[AVX512] Adding VPERMQ VPERMPD Intrinsics

[delena]

LGTM

llvm/trunk/include/llvm/IR/IntrinsicsX86.td

def int_x86_avx512_mask_psrl_qi_128: GCCBuiltin<"__builtin_ia32_psrlqi128_mask">,
                    llvm_i8_ty, llvm_v2i64_ty,  llvm_i8_ty], [IntrNoMem]>;
   def int_x86_avx512_mask_psrl_qi_256: GCCBuiltin<"__builtin_ia32_psrlqi256_mask">,
         Intrinsic<[llvm_v4i64_ty], [llvm_v4i64_ty,
                    llvm_i8_ty, llvm_v4i64_ty,  llvm_i8_ty], [IntrNoMem]>;
   def int_x86_avx512_mask_psrl_qi_512: GCCBuiltin<"__builtin_ia32_psrlqi512_mask">,
         Intrinsic<[llvm_v8i64_ty], [llvm_v8i64_ty, 
                    llvm_i8_ty, llvm_v8i64_ty,  llvm_i8_ty], [IntrNoMem]>;
 }
-
+// Permute
+let TargetPrefix = "x86" in {  // All intrinsics start with "llvm.x86.".
+  def int_x86_avx512_mask_perm_df_256 : GCCBuiltin<"__builtin_ia32_permdf256_mask">,
+              Intrinsic<[llvm_v4f64_ty], [llvm_v4f64_ty,
+                        llvm_i8_ty, llvm_v4f64_ty, llvm_i8_ty], [IntrNoMem]>;
+  def int_x86_avx512_mask_perm_df_512 : GCCBuiltin<"__builtin_ia32_permdf512_mask">,
+              Intrinsic<[llvm_v8f64_ty], [llvm_v8f64_ty,
+                         llvm_i8_ty, llvm_v8f64_ty, llvm_i8_ty], [IntrNoMem]>;
+  def int_x86_avx512_mask_perm_di_256 : GCCBuiltin<"__builtin_ia32_permdi256_mask">,
+              Intrinsic<[llvm_v4i64_ty], [llvm_v4i64_ty,
+                        llvm_i8_ty, llvm_v4i64_ty, llvm_i8_ty], [IntrNoMem]>;
+  def int_x86_avx512_mask_perm_di_512 : GCCBuiltin<"__builtin_ia32_permdi512_mask">,
+              Intrinsic<[llvm_v8i64_ty], [llvm_v8i64_ty,
+                        llvm_i8_ty, llvm_v8i64_ty, llvm_i8_ty], [IntrNoMem]>;
+}
 // Pack ops.
 let TargetPrefix = "x86" in {  // All intrinsics start with "llvm.x86.".
   def int_x86_avx2_packsswb : GCCBuiltin<"__builtin_ia32_packsswb256">,
               Intrinsic<[llvm_v32i8_ty], [llvm_v16i16_ty,
                          llvm_v16i16_ty], [IntrNoMem]>;
   def int_x86_avx2_packssdw : GCCBuiltin<"__builtin_ia32_packssdw256">,
               Intrinsic<[llvm_v16i16_ty], [llvm_v8i32_ty,
                          llvm_v8i32_ty], [IntrNoMem]>;

llvm/trunk/lib/Target/X86/X86IntrinsicsInfo.h

static const IntrinsicData  IntrinsicsWithoutChain[] = {
   X86_INTRINSIC_DATA(avx512_mask_pcmpgt_d_256,  CMP_MASK,  X86ISD::PCMPGTM, 0),
   X86_INTRINSIC_DATA(avx512_mask_pcmpgt_d_512,  CMP_MASK,  X86ISD::PCMPGTM, 0),
   X86_INTRINSIC_DATA(avx512_mask_pcmpgt_q_128,  CMP_MASK,  X86ISD::PCMPGTM, 0),
   X86_INTRINSIC_DATA(avx512_mask_pcmpgt_q_256,  CMP_MASK,  X86ISD::PCMPGTM, 0),
   X86_INTRINSIC_DATA(avx512_mask_pcmpgt_q_512,  CMP_MASK,  X86ISD::PCMPGTM, 0),
   X86_INTRINSIC_DATA(avx512_mask_pcmpgt_w_128,  CMP_MASK,  X86ISD::PCMPGTM, 0),
   X86_INTRINSIC_DATA(avx512_mask_pcmpgt_w_256,  CMP_MASK,  X86ISD::PCMPGTM, 0),
   X86_INTRINSIC_DATA(avx512_mask_pcmpgt_w_512,  CMP_MASK,  X86ISD::PCMPGTM, 0),
+  X86_INTRINSIC_DATA(avx512_mask_perm_df_256, INTR_TYPE_2OP_MASK, X86ISD::VPERMI, 0),
+  X86_INTRINSIC_DATA(avx512_mask_perm_df_512, INTR_TYPE_2OP_MASK, X86ISD::VPERMI, 0),
+  X86_INTRINSIC_DATA(avx512_mask_perm_di_256, INTR_TYPE_2OP_MASK, X86ISD::VPERMI, 0),
+  X86_INTRINSIC_DATA(avx512_mask_perm_di_512, INTR_TYPE_2OP_MASK, X86ISD::VPERMI, 0),
   X86_INTRINSIC_DATA(avx512_mask_pmaddubs_w_128, INTR_TYPE_2OP_MASK,
                      X86ISD::VPMADDUBSW, 0),
   X86_INTRINSIC_DATA(avx512_mask_pmaddubs_w_256, INTR_TYPE_2OP_MASK,
                      X86ISD::VPMADDUBSW, 0),
   X86_INTRINSIC_DATA(avx512_mask_pmaddubs_w_512, INTR_TYPE_2OP_MASK,
                      X86ISD::VPMADDUBSW, 0),
   X86_INTRINSIC_DATA(avx512_mask_pmaddw_d_128, INTR_TYPE_2OP_MASK,
                      X86ISD::VPMADDWD, 0),

