_v8i16_i
}
define void @and_v4i32_i(<4 x i32>* %c, <4 x i32>* %a) nounwind {
; CHECK: and_v4i32_i:
%1 = load <4 x i32>* %a
; CHECK-DAG: ld.w [[R1:\$w[0-9]+]], 0($5)
%2 = and <4 x i32> %1, <i32 1, i32 1, i32 1, i32 1>
; CHECK-DAG: ldi.w [[R3:\$w[0-9]+]], 1
; CHECK-DAG: and.v [[R4:\$w[0-9]+]], [[R1]], [[R3]]
store <4 x i32> %2, <4 x i32>* %c
; CHECK-DAG: st.w [[R4]], 0($4)
ret void
; CHECK: .size and_v4i32_i
}
define void @and_v2i64_i(<2 x i64>* %c, <2 x i64>* %a) nounwind {
; CHECK: and_v2i64_i:
%1 = load <2 x i64>* %a
; CHECK-DAG: ld.d [[R1:\$w[0-9]+]], 0($5)
%2 = and <2 x i64> %1, <i64 1, i64 1>
; CHECK-DAG: ldi.d [[R3:\$w[0-9]+]], 1
; CHECK-DAG: and.v [[R4:\$w[0-9]+]], [[R1]], [[R3]]
store <2 x i64> %2, <2 x i64>* %c
; CHECK-DAG: st.d [[R4]], 0($4)
ret void
; CHECK: .size and_v2i64_i
}
[mips][msa] Added support for matching and, or, and xor from normal IR (i.e. not intrinsics) git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@191194 91177308-0d34-0410-b5e6-96231b3b80d8
2013-09-23 12:57:42 +00:00
define void @or_v16i8(<16 x i8>* %c, <16 x i8>* %a, <16 x i8>* %b) nounwind {
; CHECK: or_v16i8:
%1 = load <16 x i8>* %a
; CHECK-DAG: ld.b [[R1:\$w[0-9]+]], 0($5)
%2 = load <16 x i8>* %b
; CHECK-DAG: ld.b [[R2:\$w[0-9]+]], 0($6)
%3 = or <16 x i8> %1, %2
; CHECK-DAG: or.v [[R3:\$w[0-9]+]], [[R1]], [[R2]]
store <16 x i8> %3, <16 x i8>* %c
; CHECK-DAG: st.b [[R3]], 0($4)
ret void
; CHECK: .size or_v16i8
}
define void @or_v8i16(<8 x i16>* %c, <8 x i16>* %a, <8 x i16>* %b) nounwind {
; CHECK: or_v8i16:
%1 = load <8 x i16>* %a
; CHECK-DAG: ld.h [[R1:\$w[0-9]+]], 0($5)
%2 = load <8 x i16>* %b
; CHECK-DAG: ld.h [[R2:\$w[0-9]+]], 0($6)
%3 = or <8 x i16> %1, %2
; CHECK-DAG: or.v [[R3:\$w[0-9]+]], [[R1]], [[R2]]
store <8 x i16> %3, <8 x i16>* %c
; CHECK-DAG: st.h [[R3]], 0($4)
ret void
; CHECK: .size or_v8i16
}
define void @or_v4i32(<4 x i32>* %c, <4 x i32>* %a, <4 x i32>* %b) nounwind {
; CHECK: or_v4i32:
%1 = load <4 x i32>* %a
; CHECK-DAG: ld.w [[R1:\$w[0-9]+]], 0($5)
%2 = load <4 x i32>* %b
; CHECK-DAG: ld.w [[R2:\$w[0-9]+]], 0($6)
%3 = or <4 x i32> %1, %2
; CHECK-DAG: or.v [[R3:\$w[0-9]+]], [[R1]], [[R2]]
store <4 x i32> %3, <4 x i32>* %c
; CHECK-DAG: st.w [[R3]], 0($4)
ret void
; CHECK: .size or_v4i32
}
define void @or_v2i64(<2 x i64>* %c, <2 x i64>* %a, <2 x i64>* %b) nounwind {
; CHECK: or_v2i64:
%1 = load <2 x i64>* %a
; CHECK-DAG: ld.d [[R1:\$w[0-9]+]], 0($5)
%2 = load <2 x i64>* %b
; CHECK-DAG: ld.d [[R2:\$w[0-9]+]], 0($6)
%3 = or <2 x i64> %1, %2
; CHECK-DAG: or.v [[R3:\$w[0-9]+]], [[R1]], [[R2]]
store <2 x i64> %3, <2 x i64>* %c
; CHECK-DAG: st.d [[R3]], 0($4)
ret void
; CHECK: .size or_v2i64
}
[mips][msa] Added support for matching andi, ori, nori, and xori from normal IR (i.e. not intrinsics) git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@191293 91177308-0d34-0410-b5e6-96231b3b80d8
2013-09-24 12:32:47 +00:00
define void @or_v16i8_i(<16 x i8>* %c, <16 x i8>* %a) nounwind {
; CHECK: or_v16i8_i:
%1 = load <16 x i8>* %a
; CHECK-DAG: ld.b [[R1:\$w[0-9]+]], 0($5)
[mips][msa] Change constant used in ori tests to avoid conflict with bseti (also xori to avoid bnegi) Upcoming commit(s) are going to add support for bseti and bnegi. This would cause some existing tests to (correctly) change behaviour and emit a different instruction. This patch prevents this by changing the constant used in ori and xori tests so that they will not be matchable by the bseti and bnegi patterns when these instructions are matchable from normal IR. git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@194467 91177308-0d34-0410-b5e6-96231b3b80d8
2013-11-12 10:14:18 +00:00
%2 = or <16 x i8> %1, <i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3>
; CHECK-DAG: ori.b [[R4:\$w[0-9]+]], [[R1]], 3
[mips][msa] Added support for matching andi, ori, nori, and xori from normal IR (i.e. not intrinsics) git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@191293 91177308-0d34-0410-b5e6-96231b3b80d8
2013-09-24 12:32:47 +00:00
store <16 x i8> %2, <16 x i8>* %c
; CHECK-DAG: st.b [[R4]], 0($4)
ret void
; CHECK: .size or_v16i8_i
}
define void @or_v8i16_i(<8 x i16>* %c, <8 x i16>* %a) nounwind {
; CHECK: or_v8i16_i:
%1 = load <8 x i16>* %a
; CHECK-DAG: ld.h [[R1:\$w[0-9]+]], 0($5)
[mips][msa] Change constant used in ori tests to avoid conflict with bseti (also xori to avoid bnegi) Upcoming commit(s) are going to add support for bseti and bnegi. This would cause some existing tests to (correctly) change behaviour and emit a different instruction. This patch prevents this by changing the constant used in ori and xori tests so that they will not be matchable by the bseti and bnegi patterns when these instructions are matchable from normal IR. git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@194467 91177308-0d34-0410-b5e6-96231b3b80d8
2013-11-12 10:14:18 +00:00
%2 = or <8 x i16> %1, <i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3>
; CHECK-DAG: ldi.h [[R3:\$w[0-9]+]], 3
[mips][msa] Added support for matching andi, ori, nori, and xori from normal IR (i.e. not intrinsics) git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@191293 91177308-0d34-0410-b5e6-96231b3b80d8
2013-09-24 12:32:47 +00:00
; CHECK-DAG: or.v [[R4:\$w[0-9]+]], [[R1]], [[R3]]
store <8 x i16> %2, <8 x i16>* %c
; CHECK-DAG: st.h [[R4]], 0($4)
ret void
; CHECK: .size or_v8i16_i
}
define void @or_v4i32_i(<4 x i32>* %c, <4 x i32>* %a) nounwind {
; CHECK: or_v4i32_i:
%1 = load <4 x i32>* %a
; CHECK-DAG: ld.w [[R1:\$w[0-9]+]], 0($5)
[mips][msa] Change constant used in ori tests to avoid conflict with bseti (also xori to avoid bnegi) Upcoming commit(s) are going to add support for bseti and bnegi. This would cause some existing tests to (correctly) change behaviour and emit a different instruction. This patch prevents this by changing the constant used in ori and xori tests so that they will not be matchable by the bseti and bnegi patterns when these instructions are matchable from normal IR. git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@194467 91177308-0d34-0410-b5e6-96231b3b80d8
2013-11-12 10:14:18 +00:00
%2 = or <4 x i32> %1, <i32 3, i32 3, i32 3, i32 3>
; CHECK-DAG: ldi.w [[R3:\$w[0-9]+]], 3
[mips][msa] Added support for matching andi, ori, nori, and xori from normal IR (i.e. not intrinsics) git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@191293 91177308-0d34-0410-b5e6-96231b3b80d8
2013-09-24 12:32:47 +00:00
; CHECK-DAG: or.v [[R4:\$w[0-9]+]], [[R1]], [[R3]]
store <4 x i32> %2, <4 x i32>* %c
; CHECK-DAG: st.w [[R4]], 0($4)
ret void
; CHECK: .size or_v4i32_i
}
define void @or_v2i64_i(<2 x i64>* %c, <2 x i64>* %a) nounwind {
; CHECK: or_v2i64_i:
%1 = load <2 x i64>* %a
; CHECK-DAG: ld.d [[R1:\$w[0-9]+]], 0($5)
[mips][msa] Change constant used in ori tests to avoid conflict with bseti (also xori to avoid bnegi) Upcoming commit(s) are going to add support for bseti and bnegi. This would cause some existing tests to (correctly) change behaviour and emit a different instruction. This patch prevents this by changing the constant used in ori and xori tests so that they will not be matchable by the bseti and bnegi patterns when these instructions are matchable from normal IR. git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@194467 91177308-0d34-0410-b5e6-96231b3b80d8
2013-11-12 10:14:18 +00:00
%2 = or <2 x i64> %1, <i64 3, i64 3>
; CHECK-DAG: ldi.d [[R3:\$w[0-9]+]], 3
[mips][msa] Added support for matching andi, ori, nori, and xori from normal IR (i.e. not intrinsics) git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@191293 91177308-0d34-0410-b5e6-96231b3b80d8
2013-09-24 12:32:47 +00:00
; CHECK-DAG: or.v [[R4:\$w[0-9]+]], [[R1]], [[R3]]
store <2 x i64> %2, <2 x i64>* %c
; CHECK-DAG: st.d [[R4]], 0($4)
ret void
; CHECK: .size or_v2i64_i
}
[mips][msa] Added support for matching nor from normal IR (i.e. not intrinsics) git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@191195 91177308-0d34-0410-b5e6-96231b3b80d8
2013-09-23 13:22:24 +00:00
define void @nor_v16i8(<16 x i8>* %c, <16 x i8>* %a, <16 x i8>* %b) nounwind {
; CHECK: nor_v16i8:
%1 = load <16 x i8>* %a
; CHECK-DAG: ld.b [[R1:\$w[0-9]+]], 0($5)
%2 = load <16 x i8>* %b
; CHECK-DAG: ld.b [[R2:\$w[0-9]+]], 0($6)
%3 = or <16 x i8> %1, %2
%4 = xor <16 x i8> %3, <i8 -1, i8 -1, i8 -1, i8 -1, i8 -1, i8 -1, i8 -1, i8 -1, i8 -1, i8 -1, i8 -1, i8 -1, i8 -1, i8 -1, i8 -1, i8 -1>
; CHECK-DAG: nor.v [[R3:\$w[0-9]+]], [[R1]], [[R2]]
store <16 x i8> %4, <16 x i8>* %c
; CHECK-DAG: st.b [[R3]], 0($4)
ret void
; CHECK: .size nor_v16i8
}
define void @nor_v8i16(<8 x i16>* %c, <8 x i16>* %a, <8 x i16>* %b) nounwind {
; CHECK: nor_v8i16:
%1 = load <8 x i16>* %a
; CHECK-DAG: ld.h [[R1:\$w[0-9]+]], 0($5)
%2 = load <8 x i16>* %b
; CHECK-DAG: ld.h [[R2:\$w[0-9]+]], 0($6)
%3 = or <8 x i16> %1, %2
%4 = xor <8 x i16> %3, <i16 -1, i16 -1, i16 -1, i16 -1, i16 -1, i16 -1, i16 -1, i16 -1>
; CHECK-DAG: nor.v [[R3:\$w[0-9]+]], [[R1]], [[R2]]
store <8 x i16> %4, <8 x i16>* %c
; CHECK-DAG: st.h [[R3]], 0($4)
ret void
; CHECK: .size nor_v8i16
}
define void @nor_v4i32(<4 x i32>* %c, <4 x i32>* %a, <4 x i32>* %b) nounwind {
; CHECK: nor_v4i32:
%1 = load <4 x i32>* %a
; CHECK-DAG: ld.w [[R1:\$w[0-9]+]], 0($5)
%2 = load <4 x i32>* %b
; CHECK-DAG: ld.w [[R2:\$w[0-9]+]], 0($6)
%3 = or <4 x i32> %1, %2
%4 = xor <4 x i32> %3, <i32 -1, i32 -1, i32 -1, i32 -1>
; CHECK-DAG: nor.v [[R3:\$w[0-9]+]], [[R1]], [[R2]]
store <4 x i32> %4, <4 x i32>* %c
; CHECK-DAG: st.w [[R3]], 0($4)
ret void
; CHECK: .size nor_v4i32
}
define void @nor_v2i64(<2 x i64>* %c, <2 x i64>* %a, <2 x i64>* %b) nounwind {
; CHECK: nor_v2i64:
%1 = load <2 x i64>* %a
; CHECK-DAG: ld.d [[R1:\$w[0-9]+]], 0($5)
%2 = load <2 x i64>* %b
; CHECK-DAG: ld.d [[R2:\$w[0-9]+]], 0($6)
%3 = or <2 x i64> %1, %2
%4 = xor <2 x i64> %3, <i64 -1, i64 -1>
; CHECK-DAG: nor.v [[R3:\$w[0-9]+]], [[R1]], [[R2]]
store <2 x i64> %4, <2 x i64>* %c
; CHECK-DAG: st.d [[R3]], 0($4)
ret void
; CHECK: .size nor_v2i64
}
[mips][msa] Added support for matching andi, ori, nori, and xori from normal IR (i.e. not intrinsics) git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@191293 91177308-0d34-0410-b5e6-96231b3b80d8
