[AVX512] Make the permd intrinsics take a 32-bit immediate to match the software spec.

llvm-svn: 269579

4 files changed, 28 insertions, 28 deletions

diff --git a/llvm/include/llvm/IR/IntrinsicsX86.td b/llvm/include/llvm/IR/IntrinsicsX86.td
index e86c4554e69..7a00a13740a 100644
--- a/llvm/include/llvm/IR/IntrinsicsX86.td
+++ b/llvm/include/llvm/IR/IntrinsicsX86.td
@@ -2732,16 +2732,16 @@ let TargetPrefix = "x86" in {  // All intrinsics start with "llvm.x86.".
 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]>;
+                        llvm_i32_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]>;
+                         llvm_i32_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]>;
+                        llvm_i32_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]>;
+                        llvm_i32_ty, llvm_v8i64_ty, llvm_i8_ty], [IntrNoMem]>;
 }
 // Pack ops.
 let TargetPrefix = "x86" in {  // All intrinsics start with "llvm.x86.".
diff --git a/llvm/lib/Target/X86/X86IntrinsicsInfo.h b/llvm/lib/Target/X86/X86IntrinsicsInfo.h
index 5fbc61010c7..5be0a8dae7b 100644
--- a/llvm/lib/Target/X86/X86IntrinsicsInfo.h
+++ b/llvm/lib/Target/X86/X86IntrinsicsInfo.h
@@ -1089,10 +1089,10 @@ static const IntrinsicData  IntrinsicsWithoutChain[] = {
   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_perm_df_256, INTR_TYPE_2OP_IMM8_MASK, X86ISD::VPERMI, 0),
+  X86_INTRINSIC_DATA(avx512_mask_perm_df_512, INTR_TYPE_2OP_IMM8_MASK, X86ISD::VPERMI, 0),
+  X86_INTRINSIC_DATA(avx512_mask_perm_di_256, INTR_TYPE_2OP_IMM8_MASK, X86ISD::VPERMI, 0),
+  X86_INTRINSIC_DATA(avx512_mask_perm_di_512, INTR_TYPE_2OP_IMM8_MASK, X86ISD::VPERMI, 0),
   X86_INTRINSIC_DATA(avx512_mask_permvar_df_256, INTR_TYPE_2OP_MASK, X86ISD::VPERMV, 0),
   X86_INTRINSIC_DATA(avx512_mask_permvar_df_512, INTR_TYPE_2OP_MASK, X86ISD::VPERMV, 0),
   X86_INTRINSIC_DATA(avx512_mask_permvar_di_256, INTR_TYPE_2OP_MASK, X86ISD::VPERMV, 0),
diff --git a/llvm/test/CodeGen/X86/avx512-intrinsics.ll b/llvm/test/CodeGen/X86/avx512-intrinsics.ll
index 657dd10482b..76855995976 100644
--- a/llvm/test/CodeGen/X86/avx512-intrinsics.ll
+++ b/llvm/test/CodeGen/X86/avx512-intrinsics.ll
@@ -7042,9 +7042,9 @@ define <8 x i64>@test_int_x86_avx512_mask_pmovsxw_q_512(<8 x i16> %x0, <8 x i64>
   ret <8 x i64> %res4
 }
 