llvm/trunk/test/CodeGen/X86/avx512-intrinsics.ll

 ; CHECK-NEXT:    retq 
   %res = call <8 x i64> @llvm.x86.avx512.mask.pmovsxw.q.512(<8 x i16> %x0, <8 x i64> %x1, i8 %x2)
   %res1 = call <8 x i64> @llvm.x86.avx512.mask.pmovsxw.q.512(<8 x i16> %x0, <8 x i64> zeroinitializer, i8 %x2)
   %res2 = call <8 x i64> @llvm.x86.avx512.mask.pmovsxw.q.512(<8 x i16> %x0, <8 x i64> %x1, i8 -1)
   %res3 = add <8 x i64> %res, %res1
   %res4 = add <8 x i64> %res3, %res2
   ret <8 x i64> %res4
 }
+
+declare <8 x double> @llvm.x86.avx512.mask.perm.df.512(<8 x double>, i8, <8 x double>, i8)
+
+define <8 x double>@test_int_x86_avx512_mask_perm_df_512(<8 x double> %x0, i8 %x1, <8 x double> %x2, i8 %x3) {
+; CHECK-LABEL: test_int_x86_avx512_mask_perm_df_512:
+; CHECK:       ## BB#0:
+; CHECK-NEXT:    movzbl %sil, %eax 
+; CHECK-NEXT:    kmovw %eax, %k1 
+; CHECK-NEXT:    vpermpd $3, %zmm0, %zmm1 {%k1} 
+; CHECK-NEXT:    vpermpd $3, %zmm0, %zmm2 {%k1} {z} 
+; CHECK-NEXT:    vpermpd $3, %zmm0, %zmm0 
+; CHECK-NEXT:    vaddpd %zmm2, %zmm1, %zmm1 
+; CHECK-NEXT:    vaddpd %zmm0, %zmm1, %zmm0 
+; CHECK-NEXT:    retq 
+  %res = call <8 x double> @llvm.x86.avx512.mask.perm.df.512(<8 x double> %x0, i8 3, <8 x double> %x2, i8 %x3)
+  %res1 = call <8 x double> @llvm.x86.avx512.mask.perm.df.512(<8 x double> %x0, i8 3, <8 x double> zeroinitializer, i8 %x3)
+  %res2 = call <8 x double> @llvm.x86.avx512.mask.perm.df.512(<8 x double> %x0, i8 3, <8 x double> %x2, i8 -1)
+  %res3 = fadd <8 x double> %res, %res1
+  %res4 = fadd <8 x double> %res3, %res2
+  ret <8 x double> %res4
+}
+
+declare <8 x i64> @llvm.x86.avx512.mask.perm.di.512(<8 x i64>, i8, <8 x i64>, i8)
+
+define <8 x i64>@test_int_x86_avx512_mask_perm_di_512(<8 x i64> %x0, i8 %x1, <8 x i64> %x2, i8 %x3) {
+; CHECK-LABEL: test_int_x86_avx512_mask_perm_di_512:
+; CHECK:       ## BB#0:
+; CHECK-NEXT:    movzbl %sil, %eax 
+; CHECK-NEXT:    kmovw %eax, %k1 
+; CHECK-NEXT:    vpermq $3, %zmm0, %zmm1 {%k1} 
+; CHECK-NEXT:    vpermq $3, %zmm0, %zmm2 {%k1} {z} 
+; CHECK-NEXT:    vpermq $3, %zmm0, %zmm0 
+; CHECK-NEXT:    vpaddq %zmm2, %zmm1, %zmm1 
+; CHECK-NEXT:    vpaddq %zmm0, %zmm1, %zmm0 
+; CHECK-NEXT:    retq 
+  %res = call <8 x i64> @llvm.x86.avx512.mask.perm.di.512(<8 x i64> %x0, i8 3, <8 x i64> %x2, i8 %x3)
+  %res1 = call <8 x i64> @llvm.x86.avx512.mask.perm.di.512(<8 x i64> %x0, i8 3, <8 x i64> zeroinitializer, i8 %x3)
+  %res2 = call <8 x i64> @llvm.x86.avx512.mask.perm.di.512(<8 x i64> %x0, i8 3, <8 x i64> %x2, i8 -1)
+  %res3 = add <8 x i64> %res, %res1
+  %res4 = add <8 x i64> %res3, %res2
+  ret <8 x i64> %res4
+}

