[AVX-512] Add VPTERNLOG to load folding tables.

llvm-svn: 281156

2 files changed, 36 insertions, 0 deletions

diff --git a/llvm/lib/Target/X86/X86InstrInfo.cpp b/llvm/lib/Target/X86/X86InstrInfo.cpp
index d51439c34ea..1e6b21a3653 100644
--- a/llvm/lib/Target/X86/X86InstrInfo.cpp
+++ b/llvm/lib/Target/X86/X86InstrInfo.cpp
@@ -2047,6 +2047,8 @@ X86InstrInfo::X86InstrInfo(X86Subtarget &STI)
     { X86::VPORQZrrkz,            X86::VPORQZrmkz,            0 },
     { X86::VPXORDZrrkz,           X86::VPXORDZrmkz,           0 },
     { X86::VPXORQZrrkz,           X86::VPXORQZrmkz,           0 },
+    { X86::VPTERNLOGDZrri,        X86::VPTERNLOGDZrmi,        0 },
+    { X86::VPTERNLOGQZrri,        X86::VPTERNLOGQZrmi,        0 },
     // AVX-512{F,VL} arithmetic instructions 256-bit
     { X86::VADDPSZ256rrkz,        X86::VADDPSZ256rmkz,        0 },
     { X86::VADDPDZ256rrkz,        X86::VADDPDZ256rmkz,        0 },
@@ -2080,6 +2082,8 @@ X86InstrInfo::X86InstrInfo(X86Subtarget &STI)
     { X86::VPORQZ256rrkz,         X86::VPORQZ256rmkz,         0 },
     { X86::VPXORDZ256rrkz,        X86::VPXORDZ256rmkz,        0 },
     { X86::VPXORQZ256rrkz,        X86::VPXORQZ256rmkz,        0 },
+    { X86::VPTERNLOGDZ256rri,     X86::VPTERNLOGDZ256rmi,     0 },
+    { X86::VPTERNLOGQZ256rri,     X86::VPTERNLOGQZ256rmi,     0 },
     // AVX-512{F,VL} arithmetic instructions 128-bit
     { X86::VADDPSZ128rrkz,        X86::VADDPSZ128rmkz,        0 },
     { X86::VADDPDZ128rrkz,        X86::VADDPDZ128rmkz,        0 },
@@ -2113,6 +2117,8 @@ X86InstrInfo::X86InstrInfo(X86Subtarget &STI)
     { X86::VPORQZ128rrkz,         X86::VPORQZ128rmkz,         0 },
     { X86::VPXORDZ128rrkz,        X86::VPXORDZ128rmkz,        0 },
     { X86::VPXORQZ128rrkz,        X86::VPXORQZ128rmkz,        0 },
+    { X86::VPTERNLOGDZ128rri,     X86::VPTERNLOGDZ128rmi,     0 },
+    { X86::VPTERNLOGQZ128rri,     X86::VPTERNLOGQZ128rmi,     0 },
   };
 
