Skip to content

Commit

Permalink
AVX512: Fix vpmovzxbw predicate for AVX1/2 instructions.
Browse files Browse the repository at this point in the history 
Differential Revision: http://reviews.llvm.org/D16595

llvm-svn: 258915
  • Loading branch information
Igor Breger committed Jan 27, 2016
1 parent d6c187b commit b1bd47c
Show file tree
Hide file tree
Showing 2 changed files with 53 additions and 27 deletions.
73 changes: 46 additions & 27 deletions llvm/lib/Target/X86/X86InstrSSE.td
View file
Expand Up @@ -5890,45 +5890,48 @@ multiclass SS41I_pmovx_rrrm<bits<8> opc, string OpcodeStr, X86MemOperand MemOp,
multiclass SS41I_pmovx_rm_all<bits<8> opc, string OpcodeStr,
X86MemOperand MemOp, X86MemOperand MemYOp,
OpndItins SSEItins, OpndItins AVXItins,
OpndItins AVX2Itins> {
OpndItins AVX2Itins, Predicate prd> {
defm NAME : SS41I_pmovx_rrrm<opc, OpcodeStr, MemOp, VR128, VR128, SSEItins>;
let Predicates = [HasAVX, NoVLX] in
let Predicates = [HasAVX, prd] in
defm V#NAME : SS41I_pmovx_rrrm<opc, !strconcat("v", OpcodeStr), MemOp,
VR128, VR128, AVXItins>, VEX;
let Predicates = [HasAVX2, NoVLX] in
let Predicates = [HasAVX2, prd] in
defm V#NAME#Y : SS41I_pmovx_rrrm<opc, !strconcat("v", OpcodeStr), MemYOp,
VR256, VR128, AVX2Itins>, VEX, VEX_L;
}

multiclass SS41I_pmovx_rm<bits<8> opc, string OpcodeStr,
X86MemOperand MemOp, X86MemOperand MemYOp> {
multiclass SS41I_pmovx_rm<bits<8> opc, string OpcodeStr, X86MemOperand MemOp,
X86MemOperand MemYOp, Predicate prd> {
defm PMOVSX#NAME : SS41I_pmovx_rm_all<opc, !strconcat("pmovsx", OpcodeStr),
MemOp, MemYOp,
SSE_INTALU_ITINS_SHUFF_P,
DEFAULT_ITINS_SHUFFLESCHED,
DEFAULT_ITINS_SHUFFLESCHED>;
DEFAULT_ITINS_SHUFFLESCHED, prd>;
defm PMOVZX#NAME : SS41I_pmovx_rm_all<!add(opc, 0x10),
!strconcat("pmovzx", OpcodeStr),
MemOp, MemYOp,
SSE_INTALU_ITINS_SHUFF_P,
DEFAULT_ITINS_SHUFFLESCHED,
DEFAULT_ITINS_SHUFFLESCHED>;
DEFAULT_ITINS_SHUFFLESCHED, prd>;
}

defm BW : SS41I_pmovx_rm<0x20, "bw", i64mem, i128mem>;
defm WD : SS41I_pmovx_rm<0x23, "wd", i64mem, i128mem>;
defm DQ : SS41I_pmovx_rm<0x25, "dq", i64mem, i128mem>;
defm BW : SS41I_pmovx_rm<0x20, "bw", i64mem, i128mem, NoVLX_Or_NoBWI>;
defm WD : SS41I_pmovx_rm<0x23, "wd", i64mem, i128mem, NoVLX>;
defm DQ : SS41I_pmovx_rm<0x25, "dq", i64mem, i128mem, NoVLX>;

defm BD : SS41I_pmovx_rm<0x21, "bd", i32mem, i64mem>;
defm WQ : SS41I_pmovx_rm<0x24, "wq", i32mem, i64mem>;
defm BD : SS41I_pmovx_rm<0x21, "bd", i32mem, i64mem, NoVLX>;
defm WQ : SS41I_pmovx_rm<0x24, "wq", i32mem, i64mem, NoVLX>;

defm BQ : SS41I_pmovx_rm<0x22, "bq", i16mem, i32mem>;
defm BQ : SS41I_pmovx_rm<0x22, "bq", i16mem, i32mem, NoVLX>;

// AVX2 Patterns
multiclass SS41I_pmovx_avx2_patterns<string OpcPrefix, string ExtTy, SDNode ExtOp> {
// Register-Register patterns
let Predicates = [HasAVX, NoVLX_Or_NoBWI] in {
def : Pat<(v16i16 (ExtOp (v16i8 VR128:$src))),
(!cast<I>(OpcPrefix#BWYrr) VR128:$src)>;
}
let Predicates = [HasAVX, NoVLX] in {
def : Pat<(v8i32 (ExtOp (v16i8 VR128:$src))),
(!cast<I>(OpcPrefix#BDYrr) VR128:$src)>;
def : Pat<(v4i64 (ExtOp (v16i8 VR128:$src))),
Expand All @@ -5941,10 +5944,13 @@ multiclass SS41I_pmovx_avx2_patterns<string OpcPrefix, string ExtTy, SDNode ExtO

def : Pat<(v4i64 (ExtOp (v4i32 VR128:$src))),
(!cast<I>(OpcPrefix#DQYrr) VR128:$src)>;