2013-09-24 12:32:47 +00:00
define void @nor_v16i8_i(<16 x i8>* %c, <16 x i8>* %a) nounwind {
; CHECK: nor_v16i8_i:
%1 = load <16 x i8>* %a
; CHECK-DAG: ld.b [[R1:\$w[0-9]+]], 0($5)
%2 = or <16 x i8> %1, <i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1>
%3 = xor <16 x i8> %2, <i8 -1, i8 -1, i8 -1, i8 -1, i8 -1, i8 -1, i8 -1, i8 -1, i8 -1, i8 -1, i8 -1, i8 -1, i8 -1, i8 -1, i8 -1, i8 -1>
; CHECK-DAG: ori.b [[R4:\$w[0-9]+]], [[R1]], 1
store <16 x i8> %3, <16 x i8>* %c
; CHECK-DAG: st.b [[R4]], 0($4)
ret void
; CHECK: .size nor_v16i8_i
}
define void @nor_v8i16_i(<8 x i16>* %c, <8 x i16>* %a) nounwind {
; CHECK: nor_v8i16_i:
%1 = load <8 x i16>* %a
; CHECK-DAG: ld.h [[R1:\$w[0-9]+]], 0($5)
%2 = or <8 x i16> %1, <i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1>
%3 = xor <8 x i16> %2, <i16 -1, i16 -1, i16 -1, i16 -1, i16 -1, i16 -1, i16 -1, i16 -1>
; CHECK-DAG: ldi.h [[R3:\$w[0-9]+]], 1
; CHECK-DAG: nor.v [[R4:\$w[0-9]+]], [[R1]], [[R3]]
store <8 x i16> %3, <8 x i16>* %c
; CHECK-DAG: st.h [[R4]], 0($4)
ret void
; CHECK: .size nor_v8i16_i
}
define void @nor_v4i32_i(<4 x i32>* %c, <4 x i32>* %a) nounwind {
; CHECK: nor_v4i32_i:
%1 = load <4 x i32>* %a
; CHECK-DAG: ld.w [[R1:\$w[0-9]+]], 0($5)
%2 = or <4 x i32> %1, <i32 1, i32 1, i32 1, i32 1>
%3 = xor <4 x i32> %2, <i32 -1, i32 -1, i32 -1, i32 -1>
; CHECK-DAG: ldi.w [[R3:\$w[0-9]+]], 1
; CHECK-DAG: nor.v [[R4:\$w[0-9]+]], [[R1]], [[R3]]
store <4 x i32> %3, <4 x i32>* %c
; CHECK-DAG: st.w [[R4]], 0($4)
ret void
; CHECK: .size nor_v4i32_i
}
define void @nor_v2i64_i(<2 x i64>* %c, <2 x i64>* %a) nounwind {
; CHECK: nor_v2i64_i:
%1 = load <2 x i64>* %a
; CHECK-DAG: ld.d [[R1:\$w[0-9]+]], 0($5)
%2 = or <2 x i64> %1, <i64 1, i64 1>
%3 = xor <2 x i64> %2, <i64 -1, i64 -1>
; CHECK-DAG: ldi.d [[R3:\$w[0-9]+]], 1
; CHECK-DAG: nor.v [[R4:\$w[0-9]+]], [[R1]], [[R3]]
store <2 x i64> %3, <2 x i64>* %c
; CHECK-DAG: st.d [[R4]], 0($4)
ret void
; CHECK: .size nor_v2i64_i
}
[mips][msa] Added support for matching and, or, and xor from normal IR (i.e. not intrinsics) git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@191194 91177308-0d34-0410-b5e6-96231b3b80d8
2013-09-23 12:57:42 +00:00
define void @xor_v16i8(<16 x i8>* %c, <16 x i8>* %a, <16 x i8>* %b) nounwind {
; CHECK: xor_v16i8:
%1 = load <16 x i8>* %a
; CHECK-DAG: ld.b [[R1:\$w[0-9]+]], 0($5)
%2 = load <16 x i8>* %b
; CHECK-DAG: ld.b [[R2:\$w[0-9]+]], 0($6)
%3 = xor <16 x i8> %1, %2
; CHECK-DAG: xor.v [[R3:\$w[0-9]+]], [[R1]], [[R2]]
store <16 x i8> %3, <16 x i8>* %c
; CHECK-DAG: st.b [[R3]], 0($4)
ret void
; CHECK: .size xor_v16i8
}
define void @xor_v8i16(<8 x i16>* %c, <8 x i16>* %a, <8 x i16>* %b) nounwind {
; CHECK: xor_v8i16:
%1 = load <8 x i16>* %a
; CHECK-DAG: ld.h [[R1:\$w[0-9]+]], 0($5)
%2 = load <8 x i16>* %b
; CHECK-DAG: ld.h [[R2:\$w[0-9]+]], 0($6)
%3 = xor <8 x i16> %1, %2
; CHECK-DAG: xor.v [[R3:\$w[0-9]+]], [[R1]], [[R2]]
store <8 x i16> %3, <8 x i16>* %c
; CHECK-DAG: st.h [[R3]], 0($4)
ret void
; CHECK: .size xor_v8i16
}
define void @xor_v4i32(<4 x i32>* %c, <4 x i32>* %a, <4 x i32>* %b) nounwind {
; CHECK: xor_v4i32:
%1 = load <4 x i32>* %a
; CHECK-DAG: ld.w [[R1:\$w[0-9]+]], 0($5)
%2 = load <4 x i32>* %b
; CHECK-DAG: ld.w [[R2:\$w[0-9]+]], 0($6)
%3 = xor <4 x i32> %1, %2
; CHECK-DAG: xor.v [[R3:\$w[0-9]+]], [[R1]], [[R2]]
store <4 x i32> %3, <4 x i32>* %c
; CHECK-DAG: st.w [[R3]], 0($4)
ret void
; CHECK: .size xor_v4i32
}
define void @xor_v2i64(<2 x i64>* %c, <2 x i64>* %a, <2 x i64>* %b) nounwind {
; CHECK: xor_v2i64:
%1 = load <2 x i64>* %a
; CHECK-DAG: ld.d [[R1:\$w[0-9]+]], 0($5)
%2 = load <2 x i64>* %b
; CHECK-DAG: ld.d [[R2:\$w[0-9]+]], 0($6)
%3 = xor <2 x i64> %1, %2
; CHECK-DAG: xor.v [[R3:\$w[0-9]+]], [[R1]], [[R2]]
store <2 x i64> %3, <2 x i64>* %c
; CHECK-DAG: st.d [[R3]], 0($4)
ret void
; CHECK: .size xor_v2i64
}
[mips][msa] Added support for matching andi, ori, nori, and xori from normal IR (i.e. not intrinsics) git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@191293 91177308-0d34-0410-b5e6-96231b3b80d8
2013-09-24 12:32:47 +00:00
define void @xor_v16i8_i(<16 x i8>* %c, <16 x i8>* %a) nounwind {
; CHECK: xor_v16i8_i:
%1 = load <16 x i8>* %a
; CHECK-DAG: ld.b [[R1:\$w[0-9]+]], 0($5)
[mips][msa] Change constant used in ori tests to avoid conflict with bseti (also xori to avoid bnegi) Upcoming commit(s) are going to add support for bseti and bnegi. This would cause some existing tests to (correctly) change behaviour and emit a different instruction. This patch prevents this by changing the constant used in ori and xori tests so that they will not be matchable by the bseti and bnegi patterns when these instructions are matchable from normal IR. git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@194467 91177308-0d34-0410-b5e6-96231b3b80d8
2013-11-12 10:14:18 +00:00
%2 = xor <16 x i8> %1, <i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3>
; CHECK-DAG: xori.b [[R4:\$w[0-9]+]], [[R1]], 3
[mips][msa] Added support for matching andi, ori, nori, and xori from normal IR (i.e. not intrinsics) git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@191293 91177308-0d34-0410-b5e6-96231b3b80d8
2013-09-24 12:32:47 +00:00
store <16 x i8> %2, <16 x i8>* %c
; CHECK-DAG: st.b [[R4]], 0($4)
ret void
; CHECK: .size xor_v16i8_i
}
define void @xor_v8i16_i(<8 x i16>* %c, <8 x i16>* %a) nounwind {
; CHECK: xor_v8i16_i:
%1 = load <8 x i16>* %a
; CHECK-DAG: ld.h [[R1:\$w[0-9]+]], 0($5)
[mips][msa] Change constant used in ori tests to avoid conflict with bseti (also xori to avoid bnegi) Upcoming commit(s) are going to add support for bseti and bnegi. This would cause some existing tests to (correctly) change behaviour and emit a different instruction. This patch prevents this by changing the constant used in ori and xori tests so that they will not be matchable by the bseti and bnegi patterns when these instructions are matchable from normal IR. git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@194467 91177308-0d34-0410-b5e6-96231b3b80d8
2013-11-12 10:14:18 +00:00
%2 = xor <8 x i16> %1, <i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3>
; CHECK-DAG: ldi.h [[R3:\$w[0-9]+]], 3
[mips][msa] Added support for matching andi, ori, nori, and xori from normal IR (i.e. not intrinsics) git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@191293 91177308-0d34-0410-b5e6-96231b3b80d8
2013-09-24 12:32:47 +00:00
; CHECK-DAG: xor.v [[R4:\$w[0-9]+]], [[R1]], [[R3]]
store <8 x i16> %2, <8 x i16>* %c
; CHECK-DAG: st.h [[R4]], 0($4)
ret void
; CHECK: .size xor_v8i16_i
}
define void @xor_v4i32_i(<4 x i32>* %c, <4 x i32>* %a) nounwind {
; CHECK: xor_v4i32_i:
%1 = load <4 x i32>* %a
; CHECK-DAG: ld.w [[R1:\$w[0-9]+]], 0($5)
[mips][msa] Change constant used in ori tests to avoid conflict with bseti (also xori to avoid bnegi) Upcoming commit(s) are going to add support for bseti and bnegi. This would cause some existing tests to (correctly) change behaviour and emit a different instruction. This patch prevents this by changing the constant used in ori and xori tests so that they will not be matchable by the bseti and bnegi patterns when these instructions are matchable from normal IR. git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@194467 91177308-0d34-0410-b5e6-96231b3b80d8
2013-11-12 10:14:18 +00:00
%2 = xor <4 x i32> %1, <i32 3, i32 3, i32 3, i32 3>
; CHECK-DAG: ldi.w [[R3:\$w[0-9]+]], 3
[mips][msa] Added support for matching andi, ori, nori, and xori from normal IR (i.e. not intrinsics) git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@191293 91177308-0d34-0410-b5e6-96231b3b80d8
2013-09-24 12:32:47 +00:00
; CHECK-DAG: xor.v [[R4:\$w[0-9]+]], [[R1]], [[R3]]
store <4 x i32> %2, <4 x i32>* %c
; CHECK-DAG: st.w [[R4]], 0($4)
ret void
; CHECK: .size xor_v4i32_i
}
define void @xor_v2i64_i(<2 x i64>* %c, <2 x i64>* %a) nounwind {
; CHECK: xor_v2i64_i:
%1 = load <2 x i64>* %a
; CHECK-DAG: ld.d [[R1:\$w[0-9]+]], 0($5)
[mips][msa] Change constant used in ori tests to avoid conflict with bseti (also xori to avoid bnegi) Upcoming commit(s) are going to add support for bseti and bnegi. This would cause some existing tests to (correctly) change behaviour and emit a different instruction. This patch prevents this by changing the constant used in ori and xori tests so that they will not be matchable by the bseti and bnegi patterns when these instructions are matchable from normal IR. git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@194467 91177308-0d34-0410-b5e6-96231b3b80d8
2013-11-12 10:14:18 +00:00
%2 = xor <2 x i64> %1, <i64 3, i64 3>
; CHECK-DAG: ldi.d [[R3:\$w[0-9]+]], 3
[mips][msa] Added support for matching andi, ori, nori, and xori from normal IR (i.e. not intrinsics) git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@191293 91177308-0d34-0410-b5e6-96231b3b80d8
2013-09-24 12:32:47 +00:00
; CHECK-DAG: xor.v [[R4:\$w[0-9]+]], [[R1]], [[R3]]
store <2 x i64> %2, <2 x i64>* %c
; CHECK-DAG: st.d [[R4]], 0($4)
ret void
; CHECK: .size xor_v2i64_i
}
[mips][msa] Added test cases that were supposed to be part of r190507, r190509, r190512, and r190518. git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@190522 91177308-0d34-0410-b5e6-96231b3b80d8
2013-09-11 12:39:25 +00:00
define void @sll_v16i8(<16 x i8>* %c, <16 x i8>* %a, <16 x i8>* %b) nounwind {
; CHECK: sll_v16i8:
%1 = load <16 x i8>* %a
; CHECK-DAG: ld.b [[R1:\$w[0-9]+]], 0($5)
%2 = load <16 x i8>* %b
; CHECK-DAG: ld.b [[R2:\$w[0-9]+]], 0($6)
%3 = shl <16 x i8> %1, %2
; CHECK-DAG: sll.b [[R3:\$w[0-9]+]], [[R1]], [[R2]]
store <16 x i8> %3, <16 x i8>* %c
; CHECK-DAG: st.b [[R3]], 0($4)
ret void
; CHECK: .size sll_v16i8
}
define void @sll_v8i16(<8 x i16>* %c, <8 x i16>* %a, <8 x i16>* %b) nounwind {
; CHECK: sll_v8i16:
%1 = load <8 x i16>* %a
; CHECK-DAG: ld.h [[R1:\$w[0-9]+]], 0($5)
%2 = load <8 x i16>* %b
; CHECK-DAG: ld.h [[R2:\$w[0-9]+]], 0($6)
%3 = shl <8 x i16> %1, %2
; CHECK-DAG: sll.h [[R3:\$w[0-9]+]], [[R1]], [[R2]]
store <8 x i16> %3, <8 x i16>* %c
; CHECK-DAG: st.h [[R3]], 0($4)
ret void
; CHECK: .size sll_v8i16
}
define void @sll_v4i32(<4 x i32>* %c, <4 x i32>* %a, <4 x i32>* %b) nounwind {
; CHECK: sll_v4i32:
%1 = load <4 x i32>* %a
; CHECK-DAG: ld.w [[R1:\$w[0-9]+]], 0($5)
