[llvm] 8255360 - [CostModel][X86] Add basic GFNI target test coverage for shift/rotate costs

Wed Apr 17 10:35:35 PDT 2024

Author: Simon Pilgrim
Date: 2024-04-17T18:35:17+01:00
New Revision: 825536039d667eeb933c590fe40c358fdea03a8d

URL: https://github.com/llvm/llvm-project/commit/825536039d667eeb933c590fe40c358fdea03a8d
DIFF: https://github.com/llvm/llvm-project/commit/825536039d667eeb933c590fe40c358fdea03a8d.diff

LOG: [CostModel][X86] Add basic GFNI target test coverage for shift/rotate costs

Added: 
    

Modified: 
    llvm/test/Analysis/CostModel/X86/fshl-codesize.ll
    llvm/test/Analysis/CostModel/X86/fshl-latency.ll
    llvm/test/Analysis/CostModel/X86/fshl-sizelatency.ll
    llvm/test/Analysis/CostModel/X86/fshl.ll
    llvm/test/Analysis/CostModel/X86/fshr-codesize.ll
    llvm/test/Analysis/CostModel/X86/fshr-latency.ll
    llvm/test/Analysis/CostModel/X86/fshr-sizelatency.ll
    llvm/test/Analysis/CostModel/X86/fshr.ll
    llvm/test/Analysis/CostModel/X86/vshift-ashr-codesize.ll
    llvm/test/Analysis/CostModel/X86/vshift-ashr-cost-inseltpoison.ll
    llvm/test/Analysis/CostModel/X86/vshift-ashr-cost.ll
    llvm/test/Analysis/CostModel/X86/vshift-ashr-latency.ll
    llvm/test/Analysis/CostModel/X86/vshift-ashr-sizelatency.ll
    llvm/test/Analysis/CostModel/X86/vshift-lshr-codesize.ll
    llvm/test/Analysis/CostModel/X86/vshift-lshr-cost-inseltpoison.ll
    llvm/test/Analysis/CostModel/X86/vshift-lshr-cost.ll
    llvm/test/Analysis/CostModel/X86/vshift-lshr-latency.ll
    llvm/test/Analysis/CostModel/X86/vshift-lshr-sizelatency.ll
    llvm/test/Analysis/CostModel/X86/vshift-shl-codesize.ll
    llvm/test/Analysis/CostModel/X86/vshift-shl-cost-inseltpoison.ll
    llvm/test/Analysis/CostModel/X86/vshift-shl-cost.ll
    llvm/test/Analysis/CostModel/X86/vshift-shl-latency.ll
    llvm/test/Analysis/CostModel/X86/vshift-shl-sizelatency.ll

Removed: 
    


################################################################################
diff  --git a/llvm/test/Analysis/CostModel/X86/fshl-codesize.ll b/llvm/test/Analysis/CostModel/X86/fshl-codesize.ll
index c212ecfc3eb904..c46e32ffb4ad3a 100644

--- a/llvm/test/Analysis/CostModel/X86/fshl-codesize.ll
+++ b/llvm/test/Analysis/CostModel/X86/fshl-codesize.ll
@@ -12,6 +12,7 @@
 ; RUN: opt < %s -passes="print<cost-model>" 2>&1 -disable-output -cost-kind=code-size -mtriple=x86_64-apple-macosx10.8.0 -mcpu=goldmont | FileCheck %s --check-prefixes=GLM
 ; RUN: opt < %s -passes="print<cost-model>" 2>&1 -disable-output -cost-kind=code-size -mtriple=x86_64-apple-macosx10.8.0 -mcpu=bdver2 | FileCheck %s --check-prefixes=XOP
 ; RUN: opt < %s -passes="print<cost-model>" 2>&1 -disable-output -cost-kind=code-size -mtriple=x86_64-apple-macosx10.8.0 -mcpu=btver2 | FileCheck %s --check-prefixes=AVX1
+; RUN: opt < %s -passes="print<cost-model>" 2>&1 -disable-output -cost-kind=code-size -mtriple=x86_64-apple-macosx10.8.0 -mcpu=tigerlake | FileCheck %s --check-prefixes=AVX512,AVX512GFNI
 
 target datalayout = "e-p:64:64:64-i1:8:8-i8:8:8-i16:16:16-i32:32:32-i64:64:64-f32:32:32-f64:64:64-v64:64:64-v128:128:128-a0:0:64-s0:64:64-f80:128:128-n8:16:32:64-S128"
 target triple = "x86_64-apple-macosx10.8.0"
@@ -97,6 +98,13 @@ define void @var_funnel_i64(i64 %a64, <2 x i64> %a128, <4 x i64> %a256, <8 x i64
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 23 for instruction: %V4I64 = call <4 x i64> @llvm.fshl.v4i64(<4 x i64> %a256, <4 x i64> %b256, <4 x i64> %c256)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 46 for instruction: %V8I64 = call <8 x i64> @llvm.fshl.v8i64(<8 x i64> %a512, <8 x i64> %b512, <8 x i64> %c512)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret void
+;
+; AVX512GFNI-LABEL: 'var_funnel_i64'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %I64 = call i64 @llvm.fshl.i64(i64 %a64, i64 %b64, i64 %c64)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V2I64 = call <2 x i64> @llvm.fshl.v2i64(<2 x i64> %a128, <2 x i64> %b128, <2 x i64> %c128)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V4I64 = call <4 x i64> @llvm.fshl.v4i64(<4 x i64> %a256, <4 x i64> %b256, <4 x i64> %c256)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V8I64 = call <8 x i64> @llvm.fshl.v8i64(<8 x i64> %a512, <8 x i64> %b512, <8 x i64> %c512)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret void
 ;
   %I64    = call i64 @llvm.fshl.i64(i64 %a64, i64 %b64, i64 %c64)
   %V2I64  = call <2 x i64> @llvm.fshl.v2i64(<2 x i64> %a128, <2 x i64> %b128, <2 x i64> %c128)
@@ -182,6 +190,13 @@ define void @var_funnel_i32(i32 %a32, <4 x i32> %a128, <8 x i32> %a256, <16 x i3
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 23 for instruction: %V4I32 = call <8 x i32> @llvm.fshl.v8i32(<8 x i32> %a256, <8 x i32> %b256, <8 x i32> %c256)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 46 for instruction: %V8I32 = call <16 x i32> @llvm.fshl.v16i32(<16 x i32> %a512, <16 x i32> %b512, <16 x i32> %c512)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret void
+;
+; AVX512GFNI-LABEL: 'var_funnel_i32'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %I32 = call i32 @llvm.fshl.i32(i32 %a32, i32 %b32, i32 %c32)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V2I32 = call <4 x i32> @llvm.fshl.v4i32(<4 x i32> %a128, <4 x i32> %b128, <4 x i32> %c128)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V4I32 = call <8 x i32> @llvm.fshl.v8i32(<8 x i32> %a256, <8 x i32> %b256, <8 x i32> %c256)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V8I32 = call <16 x i32> @llvm.fshl.v16i32(<16 x i32> %a512, <16 x i32> %b512, <16 x i32> %c512)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret void
 ;
   %I32   = call i32 @llvm.fshl.i32(i32 %a32, i32 %b32, i32 %c32)
   %V2I32 = call <4 x i32> @llvm.fshl.v4i32(<4 x i32> %a128, <4 x i32> %b128, <4 x i32> %c128)
@@ -267,6 +282,13 @@ define void @var_funnel_i16(i16 %a16, <8 x i16> %a128, <16 x i16> %a256, <32 x i
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 25 for instruction: %V16I16 = call <16 x i16> @llvm.fshl.v16i16(<16 x i16> %a256, <16 x i16> %b256, <16 x i16> %c256)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 50 for instruction: %V32I16 = call <32 x i16> @llvm.fshl.v32i16(<32 x i16> %a512, <32 x i16> %b512, <32 x i16> %c512)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret void
+;
+; AVX512GFNI-LABEL: 'var_funnel_i16'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %I16 = call i16 @llvm.fshl.i16(i16 %a16, i16 %b16, i16 %c16)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V8I16 = call <8 x i16> @llvm.fshl.v8i16(<8 x i16> %a128, <8 x i16> %b128, <8 x i16> %c128)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V16I16 = call <16 x i16> @llvm.fshl.v16i16(<16 x i16> %a256, <16 x i16> %b256, <16 x i16> %c256)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V32I16 = call <32 x i16> @llvm.fshl.v32i16(<32 x i16> %a512, <32 x i16> %b512, <32 x i16> %c512)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret void
 ;
   %I16    = call i16 @llvm.fshl.i16(i16 %a16, i16 %b16, i16 %c16)
   %V8I16  = call <8 x i16> @llvm.fshl.v8i16(<8 x i16> %a128, <8 x i16> %b128, <8 x i16> %c128)
@@ -352,6 +374,13 @@ define void @var_funnel_i8(i8 %a8, <16 x i8> %a128, <32 x i8> %a256, <64 x i8> %
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 25 for instruction: %V32I8 = call <32 x i8> @llvm.fshl.v32i8(<32 x i8> %a256, <32 x i8> %b256, <32 x i8> %c256)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 50 for instruction: %V64I8 = call <64 x i8> @llvm.fshl.v64i8(<64 x i8> %a512, <64 x i8> %b512, <64 x i8> %c512)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret void
+;
+; AVX512GFNI-LABEL: 'var_funnel_i8'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %I8 = call i8 @llvm.fshl.i8(i8 %a8, i8 %b8, i8 %c8)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 13 for instruction: %V16I8 = call <16 x i8> @llvm.fshl.v16i8(<16 x i8> %a128, <16 x i8> %b128, <16 x i8> %c128)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 28 for instruction: %V32I8 = call <32 x i8> @llvm.fshl.v32i8(<32 x i8> %a256, <32 x i8> %b256, <32 x i8> %c256)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 33 for instruction: %V64I8 = call <64 x i8> @llvm.fshl.v64i8(<64 x i8> %a512, <64 x i8> %b512, <64 x i8> %c512)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret void
 ;
   %I8    = call i8 @llvm.fshl.i8(i8 %a8, i8 %b8, i8 %c8)
   %V16I8 = call <16 x i8> @llvm.fshl.v16i8(<16 x i8> %a128, <16 x i8> %b128, <16 x i8> %c128)
@@ -463,6 +492,15 @@ define void @splatvar_funnel_i64(i64 %a64, <2 x i64> %a128, <4 x i64> %a256, <8
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 21 for instruction: %V4I64 = call <4 x i64> @llvm.fshl.v4i64(<4 x i64> %a256, <4 x i64> %b256, <4 x i64> %u256)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 42 for instruction: %V8I64 = call <8 x i64> @llvm.fshl.v8i64(<8 x i64> %a512, <8 x i64> %b512, <8 x i64> %u512)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret void
+;
+; AVX512GFNI-LABEL: 'splatvar_funnel_i64'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %u128 = shufflevector <2 x i64> %c128, <2 x i64> undef, <2 x i32> zeroinitializer
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %u256 = shufflevector <4 x i64> %c256, <4 x i64> undef, <4 x i32> zeroinitializer
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %u512 = shufflevector <8 x i64> %c512, <8 x i64> undef, <8 x i32> zeroinitializer
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V2I64 = call <2 x i64> @llvm.fshl.v2i64(<2 x i64> %a128, <2 x i64> %b128, <2 x i64> %u128)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V4I64 = call <4 x i64> @llvm.fshl.v4i64(<4 x i64> %a256, <4 x i64> %b256, <4 x i64> %u256)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V8I64 = call <8 x i64> @llvm.fshl.v8i64(<8 x i64> %a512, <8 x i64> %b512, <8 x i64> %u512)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret void
 ;
   %u128 = shufflevector <2 x i64> %c128, <2 x i64> undef, <2 x i32> zeroinitializer
   %u256 = shufflevector <4 x i64> %c256, <4 x i64> undef, <4 x i32> zeroinitializer
@@ -572,6 +610,15 @@ define void @splatvar_funnel_i32(i32 %a32, <4 x i32> %a128, <8 x i32> %a256, <16
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 23 for instruction: %V4I32 = call <8 x i32> @llvm.fshl.v8i32(<8 x i32> %a256, <8 x i32> %b256, <8 x i32> %u256)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 46 for instruction: %V8I32 = call <16 x i32> @llvm.fshl.v16i32(<16 x i32> %a512, <16 x i32> %b512, <16 x i32> %u512)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret void
+;
+; AVX512GFNI-LABEL: 'splatvar_funnel_i32'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %u128 = shufflevector <4 x i32> %c128, <4 x i32> undef, <4 x i32> zeroinitializer
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %u256 = shufflevector <8 x i32> %c256, <8 x i32> undef, <8 x i32> zeroinitializer
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %u512 = shufflevector <16 x i32> %c512, <16 x i32> undef, <16 x i32> zeroinitializer
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V2I32 = call <4 x i32> @llvm.fshl.v4i32(<4 x i32> %a128, <4 x i32> %b128, <4 x i32> %u128)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V4I32 = call <8 x i32> @llvm.fshl.v8i32(<8 x i32> %a256, <8 x i32> %b256, <8 x i32> %u256)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V8I32 = call <16 x i32> @llvm.fshl.v16i32(<16 x i32> %a512, <16 x i32> %b512, <16 x i32> %u512)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret void
 ;
   %u128 = shufflevector <4 x i32> %c128, <4 x i32> undef, <4 x i32> zeroinitializer
   %u256 = shufflevector <8 x i32> %c256, <8 x i32> undef, <8 x i32> zeroinitializer
@@ -681,6 +728,15 @@ define void @splatvar_funnel_i16(i16 %a16, <8 x i16> %a128, <16 x i16> %a256, <3
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 25 for instruction: %V16I16 = call <16 x i16> @llvm.fshl.v16i16(<16 x i16> %a256, <16 x i16> %b256, <16 x i16> %u256)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 50 for instruction: %V32I16 = call <32 x i16> @llvm.fshl.v32i16(<32 x i16> %a512, <32 x i16> %b512, <32 x i16> %u512)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret void
+;
+; AVX512GFNI-LABEL: 'splatvar_funnel_i16'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %u128 = shufflevector <8 x i16> %c128, <8 x i16> undef, <8 x i32> zeroinitializer
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %u256 = shufflevector <16 x i16> %c256, <16 x i16> undef, <16 x i32> zeroinitializer
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %u512 = shufflevector <32 x i16> %c512, <32 x i16> undef, <32 x i32> zeroinitializer
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V8I16 = call <8 x i16> @llvm.fshl.v8i16(<8 x i16> %a128, <8 x i16> %b128, <8 x i16> %u128)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V16I16 = call <16 x i16> @llvm.fshl.v16i16(<16 x i16> %a256, <16 x i16> %b256, <16 x i16> %u256)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V32I16 = call <32 x i16> @llvm.fshl.v32i16(<32 x i16> %a512, <32 x i16> %b512, <32 x i16> %u512)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret void
 ;
   %u128 = shufflevector <8 x i16> %c128, <8 x i16> undef, <8 x i32> zeroinitializer
   %u256 = shufflevector <16 x i16> %c256, <16 x i16> undef, <16 x i32> zeroinitializer
@@ -790,6 +846,15 @@ define void @splatvar_funnel_i8(i8 %a8, <16 x i8> %a128, <32 x i8> %a256, <64 x
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 25 for instruction: %V32I8 = call <32 x i8> @llvm.fshl.v32i8(<32 x i8> %a256, <32 x i8> %b256, <32 x i8> %u256)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 50 for instruction: %V64I8 = call <64 x i8> @llvm.fshl.v64i8(<64 x i8> %a512, <64 x i8> %b512, <64 x i8> %u512)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret void
+;
+; AVX512GFNI-LABEL: 'splatvar_funnel_i8'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %u128 = shufflevector <16 x i8> %c128, <16 x i8> undef, <16 x i32> zeroinitializer
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %u256 = shufflevector <32 x i8> %c256, <32 x i8> undef, <32 x i32> zeroinitializer
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %u512 = shufflevector <64 x i8> %c512, <64 x i8> undef, <64 x i32> zeroinitializer
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 15 for instruction: %V16I8 = call <16 x i8> @llvm.fshl.v16i8(<16 x i8> %a128, <16 x i8> %b128, <16 x i8> %u128)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 18 for instruction: %V32I8 = call <32 x i8> @llvm.fshl.v32i8(<32 x i8> %a256, <32 x i8> %b256, <32 x i8> %u256)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 18 for instruction: %V64I8 = call <64 x i8> @llvm.fshl.v64i8(<64 x i8> %a512, <64 x i8> %b512, <64 x i8> %u512)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret void
 ;
   %u128 = shufflevector <16 x i8> %c128, <16 x i8> undef, <16 x i32> zeroinitializer
   %u256 = shufflevector <32 x i8> %c256, <32 x i8> undef, <32 x i32> zeroinitializer
@@ -881,6 +946,13 @@ define void @constant_funnel_i64(i64 %a64, <2 x i64> %a128, <4 x i64> %a256, <8
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 22 for instruction: %V4I64 = call <4 x i64> @llvm.fshl.v4i64(<4 x i64> %a256, <4 x i64> %b256, <4 x i64> <i64 1, i64 7, i64 15, i64 31>)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 44 for instruction: %V8I64 = call <8 x i64> @llvm.fshl.v8i64(<8 x i64> %a512, <8 x i64> %b512, <8 x i64> <i64 1, i64 7, i64 15, i64 31, i64 1, i64 7, i64 15, i64 31>)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret void
+;
+; AVX512GFNI-LABEL: 'constant_funnel_i64'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %I64 = call i64 @llvm.fshl.i64(i64 %a64, i64 %b64, i64 7)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V2I64 = call <2 x i64> @llvm.fshl.v2i64(<2 x i64> %a128, <2 x i64> %b128, <2 x i64> <i64 1, i64 7>)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V4I64 = call <4 x i64> @llvm.fshl.v4i64(<4 x i64> %a256, <4 x i64> %b256, <4 x i64> <i64 1, i64 7, i64 15, i64 31>)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V8I64 = call <8 x i64> @llvm.fshl.v8i64(<8 x i64> %a512, <8 x i64> %b512, <8 x i64> <i64 1, i64 7, i64 15, i64 31, i64 1, i64 7, i64 15, i64 31>)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret void
 ;
   %I64    = call i64 @llvm.fshl.i64(i64 %a64, i64 %b64, i64 7)
   %V2I64  = call <2 x i64> @llvm.fshl.v2i64(<2 x i64> %a128, <2 x i64> %b128, <2 x i64> <i64 1, i64 7>)
@@ -966,6 +1038,13 @@ define void @constant_funnel_i32(i32 %a32, <4 x i32> %a128, <8 x i32> %a256, <16
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 22 for instruction: %V4I32 = call <8 x i32> @llvm.fshl.v8i32(<8 x i32> %a256, <8 x i32> %b256, <8 x i32> <i32 4, i32 5, i32 6, i32 7, i32 0, i32 1, i32 2, i32 3>)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 44 for instruction: %V8I32 = call <16 x i32> @llvm.fshl.v16i32(<16 x i32> %a512, <16 x i32> %b512, <16 x i32> <i32 4, i32 5, i32 6, i32 7, i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 0, i32 1, i32 2, i32 3>)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret void
+;
+; AVX512GFNI-LABEL: 'constant_funnel_i32'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %I32 = call i32 @llvm.fshl.i32(i32 %a32, i32 %b32, i32 7)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V2I32 = call <4 x i32> @llvm.fshl.v4i32(<4 x i32> %a128, <4 x i32> %b128, <4 x i32> <i32 4, i32 5, i32 6, i32 7>)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V4I32 = call <8 x i32> @llvm.fshl.v8i32(<8 x i32> %a256, <8 x i32> %b256, <8 x i32> <i32 4, i32 5, i32 6, i32 7, i32 0, i32 1, i32 2, i32 3>)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V8I32 = call <16 x i32> @llvm.fshl.v16i32(<16 x i32> %a512, <16 x i32> %b512, <16 x i32> <i32 4, i32 5, i32 6, i32 7, i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 0, i32 1, i32 2, i32 3>)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret void
 ;
   %I32   = call i32 @llvm.fshl.i32(i32 %a32, i32 %b32, i32 7)
   %V2I32 = call <4 x i32> @llvm.fshl.v4i32(<4 x i32> %a128, <4 x i32> %b128, <4 x i32> <i32 4, i32 5, i32 6, i32 7>)
@@ -1051,6 +1130,13 @@ define void @constant_funnel_i16(i16 %a16, <8 x i16> %a128, <16 x i16> %a256, <3
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 24 for instruction: %V16I16 = call <16 x i16> @llvm.fshl.v16i16(<16 x i16> %a256, <16 x i16> %b256, <16 x i16> <i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7>)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 48 for instruction: %V32I16 = call <32 x i16> @llvm.fshl.v32i16(<32 x i16> %a512, <32 x i16> %b512, <32 x i16> <i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7>)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret void
+;
+; AVX512GFNI-LABEL: 'constant_funnel_i16'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %I16 = call i16 @llvm.fshl.i16(i16 %a16, i16 %b16, i16 7)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V8I16 = call <8 x i16> @llvm.fshl.v8i16(<8 x i16> %a128, <8 x i16> %b128, <8 x i16> <i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7>)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V16I16 = call <16 x i16> @llvm.fshl.v16i16(<16 x i16> %a256, <16 x i16> %b256, <16 x i16> <i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7>)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V32I16 = call <32 x i16> @llvm.fshl.v32i16(<32 x i16> %a512, <32 x i16> %b512, <32 x i16> <i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7>)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret void
 ;
   %I16    = call i16 @llvm.fshl.i16(i16 %a16, i16 %b16, i16 7)
   %V8I16  = call <8 x i16> @llvm.fshl.v8i16(<8 x i16> %a128, <8 x i16> %b128, <8 x i16> <i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7>)
@@ -1136,6 +1222,13 @@ define void @constant_funnel_i8(i8 %a8, <16 x i8> %a128, <32 x i8> %a256, <64 x
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 24 for instruction: %V32I8 = call <32 x i8> @llvm.fshl.v32i8(<32 x i8> %a256, <32 x i8> %b256, <32 x i8> <i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0>)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 48 for instruction: %V64I8 = call <64 x i8> @llvm.fshl.v64i8(<64 x i8> %a512, <64 x i8> %b512, <64 x i8> <i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0>)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret void
+;
+; AVX512GFNI-LABEL: 'constant_funnel_i8'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %I8 = call i8 @llvm.fshl.i8(i8 %a8, i8 %b8, i8 7)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 12 for instruction: %V16I8 = call <16 x i8> @llvm.fshl.v16i8(<16 x i8> %a128, <16 x i8> %b128, <16 x i8> <i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0>)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 27 for instruction: %V32I8 = call <32 x i8> @llvm.fshl.v32i8(<32 x i8> %a256, <32 x i8> %b256, <32 x i8> <i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0>)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 32 for instruction: %V64I8 = call <64 x i8> @llvm.fshl.v64i8(<64 x i8> %a512, <64 x i8> %b512, <64 x i8> <i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0>)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret void
 ;
   %I8    = call i8 @llvm.fshl.i8(i8 %a8, i8 %b8, i8 7)
   %V16I8 = call <16 x i8> @llvm.fshl.v16i8(<16 x i8> %a128, <16 x i8> %b128, <16 x i8> <i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0>)
@@ -1225,6 +1318,13 @@ define void @splatconstant_funnel_i64(i64 %a64, <2 x i64> %a128, <4 x i64> %a256
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 20 for instruction: %V4I64 = call <4 x i64> @llvm.fshl.v4i64(<4 x i64> %a256, <4 x i64> %b256, <4 x i64> <i64 7, i64 7, i64 7, i64 7>)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 40 for instruction: %V8I64 = call <8 x i64> @llvm.fshl.v8i64(<8 x i64> %a512, <8 x i64> %b512, <8 x i64> <i64 7, i64 7, i64 7, i64 7, i64 7, i64 7, i64 7, i64 7>)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret void
+;
+; AVX512GFNI-LABEL: 'splatconstant_funnel_i64'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %I64 = call i64 @llvm.fshl.i64(i64 %a64, i64 %b64, i64 7)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V2I64 = call <2 x i64> @llvm.fshl.v2i64(<2 x i64> %a128, <2 x i64> %b128, <2 x i64> <i64 7, i64 7>)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V4I64 = call <4 x i64> @llvm.fshl.v4i64(<4 x i64> %a256, <4 x i64> %b256, <4 x i64> <i64 7, i64 7, i64 7, i64 7>)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V8I64 = call <8 x i64> @llvm.fshl.v8i64(<8 x i64> %a512, <8 x i64> %b512, <8 x i64> <i64 7, i64 7, i64 7, i64 7, i64 7, i64 7, i64 7, i64 7>)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret void
 ;
   %I64    = call i64 @llvm.fshl.i64(i64 %a64, i64 %b64, i64 7)
   %V2I64  = call <2 x i64> @llvm.fshl.v2i64(<2 x i64> %a128, <2 x i64> %b128, <2 x i64> <i64 7, i64 7>)
@@ -1310,6 +1410,13 @@ define void @splatconstant_funnel_i32(i32 %a32, <4 x i32> %a128, <8 x i32> %a256
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 20 for instruction: %V4I32 = call <8 x i32> @llvm.fshl.v8i32(<8 x i32> %a256, <8 x i32> %b256, <8 x i32> <i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5>)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 40 for instruction: %V8I32 = call <16 x i32> @llvm.fshl.v16i32(<16 x i32> %a512, <16 x i32> %b512, <16 x i32> <i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5>)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret void
+;
+; AVX512GFNI-LABEL: 'splatconstant_funnel_i32'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %I32 = call i32 @llvm.fshl.i32(i32 %a32, i32 %b32, i32 5)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V2I32 = call <4 x i32> @llvm.fshl.v4i32(<4 x i32> %a128, <4 x i32> %b128, <4 x i32> <i32 5, i32 5, i32 5, i32 5>)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V4I32 = call <8 x i32> @llvm.fshl.v8i32(<8 x i32> %a256, <8 x i32> %b256, <8 x i32> <i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5>)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V8I32 = call <16 x i32> @llvm.fshl.v16i32(<16 x i32> %a512, <16 x i32> %b512, <16 x i32> <i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5>)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret void
 ;
   %I32   = call i32 @llvm.fshl.i32(i32 %a32, i32 %b32, i32 5)
   %V2I32 = call <4 x i32> @llvm.fshl.v4i32(<4 x i32> %a128, <4 x i32> %b128, <4 x i32> <i32 5, i32 5, i32 5, i32 5>)
@@ -1395,6 +1502,13 @@ define void @splatconstant_funnel_i16(i16 %a16, <8 x i16> %a128, <16 x i16> %a25
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 22 for instruction: %V16I16 = call <16 x i16> @llvm.fshl.v16i16(<16 x i16> %a256, <16 x i16> %b256, <16 x i16> <i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3>)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 44 for instruction: %V32I16 = call <32 x i16> @llvm.fshl.v32i16(<32 x i16> %a512, <32 x i16> %b512, <32 x i16> <i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3>)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret void
+;
+; AVX512GFNI-LABEL: 'splatconstant_funnel_i16'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %I16 = call i16 @llvm.fshl.i16(i16 %a16, i16 %b16, i16 3)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V8I16 = call <8 x i16> @llvm.fshl.v8i16(<8 x i16> %a128, <8 x i16> %b128, <8 x i16> <i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3>)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V16I16 = call <16 x i16> @llvm.fshl.v16i16(<16 x i16> %a256, <16 x i16> %b256, <16 x i16> <i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3>)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V32I16 = call <32 x i16> @llvm.fshl.v32i16(<32 x i16> %a512, <32 x i16> %b512, <32 x i16> <i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3>)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret void
 ;
   %I16    = call i16 @llvm.fshl.i16(i16 %a16, i16 %b16, i16 3)
   %V8I16  = call <8 x i16> @llvm.fshl.v8i16(<8 x i16> %a128, <8 x i16> %b128, <8 x i16> <i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3>)
@@ -1480,6 +1594,13 @@ define void @splatconstant_funnel_i8(i8 %a8, <16 x i8> %a128, <32 x i8> %a256, <
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 24 for instruction: %V32I8 = call <32 x i8> @llvm.fshl.v32i8(<32 x i8> %a256, <32 x i8> %b256, <32 x i8> <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, 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>)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 48 for instruction: %V64I8 = call <64 x i8> @llvm.fshl.v64i8(<64 x i8> %a512, <64 x i8> %b512, <64 x i8> <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, 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, 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, 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>)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret void
+;
+; AVX512GFNI-LABEL: 'splatconstant_funnel_i8'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %I8 = call i8 @llvm.fshl.i8(i8 %a8, i8 %b8, i8 3)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %V16I8 = call <16 x i8> @llvm.fshl.v16i8(<16 x i8> %a128, <16 x i8> %b128, <16 x i8> <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>)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %V32I8 = call <32 x i8> @llvm.fshl.v32i8(<32 x i8> %a256, <32 x i8> %b256, <32 x i8> <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, 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>)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %V64I8 = call <64 x i8> @llvm.fshl.v64i8(<64 x i8> %a512, <64 x i8> %b512, <64 x i8> <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, 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, 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, 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>)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret void
 ;
   %I8    = call i8 @llvm.fshl.i8(i8 %a8, i8 %b8, i8 3)
   %V16I8 = call <16 x i8> @llvm.fshl.v16i8(<16 x i8> %a128, <16 x i8> %b128, <16 x i8> <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>)
@@ -1690,6 +1811,13 @@ define void @var_rotate_i16(i16 %a16, <8 x i16> %a128, <16 x i16> %a256, <32 x i
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 5 for instruction: %V16I16 = call <16 x i16> @llvm.fshl.v16i16(<16 x i16> %a256, <16 x i16> %a256, <16 x i16> %c256)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 10 for instruction: %V32I16 = call <32 x i16> @llvm.fshl.v32i16(<32 x i16> %a512, <32 x i16> %a512, <32 x i16> %c512)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret void
+;
+; AVX512GFNI-LABEL: 'var_rotate_i16'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %I16 = call i16 @llvm.fshl.i16(i16 %a16, i16 %a16, i16 %c16)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V8I16 = call <8 x i16> @llvm.fshl.v8i16(<8 x i16> %a128, <8 x i16> %a128, <8 x i16> %c128)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V16I16 = call <16 x i16> @llvm.fshl.v16i16(<16 x i16> %a256, <16 x i16> %a256, <16 x i16> %c256)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V32I16 = call <32 x i16> @llvm.fshl.v32i16(<32 x i16> %a512, <32 x i16> %a512, <32 x i16> %c512)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret void
 ;
   %I16    = call i16 @llvm.fshl.i16(i16 %a16, i16 %a16, i16 %c16)
   %V8I16  = call <8 x i16> @llvm.fshl.v8i16(<8 x i16> %a128, <8 x i16> %a128, <8 x i16> %c128)
@@ -1775,6 +1903,13 @@ define void @var_rotate_i8(i8 %a8, <16 x i8> %a128, <32 x i8> %a256, <64 x i8> %
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 5 for instruction: %V32I8 = call <32 x i8> @llvm.fshl.v32i8(<32 x i8> %a256, <32 x i8> %a256, <32 x i8> %c256)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 10 for instruction: %V64I8 = call <64 x i8> @llvm.fshl.v64i8(<64 x i8> %a512, <64 x i8> %a512, <64 x i8> %c512)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret void
+;
+; AVX512GFNI-LABEL: 'var_rotate_i8'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %I8 = call i8 @llvm.fshl.i8(i8 %a8, i8 %a8, i8 %c8)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 7 for instruction: %V16I8 = call <16 x i8> @llvm.fshl.v16i8(<16 x i8> %a128, <16 x i8> %a128, <16 x i8> %c128)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 7 for instruction: %V32I8 = call <32 x i8> @llvm.fshl.v32i8(<32 x i8> %a256, <32 x i8> %a256, <32 x i8> %c256)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 11 for instruction: %V64I8 = call <64 x i8> @llvm.fshl.v64i8(<64 x i8> %a512, <64 x i8> %a512, <64 x i8> %c512)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret void
 ;
   %I8    = call i8 @llvm.fshl.i8(i8 %a8, i8 %a8, i8 %c8)
   %V16I8 = call <16 x i8> @llvm.fshl.v16i8(<16 x i8> %a128, <16 x i8> %a128, <16 x i8> %c128)
@@ -2023,6 +2158,15 @@ define void @splatvar_rotate_i16(i16 %a16, <8 x i16> %a128, <16 x i16> %a256, <3
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 5 for instruction: %V16I16 = call <16 x i16> @llvm.fshl.v16i16(<16 x i16> %a256, <16 x i16> %a256, <16 x i16> %u256)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 10 for instruction: %V32I16 = call <32 x i16> @llvm.fshl.v32i16(<32 x i16> %a512, <32 x i16> %a512, <32 x i16> %u512)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret void
+;
+; AVX512GFNI-LABEL: 'splatvar_rotate_i16'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %u128 = shufflevector <8 x i16> %c128, <8 x i16> undef, <8 x i32> zeroinitializer
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %u256 = shufflevector <16 x i16> %c256, <16 x i16> undef, <16 x i32> zeroinitializer
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %u512 = shufflevector <32 x i16> %c512, <32 x i16> undef, <32 x i32> zeroinitializer
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V8I16 = call <8 x i16> @llvm.fshl.v8i16(<8 x i16> %a128, <8 x i16> %a128, <8 x i16> %u128)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V16I16 = call <16 x i16> @llvm.fshl.v16i16(<16 x i16> %a256, <16 x i16> %a256, <16 x i16> %u256)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V32I16 = call <32 x i16> @llvm.fshl.v32i16(<32 x i16> %a512, <32 x i16> %a512, <32 x i16> %u512)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret void
 ;
   %u128 = shufflevector <8 x i16> %c128, <8 x i16> undef, <8 x i32> zeroinitializer
   %u256 = shufflevector <16 x i16> %c256, <16 x i16> undef, <16 x i32> zeroinitializer
@@ -2123,6 +2267,15 @@ define void @splatvar_rotate_i8(i8 %a8, <16 x i8> %a128, <32 x i8> %a256, <64 x
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 5 for instruction: %V32I8 = call <32 x i8> @llvm.fshl.v32i8(<32 x i8> %a256, <32 x i8> %a256, <32 x i8> %u256)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 10 for instruction: %V64I8 = call <64 x i8> @llvm.fshl.v64i8(<64 x i8> %a512, <64 x i8> %a512, <64 x i8> %u512)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret void
+;
+; AVX512GFNI-LABEL: 'splatvar_rotate_i8'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %u128 = shufflevector <16 x i8> %c128, <16 x i8> undef, <16 x i32> zeroinitializer
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %u256 = shufflevector <32 x i8> %c256, <32 x i8> undef, <32 x i32> zeroinitializer
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %u512 = shufflevector <64 x i8> %c512, <64 x i8> undef, <64 x i32> zeroinitializer
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 7 for instruction: %V16I8 = call <16 x i8> @llvm.fshl.v16i8(<16 x i8> %a128, <16 x i8> %a128, <16 x i8> %u128)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 7 for instruction: %V32I8 = call <32 x i8> @llvm.fshl.v32i8(<32 x i8> %a256, <32 x i8> %a256, <32 x i8> %u256)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 11 for instruction: %V64I8 = call <64 x i8> @llvm.fshl.v64i8(<64 x i8> %a512, <64 x i8> %a512, <64 x i8> %u512)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret void
 ;
   %u128 = shufflevector <16 x i8> %c128, <16 x i8> undef, <16 x i32> zeroinitializer
   %u256 = shufflevector <32 x i8> %c256, <32 x i8> undef, <32 x i32> zeroinitializer
@@ -2335,6 +2488,13 @@ define void @constant_rotate_i16(i16 %a16, <8 x i16> %a128, <16 x i16> %a256, <3
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 5 for instruction: %V16I16 = call <16 x i16> @llvm.fshl.v16i16(<16 x i16> %a256, <16 x i16> %a256, <16 x i16> <i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7>)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 10 for instruction: %V32I16 = call <32 x i16> @llvm.fshl.v32i16(<32 x i16> %a512, <32 x i16> %a512, <32 x i16> <i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7>)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret void
+;
+; AVX512GFNI-LABEL: 'constant_rotate_i16'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %I16 = call i16 @llvm.fshl.i16(i16 %a16, i16 %a16, i16 7)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V8I16 = call <8 x i16> @llvm.fshl.v8i16(<8 x i16> %a128, <8 x i16> %a128, <8 x i16> <i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7>)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V16I16 = call <16 x i16> @llvm.fshl.v16i16(<16 x i16> %a256, <16 x i16> %a256, <16 x i16> <i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7>)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V32I16 = call <32 x i16> @llvm.fshl.v32i16(<32 x i16> %a512, <32 x i16> %a512, <32 x i16> <i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7>)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret void
 ;
   %I16    = call i16 @llvm.fshl.i16(i16 %a16, i16 %a16, i16 7)
   %V8I16  = call <8 x i16> @llvm.fshl.v8i16(<8 x i16> %a128, <8 x i16> %a128, <8 x i16> <i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7>)
@@ -2420,6 +2580,13 @@ define void @constant_rotate_i8(i8 %a8, <16 x i8> %a128, <32 x i8> %a256, <64 x
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 5 for instruction: %V32I8 = call <32 x i8> @llvm.fshl.v32i8(<32 x i8> %a256, <32 x i8> %a256, <32 x i8> <i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0>)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 10 for instruction: %V64I8 = call <64 x i8> @llvm.fshl.v64i8(<64 x i8> %a512, <64 x i8> %a512, <64 x i8> <i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0>)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret void
+;
+; AVX512GFNI-LABEL: 'constant_rotate_i8'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %I8 = call i8 @llvm.fshl.i8(i8 %a8, i8 %a8, i8 7)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 7 for instruction: %V16I8 = call <16 x i8> @llvm.fshl.v16i8(<16 x i8> %a128, <16 x i8> %a128, <16 x i8> <i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0>)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 7 for instruction: %V32I8 = call <32 x i8> @llvm.fshl.v32i8(<32 x i8> %a256, <32 x i8> %a256, <32 x i8> <i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0>)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 11 for instruction: %V64I8 = call <64 x i8> @llvm.fshl.v64i8(<64 x i8> %a512, <64 x i8> %a512, <64 x i8> <i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0>)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret void
 ;
   %I8    = call i8 @llvm.fshl.i8(i8 %a8, i8 %a8, i8 7)
   %V16I8 = call <16 x i8> @llvm.fshl.v16i8(<16 x i8> %a128, <16 x i8> %a128, <16 x i8> <i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0>)
@@ -2616,6 +2783,13 @@ define void @splatconstant_rotate_i16(i16 %a16, <8 x i16> %a128, <16 x i16> %a25
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 5 for instruction: %V16I16 = call <16 x i16> @llvm.fshl.v16i16(<16 x i16> %a256, <16 x i16> %a256, <16 x i16> <i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3>)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 10 for instruction: %V32I16 = call <32 x i16> @llvm.fshl.v32i16(<32 x i16> %a512, <32 x i16> %a512, <32 x i16> <i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3>)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret void
+;
+; AVX512GFNI-LABEL: 'splatconstant_rotate_i16'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %I16 = call i16 @llvm.fshl.i16(i16 %a16, i16 %a16, i16 3)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V8I16 = call <8 x i16> @llvm.fshl.v8i16(<8 x i16> %a128, <8 x i16> %a128, <8 x i16> <i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3>)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V16I16 = call <16 x i16> @llvm.fshl.v16i16(<16 x i16> %a256, <16 x i16> %a256, <16 x i16> <i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3>)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V32I16 = call <32 x i16> @llvm.fshl.v32i16(<32 x i16> %a512, <32 x i16> %a512, <32 x i16> <i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3>)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret void
 ;
   %I16    = call i16 @llvm.fshl.i16(i16 %a16, i16 %a16, i16 3)
   %V8I16  = call <8 x i16> @llvm.fshl.v8i16(<8 x i16> %a128, <8 x i16> %a128, <8 x i16> <i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3>)
@@ -2694,6 +2868,13 @@ define void @splatconstant_rotate_i8(i8 %a8, <16 x i8> %a128, <32 x i8> %a256, <
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 5 for instruction: %V32I8 = call <32 x i8> @llvm.fshl.v32i8(<32 x i8> %a256, <32 x i8> %a256, <32 x i8> <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, 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>)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 10 for instruction: %V64I8 = call <64 x i8> @llvm.fshl.v64i8(<64 x i8> %a512, <64 x i8> %a512, <64 x i8> <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, 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, 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, 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>)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret void
+;
+; AVX512GFNI-LABEL: 'splatconstant_rotate_i8'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %I8 = call i8 @llvm.fshl.i8(i8 %a8, i8 %a8, i8 3)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 7 for instruction: %V16I8 = call <16 x i8> @llvm.fshl.v16i8(<16 x i8> %a128, <16 x i8> %a128, <16 x i8> <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>)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 7 for instruction: %V32I8 = call <32 x i8> @llvm.fshl.v32i8(<32 x i8> %a256, <32 x i8> %a256, <32 x i8> <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, 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>)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 11 for instruction: %V64I8 = call <64 x i8> @llvm.fshl.v64i8(<64 x i8> %a512, <64 x i8> %a512, <64 x i8> <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, 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, 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, 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>)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret void
 ;
   %I8    = call i8 @llvm.fshl.i8(i8 %a8, i8 %a8, i8 3)
   %V16I8 = call <16 x i8> @llvm.fshl.v16i8(<16 x i8> %a128, <16 x i8> %a128, <16 x i8> <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>)