}
// On AVX2, we also support 256bit inputs.
let Predicates = [HasAVX, NoVLX_Or_NoBWI] in {
def : Pat<(v16i16 (ExtOp (v32i8 VR256:$src))),
(!cast<I>(OpcPrefix#BWYrr) (EXTRACT_SUBREG VR256:$src, sub_xmm))>;
}
let Predicates = [HasAVX, NoVLX] in {
def : Pat<(v8i32 (ExtOp (v32i8 VR256:$src))),
(!cast<I>(OpcPrefix#BDYrr) (EXTRACT_SUBREG VR256:$src, sub_xmm))>;
def : Pat<(v4i64 (ExtOp (v32i8 VR256:$src))),
Expand All @@ -5957,10 +5963,14 @@ multiclass SS41I_pmovx_avx2_patterns<string OpcPrefix, string ExtTy, SDNode ExtO

def : Pat<(v4i64 (ExtOp (v8i32 VR256:$src))),
(!cast<I>(OpcPrefix#DQYrr) (EXTRACT_SUBREG VR256:$src, sub_xmm))>;
}

// Simple Register-Memory patterns
let Predicates = [HasAVX, NoVLX_Or_NoBWI] in {
def : Pat<(v16i16 (!cast<PatFrag>(ExtTy#"extloadvi8") addr:$src)),
(!cast<I>(OpcPrefix#BWYrm) addr:$src)>;
}
let Predicates = [HasAVX, NoVLX] in {
def : Pat<(v8i32 (!cast<PatFrag>(ExtTy#"extloadvi8") addr:$src)),
(!cast<I>(OpcPrefix#BDYrm) addr:$src)>;
def : Pat<(v4i64 (!cast<PatFrag>(ExtTy#"extloadvi8") addr:$src)),
Expand All @@ -5973,8 +5983,10 @@ multiclass SS41I_pmovx_avx2_patterns<string OpcPrefix, string ExtTy, SDNode ExtO

def : Pat<(v4i64 (!cast<PatFrag>(ExtTy#"extloadvi32") addr:$src)),
(!cast<I>(OpcPrefix#DQYrm) addr:$src)>;

}

// AVX2 Register-Memory patterns
let Predicates = [HasAVX, NoVLX_Or_NoBWI] in {
def : Pat<(v16i16 (ExtOp (bc_v16i8 (loadv2i64 addr:$src)))),
(!cast<I>(OpcPrefix#BWYrm) addr:$src)>;
def : Pat<(v16i16 (ExtOp (v16i8 (vzmovl_v2i64 addr:$src)))),
Expand All @@ -5983,7 +5995,8 @@ multiclass SS41I_pmovx_avx2_patterns<string OpcPrefix, string ExtTy, SDNode ExtO
(!cast<I>(OpcPrefix#BWYrm) addr:$src)>;
def : Pat<(v16i16 (ExtOp (bc_v16i8 (loadv2i64 addr:$src)))),
(!cast<I>(OpcPrefix#BWYrm) addr:$src)>;

}
let Predicates = [HasAVX, NoVLX] in {
def : Pat<(v8i32 (ExtOp (bc_v16i8 (v2i64 (scalar_to_vector (loadi64 addr:$src)))))),
(!cast<I>(OpcPrefix#BDYrm) addr:$src)>;
def : Pat<(v8i32 (ExtOp (v16i8 (vzmovl_v2i64 addr:$src)))),
Expand Down Expand Up @@ -6028,18 +6041,20 @@ multiclass SS41I_pmovx_avx2_patterns<string OpcPrefix, string ExtTy, SDNode ExtO
(!cast<I>(OpcPrefix#DQYrm) addr:$src)>;
def : Pat<(v4i64 (ExtOp (bc_v4i32 (loadv2i64 addr:$src)))),
(!cast<I>(OpcPrefix#DQYrm) addr:$src)>;
}
}

let Predicates = [HasAVX2, NoVLX] in {
defm : SS41I_pmovx_avx2_patterns<"VPMOVSX", "s", X86vsext>;
defm : SS41I_pmovx_avx2_patterns<"VPMOVZX", "z", X86vzext>;
}
defm : SS41I_pmovx_avx2_patterns<"VPMOVSX", "s", X86vsext>;
defm : SS41I_pmovx_avx2_patterns<"VPMOVZX", "z", X86vzext>;