%2 = load <4 x i32>* %b
; CHECK-DAG: ld.w [[R2:\$w[0-9]+]], 0($6)
%3 = shl <4 x i32> %1, %2
; CHECK-DAG: sll.w [[R3:\$w[0-9]+]], [[R1]], [[R2]]
store <4 x i32> %3, <4 x i32>* %c
; CHECK-DAG: st.w [[R3]], 0($4)
ret void
; CHECK: .size sll_v4i32
}
define void @sll_v2i64(<2 x i64>* %c, <2 x i64>* %a, <2 x i64>* %b) nounwind {
; CHECK: sll_v2i64:
%1 = load <2 x i64>* %a
; CHECK-DAG: ld.d [[R1:\$w[0-9]+]], 0($5)
%2 = load <2 x i64>* %b
; CHECK-DAG: ld.d [[R2:\$w[0-9]+]], 0($6)
%3 = shl <2 x i64> %1, %2
; CHECK-DAG: sll.d [[R3:\$w[0-9]+]], [[R1]], [[R2]]
store <2 x i64> %3, <2 x i64>* %c
; CHECK-DAG: st.d [[R3]], 0($4)
ret void
; CHECK: .size sll_v2i64
}
[mips][msa] Added support for matching slli, srai, and srli from normal IR (i.e. not intrinsics) git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@191285 91177308-0d34-0410-b5e6-96231b3b80d8
2013-09-24 10:28:18 +00:00
define void @sll_v16i8_i(<16 x i8>* %c, <16 x i8>* %a) nounwind {
; CHECK: sll_v16i8_i:
%1 = load <16 x i8>* %a
; CHECK-DAG: ld.b [[R1:\$w[0-9]+]], 0($5)
%2 = shl <16 x i8> %1, <i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1>
; CHECK-DAG: slli.b [[R4:\$w[0-9]+]], [[R1]], 1
store <16 x i8> %2, <16 x i8>* %c
; CHECK-DAG: st.b [[R4]], 0($4)
ret void
; CHECK: .size sll_v16i8_i
}
define void @sll_v8i16_i(<8 x i16>* %c, <8 x i16>* %a) nounwind {
; CHECK: sll_v8i16_i:
%1 = load <8 x i16>* %a
; CHECK-DAG: ld.h [[R1:\$w[0-9]+]], 0($5)
%2 = shl <8 x i16> %1, <i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1>
; CHECK-DAG: slli.h [[R4:\$w[0-9]+]], [[R1]], 1
store <8 x i16> %2, <8 x i16>* %c
; CHECK-DAG: st.h [[R4]], 0($4)
ret void
; CHECK: .size sll_v8i16_i
}
define void @sll_v4i32_i(<4 x i32>* %c, <4 x i32>* %a) nounwind {
; CHECK: sll_v4i32_i:
%1 = load <4 x i32>* %a
; CHECK-DAG: ld.w [[R1:\$w[0-9]+]], 0($5)
%2 = shl <4 x i32> %1, <i32 1, i32 1, i32 1, i32 1>
; CHECK-DAG: slli.w [[R4:\$w[0-9]+]], [[R1]], 1
store <4 x i32> %2, <4 x i32>* %c
; CHECK-DAG: st.w [[R4]], 0($4)
ret void
; CHECK: .size sll_v4i32_i
}
define void @sll_v2i64_i(<2 x i64>* %c, <2 x i64>* %a) nounwind {
; CHECK: sll_v2i64_i:
%1 = load <2 x i64>* %a
; CHECK-DAG: ld.d [[R1:\$w[0-9]+]], 0($5)
%2 = shl <2 x i64> %1, <i64 1, i64 1>
; CHECK-DAG: slli.d [[R4:\$w[0-9]+]], [[R1]], 1
store <2 x i64> %2, <2 x i64>* %c
; CHECK-DAG: st.d [[R4]], 0($4)
ret void
; CHECK: .size sll_v2i64_i
}
[mips][msa] Added test cases that were supposed to be part of r190507, r190509, r190512, and r190518. git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@190522 91177308-0d34-0410-b5e6-96231b3b80d8
2013-09-11 12:39:25 +00:00
define void @sra_v16i8(<16 x i8>* %c, <16 x i8>* %a, <16 x i8>* %b) nounwind {
; CHECK: sra_v16i8:
%1 = load <16 x i8>* %a
; CHECK-DAG: ld.b [[R1:\$w[0-9]+]], 0($5)
%2 = load <16 x i8>* %b
; CHECK-DAG: ld.b [[R2:\$w[0-9]+]], 0($6)
%3 = ashr <16 x i8> %1, %2
; CHECK-DAG: sra.b [[R3:\$w[0-9]+]], [[R1]], [[R2]]
store <16 x i8> %3, <16 x i8>* %c
; CHECK-DAG: st.b [[R3]], 0($4)
ret void
; CHECK: .size sra_v16i8
}
define void @sra_v8i16(<8 x i16>* %c, <8 x i16>* %a, <8 x i16>* %b) nounwind {
; CHECK: sra_v8i16:
%1 = load <8 x i16>* %a
; CHECK-DAG: ld.h [[R1:\$w[0-9]+]], 0($5)
%2 = load <8 x i16>* %b
; CHECK-DAG: ld.h [[R2:\$w[0-9]+]], 0($6)
%3 = ashr <8 x i16> %1, %2
; CHECK-DAG: sra.h [[R3:\$w[0-9]+]], [[R1]], [[R2]]
store <8 x i16> %3, <8 x i16>* %c
; CHECK-DAG: st.h [[R3]], 0($4)
ret void
; CHECK: .size sra_v8i16
}
define void @sra_v4i32(<4 x i32>* %c, <4 x i32>* %a, <4 x i32>* %b) nounwind {
; CHECK: sra_v4i32:
%1 = load <4 x i32>* %a
; CHECK-DAG: ld.w [[R1:\$w[0-9]+]], 0($5)
%2 = load <4 x i32>* %b
; CHECK-DAG: ld.w [[R2:\$w[0-9]+]], 0($6)
%3 = ashr <4 x i32> %1, %2
; CHECK-DAG: sra.w [[R3:\$w[0-9]+]], [[R1]], [[R2]]
store <4 x i32> %3, <4 x i32>* %c
; CHECK-DAG: st.w [[R3]], 0($4)
ret void
; CHECK: .size sra_v4i32
}
define void @sra_v2i64(<2 x i64>* %c, <2 x i64>* %a, <2 x i64>* %b) nounwind {
; CHECK: sra_v2i64:
%1 = load <2 x i64>* %a
; CHECK-DAG: ld.d [[R1:\$w[0-9]+]], 0($5)
%2 = load <2 x i64>* %b
; CHECK-DAG: ld.d [[R2:\$w[0-9]+]], 0($6)
%3 = ashr <2 x i64> %1, %2
; CHECK-DAG: sra.d [[R3:\$w[0-9]+]], [[R1]], [[R2]]
store <2 x i64> %3, <2 x i64>* %c
; CHECK-DAG: st.d [[R3]], 0($4)
ret void
; CHECK: .size sra_v2i64
}
[mips][msa] Added support for matching slli, srai, and srli from normal IR (i.e. not intrinsics) git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@191285 91177308-0d34-0410-b5e6-96231b3b80d8
2013-09-24 10:28:18 +00:00
define void @sra_v16i8_i(<16 x i8>* %c, <16 x i8>* %a) nounwind {
; CHECK: sra_v16i8_i:
%1 = load <16 x i8>* %a
; CHECK-DAG: ld.b [[R1:\$w[0-9]+]], 0($5)
%2 = ashr <16 x i8> %1, <i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1>
; CHECK-DAG: srai.b [[R4:\$w[0-9]+]], [[R1]], 1
store <16 x i8> %2, <16 x i8>* %c
; CHECK-DAG: st.b [[R4]], 0($4)
ret void
; CHECK: .size sra_v16i8_i
}
define void @sra_v8i16_i(<8 x i16>* %c, <8 x i16>* %a) nounwind {
; CHECK: sra_v8i16_i:
%1 = load <8 x i16>* %a
; CHECK-DAG: ld.h [[R1:\$w[0-9]+]], 0($5)
%2 = ashr <8 x i16> %1, <i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1>
; CHECK-DAG: srai.h [[R4:\$w[0-9]+]], [[R1]], 1
store <8 x i16> %2, <8 x i16>* %c
; CHECK-DAG: st.h [[R4]], 0($4)
ret void
; CHECK: .size sra_v8i16_i
}
define void @sra_v4i32_i(<4 x i32>* %c, <4 x i32>* %a) nounwind {
; CHECK: sra_v4i32_i:
%1 = load <4 x i32>* %a
; CHECK-DAG: ld.w [[R1:\$w[0-9]+]], 0($5)
%2 = ashr <4 x i32> %1, <i32 1, i32 1, i32 1, i32 1>
; CHECK-DAG: srai.w [[R4:\$w[0-9]+]], [[R1]], 1
store <4 x i32> %2, <4 x i32>* %c
; CHECK-DAG: st.w [[R4]], 0($4)
ret void
; CHECK: .size sra_v4i32_i
}
define void @sra_v2i64_i(<2 x i64>* %c, <2 x i64>* %a) nounwind {
; CHECK: sra_v2i64_i:
%1 = load <2 x i64>* %a
; CHECK-DAG: ld.d [[R1:\$w[0-9]+]], 0($5)
%2 = ashr <2 x i64> %1, <i64 1, i64 1>
; CHECK-DAG: srai.d [[R4:\$w[0-9]+]], [[R1]], 1
store <2 x i64> %2, <2 x i64>* %c
; CHECK-DAG: st.d [[R4]], 0($4)
ret void
; CHECK: .size sra_v2i64_i
}
[mips][msa] Added test cases that were supposed to be part of r190507, r190509, r190512, and r190518. git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@190522 91177308-0d34-0410-b5e6-96231b3b80d8
2013-09-11 12:39:25 +00:00
define void @srl_v16i8(<16 x i8>* %c, <16 x i8>* %a, <16 x i8>* %b) nounwind {
; CHECK: srl_v16i8:
%1 = load <16 x i8>* %a
; CHECK-DAG: ld.b [[R1:\$w[0-9]+]], 0($5)
%2 = load <16 x i8>* %b
; CHECK-DAG: ld.b [[R2:\$w[0-9]+]], 0($6)
%3 = lshr <16 x i8> %1, %2
; CHECK-DAG: srl.b [[R3:\$w[0-9]+]], [[R1]], [[R2]]
store <16 x i8> %3, <16 x i8>* %c
; CHECK-DAG: st.b [[R3]], 0($4)
ret void
; CHECK: .size srl_v16i8
}
define void @srl_v8i16(<8 x i16>* %c, <8 x i16>* %a, <8 x i16>* %b) nounwind {
; CHECK: srl_v8i16:
%1 = load <8 x i16>* %a
; CHECK-DAG: ld.h [[R1:\$w[0-9]+]], 0($5)
%2 = load <8 x i16>* %b
; CHECK-DAG: ld.h [[R2:\$w[0-9]+]], 0($6)
%3 = lshr <8 x i16> %1, %2
; CHECK-DAG: srl.h [[R3:\$w[0-9]+]], [[R1]], [[R2]]
store <8 x i16> %3, <8 x i16>* %c
; CHECK-DAG: st.h [[R3]], 0($4)
ret void
; CHECK: .size srl_v8i16
}
define void @srl_v4i32(<4 x i32>* %c, <4 x i32>* %a, <4 x i32>* %b) nounwind {
; CHECK: srl_v4i32:
%1 = load <4 x i32>* %a
; CHECK-DAG: ld.w [[R1:\$w[0-9]+]], 0($5)
%2 = load <4 x i32>* %b
; CHECK-DAG: ld.w [[R2:\$w[0-9]+]], 0($6)
%3 = lshr <4 x i32> %1, %2
; CHECK-DAG: srl.w [[R3:\$w[0-9]+]], [[R1]], [[R2]]
store <4 x i32> %3, <4 x i32>* %c
; CHECK-DAG: st.w [[R3]], 0($4)
ret void
; CHECK: .size srl_v4i32
}
define void @srl_v2i64(<2 x i64>* %c, <2 x i64>* %a, <2 x i64>* %b) nounwind {
; CHECK: srl_v2i64:
%1 = load <2 x i64>* %a
; CHECK-DAG: ld.d [[R1:\$w[0-9]+]], 0($5)
%2 = load <2 x i64>* %b
; CHECK-DAG: ld.d [[R2:\$w[0-9]+]], 0($6)
%3 = lshr <2 x i64> %1, %2
; CHECK-DAG: srl.d [[R3:\$w[0-9]+]], [[R1]], [[R2]]
store <2 x i64> %3, <2 x i64>* %c
; CHECK-DAG: st.d [[R3]], 0($4)
ret void
; CHECK: .size srl_v2i64
}
[mips][msa] Added support for matching slli, srai, and srli from normal IR (i.e. not intrinsics) git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@191285 91177308-0d34-0410-b5e6-96231b3b80d8
2013-09-24 10:28:18 +00:00
define void @srl_v16i8_i(<16 x i8>* %c, <16 x i8>* %a) nounwind {
; CHECK: srl_v16i8_i:
%1 = load <16 x i8>* %a
; CHECK-DAG: ld.b [[R1:\$w[0-9]+]], 0($5)
%2 = lshr <16 x i8> %1, <i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1>
; CHECK-DAG: srli.b [[R4:\$w[0-9]+]], [[R1]], 1
store <16 x i8> %2, <16 x i8>* %c
; CHECK-DAG: st.b [[R4]], 0($4)
ret void
; CHECK: .size srl_v16i8_i
}
define void @srl_v8i16_i(<8 x i16>* %c, <8 x i16>* %a) nounwind {
; CHECK: srl_v8i16_i:
%1 = load <8 x i16>* %a
; CHECK-DAG: ld.h [[R1:\$w[0-9]+]], 0($5)
%2 = lshr <8 x i16> %1, <i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1>
; CHECK-DAG: srli.h [[R4:\$w[0-9]+]], [[R1]], 1
store <8 x i16> %2, <8 x i16>* %c
; CHECK-DAG: st.h [[R4]], 0($4)
ret void
; CHECK: .size srl_v8i16_i
}
define void @srl_v4i32_i(<4 x i32>* %c, <4 x i32>* %a) nounwind {
; CHECK: srl_v4i32_i:
%1 = load <4 x i32>* %a
; CHECK-DAG: ld.w [[R1:\$w[0-9]+]], 0($5)
%2 = lshr <4 x i32> %1, <i32 1, i32 1, i32 1, i32 1>
; CHECK-DAG: srli.w [[R4:\$w[0-9]+]], [[R1]], 1
store <4 x i32> %2, <4 x i32>* %c
; CHECK-DAG: st.w [[R4]], 0($4)
ret void
; CHECK: .size srl_v4i32_i
}
define void @srl_v2i64_i(<2 x i64>* %c, <2 x i64>* %a) nounwind {
; CHECK: srl_v2i64_i:
%1 = load <2 x i64>* %a
; CHECK-DAG: ld.d [[R1:\$w[0-9]+]], 0($5)
%2 = lshr <2 x i64> %1, <i64 1, i64 1>
; CHECK-DAG: srli.d [[R4:\$w[0-9]+]], [[R1]], 1
store <2 x i64> %2, <2 x i64>* %c
; CHECK-DAG: st.d [[R4]], 0($4)
ret void
; CHECK: .size srl_v2i64_i
}
[mips][msa] Added support for matching pcnt from normal IR (i.e. not intrinsics) git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@191198 91177308-0d34-0410-b5e6-96231b3b80d8
2013-09-23 13:40:21 +00:00
define void @ctpop_v16i8(<16 x i8>* %c, <16 x i8>* %a) nounwind {
; CHECK: ctpop_v16i8:
%1 = load <16 x i8>* %a
; CHECK-DAG: ld.b [[R1:\$w[0-9]+]], 0($5)
%2 = tail call <16 x i8> @llvm.ctpop.v16i8 (<16 x i8> %1)
; CHECK-DAG: pcnt.b [[R3:\$w[0-9]+]], [[R1]]
store <16 x i8> %2, <16 x i8>* %c
; CHECK-DAG: st.b [[R3]], 0($4)
ret void
; CHECK: .size ctpop_v16i8
}
define void @ctpop_v8i16(<8 x i16>* %c, <8 x i16>* %a) nounwind {