-declare <8 x double> @llvm.x86.avx512.mask.perm.df.512(<8 x double>, i8, <8 x double>, i8)
+declare <8 x double> @llvm.x86.avx512.mask.perm.df.512(<8 x double>, i32, <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) {
+define <8 x double>@test_int_x86_avx512_mask_perm_df_512(<8 x double> %x0, i32 %x1, <8 x double> %x2, i8 %x3) {
 ; CHECK-LABEL: test_int_x86_avx512_mask_perm_df_512:
 ; CHECK:       ## BB#0:
 ; CHECK-NEXT:    kmovw %esi, %k1
@@ -7054,17 +7054,17 @@ define <8 x double>@test_int_x86_avx512_mask_perm_df_512(<8 x double> %x0, i8 %x
 ; 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)
+  %res = call <8 x double> @llvm.x86.avx512.mask.perm.df.512(<8 x double> %x0, i32 3, <8 x double> %x2, i8 %x3)
+  %res1 = call <8 x double> @llvm.x86.avx512.mask.perm.df.512(<8 x double> %x0, i32 3, <8 x double> zeroinitializer, i8 %x3)
+  %res2 = call <8 x double> @llvm.x86.avx512.mask.perm.df.512(<8 x double> %x0, i32 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)
+declare <8 x i64> @llvm.x86.avx512.mask.perm.di.512(<8 x i64>, i32, <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) {
+define <8 x i64>@test_int_x86_avx512_mask_perm_di_512(<8 x i64> %x0, i32 %x1, <8 x i64> %x2, i8 %x3) {
 ; CHECK-LABEL: test_int_x86_avx512_mask_perm_di_512:
 ; CHECK:       ## BB#0:
 ; CHECK-NEXT:    kmovw %esi, %k1
@@ -7074,9 +7074,9 @@ define <8 x i64>@test_int_x86_avx512_mask_perm_di_512(<8 x i64> %x0, i8 %x1, <8
 ; 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)
+  %res = call <8 x i64> @llvm.x86.avx512.mask.perm.di.512(<8 x i64> %x0, i32 3, <8 x i64> %x2, i8 %x3)
+  %res1 = call <8 x i64> @llvm.x86.avx512.mask.perm.di.512(<8 x i64> %x0, i32 3, <8 x i64> zeroinitializer, i8 %x3)
+  %res2 = call <8 x i64> @llvm.x86.avx512.mask.perm.di.512(<8 x i64> %x0, i32 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
diff --git a/llvm/test/CodeGen/X86/avx512vl-intrinsics.ll b/llvm/test/CodeGen/X86/avx512vl-intrinsics.ll
index 3a1630fdbaa..63b01dcf75d 100644
--- a/llvm/test/CodeGen/X86/avx512vl-intrinsics.ll
+++ b/llvm/test/CodeGen/X86/avx512vl-intrinsics.ll
@@ -7442,9 +7442,9 @@ define <4 x i64>@test_int_x86_avx512_mask_pmovsxw_q_256(<8 x i16> %x0, <4 x i64>
   ret <4 x i64> %res4
 }
 
-declare <4 x double> @llvm.x86.avx512.mask.perm.df.256(<4 x double>, i8, <4 x double>, i8)
+declare <4 x double> @llvm.x86.avx512.mask.perm.df.256(<4 x double>, i32, <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) {
+define <4 x double>@test_int_x86_avx512_mask_perm_df_256(<4 x double> %x0, i32 %x1, <4 x double> %x2, i8 %x3) {
 ; CHECK-LABEL: test_int_x86_avx512_mask_perm_df_256:
 ; CHECK:       ## BB#0:
 ; CHECK-NEXT:    kmovw %esi, %k1
@@ -7455,17 +7455,17 @@ define <4 x double>@test_int_x86_avx512_mask_perm_df_256(<4 x double> %x0, i8 %x
 ; 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)
+  %res = call <4 x double> @llvm.x86.avx512.mask.perm.df.256(<4 x double> %x0, i32 3, <4 x double> %x2, i8 %x3)
+  %res1 = call <4 x double> @llvm.x86.avx512.mask.perm.df.256(<4 x double> %x0, i32 3, <4 x double> zeroinitializer, i8 %x3)
+  %res2 = call <4 x double> @llvm.x86.avx512.mask.perm.df.256(<4 x double> %x0, i32 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)
+declare <4 x i64> @llvm.x86.avx512.mask.perm.di.256(<4 x i64>, i32, <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) {
+define <4 x i64>@test_int_x86_avx512_mask_perm_di_256(<4 x i64> %x0, i32 %x1, <4 x i64> %x2, i8 %x3) {
 ; CHECK-LABEL: test_int_x86_avx512_mask_perm_di_256:
 ; CHECK:       ## BB#0:
 ; CHECK-NEXT:    kmovw %esi, %k1
@@ -7476,9 +7476,9 @@ define <4 x i64>@test_int_x86_avx512_mask_perm_di_256(<4 x i64> %x0, i8 %x1, <4
 ; 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)
+  %res = call <4 x i64> @llvm.x86.avx512.mask.perm.di.256(<4 x i64> %x0, i32 3, <4 x i64> %x2, i8 %x3)
+  %res1 = call <4 x i64> @llvm.x86.avx512.mask.perm.di.256(<4 x i64> %x0, i32 3, <4 x i64> zeroinitializer, i8 %x3)
+  %res2 = call <4 x i64> @llvm.x86.avx512.mask.perm.di.256(<4 x i64> %x0, i32 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