diff  --git a/llvm/test/Analysis/CostModel/X86/fshl-latency.ll b/llvm/test/Analysis/CostModel/X86/fshl-latency.ll
index 487adfe79d9442..fa32497c63ec7e 100644
--- a/llvm/test/Analysis/CostModel/X86/fshl-latency.ll
+++ b/llvm/test/Analysis/CostModel/X86/fshl-latency.ll
@@ -12,6 +12,7 @@
 ; RUN: opt < %s -passes="print<cost-model>" 2>&1 -disable-output -cost-kind=latency -mtriple=x86_64-apple-macosx10.8.0 -mcpu=goldmont | FileCheck %s --check-prefixes=GLM
 ; RUN: opt < %s -passes="print<cost-model>" 2>&1 -disable-output -cost-kind=latency -mtriple=x86_64-apple-macosx10.8.0 -mcpu=bdver2 | FileCheck %s --check-prefixes=XOP
 ; RUN: opt < %s -passes="print<cost-model>" 2>&1 -disable-output -cost-kind=latency -mtriple=x86_64-apple-macosx10.8.0 -mcpu=btver2 | FileCheck %s --check-prefixes=AVX1
+; RUN: opt < %s -passes="print<cost-model>" 2>&1 -disable-output -cost-kind=latency -mtriple=x86_64-apple-macosx10.8.0 -mcpu=tigerlake | FileCheck %s --check-prefixes=AVX512,AVX512GFNI
 
 target datalayout = "e-p:64:64:64-i1:8:8-i8:8:8-i16:16:16-i32:32:32-i64:64:64-f32:32:32-f64:64:64-v64:64:64-v128:128:128-a0:0:64-s0:64:64-f80:128:128-n8:16:32:64-S128"
 target triple = "x86_64-apple-macosx10.8.0"
@@ -97,6 +98,13 @@ define void @var_funnel_i64(i64 %a64, <2 x i64> %a128, <4 x i64> %a256, <8 x i64
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 23 for instruction: %V4I64 = call <4 x i64> @llvm.fshl.v4i64(<4 x i64> %a256, <4 x i64> %b256, <4 x i64> %c256)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 46 for instruction: %V8I64 = call <8 x i64> @llvm.fshl.v8i64(<8 x i64> %a512, <8 x i64> %b512, <8 x i64> %c512)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret void
+;
+; AVX512GFNI-LABEL: 'var_funnel_i64'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %I64 = call i64 @llvm.fshl.i64(i64 %a64, i64 %b64, i64 %c64)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V2I64 = call <2 x i64> @llvm.fshl.v2i64(<2 x i64> %a128, <2 x i64> %b128, <2 x i64> %c128)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V4I64 = call <4 x i64> @llvm.fshl.v4i64(<4 x i64> %a256, <4 x i64> %b256, <4 x i64> %c256)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V8I64 = call <8 x i64> @llvm.fshl.v8i64(<8 x i64> %a512, <8 x i64> %b512, <8 x i64> %c512)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret void
 ;
   %I64    = call i64 @llvm.fshl.i64(i64 %a64, i64 %b64, i64 %c64)
   %V2I64  = call <2 x i64> @llvm.fshl.v2i64(<2 x i64> %a128, <2 x i64> %b128, <2 x i64> %c128)
@@ -182,6 +190,13 @@ define void @var_funnel_i32(i32 %a32, <4 x i32> %a128, <8 x i32> %a256, <16 x i3
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 23 for instruction: %V4I32 = call <8 x i32> @llvm.fshl.v8i32(<8 x i32> %a256, <8 x i32> %b256, <8 x i32> %c256)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 46 for instruction: %V8I32 = call <16 x i32> @llvm.fshl.v16i32(<16 x i32> %a512, <16 x i32> %b512, <16 x i32> %c512)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret void
+;
+; AVX512GFNI-LABEL: 'var_funnel_i32'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %I32 = call i32 @llvm.fshl.i32(i32 %a32, i32 %b32, i32 %c32)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V2I32 = call <4 x i32> @llvm.fshl.v4i32(<4 x i32> %a128, <4 x i32> %b128, <4 x i32> %c128)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V4I32 = call <8 x i32> @llvm.fshl.v8i32(<8 x i32> %a256, <8 x i32> %b256, <8 x i32> %c256)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V8I32 = call <16 x i32> @llvm.fshl.v16i32(<16 x i32> %a512, <16 x i32> %b512, <16 x i32> %c512)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret void
 ;
   %I32   = call i32 @llvm.fshl.i32(i32 %a32, i32 %b32, i32 %c32)
   %V2I32 = call <4 x i32> @llvm.fshl.v4i32(<4 x i32> %a128, <4 x i32> %b128, <4 x i32> %c128)
@@ -267,6 +282,13 @@ define void @var_funnel_i16(i16 %a16, <8 x i16> %a128, <16 x i16> %a256, <32 x i
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 23 for instruction: %V16I16 = call <16 x i16> @llvm.fshl.v16i16(<16 x i16> %a256, <16 x i16> %b256, <16 x i16> %c256)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 46 for instruction: %V32I16 = call <32 x i16> @llvm.fshl.v32i16(<32 x i16> %a512, <32 x i16> %b512, <32 x i16> %c512)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret void
+;
+; AVX512GFNI-LABEL: 'var_funnel_i16'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %I16 = call i16 @llvm.fshl.i16(i16 %a16, i16 %b16, i16 %c16)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V8I16 = call <8 x i16> @llvm.fshl.v8i16(<8 x i16> %a128, <8 x i16> %b128, <8 x i16> %c128)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V16I16 = call <16 x i16> @llvm.fshl.v16i16(<16 x i16> %a256, <16 x i16> %b256, <16 x i16> %c256)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V32I16 = call <32 x i16> @llvm.fshl.v32i16(<32 x i16> %a512, <32 x i16> %b512, <32 x i16> %c512)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret void
 ;
   %I16    = call i16 @llvm.fshl.i16(i16 %a16, i16 %b16, i16 %c16)
   %V8I16  = call <8 x i16> @llvm.fshl.v8i16(<8 x i16> %a128, <8 x i16> %b128, <8 x i16> %c128)
@@ -352,6 +374,13 @@ define void @var_funnel_i8(i8 %a8, <16 x i8> %a128, <32 x i8> %a256, <64 x i8> %
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 23 for instruction: %V32I8 = call <32 x i8> @llvm.fshl.v32i8(<32 x i8> %a256, <32 x i8> %b256, <32 x i8> %c256)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 46 for instruction: %V64I8 = call <64 x i8> @llvm.fshl.v64i8(<64 x i8> %a512, <64 x i8> %b512, <64 x i8> %c512)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret void
+;
+; AVX512GFNI-LABEL: 'var_funnel_i8'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %I8 = call i8 @llvm.fshl.i8(i8 %a8, i8 %b8, i8 %c8)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 21 for instruction: %V16I8 = call <16 x i8> @llvm.fshl.v16i8(<16 x i8> %a128, <16 x i8> %b128, <16 x i8> %c128)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 58 for instruction: %V32I8 = call <32 x i8> @llvm.fshl.v32i8(<32 x i8> %a256, <32 x i8> %b256, <32 x i8> %c256)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 51 for instruction: %V64I8 = call <64 x i8> @llvm.fshl.v64i8(<64 x i8> %a512, <64 x i8> %b512, <64 x i8> %c512)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret void
 ;
   %I8    = call i8 @llvm.fshl.i8(i8 %a8, i8 %b8, i8 %c8)
   %V16I8 = call <16 x i8> @llvm.fshl.v16i8(<16 x i8> %a128, <16 x i8> %b128, <16 x i8> %c128)
@@ -463,6 +492,15 @@ define void @splatvar_funnel_i64(i64 %a64, <2 x i64> %a128, <4 x i64> %a256, <8
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 23 for instruction: %V4I64 = call <4 x i64> @llvm.fshl.v4i64(<4 x i64> %a256, <4 x i64> %b256, <4 x i64> %u256)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 46 for instruction: %V8I64 = call <8 x i64> @llvm.fshl.v8i64(<8 x i64> %a512, <8 x i64> %b512, <8 x i64> %u512)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret void
+;
+; AVX512GFNI-LABEL: 'splatvar_funnel_i64'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %u128 = shufflevector <2 x i64> %c128, <2 x i64> undef, <2 x i32> zeroinitializer
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %u256 = shufflevector <4 x i64> %c256, <4 x i64> undef, <4 x i32> zeroinitializer
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %u512 = shufflevector <8 x i64> %c512, <8 x i64> undef, <8 x i32> zeroinitializer
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V2I64 = call <2 x i64> @llvm.fshl.v2i64(<2 x i64> %a128, <2 x i64> %b128, <2 x i64> %u128)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V4I64 = call <4 x i64> @llvm.fshl.v4i64(<4 x i64> %a256, <4 x i64> %b256, <4 x i64> %u256)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V8I64 = call <8 x i64> @llvm.fshl.v8i64(<8 x i64> %a512, <8 x i64> %b512, <8 x i64> %u512)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret void
 ;
   %u128 = shufflevector <2 x i64> %c128, <2 x i64> undef, <2 x i32> zeroinitializer
   %u256 = shufflevector <4 x i64> %c256, <4 x i64> undef, <4 x i32> zeroinitializer
@@ -563,6 +601,15 @@ define void @splatvar_funnel_i32(i32 %a32, <4 x i32> %a128, <8 x i32> %a256, <16
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 23 for instruction: %V4I32 = call <8 x i32> @llvm.fshl.v8i32(<8 x i32> %a256, <8 x i32> %b256, <8 x i32> %u256)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 46 for instruction: %V8I32 = call <16 x i32> @llvm.fshl.v16i32(<16 x i32> %a512, <16 x i32> %b512, <16 x i32> %u512)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret void
+;
+; AVX512GFNI-LABEL: 'splatvar_funnel_i32'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %u128 = shufflevector <4 x i32> %c128, <4 x i32> undef, <4 x i32> zeroinitializer
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %u256 = shufflevector <8 x i32> %c256, <8 x i32> undef, <8 x i32> zeroinitializer
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %u512 = shufflevector <16 x i32> %c512, <16 x i32> undef, <16 x i32> zeroinitializer
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V2I32 = call <4 x i32> @llvm.fshl.v4i32(<4 x i32> %a128, <4 x i32> %b128, <4 x i32> %u128)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V4I32 = call <8 x i32> @llvm.fshl.v8i32(<8 x i32> %a256, <8 x i32> %b256, <8 x i32> %u256)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V8I32 = call <16 x i32> @llvm.fshl.v16i32(<16 x i32> %a512, <16 x i32> %b512, <16 x i32> %u512)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret void
 ;
   %u128 = shufflevector <4 x i32> %c128, <4 x i32> undef, <4 x i32> zeroinitializer
   %u256 = shufflevector <8 x i32> %c256, <8 x i32> undef, <8 x i32> zeroinitializer
@@ -663,6 +710,15 @@ define void @splatvar_funnel_i16(i16 %a16, <8 x i16> %a128, <16 x i16> %a256, <3
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 23 for instruction: %V16I16 = call <16 x i16> @llvm.fshl.v16i16(<16 x i16> %a256, <16 x i16> %b256, <16 x i16> %u256)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 46 for instruction: %V32I16 = call <32 x i16> @llvm.fshl.v32i16(<32 x i16> %a512, <32 x i16> %b512, <32 x i16> %u512)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret void
+;
+; AVX512GFNI-LABEL: 'splatvar_funnel_i16'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %u128 = shufflevector <8 x i16> %c128, <8 x i16> undef, <8 x i32> zeroinitializer
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %u256 = shufflevector <16 x i16> %c256, <16 x i16> undef, <16 x i32> zeroinitializer
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %u512 = shufflevector <32 x i16> %c512, <32 x i16> undef, <32 x i32> zeroinitializer
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V8I16 = call <8 x i16> @llvm.fshl.v8i16(<8 x i16> %a128, <8 x i16> %b128, <8 x i16> %u128)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V16I16 = call <16 x i16> @llvm.fshl.v16i16(<16 x i16> %a256, <16 x i16> %b256, <16 x i16> %u256)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V32I16 = call <32 x i16> @llvm.fshl.v32i16(<32 x i16> %a512, <32 x i16> %b512, <32 x i16> %u512)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret void
 ;
   %u128 = shufflevector <8 x i16> %c128, <8 x i16> undef, <8 x i32> zeroinitializer
   %u256 = shufflevector <16 x i16> %c256, <16 x i16> undef, <16 x i32> zeroinitializer
@@ -763,6 +819,15 @@ define void @splatvar_funnel_i8(i8 %a8, <16 x i8> %a128, <32 x i8> %a256, <64 x
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 23 for instruction: %V32I8 = call <32 x i8> @llvm.fshl.v32i8(<32 x i8> %a256, <32 x i8> %b256, <32 x i8> %u256)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 46 for instruction: %V64I8 = call <64 x i8> @llvm.fshl.v64i8(<64 x i8> %a512, <64 x i8> %b512, <64 x i8> %u512)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret void
+;
+; AVX512GFNI-LABEL: 'splatvar_funnel_i8'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %u128 = shufflevector <16 x i8> %c128, <16 x i8> undef, <16 x i32> zeroinitializer
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %u256 = shufflevector <32 x i8> %c256, <32 x i8> undef, <32 x i32> zeroinitializer
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %u512 = shufflevector <64 x i8> %c512, <64 x i8> undef, <64 x i32> zeroinitializer
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 20 for instruction: %V16I8 = call <16 x i8> @llvm.fshl.v16i8(<16 x i8> %a128, <16 x i8> %b128, <16 x i8> %u128)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 20 for instruction: %V32I8 = call <32 x i8> @llvm.fshl.v32i8(<32 x i8> %a256, <32 x i8> %b256, <32 x i8> %u256)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 20 for instruction: %V64I8 = call <64 x i8> @llvm.fshl.v64i8(<64 x i8> %a512, <64 x i8> %b512, <64 x i8> %u512)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret void
 ;
   %u128 = shufflevector <16 x i8> %c128, <16 x i8> undef, <16 x i32> zeroinitializer
   %u256 = shufflevector <32 x i8> %c256, <32 x i8> undef, <32 x i32> zeroinitializer
@@ -854,6 +919,13 @@ define void @constant_funnel_i64(i64 %a64, <2 x i64> %a128, <4 x i64> %a256, <8
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 22 for instruction: %V4I64 = call <4 x i64> @llvm.fshl.v4i64(<4 x i64> %a256, <4 x i64> %b256, <4 x i64> <i64 1, i64 7, i64 15, i64 31>)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 44 for instruction: %V8I64 = call <8 x i64> @llvm.fshl.v8i64(<8 x i64> %a512, <8 x i64> %b512, <8 x i64> <i64 1, i64 7, i64 15, i64 31, i64 1, i64 7, i64 15, i64 31>)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret void
+;
+; AVX512GFNI-LABEL: 'constant_funnel_i64'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %I64 = call i64 @llvm.fshl.i64(i64 %a64, i64 %b64, i64 7)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V2I64 = call <2 x i64> @llvm.fshl.v2i64(<2 x i64> %a128, <2 x i64> %b128, <2 x i64> <i64 1, i64 7>)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V4I64 = call <4 x i64> @llvm.fshl.v4i64(<4 x i64> %a256, <4 x i64> %b256, <4 x i64> <i64 1, i64 7, i64 15, i64 31>)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V8I64 = call <8 x i64> @llvm.fshl.v8i64(<8 x i64> %a512, <8 x i64> %b512, <8 x i64> <i64 1, i64 7, i64 15, i64 31, i64 1, i64 7, i64 15, i64 31>)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret void
 ;
   %I64    = call i64 @llvm.fshl.i64(i64 %a64, i64 %b64, i64 7)
   %V2I64  = call <2 x i64> @llvm.fshl.v2i64(<2 x i64> %a128, <2 x i64> %b128, <2 x i64> <i64 1, i64 7>)
@@ -939,6 +1011,13 @@ define void @constant_funnel_i32(i32 %a32, <4 x i32> %a128, <8 x i32> %a256, <16
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 22 for instruction: %V4I32 = call <8 x i32> @llvm.fshl.v8i32(<8 x i32> %a256, <8 x i32> %b256, <8 x i32> <i32 4, i32 5, i32 6, i32 7, i32 0, i32 1, i32 2, i32 3>)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 44 for instruction: %V8I32 = call <16 x i32> @llvm.fshl.v16i32(<16 x i32> %a512, <16 x i32> %b512, <16 x i32> <i32 4, i32 5, i32 6, i32 7, i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 0, i32 1, i32 2, i32 3>)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret void
+;
+; AVX512GFNI-LABEL: 'constant_funnel_i32'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %I32 = call i32 @llvm.fshl.i32(i32 %a32, i32 %b32, i32 7)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V2I32 = call <4 x i32> @llvm.fshl.v4i32(<4 x i32> %a128, <4 x i32> %b128, <4 x i32> <i32 4, i32 5, i32 6, i32 7>)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V4I32 = call <8 x i32> @llvm.fshl.v8i32(<8 x i32> %a256, <8 x i32> %b256, <8 x i32> <i32 4, i32 5, i32 6, i32 7, i32 0, i32 1, i32 2, i32 3>)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V8I32 = call <16 x i32> @llvm.fshl.v16i32(<16 x i32> %a512, <16 x i32> %b512, <16 x i32> <i32 4, i32 5, i32 6, i32 7, i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 0, i32 1, i32 2, i32 3>)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret void
 ;
   %I32   = call i32 @llvm.fshl.i32(i32 %a32, i32 %b32, i32 7)
   %V2I32 = call <4 x i32> @llvm.fshl.v4i32(<4 x i32> %a128, <4 x i32> %b128, <4 x i32> <i32 4, i32 5, i32 6, i32 7>)
@@ -1024,6 +1103,13 @@ define void @constant_funnel_i16(i16 %a16, <8 x i16> %a128, <16 x i16> %a256, <3
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 22 for instruction: %V16I16 = call <16 x i16> @llvm.fshl.v16i16(<16 x i16> %a256, <16 x i16> %b256, <16 x i16> <i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7>)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 44 for instruction: %V32I16 = call <32 x i16> @llvm.fshl.v32i16(<32 x i16> %a512, <32 x i16> %b512, <32 x i16> <i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7>)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret void
+;
+; AVX512GFNI-LABEL: 'constant_funnel_i16'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %I16 = call i16 @llvm.fshl.i16(i16 %a16, i16 %b16, i16 7)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V8I16 = call <8 x i16> @llvm.fshl.v8i16(<8 x i16> %a128, <8 x i16> %b128, <8 x i16> <i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7>)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V16I16 = call <16 x i16> @llvm.fshl.v16i16(<16 x i16> %a256, <16 x i16> %b256, <16 x i16> <i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7>)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V32I16 = call <32 x i16> @llvm.fshl.v32i16(<32 x i16> %a512, <32 x i16> %b512, <32 x i16> <i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7>)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret void
 ;
   %I16    = call i16 @llvm.fshl.i16(i16 %a16, i16 %b16, i16 7)
   %V8I16  = call <8 x i16> @llvm.fshl.v8i16(<8 x i16> %a128, <8 x i16> %b128, <8 x i16> <i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7>)
@@ -1109,6 +1195,13 @@ define void @constant_funnel_i8(i8 %a8, <16 x i8> %a128, <32 x i8> %a256, <64 x
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 22 for instruction: %V32I8 = call <32 x i8> @llvm.fshl.v32i8(<32 x i8> %a256, <32 x i8> %b256, <32 x i8> <i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0>)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 44 for instruction: %V64I8 = call <64 x i8> @llvm.fshl.v64i8(<64 x i8> %a512, <64 x i8> %b512, <64 x i8> <i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0>)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret void
+;
+; AVX512GFNI-LABEL: 'constant_funnel_i8'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %I8 = call i8 @llvm.fshl.i8(i8 %a8, i8 %b8, i8 7)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 20 for instruction: %V16I8 = call <16 x i8> @llvm.fshl.v16i8(<16 x i8> %a128, <16 x i8> %b128, <16 x i8> <i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0>)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 57 for instruction: %V32I8 = call <32 x i8> @llvm.fshl.v32i8(<32 x i8> %a256, <32 x i8> %b256, <32 x i8> <i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0>)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 50 for instruction: %V64I8 = call <64 x i8> @llvm.fshl.v64i8(<64 x i8> %a512, <64 x i8> %b512, <64 x i8> <i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0>)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret void
 ;
   %I8    = call i8 @llvm.fshl.i8(i8 %a8, i8 %b8, i8 7)
   %V16I8 = call <16 x i8> @llvm.fshl.v16i8(<16 x i8> %a128, <16 x i8> %b128, <16 x i8> <i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0>)
@@ -1198,6 +1291,13 @@ define void @splatconstant_funnel_i64(i64 %a64, <2 x i64> %a128, <4 x i64> %a256
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 20 for instruction: %V4I64 = call <4 x i64> @llvm.fshl.v4i64(<4 x i64> %a256, <4 x i64> %b256, <4 x i64> <i64 7, i64 7, i64 7, i64 7>)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 40 for instruction: %V8I64 = call <8 x i64> @llvm.fshl.v8i64(<8 x i64> %a512, <8 x i64> %b512, <8 x i64> <i64 7, i64 7, i64 7, i64 7, i64 7, i64 7, i64 7, i64 7>)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret void
+;
+; AVX512GFNI-LABEL: 'splatconstant_funnel_i64'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %I64 = call i64 @llvm.fshl.i64(i64 %a64, i64 %b64, i64 7)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V2I64 = call <2 x i64> @llvm.fshl.v2i64(<2 x i64> %a128, <2 x i64> %b128, <2 x i64> <i64 7, i64 7>)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V4I64 = call <4 x i64> @llvm.fshl.v4i64(<4 x i64> %a256, <4 x i64> %b256, <4 x i64> <i64 7, i64 7, i64 7, i64 7>)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V8I64 = call <8 x i64> @llvm.fshl.v8i64(<8 x i64> %a512, <8 x i64> %b512, <8 x i64> <i64 7, i64 7, i64 7, i64 7, i64 7, i64 7, i64 7, i64 7>)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret void
 ;
   %I64    = call i64 @llvm.fshl.i64(i64 %a64, i64 %b64, i64 7)
   %V2I64  = call <2 x i64> @llvm.fshl.v2i64(<2 x i64> %a128, <2 x i64> %b128, <2 x i64> <i64 7, i64 7>)
@@ -1276,6 +1376,13 @@ define void @splatconstant_funnel_i32(i32 %a32, <4 x i32> %a128, <8 x i32> %a256
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 20 for instruction: %V4I32 = call <8 x i32> @llvm.fshl.v8i32(<8 x i32> %a256, <8 x i32> %b256, <8 x i32> <i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5>)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 40 for instruction: %V8I32 = call <16 x i32> @llvm.fshl.v16i32(<16 x i32> %a512, <16 x i32> %b512, <16 x i32> <i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5>)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret void
+;
+; AVX512GFNI-LABEL: 'splatconstant_funnel_i32'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %I32 = call i32 @llvm.fshl.i32(i32 %a32, i32 %b32, i32 5)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V2I32 = call <4 x i32> @llvm.fshl.v4i32(<4 x i32> %a128, <4 x i32> %b128, <4 x i32> <i32 5, i32 5, i32 5, i32 5>)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V4I32 = call <8 x i32> @llvm.fshl.v8i32(<8 x i32> %a256, <8 x i32> %b256, <8 x i32> <i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5>)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V8I32 = call <16 x i32> @llvm.fshl.v16i32(<16 x i32> %a512, <16 x i32> %b512, <16 x i32> <i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5>)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret void
 ;
   %I32   = call i32 @llvm.fshl.i32(i32 %a32, i32 %b32, i32 5)
   %V2I32 = call <4 x i32> @llvm.fshl.v4i32(<4 x i32> %a128, <4 x i32> %b128, <4 x i32> <i32 5, i32 5, i32 5, i32 5>)
@@ -1354,6 +1461,13 @@ define void @splatconstant_funnel_i16(i16 %a16, <8 x i16> %a128, <16 x i16> %a25
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 20 for instruction: %V16I16 = call <16 x i16> @llvm.fshl.v16i16(<16 x i16> %a256, <16 x i16> %b256, <16 x i16> <i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3>)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 40 for instruction: %V32I16 = call <32 x i16> @llvm.fshl.v32i16(<32 x i16> %a512, <32 x i16> %b512, <32 x i16> <i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3>)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret void
+;
+; AVX512GFNI-LABEL: 'splatconstant_funnel_i16'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %I16 = call i16 @llvm.fshl.i16(i16 %a16, i16 %b16, i16 3)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V8I16 = call <8 x i16> @llvm.fshl.v8i16(<8 x i16> %a128, <8 x i16> %b128, <8 x i16> <i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3>)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V16I16 = call <16 x i16> @llvm.fshl.v16i16(<16 x i16> %a256, <16 x i16> %b256, <16 x i16> <i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3>)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V32I16 = call <32 x i16> @llvm.fshl.v32i16(<32 x i16> %a512, <32 x i16> %b512, <32 x i16> <i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3>)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret void
 ;
   %I16    = call i16 @llvm.fshl.i16(i16 %a16, i16 %b16, i16 3)
   %V8I16  = call <8 x i16> @llvm.fshl.v8i16(<8 x i16> %a128, <8 x i16> %b128, <8 x i16> <i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3>)
@@ -1432,6 +1546,13 @@ define void @splatconstant_funnel_i8(i8 %a8, <16 x i8> %a128, <32 x i8> %a256, <
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 22 for instruction: %V32I8 = call <32 x i8> @llvm.fshl.v32i8(<32 x i8> %a256, <32 x i8> %b256, <32 x i8> <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, 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>)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 44 for instruction: %V64I8 = call <64 x i8> @llvm.fshl.v64i8(<64 x i8> %a512, <64 x i8> %b512, <64 x i8> <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, 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, 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, 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>)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret void
+;
+; AVX512GFNI-LABEL: 'splatconstant_funnel_i8'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %I8 = call i8 @llvm.fshl.i8(i8 %a8, i8 %b8, i8 3)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 18 for instruction: %V16I8 = call <16 x i8> @llvm.fshl.v16i8(<16 x i8> %a128, <16 x i8> %b128, <16 x i8> <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>)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 20 for instruction: %V32I8 = call <32 x i8> @llvm.fshl.v32i8(<32 x i8> %a256, <32 x i8> %b256, <32 x i8> <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, 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>)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 20 for instruction: %V64I8 = call <64 x i8> @llvm.fshl.v64i8(<64 x i8> %a512, <64 x i8> %b512, <64 x i8> <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, 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, 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, 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>)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret void
 ;
   %I8    = call i8 @llvm.fshl.i8(i8 %a8, i8 %b8, i8 3)
   %V16I8 = call <16 x i8> @llvm.fshl.v16i8(<16 x i8> %a128, <16 x i8> %b128, <16 x i8> <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>)
@@ -1642,6 +1763,13 @@ define void @var_rotate_i16(i16 %a16, <8 x i16> %a128, <16 x i16> %a256, <32 x i
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 7 for instruction: %V16I16 = call <16 x i16> @llvm.fshl.v16i16(<16 x i16> %a256, <16 x i16> %a256, <16 x i16> %c256)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 14 for instruction: %V32I16 = call <32 x i16> @llvm.fshl.v32i16(<32 x i16> %a512, <32 x i16> %a512, <32 x i16> %c512)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret void
+;
+; AVX512GFNI-LABEL: 'var_rotate_i16'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %I16 = call i16 @llvm.fshl.i16(i16 %a16, i16 %a16, i16 %c16)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V8I16 = call <8 x i16> @llvm.fshl.v8i16(<8 x i16> %a128, <8 x i16> %a128, <8 x i16> %c128)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V16I16 = call <16 x i16> @llvm.fshl.v16i16(<16 x i16> %a256, <16 x i16> %a256, <16 x i16> %c256)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V32I16 = call <32 x i16> @llvm.fshl.v32i16(<32 x i16> %a512, <32 x i16> %a512, <32 x i16> %c512)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret void
 ;
   %I16    = call i16 @llvm.fshl.i16(i16 %a16, i16 %a16, i16 %c16)
   %V8I16  = call <8 x i16> @llvm.fshl.v8i16(<8 x i16> %a128, <8 x i16> %a128, <8 x i16> %c128)
@@ -1727,6 +1855,13 @@ define void @var_rotate_i8(i8 %a8, <16 x i8> %a128, <32 x i8> %a256, <64 x i8> %
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 7 for instruction: %V32I8 = call <32 x i8> @llvm.fshl.v32i8(<32 x i8> %a256, <32 x i8> %a256, <32 x i8> %c256)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 14 for instruction: %V64I8 = call <64 x i8> @llvm.fshl.v64i8(<64 x i8> %a512, <64 x i8> %a512, <64 x i8> %c512)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret void
+;
+; AVX512GFNI-LABEL: 'var_rotate_i8'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %I8 = call i8 @llvm.fshl.i8(i8 %a8, i8 %a8, i8 %c8)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 15 for instruction: %V16I8 = call <16 x i8> @llvm.fshl.v16i8(<16 x i8> %a128, <16 x i8> %a128, <16 x i8> %c128)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 15 for instruction: %V32I8 = call <32 x i8> @llvm.fshl.v32i8(<32 x i8> %a256, <32 x i8> %a256, <32 x i8> %c256)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 6 for instruction: %V64I8 = call <64 x i8> @llvm.fshl.v64i8(<64 x i8> %a512, <64 x i8> %a512, <64 x i8> %c512)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret void
 ;
   %I8    = call i8 @llvm.fshl.i8(i8 %a8, i8 %a8, i8 %c8)
   %V16I8 = call <16 x i8> @llvm.fshl.v16i8(<16 x i8> %a128, <16 x i8> %a128, <16 x i8> %c128)
@@ -1975,6 +2110,15 @@ define void @splatvar_rotate_i16(i16 %a16, <8 x i16> %a128, <16 x i16> %a256, <3
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 7 for instruction: %V16I16 = call <16 x i16> @llvm.fshl.v16i16(<16 x i16> %a256, <16 x i16> %a256, <16 x i16> %u256)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 14 for instruction: %V32I16 = call <32 x i16> @llvm.fshl.v32i16(<32 x i16> %a512, <32 x i16> %a512, <32 x i16> %u512)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret void
+;
+; AVX512GFNI-LABEL: 'splatvar_rotate_i16'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %u128 = shufflevector <8 x i16> %c128, <8 x i16> undef, <8 x i32> zeroinitializer
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %u256 = shufflevector <16 x i16> %c256, <16 x i16> undef, <16 x i32> zeroinitializer
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %u512 = shufflevector <32 x i16> %c512, <32 x i16> undef, <32 x i32> zeroinitializer
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V8I16 = call <8 x i16> @llvm.fshl.v8i16(<8 x i16> %a128, <8 x i16> %a128, <8 x i16> %u128)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V16I16 = call <16 x i16> @llvm.fshl.v16i16(<16 x i16> %a256, <16 x i16> %a256, <16 x i16> %u256)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V32I16 = call <32 x i16> @llvm.fshl.v32i16(<32 x i16> %a512, <32 x i16> %a512, <32 x i16> %u512)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret void
 ;
   %u128 = shufflevector <8 x i16> %c128, <8 x i16> undef, <8 x i32> zeroinitializer
   %u256 = shufflevector <16 x i16> %c256, <16 x i16> undef, <16 x i32> zeroinitializer
@@ -2075,6 +2219,15 @@ define void @splatvar_rotate_i8(i8 %a8, <16 x i8> %a128, <32 x i8> %a256, <64 x
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 7 for instruction: %V32I8 = call <32 x i8> @llvm.fshl.v32i8(<32 x i8> %a256, <32 x i8> %a256, <32 x i8> %u256)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 14 for instruction: %V64I8 = call <64 x i8> @llvm.fshl.v64i8(<64 x i8> %a512, <64 x i8> %a512, <64 x i8> %u512)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret void
+;
+; AVX512GFNI-LABEL: 'splatvar_rotate_i8'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %u128 = shufflevector <16 x i8> %c128, <16 x i8> undef, <16 x i32> zeroinitializer
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %u256 = shufflevector <32 x i8> %c256, <32 x i8> undef, <32 x i32> zeroinitializer
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %u512 = shufflevector <64 x i8> %c512, <64 x i8> undef, <64 x i32> zeroinitializer
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 15 for instruction: %V16I8 = call <16 x i8> @llvm.fshl.v16i8(<16 x i8> %a128, <16 x i8> %a128, <16 x i8> %u128)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 15 for instruction: %V32I8 = call <32 x i8> @llvm.fshl.v32i8(<32 x i8> %a256, <32 x i8> %a256, <32 x i8> %u256)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 6 for instruction: %V64I8 = call <64 x i8> @llvm.fshl.v64i8(<64 x i8> %a512, <64 x i8> %a512, <64 x i8> %u512)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret void
 ;
   %u128 = shufflevector <16 x i8> %c128, <16 x i8> undef, <16 x i32> zeroinitializer
   %u256 = shufflevector <32 x i8> %c256, <32 x i8> undef, <32 x i32> zeroinitializer
@@ -2287,6 +2440,13 @@ define void @constant_rotate_i16(i16 %a16, <8 x i16> %a128, <16 x i16> %a256, <3
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 7 for instruction: %V16I16 = call <16 x i16> @llvm.fshl.v16i16(<16 x i16> %a256, <16 x i16> %a256, <16 x i16> <i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7>)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 14 for instruction: %V32I16 = call <32 x i16> @llvm.fshl.v32i16(<32 x i16> %a512, <32 x i16> %a512, <32 x i16> <i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7>)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret void
+;
+; AVX512GFNI-LABEL: 'constant_rotate_i16'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %I16 = call i16 @llvm.fshl.i16(i16 %a16, i16 %a16, i16 7)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V8I16 = call <8 x i16> @llvm.fshl.v8i16(<8 x i16> %a128, <8 x i16> %a128, <8 x i16> <i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7>)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V16I16 = call <16 x i16> @llvm.fshl.v16i16(<16 x i16> %a256, <16 x i16> %a256, <16 x i16> <i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7>)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V32I16 = call <32 x i16> @llvm.fshl.v32i16(<32 x i16> %a512, <32 x i16> %a512, <32 x i16> <i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7>)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret void
 ;
   %I16    = call i16 @llvm.fshl.i16(i16 %a16, i16 %a16, i16 7)
   %V8I16  = call <8 x i16> @llvm.fshl.v8i16(<8 x i16> %a128, <8 x i16> %a128, <8 x i16> <i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7>)
@@ -2372,6 +2532,13 @@ define void @constant_rotate_i8(i8 %a8, <16 x i8> %a128, <32 x i8> %a256, <64 x
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 7 for instruction: %V32I8 = call <32 x i8> @llvm.fshl.v32i8(<32 x i8> %a256, <32 x i8> %a256, <32 x i8> <i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0>)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 14 for instruction: %V64I8 = call <64 x i8> @llvm.fshl.v64i8(<64 x i8> %a512, <64 x i8> %a512, <64 x i8> <i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0>)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret void
+;
+; AVX512GFNI-LABEL: 'constant_rotate_i8'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %I8 = call i8 @llvm.fshl.i8(i8 %a8, i8 %a8, i8 7)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 15 for instruction: %V16I8 = call <16 x i8> @llvm.fshl.v16i8(<16 x i8> %a128, <16 x i8> %a128, <16 x i8> <i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0>)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 15 for instruction: %V32I8 = call <32 x i8> @llvm.fshl.v32i8(<32 x i8> %a256, <32 x i8> %a256, <32 x i8> <i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0>)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 6 for instruction: %V64I8 = call <64 x i8> @llvm.fshl.v64i8(<64 x i8> %a512, <64 x i8> %a512, <64 x i8> <i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0>)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret void
 ;
   %I8    = call i8 @llvm.fshl.i8(i8 %a8, i8 %a8, i8 7)
   %V16I8 = call <16 x i8> @llvm.fshl.v16i8(<16 x i8> %a128, <16 x i8> %a128, <16 x i8> <i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0>)
@@ -2568,6 +2735,13 @@ define void @splatconstant_rotate_i16(i16 %a16, <8 x i16> %a128, <16 x i16> %a25
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 7 for instruction: %V16I16 = call <16 x i16> @llvm.fshl.v16i16(<16 x i16> %a256, <16 x i16> %a256, <16 x i16> <i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3>)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 14 for instruction: %V32I16 = call <32 x i16> @llvm.fshl.v32i16(<32 x i16> %a512, <32 x i16> %a512, <32 x i16> <i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3>)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret void
+;
+; AVX512GFNI-LABEL: 'splatconstant_rotate_i16'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %I16 = call i16 @llvm.fshl.i16(i16 %a16, i16 %a16, i16 3)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V8I16 = call <8 x i16> @llvm.fshl.v8i16(<8 x i16> %a128, <8 x i16> %a128, <8 x i16> <i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3>)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V16I16 = call <16 x i16> @llvm.fshl.v16i16(<16 x i16> %a256, <16 x i16> %a256, <16 x i16> <i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3>)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V32I16 = call <32 x i16> @llvm.fshl.v32i16(<32 x i16> %a512, <32 x i16> %a512, <32 x i16> <i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3>)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret void
 ;
   %I16    = call i16 @llvm.fshl.i16(i16 %a16, i16 %a16, i16 3)
   %V8I16  = call <8 x i16> @llvm.fshl.v8i16(<8 x i16> %a128, <8 x i16> %a128, <8 x i16> <i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3>)
@@ -2646,6 +2820,13 @@ define void @splatconstant_rotate_i8(i8 %a8, <16 x i8> %a128, <32 x i8> %a256, <
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 7 for instruction: %V32I8 = call <32 x i8> @llvm.fshl.v32i8(<32 x i8> %a256, <32 x i8> %a256, <32 x i8> <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, 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>)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 14 for instruction: %V64I8 = call <64 x i8> @llvm.fshl.v64i8(<64 x i8> %a512, <64 x i8> %a512, <64 x i8> <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, 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, 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, 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>)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret void
+;
+; AVX512GFNI-LABEL: 'splatconstant_rotate_i8'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %I8 = call i8 @llvm.fshl.i8(i8 %a8, i8 %a8, i8 3)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 15 for instruction: %V16I8 = call <16 x i8> @llvm.fshl.v16i8(<16 x i8> %a128, <16 x i8> %a128, <16 x i8> <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>)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 15 for instruction: %V32I8 = call <32 x i8> @llvm.fshl.v32i8(<32 x i8> %a256, <32 x i8> %a256, <32 x i8> <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, 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>)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 6 for instruction: %V64I8 = call <64 x i8> @llvm.fshl.v64i8(<64 x i8> %a512, <64 x i8> %a512, <64 x i8> <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, 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, 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, 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>)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret void
 ;
   %I8    = call i8 @llvm.fshl.i8(i8 %a8, i8 %a8, i8 3)
   %V16I8 = call <16 x i8> @llvm.fshl.v16i8(<16 x i8> %a128, <16 x i8> %a128, <16 x i8> <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>)