; CHECK: ctpop_v8i16:
%1 = load <8 x i16>* %a
; CHECK-DAG: ld.h [[R1:\$w[0-9]+]], 0($5)
%2 = tail call <8 x i16> @llvm.ctpop.v8i16 (<8 x i16> %1)
; CHECK-DAG: pcnt.h [[R3:\$w[0-9]+]], [[R1]]
store <8 x i16> %2, <8 x i16>* %c
; CHECK-DAG: st.h [[R3]], 0($4)
ret void
; CHECK: .size ctpop_v8i16
}
define void @ctpop_v4i32(<4 x i32>* %c, <4 x i32>* %a) nounwind {
; CHECK: ctpop_v4i32:
%1 = load <4 x i32>* %a
; CHECK-DAG: ld.w [[R1:\$w[0-9]+]], 0($5)
%2 = tail call <4 x i32> @llvm.ctpop.v4i32 (<4 x i32> %1)
; CHECK-DAG: pcnt.w [[R3:\$w[0-9]+]], [[R1]]
store <4 x i32> %2, <4 x i32>* %c
; CHECK-DAG: st.w [[R3]], 0($4)
ret void
; CHECK: .size ctpop_v4i32
}
define void @ctpop_v2i64(<2 x i64>* %c, <2 x i64>* %a) nounwind {
; CHECK: ctpop_v2i64:
%1 = load <2 x i64>* %a
; CHECK-DAG: ld.d [[R1:\$w[0-9]+]], 0($5)
%2 = tail call <2 x i64> @llvm.ctpop.v2i64 (<2 x i64> %1)
; CHECK-DAG: pcnt.d [[R3:\$w[0-9]+]], [[R1]]
store <2 x i64> %2, <2 x i64>* %c
; CHECK-DAG: st.d [[R3]], 0($4)
ret void
; CHECK: .size ctpop_v2i64
}
[mips][msa] Added test cases that were supposed to be part of r190507, r190509, r190512, and r190518. git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@190522 91177308-0d34-0410-b5e6-96231b3b80d8
2013-09-11 12:39:25 +00:00
define void @ctlz_v16i8(<16 x i8>* %c, <16 x i8>* %a) nounwind {
; CHECK: ctlz_v16i8:
%1 = load <16 x i8>* %a
; CHECK-DAG: ld.b [[R1:\$w[0-9]+]], 0($5)
%2 = tail call <16 x i8> @llvm.ctlz.v16i8 (<16 x i8> %1)
; CHECK-DAG: nlzc.b [[R3:\$w[0-9]+]], [[R1]]
store <16 x i8> %2, <16 x i8>* %c
; CHECK-DAG: st.b [[R3]], 0($4)
ret void
; CHECK: .size ctlz_v16i8
}
define void @ctlz_v8i16(<8 x i16>* %c, <8 x i16>* %a) nounwind {
; CHECK: ctlz_v8i16:
%1 = load <8 x i16>* %a
; CHECK-DAG: ld.h [[R1:\$w[0-9]+]], 0($5)
%2 = tail call <8 x i16> @llvm.ctlz.v8i16 (<8 x i16> %1)
; CHECK-DAG: nlzc.h [[R3:\$w[0-9]+]], [[R1]]
store <8 x i16> %2, <8 x i16>* %c
; CHECK-DAG: st.h [[R3]], 0($4)
ret void
; CHECK: .size ctlz_v8i16
}
define void @ctlz_v4i32(<4 x i32>* %c, <4 x i32>* %a) nounwind {
; CHECK: ctlz_v4i32:
%1 = load <4 x i32>* %a
; CHECK-DAG: ld.w [[R1:\$w[0-9]+]], 0($5)
%2 = tail call <4 x i32> @llvm.ctlz.v4i32 (<4 x i32> %1)
; CHECK-DAG: nlzc.w [[R3:\$w[0-9]+]], [[R1]]
store <4 x i32> %2, <4 x i32>* %c
; CHECK-DAG: st.w [[R3]], 0($4)
ret void
; CHECK: .size ctlz_v4i32
}
define void @ctlz_v2i64(<2 x i64>* %c, <2 x i64>* %a) nounwind {
; CHECK: ctlz_v2i64:
%1 = load <2 x i64>* %a
; CHECK-DAG: ld.d [[R1:\$w[0-9]+]], 0($5)
%2 = tail call <2 x i64> @llvm.ctlz.v2i64 (<2 x i64> %1)
; CHECK-DAG: nlzc.d [[R3:\$w[0-9]+]], [[R1]]
store <2 x i64> %2, <2 x i64>* %c
; CHECK-DAG: st.d [[R3]], 0($4)
ret void
; CHECK: .size ctlz_v2i64
}
[mips][msa] Added support for matching bmnz, bmnzi, bmz, and bmzi from normal IR (i.e. not intrinsics) Also corrected the definition of the intrinsics for these instructions (the result register is also the first operand), and added intrinsics for bsel and bseli to clang (they already existed in the backend). These four operations are mostly equivalent to bsel, and bseli (the difference is which operand is tied to the result). As a result some of the tests changed as described below. bitwise.ll: - bsel.v test adapted so that the mask is unknown at compile-time. This stops it emitting bmnzi.b instead of the intended bsel.v. - The bseli.b test now tests the right thing. Namely the case when one of the values is an uimm8, rather than when the condition is a uimm8 (which is covered by bmnzi.b) compare.ll: - bsel.v tests now (correctly) emits bmnz.v instead of bsel.v because this is the same operation (see MSA.txt). i8.ll - CHECK-DAG-ized test. - bmzi.b test now (correctly) emits equivalent bmnzi.b with swapped operands because this is the same operation (see MSA.txt). - bseli.b still emits bseli.b though because the immediate makes it distinguishable from bmnzi.b. vec.ll: - CHECK-DAG-ized test. - bmz.v tests now (correctly) emits bmnz.v with swapped operands (see MSA.txt). - bsel.v tests now (correctly) emits bmnz.v with swapped operands (see MSA.txt). git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@193693 91177308-0d34-0410-b5e6-96231b3b80d8
2013-10-30 15:20:38 +00:00
define void @bsel_v16i8(<16 x i8>* %c, <16 x i8>* %a, <16 x i8>* %b, <16 x i8>* %m) nounwind {
[mips][msa] Combine binsri-like DAG of AND and OR into equivalent VSELECT (or (and $a, $mask), (and $b, $inverse_mask)) => (vselect $mask, $a, $b). where $mask is a constant splat. This allows bitwise operations to make use of bsel. It's also a stepping stone towards matching bins[lr], and bins[lr]i from normal IR. Two sets of similar tests have been added in this commit. The bsel_* functions test the case where binsri cannot be used. The binsr_*_i functions will start to use the binsri instruction in the next commit. git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@193682 91177308-0d34-0410-b5e6-96231b3b80d8
2013-10-30 13:51:01 +00:00
; CHECK: bsel_v16i8:
%1 = load <16 x i8>* %a
; CHECK-DAG: ld.b [[R1:\$w[0-9]+]], 0($5)
%2 = load <16 x i8>* %b
; CHECK-DAG: ld.b [[R2:\$w[0-9]+]], 0($6)
[mips][msa] Added support for matching bmnz, bmnzi, bmz, and bmzi from normal IR (i.e. not intrinsics) Also corrected the definition of the intrinsics for these instructions (the result register is also the first operand), and added intrinsics for bsel and bseli to clang (they already existed in the backend). These four operations are mostly equivalent to bsel, and bseli (the difference is which operand is tied to the result). As a result some of the tests changed as described below. bitwise.ll: - bsel.v test adapted so that the mask is unknown at compile-time. This stops it emitting bmnzi.b instead of the intended bsel.v. - The bseli.b test now tests the right thing. Namely the case when one of the values is an uimm8, rather than when the condition is a uimm8 (which is covered by bmnzi.b) compare.ll: - bsel.v tests now (correctly) emits bmnz.v instead of bsel.v because this is the same operation (see MSA.txt). i8.ll - CHECK-DAG-ized test. - bmzi.b test now (correctly) emits equivalent bmnzi.b with swapped operands because this is the same operation (see MSA.txt). - bseli.b still emits bseli.b though because the immediate makes it distinguishable from bmnzi.b. vec.ll: - CHECK-DAG-ized test. - bmz.v tests now (correctly) emits bmnz.v with swapped operands (see MSA.txt). - bsel.v tests now (correctly) emits bmnz.v with swapped operands (see MSA.txt). git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@193693 91177308-0d34-0410-b5e6-96231b3b80d8
2013-10-30 15:20:38 +00:00
%3 = load <16 x i8>* %m
; CHECK-DAG: ld.b [[R3:\$w[0-9]+]], 0($7)
%4 = xor <16 x i8> %3, <i8 -1, i8 -1, i8 -1, i8 -1,
i8 -1, i8 -1, i8 -1, i8 -1,
i8 -1, i8 -1, i8 -1, i8 -1,
i8 -1, i8 -1, i8 -1, i8 -1>
%5 = and <16 x i8> %1, %3
%6 = and <16 x i8> %2, %4
%7 = or <16 x i8> %5, %6
; bmnz is the same operation
[mips] BSEL's and BINS[RL] operands are reversed compared to the vselect node used in the pattern. Summary: Correct the match patterns and the lowerings that made the CodeGen tests pass despite the mistakes. The original testcase that discovered the problem was SingleSource/UnitTests/SignlessType/factor.c in test-suite. During review, we also found that some of the existing CodeGen tests were incorrect and fixed them: * bitwise.ll: In bsel_v16i8 the IfSet/IfClear were reversed because bsel and bmnz have different operand orders and the test didn't correctly account for this. bmnz goes 'IfClear, IfSet, CondMask', while bsel goes 'CondMask, IfClear, IfSet'. * vec.ll: In the cases where a bsel is emitted as a bmnz (they are the same operation with a different input tied to the result) the operands were in the wrong order. * compare.ll and compare_float.ll: The bsel operand order was correct for a greater-than comparison, but a greater-than comparison instruction doesn't exist. Lowering this operation inverts the condition so the IfSet/IfClear need to be swapped to match. The differences between BSEL, BMNZ, and BMZ and how they map to/from vselect are rather confusing. I've therefore added a note to MSA.txt to explain this in a single place in addition to the comments that explain each case. Reviewers: matheusalmeida, jacksprat Reviewed By: matheusalmeida Differential Revision: http://llvm-reviews.chandlerc.com/D3028 git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@203657 91177308-0d34-0410-b5e6-96231b3b80d8
2014-03-12 11:54:00 +00:00
; (vselect Mask, IfSet, IfClr) -> (BMNZ IfClr, IfSet, Mask)
; CHECK-DAG: bmnz.v [[R2]], [[R1]], [[R3]]
[mips][msa] Added support for matching bmnz, bmnzi, bmz, and bmzi from normal IR (i.e. not intrinsics) Also corrected the definition of the intrinsics for these instructions (the result register is also the first operand), and added intrinsics for bsel and bseli to clang (they already existed in the backend). These four operations are mostly equivalent to bsel, and bseli (the difference is which operand is tied to the result). As a result some of the tests changed as described below. bitwise.ll: - bsel.v test adapted so that the mask is unknown at compile-time. This stops it emitting bmnzi.b instead of the intended bsel.v. - The bseli.b test now tests the right thing. Namely the case when one of the values is an uimm8, rather than when the condition is a uimm8 (which is covered by bmnzi.b) compare.ll: - bsel.v tests now (correctly) emits bmnz.v instead of bsel.v because this is the same operation (see MSA.txt). i8.ll - CHECK-DAG-ized test. - bmzi.b test now (correctly) emits equivalent bmnzi.b with swapped operands because this is the same operation (see MSA.txt). - bseli.b still emits bseli.b though because the immediate makes it distinguishable from bmnzi.b. vec.ll: - CHECK-DAG-ized test. - bmz.v tests now (correctly) emits bmnz.v with swapped operands (see MSA.txt). - bsel.v tests now (correctly) emits bmnz.v with swapped operands (see MSA.txt). git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@193693 91177308-0d34-0410-b5e6-96231b3b80d8
2013-10-30 15:20:38 +00:00
store <16 x i8> %7, <16 x i8>* %c