   for (X86MemoryFoldTableEntry Entry : MemoryFoldTable3) {
@@ -2163,6 +2169,10 @@ X86InstrInfo::X86InstrInfo(X86Subtarget &STI)
     { X86::VPORQZrrk,          X86::VPORQZrmk,            0 },
     { X86::VPXORDZrrk,         X86::VPXORDZrmk,           0 },
     { X86::VPXORQZrrk,         X86::VPXORQZrmk,           0 },
+    { X86::VPTERNLOGDZrrik,    X86::VPTERNLOGDZrmik,      0 },
+    { X86::VPTERNLOGDZrrikz,   X86::VPTERNLOGDZrmikz,     0 },
+    { X86::VPTERNLOGQZrrik,    X86::VPTERNLOGQZrmik,      0 },
+    { X86::VPTERNLOGQZrrikz,   X86::VPTERNLOGQZrmikz,     0 },
     // AVX-512{F,VL} foldable instructions 256-bit
     { X86::VADDPSZ256rrk,      X86::VADDPSZ256rmk,        0 },
     { X86::VADDPDZ256rrk,      X86::VADDPDZ256rmk,        0 },
@@ -2196,6 +2206,10 @@ X86InstrInfo::X86InstrInfo(X86Subtarget &STI)
     { X86::VPORQZ256rrk,       X86::VPORQZ256rmk,         0 },
     { X86::VPXORDZ256rrk,      X86::VPXORDZ256rmk,        0 },
     { X86::VPXORQZ256rrk,      X86::VPXORQZ256rmk,        0 },
+    { X86::VPTERNLOGDZ256rrik, X86::VPTERNLOGDZ256rmik,   0 },
+    { X86::VPTERNLOGDZ256rrikz,X86::VPTERNLOGDZ256rmikz,  0 },
+    { X86::VPTERNLOGQZ256rrik, X86::VPTERNLOGQZ256rmik,   0 },
+    { X86::VPTERNLOGQZ256rrikz,X86::VPTERNLOGQZ256rmikz,  0 },
     // AVX-512{F,VL} foldable instructions 128-bit
     { X86::VADDPSZ128rrk,      X86::VADDPSZ128rmk,        0 },
     { X86::VADDPDZ128rrk,      X86::VADDPDZ128rmk,        0 },
@@ -2229,6 +2243,10 @@ X86InstrInfo::X86InstrInfo(X86Subtarget &STI)
     { X86::VPORQZ128rrk,       X86::VPORQZ128rmk,         0 },
     { X86::VPXORDZ128rrk,      X86::VPXORDZ128rmk,        0 },
     { X86::VPXORQZ128rrk,      X86::VPXORQZ128rmk,        0 },
+    { X86::VPTERNLOGDZ128rrik, X86::VPTERNLOGDZ128rmik,   0 },
+    { X86::VPTERNLOGDZ128rrikz,X86::VPTERNLOGDZ128rmikz,  0 },
+    { X86::VPTERNLOGQZ128rrik, X86::VPTERNLOGQZ128rmik,   0 },
+    { X86::VPTERNLOGQZ128rrikz,X86::VPTERNLOGQZ128rmikz,  0 },
   };
 
   for (X86MemoryFoldTableEntry Entry : MemoryFoldTable4) {
diff --git a/llvm/test/CodeGen/X86/stack-folding-int-avx512.ll b/llvm/test/CodeGen/X86/stack-folding-int-avx512.ll
index 1cf1015a9b7..6b54acf28e1 100644
--- a/llvm/test/CodeGen/X86/stack-folding-int-avx512.ll
+++ b/llvm/test/CodeGen/X86/stack-folding-int-avx512.ll
@@ -180,3 +180,21 @@ define <32 x i16> @stack_fold_psubw(<32 x i16> %a0, <32 x i16> %a1) {
   ret <32 x i16> %2
 }
 
+define <16 x i32> @stack_fold_ternlogd(<16 x i32> %x0, <16 x i32> %x1, <16 x i32> %x2) {
+  ;CHECK-LABEL: stack_fold_ternlogd
+  ;CHECK:       vpternlogd $33, {{-?[0-9]*}}(%rsp), {{%zmm[0-9][0-9]*}}, {{%zmm[0-9][0-9]*}} {{.*#+}} 64-byte Folded Reload
+  %1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
+  %res = call <16 x i32> @llvm.x86.avx512.mask.pternlog.d.512(<16 x i32> %x0, <16 x i32> %x1, <16 x i32> %x2, i32 33, i16 -1)
+  ret <16 x i32> %res
+}
+declare <16 x i32> @llvm.x86.avx512.mask.pternlog.d.512(<16 x i32>, <16 x i32>, <16 x i32>, i32, i16)
+
+define <8 x i64> @stack_fold_ternlogq(<8 x i64> %x0, <8 x i64> %x1, <8 x i64> %x2) {
+  ;CHECK-LABEL: stack_fold_ternlogq
+  ;CHECK:       vpternlogq $33, {{-?[0-9]*}}(%rsp), {{%zmm[0-9][0-9]*}}, {{%zmm[0-9][0-9]*}} {{.*#+}} 64-byte Folded Reload
+  %1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
+  %res = call <8 x i64> @llvm.x86.avx512.mask.pternlog.q.512(<8 x i64> %x0, <8 x i64> %x1, <8 x i64> %x2, i32 33, i8 -1)
+  ret <8 x i64> %res
+}
+
+declare <8 x i64> @llvm.x86.avx512.mask.pternlog.q.512(<8 x i64>, <8 x i64>, <8 x i64>, i32, i8)