diff  --git a/llvm/test/Analysis/CostModel/X86/fshl-sizelatency.ll b/llvm/test/Analysis/CostModel/X86/fshl-sizelatency.ll
index 52ad7e13c84c69..832a574a9b332a 100644
--- a/llvm/test/Analysis/CostModel/X86/fshl-sizelatency.ll
+++ b/llvm/test/Analysis/CostModel/X86/fshl-sizelatency.ll
@@ -12,6 +12,7 @@
 ; RUN: opt < %s -passes="print<cost-model>" 2>&1 -disable-output -cost-kind=size-latency -mtriple=x86_64-apple-macosx10.8.0 -mcpu=goldmont | FileCheck %s --check-prefixes=GLM
 ; RUN: opt < %s -passes="print<cost-model>" 2>&1 -disable-output -cost-kind=size-latency -mtriple=x86_64-apple-macosx10.8.0 -mcpu=bdver2 | FileCheck %s --check-prefixes=XOP
 ; RUN: opt < %s -passes="print<cost-model>" 2>&1 -disable-output -cost-kind=size-latency -mtriple=x86_64-apple-macosx10.8.0 -mcpu=btver2 | FileCheck %s --check-prefixes=AVX1
+; RUN: opt < %s -passes="print<cost-model>" 2>&1 -disable-output -cost-kind=size-latency -mtriple=x86_64-apple-macosx10.8.0 -mcpu=tigerlake | FileCheck %s --check-prefixes=AVX512GFNI
 
 target datalayout = "e-p:64:64:64-i1:8:8-i8:8:8-i16:16:16-i32:32:32-i64:64:64-f32:32:32-f64:64:64-v64:64:64-v128:128:128-a0:0:64-s0:64:64-f80:128:128-n8:16:32:64-S128"
 target triple = "x86_64-apple-macosx10.8.0"
@@ -97,6 +98,13 @@ define void @var_funnel_i64(i64 %a64, <2 x i64> %a128, <4 x i64> %a256, <8 x i64
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 30 for instruction: %V4I64 = call <4 x i64> @llvm.fshl.v4i64(<4 x i64> %a256, <4 x i64> %b256, <4 x i64> %c256)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 60 for instruction: %V8I64 = call <8 x i64> @llvm.fshl.v8i64(<8 x i64> %a512, <8 x i64> %b512, <8 x i64> %c512)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret void
+;
+; AVX512GFNI-LABEL: 'var_funnel_i64'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %I64 = call i64 @llvm.fshl.i64(i64 %a64, i64 %b64, i64 %c64)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V2I64 = call <2 x i64> @llvm.fshl.v2i64(<2 x i64> %a128, <2 x i64> %b128, <2 x i64> %c128)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V4I64 = call <4 x i64> @llvm.fshl.v4i64(<4 x i64> %a256, <4 x i64> %b256, <4 x i64> %c256)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V8I64 = call <8 x i64> @llvm.fshl.v8i64(<8 x i64> %a512, <8 x i64> %b512, <8 x i64> %c512)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret void
 ;
   %I64    = call i64 @llvm.fshl.i64(i64 %a64, i64 %b64, i64 %c64)
   %V2I64  = call <2 x i64> @llvm.fshl.v2i64(<2 x i64> %a128, <2 x i64> %b128, <2 x i64> %c128)
@@ -182,6 +190,13 @@ define void @var_funnel_i32(i32 %a32, <4 x i32> %a128, <8 x i32> %a256, <16 x i3
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 30 for instruction: %V4I32 = call <8 x i32> @llvm.fshl.v8i32(<8 x i32> %a256, <8 x i32> %b256, <8 x i32> %c256)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 60 for instruction: %V8I32 = call <16 x i32> @llvm.fshl.v16i32(<16 x i32> %a512, <16 x i32> %b512, <16 x i32> %c512)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret void
+;
+; AVX512GFNI-LABEL: 'var_funnel_i32'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %I32 = call i32 @llvm.fshl.i32(i32 %a32, i32 %b32, i32 %c32)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V2I32 = call <4 x i32> @llvm.fshl.v4i32(<4 x i32> %a128, <4 x i32> %b128, <4 x i32> %c128)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V4I32 = call <8 x i32> @llvm.fshl.v8i32(<8 x i32> %a256, <8 x i32> %b256, <8 x i32> %c256)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V8I32 = call <16 x i32> @llvm.fshl.v16i32(<16 x i32> %a512, <16 x i32> %b512, <16 x i32> %c512)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret void
 ;
   %I32   = call i32 @llvm.fshl.i32(i32 %a32, i32 %b32, i32 %c32)
   %V2I32 = call <4 x i32> @llvm.fshl.v4i32(<4 x i32> %a128, <4 x i32> %b128, <4 x i32> %c128)
@@ -267,6 +282,13 @@ define void @var_funnel_i16(i16 %a16, <8 x i16> %a128, <16 x i16> %a256, <32 x i
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 31 for instruction: %V16I16 = call <16 x i16> @llvm.fshl.v16i16(<16 x i16> %a256, <16 x i16> %b256, <16 x i16> %c256)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 62 for instruction: %V32I16 = call <32 x i16> @llvm.fshl.v32i16(<32 x i16> %a512, <32 x i16> %b512, <32 x i16> %c512)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret void
+;
+; AVX512GFNI-LABEL: 'var_funnel_i16'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 5 for instruction: %I16 = call i16 @llvm.fshl.i16(i16 %a16, i16 %b16, i16 %c16)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V8I16 = call <8 x i16> @llvm.fshl.v8i16(<8 x i16> %a128, <8 x i16> %b128, <8 x i16> %c128)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V16I16 = call <16 x i16> @llvm.fshl.v16i16(<16 x i16> %a256, <16 x i16> %b256, <16 x i16> %c256)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V32I16 = call <32 x i16> @llvm.fshl.v32i16(<32 x i16> %a512, <32 x i16> %b512, <32 x i16> %c512)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret void
 ;
   %I16    = call i16 @llvm.fshl.i16(i16 %a16, i16 %b16, i16 %c16)
   %V8I16  = call <8 x i16> @llvm.fshl.v8i16(<8 x i16> %a128, <8 x i16> %b128, <8 x i16> %c128)
@@ -352,6 +374,13 @@ define void @var_funnel_i8(i8 %a8, <16 x i8> %a128, <32 x i8> %a256, <64 x i8> %
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 31 for instruction: %V32I8 = call <32 x i8> @llvm.fshl.v32i8(<32 x i8> %a256, <32 x i8> %b256, <32 x i8> %c256)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 62 for instruction: %V64I8 = call <64 x i8> @llvm.fshl.v64i8(<64 x i8> %a512, <64 x i8> %b512, <64 x i8> %c512)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret void
+;
+; AVX512GFNI-LABEL: 'var_funnel_i8'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 5 for instruction: %I8 = call i8 @llvm.fshl.i8(i8 %a8, i8 %b8, i8 %c8)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 15 for instruction: %V16I8 = call <16 x i8> @llvm.fshl.v16i8(<16 x i8> %a128, <16 x i8> %b128, <16 x i8> %c128)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 39 for instruction: %V32I8 = call <32 x i8> @llvm.fshl.v32i8(<32 x i8> %a256, <32 x i8> %b256, <32 x i8> %c256)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 38 for instruction: %V64I8 = call <64 x i8> @llvm.fshl.v64i8(<64 x i8> %a512, <64 x i8> %b512, <64 x i8> %c512)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret void
 ;
   %I8    = call i8 @llvm.fshl.i8(i8 %a8, i8 %b8, i8 %c8)
   %V16I8 = call <16 x i8> @llvm.fshl.v16i8(<16 x i8> %a128, <16 x i8> %b128, <16 x i8> %c128)
@@ -463,6 +492,15 @@ define void @splatvar_funnel_i64(i64 %a64, <2 x i64> %a128, <4 x i64> %a256, <8
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 30 for instruction: %V4I64 = call <4 x i64> @llvm.fshl.v4i64(<4 x i64> %a256, <4 x i64> %b256, <4 x i64> %u256)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 60 for instruction: %V8I64 = call <8 x i64> @llvm.fshl.v8i64(<8 x i64> %a512, <8 x i64> %b512, <8 x i64> %u512)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret void
+;
+; AVX512GFNI-LABEL: 'splatvar_funnel_i64'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %u128 = shufflevector <2 x i64> %c128, <2 x i64> undef, <2 x i32> zeroinitializer
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %u256 = shufflevector <4 x i64> %c256, <4 x i64> undef, <4 x i32> zeroinitializer
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %u512 = shufflevector <8 x i64> %c512, <8 x i64> undef, <8 x i32> zeroinitializer
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V2I64 = call <2 x i64> @llvm.fshl.v2i64(<2 x i64> %a128, <2 x i64> %b128, <2 x i64> %u128)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V4I64 = call <4 x i64> @llvm.fshl.v4i64(<4 x i64> %a256, <4 x i64> %b256, <4 x i64> %u256)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V8I64 = call <8 x i64> @llvm.fshl.v8i64(<8 x i64> %a512, <8 x i64> %b512, <8 x i64> %u512)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret void
 ;
   %u128 = shufflevector <2 x i64> %c128, <2 x i64> undef, <2 x i32> zeroinitializer
   %u256 = shufflevector <4 x i64> %c256, <4 x i64> undef, <4 x i32> zeroinitializer
@@ -572,6 +610,15 @@ define void @splatvar_funnel_i32(i32 %a32, <4 x i32> %a128, <8 x i32> %a256, <16
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 32 for instruction: %V4I32 = call <8 x i32> @llvm.fshl.v8i32(<8 x i32> %a256, <8 x i32> %b256, <8 x i32> %u256)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 64 for instruction: %V8I32 = call <16 x i32> @llvm.fshl.v16i32(<16 x i32> %a512, <16 x i32> %b512, <16 x i32> %u512)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret void
+;
+; AVX512GFNI-LABEL: 'splatvar_funnel_i32'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %u128 = shufflevector <4 x i32> %c128, <4 x i32> undef, <4 x i32> zeroinitializer
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %u256 = shufflevector <8 x i32> %c256, <8 x i32> undef, <8 x i32> zeroinitializer
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %u512 = shufflevector <16 x i32> %c512, <16 x i32> undef, <16 x i32> zeroinitializer
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V2I32 = call <4 x i32> @llvm.fshl.v4i32(<4 x i32> %a128, <4 x i32> %b128, <4 x i32> %u128)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V4I32 = call <8 x i32> @llvm.fshl.v8i32(<8 x i32> %a256, <8 x i32> %b256, <8 x i32> %u256)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V8I32 = call <16 x i32> @llvm.fshl.v16i32(<16 x i32> %a512, <16 x i32> %b512, <16 x i32> %u512)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret void
 ;
   %u128 = shufflevector <4 x i32> %c128, <4 x i32> undef, <4 x i32> zeroinitializer
   %u256 = shufflevector <8 x i32> %c256, <8 x i32> undef, <8 x i32> zeroinitializer
@@ -681,6 +728,15 @@ define void @splatvar_funnel_i16(i16 %a16, <8 x i16> %a128, <16 x i16> %a256, <3
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 33 for instruction: %V16I16 = call <16 x i16> @llvm.fshl.v16i16(<16 x i16> %a256, <16 x i16> %b256, <16 x i16> %u256)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 66 for instruction: %V32I16 = call <32 x i16> @llvm.fshl.v32i16(<32 x i16> %a512, <32 x i16> %b512, <32 x i16> %u512)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret void
+;
+; AVX512GFNI-LABEL: 'splatvar_funnel_i16'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %u128 = shufflevector <8 x i16> %c128, <8 x i16> undef, <8 x i32> zeroinitializer
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %u256 = shufflevector <16 x i16> %c256, <16 x i16> undef, <16 x i32> zeroinitializer
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %u512 = shufflevector <32 x i16> %c512, <32 x i16> undef, <32 x i32> zeroinitializer
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V8I16 = call <8 x i16> @llvm.fshl.v8i16(<8 x i16> %a128, <8 x i16> %b128, <8 x i16> %u128)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V16I16 = call <16 x i16> @llvm.fshl.v16i16(<16 x i16> %a256, <16 x i16> %b256, <16 x i16> %u256)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V32I16 = call <32 x i16> @llvm.fshl.v32i16(<32 x i16> %a512, <32 x i16> %b512, <32 x i16> %u512)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret void
 ;
   %u128 = shufflevector <8 x i16> %c128, <8 x i16> undef, <8 x i32> zeroinitializer
   %u256 = shufflevector <16 x i16> %c256, <16 x i16> undef, <16 x i32> zeroinitializer
@@ -790,6 +846,15 @@ define void @splatvar_funnel_i8(i8 %a8, <16 x i8> %a128, <32 x i8> %a256, <64 x
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 31 for instruction: %V32I8 = call <32 x i8> @llvm.fshl.v32i8(<32 x i8> %a256, <32 x i8> %b256, <32 x i8> %u256)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 62 for instruction: %V64I8 = call <64 x i8> @llvm.fshl.v64i8(<64 x i8> %a512, <64 x i8> %b512, <64 x i8> %u512)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret void
+;
+; AVX512GFNI-LABEL: 'splatvar_funnel_i8'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %u128 = shufflevector <16 x i8> %c128, <16 x i8> undef, <16 x i32> zeroinitializer
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %u256 = shufflevector <32 x i8> %c256, <32 x i8> undef, <32 x i32> zeroinitializer
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %u512 = shufflevector <64 x i8> %c512, <64 x i8> undef, <64 x i32> zeroinitializer
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 20 for instruction: %V16I8 = call <16 x i8> @llvm.fshl.v16i8(<16 x i8> %a128, <16 x i8> %b128, <16 x i8> %u128)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 22 for instruction: %V32I8 = call <32 x i8> @llvm.fshl.v32i8(<32 x i8> %a256, <32 x i8> %b256, <32 x i8> %u256)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 23 for instruction: %V64I8 = call <64 x i8> @llvm.fshl.v64i8(<64 x i8> %a512, <64 x i8> %b512, <64 x i8> %u512)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret void
 ;
   %u128 = shufflevector <16 x i8> %c128, <16 x i8> undef, <16 x i32> zeroinitializer
   %u256 = shufflevector <32 x i8> %c256, <32 x i8> undef, <32 x i32> zeroinitializer
@@ -881,6 +946,13 @@ define void @constant_funnel_i64(i64 %a64, <2 x i64> %a128, <4 x i64> %a256, <8
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 28 for instruction: %V4I64 = call <4 x i64> @llvm.fshl.v4i64(<4 x i64> %a256, <4 x i64> %b256, <4 x i64> <i64 1, i64 7, i64 15, i64 31>)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 56 for instruction: %V8I64 = call <8 x i64> @llvm.fshl.v8i64(<8 x i64> %a512, <8 x i64> %b512, <8 x i64> <i64 1, i64 7, i64 15, i64 31, i64 1, i64 7, i64 15, i64 31>)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret void
+;
+; AVX512GFNI-LABEL: 'constant_funnel_i64'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %I64 = call i64 @llvm.fshl.i64(i64 %a64, i64 %b64, i64 7)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V2I64 = call <2 x i64> @llvm.fshl.v2i64(<2 x i64> %a128, <2 x i64> %b128, <2 x i64> <i64 1, i64 7>)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V4I64 = call <4 x i64> @llvm.fshl.v4i64(<4 x i64> %a256, <4 x i64> %b256, <4 x i64> <i64 1, i64 7, i64 15, i64 31>)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V8I64 = call <8 x i64> @llvm.fshl.v8i64(<8 x i64> %a512, <8 x i64> %b512, <8 x i64> <i64 1, i64 7, i64 15, i64 31, i64 1, i64 7, i64 15, i64 31>)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret void
 ;
   %I64    = call i64 @llvm.fshl.i64(i64 %a64, i64 %b64, i64 7)
   %V2I64  = call <2 x i64> @llvm.fshl.v2i64(<2 x i64> %a128, <2 x i64> %b128, <2 x i64> <i64 1, i64 7>)
@@ -966,6 +1038,13 @@ define void @constant_funnel_i32(i32 %a32, <4 x i32> %a128, <8 x i32> %a256, <16
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 28 for instruction: %V4I32 = call <8 x i32> @llvm.fshl.v8i32(<8 x i32> %a256, <8 x i32> %b256, <8 x i32> <i32 4, i32 5, i32 6, i32 7, i32 0, i32 1, i32 2, i32 3>)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 56 for instruction: %V8I32 = call <16 x i32> @llvm.fshl.v16i32(<16 x i32> %a512, <16 x i32> %b512, <16 x i32> <i32 4, i32 5, i32 6, i32 7, i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 0, i32 1, i32 2, i32 3>)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret void
+;
+; AVX512GFNI-LABEL: 'constant_funnel_i32'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %I32 = call i32 @llvm.fshl.i32(i32 %a32, i32 %b32, i32 7)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V2I32 = call <4 x i32> @llvm.fshl.v4i32(<4 x i32> %a128, <4 x i32> %b128, <4 x i32> <i32 4, i32 5, i32 6, i32 7>)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V4I32 = call <8 x i32> @llvm.fshl.v8i32(<8 x i32> %a256, <8 x i32> %b256, <8 x i32> <i32 4, i32 5, i32 6, i32 7, i32 0, i32 1, i32 2, i32 3>)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V8I32 = call <16 x i32> @llvm.fshl.v16i32(<16 x i32> %a512, <16 x i32> %b512, <16 x i32> <i32 4, i32 5, i32 6, i32 7, i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 0, i32 1, i32 2, i32 3>)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret void
 ;
   %I32   = call i32 @llvm.fshl.i32(i32 %a32, i32 %b32, i32 7)
   %V2I32 = call <4 x i32> @llvm.fshl.v4i32(<4 x i32> %a128, <4 x i32> %b128, <4 x i32> <i32 4, i32 5, i32 6, i32 7>)
@@ -1051,6 +1130,13 @@ define void @constant_funnel_i16(i16 %a16, <8 x i16> %a128, <16 x i16> %a256, <3
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 29 for instruction: %V16I16 = call <16 x i16> @llvm.fshl.v16i16(<16 x i16> %a256, <16 x i16> %b256, <16 x i16> <i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7>)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 58 for instruction: %V32I16 = call <32 x i16> @llvm.fshl.v32i16(<32 x i16> %a512, <32 x i16> %b512, <32 x i16> <i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7>)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret void
+;
+; AVX512GFNI-LABEL: 'constant_funnel_i16'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 5 for instruction: %I16 = call i16 @llvm.fshl.i16(i16 %a16, i16 %b16, i16 7)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V8I16 = call <8 x i16> @llvm.fshl.v8i16(<8 x i16> %a128, <8 x i16> %b128, <8 x i16> <i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7>)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V16I16 = call <16 x i16> @llvm.fshl.v16i16(<16 x i16> %a256, <16 x i16> %b256, <16 x i16> <i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7>)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V32I16 = call <32 x i16> @llvm.fshl.v32i16(<32 x i16> %a512, <32 x i16> %b512, <32 x i16> <i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7>)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret void
 ;
   %I16    = call i16 @llvm.fshl.i16(i16 %a16, i16 %b16, i16 7)
   %V8I16  = call <8 x i16> @llvm.fshl.v8i16(<8 x i16> %a128, <8 x i16> %b128, <8 x i16> <i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7>)
@@ -1136,6 +1222,13 @@ define void @constant_funnel_i8(i8 %a8, <16 x i8> %a128, <32 x i8> %a256, <64 x
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 29 for instruction: %V32I8 = call <32 x i8> @llvm.fshl.v32i8(<32 x i8> %a256, <32 x i8> %b256, <32 x i8> <i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0>)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 58 for instruction: %V64I8 = call <64 x i8> @llvm.fshl.v64i8(<64 x i8> %a512, <64 x i8> %b512, <64 x i8> <i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0>)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret void
+;
+; AVX512GFNI-LABEL: 'constant_funnel_i8'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 5 for instruction: %I8 = call i8 @llvm.fshl.i8(i8 %a8, i8 %b8, i8 7)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 14 for instruction: %V16I8 = call <16 x i8> @llvm.fshl.v16i8(<16 x i8> %a128, <16 x i8> %b128, <16 x i8> <i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0>)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 38 for instruction: %V32I8 = call <32 x i8> @llvm.fshl.v32i8(<32 x i8> %a256, <32 x i8> %b256, <32 x i8> <i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0>)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 37 for instruction: %V64I8 = call <64 x i8> @llvm.fshl.v64i8(<64 x i8> %a512, <64 x i8> %b512, <64 x i8> <i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0>)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret void
 ;
   %I8    = call i8 @llvm.fshl.i8(i8 %a8, i8 %b8, i8 7)
   %V16I8 = call <16 x i8> @llvm.fshl.v16i8(<16 x i8> %a128, <16 x i8> %b128, <16 x i8> <i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0>)
@@ -1225,6 +1318,13 @@ define void @splatconstant_funnel_i64(i64 %a64, <2 x i64> %a128, <4 x i64> %a256
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 26 for instruction: %V4I64 = call <4 x i64> @llvm.fshl.v4i64(<4 x i64> %a256, <4 x i64> %b256, <4 x i64> <i64 7, i64 7, i64 7, i64 7>)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 52 for instruction: %V8I64 = call <8 x i64> @llvm.fshl.v8i64(<8 x i64> %a512, <8 x i64> %b512, <8 x i64> <i64 7, i64 7, i64 7, i64 7, i64 7, i64 7, i64 7, i64 7>)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret void
+;
+; AVX512GFNI-LABEL: 'splatconstant_funnel_i64'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %I64 = call i64 @llvm.fshl.i64(i64 %a64, i64 %b64, i64 7)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V2I64 = call <2 x i64> @llvm.fshl.v2i64(<2 x i64> %a128, <2 x i64> %b128, <2 x i64> <i64 7, i64 7>)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V4I64 = call <4 x i64> @llvm.fshl.v4i64(<4 x i64> %a256, <4 x i64> %b256, <4 x i64> <i64 7, i64 7, i64 7, i64 7>)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V8I64 = call <8 x i64> @llvm.fshl.v8i64(<8 x i64> %a512, <8 x i64> %b512, <8 x i64> <i64 7, i64 7, i64 7, i64 7, i64 7, i64 7, i64 7, i64 7>)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret void
 ;
   %I64    = call i64 @llvm.fshl.i64(i64 %a64, i64 %b64, i64 7)
   %V2I64  = call <2 x i64> @llvm.fshl.v2i64(<2 x i64> %a128, <2 x i64> %b128, <2 x i64> <i64 7, i64 7>)
@@ -1310,6 +1410,13 @@ define void @splatconstant_funnel_i32(i32 %a32, <4 x i32> %a128, <8 x i32> %a256
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 26 for instruction: %V4I32 = call <8 x i32> @llvm.fshl.v8i32(<8 x i32> %a256, <8 x i32> %b256, <8 x i32> <i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5>)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 52 for instruction: %V8I32 = call <16 x i32> @llvm.fshl.v16i32(<16 x i32> %a512, <16 x i32> %b512, <16 x i32> <i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5>)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret void
+;
+; AVX512GFNI-LABEL: 'splatconstant_funnel_i32'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %I32 = call i32 @llvm.fshl.i32(i32 %a32, i32 %b32, i32 5)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V2I32 = call <4 x i32> @llvm.fshl.v4i32(<4 x i32> %a128, <4 x i32> %b128, <4 x i32> <i32 5, i32 5, i32 5, i32 5>)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V4I32 = call <8 x i32> @llvm.fshl.v8i32(<8 x i32> %a256, <8 x i32> %b256, <8 x i32> <i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5>)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V8I32 = call <16 x i32> @llvm.fshl.v16i32(<16 x i32> %a512, <16 x i32> %b512, <16 x i32> <i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5>)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret void
 ;
   %I32   = call i32 @llvm.fshl.i32(i32 %a32, i32 %b32, i32 5)
   %V2I32 = call <4 x i32> @llvm.fshl.v4i32(<4 x i32> %a128, <4 x i32> %b128, <4 x i32> <i32 5, i32 5, i32 5, i32 5>)
@@ -1395,6 +1502,13 @@ define void @splatconstant_funnel_i16(i16 %a16, <8 x i16> %a128, <16 x i16> %a25
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 27 for instruction: %V16I16 = call <16 x i16> @llvm.fshl.v16i16(<16 x i16> %a256, <16 x i16> %b256, <16 x i16> <i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3>)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 54 for instruction: %V32I16 = call <32 x i16> @llvm.fshl.v32i16(<32 x i16> %a512, <32 x i16> %b512, <32 x i16> <i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3>)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret void
+;
+; AVX512GFNI-LABEL: 'splatconstant_funnel_i16'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 5 for instruction: %I16 = call i16 @llvm.fshl.i16(i16 %a16, i16 %b16, i16 3)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V8I16 = call <8 x i16> @llvm.fshl.v8i16(<8 x i16> %a128, <8 x i16> %b128, <8 x i16> <i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3>)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V16I16 = call <16 x i16> @llvm.fshl.v16i16(<16 x i16> %a256, <16 x i16> %b256, <16 x i16> <i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3>)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V32I16 = call <32 x i16> @llvm.fshl.v32i16(<32 x i16> %a512, <32 x i16> %b512, <32 x i16> <i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3>)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret void
 ;
   %I16    = call i16 @llvm.fshl.i16(i16 %a16, i16 %b16, i16 3)
   %V8I16  = call <8 x i16> @llvm.fshl.v8i16(<8 x i16> %a128, <8 x i16> %b128, <8 x i16> <i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3>)
@@ -1480,6 +1594,13 @@ define void @splatconstant_funnel_i8(i8 %a8, <16 x i8> %a128, <32 x i8> %a256, <
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 29 for instruction: %V32I8 = call <32 x i8> @llvm.fshl.v32i8(<32 x i8> %a256, <32 x i8> %b256, <32 x i8> <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, 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>)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 58 for instruction: %V64I8 = call <64 x i8> @llvm.fshl.v64i8(<64 x i8> %a512, <64 x i8> %b512, <64 x i8> <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, 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, 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, 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>)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret void
+;
+; AVX512GFNI-LABEL: 'splatconstant_funnel_i8'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 5 for instruction: %I8 = call i8 @llvm.fshl.i8(i8 %a8, i8 %b8, i8 3)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 10 for instruction: %V16I8 = call <16 x i8> @llvm.fshl.v16i8(<16 x i8> %a128, <16 x i8> %b128, <16 x i8> <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>)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 10 for instruction: %V32I8 = call <32 x i8> @llvm.fshl.v32i8(<32 x i8> %a256, <32 x i8> %b256, <32 x i8> <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, 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>)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 10 for instruction: %V64I8 = call <64 x i8> @llvm.fshl.v64i8(<64 x i8> %a512, <64 x i8> %b512, <64 x i8> <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, 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, 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, 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>)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret void
 ;
   %I8    = call i8 @llvm.fshl.i8(i8 %a8, i8 %b8, i8 3)
   %V16I8 = call <16 x i8> @llvm.fshl.v16i8(<16 x i8> %a128, <16 x i8> %b128, <16 x i8> <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>)
@@ -1562,6 +1683,13 @@ define void @var_rotate_i64(i64 %a64, <2 x i64> %a128, <4 x i64> %a256, <8 x i64
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 6 for instruction: %V4I64 = call <4 x i64> @llvm.fshl.v4i64(<4 x i64> %a256, <4 x i64> %a256, <4 x i64> %c256)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 12 for instruction: %V8I64 = call <8 x i64> @llvm.fshl.v8i64(<8 x i64> %a512, <8 x i64> %a512, <8 x i64> %c512)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret void
+;
+; AVX512GFNI-LABEL: 'var_rotate_i64'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %I64 = call i64 @llvm.fshl.i64(i64 %a64, i64 %a64, i64 %c64)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V2I64 = call <2 x i64> @llvm.fshl.v2i64(<2 x i64> %a128, <2 x i64> %a128, <2 x i64> %c128)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V4I64 = call <4 x i64> @llvm.fshl.v4i64(<4 x i64> %a256, <4 x i64> %a256, <4 x i64> %c256)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V8I64 = call <8 x i64> @llvm.fshl.v8i64(<8 x i64> %a512, <8 x i64> %a512, <8 x i64> %c512)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret void
 ;
   %I64    = call i64 @llvm.fshl.i64(i64 %a64, i64 %a64, i64 %c64)
   %V2I64  = call <2 x i64> @llvm.fshl.v2i64(<2 x i64> %a128, <2 x i64> %a128, <2 x i64> %c128)
@@ -1647,6 +1775,13 @@ define void @var_rotate_i32(i32 %a32, <4 x i32> %a128, <8 x i32> %a256, <16 x i3
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 6 for instruction: %V4I32 = call <8 x i32> @llvm.fshl.v8i32(<8 x i32> %a256, <8 x i32> %a256, <8 x i32> %c256)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 12 for instruction: %V8I32 = call <16 x i32> @llvm.fshl.v16i32(<16 x i32> %a512, <16 x i32> %a512, <16 x i32> %c512)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret void
+;
+; AVX512GFNI-LABEL: 'var_rotate_i32'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %I32 = call i32 @llvm.fshl.i32(i32 %a32, i32 %a32, i32 %c32)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V2I32 = call <4 x i32> @llvm.fshl.v4i32(<4 x i32> %a128, <4 x i32> %a128, <4 x i32> %c128)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V4I32 = call <8 x i32> @llvm.fshl.v8i32(<8 x i32> %a256, <8 x i32> %a256, <8 x i32> %c256)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V8I32 = call <16 x i32> @llvm.fshl.v16i32(<16 x i32> %a512, <16 x i32> %a512, <16 x i32> %c512)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret void
 ;
   %I32   = call i32 @llvm.fshl.i32(i32 %a32, i32 %a32, i32 %c32)
   %V2I32 = call <4 x i32> @llvm.fshl.v4i32(<4 x i32> %a128, <4 x i32> %a128, <4 x i32> %c128)
@@ -1732,6 +1867,13 @@ define void @var_rotate_i16(i16 %a16, <8 x i16> %a128, <16 x i16> %a256, <32 x i
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 6 for instruction: %V16I16 = call <16 x i16> @llvm.fshl.v16i16(<16 x i16> %a256, <16 x i16> %a256, <16 x i16> %c256)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 12 for instruction: %V32I16 = call <32 x i16> @llvm.fshl.v32i16(<32 x i16> %a512, <32 x i16> %a512, <32 x i16> %c512)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret void
+;
+; AVX512GFNI-LABEL: 'var_rotate_i16'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %I16 = call i16 @llvm.fshl.i16(i16 %a16, i16 %a16, i16 %c16)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V8I16 = call <8 x i16> @llvm.fshl.v8i16(<8 x i16> %a128, <8 x i16> %a128, <8 x i16> %c128)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V16I16 = call <16 x i16> @llvm.fshl.v16i16(<16 x i16> %a256, <16 x i16> %a256, <16 x i16> %c256)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V32I16 = call <32 x i16> @llvm.fshl.v32i16(<32 x i16> %a512, <32 x i16> %a512, <32 x i16> %c512)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret void
 ;
   %I16    = call i16 @llvm.fshl.i16(i16 %a16, i16 %a16, i16 %c16)
   %V8I16  = call <8 x i16> @llvm.fshl.v8i16(<8 x i16> %a128, <8 x i16> %a128, <8 x i16> %c128)
@@ -1817,6 +1959,13 @@ define void @var_rotate_i8(i8 %a8, <16 x i8> %a128, <32 x i8> %a256, <64 x i8> %
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 6 for instruction: %V32I8 = call <32 x i8> @llvm.fshl.v32i8(<32 x i8> %a256, <32 x i8> %a256, <32 x i8> %c256)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 12 for instruction: %V64I8 = call <64 x i8> @llvm.fshl.v64i8(<64 x i8> %a512, <64 x i8> %a512, <64 x i8> %c512)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret void
+;
+; AVX512GFNI-LABEL: 'var_rotate_i8'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %I8 = call i8 @llvm.fshl.i8(i8 %a8, i8 %a8, i8 %c8)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 10 for instruction: %V16I8 = call <16 x i8> @llvm.fshl.v16i8(<16 x i8> %a128, <16 x i8> %a128, <16 x i8> %c128)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 10 for instruction: %V32I8 = call <32 x i8> @llvm.fshl.v32i8(<32 x i8> %a256, <32 x i8> %a256, <32 x i8> %c256)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 12 for instruction: %V64I8 = call <64 x i8> @llvm.fshl.v64i8(<64 x i8> %a512, <64 x i8> %a512, <64 x i8> %c512)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret void
 ;
   %I8    = call i8 @llvm.fshl.i8(i8 %a8, i8 %a8, i8 %c8)
   %V16I8 = call <16 x i8> @llvm.fshl.v16i8(<16 x i8> %a128, <16 x i8> %a128, <16 x i8> %c128)
@@ -1919,6 +2068,15 @@ define void @splatvar_rotate_i64(i64 %a64, <2 x i64> %a128, <4 x i64> %a256, <8
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 6 for instruction: %V4I64 = call <4 x i64> @llvm.fshl.v4i64(<4 x i64> %a256, <4 x i64> %a256, <4 x i64> %u256)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 12 for instruction: %V8I64 = call <8 x i64> @llvm.fshl.v8i64(<8 x i64> %a512, <8 x i64> %a512, <8 x i64> %u512)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret void
+;
+; AVX512GFNI-LABEL: 'splatvar_rotate_i64'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %u128 = shufflevector <2 x i64> %c128, <2 x i64> undef, <2 x i32> zeroinitializer
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %u256 = shufflevector <4 x i64> %c256, <4 x i64> undef, <4 x i32> zeroinitializer
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %u512 = shufflevector <8 x i64> %c512, <8 x i64> undef, <8 x i32> zeroinitializer
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V2I64 = call <2 x i64> @llvm.fshl.v2i64(<2 x i64> %a128, <2 x i64> %a128, <2 x i64> %u128)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V4I64 = call <4 x i64> @llvm.fshl.v4i64(<4 x i64> %a256, <4 x i64> %a256, <4 x i64> %u256)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V8I64 = call <8 x i64> @llvm.fshl.v8i64(<8 x i64> %a512, <8 x i64> %a512, <8 x i64> %u512)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret void
 ;
   %u128 = shufflevector <2 x i64> %c128, <2 x i64> undef, <2 x i32> zeroinitializer
   %u256 = shufflevector <4 x i64> %c256, <4 x i64> undef, <4 x i32> zeroinitializer
@@ -2019,6 +2177,15 @@ define void @splatvar_rotate_i32(i32 %a32, <4 x i32> %a128, <8 x i32> %a256, <16
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 6 for instruction: %V4I32 = call <8 x i32> @llvm.fshl.v8i32(<8 x i32> %a256, <8 x i32> %a256, <8 x i32> %u256)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 12 for instruction: %V8I32 = call <16 x i32> @llvm.fshl.v16i32(<16 x i32> %a512, <16 x i32> %a512, <16 x i32> %u512)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret void
+;
+; AVX512GFNI-LABEL: 'splatvar_rotate_i32'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %u128 = shufflevector <4 x i32> %c128, <4 x i32> undef, <4 x i32> zeroinitializer
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %u256 = shufflevector <8 x i32> %c256, <8 x i32> undef, <8 x i32> zeroinitializer
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %u512 = shufflevector <16 x i32> %c512, <16 x i32> undef, <16 x i32> zeroinitializer
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V2I32 = call <4 x i32> @llvm.fshl.v4i32(<4 x i32> %a128, <4 x i32> %a128, <4 x i32> %u128)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V4I32 = call <8 x i32> @llvm.fshl.v8i32(<8 x i32> %a256, <8 x i32> %a256, <8 x i32> %u256)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V8I32 = call <16 x i32> @llvm.fshl.v16i32(<16 x i32> %a512, <16 x i32> %a512, <16 x i32> %u512)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret void
 ;
   %u128 = shufflevector <4 x i32> %c128, <4 x i32> undef, <4 x i32> zeroinitializer
   %u256 = shufflevector <8 x i32> %c256, <8 x i32> undef, <8 x i32> zeroinitializer
@@ -2119,6 +2286,15 @@ define void @splatvar_rotate_i16(i16 %a16, <8 x i16> %a128, <16 x i16> %a256, <3
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 6 for instruction: %V16I16 = call <16 x i16> @llvm.fshl.v16i16(<16 x i16> %a256, <16 x i16> %a256, <16 x i16> %u256)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 12 for instruction: %V32I16 = call <32 x i16> @llvm.fshl.v32i16(<32 x i16> %a512, <32 x i16> %a512, <32 x i16> %u512)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret void
+;
+; AVX512GFNI-LABEL: 'splatvar_rotate_i16'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %u128 = shufflevector <8 x i16> %c128, <8 x i16> undef, <8 x i32> zeroinitializer
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %u256 = shufflevector <16 x i16> %c256, <16 x i16> undef, <16 x i32> zeroinitializer
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %u512 = shufflevector <32 x i16> %c512, <32 x i16> undef, <32 x i32> zeroinitializer
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V8I16 = call <8 x i16> @llvm.fshl.v8i16(<8 x i16> %a128, <8 x i16> %a128, <8 x i16> %u128)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V16I16 = call <16 x i16> @llvm.fshl.v16i16(<16 x i16> %a256, <16 x i16> %a256, <16 x i16> %u256)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V32I16 = call <32 x i16> @llvm.fshl.v32i16(<32 x i16> %a512, <32 x i16> %a512, <32 x i16> %u512)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret void
 ;
   %u128 = shufflevector <8 x i16> %c128, <8 x i16> undef, <8 x i32> zeroinitializer
   %u256 = shufflevector <16 x i16> %c256, <16 x i16> undef, <16 x i32> zeroinitializer
@@ -2219,6 +2395,15 @@ define void @splatvar_rotate_i8(i8 %a8, <16 x i8> %a128, <32 x i8> %a256, <64 x
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 6 for instruction: %V32I8 = call <32 x i8> @llvm.fshl.v32i8(<32 x i8> %a256, <32 x i8> %a256, <32 x i8> %u256)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 12 for instruction: %V64I8 = call <64 x i8> @llvm.fshl.v64i8(<64 x i8> %a512, <64 x i8> %a512, <64 x i8> %u512)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret void
+;
+; AVX512GFNI-LABEL: 'splatvar_rotate_i8'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %u128 = shufflevector <16 x i8> %c128, <16 x i8> undef, <16 x i32> zeroinitializer
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %u256 = shufflevector <32 x i8> %c256, <32 x i8> undef, <32 x i32> zeroinitializer
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %u512 = shufflevector <64 x i8> %c512, <64 x i8> undef, <64 x i32> zeroinitializer
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 10 for instruction: %V16I8 = call <16 x i8> @llvm.fshl.v16i8(<16 x i8> %a128, <16 x i8> %a128, <16 x i8> %u128)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 10 for instruction: %V32I8 = call <32 x i8> @llvm.fshl.v32i8(<32 x i8> %a256, <32 x i8> %a256, <32 x i8> %u256)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 12 for instruction: %V64I8 = call <64 x i8> @llvm.fshl.v64i8(<64 x i8> %a512, <64 x i8> %a512, <64 x i8> %u512)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret void
 ;
   %u128 = shufflevector <16 x i8> %c128, <16 x i8> undef, <16 x i32> zeroinitializer
   %u256 = shufflevector <32 x i8> %c256, <32 x i8> undef, <32 x i32> zeroinitializer
@@ -2303,6 +2488,13 @@ define void @constant_rotate_i64(i64 %a64, <2 x i64> %a128, <4 x i64> %a256, <8
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 6 for instruction: %V4I64 = call <4 x i64> @llvm.fshl.v4i64(<4 x i64> %a256, <4 x i64> %a256, <4 x i64> <i64 1, i64 7, i64 15, i64 31>)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 12 for instruction: %V8I64 = call <8 x i64> @llvm.fshl.v8i64(<8 x i64> %a512, <8 x i64> %a512, <8 x i64> <i64 1, i64 7, i64 15, i64 31, i64 1, i64 7, i64 15, i64 31>)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret void
+;
+; AVX512GFNI-LABEL: 'constant_rotate_i64'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %I64 = call i64 @llvm.fshl.i64(i64 %a64, i64 %a64, i64 7)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V2I64 = call <2 x i64> @llvm.fshl.v2i64(<2 x i64> %a128, <2 x i64> %a128, <2 x i64> <i64 1, i64 7>)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V4I64 = call <4 x i64> @llvm.fshl.v4i64(<4 x i64> %a256, <4 x i64> %a256, <4 x i64> <i64 1, i64 7, i64 15, i64 31>)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V8I64 = call <8 x i64> @llvm.fshl.v8i64(<8 x i64> %a512, <8 x i64> %a512, <8 x i64> <i64 1, i64 7, i64 15, i64 31, i64 1, i64 7, i64 15, i64 31>)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret void
 ;
   %I64    = call i64 @llvm.fshl.i64(i64 %a64, i64 %a64, i64 7)
   %V2I64  = call <2 x i64> @llvm.fshl.v2i64(<2 x i64> %a128, <2 x i64> %a128, <2 x i64> <i64 1, i64 7>)
@@ -2388,6 +2580,13 @@ define void @constant_rotate_i32(i32 %a32, <4 x i32> %a128, <8 x i32> %a256, <16
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 6 for instruction: %V4I32 = call <8 x i32> @llvm.fshl.v8i32(<8 x i32> %a256, <8 x i32> %a256, <8 x i32> <i32 4, i32 5, i32 6, i32 7, i32 0, i32 1, i32 2, i32 3>)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 12 for instruction: %V8I32 = call <16 x i32> @llvm.fshl.v16i32(<16 x i32> %a512, <16 x i32> %a512, <16 x i32> <i32 4, i32 5, i32 6, i32 7, i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 0, i32 1, i32 2, i32 3>)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret void
+;
+; AVX512GFNI-LABEL: 'constant_rotate_i32'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %I32 = call i32 @llvm.fshl.i32(i32 %a32, i32 %a32, i32 7)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V2I32 = call <4 x i32> @llvm.fshl.v4i32(<4 x i32> %a128, <4 x i32> %a128, <4 x i32> <i32 4, i32 5, i32 6, i32 7>)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V4I32 = call <8 x i32> @llvm.fshl.v8i32(<8 x i32> %a256, <8 x i32> %a256, <8 x i32> <i32 4, i32 5, i32 6, i32 7, i32 0, i32 1, i32 2, i32 3>)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V8I32 = call <16 x i32> @llvm.fshl.v16i32(<16 x i32> %a512, <16 x i32> %a512, <16 x i32> <i32 4, i32 5, i32 6, i32 7, i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 0, i32 1, i32 2, i32 3>)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret void
 ;
   %I32   = call i32 @llvm.fshl.i32(i32 %a32, i32 %a32, i32 7)
   %V2I32 = call <4 x i32> @llvm.fshl.v4i32(<4 x i32> %a128, <4 x i32> %a128, <4 x i32> <i32 4, i32 5, i32 6, i32 7>)
@@ -2473,6 +2672,13 @@ define void @constant_rotate_i16(i16 %a16, <8 x i16> %a128, <16 x i16> %a256, <3
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 6 for instruction: %V16I16 = call <16 x i16> @llvm.fshl.v16i16(<16 x i16> %a256, <16 x i16> %a256, <16 x i16> <i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7>)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 12 for instruction: %V32I16 = call <32 x i16> @llvm.fshl.v32i16(<32 x i16> %a512, <32 x i16> %a512, <32 x i16> <i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7>)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret void
+;
+; AVX512GFNI-LABEL: 'constant_rotate_i16'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %I16 = call i16 @llvm.fshl.i16(i16 %a16, i16 %a16, i16 7)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V8I16 = call <8 x i16> @llvm.fshl.v8i16(<8 x i16> %a128, <8 x i16> %a128, <8 x i16> <i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7>)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V16I16 = call <16 x i16> @llvm.fshl.v16i16(<16 x i16> %a256, <16 x i16> %a256, <16 x i16> <i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7>)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V32I16 = call <32 x i16> @llvm.fshl.v32i16(<32 x i16> %a512, <32 x i16> %a512, <32 x i16> <i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7>)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret void
 ;
   %I16    = call i16 @llvm.fshl.i16(i16 %a16, i16 %a16, i16 7)
   %V8I16  = call <8 x i16> @llvm.fshl.v8i16(<8 x i16> %a128, <8 x i16> %a128, <8 x i16> <i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7>)
@@ -2558,6 +2764,13 @@ define void @constant_rotate_i8(i8 %a8, <16 x i8> %a128, <32 x i8> %a256, <64 x
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 6 for instruction: %V32I8 = call <32 x i8> @llvm.fshl.v32i8(<32 x i8> %a256, <32 x i8> %a256, <32 x i8> <i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0>)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 12 for instruction: %V64I8 = call <64 x i8> @llvm.fshl.v64i8(<64 x i8> %a512, <64 x i8> %a512, <64 x i8> <i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0>)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret void
+;
+; AVX512GFNI-LABEL: 'constant_rotate_i8'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %I8 = call i8 @llvm.fshl.i8(i8 %a8, i8 %a8, i8 7)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 10 for instruction: %V16I8 = call <16 x i8> @llvm.fshl.v16i8(<16 x i8> %a128, <16 x i8> %a128, <16 x i8> <i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0>)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 10 for instruction: %V32I8 = call <32 x i8> @llvm.fshl.v32i8(<32 x i8> %a256, <32 x i8> %a256, <32 x i8> <i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0>)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 12 for instruction: %V64I8 = call <64 x i8> @llvm.fshl.v64i8(<64 x i8> %a512, <64 x i8> %a512, <64 x i8> <i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0>)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret void
 ;
   %I8    = call i8 @llvm.fshl.i8(i8 %a8, i8 %a8, i8 7)
   %V16I8 = call <16 x i8> @llvm.fshl.v16i8(<16 x i8> %a128, <16 x i8> %a128, <16 x i8> <i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0>)
@@ -2640,6 +2853,13 @@ define void @splatconstant_rotate_i64(i64 %a64, <2 x i64> %a128, <4 x i64> %a256
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 6 for instruction: %V4I64 = call <4 x i64> @llvm.fshl.v4i64(<4 x i64> %a256, <4 x i64> %a256, <4 x i64> <i64 7, i64 7, i64 7, i64 7>)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 12 for instruction: %V8I64 = call <8 x i64> @llvm.fshl.v8i64(<8 x i64> %a512, <8 x i64> %a512, <8 x i64> <i64 7, i64 7, i64 7, i64 7, i64 7, i64 7, i64 7, i64 7>)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret void
+;
+; AVX512GFNI-LABEL: 'splatconstant_rotate_i64'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %I64 = call i64 @llvm.fshl.i64(i64 %a64, i64 %a64, i64 7)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V2I64 = call <2 x i64> @llvm.fshl.v2i64(<2 x i64> %a128, <2 x i64> %a128, <2 x i64> <i64 7, i64 7>)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V4I64 = call <4 x i64> @llvm.fshl.v4i64(<4 x i64> %a256, <4 x i64> %a256, <4 x i64> <i64 7, i64 7, i64 7, i64 7>)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V8I64 = call <8 x i64> @llvm.fshl.v8i64(<8 x i64> %a512, <8 x i64> %a512, <8 x i64> <i64 7, i64 7, i64 7, i64 7, i64 7, i64 7, i64 7, i64 7>)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret void
 ;
   %I64    = call i64 @llvm.fshl.i64(i64 %a64, i64 %a64, i64 7)
   %V2I64  = call <2 x i64> @llvm.fshl.v2i64(<2 x i64> %a128, <2 x i64> %a128, <2 x i64> <i64 7, i64 7>)
@@ -2718,6 +2938,13 @@ define void @splatconstant_rotate_i32(i32 %a32, <4 x i32> %a128, <8 x i32> %a256
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 6 for instruction: %V4I32 = call <8 x i32> @llvm.fshl.v8i32(<8 x i32> %a256, <8 x i32> %a256, <8 x i32> <i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5>)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 12 for instruction: %V8I32 = call <16 x i32> @llvm.fshl.v16i32(<16 x i32> %a512, <16 x i32> %a512, <16 x i32> <i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5>)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret void
+;
+; AVX512GFNI-LABEL: 'splatconstant_rotate_i32'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %I32 = call i32 @llvm.fshl.i32(i32 %a32, i32 %a32, i32 5)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V2I32 = call <4 x i32> @llvm.fshl.v4i32(<4 x i32> %a128, <4 x i32> %a128, <4 x i32> <i32 5, i32 5, i32 5, i32 5>)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V4I32 = call <8 x i32> @llvm.fshl.v8i32(<8 x i32> %a256, <8 x i32> %a256, <8 x i32> <i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5>)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V8I32 = call <16 x i32> @llvm.fshl.v16i32(<16 x i32> %a512, <16 x i32> %a512, <16 x i32> <i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5>)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret void
 ;
   %I32   = call i32 @llvm.fshl.i32(i32 %a32, i32 %a32, i32 5)
   %V2I32 = call <4 x i32> @llvm.fshl.v4i32(<4 x i32> %a128, <4 x i32> %a128, <4 x i32> <i32 5, i32 5, i32 5, i32 5>)
@@ -2796,6 +3023,13 @@ define void @splatconstant_rotate_i16(i16 %a16, <8 x i16> %a128, <16 x i16> %a25
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 6 for instruction: %V16I16 = call <16 x i16> @llvm.fshl.v16i16(<16 x i16> %a256, <16 x i16> %a256, <16 x i16> <i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3>)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 12 for instruction: %V32I16 = call <32 x i16> @llvm.fshl.v32i16(<32 x i16> %a512, <32 x i16> %a512, <32 x i16> <i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3>)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret void
+;
+; AVX512GFNI-LABEL: 'splatconstant_rotate_i16'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %I16 = call i16 @llvm.fshl.i16(i16 %a16, i16 %a16, i16 3)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V8I16 = call <8 x i16> @llvm.fshl.v8i16(<8 x i16> %a128, <8 x i16> %a128, <8 x i16> <i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3>)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V16I16 = call <16 x i16> @llvm.fshl.v16i16(<16 x i16> %a256, <16 x i16> %a256, <16 x i16> <i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3>)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V32I16 = call <32 x i16> @llvm.fshl.v32i16(<32 x i16> %a512, <32 x i16> %a512, <32 x i16> <i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3>)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret void
 ;
   %I16    = call i16 @llvm.fshl.i16(i16 %a16, i16 %a16, i16 3)
   %V8I16  = call <8 x i16> @llvm.fshl.v8i16(<8 x i16> %a128, <8 x i16> %a128, <8 x i16> <i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3>)
@@ -2874,6 +3108,13 @@ define void @splatconstant_rotate_i8(i8 %a8, <16 x i8> %a128, <32 x i8> %a256, <
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 6 for instruction: %V32I8 = call <32 x i8> @llvm.fshl.v32i8(<32 x i8> %a256, <32 x i8> %a256, <32 x i8> <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, 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>)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 12 for instruction: %V64I8 = call <64 x i8> @llvm.fshl.v64i8(<64 x i8> %a512, <64 x i8> %a512, <64 x i8> <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, 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, 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, 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>)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret void
+;
+; AVX512GFNI-LABEL: 'splatconstant_rotate_i8'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %I8 = call i8 @llvm.fshl.i8(i8 %a8, i8 %a8, i8 3)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 10 for instruction: %V16I8 = call <16 x i8> @llvm.fshl.v16i8(<16 x i8> %a128, <16 x i8> %a128, <16 x i8> <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>)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 10 for instruction: %V32I8 = call <32 x i8> @llvm.fshl.v32i8(<32 x i8> %a256, <32 x i8> %a256, <32 x i8> <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, 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>)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 12 for instruction: %V64I8 = call <64 x i8> @llvm.fshl.v64i8(<64 x i8> %a512, <64 x i8> %a512, <64 x i8> <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, 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, 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, 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>)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret void
 ;
   %I8    = call i8 @llvm.fshl.i8(i8 %a8, i8 %a8, i8 3)
   %V16I8 = call <16 x i8> @llvm.fshl.v16i8(<16 x i8> %a128, <16 x i8> %a128, <16 x i8> <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>)