[mips] BSEL's and BINS[RL] operands are reversed compared to the vselect node used in the pattern. Summary: Correct the match patterns and the lowerings that made the CodeGen tests pass despite the mistakes. The original testcase that discovered the problem was SingleSource/UnitTests/SignlessType/factor.c in test-suite. During review, we also found that some of the existing CodeGen tests were incorrect and fixed them: * bitwise.ll: In bsel_v16i8 the IfSet/IfClear were reversed because bsel and bmnz have different operand orders and the test didn't correctly account for this. bmnz goes 'IfClear, IfSet, CondMask', while bsel goes 'CondMask, IfClear, IfSet'. * vec.ll: In the cases where a bsel is emitted as a bmnz (they are the same operation with a different input tied to the result) the operands were in the wrong order. * compare.ll and compare_float.ll: The bsel operand order was correct for a greater-than comparison, but a greater-than comparison instruction doesn't exist. Lowering this operation inverts the condition so the IfSet/IfClear need to be swapped to match. The differences between BSEL, BMNZ, and BMZ and how they map to/from vselect are rather confusing. I've therefore added a note to MSA.txt to explain this in a single place in addition to the comments that explain each case. Reviewers: matheusalmeida, jacksprat Reviewed By: matheusalmeida Differential Revision: http://llvm-reviews.chandlerc.com/D3028 git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@203657 91177308-0d34-0410-b5e6-96231b3b80d8
2014-03-12 11:54:00 +00:00
; CHECK-DAG: st.b [[R2]], 0($4)
[mips][msa] Combine binsri-like DAG of AND and OR into equivalent VSELECT (or (and $a, $mask), (and $b, $inverse_mask)) => (vselect $mask, $a, $b). where $mask is a constant splat. This allows bitwise operations to make use of bsel. It's also a stepping stone towards matching bins[lr], and bins[lr]i from normal IR. Two sets of similar tests have been added in this commit. The bsel_* functions test the case where binsri cannot be used. The binsr_*_i functions will start to use the binsri instruction in the next commit. git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@193682 91177308-0d34-0410-b5e6-96231b3b80d8
2013-10-30 13:51:01 +00:00
ret void
; CHECK: .size bsel_v16i8
}
[mips][msa] Added support for matching bmnz, bmnzi, bmz, and bmzi from normal IR (i.e. not intrinsics) Also corrected the definition of the intrinsics for these instructions (the result register is also the first operand), and added intrinsics for bsel and bseli to clang (they already existed in the backend). These four operations are mostly equivalent to bsel, and bseli (the difference is which operand is tied to the result). As a result some of the tests changed as described below. bitwise.ll: - bsel.v test adapted so that the mask is unknown at compile-time. This stops it emitting bmnzi.b instead of the intended bsel.v. - The bseli.b test now tests the right thing. Namely the case when one of the values is an uimm8, rather than when the condition is a uimm8 (which is covered by bmnzi.b) compare.ll: - bsel.v tests now (correctly) emits bmnz.v instead of bsel.v because this is the same operation (see MSA.txt). i8.ll - CHECK-DAG-ized test. - bmzi.b test now (correctly) emits equivalent bmnzi.b with swapped operands because this is the same operation (see MSA.txt). - bseli.b still emits bseli.b though because the immediate makes it distinguishable from bmnzi.b. vec.ll: - CHECK-DAG-ized test. - bmz.v tests now (correctly) emits bmnz.v with swapped operands (see MSA.txt). - bsel.v tests now (correctly) emits bmnz.v with swapped operands (see MSA.txt). git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@193693 91177308-0d34-0410-b5e6-96231b3b80d8
2013-10-30 15:20:38 +00:00
define void @bsel_v16i8_i(<16 x i8>* %c, <16 x i8>* %a, <16 x i8>* %m) nounwind {
; CHECK: bsel_v16i8_i:
%1 = load <16 x i8>* %a
; CHECK-DAG: ld.b [[R1:\$w[0-9]+]], 0($5)
%2 = load <16 x i8>* %m
; CHECK-DAG: ld.b [[R3:\$w[0-9]+]], 0($6)
%3 = xor <16 x i8> %2, <i8 -1, i8 -1, i8 -1, i8 -1,
i8 -1, i8 -1, i8 -1, i8 -1,
i8 -1, i8 -1, i8 -1, i8 -1,
i8 -1, i8 -1, i8 -1, i8 -1>
%4 = and <16 x i8> %1, %3
%5 = and <16 x i8> <i8 6, i8 6, i8 6, i8 6,
i8 6, i8 6, i8 6, i8 6,
i8 6, i8 6, i8 6, i8 6,
i8 6, i8 6, i8 6, i8 6>, %2
%6 = or <16 x i8> %4, %5
; CHECK-DAG: bseli.b [[R3]], [[R1]], 6
store <16 x i8> %6, <16 x i8>* %c
; CHECK-DAG: st.b [[R3]], 0($4)
ret void
; CHECK: .size bsel_v16i8_i
}
[mips][msa] Combine binsri-like DAG of AND and OR into equivalent VSELECT (or (and $a, $mask), (and $b, $inverse_mask)) => (vselect $mask, $a, $b). where $mask is a constant splat. This allows bitwise operations to make use of bsel. It's also a stepping stone towards matching bins[lr], and bins[lr]i from normal IR. Two sets of similar tests have been added in this commit. The bsel_* functions test the case where binsri cannot be used. The binsr_*_i functions will start to use the binsri instruction in the next commit. git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@193682 91177308-0d34-0410-b5e6-96231b3b80d8
2013-10-30 13:51:01 +00:00
define void @bsel_v8i16(<8 x i16>* %c, <8 x i16>* %a, <8 x i16>* %b) nounwind {
; CHECK: bsel_v8i16:
%1 = load <8 x i16>* %a
; CHECK-DAG: ld.h [[R1:\$w[0-9]+]], 0($5)
%2 = load <8 x i16>* %b
; CHECK-DAG: ld.h [[R2:\$w[0-9]+]], 0($6)
%3 = and <8 x i16> %1, <i16 6, i16 6, i16 6, i16 6,
i16 6, i16 6, i16 6, i16 6>
%4 = and <8 x i16> %2, <i16 65529, i16 65529, i16 65529, i16 65529,
i16 65529, i16 65529, i16 65529, i16 65529>
%5 = or <8 x i16> %3, %4
; CHECK-DAG: ldi.h [[R3:\$w[0-9]+]], 6
; CHECK-DAG: bsel.v [[R3]], [[R2]], [[R1]]
store <8 x i16> %5, <8 x i16>* %c
; CHECK-DAG: st.h [[R3]], 0($4)
ret void
; CHECK: .size bsel_v8i16
}
define void @bsel_v4i32(<4 x i32>* %c, <4 x i32>* %a, <4 x i32>* %b) nounwind {
; CHECK: bsel_v4i32:
%1 = load <4 x i32>* %a
; CHECK-DAG: ld.w [[R1:\$w[0-9]+]], 0($5)
%2 = load <4 x i32>* %b
; CHECK-DAG: ld.w [[R2:\$w[0-9]+]], 0($6)
%3 = and <4 x i32> %1, <i32 6, i32 6, i32 6, i32 6>
%4 = and <4 x i32> %2, <i32 4294967289, i32 4294967289, i32 4294967289, i32 4294967289>
%5 = or <4 x i32> %3, %4
; CHECK-DAG: ldi.w [[R3:\$w[0-9]+]], 6
; CHECK-DAG: bsel.v [[R3]], [[R2]], [[R1]]
store <4 x i32> %5, <4 x i32>* %c
; CHECK-DAG: st.w [[R3]], 0($4)
ret void
; CHECK: .size bsel_v4i32
}
define void @bsel_v2i64(<2 x i64>* %c, <2 x i64>* %a, <2 x i64>* %b) nounwind {
; CHECK: bsel_v2i64:
%1 = load <2 x i64>* %a
; CHECK-DAG: ld.d [[R1:\$w[0-9]+]], 0($5)
%2 = load <2 x i64>* %b
; CHECK-DAG: ld.d [[R2:\$w[0-9]+]], 0($6)
%3 = and <2 x i64> %1, <i64 6, i64 6>
%4 = and <2 x i64> %2, <i64 18446744073709551609, i64 18446744073709551609>
%5 = or <2 x i64> %3, %4
; CHECK-DAG: ldi.d [[R3:\$w[0-9]+]], 6
; CHECK-DAG: bsel.v [[R3]], [[R2]], [[R1]]
store <2 x i64> %5, <2 x i64>* %c
; CHECK-DAG: st.d [[R3]], 0($4)
ret void
; CHECK: .size bsel_v2i64
}
[mips][msa] Added support for matching bins[lr]i.[bhwd] from normal IR (i.e. not intrinsics) This required correcting the definition of the bins[lr]i intrinsics because the result is also the first operand. It also required removing the (arbitrary) check for 32-bit immediates in MipsSEDAGToDAGISel::selectVSplat(). Currently using binsli.d with 2 bits set in the mask doesn't select binsli.d because the constant is legalized into a ConstantPool. Similar things can happen with binsri.d with more than 10 bits set in the mask. The resulting code when this happens is correct but not optimal. git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@193687 91177308-0d34-0410-b5e6-96231b3b80d8
2013-10-30 14:45:14 +00:00
define void @binsl_v16i8_i(<16 x i8>* %c, <16 x i8>* %a, <16 x i8>* %b) nounwind {
; CHECK: binsl_v16i8_i:
%1 = load <16 x i8>* %a
; CHECK-DAG: ld.b [[R1:\$w[0-9]+]], 0($5)
%2 = load <16 x i8>* %b
; CHECK-DAG: ld.b [[R2:\$w[0-9]+]], 0($6)
%3 = and <16 x i8> %1, <i8 192, i8 192, i8 192, i8 192,
i8 192, i8 192, i8 192, i8 192,
i8 192, i8 192, i8 192, i8 192,
i8 192, i8 192, i8 192, i8 192>
%4 = and <16 x i8> %2, <i8 63, i8 63, i8 63, i8 63,
i8 63, i8 63, i8 63, i8 63,
i8 63, i8 63, i8 63, i8 63,
i8 63, i8 63, i8 63, i8 63>
%5 = or <16 x i8> %3, %4
; CHECK-DAG: binsli.b [[R2]], [[R1]], 2
store <16 x i8> %5, <16 x i8>* %c
; CHECK-DAG: st.b [[R2]], 0($4)
ret void
; CHECK: .size binsl_v16i8_i
}
define void @binsl_v8i16_i(<8 x i16>* %c, <8 x i16>* %a, <8 x i16>* %b) nounwind {
; CHECK: binsl_v8i16_i:
%1 = load <8 x i16>* %a
; CHECK-DAG: ld.h [[R1:\$w[0-9]+]], 0($5)
%2 = load <8 x i16>* %b
; CHECK-DAG: ld.h [[R2:\$w[0-9]+]], 0($6)
%3 = and <8 x i16> %1, <i16 49152, i16 49152, i16 49152, i16 49152,
i16 49152, i16 49152, i16 49152, i16 49152>
%4 = and <8 x i16> %2, <i16 16383, i16 16383, i16 16383, i16 16383,
i16 16383, i16 16383, i16 16383, i16 16383>
%5 = or <8 x i16> %3, %4
; CHECK-DAG: binsli.h [[R2]], [[R1]], 2
store <8 x i16> %5, <8 x i16>* %c
; CHECK-DAG: st.h [[R2]], 0($4)
ret void
; CHECK: .size binsl_v8i16_i
}
define void @binsl_v4i32_i(<4 x i32>* %c, <4 x i32>* %a, <4 x i32>* %b) nounwind {
; CHECK: binsl_v4i32_i:
%1 = load <4 x i32>* %a
; CHECK-DAG: ld.w [[R1:\$w[0-9]+]], 0($5)
%2 = load <4 x i32>* %b
; CHECK-DAG: ld.w [[R2:\$w[0-9]+]], 0($6)
%3 = and <4 x i32> %1, <i32 3221225472, i32 3221225472, i32 3221225472, i32 3221225472>
%4 = and <4 x i32> %2, <i32 1073741823, i32 1073741823, i32 1073741823, i32 1073741823>
%5 = or <4 x i32> %3, %4
; CHECK-DAG: binsli.w [[R2]], [[R1]], 2
store <4 x i32> %5, <4 x i32>* %c
; CHECK-DAG: st.w [[R2]], 0($4)
ret void
; CHECK: .size binsl_v4i32_i
}
define void @binsl_v2i64_i(<2 x i64>* %c, <2 x i64>* %a, <2 x i64>* %b) nounwind {
; CHECK: binsl_v2i64_i:
%1 = load <2 x i64>* %a
; CHECK-DAG: ld.d [[R1:\$w[0-9]+]], 0($5)
%2 = load <2 x i64>* %b
; CHECK-DAG: ld.d [[R2:\$w[0-9]+]], 0($6)
%3 = and <2 x i64> %1, <i64 18446744073709551608, i64 18446744073709551608>
%4 = and <2 x i64> %2, <i64 7, i64 7>
%5 = or <2 x i64> %3, %4
; TODO: We use a particularly wide mask here to work around a legalization
; issue. If the mask doesn't fit within a 10-bit immediate, it gets
; legalized into a constant pool. We should add a test to cover the
; other cases once they correctly select binsli.d.