llvm/trunk/test/CodeGen/X86/avx512vl-intrinsics.ll

 ; CHECK-NEXT:    retq 
   %res = call <4 x i64> @llvm.x86.avx512.mask.pmovsxw.q.256(<8 x i16> %x0, <4 x i64> %x1, i8 %x2)
   %res1 = call <4 x i64> @llvm.x86.avx512.mask.pmovsxw.q.256(<8 x i16> %x0, <4 x i64> zeroinitializer, i8 %x2)
   %res2 = call <4 x i64> @llvm.x86.avx512.mask.pmovsxw.q.256(<8 x i16> %x0, <4 x i64> %x1, i8 -1)
   %res3 = add <4 x i64> %res, %res1
   %res4 = add <4 x i64> %res3, %res2
   ret <4 x i64> %res4
 }
+
+declare <4 x double> @llvm.x86.avx512.mask.perm.df.256(<4 x double>, i8, <4 x double>, i8)
+
+define <4 x double>@test_int_x86_avx512_mask_perm_df_256(<4 x double> %x0, i8 %x1, <4 x double> %x2, i8 %x3) {
+; CHECK-LABEL: test_int_x86_avx512_mask_perm_df_256:
+; CHECK:       ## BB#0:
+; CHECK-NEXT:    movzbl %sil, %eax 
+; CHECK-NEXT:    kmovw %eax, %k1 
+; CHECK-NEXT:    vpermpd $3, %ymm0, %ymm1 {%k1} 
+; CHECK-NEXT:    vpermpd $3, %ymm0, %ymm2 {%k1} {z} 
+; CHECK-NEXT:    vpermpd $3, %ymm0, %ymm0 
+; CHECK-NEXT:    ## ymm0 = ymm0[3,0,0,0]
+; CHECK-NEXT:    vaddpd %ymm2, %ymm1, %ymm1 
+; CHECK-NEXT:    vaddpd %ymm0, %ymm1, %ymm0 
+; CHECK-NEXT:    retq 
+  %res = call <4 x double> @llvm.x86.avx512.mask.perm.df.256(<4 x double> %x0, i8 3, <4 x double> %x2, i8 %x3)
+  %res1 = call <4 x double> @llvm.x86.avx512.mask.perm.df.256(<4 x double> %x0, i8 3, <4 x double> zeroinitializer, i8 %x3)
+  %res2 = call <4 x double> @llvm.x86.avx512.mask.perm.df.256(<4 x double> %x0, i8 3, <4 x double> %x2, i8 -1)
+  %res3 = fadd <4 x double> %res, %res1
+  %res4 = fadd <4 x double> %res3, %res2
+  ret <4 x double> %res4
+}
+
+declare <4 x i64> @llvm.x86.avx512.mask.perm.di.256(<4 x i64>, i8, <4 x i64>, i8)
+
+define <4 x i64>@test_int_x86_avx512_mask_perm_di_256(<4 x i64> %x0, i8 %x1, <4 x i64> %x2, i8 %x3) {
+; CHECK-LABEL: test_int_x86_avx512_mask_perm_di_256:
+; CHECK:       ## BB#0:
+; CHECK-NEXT:    movzbl %sil, %eax 
+; CHECK-NEXT:    kmovw %eax, %k1 
+; CHECK-NEXT:    vpermq $3, %ymm0, %ymm1 {%k1} 
+; CHECK-NEXT:    vpermq $3, %ymm0, %ymm2 {%k1} {z} 
+; CHECK-NEXT:    vpermq $3, %ymm0, %ymm0 
+; CHECK-NEXT:    ## ymm0 = ymm0[3,0,0,0]
+; CHECK-NEXT:    vpaddq %ymm2, %ymm1, %ymm1 
+; CHECK-NEXT:    vpaddq %ymm0, %ymm1, %ymm0 
+; CHECK-NEXT:    retq 
+  %res = call <4 x i64> @llvm.x86.avx512.mask.perm.di.256(<4 x i64> %x0, i8 3, <4 x i64> %x2, i8 %x3)
+  %res1 = call <4 x i64> @llvm.x86.avx512.mask.perm.di.256(<4 x i64> %x0, i8 3, <4 x i64> zeroinitializer, i8 %x3)
+  %res2 = call <4 x i64> @llvm.x86.avx512.mask.perm.di.256(<4 x i64> %x0, i8 3, <4 x i64> %x2, i8 -1)
+  %res3 = add <4 x i64> %res, %res1
+  %res4 = add <4 x i64> %res3, %res2
+  ret <4 x i64> %res4
+}