diff  --git a/llvm/test/Analysis/CostModel/X86/fshl.ll b/llvm/test/Analysis/CostModel/X86/fshl.ll
index 4e688c29c7ea33..311d8d5ed7d2a9 100644
--- a/llvm/test/Analysis/CostModel/X86/fshl.ll
+++ b/llvm/test/Analysis/CostModel/X86/fshl.ll
@@ -12,6 +12,7 @@
 ; RUN: opt < %s -passes="print<cost-model>" 2>&1 -disable-output -mtriple=x86_64-apple-macosx10.8.0 -mcpu=goldmont | FileCheck %s --check-prefixes=GLM
 ; RUN: opt < %s -passes="print<cost-model>" 2>&1 -disable-output -mtriple=x86_64-apple-macosx10.8.0 -mcpu=bdver2 | FileCheck %s --check-prefixes=XOP
 ; RUN: opt < %s -passes="print<cost-model>" 2>&1 -disable-output -mtriple=x86_64-apple-macosx10.8.0 -mcpu=btver2 | FileCheck %s --check-prefixes=AVX1
+; RUN: opt < %s -passes="print<cost-model>" 2>&1 -disable-output -mtriple=x86_64-apple-macosx10.8.0 -mcpu=tigerlake | FileCheck %s --check-prefixes=AVX512,AVX512GFNI
 
 target datalayout = "e-p:64:64:64-i1:8:8-i8:8:8-i16:16:16-i32:32:32-i64:64:64-f32:32:32-f64:64:64-v64:64:64-v128:128:128-a0:0:64-s0:64:64-f80:128:128-n8:16:32:64-S128"
 target triple = "x86_64-apple-macosx10.8.0"
@@ -97,6 +98,13 @@ define void @var_funnel_i64(i64 %a64, <2 x i64> %a128, <4 x i64> %a256, <8 x i64
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 23 for instruction: %V4I64 = call <4 x i64> @llvm.fshl.v4i64(<4 x i64> %a256, <4 x i64> %b256, <4 x i64> %c256)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 46 for instruction: %V8I64 = call <8 x i64> @llvm.fshl.v8i64(<8 x i64> %a512, <8 x i64> %b512, <8 x i64> %c512)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret void
+;
+; AVX512GFNI-LABEL: 'var_funnel_i64'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %I64 = call i64 @llvm.fshl.i64(i64 %a64, i64 %b64, i64 %c64)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V2I64 = call <2 x i64> @llvm.fshl.v2i64(<2 x i64> %a128, <2 x i64> %b128, <2 x i64> %c128)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V4I64 = call <4 x i64> @llvm.fshl.v4i64(<4 x i64> %a256, <4 x i64> %b256, <4 x i64> %c256)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V8I64 = call <8 x i64> @llvm.fshl.v8i64(<8 x i64> %a512, <8 x i64> %b512, <8 x i64> %c512)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret void
 ;
   %I64    = call i64 @llvm.fshl.i64(i64 %a64, i64 %b64, i64 %c64)
   %V2I64  = call <2 x i64> @llvm.fshl.v2i64(<2 x i64> %a128, <2 x i64> %b128, <2 x i64> %c128)
@@ -182,6 +190,13 @@ define void @var_funnel_i32(i32 %a32, <4 x i32> %a128, <8 x i32> %a256, <16 x i3
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 23 for instruction: %V4I32 = call <8 x i32> @llvm.fshl.v8i32(<8 x i32> %a256, <8 x i32> %b256, <8 x i32> %c256)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 46 for instruction: %V8I32 = call <16 x i32> @llvm.fshl.v16i32(<16 x i32> %a512, <16 x i32> %b512, <16 x i32> %c512)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret void
+;
+; AVX512GFNI-LABEL: 'var_funnel_i32'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %I32 = call i32 @llvm.fshl.i32(i32 %a32, i32 %b32, i32 %c32)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V2I32 = call <4 x i32> @llvm.fshl.v4i32(<4 x i32> %a128, <4 x i32> %b128, <4 x i32> %c128)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V4I32 = call <8 x i32> @llvm.fshl.v8i32(<8 x i32> %a256, <8 x i32> %b256, <8 x i32> %c256)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V8I32 = call <16 x i32> @llvm.fshl.v16i32(<16 x i32> %a512, <16 x i32> %b512, <16 x i32> %c512)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret void
 ;
   %I32   = call i32 @llvm.fshl.i32(i32 %a32, i32 %b32, i32 %c32)
   %V2I32 = call <4 x i32> @llvm.fshl.v4i32(<4 x i32> %a128, <4 x i32> %b128, <4 x i32> %c128)
@@ -267,6 +282,13 @@ define void @var_funnel_i16(i16 %a16, <8 x i16> %a128, <16 x i16> %a256, <32 x i
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 23 for instruction: %V16I16 = call <16 x i16> @llvm.fshl.v16i16(<16 x i16> %a256, <16 x i16> %b256, <16 x i16> %c256)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 46 for instruction: %V32I16 = call <32 x i16> @llvm.fshl.v32i16(<32 x i16> %a512, <32 x i16> %b512, <32 x i16> %c512)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret void
+;
+; AVX512GFNI-LABEL: 'var_funnel_i16'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %I16 = call i16 @llvm.fshl.i16(i16 %a16, i16 %b16, i16 %c16)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V8I16 = call <8 x i16> @llvm.fshl.v8i16(<8 x i16> %a128, <8 x i16> %b128, <8 x i16> %c128)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V16I16 = call <16 x i16> @llvm.fshl.v16i16(<16 x i16> %a256, <16 x i16> %b256, <16 x i16> %c256)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V32I16 = call <32 x i16> @llvm.fshl.v32i16(<32 x i16> %a512, <32 x i16> %b512, <32 x i16> %c512)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret void
 ;
   %I16    = call i16 @llvm.fshl.i16(i16 %a16, i16 %b16, i16 %c16)
   %V8I16  = call <8 x i16> @llvm.fshl.v8i16(<8 x i16> %a128, <8 x i16> %b128, <8 x i16> %c128)
@@ -352,6 +374,13 @@ define void @var_funnel_i8(i8 %a8, <16 x i8> %a128, <32 x i8> %a256, <64 x i8> %
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 23 for instruction: %V32I8 = call <32 x i8> @llvm.fshl.v32i8(<32 x i8> %a256, <32 x i8> %b256, <32 x i8> %c256)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 46 for instruction: %V64I8 = call <64 x i8> @llvm.fshl.v64i8(<64 x i8> %a512, <64 x i8> %b512, <64 x i8> %c512)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret void
+;
+; AVX512GFNI-LABEL: 'var_funnel_i8'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %I8 = call i8 @llvm.fshl.i8(i8 %a8, i8 %b8, i8 %c8)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 13 for instruction: %V16I8 = call <16 x i8> @llvm.fshl.v16i8(<16 x i8> %a128, <16 x i8> %b128, <16 x i8> %c128)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 13 for instruction: %V32I8 = call <32 x i8> @llvm.fshl.v32i8(<32 x i8> %a256, <32 x i8> %b256, <32 x i8> %c256)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 18 for instruction: %V64I8 = call <64 x i8> @llvm.fshl.v64i8(<64 x i8> %a512, <64 x i8> %b512, <64 x i8> %c512)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret void
 ;
   %I8    = call i8 @llvm.fshl.i8(i8 %a8, i8 %b8, i8 %c8)
   %V16I8 = call <16 x i8> @llvm.fshl.v16i8(<16 x i8> %a128, <16 x i8> %b128, <16 x i8> %c128)
@@ -463,6 +492,15 @@ define void @splatvar_funnel_i64(i64 %a64, <2 x i64> %a128, <4 x i64> %a256, <8
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 19 for instruction: %V4I64 = call <4 x i64> @llvm.fshl.v4i64(<4 x i64> %a256, <4 x i64> %b256, <4 x i64> %u256)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 38 for instruction: %V8I64 = call <8 x i64> @llvm.fshl.v8i64(<8 x i64> %a512, <8 x i64> %b512, <8 x i64> %u512)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret void
+;
+; AVX512GFNI-LABEL: 'splatvar_funnel_i64'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %u128 = shufflevector <2 x i64> %c128, <2 x i64> undef, <2 x i32> zeroinitializer
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %u256 = shufflevector <4 x i64> %c256, <4 x i64> undef, <4 x i32> zeroinitializer
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %u512 = shufflevector <8 x i64> %c512, <8 x i64> undef, <8 x i32> zeroinitializer
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V2I64 = call <2 x i64> @llvm.fshl.v2i64(<2 x i64> %a128, <2 x i64> %b128, <2 x i64> %u128)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V4I64 = call <4 x i64> @llvm.fshl.v4i64(<4 x i64> %a256, <4 x i64> %b256, <4 x i64> %u256)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V8I64 = call <8 x i64> @llvm.fshl.v8i64(<8 x i64> %a512, <8 x i64> %b512, <8 x i64> %u512)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret void
 ;
   %u128 = shufflevector <2 x i64> %c128, <2 x i64> undef, <2 x i32> zeroinitializer
   %u256 = shufflevector <4 x i64> %c256, <4 x i64> undef, <4 x i32> zeroinitializer
@@ -563,6 +601,15 @@ define void @splatvar_funnel_i32(i32 %a32, <4 x i32> %a128, <8 x i32> %a256, <16
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 19 for instruction: %V4I32 = call <8 x i32> @llvm.fshl.v8i32(<8 x i32> %a256, <8 x i32> %b256, <8 x i32> %u256)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 38 for instruction: %V8I32 = call <16 x i32> @llvm.fshl.v16i32(<16 x i32> %a512, <16 x i32> %b512, <16 x i32> %u512)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret void
+;
+; AVX512GFNI-LABEL: 'splatvar_funnel_i32'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %u128 = shufflevector <4 x i32> %c128, <4 x i32> undef, <4 x i32> zeroinitializer
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %u256 = shufflevector <8 x i32> %c256, <8 x i32> undef, <8 x i32> zeroinitializer
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %u512 = shufflevector <16 x i32> %c512, <16 x i32> undef, <16 x i32> zeroinitializer
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V2I32 = call <4 x i32> @llvm.fshl.v4i32(<4 x i32> %a128, <4 x i32> %b128, <4 x i32> %u128)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V4I32 = call <8 x i32> @llvm.fshl.v8i32(<8 x i32> %a256, <8 x i32> %b256, <8 x i32> %u256)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V8I32 = call <16 x i32> @llvm.fshl.v16i32(<16 x i32> %a512, <16 x i32> %b512, <16 x i32> %u512)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret void
 ;
   %u128 = shufflevector <4 x i32> %c128, <4 x i32> undef, <4 x i32> zeroinitializer
   %u256 = shufflevector <8 x i32> %c256, <8 x i32> undef, <8 x i32> zeroinitializer
@@ -663,6 +710,15 @@ define void @splatvar_funnel_i16(i16 %a16, <8 x i16> %a128, <16 x i16> %a256, <3
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 19 for instruction: %V16I16 = call <16 x i16> @llvm.fshl.v16i16(<16 x i16> %a256, <16 x i16> %b256, <16 x i16> %u256)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 38 for instruction: %V32I16 = call <32 x i16> @llvm.fshl.v32i16(<32 x i16> %a512, <32 x i16> %b512, <32 x i16> %u512)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret void
+;
+; AVX512GFNI-LABEL: 'splatvar_funnel_i16'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %u128 = shufflevector <8 x i16> %c128, <8 x i16> undef, <8 x i32> zeroinitializer
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %u256 = shufflevector <16 x i16> %c256, <16 x i16> undef, <16 x i32> zeroinitializer
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %u512 = shufflevector <32 x i16> %c512, <32 x i16> undef, <32 x i32> zeroinitializer
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V8I16 = call <8 x i16> @llvm.fshl.v8i16(<8 x i16> %a128, <8 x i16> %b128, <8 x i16> %u128)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V16I16 = call <16 x i16> @llvm.fshl.v16i16(<16 x i16> %a256, <16 x i16> %b256, <16 x i16> %u256)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V32I16 = call <32 x i16> @llvm.fshl.v32i16(<32 x i16> %a512, <32 x i16> %b512, <32 x i16> %u512)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret void
 ;
   %u128 = shufflevector <8 x i16> %c128, <8 x i16> undef, <8 x i32> zeroinitializer
   %u256 = shufflevector <16 x i16> %c256, <16 x i16> undef, <16 x i32> zeroinitializer
@@ -763,6 +819,15 @@ define void @splatvar_funnel_i8(i8 %a8, <16 x i8> %a128, <32 x i8> %a256, <64 x
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 23 for instruction: %V32I8 = call <32 x i8> @llvm.fshl.v32i8(<32 x i8> %a256, <32 x i8> %b256, <32 x i8> %u256)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 46 for instruction: %V64I8 = call <64 x i8> @llvm.fshl.v64i8(<64 x i8> %a512, <64 x i8> %b512, <64 x i8> %u512)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret void
+;
+; AVX512GFNI-LABEL: 'splatvar_funnel_i8'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %u128 = shufflevector <16 x i8> %c128, <16 x i8> undef, <16 x i32> zeroinitializer
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %u256 = shufflevector <32 x i8> %c256, <32 x i8> undef, <32 x i32> zeroinitializer
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %u512 = shufflevector <64 x i8> %c512, <64 x i8> undef, <64 x i32> zeroinitializer
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 11 for instruction: %V16I8 = call <16 x i8> @llvm.fshl.v16i8(<16 x i8> %a128, <16 x i8> %b128, <16 x i8> %u128)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 13 for instruction: %V32I8 = call <32 x i8> @llvm.fshl.v32i8(<32 x i8> %a256, <32 x i8> %b256, <32 x i8> %u256)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 13 for instruction: %V64I8 = call <64 x i8> @llvm.fshl.v64i8(<64 x i8> %a512, <64 x i8> %b512, <64 x i8> %u512)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret void
 ;
   %u128 = shufflevector <16 x i8> %c128, <16 x i8> undef, <16 x i32> zeroinitializer
   %u256 = shufflevector <32 x i8> %c256, <32 x i8> undef, <32 x i32> zeroinitializer
@@ -854,6 +919,13 @@ define void @constant_funnel_i64(i64 %a64, <2 x i64> %a128, <4 x i64> %a256, <8
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 20 for instruction: %V4I64 = call <4 x i64> @llvm.fshl.v4i64(<4 x i64> %a256, <4 x i64> %b256, <4 x i64> <i64 1, i64 7, i64 15, i64 31>)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 40 for instruction: %V8I64 = call <8 x i64> @llvm.fshl.v8i64(<8 x i64> %a512, <8 x i64> %b512, <8 x i64> <i64 1, i64 7, i64 15, i64 31, i64 1, i64 7, i64 15, i64 31>)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret void
+;
+; AVX512GFNI-LABEL: 'constant_funnel_i64'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %I64 = call i64 @llvm.fshl.i64(i64 %a64, i64 %b64, i64 7)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V2I64 = call <2 x i64> @llvm.fshl.v2i64(<2 x i64> %a128, <2 x i64> %b128, <2 x i64> <i64 1, i64 7>)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V4I64 = call <4 x i64> @llvm.fshl.v4i64(<4 x i64> %a256, <4 x i64> %b256, <4 x i64> <i64 1, i64 7, i64 15, i64 31>)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V8I64 = call <8 x i64> @llvm.fshl.v8i64(<8 x i64> %a512, <8 x i64> %b512, <8 x i64> <i64 1, i64 7, i64 15, i64 31, i64 1, i64 7, i64 15, i64 31>)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret void
 ;
   %I64    = call i64 @llvm.fshl.i64(i64 %a64, i64 %b64, i64 7)
   %V2I64  = call <2 x i64> @llvm.fshl.v2i64(<2 x i64> %a128, <2 x i64> %b128, <2 x i64> <i64 1, i64 7>)
@@ -932,6 +1004,13 @@ define void @constant_funnel_i32(i32 %a32, <4 x i32> %a128, <8 x i32> %a256, <16
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 20 for instruction: %V4I32 = call <8 x i32> @llvm.fshl.v8i32(<8 x i32> %a256, <8 x i32> %b256, <8 x i32> <i32 4, i32 5, i32 6, i32 7, i32 0, i32 1, i32 2, i32 3>)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 40 for instruction: %V8I32 = call <16 x i32> @llvm.fshl.v16i32(<16 x i32> %a512, <16 x i32> %b512, <16 x i32> <i32 4, i32 5, i32 6, i32 7, i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 0, i32 1, i32 2, i32 3>)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret void
+;
+; AVX512GFNI-LABEL: 'constant_funnel_i32'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %I32 = call i32 @llvm.fshl.i32(i32 %a32, i32 %b32, i32 7)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V2I32 = call <4 x i32> @llvm.fshl.v4i32(<4 x i32> %a128, <4 x i32> %b128, <4 x i32> <i32 4, i32 5, i32 6, i32 7>)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V4I32 = call <8 x i32> @llvm.fshl.v8i32(<8 x i32> %a256, <8 x i32> %b256, <8 x i32> <i32 4, i32 5, i32 6, i32 7, i32 0, i32 1, i32 2, i32 3>)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V8I32 = call <16 x i32> @llvm.fshl.v16i32(<16 x i32> %a512, <16 x i32> %b512, <16 x i32> <i32 4, i32 5, i32 6, i32 7, i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 0, i32 1, i32 2, i32 3>)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret void
 ;
   %I32   = call i32 @llvm.fshl.i32(i32 %a32, i32 %b32, i32 7)
   %V2I32 = call <4 x i32> @llvm.fshl.v4i32(<4 x i32> %a128, <4 x i32> %b128, <4 x i32> <i32 4, i32 5, i32 6, i32 7>)
@@ -1017,6 +1096,13 @@ define void @constant_funnel_i16(i16 %a16, <8 x i16> %a128, <16 x i16> %a256, <3
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 20 for instruction: %V16I16 = call <16 x i16> @llvm.fshl.v16i16(<16 x i16> %a256, <16 x i16> %b256, <16 x i16> <i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7>)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 40 for instruction: %V32I16 = call <32 x i16> @llvm.fshl.v32i16(<32 x i16> %a512, <32 x i16> %b512, <32 x i16> <i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7>)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret void
+;
+; AVX512GFNI-LABEL: 'constant_funnel_i16'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %I16 = call i16 @llvm.fshl.i16(i16 %a16, i16 %b16, i16 7)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V8I16 = call <8 x i16> @llvm.fshl.v8i16(<8 x i16> %a128, <8 x i16> %b128, <8 x i16> <i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7>)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V16I16 = call <16 x i16> @llvm.fshl.v16i16(<16 x i16> %a256, <16 x i16> %b256, <16 x i16> <i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7>)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V32I16 = call <32 x i16> @llvm.fshl.v32i16(<32 x i16> %a512, <32 x i16> %b512, <32 x i16> <i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7>)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret void
 ;
   %I16    = call i16 @llvm.fshl.i16(i16 %a16, i16 %b16, i16 7)
   %V8I16  = call <8 x i16> @llvm.fshl.v8i16(<8 x i16> %a128, <8 x i16> %b128, <8 x i16> <i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7>)
@@ -1102,6 +1188,13 @@ define void @constant_funnel_i8(i8 %a8, <16 x i8> %a128, <32 x i8> %a256, <64 x
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 20 for instruction: %V32I8 = call <32 x i8> @llvm.fshl.v32i8(<32 x i8> %a256, <32 x i8> %b256, <32 x i8> <i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0>)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 40 for instruction: %V64I8 = call <64 x i8> @llvm.fshl.v64i8(<64 x i8> %a512, <64 x i8> %b512, <64 x i8> <i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0>)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret void
+;
+; AVX512GFNI-LABEL: 'constant_funnel_i8'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %I8 = call i8 @llvm.fshl.i8(i8 %a8, i8 %b8, i8 7)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 12 for instruction: %V16I8 = call <16 x i8> @llvm.fshl.v16i8(<16 x i8> %a128, <16 x i8> %b128, <16 x i8> <i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0>)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 12 for instruction: %V32I8 = call <32 x i8> @llvm.fshl.v32i8(<32 x i8> %a256, <32 x i8> %b256, <32 x i8> <i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0>)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 17 for instruction: %V64I8 = call <64 x i8> @llvm.fshl.v64i8(<64 x i8> %a512, <64 x i8> %b512, <64 x i8> <i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0>)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret void
 ;
   %I8    = call i8 @llvm.fshl.i8(i8 %a8, i8 %b8, i8 7)
   %V16I8 = call <16 x i8> @llvm.fshl.v16i8(<16 x i8> %a128, <16 x i8> %b128, <16 x i8> <i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0>)
@@ -1191,6 +1284,13 @@ define void @splatconstant_funnel_i64(i64 %a64, <2 x i64> %a128, <4 x i64> %a256
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 18 for instruction: %V4I64 = call <4 x i64> @llvm.fshl.v4i64(<4 x i64> %a256, <4 x i64> %b256, <4 x i64> <i64 7, i64 7, i64 7, i64 7>)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 36 for instruction: %V8I64 = call <8 x i64> @llvm.fshl.v8i64(<8 x i64> %a512, <8 x i64> %b512, <8 x i64> <i64 7, i64 7, i64 7, i64 7, i64 7, i64 7, i64 7, i64 7>)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret void
+;
+; AVX512GFNI-LABEL: 'splatconstant_funnel_i64'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %I64 = call i64 @llvm.fshl.i64(i64 %a64, i64 %b64, i64 7)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V2I64 = call <2 x i64> @llvm.fshl.v2i64(<2 x i64> %a128, <2 x i64> %b128, <2 x i64> <i64 7, i64 7>)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V4I64 = call <4 x i64> @llvm.fshl.v4i64(<4 x i64> %a256, <4 x i64> %b256, <4 x i64> <i64 7, i64 7, i64 7, i64 7>)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V8I64 = call <8 x i64> @llvm.fshl.v8i64(<8 x i64> %a512, <8 x i64> %b512, <8 x i64> <i64 7, i64 7, i64 7, i64 7, i64 7, i64 7, i64 7, i64 7>)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret void
 ;
   %I64    = call i64 @llvm.fshl.i64(i64 %a64, i64 %b64, i64 7)
   %V2I64  = call <2 x i64> @llvm.fshl.v2i64(<2 x i64> %a128, <2 x i64> %b128, <2 x i64> <i64 7, i64 7>)
@@ -1269,6 +1369,13 @@ define void @splatconstant_funnel_i32(i32 %a32, <4 x i32> %a128, <8 x i32> %a256
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 18 for instruction: %V4I32 = call <8 x i32> @llvm.fshl.v8i32(<8 x i32> %a256, <8 x i32> %b256, <8 x i32> <i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5>)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 36 for instruction: %V8I32 = call <16 x i32> @llvm.fshl.v16i32(<16 x i32> %a512, <16 x i32> %b512, <16 x i32> <i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5>)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret void
+;
+; AVX512GFNI-LABEL: 'splatconstant_funnel_i32'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %I32 = call i32 @llvm.fshl.i32(i32 %a32, i32 %b32, i32 5)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V2I32 = call <4 x i32> @llvm.fshl.v4i32(<4 x i32> %a128, <4 x i32> %b128, <4 x i32> <i32 5, i32 5, i32 5, i32 5>)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V4I32 = call <8 x i32> @llvm.fshl.v8i32(<8 x i32> %a256, <8 x i32> %b256, <8 x i32> <i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5>)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V8I32 = call <16 x i32> @llvm.fshl.v16i32(<16 x i32> %a512, <16 x i32> %b512, <16 x i32> <i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5>)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret void
 ;
   %I32   = call i32 @llvm.fshl.i32(i32 %a32, i32 %b32, i32 5)
   %V2I32 = call <4 x i32> @llvm.fshl.v4i32(<4 x i32> %a128, <4 x i32> %b128, <4 x i32> <i32 5, i32 5, i32 5, i32 5>)
@@ -1347,6 +1454,13 @@ define void @splatconstant_funnel_i16(i16 %a16, <8 x i16> %a128, <16 x i16> %a25
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 18 for instruction: %V16I16 = call <16 x i16> @llvm.fshl.v16i16(<16 x i16> %a256, <16 x i16> %b256, <16 x i16> <i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3>)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 36 for instruction: %V32I16 = call <32 x i16> @llvm.fshl.v32i16(<32 x i16> %a512, <32 x i16> %b512, <32 x i16> <i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3>)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret void
+;
+; AVX512GFNI-LABEL: 'splatconstant_funnel_i16'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %I16 = call i16 @llvm.fshl.i16(i16 %a16, i16 %b16, i16 3)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V8I16 = call <8 x i16> @llvm.fshl.v8i16(<8 x i16> %a128, <8 x i16> %b128, <8 x i16> <i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3>)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V16I16 = call <16 x i16> @llvm.fshl.v16i16(<16 x i16> %a256, <16 x i16> %b256, <16 x i16> <i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3>)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V32I16 = call <32 x i16> @llvm.fshl.v32i16(<32 x i16> %a512, <32 x i16> %b512, <32 x i16> <i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3>)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret void
 ;
   %I16    = call i16 @llvm.fshl.i16(i16 %a16, i16 %b16, i16 3)
   %V8I16  = call <8 x i16> @llvm.fshl.v8i16(<8 x i16> %a128, <8 x i16> %b128, <8 x i16> <i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3>)
@@ -1425,6 +1539,13 @@ define void @splatconstant_funnel_i8(i8 %a8, <16 x i8> %a128, <32 x i8> %a256, <
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 20 for instruction: %V32I8 = call <32 x i8> @llvm.fshl.v32i8(<32 x i8> %a256, <32 x i8> %b256, <32 x i8> <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, 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>)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 40 for instruction: %V64I8 = call <64 x i8> @llvm.fshl.v64i8(<64 x i8> %a512, <64 x i8> %b512, <64 x i8> <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, 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, 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, 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>)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret void
+;
+; AVX512GFNI-LABEL: 'splatconstant_funnel_i8'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %I8 = call i8 @llvm.fshl.i8(i8 %a8, i8 %b8, i8 3)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 6 for instruction: %V16I8 = call <16 x i8> @llvm.fshl.v16i8(<16 x i8> %a128, <16 x i8> %b128, <16 x i8> <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>)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 6 for instruction: %V32I8 = call <32 x i8> @llvm.fshl.v32i8(<32 x i8> %a256, <32 x i8> %b256, <32 x i8> <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, 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>)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 6 for instruction: %V64I8 = call <64 x i8> @llvm.fshl.v64i8(<64 x i8> %a512, <64 x i8> %b512, <64 x i8> <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, 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, 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, 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>)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret void
 ;
   %I8    = call i8 @llvm.fshl.i8(i8 %a8, i8 %b8, i8 3)
   %V16I8 = call <16 x i8> @llvm.fshl.v16i8(<16 x i8> %a128, <16 x i8> %b128, <16 x i8> <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>)
@@ -1635,6 +1756,13 @@ define void @var_rotate_i16(i16 %a16, <8 x i16> %a128, <16 x i16> %a256, <32 x i
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V16I16 = call <16 x i16> @llvm.fshl.v16i16(<16 x i16> %a256, <16 x i16> %a256, <16 x i16> %c256)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %V32I16 = call <32 x i16> @llvm.fshl.v32i16(<32 x i16> %a512, <32 x i16> %a512, <32 x i16> %c512)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret void
+;
+; AVX512GFNI-LABEL: 'var_rotate_i16'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %I16 = call i16 @llvm.fshl.i16(i16 %a16, i16 %a16, i16 %c16)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V8I16 = call <8 x i16> @llvm.fshl.v8i16(<8 x i16> %a128, <8 x i16> %a128, <8 x i16> %c128)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V16I16 = call <16 x i16> @llvm.fshl.v16i16(<16 x i16> %a256, <16 x i16> %a256, <16 x i16> %c256)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V32I16 = call <32 x i16> @llvm.fshl.v32i16(<32 x i16> %a512, <32 x i16> %a512, <32 x i16> %c512)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret void
 ;
   %I16    = call i16 @llvm.fshl.i16(i16 %a16, i16 %a16, i16 %c16)
   %V8I16  = call <8 x i16> @llvm.fshl.v8i16(<8 x i16> %a128, <8 x i16> %a128, <8 x i16> %c128)
@@ -1720,6 +1848,13 @@ define void @var_rotate_i8(i8 %a8, <16 x i8> %a128, <32 x i8> %a256, <64 x i8> %
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V32I8 = call <32 x i8> @llvm.fshl.v32i8(<32 x i8> %a256, <32 x i8> %a256, <32 x i8> %c256)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %V64I8 = call <64 x i8> @llvm.fshl.v64i8(<64 x i8> %a512, <64 x i8> %a512, <64 x i8> %c512)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret void
+;
+; AVX512GFNI-LABEL: 'var_rotate_i8'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %I8 = call i8 @llvm.fshl.i8(i8 %a8, i8 %a8, i8 %c8)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 5 for instruction: %V16I8 = call <16 x i8> @llvm.fshl.v16i8(<16 x i8> %a128, <16 x i8> %a128, <16 x i8> %c128)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 5 for instruction: %V32I8 = call <32 x i8> @llvm.fshl.v32i8(<32 x i8> %a256, <32 x i8> %a256, <32 x i8> %c256)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 5 for instruction: %V64I8 = call <64 x i8> @llvm.fshl.v64i8(<64 x i8> %a512, <64 x i8> %a512, <64 x i8> %c512)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret void
 ;
   %I8    = call i8 @llvm.fshl.i8(i8 %a8, i8 %a8, i8 %c8)
   %V16I8 = call <16 x i8> @llvm.fshl.v16i8(<16 x i8> %a128, <16 x i8> %a128, <16 x i8> %c128)
@@ -1968,6 +2103,15 @@ define void @splatvar_rotate_i16(i16 %a16, <8 x i16> %a128, <16 x i16> %a256, <3
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V16I16 = call <16 x i16> @llvm.fshl.v16i16(<16 x i16> %a256, <16 x i16> %a256, <16 x i16> %u256)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %V32I16 = call <32 x i16> @llvm.fshl.v32i16(<32 x i16> %a512, <32 x i16> %a512, <32 x i16> %u512)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret void
+;
+; AVX512GFNI-LABEL: 'splatvar_rotate_i16'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %u128 = shufflevector <8 x i16> %c128, <8 x i16> undef, <8 x i32> zeroinitializer
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %u256 = shufflevector <16 x i16> %c256, <16 x i16> undef, <16 x i32> zeroinitializer
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %u512 = shufflevector <32 x i16> %c512, <32 x i16> undef, <32 x i32> zeroinitializer
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V8I16 = call <8 x i16> @llvm.fshl.v8i16(<8 x i16> %a128, <8 x i16> %a128, <8 x i16> %u128)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V16I16 = call <16 x i16> @llvm.fshl.v16i16(<16 x i16> %a256, <16 x i16> %a256, <16 x i16> %u256)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V32I16 = call <32 x i16> @llvm.fshl.v32i16(<32 x i16> %a512, <32 x i16> %a512, <32 x i16> %u512)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret void
 ;
   %u128 = shufflevector <8 x i16> %c128, <8 x i16> undef, <8 x i32> zeroinitializer
   %u256 = shufflevector <16 x i16> %c256, <16 x i16> undef, <16 x i32> zeroinitializer
@@ -2068,6 +2212,15 @@ define void @splatvar_rotate_i8(i8 %a8, <16 x i8> %a128, <32 x i8> %a256, <64 x
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V32I8 = call <32 x i8> @llvm.fshl.v32i8(<32 x i8> %a256, <32 x i8> %a256, <32 x i8> %u256)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %V64I8 = call <64 x i8> @llvm.fshl.v64i8(<64 x i8> %a512, <64 x i8> %a512, <64 x i8> %u512)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret void
+;
+; AVX512GFNI-LABEL: 'splatvar_rotate_i8'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %u128 = shufflevector <16 x i8> %c128, <16 x i8> undef, <16 x i32> zeroinitializer
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %u256 = shufflevector <32 x i8> %c256, <32 x i8> undef, <32 x i32> zeroinitializer
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %u512 = shufflevector <64 x i8> %c512, <64 x i8> undef, <64 x i32> zeroinitializer
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 5 for instruction: %V16I8 = call <16 x i8> @llvm.fshl.v16i8(<16 x i8> %a128, <16 x i8> %a128, <16 x i8> %u128)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 5 for instruction: %V32I8 = call <32 x i8> @llvm.fshl.v32i8(<32 x i8> %a256, <32 x i8> %a256, <32 x i8> %u256)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 5 for instruction: %V64I8 = call <64 x i8> @llvm.fshl.v64i8(<64 x i8> %a512, <64 x i8> %a512, <64 x i8> %u512)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret void
 ;
   %u128 = shufflevector <16 x i8> %c128, <16 x i8> undef, <16 x i32> zeroinitializer
   %u256 = shufflevector <32 x i8> %c256, <32 x i8> undef, <32 x i32> zeroinitializer
@@ -2273,6 +2426,13 @@ define void @constant_rotate_i16(i16 %a16, <8 x i16> %a128, <16 x i16> %a256, <3
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V16I16 = call <16 x i16> @llvm.fshl.v16i16(<16 x i16> %a256, <16 x i16> %a256, <16 x i16> <i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7>)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %V32I16 = call <32 x i16> @llvm.fshl.v32i16(<32 x i16> %a512, <32 x i16> %a512, <32 x i16> <i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7>)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret void
+;
+; AVX512GFNI-LABEL: 'constant_rotate_i16'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %I16 = call i16 @llvm.fshl.i16(i16 %a16, i16 %a16, i16 7)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V8I16 = call <8 x i16> @llvm.fshl.v8i16(<8 x i16> %a128, <8 x i16> %a128, <8 x i16> <i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7>)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V16I16 = call <16 x i16> @llvm.fshl.v16i16(<16 x i16> %a256, <16 x i16> %a256, <16 x i16> <i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7>)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V32I16 = call <32 x i16> @llvm.fshl.v32i16(<32 x i16> %a512, <32 x i16> %a512, <32 x i16> <i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7>)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret void
 ;
   %I16    = call i16 @llvm.fshl.i16(i16 %a16, i16 %a16, i16 7)
   %V8I16  = call <8 x i16> @llvm.fshl.v8i16(<8 x i16> %a128, <8 x i16> %a128, <8 x i16> <i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7>)
@@ -2358,6 +2518,13 @@ define void @constant_rotate_i8(i8 %a8, <16 x i8> %a128, <32 x i8> %a256, <64 x
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V32I8 = call <32 x i8> @llvm.fshl.v32i8(<32 x i8> %a256, <32 x i8> %a256, <32 x i8> <i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0>)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %V64I8 = call <64 x i8> @llvm.fshl.v64i8(<64 x i8> %a512, <64 x i8> %a512, <64 x i8> <i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0>)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret void
+;
+; AVX512GFNI-LABEL: 'constant_rotate_i8'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %I8 = call i8 @llvm.fshl.i8(i8 %a8, i8 %a8, i8 7)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 5 for instruction: %V16I8 = call <16 x i8> @llvm.fshl.v16i8(<16 x i8> %a128, <16 x i8> %a128, <16 x i8> <i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0>)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 5 for instruction: %V32I8 = call <32 x i8> @llvm.fshl.v32i8(<32 x i8> %a256, <32 x i8> %a256, <32 x i8> <i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0>)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 5 for instruction: %V64I8 = call <64 x i8> @llvm.fshl.v64i8(<64 x i8> %a512, <64 x i8> %a512, <64 x i8> <i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0>)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret void
 ;
   %I8    = call i8 @llvm.fshl.i8(i8 %a8, i8 %a8, i8 7)
   %V16I8 = call <16 x i8> @llvm.fshl.v16i8(<16 x i8> %a128, <16 x i8> %a128, <16 x i8> <i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0>)
@@ -2554,6 +2721,13 @@ define void @splatconstant_rotate_i16(i16 %a16, <8 x i16> %a128, <16 x i16> %a25
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V16I16 = call <16 x i16> @llvm.fshl.v16i16(<16 x i16> %a256, <16 x i16> %a256, <16 x i16> <i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3>)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %V32I16 = call <32 x i16> @llvm.fshl.v32i16(<32 x i16> %a512, <32 x i16> %a512, <32 x i16> <i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3>)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret void
+;
+; AVX512GFNI-LABEL: 'splatconstant_rotate_i16'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %I16 = call i16 @llvm.fshl.i16(i16 %a16, i16 %a16, i16 3)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V8I16 = call <8 x i16> @llvm.fshl.v8i16(<8 x i16> %a128, <8 x i16> %a128, <8 x i16> <i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3>)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V16I16 = call <16 x i16> @llvm.fshl.v16i16(<16 x i16> %a256, <16 x i16> %a256, <16 x i16> <i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3>)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V32I16 = call <32 x i16> @llvm.fshl.v32i16(<32 x i16> %a512, <32 x i16> %a512, <32 x i16> <i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3>)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret void
 ;
   %I16    = call i16 @llvm.fshl.i16(i16 %a16, i16 %a16, i16 3)
   %V8I16  = call <8 x i16> @llvm.fshl.v8i16(<8 x i16> %a128, <8 x i16> %a128, <8 x i16> <i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3>)
@@ -2632,6 +2806,13 @@ define void @splatconstant_rotate_i8(i8 %a8, <16 x i8> %a128, <32 x i8> %a256, <
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V32I8 = call <32 x i8> @llvm.fshl.v32i8(<32 x i8> %a256, <32 x i8> %a256, <32 x i8> <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, 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>)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %V64I8 = call <64 x i8> @llvm.fshl.v64i8(<64 x i8> %a512, <64 x i8> %a512, <64 x i8> <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, 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, 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, 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>)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret void
+;
+; AVX512GFNI-LABEL: 'splatconstant_rotate_i8'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %I8 = call i8 @llvm.fshl.i8(i8 %a8, i8 %a8, i8 3)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 5 for instruction: %V16I8 = call <16 x i8> @llvm.fshl.v16i8(<16 x i8> %a128, <16 x i8> %a128, <16 x i8> <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>)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 5 for instruction: %V32I8 = call <32 x i8> @llvm.fshl.v32i8(<32 x i8> %a256, <32 x i8> %a256, <32 x i8> <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, 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>)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 5 for instruction: %V64I8 = call <64 x i8> @llvm.fshl.v64i8(<64 x i8> %a512, <64 x i8> %a512, <64 x i8> <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, 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, 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, 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>)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret void
 ;
   %I8    = call i8 @llvm.fshl.i8(i8 %a8, i8 %a8, i8 3)
   %V16I8 = call <16 x i8> @llvm.fshl.v16i8(<16 x i8> %a128, <16 x i8> %a128, <16 x i8> <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>)