; CHECK-DAG: binsli.d [[R2]], [[R1]], 61
store <2 x i64> %5, <2 x i64>* %c
; CHECK-DAG: st.d [[R2]], 0($4)
ret void
; CHECK: .size binsl_v2i64_i
}
[mips][msa] Combine binsri-like DAG of AND and OR into equivalent VSELECT (or (and $a, $mask), (and $b, $inverse_mask)) => (vselect $mask, $a, $b). where $mask is a constant splat. This allows bitwise operations to make use of bsel. It's also a stepping stone towards matching bins[lr], and bins[lr]i from normal IR. Two sets of similar tests have been added in this commit. The bsel_* functions test the case where binsri cannot be used. The binsr_*_i functions will start to use the binsri instruction in the next commit. git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@193682 91177308-0d34-0410-b5e6-96231b3b80d8
2013-10-30 13:51:01 +00:00
define void @binsr_v16i8_i(<16 x i8>* %c, <16 x i8>* %a, <16 x i8>* %b) nounwind {
; CHECK: binsr_v16i8_i:
%1 = load <16 x i8>* %a
; CHECK-DAG: ld.b [[R1:\$w[0-9]+]], 0($5)
%2 = load <16 x i8>* %b
; CHECK-DAG: ld.b [[R2:\$w[0-9]+]], 0($6)
%3 = and <16 x i8> %1, <i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3,
i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3>
%4 = and <16 x i8> %2, <i8 252, i8 252, i8 252, i8 252,
i8 252, i8 252, i8 252, i8 252,
i8 252, i8 252, i8 252, i8 252,
i8 252, i8 252, i8 252, i8 252>
%5 = or <16 x i8> %3, %4
[mips][msa] Added support for matching bins[lr]i.[bhwd] from normal IR (i.e. not intrinsics) This required correcting the definition of the bins[lr]i intrinsics because the result is also the first operand. It also required removing the (arbitrary) check for 32-bit immediates in MipsSEDAGToDAGISel::selectVSplat(). Currently using binsli.d with 2 bits set in the mask doesn't select binsli.d because the constant is legalized into a ConstantPool. Similar things can happen with binsri.d with more than 10 bits set in the mask. The resulting code when this happens is correct but not optimal. git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@193687 91177308-0d34-0410-b5e6-96231b3b80d8
2013-10-30 14:45:14 +00:00
; CHECK-DAG: binsri.b [[R2]], [[R1]], 2
[mips][msa] Combine binsri-like DAG of AND and OR into equivalent VSELECT (or (and $a, $mask), (and $b, $inverse_mask)) => (vselect $mask, $a, $b). where $mask is a constant splat. This allows bitwise operations to make use of bsel. It's also a stepping stone towards matching bins[lr], and bins[lr]i from normal IR. Two sets of similar tests have been added in this commit. The bsel_* functions test the case where binsri cannot be used. The binsr_*_i functions will start to use the binsri instruction in the next commit. git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@193682 91177308-0d34-0410-b5e6-96231b3b80d8
2013-10-30 13:51:01 +00:00
store <16 x i8> %5, <16 x i8>* %c
[mips][msa] Added support for matching bins[lr]i.[bhwd] from normal IR (i.e. not intrinsics) This required correcting the definition of the bins[lr]i intrinsics because the result is also the first operand. It also required removing the (arbitrary) check for 32-bit immediates in MipsSEDAGToDAGISel::selectVSplat(). Currently using binsli.d with 2 bits set in the mask doesn't select binsli.d because the constant is legalized into a ConstantPool. Similar things can happen with binsri.d with more than 10 bits set in the mask. The resulting code when this happens is correct but not optimal. git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@193687 91177308-0d34-0410-b5e6-96231b3b80d8
2013-10-30 14:45:14 +00:00
; CHECK-DAG: st.b [[R2]], 0($4)
[mips][msa] Combine binsri-like DAG of AND and OR into equivalent VSELECT (or (and $a, $mask), (and $b, $inverse_mask)) => (vselect $mask, $a, $b). where $mask is a constant splat. This allows bitwise operations to make use of bsel. It's also a stepping stone towards matching bins[lr], and bins[lr]i from normal IR. Two sets of similar tests have been added in this commit. The bsel_* functions test the case where binsri cannot be used. The binsr_*_i functions will start to use the binsri instruction in the next commit. git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@193682 91177308-0d34-0410-b5e6-96231b3b80d8
2013-10-30 13:51:01 +00:00
ret void
; CHECK: .size binsr_v16i8_i
}
define void @binsr_v8i16_i(<8 x i16>* %c, <8 x i16>* %a, <8 x i16>* %b) nounwind {
; CHECK: binsr_v8i16_i:
%1 = load <8 x i16>* %a
; CHECK-DAG: ld.h [[R1:\$w[0-9]+]], 0($5)
%2 = load <8 x i16>* %b
; CHECK-DAG: ld.h [[R2:\$w[0-9]+]], 0($6)
%3 = and <8 x i16> %1, <i16 3, i16 3, i16 3, i16 3,
i16 3, i16 3, i16 3, i16 3>
%4 = and <8 x i16> %2, <i16 65532, i16 65532, i16 65532, i16 65532,
i16 65532, i16 65532, i16 65532, i16 65532>
%5 = or <8 x i16> %3, %4
[mips][msa] Added support for matching bins[lr]i.[bhwd] from normal IR (i.e. not intrinsics) This required correcting the definition of the bins[lr]i intrinsics because the result is also the first operand. It also required removing the (arbitrary) check for 32-bit immediates in MipsSEDAGToDAGISel::selectVSplat(). Currently using binsli.d with 2 bits set in the mask doesn't select binsli.d because the constant is legalized into a ConstantPool. Similar things can happen with binsri.d with more than 10 bits set in the mask. The resulting code when this happens is correct but not optimal. git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@193687 91177308-0d34-0410-b5e6-96231b3b80d8
2013-10-30 14:45:14 +00:00
; CHECK-DAG: binsri.h [[R2]], [[R1]], 2
[mips][msa] Combine binsri-like DAG of AND and OR into equivalent VSELECT (or (and $a, $mask), (and $b, $inverse_mask)) => (vselect $mask, $a, $b). where $mask is a constant splat. This allows bitwise operations to make use of bsel. It's also a stepping stone towards matching bins[lr], and bins[lr]i from normal IR. Two sets of similar tests have been added in this commit. The bsel_* functions test the case where binsri cannot be used. The binsr_*_i functions will start to use the binsri instruction in the next commit. git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@193682 91177308-0d34-0410-b5e6-96231b3b80d8
2013-10-30 13:51:01 +00:00
store <8 x i16> %5, <8 x i16>* %c
[mips][msa] Added support for matching bins[lr]i.[bhwd] from normal IR (i.e. not intrinsics) This required correcting the definition of the bins[lr]i intrinsics because the result is also the first operand. It also required removing the (arbitrary) check for 32-bit immediates in MipsSEDAGToDAGISel::selectVSplat(). Currently using binsli.d with 2 bits set in the mask doesn't select binsli.d because the constant is legalized into a ConstantPool. Similar things can happen with binsri.d with more than 10 bits set in the mask. The resulting code when this happens is correct but not optimal. git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@193687 91177308-0d34-0410-b5e6-96231b3b80d8
2013-10-30 14:45:14 +00:00
; CHECK-DAG: st.h [[R2]], 0($4)
[mips][msa] Combine binsri-like DAG of AND and OR into equivalent VSELECT (or (and $a, $mask), (and $b, $inverse_mask)) => (vselect $mask, $a, $b). where $mask is a constant splat. This allows bitwise operations to make use of bsel. It's also a stepping stone towards matching bins[lr], and bins[lr]i from normal IR. Two sets of similar tests have been added in this commit. The bsel_* functions test the case where binsri cannot be used. The binsr_*_i functions will start to use the binsri instruction in the next commit. git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@193682 91177308-0d34-0410-b5e6-96231b3b80d8
2013-10-30 13:51:01 +00:00
ret void
; CHECK: .size binsr_v8i16_i
}
define void @binsr_v4i32_i(<4 x i32>* %c, <4 x i32>* %a, <4 x i32>* %b) nounwind {
; CHECK: binsr_v4i32_i:
%1 = load <4 x i32>* %a
; CHECK-DAG: ld.w [[R1:\$w[0-9]+]], 0($5)
%2 = load <4 x i32>* %b
; CHECK-DAG: ld.w [[R2:\$w[0-9]+]], 0($6)
%3 = and <4 x i32> %1, <i32 3, i32 3, i32 3, i32 3>
%4 = and <4 x i32> %2, <i32 4294967292, i32 4294967292, i32 4294967292, i32 4294967292>
%5 = or <4 x i32> %3, %4
[mips][msa] Added support for matching bins[lr]i.[bhwd] from normal IR (i.e. not intrinsics) This required correcting the definition of the bins[lr]i intrinsics because the result is also the first operand. It also required removing the (arbitrary) check for 32-bit immediates in MipsSEDAGToDAGISel::selectVSplat(). Currently using binsli.d with 2 bits set in the mask doesn't select binsli.d because the constant is legalized into a ConstantPool. Similar things can happen with binsri.d with more than 10 bits set in the mask. The resulting code when this happens is correct but not optimal. git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@193687 91177308-0d34-0410-b5e6-96231b3b80d8
2013-10-30 14:45:14 +00:00
; CHECK-DAG: binsri.w [[R2]], [[R1]], 2