// SSE4.1/AVX patterns.
multiclass SS41I_pmovx_patterns<string OpcPrefix, string ExtTy,
SDNode ExtOp, PatFrag ExtLoad16> {
let Predicates = [HasAVX, NoVLX_Or_NoBWI] in {
def : Pat<(v8i16 (ExtOp (v16i8 VR128:$src))),
(!cast<I>(OpcPrefix#BWrr) VR128:$src)>;
}
let Predicates = [HasAVX, NoVLX] in {
def : Pat<(v4i32 (ExtOp (v16i8 VR128:$src))),
(!cast<I>(OpcPrefix#BDrr) VR128:$src)>;
def : Pat<(v2i64 (ExtOp (v16i8 VR128:$src))),
Expand All @@ -6052,9 +6067,12 @@ multiclass SS41I_pmovx_patterns<string OpcPrefix, string ExtTy,

def : Pat<(v2i64 (ExtOp (v4i32 VR128:$src))),
(!cast<I>(OpcPrefix#DQrr) VR128:$src)>;

}
let Predicates = [HasAVX, NoVLX_Or_NoBWI] in {
def : Pat<(v8i16 (!cast<PatFrag>(ExtTy#"extloadvi8") addr:$src)),
(!cast<I>(OpcPrefix#BWrm) addr:$src)>;
}
let Predicates = [HasAVX, NoVLX] in {
def : Pat<(v4i32 (!cast<PatFrag>(ExtTy#"extloadvi8") addr:$src)),
(!cast<I>(OpcPrefix#BDrm) addr:$src)>;
def : Pat<(v2i64 (!cast<PatFrag>(ExtTy#"extloadvi8") addr:$src)),
Expand All @@ -6067,7 +6085,8 @@ multiclass SS41I_pmovx_patterns<string OpcPrefix, string ExtTy,

def : Pat<(v2i64 (!cast<PatFrag>(ExtTy#"extloadvi32") addr:$src)),
(!cast<I>(OpcPrefix#DQrm) addr:$src)>;

}
let Predicates = [HasAVX, NoVLX_Or_NoBWI] in {
def : Pat<(v8i16 (ExtOp (bc_v16i8 (v2i64 (scalar_to_vector (loadi64 addr:$src)))))),
(!cast<I>(OpcPrefix#BWrm) addr:$src)>;
def : Pat<(v8i16 (ExtOp (bc_v16i8 (v2f64 (scalar_to_vector (loadf64 addr:$src)))))),
Expand All @@ -6078,7 +6097,8 @@ multiclass SS41I_pmovx_patterns<string OpcPrefix, string ExtTy,
(!cast<I>(OpcPrefix#BWrm) addr:$src)>;
def : Pat<(v8i16 (ExtOp (bc_v16i8 (loadv2i64 addr:$src)))),
(!cast<I>(OpcPrefix#BWrm) addr:$src)>;

}
let Predicates = [HasAVX, NoVLX] in {
def : Pat<(v4i32 (ExtOp (bc_v16i8 (v4i32 (scalar_to_vector (loadi32 addr:$src)))))),
(!cast<I>(OpcPrefix#BDrm) addr:$src)>;
def : Pat<(v4i32 (ExtOp (v16i8 (vzmovl_v4i32 addr:$src)))),
Expand Down Expand Up @@ -6127,12 +6147,11 @@ multiclass SS41I_pmovx_patterns<string OpcPrefix, string ExtTy,
(!cast<I>(OpcPrefix#DQrm) addr:$src)>;
def : Pat<(v2i64 (ExtOp (bc_v4i32 (loadv2i64 addr:$src)))),
(!cast<I>(OpcPrefix#DQrm) addr:$src)>;
}
}

let Predicates = [HasAVX, NoVLX] in {
defm : SS41I_pmovx_patterns<"VPMOVSX", "s", X86vsext, extloadi32i16>;
defm : SS41I_pmovx_patterns<"VPMOVZX", "z", X86vzext, loadi16_anyext>;
}
defm : SS41I_pmovx_patterns<"VPMOVSX", "s", X86vsext, extloadi32i16>;
defm : SS41I_pmovx_patterns<"VPMOVZX", "z", X86vzext, loadi16_anyext>;

let Predicates = [UseSSE41] in {
defm : SS41I_pmovx_patterns<"PMOVSX", "s", X86vsext, extloadi32i16>;
Expand Down
7 changes: 7 additions & 0 deletions llvm/test/CodeGen/X86/avx-isa-check.ll
View file
Expand Up @@ -568,3 +568,10 @@ define <8 x i16> @shl_const_v8i16(<8 x i16> %a) {
%shift = shl <8 x i16> %a, <i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3>
ret <8 x i16> %shift
}

define <8 x i16> @zext_16i8_to_8i16(<16 x i8> %A) nounwind uwtable readnone ssp {
entry:
%B = shufflevector <16 x i8> %A, <16 x i8> undef, <8 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7>
%C = zext <8 x i8> %B to <8 x i16>
ret <8 x i16> %C
}

0 comments on commit b1bd47c

Please sign in to comment.