diff  --git a/llvm/test/Analysis/CostModel/X86/fshr-codesize.ll b/llvm/test/Analysis/CostModel/X86/fshr-codesize.ll
index 0e76246fad85f0..f9d30e4ced3ec9 100644
--- a/llvm/test/Analysis/CostModel/X86/fshr-codesize.ll
+++ b/llvm/test/Analysis/CostModel/X86/fshr-codesize.ll
@@ -12,6 +12,7 @@
 ; RUN: opt < %s -passes="print<cost-model>" 2>&1 -disable-output -cost-kind=code-size -mtriple=x86_64-apple-macosx10.8.0 -mcpu=goldmont | FileCheck %s --check-prefixes=GLM
 ; RUN: opt < %s -passes="print<cost-model>" 2>&1 -disable-output -cost-kind=code-size -mtriple=x86_64-apple-macosx10.8.0 -mcpu=bdver2 | FileCheck %s --check-prefixes=XOP
 ; RUN: opt < %s -passes="print<cost-model>" 2>&1 -disable-output -cost-kind=code-size -mtriple=x86_64-apple-macosx10.8.0 -mcpu=btver2 | FileCheck %s --check-prefixes=AVX1
+; RUN: opt < %s -passes="print<cost-model>" 2>&1 -disable-output -cost-kind=code-size -mtriple=x86_64-apple-macosx10.8.0 -mcpu=tigerlake | FileCheck %s --check-prefixes=AVX512,AVX512GFNI
 
 target datalayout = "e-p:64:64:64-i1:8:8-i8:8:8-i16:16:16-i32:32:32-i64:64:64-f32:32:32-f64:64:64-v64:64:64-v128:128:128-a0:0:64-s0:64:64-f80:128:128-n8:16:32:64-S128"
 target triple = "x86_64-apple-macosx10.8.0"
@@ -97,6 +98,13 @@ define void @var_funnel_i64(i64 %a64, <2 x i64> %a128, <4 x i64> %a256, <8 x i64
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 23 for instruction: %V4I64 = call <4 x i64> @llvm.fshr.v4i64(<4 x i64> %a256, <4 x i64> %b256, <4 x i64> %c256)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 46 for instruction: %V8I64 = call <8 x i64> @llvm.fshr.v8i64(<8 x i64> %a512, <8 x i64> %b512, <8 x i64> %c512)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret void
+;
+; AVX512GFNI-LABEL: 'var_funnel_i64'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %I64 = call i64 @llvm.fshr.i64(i64 %a64, i64 %b64, i64 %c64)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V2I64 = call <2 x i64> @llvm.fshr.v2i64(<2 x i64> %a128, <2 x i64> %b128, <2 x i64> %c128)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V4I64 = call <4 x i64> @llvm.fshr.v4i64(<4 x i64> %a256, <4 x i64> %b256, <4 x i64> %c256)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V8I64 = call <8 x i64> @llvm.fshr.v8i64(<8 x i64> %a512, <8 x i64> %b512, <8 x i64> %c512)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret void
 ;
   %I64    = call i64 @llvm.fshr.i64(i64 %a64, i64 %b64, i64 %c64)
   %V2I64  = call <2 x i64> @llvm.fshr.v2i64(<2 x i64> %a128, <2 x i64> %b128, <2 x i64> %c128)
@@ -182,6 +190,13 @@ define void @var_funnel_i32(i32 %a32, <4 x i32> %a128, <8 x i32> %a256, <16 x i3
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 23 for instruction: %V4I32 = call <8 x i32> @llvm.fshr.v8i32(<8 x i32> %a256, <8 x i32> %b256, <8 x i32> %c256)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 46 for instruction: %V8I32 = call <16 x i32> @llvm.fshr.v16i32(<16 x i32> %a512, <16 x i32> %b512, <16 x i32> %c512)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret void
+;
+; AVX512GFNI-LABEL: 'var_funnel_i32'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %I32 = call i32 @llvm.fshr.i32(i32 %a32, i32 %b32, i32 %c32)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V2I32 = call <4 x i32> @llvm.fshr.v4i32(<4 x i32> %a128, <4 x i32> %b128, <4 x i32> %c128)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V4I32 = call <8 x i32> @llvm.fshr.v8i32(<8 x i32> %a256, <8 x i32> %b256, <8 x i32> %c256)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V8I32 = call <16 x i32> @llvm.fshr.v16i32(<16 x i32> %a512, <16 x i32> %b512, <16 x i32> %c512)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret void
 ;
   %I32   = call i32 @llvm.fshr.i32(i32 %a32, i32 %b32, i32 %c32)
   %V2I32 = call <4 x i32> @llvm.fshr.v4i32(<4 x i32> %a128, <4 x i32> %b128, <4 x i32> %c128)
@@ -267,6 +282,13 @@ define void @var_funnel_i16(i16 %a16, <8 x i16> %a128, <16 x i16> %a256, <32 x i
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 25 for instruction: %V16I16 = call <16 x i16> @llvm.fshr.v16i16(<16 x i16> %a256, <16 x i16> %b256, <16 x i16> %c256)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 50 for instruction: %V32I16 = call <32 x i16> @llvm.fshr.v32i16(<32 x i16> %a512, <32 x i16> %b512, <32 x i16> %c512)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret void
+;
+; AVX512GFNI-LABEL: 'var_funnel_i16'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %I16 = call i16 @llvm.fshr.i16(i16 %a16, i16 %b16, i16 %c16)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V8I16 = call <8 x i16> @llvm.fshr.v8i16(<8 x i16> %a128, <8 x i16> %b128, <8 x i16> %c128)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V16I16 = call <16 x i16> @llvm.fshr.v16i16(<16 x i16> %a256, <16 x i16> %b256, <16 x i16> %c256)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V32I16 = call <32 x i16> @llvm.fshr.v32i16(<32 x i16> %a512, <32 x i16> %b512, <32 x i16> %c512)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret void
 ;
   %I16    = call i16 @llvm.fshr.i16(i16 %a16, i16 %b16, i16 %c16)
   %V8I16  = call <8 x i16> @llvm.fshr.v8i16(<8 x i16> %a128, <8 x i16> %b128, <8 x i16> %c128)
@@ -352,6 +374,13 @@ define void @var_funnel_i8(i8 %a8, <16 x i8> %a128, <32 x i8> %a256, <64 x i8> %
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 25 for instruction: %V32I8 = call <32 x i8> @llvm.fshr.v32i8(<32 x i8> %a256, <32 x i8> %b256, <32 x i8> %c256)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 50 for instruction: %V64I8 = call <64 x i8> @llvm.fshr.v64i8(<64 x i8> %a512, <64 x i8> %b512, <64 x i8> %c512)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret void
+;
+; AVX512GFNI-LABEL: 'var_funnel_i8'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %I8 = call i8 @llvm.fshr.i8(i8 %a8, i8 %b8, i8 %c8)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 13 for instruction: %V16I8 = call <16 x i8> @llvm.fshr.v16i8(<16 x i8> %a128, <16 x i8> %b128, <16 x i8> %c128)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 28 for instruction: %V32I8 = call <32 x i8> @llvm.fshr.v32i8(<32 x i8> %a256, <32 x i8> %b256, <32 x i8> %c256)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 33 for instruction: %V64I8 = call <64 x i8> @llvm.fshr.v64i8(<64 x i8> %a512, <64 x i8> %b512, <64 x i8> %c512)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret void
 ;
   %I8    = call i8 @llvm.fshr.i8(i8 %a8, i8 %b8, i8 %c8)
   %V16I8 = call <16 x i8> @llvm.fshr.v16i8(<16 x i8> %a128, <16 x i8> %b128, <16 x i8> %c128)
@@ -463,6 +492,15 @@ define void @splatvar_funnel_i64(i64 %a64, <2 x i64> %a128, <4 x i64> %a256, <8
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 21 for instruction: %V4I64 = call <4 x i64> @llvm.fshr.v4i64(<4 x i64> %a256, <4 x i64> %b256, <4 x i64> %u256)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 42 for instruction: %V8I64 = call <8 x i64> @llvm.fshr.v8i64(<8 x i64> %a512, <8 x i64> %b512, <8 x i64> %u512)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret void
+;
+; AVX512GFNI-LABEL: 'splatvar_funnel_i64'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %u128 = shufflevector <2 x i64> %c128, <2 x i64> undef, <2 x i32> zeroinitializer
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %u256 = shufflevector <4 x i64> %c256, <4 x i64> undef, <4 x i32> zeroinitializer
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %u512 = shufflevector <8 x i64> %c512, <8 x i64> undef, <8 x i32> zeroinitializer
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V2I64 = call <2 x i64> @llvm.fshr.v2i64(<2 x i64> %a128, <2 x i64> %b128, <2 x i64> %u128)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V4I64 = call <4 x i64> @llvm.fshr.v4i64(<4 x i64> %a256, <4 x i64> %b256, <4 x i64> %u256)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V8I64 = call <8 x i64> @llvm.fshr.v8i64(<8 x i64> %a512, <8 x i64> %b512, <8 x i64> %u512)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret void
 ;
   %u128 = shufflevector <2 x i64> %c128, <2 x i64> undef, <2 x i32> zeroinitializer
   %u256 = shufflevector <4 x i64> %c256, <4 x i64> undef, <4 x i32> zeroinitializer
@@ -572,6 +610,15 @@ define void @splatvar_funnel_i32(i32 %a32, <4 x i32> %a128, <8 x i32> %a256, <16
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 23 for instruction: %V4I32 = call <8 x i32> @llvm.fshr.v8i32(<8 x i32> %a256, <8 x i32> %b256, <8 x i32> %u256)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 46 for instruction: %V8I32 = call <16 x i32> @llvm.fshr.v16i32(<16 x i32> %a512, <16 x i32> %b512, <16 x i32> %u512)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret void
+;
+; AVX512GFNI-LABEL: 'splatvar_funnel_i32'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %u128 = shufflevector <4 x i32> %c128, <4 x i32> undef, <4 x i32> zeroinitializer
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %u256 = shufflevector <8 x i32> %c256, <8 x i32> undef, <8 x i32> zeroinitializer
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %u512 = shufflevector <16 x i32> %c512, <16 x i32> undef, <16 x i32> zeroinitializer
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V2I32 = call <4 x i32> @llvm.fshr.v4i32(<4 x i32> %a128, <4 x i32> %b128, <4 x i32> %u128)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V4I32 = call <8 x i32> @llvm.fshr.v8i32(<8 x i32> %a256, <8 x i32> %b256, <8 x i32> %u256)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V8I32 = call <16 x i32> @llvm.fshr.v16i32(<16 x i32> %a512, <16 x i32> %b512, <16 x i32> %u512)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret void
 ;
   %u128 = shufflevector <4 x i32> %c128, <4 x i32> undef, <4 x i32> zeroinitializer
   %u256 = shufflevector <8 x i32> %c256, <8 x i32> undef, <8 x i32> zeroinitializer
@@ -681,6 +728,15 @@ define void @splatvar_funnel_i16(i16 %a16, <8 x i16> %a128, <16 x i16> %a256, <3
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 25 for instruction: %V16I16 = call <16 x i16> @llvm.fshr.v16i16(<16 x i16> %a256, <16 x i16> %b256, <16 x i16> %u256)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 50 for instruction: %V32I16 = call <32 x i16> @llvm.fshr.v32i16(<32 x i16> %a512, <32 x i16> %b512, <32 x i16> %u512)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret void
+;
+; AVX512GFNI-LABEL: 'splatvar_funnel_i16'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %u128 = shufflevector <8 x i16> %c128, <8 x i16> undef, <8 x i32> zeroinitializer
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %u256 = shufflevector <16 x i16> %c256, <16 x i16> undef, <16 x i32> zeroinitializer
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %u512 = shufflevector <32 x i16> %c512, <32 x i16> undef, <32 x i32> zeroinitializer
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V8I16 = call <8 x i16> @llvm.fshr.v8i16(<8 x i16> %a128, <8 x i16> %b128, <8 x i16> %u128)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V16I16 = call <16 x i16> @llvm.fshr.v16i16(<16 x i16> %a256, <16 x i16> %b256, <16 x i16> %u256)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V32I16 = call <32 x i16> @llvm.fshr.v32i16(<32 x i16> %a512, <32 x i16> %b512, <32 x i16> %u512)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret void
 ;
   %u128 = shufflevector <8 x i16> %c128, <8 x i16> undef, <8 x i32> zeroinitializer
   %u256 = shufflevector <16 x i16> %c256, <16 x i16> undef, <16 x i32> zeroinitializer
@@ -790,6 +846,15 @@ define void @splatvar_funnel_i8(i8 %a8, <16 x i8> %a128, <32 x i8> %a256, <64 x
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 25 for instruction: %V32I8 = call <32 x i8> @llvm.fshr.v32i8(<32 x i8> %a256, <32 x i8> %b256, <32 x i8> %u256)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 50 for instruction: %V64I8 = call <64 x i8> @llvm.fshr.v64i8(<64 x i8> %a512, <64 x i8> %b512, <64 x i8> %u512)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret void
+;
+; AVX512GFNI-LABEL: 'splatvar_funnel_i8'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %u128 = shufflevector <16 x i8> %c128, <16 x i8> undef, <16 x i32> zeroinitializer
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %u256 = shufflevector <32 x i8> %c256, <32 x i8> undef, <32 x i32> zeroinitializer
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %u512 = shufflevector <64 x i8> %c512, <64 x i8> undef, <64 x i32> zeroinitializer
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 15 for instruction: %V16I8 = call <16 x i8> @llvm.fshr.v16i8(<16 x i8> %a128, <16 x i8> %b128, <16 x i8> %u128)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 18 for instruction: %V32I8 = call <32 x i8> @llvm.fshr.v32i8(<32 x i8> %a256, <32 x i8> %b256, <32 x i8> %u256)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 18 for instruction: %V64I8 = call <64 x i8> @llvm.fshr.v64i8(<64 x i8> %a512, <64 x i8> %b512, <64 x i8> %u512)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret void
 ;
   %u128 = shufflevector <16 x i8> %c128, <16 x i8> undef, <16 x i32> zeroinitializer
   %u256 = shufflevector <32 x i8> %c256, <32 x i8> undef, <32 x i32> zeroinitializer
@@ -881,6 +946,13 @@ define void @constant_funnel_i64(i64 %a64, <2 x i64> %a128, <4 x i64> %a256, <8
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 22 for instruction: %V4I64 = call <4 x i64> @llvm.fshr.v4i64(<4 x i64> %a256, <4 x i64> %b256, <4 x i64> <i64 1, i64 7, i64 15, i64 31>)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 44 for instruction: %V8I64 = call <8 x i64> @llvm.fshr.v8i64(<8 x i64> %a512, <8 x i64> %b512, <8 x i64> <i64 1, i64 7, i64 15, i64 31, i64 1, i64 7, i64 15, i64 31>)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret void
+;
+; AVX512GFNI-LABEL: 'constant_funnel_i64'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %I64 = call i64 @llvm.fshr.i64(i64 %a64, i64 %b64, i64 7)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V2I64 = call <2 x i64> @llvm.fshr.v2i64(<2 x i64> %a128, <2 x i64> %b128, <2 x i64> <i64 1, i64 7>)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V4I64 = call <4 x i64> @llvm.fshr.v4i64(<4 x i64> %a256, <4 x i64> %b256, <4 x i64> <i64 1, i64 7, i64 15, i64 31>)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V8I64 = call <8 x i64> @llvm.fshr.v8i64(<8 x i64> %a512, <8 x i64> %b512, <8 x i64> <i64 1, i64 7, i64 15, i64 31, i64 1, i64 7, i64 15, i64 31>)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret void
 ;
   %I64    = call i64 @llvm.fshr.i64(i64 %a64, i64 %b64, i64 7)
   %V2I64  = call <2 x i64> @llvm.fshr.v2i64(<2 x i64> %a128, <2 x i64> %b128, <2 x i64> <i64 1, i64 7>)
@@ -966,6 +1038,13 @@ define void @constant_funnel_i32(i32 %a32, <4 x i32> %a128, <8 x i32> %a256, <16
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 22 for instruction: %V4I32 = call <8 x i32> @llvm.fshr.v8i32(<8 x i32> %a256, <8 x i32> %b256, <8 x i32> <i32 4, i32 5, i32 6, i32 7, i32 0, i32 1, i32 2, i32 3>)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 44 for instruction: %V8I32 = call <16 x i32> @llvm.fshr.v16i32(<16 x i32> %a512, <16 x i32> %b512, <16 x i32> <i32 4, i32 5, i32 6, i32 7, i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 0, i32 1, i32 2, i32 3>)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret void
+;
+; AVX512GFNI-LABEL: 'constant_funnel_i32'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %I32 = call i32 @llvm.fshr.i32(i32 %a32, i32 %b32, i32 7)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V2I32 = call <4 x i32> @llvm.fshr.v4i32(<4 x i32> %a128, <4 x i32> %b128, <4 x i32> <i32 4, i32 5, i32 6, i32 7>)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V4I32 = call <8 x i32> @llvm.fshr.v8i32(<8 x i32> %a256, <8 x i32> %b256, <8 x i32> <i32 4, i32 5, i32 6, i32 7, i32 0, i32 1, i32 2, i32 3>)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V8I32 = call <16 x i32> @llvm.fshr.v16i32(<16 x i32> %a512, <16 x i32> %b512, <16 x i32> <i32 4, i32 5, i32 6, i32 7, i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 0, i32 1, i32 2, i32 3>)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret void
 ;
   %I32   = call i32 @llvm.fshr.i32(i32 %a32, i32 %b32, i32 7)
   %V2I32 = call <4 x i32> @llvm.fshr.v4i32(<4 x i32> %a128, <4 x i32> %b128, <4 x i32> <i32 4, i32 5, i32 6, i32 7>)
@@ -1051,6 +1130,13 @@ define void @constant_funnel_i16(i16 %a16, <8 x i16> %a128, <16 x i16> %a256, <3
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 24 for instruction: %V16I16 = call <16 x i16> @llvm.fshr.v16i16(<16 x i16> %a256, <16 x i16> %b256, <16 x i16> <i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7>)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 48 for instruction: %V32I16 = call <32 x i16> @llvm.fshr.v32i16(<32 x i16> %a512, <32 x i16> %b512, <32 x i16> <i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7>)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret void
+;
+; AVX512GFNI-LABEL: 'constant_funnel_i16'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %I16 = call i16 @llvm.fshr.i16(i16 %a16, i16 %b16, i16 7)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V8I16 = call <8 x i16> @llvm.fshr.v8i16(<8 x i16> %a128, <8 x i16> %b128, <8 x i16> <i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7>)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V16I16 = call <16 x i16> @llvm.fshr.v16i16(<16 x i16> %a256, <16 x i16> %b256, <16 x i16> <i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7>)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V32I16 = call <32 x i16> @llvm.fshr.v32i16(<32 x i16> %a512, <32 x i16> %b512, <32 x i16> <i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7>)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret void
 ;
   %I16    = call i16 @llvm.fshr.i16(i16 %a16, i16 %b16, i16 7)
   %V8I16  = call <8 x i16> @llvm.fshr.v8i16(<8 x i16> %a128, <8 x i16> %b128, <8 x i16> <i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7>)
@@ -1136,6 +1222,13 @@ define void @constant_funnel_i8(i8 %a8, <16 x i8> %a128, <32 x i8> %a256, <64 x
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 24 for instruction: %V32I8 = call <32 x i8> @llvm.fshr.v32i8(<32 x i8> %a256, <32 x i8> %b256, <32 x i8> <i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0>)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 48 for instruction: %V64I8 = call <64 x i8> @llvm.fshr.v64i8(<64 x i8> %a512, <64 x i8> %b512, <64 x i8> <i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0>)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret void
+;
+; AVX512GFNI-LABEL: 'constant_funnel_i8'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %I8 = call i8 @llvm.fshr.i8(i8 %a8, i8 %b8, i8 7)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 12 for instruction: %V16I8 = call <16 x i8> @llvm.fshr.v16i8(<16 x i8> %a128, <16 x i8> %b128, <16 x i8> <i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0>)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 27 for instruction: %V32I8 = call <32 x i8> @llvm.fshr.v32i8(<32 x i8> %a256, <32 x i8> %b256, <32 x i8> <i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0>)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 32 for instruction: %V64I8 = call <64 x i8> @llvm.fshr.v64i8(<64 x i8> %a512, <64 x i8> %b512, <64 x i8> <i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0>)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret void
 ;
   %I8    = call i8 @llvm.fshr.i8(i8 %a8, i8 %b8, i8 7)
   %V16I8 = call <16 x i8> @llvm.fshr.v16i8(<16 x i8> %a128, <16 x i8> %b128, <16 x i8> <i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0>)
@@ -1225,6 +1318,13 @@ define void @splatconstant_funnel_i64(i64 %a64, <2 x i64> %a128, <4 x i64> %a256
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 20 for instruction: %V4I64 = call <4 x i64> @llvm.fshr.v4i64(<4 x i64> %a256, <4 x i64> %b256, <4 x i64> <i64 7, i64 7, i64 7, i64 7>)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 40 for instruction: %V8I64 = call <8 x i64> @llvm.fshr.v8i64(<8 x i64> %a512, <8 x i64> %b512, <8 x i64> <i64 7, i64 7, i64 7, i64 7, i64 7, i64 7, i64 7, i64 7>)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret void
+;
+; AVX512GFNI-LABEL: 'splatconstant_funnel_i64'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %I64 = call i64 @llvm.fshr.i64(i64 %a64, i64 %b64, i64 7)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V2I64 = call <2 x i64> @llvm.fshr.v2i64(<2 x i64> %a128, <2 x i64> %b128, <2 x i64> <i64 7, i64 7>)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V4I64 = call <4 x i64> @llvm.fshr.v4i64(<4 x i64> %a256, <4 x i64> %b256, <4 x i64> <i64 7, i64 7, i64 7, i64 7>)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V8I64 = call <8 x i64> @llvm.fshr.v8i64(<8 x i64> %a512, <8 x i64> %b512, <8 x i64> <i64 7, i64 7, i64 7, i64 7, i64 7, i64 7, i64 7, i64 7>)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret void
 ;
   %I64    = call i64 @llvm.fshr.i64(i64 %a64, i64 %b64, i64 7)
   %V2I64  = call <2 x i64> @llvm.fshr.v2i64(<2 x i64> %a128, <2 x i64> %b128, <2 x i64> <i64 7, i64 7>)
@@ -1310,6 +1410,13 @@ define void @splatconstant_funnel_i32(i32 %a32, <4 x i32> %a128, <8 x i32> %a256
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 20 for instruction: %V4I32 = call <8 x i32> @llvm.fshr.v8i32(<8 x i32> %a256, <8 x i32> %b256, <8 x i32> <i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5>)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 40 for instruction: %V8I32 = call <16 x i32> @llvm.fshr.v16i32(<16 x i32> %a512, <16 x i32> %b512, <16 x i32> <i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5>)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret void
+;
+; AVX512GFNI-LABEL: 'splatconstant_funnel_i32'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %I32 = call i32 @llvm.fshr.i32(i32 %a32, i32 %b32, i32 5)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V2I32 = call <4 x i32> @llvm.fshr.v4i32(<4 x i32> %a128, <4 x i32> %b128, <4 x i32> <i32 5, i32 5, i32 5, i32 5>)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V4I32 = call <8 x i32> @llvm.fshr.v8i32(<8 x i32> %a256, <8 x i32> %b256, <8 x i32> <i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5>)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V8I32 = call <16 x i32> @llvm.fshr.v16i32(<16 x i32> %a512, <16 x i32> %b512, <16 x i32> <i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5>)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret void
 ;
   %I32   = call i32 @llvm.fshr.i32(i32 %a32, i32 %b32, i32 5)
   %V2I32 = call <4 x i32> @llvm.fshr.v4i32(<4 x i32> %a128, <4 x i32> %b128, <4 x i32> <i32 5, i32 5, i32 5, i32 5>)
@@ -1395,6 +1502,13 @@ define void @splatconstant_funnel_i16(i16 %a16, <8 x i16> %a128, <16 x i16> %a25
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 22 for instruction: %V16I16 = call <16 x i16> @llvm.fshr.v16i16(<16 x i16> %a256, <16 x i16> %b256, <16 x i16> <i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3>)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 44 for instruction: %V32I16 = call <32 x i16> @llvm.fshr.v32i16(<32 x i16> %a512, <32 x i16> %b512, <32 x i16> <i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3>)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret void
+;
+; AVX512GFNI-LABEL: 'splatconstant_funnel_i16'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %I16 = call i16 @llvm.fshr.i16(i16 %a16, i16 %b16, i16 3)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V8I16 = call <8 x i16> @llvm.fshr.v8i16(<8 x i16> %a128, <8 x i16> %b128, <8 x i16> <i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3>)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V16I16 = call <16 x i16> @llvm.fshr.v16i16(<16 x i16> %a256, <16 x i16> %b256, <16 x i16> <i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3>)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V32I16 = call <32 x i16> @llvm.fshr.v32i16(<32 x i16> %a512, <32 x i16> %b512, <32 x i16> <i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3>)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret void
 ;
   %I16    = call i16 @llvm.fshr.i16(i16 %a16, i16 %b16, i16 3)
   %V8I16  = call <8 x i16> @llvm.fshr.v8i16(<8 x i16> %a128, <8 x i16> %b128, <8 x i16> <i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3>)
@@ -1480,6 +1594,13 @@ define void @splatconstant_funnel_i8(i8 %a8, <16 x i8> %a128, <32 x i8> %a256, <
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 24 for instruction: %V32I8 = call <32 x i8> @llvm.fshr.v32i8(<32 x i8> %a256, <32 x i8> %b256, <32 x i8> <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, 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>)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 48 for instruction: %V64I8 = call <64 x i8> @llvm.fshr.v64i8(<64 x i8> %a512, <64 x i8> %b512, <64 x i8> <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, 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, 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, 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>)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret void
+;
+; AVX512GFNI-LABEL: 'splatconstant_funnel_i8'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %I8 = call i8 @llvm.fshr.i8(i8 %a8, i8 %b8, i8 3)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %V16I8 = call <16 x i8> @llvm.fshr.v16i8(<16 x i8> %a128, <16 x i8> %b128, <16 x i8> <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>)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %V32I8 = call <32 x i8> @llvm.fshr.v32i8(<32 x i8> %a256, <32 x i8> %b256, <32 x i8> <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, 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>)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %V64I8 = call <64 x i8> @llvm.fshr.v64i8(<64 x i8> %a512, <64 x i8> %b512, <64 x i8> <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, 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, 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, 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>)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret void
 ;
   %I8    = call i8 @llvm.fshr.i8(i8 %a8, i8 %b8, i8 3)
   %V16I8 = call <16 x i8> @llvm.fshr.v16i8(<16 x i8> %a128, <16 x i8> %b128, <16 x i8> <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>)
@@ -1690,6 +1811,13 @@ define void @var_rotate_i16(i16 %a16, <8 x i16> %a128, <16 x i16> %a256, <32 x i
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %V16I16 = call <16 x i16> @llvm.fshr.v16i16(<16 x i16> %a256, <16 x i16> %a256, <16 x i16> %c256)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 16 for instruction: %V32I16 = call <32 x i16> @llvm.fshr.v32i16(<32 x i16> %a512, <32 x i16> %a512, <32 x i16> %c512)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret void
+;
+; AVX512GFNI-LABEL: 'var_rotate_i16'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %I16 = call i16 @llvm.fshr.i16(i16 %a16, i16 %a16, i16 %c16)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V8I16 = call <8 x i16> @llvm.fshr.v8i16(<8 x i16> %a128, <8 x i16> %a128, <8 x i16> %c128)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V16I16 = call <16 x i16> @llvm.fshr.v16i16(<16 x i16> %a256, <16 x i16> %a256, <16 x i16> %c256)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V32I16 = call <32 x i16> @llvm.fshr.v32i16(<32 x i16> %a512, <32 x i16> %a512, <32 x i16> %c512)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret void
 ;
   %I16    = call i16 @llvm.fshr.i16(i16 %a16, i16 %a16, i16 %c16)
   %V8I16  = call <8 x i16> @llvm.fshr.v8i16(<8 x i16> %a128, <8 x i16> %a128, <8 x i16> %c128)
@@ -1775,6 +1903,13 @@ define void @var_rotate_i8(i8 %a8, <16 x i8> %a128, <32 x i8> %a256, <64 x i8> %
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %V32I8 = call <32 x i8> @llvm.fshr.v32i8(<32 x i8> %a256, <32 x i8> %a256, <32 x i8> %c256)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 16 for instruction: %V64I8 = call <64 x i8> @llvm.fshr.v64i8(<64 x i8> %a512, <64 x i8> %a512, <64 x i8> %c512)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret void
+;
+; AVX512GFNI-LABEL: 'var_rotate_i8'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %I8 = call i8 @llvm.fshr.i8(i8 %a8, i8 %a8, i8 %c8)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 6 for instruction: %V16I8 = call <16 x i8> @llvm.fshr.v16i8(<16 x i8> %a128, <16 x i8> %a128, <16 x i8> %c128)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 6 for instruction: %V32I8 = call <32 x i8> @llvm.fshr.v32i8(<32 x i8> %a256, <32 x i8> %a256, <32 x i8> %c256)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 12 for instruction: %V64I8 = call <64 x i8> @llvm.fshr.v64i8(<64 x i8> %a512, <64 x i8> %a512, <64 x i8> %c512)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret void
 ;
   %I8    = call i8 @llvm.fshr.i8(i8 %a8, i8 %a8, i8 %c8)
   %V16I8 = call <16 x i8> @llvm.fshr.v16i8(<16 x i8> %a128, <16 x i8> %a128, <16 x i8> %c128)
@@ -2023,6 +2158,15 @@ define void @splatvar_rotate_i16(i16 %a16, <8 x i16> %a128, <16 x i16> %a256, <3
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %V16I16 = call <16 x i16> @llvm.fshr.v16i16(<16 x i16> %a256, <16 x i16> %a256, <16 x i16> %u256)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 16 for instruction: %V32I16 = call <32 x i16> @llvm.fshr.v32i16(<32 x i16> %a512, <32 x i16> %a512, <32 x i16> %u512)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret void
+;
+; AVX512GFNI-LABEL: 'splatvar_rotate_i16'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %u128 = shufflevector <8 x i16> %c128, <8 x i16> undef, <8 x i32> zeroinitializer
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %u256 = shufflevector <16 x i16> %c256, <16 x i16> undef, <16 x i32> zeroinitializer
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %u512 = shufflevector <32 x i16> %c512, <32 x i16> undef, <32 x i32> zeroinitializer
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V8I16 = call <8 x i16> @llvm.fshr.v8i16(<8 x i16> %a128, <8 x i16> %a128, <8 x i16> %u128)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V16I16 = call <16 x i16> @llvm.fshr.v16i16(<16 x i16> %a256, <16 x i16> %a256, <16 x i16> %u256)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V32I16 = call <32 x i16> @llvm.fshr.v32i16(<32 x i16> %a512, <32 x i16> %a512, <32 x i16> %u512)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret void
 ;
   %u128 = shufflevector <8 x i16> %c128, <8 x i16> undef, <8 x i32> zeroinitializer
   %u256 = shufflevector <16 x i16> %c256, <16 x i16> undef, <16 x i32> zeroinitializer
@@ -2123,6 +2267,15 @@ define void @splatvar_rotate_i8(i8 %a8, <16 x i8> %a128, <32 x i8> %a256, <64 x
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %V32I8 = call <32 x i8> @llvm.fshr.v32i8(<32 x i8> %a256, <32 x i8> %a256, <32 x i8> %u256)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 16 for instruction: %V64I8 = call <64 x i8> @llvm.fshr.v64i8(<64 x i8> %a512, <64 x i8> %a512, <64 x i8> %u512)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret void
+;
+; AVX512GFNI-LABEL: 'splatvar_rotate_i8'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %u128 = shufflevector <16 x i8> %c128, <16 x i8> undef, <16 x i32> zeroinitializer
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %u256 = shufflevector <32 x i8> %c256, <32 x i8> undef, <32 x i32> zeroinitializer
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %u512 = shufflevector <64 x i8> %c512, <64 x i8> undef, <64 x i32> zeroinitializer
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 6 for instruction: %V16I8 = call <16 x i8> @llvm.fshr.v16i8(<16 x i8> %a128, <16 x i8> %a128, <16 x i8> %u128)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 6 for instruction: %V32I8 = call <32 x i8> @llvm.fshr.v32i8(<32 x i8> %a256, <32 x i8> %a256, <32 x i8> %u256)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 12 for instruction: %V64I8 = call <64 x i8> @llvm.fshr.v64i8(<64 x i8> %a512, <64 x i8> %a512, <64 x i8> %u512)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret void
 ;
   %u128 = shufflevector <16 x i8> %c128, <16 x i8> undef, <16 x i32> zeroinitializer
   %u256 = shufflevector <32 x i8> %c256, <32 x i8> undef, <32 x i32> zeroinitializer
@@ -2335,6 +2488,13 @@ define void @constant_rotate_i16(i16 %a16, <8 x i16> %a128, <16 x i16> %a256, <3
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %V16I16 = call <16 x i16> @llvm.fshr.v16i16(<16 x i16> %a256, <16 x i16> %a256, <16 x i16> <i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7>)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 16 for instruction: %V32I16 = call <32 x i16> @llvm.fshr.v32i16(<32 x i16> %a512, <32 x i16> %a512, <32 x i16> <i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7>)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret void
+;
+; AVX512GFNI-LABEL: 'constant_rotate_i16'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %I16 = call i16 @llvm.fshr.i16(i16 %a16, i16 %a16, i16 7)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V8I16 = call <8 x i16> @llvm.fshr.v8i16(<8 x i16> %a128, <8 x i16> %a128, <8 x i16> <i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7>)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V16I16 = call <16 x i16> @llvm.fshr.v16i16(<16 x i16> %a256, <16 x i16> %a256, <16 x i16> <i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7>)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V32I16 = call <32 x i16> @llvm.fshr.v32i16(<32 x i16> %a512, <32 x i16> %a512, <32 x i16> <i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7>)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret void
 ;
   %I16    = call i16 @llvm.fshr.i16(i16 %a16, i16 %a16, i16 7)
   %V8I16  = call <8 x i16> @llvm.fshr.v8i16(<8 x i16> %a128, <8 x i16> %a128, <8 x i16> <i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7>)
@@ -2420,6 +2580,13 @@ define void @constant_rotate_i8(i8 %a8, <16 x i8> %a128, <32 x i8> %a256, <64 x
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %V32I8 = call <32 x i8> @llvm.fshr.v32i8(<32 x i8> %a256, <32 x i8> %a256, <32 x i8> <i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0>)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 16 for instruction: %V64I8 = call <64 x i8> @llvm.fshr.v64i8(<64 x i8> %a512, <64 x i8> %a512, <64 x i8> <i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0>)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret void
+;
+; AVX512GFNI-LABEL: 'constant_rotate_i8'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %I8 = call i8 @llvm.fshr.i8(i8 %a8, i8 %a8, i8 7)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 6 for instruction: %V16I8 = call <16 x i8> @llvm.fshr.v16i8(<16 x i8> %a128, <16 x i8> %a128, <16 x i8> <i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0>)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 6 for instruction: %V32I8 = call <32 x i8> @llvm.fshr.v32i8(<32 x i8> %a256, <32 x i8> %a256, <32 x i8> <i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0>)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 12 for instruction: %V64I8 = call <64 x i8> @llvm.fshr.v64i8(<64 x i8> %a512, <64 x i8> %a512, <64 x i8> <i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0>)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret void
 ;
   %I8    = call i8 @llvm.fshr.i8(i8 %a8, i8 %a8, i8 7)
   %V16I8 = call <16 x i8> @llvm.fshr.v16i8(<16 x i8> %a128, <16 x i8> %a128, <16 x i8> <i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0>)
@@ -2616,6 +2783,13 @@ define void @splatconstant_rotate_i16(i16 %a16, <8 x i16> %a128, <16 x i16> %a25
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %V16I16 = call <16 x i16> @llvm.fshr.v16i16(<16 x i16> %a256, <16 x i16> %a256, <16 x i16> <i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3>)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 16 for instruction: %V32I16 = call <32 x i16> @llvm.fshr.v32i16(<32 x i16> %a512, <32 x i16> %a512, <32 x i16> <i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3>)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret void
+;
+; AVX512GFNI-LABEL: 'splatconstant_rotate_i16'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %I16 = call i16 @llvm.fshr.i16(i16 %a16, i16 %a16, i16 3)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V8I16 = call <8 x i16> @llvm.fshr.v8i16(<8 x i16> %a128, <8 x i16> %a128, <8 x i16> <i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3>)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V16I16 = call <16 x i16> @llvm.fshr.v16i16(<16 x i16> %a256, <16 x i16> %a256, <16 x i16> <i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3>)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V32I16 = call <32 x i16> @llvm.fshr.v32i16(<32 x i16> %a512, <32 x i16> %a512, <32 x i16> <i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3>)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret void
 ;
   %I16    = call i16 @llvm.fshr.i16(i16 %a16, i16 %a16, i16 3)
   %V8I16  = call <8 x i16> @llvm.fshr.v8i16(<8 x i16> %a128, <8 x i16> %a128, <8 x i16> <i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3>)
@@ -2694,6 +2868,13 @@ define void @splatconstant_rotate_i8(i8 %a8, <16 x i8> %a128, <32 x i8> %a256, <
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %V32I8 = call <32 x i8> @llvm.fshr.v32i8(<32 x i8> %a256, <32 x i8> %a256, <32 x i8> <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, 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>)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 16 for instruction: %V64I8 = call <64 x i8> @llvm.fshr.v64i8(<64 x i8> %a512, <64 x i8> %a512, <64 x i8> <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, 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, 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, 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>)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret void
+;
+; AVX512GFNI-LABEL: 'splatconstant_rotate_i8'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %I8 = call i8 @llvm.fshr.i8(i8 %a8, i8 %a8, i8 3)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 6 for instruction: %V16I8 = call <16 x i8> @llvm.fshr.v16i8(<16 x i8> %a128, <16 x i8> %a128, <16 x i8> <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>)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 6 for instruction: %V32I8 = call <32 x i8> @llvm.fshr.v32i8(<32 x i8> %a256, <32 x i8> %a256, <32 x i8> <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, 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>)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 12 for instruction: %V64I8 = call <64 x i8> @llvm.fshr.v64i8(<64 x i8> %a512, <64 x i8> %a512, <64 x i8> <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, 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, 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, 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>)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret void
 ;
   %I8    = call i8 @llvm.fshr.i8(i8 %a8, i8 %a8, i8 3)
   %V16I8 = call <16 x i8> @llvm.fshr.v16i8(<16 x i8> %a128, <16 x i8> %a128, <16 x i8> <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>)