[mips][msa] Combine binsri-like DAG of AND and OR into equivalent VSELECT (or (and $a, $mask), (and $b, $inverse_mask)) => (vselect $mask, $a, $b). where $mask is a constant splat. This allows bitwise operations to make use of bsel. It's also a stepping stone towards matching bins[lr], and bins[lr]i from normal IR. Two sets of similar tests have been added in this commit. The bsel_* functions test the case where binsri cannot be used. The binsr_*_i functions will start to use the binsri instruction in the next commit. git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@193682 91177308-0d34-0410-b5e6-96231b3b80d8
2013-10-30 13:51:01 +00:00
store <4 x i32> %5, <4 x i32>* %c
[mips][msa] Added support for matching bins[lr]i.[bhwd] from normal IR (i.e. not intrinsics) This required correcting the definition of the bins[lr]i intrinsics because the result is also the first operand. It also required removing the (arbitrary) check for 32-bit immediates in MipsSEDAGToDAGISel::selectVSplat(). Currently using binsli.d with 2 bits set in the mask doesn't select binsli.d because the constant is legalized into a ConstantPool. Similar things can happen with binsri.d with more than 10 bits set in the mask. The resulting code when this happens is correct but not optimal. git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@193687 91177308-0d34-0410-b5e6-96231b3b80d8
2013-10-30 14:45:14 +00:00
; CHECK-DAG: st.w [[R2]], 0($4)
[mips][msa] Combine binsri-like DAG of AND and OR into equivalent VSELECT (or (and $a, $mask), (and $b, $inverse_mask)) => (vselect $mask, $a, $b). where $mask is a constant splat. This allows bitwise operations to make use of bsel. It's also a stepping stone towards matching bins[lr], and bins[lr]i from normal IR. Two sets of similar tests have been added in this commit. The bsel_* functions test the case where binsri cannot be used. The binsr_*_i functions will start to use the binsri instruction in the next commit. git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@193682 91177308-0d34-0410-b5e6-96231b3b80d8
2013-10-30 13:51:01 +00:00
ret void
; CHECK: .size binsr_v4i32_i
}
define void @binsr_v2i64_i(<2 x i64>* %c, <2 x i64>* %a, <2 x i64>* %b) nounwind {
; CHECK: binsr_v2i64_i:
%1 = load <2 x i64>* %a
; CHECK-DAG: ld.d [[R1:\$w[0-9]+]], 0($5)
%2 = load <2 x i64>* %b
; CHECK-DAG: ld.d [[R2:\$w[0-9]+]], 0($6)
%3 = and <2 x i64> %1, <i64 3, i64 3>
%4 = and <2 x i64> %2, <i64 18446744073709551612, i64 18446744073709551612>
%5 = or <2 x i64> %3, %4
[mips][msa] Added support for matching bins[lr]i.[bhwd] from normal IR (i.e. not intrinsics) This required correcting the definition of the bins[lr]i intrinsics because the result is also the first operand. It also required removing the (arbitrary) check for 32-bit immediates in MipsSEDAGToDAGISel::selectVSplat(). Currently using binsli.d with 2 bits set in the mask doesn't select binsli.d because the constant is legalized into a ConstantPool. Similar things can happen with binsri.d with more than 10 bits set in the mask. The resulting code when this happens is correct but not optimal. git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@193687 91177308-0d34-0410-b5e6-96231b3b80d8
2013-10-30 14:45:14 +00:00
; CHECK-DAG: binsri.d [[R2]], [[R1]], 2
[mips][msa] Combine binsri-like DAG of AND and OR into equivalent VSELECT (or (and $a, $mask), (and $b, $inverse_mask)) => (vselect $mask, $a, $b). where $mask is a constant splat. This allows bitwise operations to make use of bsel. It's also a stepping stone towards matching bins[lr], and bins[lr]i from normal IR. Two sets of similar tests have been added in this commit. The bsel_* functions test the case where binsri cannot be used. The binsr_*_i functions will start to use the binsri instruction in the next commit. git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@193682 91177308-0d34-0410-b5e6-96231b3b80d8
2013-10-30 13:51:01 +00:00
store <2 x i64> %5, <2 x i64>* %c
[mips][msa] Added support for matching bins[lr]i.[bhwd] from normal IR (i.e. not intrinsics) This required correcting the definition of the bins[lr]i intrinsics because the result is also the first operand. It also required removing the (arbitrary) check for 32-bit immediates in MipsSEDAGToDAGISel::selectVSplat(). Currently using binsli.d with 2 bits set in the mask doesn't select binsli.d because the constant is legalized into a ConstantPool. Similar things can happen with binsri.d with more than 10 bits set in the mask. The resulting code when this happens is correct but not optimal. git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@193687 91177308-0d34-0410-b5e6-96231b3b80d8
2013-10-30 14:45:14 +00:00
; CHECK-DAG: st.d [[R2]], 0($4)
[mips][msa] Combine binsri-like DAG of AND and OR into equivalent VSELECT (or (and $a, $mask), (and $b, $inverse_mask)) => (vselect $mask, $a, $b). where $mask is a constant splat. This allows bitwise operations to make use of bsel. It's also a stepping stone towards matching bins[lr], and bins[lr]i from normal IR. Two sets of similar tests have been added in this commit. The bsel_* functions test the case where binsri cannot be used. The binsr_*_i functions will start to use the binsri instruction in the next commit. git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@193682 91177308-0d34-0410-b5e6-96231b3b80d8
2013-10-30 13:51:01 +00:00
ret void
; CHECK: .size binsr_v2i64_i
}
[mips][msa] Added support for matching bclr, and bclri from normal IR (i.e. not intrinsics) git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@194471 91177308-0d34-0410-b5e6-96231b3b80d8
2013-11-12 10:45:18 +00:00
define void @bclr_v16i8(<16 x i8>* %c, <16 x i8>* %a, <16 x i8>* %b) nounwind {
; CHECK: bclr_v16i8:
%1 = load <16 x i8>* %a
; CHECK-DAG: ld.b [[R1:\$w[0-9]+]], 0($5)
%2 = load <16 x i8>* %b
; CHECK-DAG: ld.b [[R2:\$w[0-9]+]], 0($6)
%3 = shl <16 x i8> <i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1>, %2
%4 = xor <16 x i8> %3, <i8 -1, i8 -1, i8 -1, i8 -1, i8 -1, i8 -1, i8 -1, i8 -1, i8 -1, i8 -1, i8 -1, i8 -1, i8 -1, i8 -1, i8 -1, i8 -1>
%5 = and <16 x i8> %1, %4
; CHECK-DAG: bclr.b [[R3:\$w[0-9]+]], [[R1]], [[R2]]
store <16 x i8> %5, <16 x i8>* %c
; CHECK-DAG: st.b [[R3]], 0($4)
ret void
; CHECK: .size bclr_v16i8
}
define void @bclr_v8i16(<8 x i16>* %c, <8 x i16>* %a, <8 x i16>* %b) nounwind {
; CHECK: bclr_v8i16:
%1 = load <8 x i16>* %a
; CHECK-DAG: ld.h [[R1:\$w[0-9]+]], 0($5)
%2 = load <8 x i16>* %b
; CHECK-DAG: ld.h [[R2:\$w[0-9]+]], 0($6)
%3 = shl <8 x i16> <i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1>, %2
%4 = xor <8 x i16> %3, <i16 -1, i16 -1, i16 -1, i16 -1, i16 -1, i16 -1, i16 -1, i16 -1>
%5 = and <8 x i16> %1, %4
; CHECK-DAG: bclr.h [[R3:\$w[0-9]+]], [[R1]], [[R2]]
store <8 x i16> %5, <8 x i16>* %c
; CHECK-DAG: st.h [[R3]], 0($4)
ret void
; CHECK: .size bclr_v8i16
}
define void @bclr_v4i32(<4 x i32>* %c, <4 x i32>* %a, <4 x i32>* %b) nounwind {
; CHECK: bclr_v4i32:
%1 = load <4 x i32>* %a
; CHECK-DAG: ld.w [[R1:\$w[0-9]+]], 0($5)
%2 = load <4 x i32>* %b
; CHECK-DAG: ld.w [[R2:\$w[0-9]+]], 0($6)
%3 = shl <4 x i32> <i32 1, i32 1, i32 1, i32 1>, %2
%4 = xor <4 x i32> %3, <i32 -1, i32 -1, i32 -1, i32 -1>
%5 = and <4 x i32> %1, %4
; CHECK-DAG: bclr.w [[R3:\$w[0-9]+]], [[R1]], [[R2]]
store <4 x i32> %5, <4 x i32>* %c
; CHECK-DAG: st.w [[R3]], 0($4)
ret void
; CHECK: .size bclr_v4i32
}
define void @bclr_v2i64(<2 x i64>* %c, <2 x i64>* %a, <2 x i64>* %b) nounwind {
; CHECK: bclr_v2i64:
%1 = load <2 x i64>* %a
; CHECK-DAG: ld.d [[R1:\$w[0-9]+]], 0($5)
%2 = load <2 x i64>* %b
; CHECK-DAG: ld.d [[R2:\$w[0-9]+]], 0($6)
%3 = shl <2 x i64> <i64 1, i64 1>, %2
%4 = xor <2 x i64> %3, <i64 -1, i64 -1>
%5 = and <2 x i64> %1, %4
; CHECK-DAG: bclr.d [[R3:\$w[0-9]+]], [[R1]], [[R2]]
store <2 x i64> %5, <2 x i64>* %c
; CHECK-DAG: st.d [[R3]], 0($4)
ret void
; CHECK: .size bclr_v2i64
}
[mips][msa] Added support for matching bset, bseti, bneg, and bnegi from normal IR (i.e. not intrinsics) git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@194469 91177308-0d34-0410-b5e6-96231b3b80d8
2013-11-12 10:31:49 +00:00
define void @bset_v16i8(<16 x i8>* %c, <16 x i8>* %a, <16 x i8>* %b) nounwind {
; CHECK: bset_v16i8:
%1 = load <16 x i8>* %a
; CHECK-DAG: ld.b [[R1:\$w[0-9]+]], 0($5)
%2 = load <16 x i8>* %b
; CHECK-DAG: ld.b [[R2:\$w[0-9]+]], 0($6)
%3 = shl <16 x i8> <i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1>, %2
%4 = or <16 x i8> %1, %3
; CHECK-DAG: bset.b [[R3:\$w[0-9]+]], [[R1]], [[R2]]
store <16 x i8> %4, <16 x i8>* %c
; CHECK-DAG: st.b [[R3]], 0($4)
ret void
; CHECK: .size bset_v16i8
}
define void @bset_v8i16(<8 x i16>* %c, <8 x i16>* %a, <8 x i16>* %b) nounwind {
; CHECK: bset_v8i16:
%1 = load <8 x i16>* %a
; CHECK-DAG: ld.h [[R1:\$w[0-9]+]], 0($5)
%2 = load <8 x i16>* %b
; CHECK-DAG: ld.h [[R2:\$w[0-9]+]], 0($6)
%3 = shl <8 x i16> <i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1>, %2
%4 = or <8 x i16> %1, %3
; CHECK-DAG: bset.h [[R3:\$w[0-9]+]], [[R1]], [[R2]]
store <8 x i16> %4, <8 x i16>* %c
; CHECK-DAG: st.h [[R3]], 0($4)
ret void
; CHECK: .size bset_v8i16
}
define void @bset_v4i32(<4 x i32>* %c, <4 x i32>* %a, <4 x i32>* %b) nounwind {
; CHECK: bset_v4i32:
%1 = load <4 x i32>* %a