diff  --git a/llvm/test/Analysis/CostModel/X86/fshr-latency.ll b/llvm/test/Analysis/CostModel/X86/fshr-latency.ll
index 73e32dc92a69b7..ed2227591847a6 100644
--- a/llvm/test/Analysis/CostModel/X86/fshr-latency.ll
+++ b/llvm/test/Analysis/CostModel/X86/fshr-latency.ll
@@ -12,6 +12,7 @@
 ; RUN: opt < %s -passes="print<cost-model>" 2>&1 -disable-output -cost-kind=latency -mtriple=x86_64-apple-macosx10.8.0 -mcpu=goldmont | FileCheck %s --check-prefixes=GLM
 ; RUN: opt < %s -passes="print<cost-model>" 2>&1 -disable-output -cost-kind=latency -mtriple=x86_64-apple-macosx10.8.0 -mcpu=bdver2 | FileCheck %s --check-prefixes=XOP
 ; RUN: opt < %s -passes="print<cost-model>" 2>&1 -disable-output -cost-kind=latency -mtriple=x86_64-apple-macosx10.8.0 -mcpu=btver2 | FileCheck %s --check-prefixes=AVX1
+; RUN: opt < %s -passes="print<cost-model>" 2>&1 -disable-output -cost-kind=latency -mtriple=x86_64-apple-macosx10.8.0 -mcpu=tigerlake | FileCheck %s --check-prefixes=AVX512,AVX512GFNI
 
 target datalayout = "e-p:64:64:64-i1:8:8-i8:8:8-i16:16:16-i32:32:32-i64:64:64-f32:32:32-f64:64:64-v64:64:64-v128:128:128-a0:0:64-s0:64:64-f80:128:128-n8:16:32:64-S128"
 target triple = "x86_64-apple-macosx10.8.0"
@@ -97,6 +98,13 @@ define void @var_funnel_i64(i64 %a64, <2 x i64> %a128, <4 x i64> %a256, <8 x i64
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 23 for instruction: %V4I64 = call <4 x i64> @llvm.fshr.v4i64(<4 x i64> %a256, <4 x i64> %b256, <4 x i64> %c256)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 46 for instruction: %V8I64 = call <8 x i64> @llvm.fshr.v8i64(<8 x i64> %a512, <8 x i64> %b512, <8 x i64> %c512)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret void
+;
+; AVX512GFNI-LABEL: 'var_funnel_i64'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %I64 = call i64 @llvm.fshr.i64(i64 %a64, i64 %b64, i64 %c64)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V2I64 = call <2 x i64> @llvm.fshr.v2i64(<2 x i64> %a128, <2 x i64> %b128, <2 x i64> %c128)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V4I64 = call <4 x i64> @llvm.fshr.v4i64(<4 x i64> %a256, <4 x i64> %b256, <4 x i64> %c256)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V8I64 = call <8 x i64> @llvm.fshr.v8i64(<8 x i64> %a512, <8 x i64> %b512, <8 x i64> %c512)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret void
 ;
   %I64    = call i64 @llvm.fshr.i64(i64 %a64, i64 %b64, i64 %c64)
   %V2I64  = call <2 x i64> @llvm.fshr.v2i64(<2 x i64> %a128, <2 x i64> %b128, <2 x i64> %c128)
@@ -182,6 +190,13 @@ define void @var_funnel_i32(i32 %a32, <4 x i32> %a128, <8 x i32> %a256, <16 x i3
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 23 for instruction: %V4I32 = call <8 x i32> @llvm.fshr.v8i32(<8 x i32> %a256, <8 x i32> %b256, <8 x i32> %c256)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 46 for instruction: %V8I32 = call <16 x i32> @llvm.fshr.v16i32(<16 x i32> %a512, <16 x i32> %b512, <16 x i32> %c512)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret void
+;
+; AVX512GFNI-LABEL: 'var_funnel_i32'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %I32 = call i32 @llvm.fshr.i32(i32 %a32, i32 %b32, i32 %c32)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V2I32 = call <4 x i32> @llvm.fshr.v4i32(<4 x i32> %a128, <4 x i32> %b128, <4 x i32> %c128)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V4I32 = call <8 x i32> @llvm.fshr.v8i32(<8 x i32> %a256, <8 x i32> %b256, <8 x i32> %c256)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V8I32 = call <16 x i32> @llvm.fshr.v16i32(<16 x i32> %a512, <16 x i32> %b512, <16 x i32> %c512)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret void
 ;
   %I32   = call i32 @llvm.fshr.i32(i32 %a32, i32 %b32, i32 %c32)
   %V2I32 = call <4 x i32> @llvm.fshr.v4i32(<4 x i32> %a128, <4 x i32> %b128, <4 x i32> %c128)
@@ -267,6 +282,13 @@ define void @var_funnel_i16(i16 %a16, <8 x i16> %a128, <16 x i16> %a256, <32 x i
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 23 for instruction: %V16I16 = call <16 x i16> @llvm.fshr.v16i16(<16 x i16> %a256, <16 x i16> %b256, <16 x i16> %c256)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 46 for instruction: %V32I16 = call <32 x i16> @llvm.fshr.v32i16(<32 x i16> %a512, <32 x i16> %b512, <32 x i16> %c512)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret void
+;
+; AVX512GFNI-LABEL: 'var_funnel_i16'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %I16 = call i16 @llvm.fshr.i16(i16 %a16, i16 %b16, i16 %c16)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V8I16 = call <8 x i16> @llvm.fshr.v8i16(<8 x i16> %a128, <8 x i16> %b128, <8 x i16> %c128)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V16I16 = call <16 x i16> @llvm.fshr.v16i16(<16 x i16> %a256, <16 x i16> %b256, <16 x i16> %c256)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V32I16 = call <32 x i16> @llvm.fshr.v32i16(<32 x i16> %a512, <32 x i16> %b512, <32 x i16> %c512)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret void
 ;
   %I16    = call i16 @llvm.fshr.i16(i16 %a16, i16 %b16, i16 %c16)
   %V8I16  = call <8 x i16> @llvm.fshr.v8i16(<8 x i16> %a128, <8 x i16> %b128, <8 x i16> %c128)
@@ -352,6 +374,13 @@ define void @var_funnel_i8(i8 %a8, <16 x i8> %a128, <32 x i8> %a256, <64 x i8> %
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 23 for instruction: %V32I8 = call <32 x i8> @llvm.fshr.v32i8(<32 x i8> %a256, <32 x i8> %b256, <32 x i8> %c256)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 46 for instruction: %V64I8 = call <64 x i8> @llvm.fshr.v64i8(<64 x i8> %a512, <64 x i8> %b512, <64 x i8> %c512)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret void
+;
+; AVX512GFNI-LABEL: 'var_funnel_i8'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %I8 = call i8 @llvm.fshr.i8(i8 %a8, i8 %b8, i8 %c8)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 21 for instruction: %V16I8 = call <16 x i8> @llvm.fshr.v16i8(<16 x i8> %a128, <16 x i8> %b128, <16 x i8> %c128)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 58 for instruction: %V32I8 = call <32 x i8> @llvm.fshr.v32i8(<32 x i8> %a256, <32 x i8> %b256, <32 x i8> %c256)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 51 for instruction: %V64I8 = call <64 x i8> @llvm.fshr.v64i8(<64 x i8> %a512, <64 x i8> %b512, <64 x i8> %c512)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret void
 ;
   %I8    = call i8 @llvm.fshr.i8(i8 %a8, i8 %b8, i8 %c8)
   %V16I8 = call <16 x i8> @llvm.fshr.v16i8(<16 x i8> %a128, <16 x i8> %b128, <16 x i8> %c128)
@@ -463,6 +492,15 @@ define void @splatvar_funnel_i64(i64 %a64, <2 x i64> %a128, <4 x i64> %a256, <8
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 23 for instruction: %V4I64 = call <4 x i64> @llvm.fshr.v4i64(<4 x i64> %a256, <4 x i64> %b256, <4 x i64> %u256)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 46 for instruction: %V8I64 = call <8 x i64> @llvm.fshr.v8i64(<8 x i64> %a512, <8 x i64> %b512, <8 x i64> %u512)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret void
+;
+; AVX512GFNI-LABEL: 'splatvar_funnel_i64'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %u128 = shufflevector <2 x i64> %c128, <2 x i64> undef, <2 x i32> zeroinitializer
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %u256 = shufflevector <4 x i64> %c256, <4 x i64> undef, <4 x i32> zeroinitializer
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %u512 = shufflevector <8 x i64> %c512, <8 x i64> undef, <8 x i32> zeroinitializer
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V2I64 = call <2 x i64> @llvm.fshr.v2i64(<2 x i64> %a128, <2 x i64> %b128, <2 x i64> %u128)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V4I64 = call <4 x i64> @llvm.fshr.v4i64(<4 x i64> %a256, <4 x i64> %b256, <4 x i64> %u256)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V8I64 = call <8 x i64> @llvm.fshr.v8i64(<8 x i64> %a512, <8 x i64> %b512, <8 x i64> %u512)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret void
 ;
   %u128 = shufflevector <2 x i64> %c128, <2 x i64> undef, <2 x i32> zeroinitializer
   %u256 = shufflevector <4 x i64> %c256, <4 x i64> undef, <4 x i32> zeroinitializer
@@ -563,6 +601,15 @@ define void @splatvar_funnel_i32(i32 %a32, <4 x i32> %a128, <8 x i32> %a256, <16
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 23 for instruction: %V4I32 = call <8 x i32> @llvm.fshr.v8i32(<8 x i32> %a256, <8 x i32> %b256, <8 x i32> %u256)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 46 for instruction: %V8I32 = call <16 x i32> @llvm.fshr.v16i32(<16 x i32> %a512, <16 x i32> %b512, <16 x i32> %u512)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret void
+;
+; AVX512GFNI-LABEL: 'splatvar_funnel_i32'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %u128 = shufflevector <4 x i32> %c128, <4 x i32> undef, <4 x i32> zeroinitializer
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %u256 = shufflevector <8 x i32> %c256, <8 x i32> undef, <8 x i32> zeroinitializer
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %u512 = shufflevector <16 x i32> %c512, <16 x i32> undef, <16 x i32> zeroinitializer
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V2I32 = call <4 x i32> @llvm.fshr.v4i32(<4 x i32> %a128, <4 x i32> %b128, <4 x i32> %u128)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V4I32 = call <8 x i32> @llvm.fshr.v8i32(<8 x i32> %a256, <8 x i32> %b256, <8 x i32> %u256)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V8I32 = call <16 x i32> @llvm.fshr.v16i32(<16 x i32> %a512, <16 x i32> %b512, <16 x i32> %u512)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret void
 ;
   %u128 = shufflevector <4 x i32> %c128, <4 x i32> undef, <4 x i32> zeroinitializer
   %u256 = shufflevector <8 x i32> %c256, <8 x i32> undef, <8 x i32> zeroinitializer
@@ -663,6 +710,15 @@ define void @splatvar_funnel_i16(i16 %a16, <8 x i16> %a128, <16 x i16> %a256, <3
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 23 for instruction: %V16I16 = call <16 x i16> @llvm.fshr.v16i16(<16 x i16> %a256, <16 x i16> %b256, <16 x i16> %u256)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 46 for instruction: %V32I16 = call <32 x i16> @llvm.fshr.v32i16(<32 x i16> %a512, <32 x i16> %b512, <32 x i16> %u512)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret void
+;
+; AVX512GFNI-LABEL: 'splatvar_funnel_i16'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %u128 = shufflevector <8 x i16> %c128, <8 x i16> undef, <8 x i32> zeroinitializer
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %u256 = shufflevector <16 x i16> %c256, <16 x i16> undef, <16 x i32> zeroinitializer
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %u512 = shufflevector <32 x i16> %c512, <32 x i16> undef, <32 x i32> zeroinitializer
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V8I16 = call <8 x i16> @llvm.fshr.v8i16(<8 x i16> %a128, <8 x i16> %b128, <8 x i16> %u128)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V16I16 = call <16 x i16> @llvm.fshr.v16i16(<16 x i16> %a256, <16 x i16> %b256, <16 x i16> %u256)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V32I16 = call <32 x i16> @llvm.fshr.v32i16(<32 x i16> %a512, <32 x i16> %b512, <32 x i16> %u512)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret void
 ;
   %u128 = shufflevector <8 x i16> %c128, <8 x i16> undef, <8 x i32> zeroinitializer
   %u256 = shufflevector <16 x i16> %c256, <16 x i16> undef, <16 x i32> zeroinitializer
@@ -763,6 +819,15 @@ define void @splatvar_funnel_i8(i8 %a8, <16 x i8> %a128, <32 x i8> %a256, <64 x
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 23 for instruction: %V32I8 = call <32 x i8> @llvm.fshr.v32i8(<32 x i8> %a256, <32 x i8> %b256, <32 x i8> %u256)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 46 for instruction: %V64I8 = call <64 x i8> @llvm.fshr.v64i8(<64 x i8> %a512, <64 x i8> %b512, <64 x i8> %u512)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret void
+;
+; AVX512GFNI-LABEL: 'splatvar_funnel_i8'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %u128 = shufflevector <16 x i8> %c128, <16 x i8> undef, <16 x i32> zeroinitializer
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %u256 = shufflevector <32 x i8> %c256, <32 x i8> undef, <32 x i32> zeroinitializer
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %u512 = shufflevector <64 x i8> %c512, <64 x i8> undef, <64 x i32> zeroinitializer
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 20 for instruction: %V16I8 = call <16 x i8> @llvm.fshr.v16i8(<16 x i8> %a128, <16 x i8> %b128, <16 x i8> %u128)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 20 for instruction: %V32I8 = call <32 x i8> @llvm.fshr.v32i8(<32 x i8> %a256, <32 x i8> %b256, <32 x i8> %u256)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 20 for instruction: %V64I8 = call <64 x i8> @llvm.fshr.v64i8(<64 x i8> %a512, <64 x i8> %b512, <64 x i8> %u512)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret void
 ;
   %u128 = shufflevector <16 x i8> %c128, <16 x i8> undef, <16 x i32> zeroinitializer
   %u256 = shufflevector <32 x i8> %c256, <32 x i8> undef, <32 x i32> zeroinitializer
@@ -854,6 +919,13 @@ define void @constant_funnel_i64(i64 %a64, <2 x i64> %a128, <4 x i64> %a256, <8
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 22 for instruction: %V4I64 = call <4 x i64> @llvm.fshr.v4i64(<4 x i64> %a256, <4 x i64> %b256, <4 x i64> <i64 1, i64 7, i64 15, i64 31>)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 44 for instruction: %V8I64 = call <8 x i64> @llvm.fshr.v8i64(<8 x i64> %a512, <8 x i64> %b512, <8 x i64> <i64 1, i64 7, i64 15, i64 31, i64 1, i64 7, i64 15, i64 31>)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret void
+;
+; AVX512GFNI-LABEL: 'constant_funnel_i64'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %I64 = call i64 @llvm.fshr.i64(i64 %a64, i64 %b64, i64 7)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V2I64 = call <2 x i64> @llvm.fshr.v2i64(<2 x i64> %a128, <2 x i64> %b128, <2 x i64> <i64 1, i64 7>)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V4I64 = call <4 x i64> @llvm.fshr.v4i64(<4 x i64> %a256, <4 x i64> %b256, <4 x i64> <i64 1, i64 7, i64 15, i64 31>)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V8I64 = call <8 x i64> @llvm.fshr.v8i64(<8 x i64> %a512, <8 x i64> %b512, <8 x i64> <i64 1, i64 7, i64 15, i64 31, i64 1, i64 7, i64 15, i64 31>)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret void
 ;
   %I64    = call i64 @llvm.fshr.i64(i64 %a64, i64 %b64, i64 7)
   %V2I64  = call <2 x i64> @llvm.fshr.v2i64(<2 x i64> %a128, <2 x i64> %b128, <2 x i64> <i64 1, i64 7>)
@@ -939,6 +1011,13 @@ define void @constant_funnel_i32(i32 %a32, <4 x i32> %a128, <8 x i32> %a256, <16
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 22 for instruction: %V4I32 = call <8 x i32> @llvm.fshr.v8i32(<8 x i32> %a256, <8 x i32> %b256, <8 x i32> <i32 4, i32 5, i32 6, i32 7, i32 0, i32 1, i32 2, i32 3>)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 44 for instruction: %V8I32 = call <16 x i32> @llvm.fshr.v16i32(<16 x i32> %a512, <16 x i32> %b512, <16 x i32> <i32 4, i32 5, i32 6, i32 7, i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 0, i32 1, i32 2, i32 3>)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret void
+;
+; AVX512GFNI-LABEL: 'constant_funnel_i32'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %I32 = call i32 @llvm.fshr.i32(i32 %a32, i32 %b32, i32 7)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V2I32 = call <4 x i32> @llvm.fshr.v4i32(<4 x i32> %a128, <4 x i32> %b128, <4 x i32> <i32 4, i32 5, i32 6, i32 7>)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V4I32 = call <8 x i32> @llvm.fshr.v8i32(<8 x i32> %a256, <8 x i32> %b256, <8 x i32> <i32 4, i32 5, i32 6, i32 7, i32 0, i32 1, i32 2, i32 3>)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V8I32 = call <16 x i32> @llvm.fshr.v16i32(<16 x i32> %a512, <16 x i32> %b512, <16 x i32> <i32 4, i32 5, i32 6, i32 7, i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 0, i32 1, i32 2, i32 3>)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret void
 ;
   %I32   = call i32 @llvm.fshr.i32(i32 %a32, i32 %b32, i32 7)
   %V2I32 = call <4 x i32> @llvm.fshr.v4i32(<4 x i32> %a128, <4 x i32> %b128, <4 x i32> <i32 4, i32 5, i32 6, i32 7>)
@@ -1024,6 +1103,13 @@ define void @constant_funnel_i16(i16 %a16, <8 x i16> %a128, <16 x i16> %a256, <3
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 22 for instruction: %V16I16 = call <16 x i16> @llvm.fshr.v16i16(<16 x i16> %a256, <16 x i16> %b256, <16 x i16> <i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7>)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 44 for instruction: %V32I16 = call <32 x i16> @llvm.fshr.v32i16(<32 x i16> %a512, <32 x i16> %b512, <32 x i16> <i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7>)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret void
+;
+; AVX512GFNI-LABEL: 'constant_funnel_i16'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %I16 = call i16 @llvm.fshr.i16(i16 %a16, i16 %b16, i16 7)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V8I16 = call <8 x i16> @llvm.fshr.v8i16(<8 x i16> %a128, <8 x i16> %b128, <8 x i16> <i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7>)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V16I16 = call <16 x i16> @llvm.fshr.v16i16(<16 x i16> %a256, <16 x i16> %b256, <16 x i16> <i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7>)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V32I16 = call <32 x i16> @llvm.fshr.v32i16(<32 x i16> %a512, <32 x i16> %b512, <32 x i16> <i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7>)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret void
 ;
   %I16    = call i16 @llvm.fshr.i16(i16 %a16, i16 %b16, i16 7)
   %V8I16  = call <8 x i16> @llvm.fshr.v8i16(<8 x i16> %a128, <8 x i16> %b128, <8 x i16> <i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7>)
@@ -1109,6 +1195,13 @@ define void @constant_funnel_i8(i8 %a8, <16 x i8> %a128, <32 x i8> %a256, <64 x
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 22 for instruction: %V32I8 = call <32 x i8> @llvm.fshr.v32i8(<32 x i8> %a256, <32 x i8> %b256, <32 x i8> <i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0>)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 44 for instruction: %V64I8 = call <64 x i8> @llvm.fshr.v64i8(<64 x i8> %a512, <64 x i8> %b512, <64 x i8> <i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0>)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret void
+;
+; AVX512GFNI-LABEL: 'constant_funnel_i8'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %I8 = call i8 @llvm.fshr.i8(i8 %a8, i8 %b8, i8 7)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 20 for instruction: %V16I8 = call <16 x i8> @llvm.fshr.v16i8(<16 x i8> %a128, <16 x i8> %b128, <16 x i8> <i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0>)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 57 for instruction: %V32I8 = call <32 x i8> @llvm.fshr.v32i8(<32 x i8> %a256, <32 x i8> %b256, <32 x i8> <i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0>)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 50 for instruction: %V64I8 = call <64 x i8> @llvm.fshr.v64i8(<64 x i8> %a512, <64 x i8> %b512, <64 x i8> <i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0>)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret void
 ;
   %I8    = call i8 @llvm.fshr.i8(i8 %a8, i8 %b8, i8 7)
   %V16I8 = call <16 x i8> @llvm.fshr.v16i8(<16 x i8> %a128, <16 x i8> %b128, <16 x i8> <i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0>)
@@ -1198,6 +1291,13 @@ define void @splatconstant_funnel_i64(i64 %a64, <2 x i64> %a128, <4 x i64> %a256
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 20 for instruction: %V4I64 = call <4 x i64> @llvm.fshr.v4i64(<4 x i64> %a256, <4 x i64> %b256, <4 x i64> <i64 7, i64 7, i64 7, i64 7>)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 40 for instruction: %V8I64 = call <8 x i64> @llvm.fshr.v8i64(<8 x i64> %a512, <8 x i64> %b512, <8 x i64> <i64 7, i64 7, i64 7, i64 7, i64 7, i64 7, i64 7, i64 7>)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret void
+;
+; AVX512GFNI-LABEL: 'splatconstant_funnel_i64'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %I64 = call i64 @llvm.fshr.i64(i64 %a64, i64 %b64, i64 7)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V2I64 = call <2 x i64> @llvm.fshr.v2i64(<2 x i64> %a128, <2 x i64> %b128, <2 x i64> <i64 7, i64 7>)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V4I64 = call <4 x i64> @llvm.fshr.v4i64(<4 x i64> %a256, <4 x i64> %b256, <4 x i64> <i64 7, i64 7, i64 7, i64 7>)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V8I64 = call <8 x i64> @llvm.fshr.v8i64(<8 x i64> %a512, <8 x i64> %b512, <8 x i64> <i64 7, i64 7, i64 7, i64 7, i64 7, i64 7, i64 7, i64 7>)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret void
 ;
   %I64    = call i64 @llvm.fshr.i64(i64 %a64, i64 %b64, i64 7)
   %V2I64  = call <2 x i64> @llvm.fshr.v2i64(<2 x i64> %a128, <2 x i64> %b128, <2 x i64> <i64 7, i64 7>)
@@ -1276,6 +1376,13 @@ define void @splatconstant_funnel_i32(i32 %a32, <4 x i32> %a128, <8 x i32> %a256
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 20 for instruction: %V4I32 = call <8 x i32> @llvm.fshr.v8i32(<8 x i32> %a256, <8 x i32> %b256, <8 x i32> <i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5>)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 40 for instruction: %V8I32 = call <16 x i32> @llvm.fshr.v16i32(<16 x i32> %a512, <16 x i32> %b512, <16 x i32> <i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5>)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret void
+;
+; AVX512GFNI-LABEL: 'splatconstant_funnel_i32'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %I32 = call i32 @llvm.fshr.i32(i32 %a32, i32 %b32, i32 5)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V2I32 = call <4 x i32> @llvm.fshr.v4i32(<4 x i32> %a128, <4 x i32> %b128, <4 x i32> <i32 5, i32 5, i32 5, i32 5>)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V4I32 = call <8 x i32> @llvm.fshr.v8i32(<8 x i32> %a256, <8 x i32> %b256, <8 x i32> <i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5>)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V8I32 = call <16 x i32> @llvm.fshr.v16i32(<16 x i32> %a512, <16 x i32> %b512, <16 x i32> <i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5>)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret void
 ;
   %I32   = call i32 @llvm.fshr.i32(i32 %a32, i32 %b32, i32 5)
   %V2I32 = call <4 x i32> @llvm.fshr.v4i32(<4 x i32> %a128, <4 x i32> %b128, <4 x i32> <i32 5, i32 5, i32 5, i32 5>)
@@ -1354,6 +1461,13 @@ define void @splatconstant_funnel_i16(i16 %a16, <8 x i16> %a128, <16 x i16> %a25
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 20 for instruction: %V16I16 = call <16 x i16> @llvm.fshr.v16i16(<16 x i16> %a256, <16 x i16> %b256, <16 x i16> <i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3>)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 40 for instruction: %V32I16 = call <32 x i16> @llvm.fshr.v32i16(<32 x i16> %a512, <32 x i16> %b512, <32 x i16> <i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3>)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret void
+;
+; AVX512GFNI-LABEL: 'splatconstant_funnel_i16'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %I16 = call i16 @llvm.fshr.i16(i16 %a16, i16 %b16, i16 3)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V8I16 = call <8 x i16> @llvm.fshr.v8i16(<8 x i16> %a128, <8 x i16> %b128, <8 x i16> <i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3>)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V16I16 = call <16 x i16> @llvm.fshr.v16i16(<16 x i16> %a256, <16 x i16> %b256, <16 x i16> <i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3>)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V32I16 = call <32 x i16> @llvm.fshr.v32i16(<32 x i16> %a512, <32 x i16> %b512, <32 x i16> <i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3>)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret void
 ;
   %I16    = call i16 @llvm.fshr.i16(i16 %a16, i16 %b16, i16 3)
   %V8I16  = call <8 x i16> @llvm.fshr.v8i16(<8 x i16> %a128, <8 x i16> %b128, <8 x i16> <i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3>)
@@ -1432,6 +1546,13 @@ define void @splatconstant_funnel_i8(i8 %a8, <16 x i8> %a128, <32 x i8> %a256, <
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 22 for instruction: %V32I8 = call <32 x i8> @llvm.fshr.v32i8(<32 x i8> %a256, <32 x i8> %b256, <32 x i8> <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, 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>)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 44 for instruction: %V64I8 = call <64 x i8> @llvm.fshr.v64i8(<64 x i8> %a512, <64 x i8> %b512, <64 x i8> <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, 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, 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, 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>)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret void
+;
+; AVX512GFNI-LABEL: 'splatconstant_funnel_i8'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %I8 = call i8 @llvm.fshr.i8(i8 %a8, i8 %b8, i8 3)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 18 for instruction: %V16I8 = call <16 x i8> @llvm.fshr.v16i8(<16 x i8> %a128, <16 x i8> %b128, <16 x i8> <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>)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 20 for instruction: %V32I8 = call <32 x i8> @llvm.fshr.v32i8(<32 x i8> %a256, <32 x i8> %b256, <32 x i8> <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, 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>)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 20 for instruction: %V64I8 = call <64 x i8> @llvm.fshr.v64i8(<64 x i8> %a512, <64 x i8> %b512, <64 x i8> <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, 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, 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, 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>)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret void
 ;
   %I8    = call i8 @llvm.fshr.i8(i8 %a8, i8 %b8, i8 3)
   %V16I8 = call <16 x i8> @llvm.fshr.v16i8(<16 x i8> %a128, <16 x i8> %b128, <16 x i8> <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>)
@@ -1642,6 +1763,13 @@ define void @var_rotate_i16(i16 %a16, <8 x i16> %a128, <16 x i16> %a256, <32 x i
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 7 for instruction: %V16I16 = call <16 x i16> @llvm.fshr.v16i16(<16 x i16> %a256, <16 x i16> %a256, <16 x i16> %c256)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 14 for instruction: %V32I16 = call <32 x i16> @llvm.fshr.v32i16(<32 x i16> %a512, <32 x i16> %a512, <32 x i16> %c512)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret void
+;
+; AVX512GFNI-LABEL: 'var_rotate_i16'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %I16 = call i16 @llvm.fshr.i16(i16 %a16, i16 %a16, i16 %c16)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V8I16 = call <8 x i16> @llvm.fshr.v8i16(<8 x i16> %a128, <8 x i16> %a128, <8 x i16> %c128)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V16I16 = call <16 x i16> @llvm.fshr.v16i16(<16 x i16> %a256, <16 x i16> %a256, <16 x i16> %c256)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V32I16 = call <32 x i16> @llvm.fshr.v32i16(<32 x i16> %a512, <32 x i16> %a512, <32 x i16> %c512)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret void
 ;
   %I16    = call i16 @llvm.fshr.i16(i16 %a16, i16 %a16, i16 %c16)
   %V8I16  = call <8 x i16> @llvm.fshr.v8i16(<8 x i16> %a128, <8 x i16> %a128, <8 x i16> %c128)
@@ -1727,6 +1855,13 @@ define void @var_rotate_i8(i8 %a8, <16 x i8> %a128, <32 x i8> %a256, <64 x i8> %
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 7 for instruction: %V32I8 = call <32 x i8> @llvm.fshr.v32i8(<32 x i8> %a256, <32 x i8> %a256, <32 x i8> %c256)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 14 for instruction: %V64I8 = call <64 x i8> @llvm.fshr.v64i8(<64 x i8> %a512, <64 x i8> %a512, <64 x i8> %c512)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret void
+;
+; AVX512GFNI-LABEL: 'var_rotate_i8'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %I8 = call i8 @llvm.fshr.i8(i8 %a8, i8 %a8, i8 %c8)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 14 for instruction: %V16I8 = call <16 x i8> @llvm.fshr.v16i8(<16 x i8> %a128, <16 x i8> %a128, <16 x i8> %c128)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 14 for instruction: %V32I8 = call <32 x i8> @llvm.fshr.v32i8(<32 x i8> %a256, <32 x i8> %a256, <32 x i8> %c256)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 6 for instruction: %V64I8 = call <64 x i8> @llvm.fshr.v64i8(<64 x i8> %a512, <64 x i8> %a512, <64 x i8> %c512)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret void
 ;
   %I8    = call i8 @llvm.fshr.i8(i8 %a8, i8 %a8, i8 %c8)
   %V16I8 = call <16 x i8> @llvm.fshr.v16i8(<16 x i8> %a128, <16 x i8> %a128, <16 x i8> %c128)
@@ -1975,6 +2110,15 @@ define void @splatvar_rotate_i16(i16 %a16, <8 x i16> %a128, <16 x i16> %a256, <3
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 7 for instruction: %V16I16 = call <16 x i16> @llvm.fshr.v16i16(<16 x i16> %a256, <16 x i16> %a256, <16 x i16> %u256)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 14 for instruction: %V32I16 = call <32 x i16> @llvm.fshr.v32i16(<32 x i16> %a512, <32 x i16> %a512, <32 x i16> %u512)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret void
+;
+; AVX512GFNI-LABEL: 'splatvar_rotate_i16'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %u128 = shufflevector <8 x i16> %c128, <8 x i16> undef, <8 x i32> zeroinitializer
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %u256 = shufflevector <16 x i16> %c256, <16 x i16> undef, <16 x i32> zeroinitializer
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %u512 = shufflevector <32 x i16> %c512, <32 x i16> undef, <32 x i32> zeroinitializer
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V8I16 = call <8 x i16> @llvm.fshr.v8i16(<8 x i16> %a128, <8 x i16> %a128, <8 x i16> %u128)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V16I16 = call <16 x i16> @llvm.fshr.v16i16(<16 x i16> %a256, <16 x i16> %a256, <16 x i16> %u256)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V32I16 = call <32 x i16> @llvm.fshr.v32i16(<32 x i16> %a512, <32 x i16> %a512, <32 x i16> %u512)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret void
 ;
   %u128 = shufflevector <8 x i16> %c128, <8 x i16> undef, <8 x i32> zeroinitializer
   %u256 = shufflevector <16 x i16> %c256, <16 x i16> undef, <16 x i32> zeroinitializer
@@ -2075,6 +2219,15 @@ define void @splatvar_rotate_i8(i8 %a8, <16 x i8> %a128, <32 x i8> %a256, <64 x
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 7 for instruction: %V32I8 = call <32 x i8> @llvm.fshr.v32i8(<32 x i8> %a256, <32 x i8> %a256, <32 x i8> %u256)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 14 for instruction: %V64I8 = call <64 x i8> @llvm.fshr.v64i8(<64 x i8> %a512, <64 x i8> %a512, <64 x i8> %u512)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret void
+;
+; AVX512GFNI-LABEL: 'splatvar_rotate_i8'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %u128 = shufflevector <16 x i8> %c128, <16 x i8> undef, <16 x i32> zeroinitializer
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %u256 = shufflevector <32 x i8> %c256, <32 x i8> undef, <32 x i32> zeroinitializer
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %u512 = shufflevector <64 x i8> %c512, <64 x i8> undef, <64 x i32> zeroinitializer
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 14 for instruction: %V16I8 = call <16 x i8> @llvm.fshr.v16i8(<16 x i8> %a128, <16 x i8> %a128, <16 x i8> %u128)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 14 for instruction: %V32I8 = call <32 x i8> @llvm.fshr.v32i8(<32 x i8> %a256, <32 x i8> %a256, <32 x i8> %u256)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 6 for instruction: %V64I8 = call <64 x i8> @llvm.fshr.v64i8(<64 x i8> %a512, <64 x i8> %a512, <64 x i8> %u512)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret void
 ;
   %u128 = shufflevector <16 x i8> %c128, <16 x i8> undef, <16 x i32> zeroinitializer
   %u256 = shufflevector <32 x i8> %c256, <32 x i8> undef, <32 x i32> zeroinitializer
@@ -2287,6 +2440,13 @@ define void @constant_rotate_i16(i16 %a16, <8 x i16> %a128, <16 x i16> %a256, <3
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 7 for instruction: %V16I16 = call <16 x i16> @llvm.fshr.v16i16(<16 x i16> %a256, <16 x i16> %a256, <16 x i16> <i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7>)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 14 for instruction: %V32I16 = call <32 x i16> @llvm.fshr.v32i16(<32 x i16> %a512, <32 x i16> %a512, <32 x i16> <i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7>)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret void
+;
+; AVX512GFNI-LABEL: 'constant_rotate_i16'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %I16 = call i16 @llvm.fshr.i16(i16 %a16, i16 %a16, i16 7)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V8I16 = call <8 x i16> @llvm.fshr.v8i16(<8 x i16> %a128, <8 x i16> %a128, <8 x i16> <i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7>)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V16I16 = call <16 x i16> @llvm.fshr.v16i16(<16 x i16> %a256, <16 x i16> %a256, <16 x i16> <i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7>)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V32I16 = call <32 x i16> @llvm.fshr.v32i16(<32 x i16> %a512, <32 x i16> %a512, <32 x i16> <i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7>)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret void
 ;
   %I16    = call i16 @llvm.fshr.i16(i16 %a16, i16 %a16, i16 7)
   %V8I16  = call <8 x i16> @llvm.fshr.v8i16(<8 x i16> %a128, <8 x i16> %a128, <8 x i16> <i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7>)
@@ -2372,6 +2532,13 @@ define void @constant_rotate_i8(i8 %a8, <16 x i8> %a128, <32 x i8> %a256, <64 x
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 7 for instruction: %V32I8 = call <32 x i8> @llvm.fshr.v32i8(<32 x i8> %a256, <32 x i8> %a256, <32 x i8> <i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0>)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 14 for instruction: %V64I8 = call <64 x i8> @llvm.fshr.v64i8(<64 x i8> %a512, <64 x i8> %a512, <64 x i8> <i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0>)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret void
+;
+; AVX512GFNI-LABEL: 'constant_rotate_i8'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %I8 = call i8 @llvm.fshr.i8(i8 %a8, i8 %a8, i8 7)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 14 for instruction: %V16I8 = call <16 x i8> @llvm.fshr.v16i8(<16 x i8> %a128, <16 x i8> %a128, <16 x i8> <i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0>)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 14 for instruction: %V32I8 = call <32 x i8> @llvm.fshr.v32i8(<32 x i8> %a256, <32 x i8> %a256, <32 x i8> <i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0>)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 6 for instruction: %V64I8 = call <64 x i8> @llvm.fshr.v64i8(<64 x i8> %a512, <64 x i8> %a512, <64 x i8> <i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0>)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret void
 ;
   %I8    = call i8 @llvm.fshr.i8(i8 %a8, i8 %a8, i8 7)
   %V16I8 = call <16 x i8> @llvm.fshr.v16i8(<16 x i8> %a128, <16 x i8> %a128, <16 x i8> <i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0>)
@@ -2568,6 +2735,13 @@ define void @splatconstant_rotate_i16(i16 %a16, <8 x i16> %a128, <16 x i16> %a25
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 7 for instruction: %V16I16 = call <16 x i16> @llvm.fshr.v16i16(<16 x i16> %a256, <16 x i16> %a256, <16 x i16> <i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3>)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 14 for instruction: %V32I16 = call <32 x i16> @llvm.fshr.v32i16(<32 x i16> %a512, <32 x i16> %a512, <32 x i16> <i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3>)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret void
+;
+; AVX512GFNI-LABEL: 'splatconstant_rotate_i16'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %I16 = call i16 @llvm.fshr.i16(i16 %a16, i16 %a16, i16 3)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V8I16 = call <8 x i16> @llvm.fshr.v8i16(<8 x i16> %a128, <8 x i16> %a128, <8 x i16> <i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3>)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V16I16 = call <16 x i16> @llvm.fshr.v16i16(<16 x i16> %a256, <16 x i16> %a256, <16 x i16> <i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3>)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V32I16 = call <32 x i16> @llvm.fshr.v32i16(<32 x i16> %a512, <32 x i16> %a512, <32 x i16> <i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3>)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret void
 ;
   %I16    = call i16 @llvm.fshr.i16(i16 %a16, i16 %a16, i16 3)
   %V8I16  = call <8 x i16> @llvm.fshr.v8i16(<8 x i16> %a128, <8 x i16> %a128, <8 x i16> <i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3>)
@@ -2646,6 +2820,13 @@ define void @splatconstant_rotate_i8(i8 %a8, <16 x i8> %a128, <32 x i8> %a256, <
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 7 for instruction: %V32I8 = call <32 x i8> @llvm.fshr.v32i8(<32 x i8> %a256, <32 x i8> %a256, <32 x i8> <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, 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>)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 14 for instruction: %V64I8 = call <64 x i8> @llvm.fshr.v64i8(<64 x i8> %a512, <64 x i8> %a512, <64 x i8> <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, 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, 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, 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>)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret void
+;
+; AVX512GFNI-LABEL: 'splatconstant_rotate_i8'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %I8 = call i8 @llvm.fshr.i8(i8 %a8, i8 %a8, i8 3)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 14 for instruction: %V16I8 = call <16 x i8> @llvm.fshr.v16i8(<16 x i8> %a128, <16 x i8> %a128, <16 x i8> <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>)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 14 for instruction: %V32I8 = call <32 x i8> @llvm.fshr.v32i8(<32 x i8> %a256, <32 x i8> %a256, <32 x i8> <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, 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>)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 6 for instruction: %V64I8 = call <64 x i8> @llvm.fshr.v64i8(<64 x i8> %a512, <64 x i8> %a512, <64 x i8> <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, 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, 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, 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>)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret void
 ;
   %I8    = call i8 @llvm.fshr.i8(i8 %a8, i8 %a8, i8 3)
   %V16I8 = call <16 x i8> @llvm.fshr.v16i8(<16 x i8> %a128, <16 x i8> %a128, <16 x i8> <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>)

diff  --git a/llvm/test/Analysis/CostModel/X86/fshr-sizelatency.ll b/llvm/test/Analysis/CostModel/X86/fshr-sizelatency.ll
index 2d106c6dd30691..8931781f70bdce 100644
--- a/llvm/test/Analysis/CostModel/X86/fshr-sizelatency.ll
+++ b/llvm/test/Analysis/CostModel/X86/fshr-sizelatency.ll
@@ -12,6 +12,7 @@
 ; RUN: opt < %s -passes="print<cost-model>" 2>&1 -disable-output -cost-kind=size-latency -mtriple=x86_64-apple-macosx10.8.0 -mcpu=goldmont | FileCheck %s --check-prefixes=GLM
 ; RUN: opt < %s -passes="print<cost-model>" 2>&1 -disable-output -cost-kind=size-latency -mtriple=x86_64-apple-macosx10.8.0 -mcpu=bdver2 | FileCheck %s --check-prefixes=XOP
 ; RUN: opt < %s -passes="print<cost-model>" 2>&1 -disable-output -cost-kind=size-latency -mtriple=x86_64-apple-macosx10.8.0 -mcpu=btver2 | FileCheck %s --check-prefixes=AVX1
+; RUN: opt < %s -passes="print<cost-model>" 2>&1 -disable-output -cost-kind=size-latency -mtriple=x86_64-apple-macosx10.8.0 -mcpu=tigerlake | FileCheck %s --check-prefixes=AVX512GFNI
 
 target datalayout = "e-p:64:64:64-i1:8:8-i8:8:8-i16:16:16-i32:32:32-i64:64:64-f32:32:32-f64:64:64-v64:64:64-v128:128:128-a0:0:64-s0:64:64-f80:128:128-n8:16:32:64-S128"
 target triple = "x86_64-apple-macosx10.8.0"
@@ -97,6 +98,13 @@ define void @var_funnel_i64(i64 %a64, <2 x i64> %a128, <4 x i64> %a256, <8 x i64
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 30 for instruction: %V4I64 = call <4 x i64> @llvm.fshr.v4i64(<4 x i64> %a256, <4 x i64> %b256, <4 x i64> %c256)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 60 for instruction: %V8I64 = call <8 x i64> @llvm.fshr.v8i64(<8 x i64> %a512, <8 x i64> %b512, <8 x i64> %c512)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret void
+;
+; AVX512GFNI-LABEL: 'var_funnel_i64'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %I64 = call i64 @llvm.fshr.i64(i64 %a64, i64 %b64, i64 %c64)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V2I64 = call <2 x i64> @llvm.fshr.v2i64(<2 x i64> %a128, <2 x i64> %b128, <2 x i64> %c128)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V4I64 = call <4 x i64> @llvm.fshr.v4i64(<4 x i64> %a256, <4 x i64> %b256, <4 x i64> %c256)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V8I64 = call <8 x i64> @llvm.fshr.v8i64(<8 x i64> %a512, <8 x i64> %b512, <8 x i64> %c512)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret void
 ;
   %I64    = call i64 @llvm.fshr.i64(i64 %a64, i64 %b64, i64 %c64)
   %V2I64  = call <2 x i64> @llvm.fshr.v2i64(<2 x i64> %a128, <2 x i64> %b128, <2 x i64> %c128)
@@ -182,6 +190,13 @@ define void @var_funnel_i32(i32 %a32, <4 x i32> %a128, <8 x i32> %a256, <16 x i3
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 30 for instruction: %V4I32 = call <8 x i32> @llvm.fshr.v8i32(<8 x i32> %a256, <8 x i32> %b256, <8 x i32> %c256)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 60 for instruction: %V8I32 = call <16 x i32> @llvm.fshr.v16i32(<16 x i32> %a512, <16 x i32> %b512, <16 x i32> %c512)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret void
+;
+; AVX512GFNI-LABEL: 'var_funnel_i32'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %I32 = call i32 @llvm.fshr.i32(i32 %a32, i32 %b32, i32 %c32)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V2I32 = call <4 x i32> @llvm.fshr.v4i32(<4 x i32> %a128, <4 x i32> %b128, <4 x i32> %c128)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V4I32 = call <8 x i32> @llvm.fshr.v8i32(<8 x i32> %a256, <8 x i32> %b256, <8 x i32> %c256)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V8I32 = call <16 x i32> @llvm.fshr.v16i32(<16 x i32> %a512, <16 x i32> %b512, <16 x i32> %c512)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret void
 ;
   %I32   = call i32 @llvm.fshr.i32(i32 %a32, i32 %b32, i32 %c32)
   %V2I32 = call <4 x i32> @llvm.fshr.v4i32(<4 x i32> %a128, <4 x i32> %b128, <4 x i32> %c128)
@@ -267,6 +282,13 @@ define void @var_funnel_i16(i16 %a16, <8 x i16> %a128, <16 x i16> %a256, <32 x i
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 31 for instruction: %V16I16 = call <16 x i16> @llvm.fshr.v16i16(<16 x i16> %a256, <16 x i16> %b256, <16 x i16> %c256)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 62 for instruction: %V32I16 = call <32 x i16> @llvm.fshr.v32i16(<32 x i16> %a512, <32 x i16> %b512, <32 x i16> %c512)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret void
+;
+; AVX512GFNI-LABEL: 'var_funnel_i16'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 5 for instruction: %I16 = call i16 @llvm.fshr.i16(i16 %a16, i16 %b16, i16 %c16)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V8I16 = call <8 x i16> @llvm.fshr.v8i16(<8 x i16> %a128, <8 x i16> %b128, <8 x i16> %c128)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V16I16 = call <16 x i16> @llvm.fshr.v16i16(<16 x i16> %a256, <16 x i16> %b256, <16 x i16> %c256)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V32I16 = call <32 x i16> @llvm.fshr.v32i16(<32 x i16> %a512, <32 x i16> %b512, <32 x i16> %c512)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret void
 ;
   %I16    = call i16 @llvm.fshr.i16(i16 %a16, i16 %b16, i16 %c16)
   %V8I16  = call <8 x i16> @llvm.fshr.v8i16(<8 x i16> %a128, <8 x i16> %b128, <8 x i16> %c128)
@@ -352,6 +374,13 @@ define void @var_funnel_i8(i8 %a8, <16 x i8> %a128, <32 x i8> %a256, <64 x i8> %
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 31 for instruction: %V32I8 = call <32 x i8> @llvm.fshr.v32i8(<32 x i8> %a256, <32 x i8> %b256, <32 x i8> %c256)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 62 for instruction: %V64I8 = call <64 x i8> @llvm.fshr.v64i8(<64 x i8> %a512, <64 x i8> %b512, <64 x i8> %c512)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret void
+;
+; AVX512GFNI-LABEL: 'var_funnel_i8'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 5 for instruction: %I8 = call i8 @llvm.fshr.i8(i8 %a8, i8 %b8, i8 %c8)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 15 for instruction: %V16I8 = call <16 x i8> @llvm.fshr.v16i8(<16 x i8> %a128, <16 x i8> %b128, <16 x i8> %c128)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 39 for instruction: %V32I8 = call <32 x i8> @llvm.fshr.v32i8(<32 x i8> %a256, <32 x i8> %b256, <32 x i8> %c256)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 38 for instruction: %V64I8 = call <64 x i8> @llvm.fshr.v64i8(<64 x i8> %a512, <64 x i8> %b512, <64 x i8> %c512)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret void
 ;
   %I8    = call i8 @llvm.fshr.i8(i8 %a8, i8 %b8, i8 %c8)
   %V16I8 = call <16 x i8> @llvm.fshr.v16i8(<16 x i8> %a128, <16 x i8> %b128, <16 x i8> %c128)
@@ -463,6 +492,15 @@ define void @splatvar_funnel_i64(i64 %a64, <2 x i64> %a128, <4 x i64> %a256, <8
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 30 for instruction: %V4I64 = call <4 x i64> @llvm.fshr.v4i64(<4 x i64> %a256, <4 x i64> %b256, <4 x i64> %u256)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 60 for instruction: %V8I64 = call <8 x i64> @llvm.fshr.v8i64(<8 x i64> %a512, <8 x i64> %b512, <8 x i64> %u512)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret void
+;
+; AVX512GFNI-LABEL: 'splatvar_funnel_i64'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %u128 = shufflevector <2 x i64> %c128, <2 x i64> undef, <2 x i32> zeroinitializer
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %u256 = shufflevector <4 x i64> %c256, <4 x i64> undef, <4 x i32> zeroinitializer
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %u512 = shufflevector <8 x i64> %c512, <8 x i64> undef, <8 x i32> zeroinitializer
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V2I64 = call <2 x i64> @llvm.fshr.v2i64(<2 x i64> %a128, <2 x i64> %b128, <2 x i64> %u128)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V4I64 = call <4 x i64> @llvm.fshr.v4i64(<4 x i64> %a256, <4 x i64> %b256, <4 x i64> %u256)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V8I64 = call <8 x i64> @llvm.fshr.v8i64(<8 x i64> %a512, <8 x i64> %b512, <8 x i64> %u512)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret void
 ;
   %u128 = shufflevector <2 x i64> %c128, <2 x i64> undef, <2 x i32> zeroinitializer
   %u256 = shufflevector <4 x i64> %c256, <4 x i64> undef, <4 x i32> zeroinitializer
@@ -572,6 +610,15 @@ define void @splatvar_funnel_i32(i32 %a32, <4 x i32> %a128, <8 x i32> %a256, <16
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 32 for instruction: %V4I32 = call <8 x i32> @llvm.fshr.v8i32(<8 x i32> %a256, <8 x i32> %b256, <8 x i32> %u256)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 64 for instruction: %V8I32 = call <16 x i32> @llvm.fshr.v16i32(<16 x i32> %a512, <16 x i32> %b512, <16 x i32> %u512)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret void
+;
+; AVX512GFNI-LABEL: 'splatvar_funnel_i32'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %u128 = shufflevector <4 x i32> %c128, <4 x i32> undef, <4 x i32> zeroinitializer
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %u256 = shufflevector <8 x i32> %c256, <8 x i32> undef, <8 x i32> zeroinitializer
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %u512 = shufflevector <16 x i32> %c512, <16 x i32> undef, <16 x i32> zeroinitializer
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V2I32 = call <4 x i32> @llvm.fshr.v4i32(<4 x i32> %a128, <4 x i32> %b128, <4 x i32> %u128)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V4I32 = call <8 x i32> @llvm.fshr.v8i32(<8 x i32> %a256, <8 x i32> %b256, <8 x i32> %u256)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V8I32 = call <16 x i32> @llvm.fshr.v16i32(<16 x i32> %a512, <16 x i32> %b512, <16 x i32> %u512)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret void
 ;
   %u128 = shufflevector <4 x i32> %c128, <4 x i32> undef, <4 x i32> zeroinitializer
   %u256 = shufflevector <8 x i32> %c256, <8 x i32> undef, <8 x i32> zeroinitializer
@@ -681,6 +728,15 @@ define void @splatvar_funnel_i16(i16 %a16, <8 x i16> %a128, <16 x i16> %a256, <3
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 33 for instruction: %V16I16 = call <16 x i16> @llvm.fshr.v16i16(<16 x i16> %a256, <16 x i16> %b256, <16 x i16> %u256)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 66 for instruction: %V32I16 = call <32 x i16> @llvm.fshr.v32i16(<32 x i16> %a512, <32 x i16> %b512, <32 x i16> %u512)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret void
+;
+; AVX512GFNI-LABEL: 'splatvar_funnel_i16'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %u128 = shufflevector <8 x i16> %c128, <8 x i16> undef, <8 x i32> zeroinitializer
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %u256 = shufflevector <16 x i16> %c256, <16 x i16> undef, <16 x i32> zeroinitializer
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %u512 = shufflevector <32 x i16> %c512, <32 x i16> undef, <32 x i32> zeroinitializer
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V8I16 = call <8 x i16> @llvm.fshr.v8i16(<8 x i16> %a128, <8 x i16> %b128, <8 x i16> %u128)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V16I16 = call <16 x i16> @llvm.fshr.v16i16(<16 x i16> %a256, <16 x i16> %b256, <16 x i16> %u256)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V32I16 = call <32 x i16> @llvm.fshr.v32i16(<32 x i16> %a512, <32 x i16> %b512, <32 x i16> %u512)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret void
 ;
   %u128 = shufflevector <8 x i16> %c128, <8 x i16> undef, <8 x i32> zeroinitializer
   %u256 = shufflevector <16 x i16> %c256, <16 x i16> undef, <16 x i32> zeroinitializer
@@ -790,6 +846,15 @@ define void @splatvar_funnel_i8(i8 %a8, <16 x i8> %a128, <32 x i8> %a256, <64 x
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 31 for instruction: %V32I8 = call <32 x i8> @llvm.fshr.v32i8(<32 x i8> %a256, <32 x i8> %b256, <32 x i8> %u256)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 62 for instruction: %V64I8 = call <64 x i8> @llvm.fshr.v64i8(<64 x i8> %a512, <64 x i8> %b512, <64 x i8> %u512)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret void
+;
+; AVX512GFNI-LABEL: 'splatvar_funnel_i8'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %u128 = shufflevector <16 x i8> %c128, <16 x i8> undef, <16 x i32> zeroinitializer
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %u256 = shufflevector <32 x i8> %c256, <32 x i8> undef, <32 x i32> zeroinitializer
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %u512 = shufflevector <64 x i8> %c512, <64 x i8> undef, <64 x i32> zeroinitializer
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 20 for instruction: %V16I8 = call <16 x i8> @llvm.fshr.v16i8(<16 x i8> %a128, <16 x i8> %b128, <16 x i8> %u128)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 22 for instruction: %V32I8 = call <32 x i8> @llvm.fshr.v32i8(<32 x i8> %a256, <32 x i8> %b256, <32 x i8> %u256)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 23 for instruction: %V64I8 = call <64 x i8> @llvm.fshr.v64i8(<64 x i8> %a512, <64 x i8> %b512, <64 x i8> %u512)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret void
 ;
   %u128 = shufflevector <16 x i8> %c128, <16 x i8> undef, <16 x i32> zeroinitializer
   %u256 = shufflevector <32 x i8> %c256, <32 x i8> undef, <32 x i32> zeroinitializer
@@ -881,6 +946,13 @@ define void @constant_funnel_i64(i64 %a64, <2 x i64> %a128, <4 x i64> %a256, <8
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 28 for instruction: %V4I64 = call <4 x i64> @llvm.fshr.v4i64(<4 x i64> %a256, <4 x i64> %b256, <4 x i64> <i64 1, i64 7, i64 15, i64 31>)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 56 for instruction: %V8I64 = call <8 x i64> @llvm.fshr.v8i64(<8 x i64> %a512, <8 x i64> %b512, <8 x i64> <i64 1, i64 7, i64 15, i64 31, i64 1, i64 7, i64 15, i64 31>)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret void
+;
+; AVX512GFNI-LABEL: 'constant_funnel_i64'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %I64 = call i64 @llvm.fshr.i64(i64 %a64, i64 %b64, i64 7)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V2I64 = call <2 x i64> @llvm.fshr.v2i64(<2 x i64> %a128, <2 x i64> %b128, <2 x i64> <i64 1, i64 7>)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V4I64 = call <4 x i64> @llvm.fshr.v4i64(<4 x i64> %a256, <4 x i64> %b256, <4 x i64> <i64 1, i64 7, i64 15, i64 31>)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V8I64 = call <8 x i64> @llvm.fshr.v8i64(<8 x i64> %a512, <8 x i64> %b512, <8 x i64> <i64 1, i64 7, i64 15, i64 31, i64 1, i64 7, i64 15, i64 31>)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret void
 ;
   %I64    = call i64 @llvm.fshr.i64(i64 %a64, i64 %b64, i64 7)
   %V2I64  = call <2 x i64> @llvm.fshr.v2i64(<2 x i64> %a128, <2 x i64> %b128, <2 x i64> <i64 1, i64 7>)
@@ -966,6 +1038,13 @@ define void @constant_funnel_i32(i32 %a32, <4 x i32> %a128, <8 x i32> %a256, <16
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 28 for instruction: %V4I32 = call <8 x i32> @llvm.fshr.v8i32(<8 x i32> %a256, <8 x i32> %b256, <8 x i32> <i32 4, i32 5, i32 6, i32 7, i32 0, i32 1, i32 2, i32 3>)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 56 for instruction: %V8I32 = call <16 x i32> @llvm.fshr.v16i32(<16 x i32> %a512, <16 x i32> %b512, <16 x i32> <i32 4, i32 5, i32 6, i32 7, i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 0, i32 1, i32 2, i32 3>)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret void
+;
+; AVX512GFNI-LABEL: 'constant_funnel_i32'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %I32 = call i32 @llvm.fshr.i32(i32 %a32, i32 %b32, i32 7)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V2I32 = call <4 x i32> @llvm.fshr.v4i32(<4 x i32> %a128, <4 x i32> %b128, <4 x i32> <i32 4, i32 5, i32 6, i32 7>)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V4I32 = call <8 x i32> @llvm.fshr.v8i32(<8 x i32> %a256, <8 x i32> %b256, <8 x i32> <i32 4, i32 5, i32 6, i32 7, i32 0, i32 1, i32 2, i32 3>)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V8I32 = call <16 x i32> @llvm.fshr.v16i32(<16 x i32> %a512, <16 x i32> %b512, <16 x i32> <i32 4, i32 5, i32 6, i32 7, i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 0, i32 1, i32 2, i32 3>)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret void
 ;
   %I32   = call i32 @llvm.fshr.i32(i32 %a32, i32 %b32, i32 7)
   %V2I32 = call <4 x i32> @llvm.fshr.v4i32(<4 x i32> %a128, <4 x i32> %b128, <4 x i32> <i32 4, i32 5, i32 6, i32 7>)
@@ -1051,6 +1130,13 @@ define void @constant_funnel_i16(i16 %a16, <8 x i16> %a128, <16 x i16> %a256, <3
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 29 for instruction: %V16I16 = call <16 x i16> @llvm.fshr.v16i16(<16 x i16> %a256, <16 x i16> %b256, <16 x i16> <i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7>)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 58 for instruction: %V32I16 = call <32 x i16> @llvm.fshr.v32i16(<32 x i16> %a512, <32 x i16> %b512, <32 x i16> <i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7>)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret void
+;
+; AVX512GFNI-LABEL: 'constant_funnel_i16'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 5 for instruction: %I16 = call i16 @llvm.fshr.i16(i16 %a16, i16 %b16, i16 7)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V8I16 = call <8 x i16> @llvm.fshr.v8i16(<8 x i16> %a128, <8 x i16> %b128, <8 x i16> <i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7>)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V16I16 = call <16 x i16> @llvm.fshr.v16i16(<16 x i16> %a256, <16 x i16> %b256, <16 x i16> <i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7>)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V32I16 = call <32 x i16> @llvm.fshr.v32i16(<32 x i16> %a512, <32 x i16> %b512, <32 x i16> <i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7>)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret void
 ;
   %I16    = call i16 @llvm.fshr.i16(i16 %a16, i16 %b16, i16 7)
   %V8I16  = call <8 x i16> @llvm.fshr.v8i16(<8 x i16> %a128, <8 x i16> %b128, <8 x i16> <i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7>)
@@ -1136,6 +1222,13 @@ define void @constant_funnel_i8(i8 %a8, <16 x i8> %a128, <32 x i8> %a256, <64 x
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 29 for instruction: %V32I8 = call <32 x i8> @llvm.fshr.v32i8(<32 x i8> %a256, <32 x i8> %b256, <32 x i8> <i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0>)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 58 for instruction: %V64I8 = call <64 x i8> @llvm.fshr.v64i8(<64 x i8> %a512, <64 x i8> %b512, <64 x i8> <i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0>)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret void
+;
+; AVX512GFNI-LABEL: 'constant_funnel_i8'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 5 for instruction: %I8 = call i8 @llvm.fshr.i8(i8 %a8, i8 %b8, i8 7)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 14 for instruction: %V16I8 = call <16 x i8> @llvm.fshr.v16i8(<16 x i8> %a128, <16 x i8> %b128, <16 x i8> <i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0>)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 38 for instruction: %V32I8 = call <32 x i8> @llvm.fshr.v32i8(<32 x i8> %a256, <32 x i8> %b256, <32 x i8> <i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0>)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 37 for instruction: %V64I8 = call <64 x i8> @llvm.fshr.v64i8(<64 x i8> %a512, <64 x i8> %b512, <64 x i8> <i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0>)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret void
 ;
   %I8    = call i8 @llvm.fshr.i8(i8 %a8, i8 %b8, i8 7)
   %V16I8 = call <16 x i8> @llvm.fshr.v16i8(<16 x i8> %a128, <16 x i8> %b128, <16 x i8> <i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0>)
@@ -1225,6 +1318,13 @@ define void @splatconstant_funnel_i64(i64 %a64, <2 x i64> %a128, <4 x i64> %a256
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 26 for instruction: %V4I64 = call <4 x i64> @llvm.fshr.v4i64(<4 x i64> %a256, <4 x i64> %b256, <4 x i64> <i64 7, i64 7, i64 7, i64 7>)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 52 for instruction: %V8I64 = call <8 x i64> @llvm.fshr.v8i64(<8 x i64> %a512, <8 x i64> %b512, <8 x i64> <i64 7, i64 7, i64 7, i64 7, i64 7, i64 7, i64 7, i64 7>)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret void
+;
+; AVX512GFNI-LABEL: 'splatconstant_funnel_i64'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %I64 = call i64 @llvm.fshr.i64(i64 %a64, i64 %b64, i64 7)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V2I64 = call <2 x i64> @llvm.fshr.v2i64(<2 x i64> %a128, <2 x i64> %b128, <2 x i64> <i64 7, i64 7>)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V4I64 = call <4 x i64> @llvm.fshr.v4i64(<4 x i64> %a256, <4 x i64> %b256, <4 x i64> <i64 7, i64 7, i64 7, i64 7>)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V8I64 = call <8 x i64> @llvm.fshr.v8i64(<8 x i64> %a512, <8 x i64> %b512, <8 x i64> <i64 7, i64 7, i64 7, i64 7, i64 7, i64 7, i64 7, i64 7>)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret void
 ;
   %I64    = call i64 @llvm.fshr.i64(i64 %a64, i64 %b64, i64 7)
   %V2I64  = call <2 x i64> @llvm.fshr.v2i64(<2 x i64> %a128, <2 x i64> %b128, <2 x i64> <i64 7, i64 7>)
@@ -1310,6 +1410,13 @@ define void @splatconstant_funnel_i32(i32 %a32, <4 x i32> %a128, <8 x i32> %a256
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 26 for instruction: %V4I32 = call <8 x i32> @llvm.fshr.v8i32(<8 x i32> %a256, <8 x i32> %b256, <8 x i32> <i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5>)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 52 for instruction: %V8I32 = call <16 x i32> @llvm.fshr.v16i32(<16 x i32> %a512, <16 x i32> %b512, <16 x i32> <i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5>)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret void
+;
+; AVX512GFNI-LABEL: 'splatconstant_funnel_i32'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %I32 = call i32 @llvm.fshr.i32(i32 %a32, i32 %b32, i32 5)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V2I32 = call <4 x i32> @llvm.fshr.v4i32(<4 x i32> %a128, <4 x i32> %b128, <4 x i32> <i32 5, i32 5, i32 5, i32 5>)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V4I32 = call <8 x i32> @llvm.fshr.v8i32(<8 x i32> %a256, <8 x i32> %b256, <8 x i32> <i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5>)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V8I32 = call <16 x i32> @llvm.fshr.v16i32(<16 x i32> %a512, <16 x i32> %b512, <16 x i32> <i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5>)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret void
 ;
   %I32   = call i32 @llvm.fshr.i32(i32 %a32, i32 %b32, i32 5)
   %V2I32 = call <4 x i32> @llvm.fshr.v4i32(<4 x i32> %a128, <4 x i32> %b128, <4 x i32> <i32 5, i32 5, i32 5, i32 5>)
@@ -1395,6 +1502,13 @@ define void @splatconstant_funnel_i16(i16 %a16, <8 x i16> %a128, <16 x i16> %a25
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 27 for instruction: %V16I16 = call <16 x i16> @llvm.fshr.v16i16(<16 x i16> %a256, <16 x i16> %b256, <16 x i16> <i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3>)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 54 for instruction: %V32I16 = call <32 x i16> @llvm.fshr.v32i16(<32 x i16> %a512, <32 x i16> %b512, <32 x i16> <i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3>)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret void
+;
+; AVX512GFNI-LABEL: 'splatconstant_funnel_i16'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 5 for instruction: %I16 = call i16 @llvm.fshr.i16(i16 %a16, i16 %b16, i16 3)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V8I16 = call <8 x i16> @llvm.fshr.v8i16(<8 x i16> %a128, <8 x i16> %b128, <8 x i16> <i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3>)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V16I16 = call <16 x i16> @llvm.fshr.v16i16(<16 x i16> %a256, <16 x i16> %b256, <16 x i16> <i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3>)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V32I16 = call <32 x i16> @llvm.fshr.v32i16(<32 x i16> %a512, <32 x i16> %b512, <32 x i16> <i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3>)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret void
 ;
   %I16    = call i16 @llvm.fshr.i16(i16 %a16, i16 %b16, i16 3)
   %V8I16  = call <8 x i16> @llvm.fshr.v8i16(<8 x i16> %a128, <8 x i16> %b128, <8 x i16> <i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3>)
@@ -1480,6 +1594,13 @@ define void @splatconstant_funnel_i8(i8 %a8, <16 x i8> %a128, <32 x i8> %a256, <
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 29 for instruction: %V32I8 = call <32 x i8> @llvm.fshr.v32i8(<32 x i8> %a256, <32 x i8> %b256, <32 x i8> <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, 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>)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 58 for instruction: %V64I8 = call <64 x i8> @llvm.fshr.v64i8(<64 x i8> %a512, <64 x i8> %b512, <64 x i8> <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, 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, 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, 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>)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret void
+;
+; AVX512GFNI-LABEL: 'splatconstant_funnel_i8'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 5 for instruction: %I8 = call i8 @llvm.fshr.i8(i8 %a8, i8 %b8, i8 3)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 10 for instruction: %V16I8 = call <16 x i8> @llvm.fshr.v16i8(<16 x i8> %a128, <16 x i8> %b128, <16 x i8> <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>)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 10 for instruction: %V32I8 = call <32 x i8> @llvm.fshr.v32i8(<32 x i8> %a256, <32 x i8> %b256, <32 x i8> <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, 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>)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 10 for instruction: %V64I8 = call <64 x i8> @llvm.fshr.v64i8(<64 x i8> %a512, <64 x i8> %b512, <64 x i8> <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, 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, 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, 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>)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret void
 ;
   %I8    = call i8 @llvm.fshr.i8(i8 %a8, i8 %b8, i8 3)
   %V16I8 = call <16 x i8> @llvm.fshr.v16i8(<16 x i8> %a128, <16 x i8> %b128, <16 x i8> <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>)
@@ -1562,6 +1683,13 @@ define void @var_rotate_i64(i64 %a64, <2 x i64> %a128, <4 x i64> %a256, <8 x i64
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 9 for instruction: %V4I64 = call <4 x i64> @llvm.fshr.v4i64(<4 x i64> %a256, <4 x i64> %a256, <4 x i64> %c256)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 18 for instruction: %V8I64 = call <8 x i64> @llvm.fshr.v8i64(<8 x i64> %a512, <8 x i64> %a512, <8 x i64> %c512)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret void
+;
+; AVX512GFNI-LABEL: 'var_rotate_i64'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %I64 = call i64 @llvm.fshr.i64(i64 %a64, i64 %a64, i64 %c64)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V2I64 = call <2 x i64> @llvm.fshr.v2i64(<2 x i64> %a128, <2 x i64> %a128, <2 x i64> %c128)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V4I64 = call <4 x i64> @llvm.fshr.v4i64(<4 x i64> %a256, <4 x i64> %a256, <4 x i64> %c256)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V8I64 = call <8 x i64> @llvm.fshr.v8i64(<8 x i64> %a512, <8 x i64> %a512, <8 x i64> %c512)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret void
 ;
   %I64    = call i64 @llvm.fshr.i64(i64 %a64, i64 %a64, i64 %c64)
   %V2I64  = call <2 x i64> @llvm.fshr.v2i64(<2 x i64> %a128, <2 x i64> %a128, <2 x i64> %c128)
@@ -1647,6 +1775,13 @@ define void @var_rotate_i32(i32 %a32, <4 x i32> %a128, <8 x i32> %a256, <16 x i3
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 9 for instruction: %V4I32 = call <8 x i32> @llvm.fshr.v8i32(<8 x i32> %a256, <8 x i32> %a256, <8 x i32> %c256)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 18 for instruction: %V8I32 = call <16 x i32> @llvm.fshr.v16i32(<16 x i32> %a512, <16 x i32> %a512, <16 x i32> %c512)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret void
+;
+; AVX512GFNI-LABEL: 'var_rotate_i32'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %I32 = call i32 @llvm.fshr.i32(i32 %a32, i32 %a32, i32 %c32)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V2I32 = call <4 x i32> @llvm.fshr.v4i32(<4 x i32> %a128, <4 x i32> %a128, <4 x i32> %c128)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V4I32 = call <8 x i32> @llvm.fshr.v8i32(<8 x i32> %a256, <8 x i32> %a256, <8 x i32> %c256)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V8I32 = call <16 x i32> @llvm.fshr.v16i32(<16 x i32> %a512, <16 x i32> %a512, <16 x i32> %c512)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret void
 ;
   %I32   = call i32 @llvm.fshr.i32(i32 %a32, i32 %a32, i32 %c32)
   %V2I32 = call <4 x i32> @llvm.fshr.v4i32(<4 x i32> %a128, <4 x i32> %a128, <4 x i32> %c128)
@@ -1732,6 +1867,13 @@ define void @var_rotate_i16(i16 %a16, <8 x i16> %a128, <16 x i16> %a256, <32 x i
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 9 for instruction: %V16I16 = call <16 x i16> @llvm.fshr.v16i16(<16 x i16> %a256, <16 x i16> %a256, <16 x i16> %c256)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 18 for instruction: %V32I16 = call <32 x i16> @llvm.fshr.v32i16(<32 x i16> %a512, <32 x i16> %a512, <32 x i16> %c512)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret void
+;
+; AVX512GFNI-LABEL: 'var_rotate_i16'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %I16 = call i16 @llvm.fshr.i16(i16 %a16, i16 %a16, i16 %c16)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V8I16 = call <8 x i16> @llvm.fshr.v8i16(<8 x i16> %a128, <8 x i16> %a128, <8 x i16> %c128)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V16I16 = call <16 x i16> @llvm.fshr.v16i16(<16 x i16> %a256, <16 x i16> %a256, <16 x i16> %c256)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V32I16 = call <32 x i16> @llvm.fshr.v32i16(<32 x i16> %a512, <32 x i16> %a512, <32 x i16> %c512)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret void
 ;
   %I16    = call i16 @llvm.fshr.i16(i16 %a16, i16 %a16, i16 %c16)
   %V8I16  = call <8 x i16> @llvm.fshr.v8i16(<8 x i16> %a128, <8 x i16> %a128, <8 x i16> %c128)
@@ -1817,6 +1959,13 @@ define void @var_rotate_i8(i8 %a8, <16 x i8> %a128, <32 x i8> %a256, <64 x i8> %
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 9 for instruction: %V32I8 = call <32 x i8> @llvm.fshr.v32i8(<32 x i8> %a256, <32 x i8> %a256, <32 x i8> %c256)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 18 for instruction: %V64I8 = call <64 x i8> @llvm.fshr.v64i8(<64 x i8> %a512, <64 x i8> %a512, <64 x i8> %c512)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret void
+;
+; AVX512GFNI-LABEL: 'var_rotate_i8'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %I8 = call i8 @llvm.fshr.i8(i8 %a8, i8 %a8, i8 %c8)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 9 for instruction: %V16I8 = call <16 x i8> @llvm.fshr.v16i8(<16 x i8> %a128, <16 x i8> %a128, <16 x i8> %c128)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 9 for instruction: %V32I8 = call <32 x i8> @llvm.fshr.v32i8(<32 x i8> %a256, <32 x i8> %a256, <32 x i8> %c256)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 14 for instruction: %V64I8 = call <64 x i8> @llvm.fshr.v64i8(<64 x i8> %a512, <64 x i8> %a512, <64 x i8> %c512)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret void
 ;
   %I8    = call i8 @llvm.fshr.i8(i8 %a8, i8 %a8, i8 %c8)
   %V16I8 = call <16 x i8> @llvm.fshr.v16i8(<16 x i8> %a128, <16 x i8> %a128, <16 x i8> %c128)
@@ -1919,6 +2068,15 @@ define void @splatvar_rotate_i64(i64 %a64, <2 x i64> %a128, <4 x i64> %a256, <8
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 9 for instruction: %V4I64 = call <4 x i64> @llvm.fshr.v4i64(<4 x i64> %a256, <4 x i64> %a256, <4 x i64> %u256)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 18 for instruction: %V8I64 = call <8 x i64> @llvm.fshr.v8i64(<8 x i64> %a512, <8 x i64> %a512, <8 x i64> %u512)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret void
+;
+; AVX512GFNI-LABEL: 'splatvar_rotate_i64'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %u128 = shufflevector <2 x i64> %c128, <2 x i64> undef, <2 x i32> zeroinitializer
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %u256 = shufflevector <4 x i64> %c256, <4 x i64> undef, <4 x i32> zeroinitializer
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %u512 = shufflevector <8 x i64> %c512, <8 x i64> undef, <8 x i32> zeroinitializer
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V2I64 = call <2 x i64> @llvm.fshr.v2i64(<2 x i64> %a128, <2 x i64> %a128, <2 x i64> %u128)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V4I64 = call <4 x i64> @llvm.fshr.v4i64(<4 x i64> %a256, <4 x i64> %a256, <4 x i64> %u256)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V8I64 = call <8 x i64> @llvm.fshr.v8i64(<8 x i64> %a512, <8 x i64> %a512, <8 x i64> %u512)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret void
 ;
   %u128 = shufflevector <2 x i64> %c128, <2 x i64> undef, <2 x i32> zeroinitializer
   %u256 = shufflevector <4 x i64> %c256, <4 x i64> undef, <4 x i32> zeroinitializer
@@ -2019,6 +2177,15 @@ define void @splatvar_rotate_i32(i32 %a32, <4 x i32> %a128, <8 x i32> %a256, <16
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 9 for instruction: %V4I32 = call <8 x i32> @llvm.fshr.v8i32(<8 x i32> %a256, <8 x i32> %a256, <8 x i32> %u256)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 18 for instruction: %V8I32 = call <16 x i32> @llvm.fshr.v16i32(<16 x i32> %a512, <16 x i32> %a512, <16 x i32> %u512)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret void
+;
+; AVX512GFNI-LABEL: 'splatvar_rotate_i32'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %u128 = shufflevector <4 x i32> %c128, <4 x i32> undef, <4 x i32> zeroinitializer
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %u256 = shufflevector <8 x i32> %c256, <8 x i32> undef, <8 x i32> zeroinitializer
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %u512 = shufflevector <16 x i32> %c512, <16 x i32> undef, <16 x i32> zeroinitializer
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V2I32 = call <4 x i32> @llvm.fshr.v4i32(<4 x i32> %a128, <4 x i32> %a128, <4 x i32> %u128)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V4I32 = call <8 x i32> @llvm.fshr.v8i32(<8 x i32> %a256, <8 x i32> %a256, <8 x i32> %u256)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V8I32 = call <16 x i32> @llvm.fshr.v16i32(<16 x i32> %a512, <16 x i32> %a512, <16 x i32> %u512)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret void
 ;
   %u128 = shufflevector <4 x i32> %c128, <4 x i32> undef, <4 x i32> zeroinitializer
   %u256 = shufflevector <8 x i32> %c256, <8 x i32> undef, <8 x i32> zeroinitializer
@@ -2119,6 +2286,15 @@ define void @splatvar_rotate_i16(i16 %a16, <8 x i16> %a128, <16 x i16> %a256, <3
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 9 for instruction: %V16I16 = call <16 x i16> @llvm.fshr.v16i16(<16 x i16> %a256, <16 x i16> %a256, <16 x i16> %u256)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 18 for instruction: %V32I16 = call <32 x i16> @llvm.fshr.v32i16(<32 x i16> %a512, <32 x i16> %a512, <32 x i16> %u512)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret void
+;
+; AVX512GFNI-LABEL: 'splatvar_rotate_i16'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %u128 = shufflevector <8 x i16> %c128, <8 x i16> undef, <8 x i32> zeroinitializer
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %u256 = shufflevector <16 x i16> %c256, <16 x i16> undef, <16 x i32> zeroinitializer
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %u512 = shufflevector <32 x i16> %c512, <32 x i16> undef, <32 x i32> zeroinitializer
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V8I16 = call <8 x i16> @llvm.fshr.v8i16(<8 x i16> %a128, <8 x i16> %a128, <8 x i16> %u128)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V16I16 = call <16 x i16> @llvm.fshr.v16i16(<16 x i16> %a256, <16 x i16> %a256, <16 x i16> %u256)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V32I16 = call <32 x i16> @llvm.fshr.v32i16(<32 x i16> %a512, <32 x i16> %a512, <32 x i16> %u512)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret void
 ;
   %u128 = shufflevector <8 x i16> %c128, <8 x i16> undef, <8 x i32> zeroinitializer
   %u256 = shufflevector <16 x i16> %c256, <16 x i16> undef, <16 x i32> zeroinitializer
@@ -2219,6 +2395,15 @@ define void @splatvar_rotate_i8(i8 %a8, <16 x i8> %a128, <32 x i8> %a256, <64 x
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 9 for instruction: %V32I8 = call <32 x i8> @llvm.fshr.v32i8(<32 x i8> %a256, <32 x i8> %a256, <32 x i8> %u256)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 18 for instruction: %V64I8 = call <64 x i8> @llvm.fshr.v64i8(<64 x i8> %a512, <64 x i8> %a512, <64 x i8> %u512)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret void
+;
+; AVX512GFNI-LABEL: 'splatvar_rotate_i8'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %u128 = shufflevector <16 x i8> %c128, <16 x i8> undef, <16 x i32> zeroinitializer
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %u256 = shufflevector <32 x i8> %c256, <32 x i8> undef, <32 x i32> zeroinitializer
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %u512 = shufflevector <64 x i8> %c512, <64 x i8> undef, <64 x i32> zeroinitializer
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 9 for instruction: %V16I8 = call <16 x i8> @llvm.fshr.v16i8(<16 x i8> %a128, <16 x i8> %a128, <16 x i8> %u128)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 9 for instruction: %V32I8 = call <32 x i8> @llvm.fshr.v32i8(<32 x i8> %a256, <32 x i8> %a256, <32 x i8> %u256)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 14 for instruction: %V64I8 = call <64 x i8> @llvm.fshr.v64i8(<64 x i8> %a512, <64 x i8> %a512, <64 x i8> %u512)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret void
 ;
   %u128 = shufflevector <16 x i8> %c128, <16 x i8> undef, <16 x i32> zeroinitializer
   %u256 = shufflevector <32 x i8> %c256, <32 x i8> undef, <32 x i32> zeroinitializer
@@ -2303,6 +2488,13 @@ define void @constant_rotate_i64(i64 %a64, <2 x i64> %a128, <4 x i64> %a256, <8
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 9 for instruction: %V4I64 = call <4 x i64> @llvm.fshr.v4i64(<4 x i64> %a256, <4 x i64> %a256, <4 x i64> <i64 1, i64 7, i64 15, i64 31>)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 18 for instruction: %V8I64 = call <8 x i64> @llvm.fshr.v8i64(<8 x i64> %a512, <8 x i64> %a512, <8 x i64> <i64 1, i64 7, i64 15, i64 31, i64 1, i64 7, i64 15, i64 31>)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret void
+;
+; AVX512GFNI-LABEL: 'constant_rotate_i64'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %I64 = call i64 @llvm.fshr.i64(i64 %a64, i64 %a64, i64 7)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V2I64 = call <2 x i64> @llvm.fshr.v2i64(<2 x i64> %a128, <2 x i64> %a128, <2 x i64> <i64 1, i64 7>)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V4I64 = call <4 x i64> @llvm.fshr.v4i64(<4 x i64> %a256, <4 x i64> %a256, <4 x i64> <i64 1, i64 7, i64 15, i64 31>)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V8I64 = call <8 x i64> @llvm.fshr.v8i64(<8 x i64> %a512, <8 x i64> %a512, <8 x i64> <i64 1, i64 7, i64 15, i64 31, i64 1, i64 7, i64 15, i64 31>)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret void
 ;
   %I64    = call i64 @llvm.fshr.i64(i64 %a64, i64 %a64, i64 7)
   %V2I64  = call <2 x i64> @llvm.fshr.v2i64(<2 x i64> %a128, <2 x i64> %a128, <2 x i64> <i64 1, i64 7>)
@@ -2388,6 +2580,13 @@ define void @constant_rotate_i32(i32 %a32, <4 x i32> %a128, <8 x i32> %a256, <16
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 9 for instruction: %V4I32 = call <8 x i32> @llvm.fshr.v8i32(<8 x i32> %a256, <8 x i32> %a256, <8 x i32> <i32 4, i32 5, i32 6, i32 7, i32 0, i32 1, i32 2, i32 3>)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 18 for instruction: %V8I32 = call <16 x i32> @llvm.fshr.v16i32(<16 x i32> %a512, <16 x i32> %a512, <16 x i32> <i32 4, i32 5, i32 6, i32 7, i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 0, i32 1, i32 2, i32 3>)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret void
+;
+; AVX512GFNI-LABEL: 'constant_rotate_i32'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %I32 = call i32 @llvm.fshr.i32(i32 %a32, i32 %a32, i32 7)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V2I32 = call <4 x i32> @llvm.fshr.v4i32(<4 x i32> %a128, <4 x i32> %a128, <4 x i32> <i32 4, i32 5, i32 6, i32 7>)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V4I32 = call <8 x i32> @llvm.fshr.v8i32(<8 x i32> %a256, <8 x i32> %a256, <8 x i32> <i32 4, i32 5, i32 6, i32 7, i32 0, i32 1, i32 2, i32 3>)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V8I32 = call <16 x i32> @llvm.fshr.v16i32(<16 x i32> %a512, <16 x i32> %a512, <16 x i32> <i32 4, i32 5, i32 6, i32 7, i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 0, i32 1, i32 2, i32 3>)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret void
 ;
   %I32   = call i32 @llvm.fshr.i32(i32 %a32, i32 %a32, i32 7)
   %V2I32 = call <4 x i32> @llvm.fshr.v4i32(<4 x i32> %a128, <4 x i32> %a128, <4 x i32> <i32 4, i32 5, i32 6, i32 7>)
@@ -2473,6 +2672,13 @@ define void @constant_rotate_i16(i16 %a16, <8 x i16> %a128, <16 x i16> %a256, <3
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 9 for instruction: %V16I16 = call <16 x i16> @llvm.fshr.v16i16(<16 x i16> %a256, <16 x i16> %a256, <16 x i16> <i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7>)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 18 for instruction: %V32I16 = call <32 x i16> @llvm.fshr.v32i16(<32 x i16> %a512, <32 x i16> %a512, <32 x i16> <i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7>)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret void
+;
+; AVX512GFNI-LABEL: 'constant_rotate_i16'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %I16 = call i16 @llvm.fshr.i16(i16 %a16, i16 %a16, i16 7)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V8I16 = call <8 x i16> @llvm.fshr.v8i16(<8 x i16> %a128, <8 x i16> %a128, <8 x i16> <i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7>)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V16I16 = call <16 x i16> @llvm.fshr.v16i16(<16 x i16> %a256, <16 x i16> %a256, <16 x i16> <i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7>)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V32I16 = call <32 x i16> @llvm.fshr.v32i16(<32 x i16> %a512, <32 x i16> %a512, <32 x i16> <i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7>)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret void
 ;
   %I16    = call i16 @llvm.fshr.i16(i16 %a16, i16 %a16, i16 7)
   %V8I16  = call <8 x i16> @llvm.fshr.v8i16(<8 x i16> %a128, <8 x i16> %a128, <8 x i16> <i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7>)
@@ -2558,6 +2764,13 @@ define void @constant_rotate_i8(i8 %a8, <16 x i8> %a128, <32 x i8> %a256, <64 x
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 9 for instruction: %V32I8 = call <32 x i8> @llvm.fshr.v32i8(<32 x i8> %a256, <32 x i8> %a256, <32 x i8> <i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0>)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 18 for instruction: %V64I8 = call <64 x i8> @llvm.fshr.v64i8(<64 x i8> %a512, <64 x i8> %a512, <64 x i8> <i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0>)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret void
+;
+; AVX512GFNI-LABEL: 'constant_rotate_i8'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %I8 = call i8 @llvm.fshr.i8(i8 %a8, i8 %a8, i8 7)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 9 for instruction: %V16I8 = call <16 x i8> @llvm.fshr.v16i8(<16 x i8> %a128, <16 x i8> %a128, <16 x i8> <i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0>)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 9 for instruction: %V32I8 = call <32 x i8> @llvm.fshr.v32i8(<32 x i8> %a256, <32 x i8> %a256, <32 x i8> <i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0>)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 14 for instruction: %V64I8 = call <64 x i8> @llvm.fshr.v64i8(<64 x i8> %a512, <64 x i8> %a512, <64 x i8> <i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0>)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret void
 ;
   %I8    = call i8 @llvm.fshr.i8(i8 %a8, i8 %a8, i8 7)
   %V16I8 = call <16 x i8> @llvm.fshr.v16i8(<16 x i8> %a128, <16 x i8> %a128, <16 x i8> <i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0>)
@@ -2640,6 +2853,13 @@ define void @splatconstant_rotate_i64(i64 %a64, <2 x i64> %a128, <4 x i64> %a256
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 9 for instruction: %V4I64 = call <4 x i64> @llvm.fshr.v4i64(<4 x i64> %a256, <4 x i64> %a256, <4 x i64> <i64 7, i64 7, i64 7, i64 7>)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 18 for instruction: %V8I64 = call <8 x i64> @llvm.fshr.v8i64(<8 x i64> %a512, <8 x i64> %a512, <8 x i64> <i64 7, i64 7, i64 7, i64 7, i64 7, i64 7, i64 7, i64 7>)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret void
+;
+; AVX512GFNI-LABEL: 'splatconstant_rotate_i64'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %I64 = call i64 @llvm.fshr.i64(i64 %a64, i64 %a64, i64 7)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V2I64 = call <2 x i64> @llvm.fshr.v2i64(<2 x i64> %a128, <2 x i64> %a128, <2 x i64> <i64 7, i64 7>)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V4I64 = call <4 x i64> @llvm.fshr.v4i64(<4 x i64> %a256, <4 x i64> %a256, <4 x i64> <i64 7, i64 7, i64 7, i64 7>)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V8I64 = call <8 x i64> @llvm.fshr.v8i64(<8 x i64> %a512, <8 x i64> %a512, <8 x i64> <i64 7, i64 7, i64 7, i64 7, i64 7, i64 7, i64 7, i64 7>)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret void
 ;
   %I64    = call i64 @llvm.fshr.i64(i64 %a64, i64 %a64, i64 7)
   %V2I64  = call <2 x i64> @llvm.fshr.v2i64(<2 x i64> %a128, <2 x i64> %a128, <2 x i64> <i64 7, i64 7>)
@@ -2718,6 +2938,13 @@ define void @splatconstant_rotate_i32(i32 %a32, <4 x i32> %a128, <8 x i32> %a256
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 9 for instruction: %V4I32 = call <8 x i32> @llvm.fshr.v8i32(<8 x i32> %a256, <8 x i32> %a256, <8 x i32> <i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5>)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 18 for instruction: %V8I32 = call <16 x i32> @llvm.fshr.v16i32(<16 x i32> %a512, <16 x i32> %a512, <16 x i32> <i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5>)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret void
+;
+; AVX512GFNI-LABEL: 'splatconstant_rotate_i32'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %I32 = call i32 @llvm.fshr.i32(i32 %a32, i32 %a32, i32 5)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V2I32 = call <4 x i32> @llvm.fshr.v4i32(<4 x i32> %a128, <4 x i32> %a128, <4 x i32> <i32 5, i32 5, i32 5, i32 5>)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V4I32 = call <8 x i32> @llvm.fshr.v8i32(<8 x i32> %a256, <8 x i32> %a256, <8 x i32> <i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5>)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V8I32 = call <16 x i32> @llvm.fshr.v16i32(<16 x i32> %a512, <16 x i32> %a512, <16 x i32> <i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5>)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret void
 ;
   %I32   = call i32 @llvm.fshr.i32(i32 %a32, i32 %a32, i32 5)
   %V2I32 = call <4 x i32> @llvm.fshr.v4i32(<4 x i32> %a128, <4 x i32> %a128, <4 x i32> <i32 5, i32 5, i32 5, i32 5>)
@@ -2796,6 +3023,13 @@ define void @splatconstant_rotate_i16(i16 %a16, <8 x i16> %a128, <16 x i16> %a25
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 9 for instruction: %V16I16 = call <16 x i16> @llvm.fshr.v16i16(<16 x i16> %a256, <16 x i16> %a256, <16 x i16> <i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3>)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 18 for instruction: %V32I16 = call <32 x i16> @llvm.fshr.v32i16(<32 x i16> %a512, <32 x i16> %a512, <32 x i16> <i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3>)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret void
+;
+; AVX512GFNI-LABEL: 'splatconstant_rotate_i16'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %I16 = call i16 @llvm.fshr.i16(i16 %a16, i16 %a16, i16 3)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V8I16 = call <8 x i16> @llvm.fshr.v8i16(<8 x i16> %a128, <8 x i16> %a128, <8 x i16> <i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3>)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V16I16 = call <16 x i16> @llvm.fshr.v16i16(<16 x i16> %a256, <16 x i16> %a256, <16 x i16> <i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3>)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V32I16 = call <32 x i16> @llvm.fshr.v32i16(<32 x i16> %a512, <32 x i16> %a512, <32 x i16> <i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3>)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret void
 ;
   %I16    = call i16 @llvm.fshr.i16(i16 %a16, i16 %a16, i16 3)
   %V8I16  = call <8 x i16> @llvm.fshr.v8i16(<8 x i16> %a128, <8 x i16> %a128, <8 x i16> <i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3>)
@@ -2874,6 +3108,13 @@ define void @splatconstant_rotate_i8(i8 %a8, <16 x i8> %a128, <32 x i8> %a256, <
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 9 for instruction: %V32I8 = call <32 x i8> @llvm.fshr.v32i8(<32 x i8> %a256, <32 x i8> %a256, <32 x i8> <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, 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>)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 18 for instruction: %V64I8 = call <64 x i8> @llvm.fshr.v64i8(<64 x i8> %a512, <64 x i8> %a512, <64 x i8> <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, 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, 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, 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>)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret void
+;
+; AVX512GFNI-LABEL: 'splatconstant_rotate_i8'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %I8 = call i8 @llvm.fshr.i8(i8 %a8, i8 %a8, i8 3)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 9 for instruction: %V16I8 = call <16 x i8> @llvm.fshr.v16i8(<16 x i8> %a128, <16 x i8> %a128, <16 x i8> <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>)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 9 for instruction: %V32I8 = call <32 x i8> @llvm.fshr.v32i8(<32 x i8> %a256, <32 x i8> %a256, <32 x i8> <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, 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>)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 14 for instruction: %V64I8 = call <64 x i8> @llvm.fshr.v64i8(<64 x i8> %a512, <64 x i8> %a512, <64 x i8> <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, 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, 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, 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>)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret void
 ;
   %I8    = call i8 @llvm.fshr.i8(i8 %a8, i8 %a8, i8 3)
   %V16I8 = call <16 x i8> @llvm.fshr.v16i8(<16 x i8> %a128, <16 x i8> %a128, <16 x i8> <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>)