; CHECK-DAG: ld.w [[R1:\$w[0-9]+]], 0($5)
%2 = load <4 x i32>* %b
; CHECK-DAG: ld.w [[R2:\$w[0-9]+]], 0($6)
%3 = shl <4 x i32> <i32 1, i32 1, i32 1, i32 1>, %2
%4 = or <4 x i32> %1, %3
; CHECK-DAG: bset.w [[R3:\$w[0-9]+]], [[R1]], [[R2]]
store <4 x i32> %4, <4 x i32>* %c
; CHECK-DAG: st.w [[R3]], 0($4)
ret void
; CHECK: .size bset_v4i32
}
define void @bset_v2i64(<2 x i64>* %c, <2 x i64>* %a, <2 x i64>* %b) nounwind {
; CHECK: bset_v2i64:
%1 = load <2 x i64>* %a
; CHECK-DAG: ld.d [[R1:\$w[0-9]+]], 0($5)
%2 = load <2 x i64>* %b
; CHECK-DAG: ld.d [[R2:\$w[0-9]+]], 0($6)
%3 = shl <2 x i64> <i64 1, i64 1>, %2
%4 = or <2 x i64> %1, %3
; CHECK-DAG: bset.d [[R3:\$w[0-9]+]], [[R1]], [[R2]]
store <2 x i64> %4, <2 x i64>* %c
; CHECK-DAG: st.d [[R3]], 0($4)
ret void
; CHECK: .size bset_v2i64
}
define void @bneg_v16i8(<16 x i8>* %c, <16 x i8>* %a, <16 x i8>* %b) nounwind {
; CHECK: bneg_v16i8:
%1 = load <16 x i8>* %a
; CHECK-DAG: ld.b [[R1:\$w[0-9]+]], 0($5)
%2 = load <16 x i8>* %b
; CHECK-DAG: ld.b [[R2:\$w[0-9]+]], 0($6)
%3 = shl <16 x i8> <i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1>, %2
%4 = xor <16 x i8> %1, %3
; CHECK-DAG: bneg.b [[R3:\$w[0-9]+]], [[R1]], [[R2]]
store <16 x i8> %4, <16 x i8>* %c
; CHECK-DAG: st.b [[R3]], 0($4)
ret void
; CHECK: .size bneg_v16i8
}
define void @bneg_v8i16(<8 x i16>* %c, <8 x i16>* %a, <8 x i16>* %b) nounwind {
; CHECK: bneg_v8i16:
%1 = load <8 x i16>* %a
; CHECK-DAG: ld.h [[R1:\$w[0-9]+]], 0($5)
%2 = load <8 x i16>* %b
; CHECK-DAG: ld.h [[R2:\$w[0-9]+]], 0($6)
%3 = shl <8 x i16> <i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1>, %2
%4 = xor <8 x i16> %1, %3
; CHECK-DAG: bneg.h [[R3:\$w[0-9]+]], [[R1]], [[R2]]
store <8 x i16> %4, <8 x i16>* %c
; CHECK-DAG: st.h [[R3]], 0($4)
ret void
; CHECK: .size bneg_v8i16
}
define void @bneg_v4i32(<4 x i32>* %c, <4 x i32>* %a, <4 x i32>* %b) nounwind {
; CHECK: bneg_v4i32:
%1 = load <4 x i32>* %a
; CHECK-DAG: ld.w [[R1:\$w[0-9]+]], 0($5)
%2 = load <4 x i32>* %b
; CHECK-DAG: ld.w [[R2:\$w[0-9]+]], 0($6)
%3 = shl <4 x i32> <i32 1, i32 1, i32 1, i32 1>, %2
%4 = xor <4 x i32> %1, %3
; CHECK-DAG: bneg.w [[R3:\$w[0-9]+]], [[R1]], [[R2]]
store <4 x i32> %4, <4 x i32>* %c
; CHECK-DAG: st.w [[R3]], 0($4)
ret void
; CHECK: .size bneg_v4i32
}
define void @bneg_v2i64(<2 x i64>* %c, <2 x i64>* %a, <2 x i64>* %b) nounwind {
; CHECK: bneg_v2i64:
%1 = load <2 x i64>* %a
; CHECK-DAG: ld.d [[R1:\$w[0-9]+]], 0($5)
%2 = load <2 x i64>* %b
; CHECK-DAG: ld.d [[R2:\$w[0-9]+]], 0($6)
%3 = shl <2 x i64> <i64 1, i64 1>, %2
%4 = xor <2 x i64> %1, %3
; CHECK-DAG: bneg.d [[R3:\$w[0-9]+]], [[R1]], [[R2]]
store <2 x i64> %4, <2 x i64>* %c
; CHECK-DAG: st.d [[R3]], 0($4)
ret void
; CHECK: .size bneg_v2i64
}
[mips][msa] Added support for matching bclr, and bclri from normal IR (i.e. not intrinsics) git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@194471 91177308-0d34-0410-b5e6-96231b3b80d8
2013-11-12 10:45:18 +00:00
define void @bclri_v16i8(<16 x i8>* %c, <16 x i8>* %a) nounwind {
; CHECK: bclri_v16i8:
%1 = load <16 x i8>* %a
; CHECK-DAG: ld.b [[R1:\$w[0-9]+]], 0($5)
%2 = xor <16 x i8> <i8 8, i8 8, i8 8, i8 8, i8 8, i8 8, i8 8, i8 8, i8 8, i8 8, i8 8, i8 8, i8 8, i8 8, i8 8, i8 8>,
<i8 -1, i8 -1, i8 -1, i8 -1, i8 -1, i8 -1, i8 -1, i8 -1, i8 -1, i8 -1, i8 -1, i8 -1, i8 -1, i8 -1, i8 -1, i8 -1>
%3 = and <16 x i8> %1, %2
; bclri.b and andi.b are exactly equivalent.
; CHECK-DAG: andi.b [[R3:\$w[0-9]+]], [[R1]], 247
store <16 x i8> %3, <16 x i8>* %c
; CHECK-DAG: st.b [[R3]], 0($4)
ret void
; CHECK: .size bclri_v16i8
}
define void @bclri_v8i16(<8 x i16>* %c, <8 x i16>* %a) nounwind {
; CHECK: bclri_v8i16:
%1 = load <8 x i16>* %a
; CHECK-DAG: ld.h [[R1:\$w[0-9]+]], 0($5)
%2 = xor <8 x i16> <i16 8, i16 8, i16 8, i16 8, i16 8, i16 8, i16 8, i16 8>,
<i16 -1, i16 -1, i16 -1, i16 -1, i16 -1, i16 -1, i16 -1, i16 -1>
%3 = and <8 x i16> %1, %2
; CHECK-DAG: bclri.h [[R3:\$w[0-9]+]], [[R1]], 3
store <8 x i16> %3, <8 x i16>* %c
; CHECK-DAG: st.h [[R3]], 0($4)
ret void
; CHECK: .size bclri_v8i16
}
define void @bclri_v4i32(<4 x i32>* %c, <4 x i32>* %a) nounwind {
; CHECK: bclri_v4i32:
%1 = load <4 x i32>* %a
; CHECK-DAG: ld.w [[R1:\$w[0-9]+]], 0($5)
%2 = xor <4 x i32> <i32 8, i32 8, i32 8, i32 8>,
<i32 -1, i32 -1, i32 -1, i32 -1>
%3 = and <4 x i32> %1, %2
; CHECK-DAG: bclri.w [[R3:\$w[0-9]+]], [[R1]], 3
store <4 x i32> %3, <4 x i32>* %c
; CHECK-DAG: st.w [[R3]], 0($4)
ret void
; CHECK: .size bclri_v4i32
}
define void @bclri_v2i64(<2 x i64>* %c, <2 x i64>* %a) nounwind {
; CHECK: bclri_v2i64:
%1 = load <2 x i64>* %a
; CHECK-DAG: ld.d [[R1:\$w[0-9]+]], 0($5)
%2 = xor <2 x i64> <i64 8, i64 8>,
<i64 -1, i64 -1>
%3 = and <2 x i64> %1, %2
; CHECK-DAG: bclri.d [[R3:\$w[0-9]+]], [[R1]], 3
store <2 x i64> %3, <2 x i64>* %c
; CHECK-DAG: st.d [[R3]], 0($4)
ret void
; CHECK: .size bclri_v2i64
}
[mips][msa] Added support for matching bset, bseti, bneg, and bnegi from normal IR (i.e. not intrinsics) git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@194469 91177308-0d34-0410-b5e6-96231b3b80d8
2013-11-12 10:31:49 +00:00
define void @bseti_v16i8(<16 x i8>* %c, <16 x i8>* %a) nounwind {
; CHECK: bseti_v16i8:
%1 = load <16 x i8>* %a
; CHECK-DAG: ld.b [[R1:\$w[0-9]+]], 0($5)
%2 = or <16 x i8> %1, <i8 8, i8 8, i8 8, i8 8, i8 8, i8 8, i8 8, i8 8, i8 8, i8 8, i8 8, i8 8, i8 8, i8 8, i8 8, i8 8>
; CHECK-DAG: bseti.b [[R3:\$w[0-9]+]], [[R1]], 3
store <16 x i8> %2, <16 x i8>* %c
; CHECK-DAG: st.b [[R3]], 0($4)
ret void
; CHECK: .size bseti_v16i8
}
define void @bseti_v8i16(<8 x i16>* %c, <8 x i16>* %a) nounwind {
; CHECK: bseti_v8i16:
%1 = load <8 x i16>* %a
; CHECK-DAG: ld.h [[R1:\$w[0-9]+]], 0($5)
%2 = or <8 x i16> %1, <i16 8, i16 8, i16 8, i16 8, i16 8, i16 8, i16 8, i16 8>
; CHECK-DAG: bseti.h [[R3:\$w[0-9]+]], [[R1]], 3
store <8 x i16> %2, <8 x i16>* %c
; CHECK-DAG: st.h [[R3]], 0($4)
ret void
; CHECK: .size bseti_v8i16
}
define void @bseti_v4i32(<4 x i32>* %c, <4 x i32>* %a) nounwind {
; CHECK: bseti_v4i32:
%1 = load <4 x i32>* %a
; CHECK-DAG: ld.w [[R1:\$w[0-9]+]], 0($5)
%2 = or <4 x i32> %1, <i32 8, i32 8, i32 8, i32 8>
; CHECK-DAG: bseti.w [[R3:\$w[0-9]+]], [[R1]], 3
store <4 x i32> %2, <4 x i32>* %c
; CHECK-DAG: st.w [[R3]], 0($4)
ret void
; CHECK: .size bseti_v4i32
}
define void @bseti_v2i64(<2 x i64>* %c, <2 x i64>* %a) nounwind {
; CHECK: bseti_v2i64:
%1 = load <2 x i64>* %a
; CHECK-DAG: ld.d [[R1:\$w[0-9]+]], 0($5)
%2 = or <2 x i64> %1, <i64 8, i64 8>
; CHECK-DAG: bseti.d [[R3:\$w[0-9]+]], [[R1]], 3
store <2 x i64> %2, <2 x i64>* %c
; CHECK-DAG: st.d [[R3]], 0($4)
ret void
; CHECK: .size bseti_v2i64
}
define void @bnegi_v16i8(<16 x i8>* %c, <16 x i8>* %a) nounwind {
; CHECK: bnegi_v16i8:
%1 = load <16 x i8>* %a
; CHECK-DAG: ld.b [[R1:\$w[0-9]+]], 0($5)
%2 = xor <16 x i8> %1, <i8 8, i8 8, i8 8, i8 8, i8 8, i8 8, i8 8, i8 8, i8 8, i8 8, i8 8, i8 8, i8 8, i8 8, i8 8, i8 8>
; CHECK-DAG: bnegi.b [[R3:\$w[0-9]+]], [[R1]], 3
store <16 x i8> %2, <16 x i8>* %c
; CHECK-DAG: st.b [[R3]], 0($4)
ret void
; CHECK: .size bnegi_v16i8
}
define void @bnegi_v8i16(<8 x i16>* %c, <8 x i16>* %a) nounwind {
; CHECK: bnegi_v8i16:
%1 = load <8 x i16>* %a
; CHECK-DAG: ld.h [[R1:\$w[0-9]+]], 0($5)
%2 = xor <8 x i16> %1, <i16 8, i16 8, i16 8, i16 8, i16 8, i16 8, i16 8, i16 8>
; CHECK-DAG: bnegi.h [[R3:\$w[0-9]+]], [[R1]], 3
store <8 x i16> %2, <8 x i16>* %c
; CHECK-DAG: st.h [[R3]], 0($4)
ret void
; CHECK: .size bnegi_v8i16
}
define void @bnegi_v4i32(<4 x i32>* %c, <4 x i32>* %a) nounwind {
; CHECK: bnegi_v4i32:
%1 = load <4 x i32>* %a
; CHECK-DAG: ld.w [[R1:\$w[0-9]+]], 0($5)
%2 = xor <4 x i32> %1, <i32 8, i32 8, i32 8, i32 8>
; CHECK-DAG: bnegi.w [[R3:\$w[0-9]+]], [[R1]], 3
store <4 x i32> %2, <4 x i32>* %c
; CHECK-DAG: st.w [[R3]], 0($4)
ret void
; CHECK: .size bnegi_v4i32
}
define void @bnegi_v2i64(<2 x i64>* %c, <2 x i64>* %a) nounwind {
; CHECK: bnegi_v2i64:
%1 = load <2 x i64>* %a
; CHECK-DAG: ld.d [[R1:\$w[0-9]+]], 0($5)
%2 = xor <2 x i64> %1, <i64 8, i64 8>
; CHECK-DAG: bnegi.d [[R3:\$w[0-9]+]], [[R1]], 3
store <2 x i64> %2, <2 x i64>* %c
; CHECK-DAG: st.d [[R3]], 0($4)
ret void
; CHECK: .size bnegi_v2i64
}
[mips][msa] Added support for matching pcnt from normal IR (i.e. not intrinsics) git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@191198 91177308-0d34-0410-b5e6-96231b3b80d8
2013-09-23 13:40:21 +00:00
declare <16 x i8> @llvm.ctpop.v16i8(<16 x i8> %val)
declare <8 x i16> @llvm.ctpop.v8i16(<8 x i16> %val)
declare <4 x i32> @llvm.ctpop.v4i32(<4 x i32> %val)
declare <2 x i64> @llvm.ctpop.v2i64(<2 x i64> %val)
[mips][msa] Added test cases that were supposed to be part of r190507, r190509, r190512, and r190518. git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@190522 91177308-0d34-0410-b5e6-96231b3b80d8
2013-09-11 12:39:25 +00:00
declare <16 x i8> @llvm.ctlz.v16i8(<16 x i8> %val)
declare <8 x i16> @llvm.ctlz.v8i16(<8 x i16> %val)
declare <4 x i32> @llvm.ctlz.v4i32(<4 x i32> %val)
declare <2 x i64> @llvm.ctlz.v2i64(<2 x i64> %val)
Reference in New Issue Copy Permalink