diff  --git a/llvm/test/Analysis/CostModel/X86/fshr.ll b/llvm/test/Analysis/CostModel/X86/fshr.ll
index 9565630677d52c..ca9ddcc52938d7 100644
--- a/llvm/test/Analysis/CostModel/X86/fshr.ll
+++ b/llvm/test/Analysis/CostModel/X86/fshr.ll
@@ -12,6 +12,7 @@
 ; RUN: opt < %s -passes="print<cost-model>" 2>&1 -disable-output -mtriple=x86_64-apple-macosx10.8.0 -mcpu=goldmont | FileCheck %s --check-prefixes=GLM
 ; RUN: opt < %s -passes="print<cost-model>" 2>&1 -disable-output -mtriple=x86_64-apple-macosx10.8.0 -mcpu=bdver2 | FileCheck %s --check-prefixes=XOP
 ; RUN: opt < %s -passes="print<cost-model>" 2>&1 -disable-output -mtriple=x86_64-apple-macosx10.8.0 -mcpu=btver2 | FileCheck %s --check-prefixes=AVX1
+; RUN: opt < %s -passes="print<cost-model>" 2>&1 -disable-output -mtriple=x86_64-apple-macosx10.8.0 -mcpu=tigerlake | FileCheck %s --check-prefixes=AVX512,AVX512GFNI
 
 target datalayout = "e-p:64:64:64-i1:8:8-i8:8:8-i16:16:16-i32:32:32-i64:64:64-f32:32:32-f64:64:64-v64:64:64-v128:128:128-a0:0:64-s0:64:64-f80:128:128-n8:16:32:64-S128"
 target triple = "x86_64-apple-macosx10.8.0"
@@ -97,6 +98,13 @@ define void @var_funnel_i64(i64 %a64, <2 x i64> %a128, <4 x i64> %a256, <8 x i64
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 23 for instruction: %V4I64 = call <4 x i64> @llvm.fshr.v4i64(<4 x i64> %a256, <4 x i64> %b256, <4 x i64> %c256)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 46 for instruction: %V8I64 = call <8 x i64> @llvm.fshr.v8i64(<8 x i64> %a512, <8 x i64> %b512, <8 x i64> %c512)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret void
+;
+; AVX512GFNI-LABEL: 'var_funnel_i64'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %I64 = call i64 @llvm.fshr.i64(i64 %a64, i64 %b64, i64 %c64)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V2I64 = call <2 x i64> @llvm.fshr.v2i64(<2 x i64> %a128, <2 x i64> %b128, <2 x i64> %c128)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V4I64 = call <4 x i64> @llvm.fshr.v4i64(<4 x i64> %a256, <4 x i64> %b256, <4 x i64> %c256)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V8I64 = call <8 x i64> @llvm.fshr.v8i64(<8 x i64> %a512, <8 x i64> %b512, <8 x i64> %c512)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret void
 ;
   %I64    = call i64 @llvm.fshr.i64(i64 %a64, i64 %b64, i64 %c64)
   %V2I64  = call <2 x i64> @llvm.fshr.v2i64(<2 x i64> %a128, <2 x i64> %b128, <2 x i64> %c128)
@@ -182,6 +190,13 @@ define void @var_funnel_i32(i32 %a32, <4 x i32> %a128, <8 x i32> %a256, <16 x i3
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 23 for instruction: %V4I32 = call <8 x i32> @llvm.fshr.v8i32(<8 x i32> %a256, <8 x i32> %b256, <8 x i32> %c256)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 46 for instruction: %V8I32 = call <16 x i32> @llvm.fshr.v16i32(<16 x i32> %a512, <16 x i32> %b512, <16 x i32> %c512)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret void
+;
+; AVX512GFNI-LABEL: 'var_funnel_i32'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %I32 = call i32 @llvm.fshr.i32(i32 %a32, i32 %b32, i32 %c32)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V2I32 = call <4 x i32> @llvm.fshr.v4i32(<4 x i32> %a128, <4 x i32> %b128, <4 x i32> %c128)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V4I32 = call <8 x i32> @llvm.fshr.v8i32(<8 x i32> %a256, <8 x i32> %b256, <8 x i32> %c256)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V8I32 = call <16 x i32> @llvm.fshr.v16i32(<16 x i32> %a512, <16 x i32> %b512, <16 x i32> %c512)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret void
 ;
   %I32   = call i32 @llvm.fshr.i32(i32 %a32, i32 %b32, i32 %c32)
   %V2I32 = call <4 x i32> @llvm.fshr.v4i32(<4 x i32> %a128, <4 x i32> %b128, <4 x i32> %c128)
@@ -267,6 +282,13 @@ define void @var_funnel_i16(i16 %a16, <8 x i16> %a128, <16 x i16> %a256, <32 x i
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 23 for instruction: %V16I16 = call <16 x i16> @llvm.fshr.v16i16(<16 x i16> %a256, <16 x i16> %b256, <16 x i16> %c256)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 46 for instruction: %V32I16 = call <32 x i16> @llvm.fshr.v32i16(<32 x i16> %a512, <32 x i16> %b512, <32 x i16> %c512)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret void
+;
+; AVX512GFNI-LABEL: 'var_funnel_i16'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %I16 = call i16 @llvm.fshr.i16(i16 %a16, i16 %b16, i16 %c16)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V8I16 = call <8 x i16> @llvm.fshr.v8i16(<8 x i16> %a128, <8 x i16> %b128, <8 x i16> %c128)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V16I16 = call <16 x i16> @llvm.fshr.v16i16(<16 x i16> %a256, <16 x i16> %b256, <16 x i16> %c256)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V32I16 = call <32 x i16> @llvm.fshr.v32i16(<32 x i16> %a512, <32 x i16> %b512, <32 x i16> %c512)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret void
 ;
   %I16    = call i16 @llvm.fshr.i16(i16 %a16, i16 %b16, i16 %c16)
   %V8I16  = call <8 x i16> @llvm.fshr.v8i16(<8 x i16> %a128, <8 x i16> %b128, <8 x i16> %c128)
@@ -352,6 +374,13 @@ define void @var_funnel_i8(i8 %a8, <16 x i8> %a128, <32 x i8> %a256, <64 x i8> %
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 23 for instruction: %V32I8 = call <32 x i8> @llvm.fshr.v32i8(<32 x i8> %a256, <32 x i8> %b256, <32 x i8> %c256)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 46 for instruction: %V64I8 = call <64 x i8> @llvm.fshr.v64i8(<64 x i8> %a512, <64 x i8> %b512, <64 x i8> %c512)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret void
+;
+; AVX512GFNI-LABEL: 'var_funnel_i8'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %I8 = call i8 @llvm.fshr.i8(i8 %a8, i8 %b8, i8 %c8)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 13 for instruction: %V16I8 = call <16 x i8> @llvm.fshr.v16i8(<16 x i8> %a128, <16 x i8> %b128, <16 x i8> %c128)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 13 for instruction: %V32I8 = call <32 x i8> @llvm.fshr.v32i8(<32 x i8> %a256, <32 x i8> %b256, <32 x i8> %c256)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 18 for instruction: %V64I8 = call <64 x i8> @llvm.fshr.v64i8(<64 x i8> %a512, <64 x i8> %b512, <64 x i8> %c512)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret void
 ;
   %I8    = call i8 @llvm.fshr.i8(i8 %a8, i8 %b8, i8 %c8)
   %V16I8 = call <16 x i8> @llvm.fshr.v16i8(<16 x i8> %a128, <16 x i8> %b128, <16 x i8> %c128)
@@ -463,6 +492,15 @@ define void @splatvar_funnel_i64(i64 %a64, <2 x i64> %a128, <4 x i64> %a256, <8
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 19 for instruction: %V4I64 = call <4 x i64> @llvm.fshr.v4i64(<4 x i64> %a256, <4 x i64> %b256, <4 x i64> %u256)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 38 for instruction: %V8I64 = call <8 x i64> @llvm.fshr.v8i64(<8 x i64> %a512, <8 x i64> %b512, <8 x i64> %u512)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret void
+;
+; AVX512GFNI-LABEL: 'splatvar_funnel_i64'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %u128 = shufflevector <2 x i64> %c128, <2 x i64> undef, <2 x i32> zeroinitializer
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %u256 = shufflevector <4 x i64> %c256, <4 x i64> undef, <4 x i32> zeroinitializer
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %u512 = shufflevector <8 x i64> %c512, <8 x i64> undef, <8 x i32> zeroinitializer
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V2I64 = call <2 x i64> @llvm.fshr.v2i64(<2 x i64> %a128, <2 x i64> %b128, <2 x i64> %u128)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V4I64 = call <4 x i64> @llvm.fshr.v4i64(<4 x i64> %a256, <4 x i64> %b256, <4 x i64> %u256)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V8I64 = call <8 x i64> @llvm.fshr.v8i64(<8 x i64> %a512, <8 x i64> %b512, <8 x i64> %u512)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret void
 ;
   %u128 = shufflevector <2 x i64> %c128, <2 x i64> undef, <2 x i32> zeroinitializer
   %u256 = shufflevector <4 x i64> %c256, <4 x i64> undef, <4 x i32> zeroinitializer
@@ -563,6 +601,15 @@ define void @splatvar_funnel_i32(i32 %a32, <4 x i32> %a128, <8 x i32> %a256, <16
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 19 for instruction: %V4I32 = call <8 x i32> @llvm.fshr.v8i32(<8 x i32> %a256, <8 x i32> %b256, <8 x i32> %u256)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 38 for instruction: %V8I32 = call <16 x i32> @llvm.fshr.v16i32(<16 x i32> %a512, <16 x i32> %b512, <16 x i32> %u512)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret void
+;
+; AVX512GFNI-LABEL: 'splatvar_funnel_i32'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %u128 = shufflevector <4 x i32> %c128, <4 x i32> undef, <4 x i32> zeroinitializer
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %u256 = shufflevector <8 x i32> %c256, <8 x i32> undef, <8 x i32> zeroinitializer
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %u512 = shufflevector <16 x i32> %c512, <16 x i32> undef, <16 x i32> zeroinitializer
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V2I32 = call <4 x i32> @llvm.fshr.v4i32(<4 x i32> %a128, <4 x i32> %b128, <4 x i32> %u128)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V4I32 = call <8 x i32> @llvm.fshr.v8i32(<8 x i32> %a256, <8 x i32> %b256, <8 x i32> %u256)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V8I32 = call <16 x i32> @llvm.fshr.v16i32(<16 x i32> %a512, <16 x i32> %b512, <16 x i32> %u512)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret void
 ;
   %u128 = shufflevector <4 x i32> %c128, <4 x i32> undef, <4 x i32> zeroinitializer
   %u256 = shufflevector <8 x i32> %c256, <8 x i32> undef, <8 x i32> zeroinitializer
@@ -663,6 +710,15 @@ define void @splatvar_funnel_i16(i16 %a16, <8 x i16> %a128, <16 x i16> %a256, <3
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 19 for instruction: %V16I16 = call <16 x i16> @llvm.fshr.v16i16(<16 x i16> %a256, <16 x i16> %b256, <16 x i16> %u256)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 38 for instruction: %V32I16 = call <32 x i16> @llvm.fshr.v32i16(<32 x i16> %a512, <32 x i16> %b512, <32 x i16> %u512)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret void
+;
+; AVX512GFNI-LABEL: 'splatvar_funnel_i16'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %u128 = shufflevector <8 x i16> %c128, <8 x i16> undef, <8 x i32> zeroinitializer
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %u256 = shufflevector <16 x i16> %c256, <16 x i16> undef, <16 x i32> zeroinitializer
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %u512 = shufflevector <32 x i16> %c512, <32 x i16> undef, <32 x i32> zeroinitializer
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V8I16 = call <8 x i16> @llvm.fshr.v8i16(<8 x i16> %a128, <8 x i16> %b128, <8 x i16> %u128)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V16I16 = call <16 x i16> @llvm.fshr.v16i16(<16 x i16> %a256, <16 x i16> %b256, <16 x i16> %u256)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V32I16 = call <32 x i16> @llvm.fshr.v32i16(<32 x i16> %a512, <32 x i16> %b512, <32 x i16> %u512)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret void
 ;
   %u128 = shufflevector <8 x i16> %c128, <8 x i16> undef, <8 x i32> zeroinitializer
   %u256 = shufflevector <16 x i16> %c256, <16 x i16> undef, <16 x i32> zeroinitializer
@@ -763,6 +819,15 @@ define void @splatvar_funnel_i8(i8 %a8, <16 x i8> %a128, <32 x i8> %a256, <64 x
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 23 for instruction: %V32I8 = call <32 x i8> @llvm.fshr.v32i8(<32 x i8> %a256, <32 x i8> %b256, <32 x i8> %u256)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 46 for instruction: %V64I8 = call <64 x i8> @llvm.fshr.v64i8(<64 x i8> %a512, <64 x i8> %b512, <64 x i8> %u512)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret void
+;
+; AVX512GFNI-LABEL: 'splatvar_funnel_i8'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %u128 = shufflevector <16 x i8> %c128, <16 x i8> undef, <16 x i32> zeroinitializer
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %u256 = shufflevector <32 x i8> %c256, <32 x i8> undef, <32 x i32> zeroinitializer
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %u512 = shufflevector <64 x i8> %c512, <64 x i8> undef, <64 x i32> zeroinitializer
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 11 for instruction: %V16I8 = call <16 x i8> @llvm.fshr.v16i8(<16 x i8> %a128, <16 x i8> %b128, <16 x i8> %u128)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 13 for instruction: %V32I8 = call <32 x i8> @llvm.fshr.v32i8(<32 x i8> %a256, <32 x i8> %b256, <32 x i8> %u256)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 13 for instruction: %V64I8 = call <64 x i8> @llvm.fshr.v64i8(<64 x i8> %a512, <64 x i8> %b512, <64 x i8> %u512)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret void
 ;
   %u128 = shufflevector <16 x i8> %c128, <16 x i8> undef, <16 x i32> zeroinitializer
   %u256 = shufflevector <32 x i8> %c256, <32 x i8> undef, <32 x i32> zeroinitializer
@@ -854,6 +919,13 @@ define void @constant_funnel_i64(i64 %a64, <2 x i64> %a128, <4 x i64> %a256, <8
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 20 for instruction: %V4I64 = call <4 x i64> @llvm.fshr.v4i64(<4 x i64> %a256, <4 x i64> %b256, <4 x i64> <i64 1, i64 7, i64 15, i64 31>)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 40 for instruction: %V8I64 = call <8 x i64> @llvm.fshr.v8i64(<8 x i64> %a512, <8 x i64> %b512, <8 x i64> <i64 1, i64 7, i64 15, i64 31, i64 1, i64 7, i64 15, i64 31>)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret void
+;
+; AVX512GFNI-LABEL: 'constant_funnel_i64'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %I64 = call i64 @llvm.fshr.i64(i64 %a64, i64 %b64, i64 7)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V2I64 = call <2 x i64> @llvm.fshr.v2i64(<2 x i64> %a128, <2 x i64> %b128, <2 x i64> <i64 1, i64 7>)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V4I64 = call <4 x i64> @llvm.fshr.v4i64(<4 x i64> %a256, <4 x i64> %b256, <4 x i64> <i64 1, i64 7, i64 15, i64 31>)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V8I64 = call <8 x i64> @llvm.fshr.v8i64(<8 x i64> %a512, <8 x i64> %b512, <8 x i64> <i64 1, i64 7, i64 15, i64 31, i64 1, i64 7, i64 15, i64 31>)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret void
 ;
   %I64    = call i64 @llvm.fshr.i64(i64 %a64, i64 %b64, i64 7)
   %V2I64  = call <2 x i64> @llvm.fshr.v2i64(<2 x i64> %a128, <2 x i64> %b128, <2 x i64> <i64 1, i64 7>)
@@ -932,6 +1004,13 @@ define void @constant_funnel_i32(i32 %a32, <4 x i32> %a128, <8 x i32> %a256, <16
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 20 for instruction: %V4I32 = call <8 x i32> @llvm.fshr.v8i32(<8 x i32> %a256, <8 x i32> %b256, <8 x i32> <i32 4, i32 5, i32 6, i32 7, i32 0, i32 1, i32 2, i32 3>)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 40 for instruction: %V8I32 = call <16 x i32> @llvm.fshr.v16i32(<16 x i32> %a512, <16 x i32> %b512, <16 x i32> <i32 4, i32 5, i32 6, i32 7, i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 0, i32 1, i32 2, i32 3>)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret void
+;
+; AVX512GFNI-LABEL: 'constant_funnel_i32'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %I32 = call i32 @llvm.fshr.i32(i32 %a32, i32 %b32, i32 7)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V2I32 = call <4 x i32> @llvm.fshr.v4i32(<4 x i32> %a128, <4 x i32> %b128, <4 x i32> <i32 4, i32 5, i32 6, i32 7>)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V4I32 = call <8 x i32> @llvm.fshr.v8i32(<8 x i32> %a256, <8 x i32> %b256, <8 x i32> <i32 4, i32 5, i32 6, i32 7, i32 0, i32 1, i32 2, i32 3>)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V8I32 = call <16 x i32> @llvm.fshr.v16i32(<16 x i32> %a512, <16 x i32> %b512, <16 x i32> <i32 4, i32 5, i32 6, i32 7, i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 0, i32 1, i32 2, i32 3>)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret void
 ;
   %I32   = call i32 @llvm.fshr.i32(i32 %a32, i32 %b32, i32 7)
   %V2I32 = call <4 x i32> @llvm.fshr.v4i32(<4 x i32> %a128, <4 x i32> %b128, <4 x i32> <i32 4, i32 5, i32 6, i32 7>)
@@ -1017,6 +1096,13 @@ define void @constant_funnel_i16(i16 %a16, <8 x i16> %a128, <16 x i16> %a256, <3
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 20 for instruction: %V16I16 = call <16 x i16> @llvm.fshr.v16i16(<16 x i16> %a256, <16 x i16> %b256, <16 x i16> <i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7>)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 40 for instruction: %V32I16 = call <32 x i16> @llvm.fshr.v32i16(<32 x i16> %a512, <32 x i16> %b512, <32 x i16> <i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7>)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret void
+;
+; AVX512GFNI-LABEL: 'constant_funnel_i16'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %I16 = call i16 @llvm.fshr.i16(i16 %a16, i16 %b16, i16 7)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V8I16 = call <8 x i16> @llvm.fshr.v8i16(<8 x i16> %a128, <8 x i16> %b128, <8 x i16> <i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7>)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V16I16 = call <16 x i16> @llvm.fshr.v16i16(<16 x i16> %a256, <16 x i16> %b256, <16 x i16> <i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7>)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V32I16 = call <32 x i16> @llvm.fshr.v32i16(<32 x i16> %a512, <32 x i16> %b512, <32 x i16> <i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7>)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret void
 ;
   %I16    = call i16 @llvm.fshr.i16(i16 %a16, i16 %b16, i16 7)
   %V8I16  = call <8 x i16> @llvm.fshr.v8i16(<8 x i16> %a128, <8 x i16> %b128, <8 x i16> <i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7>)
@@ -1102,6 +1188,13 @@ define void @constant_funnel_i8(i8 %a8, <16 x i8> %a128, <32 x i8> %a256, <64 x
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 20 for instruction: %V32I8 = call <32 x i8> @llvm.fshr.v32i8(<32 x i8> %a256, <32 x i8> %b256, <32 x i8> <i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0>)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 40 for instruction: %V64I8 = call <64 x i8> @llvm.fshr.v64i8(<64 x i8> %a512, <64 x i8> %b512, <64 x i8> <i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0>)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret void
+;
+; AVX512GFNI-LABEL: 'constant_funnel_i8'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %I8 = call i8 @llvm.fshr.i8(i8 %a8, i8 %b8, i8 7)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 12 for instruction: %V16I8 = call <16 x i8> @llvm.fshr.v16i8(<16 x i8> %a128, <16 x i8> %b128, <16 x i8> <i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0>)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 12 for instruction: %V32I8 = call <32 x i8> @llvm.fshr.v32i8(<32 x i8> %a256, <32 x i8> %b256, <32 x i8> <i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0>)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 17 for instruction: %V64I8 = call <64 x i8> @llvm.fshr.v64i8(<64 x i8> %a512, <64 x i8> %b512, <64 x i8> <i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0>)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret void
 ;
   %I8    = call i8 @llvm.fshr.i8(i8 %a8, i8 %b8, i8 7)
   %V16I8 = call <16 x i8> @llvm.fshr.v16i8(<16 x i8> %a128, <16 x i8> %b128, <16 x i8> <i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0>)
@@ -1191,6 +1284,13 @@ define void @splatconstant_funnel_i64(i64 %a64, <2 x i64> %a128, <4 x i64> %a256
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 18 for instruction: %V4I64 = call <4 x i64> @llvm.fshr.v4i64(<4 x i64> %a256, <4 x i64> %b256, <4 x i64> <i64 7, i64 7, i64 7, i64 7>)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 36 for instruction: %V8I64 = call <8 x i64> @llvm.fshr.v8i64(<8 x i64> %a512, <8 x i64> %b512, <8 x i64> <i64 7, i64 7, i64 7, i64 7, i64 7, i64 7, i64 7, i64 7>)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret void
+;
+; AVX512GFNI-LABEL: 'splatconstant_funnel_i64'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %I64 = call i64 @llvm.fshr.i64(i64 %a64, i64 %b64, i64 7)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V2I64 = call <2 x i64> @llvm.fshr.v2i64(<2 x i64> %a128, <2 x i64> %b128, <2 x i64> <i64 7, i64 7>)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V4I64 = call <4 x i64> @llvm.fshr.v4i64(<4 x i64> %a256, <4 x i64> %b256, <4 x i64> <i64 7, i64 7, i64 7, i64 7>)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V8I64 = call <8 x i64> @llvm.fshr.v8i64(<8 x i64> %a512, <8 x i64> %b512, <8 x i64> <i64 7, i64 7, i64 7, i64 7, i64 7, i64 7, i64 7, i64 7>)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret void
 ;
   %I64    = call i64 @llvm.fshr.i64(i64 %a64, i64 %b64, i64 7)
   %V2I64  = call <2 x i64> @llvm.fshr.v2i64(<2 x i64> %a128, <2 x i64> %b128, <2 x i64> <i64 7, i64 7>)
@@ -1269,6 +1369,13 @@ define void @splatconstant_funnel_i32(i32 %a32, <4 x i32> %a128, <8 x i32> %a256
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 18 for instruction: %V4I32 = call <8 x i32> @llvm.fshr.v8i32(<8 x i32> %a256, <8 x i32> %b256, <8 x i32> <i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5>)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 36 for instruction: %V8I32 = call <16 x i32> @llvm.fshr.v16i32(<16 x i32> %a512, <16 x i32> %b512, <16 x i32> <i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5>)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret void
+;
+; AVX512GFNI-LABEL: 'splatconstant_funnel_i32'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %I32 = call i32 @llvm.fshr.i32(i32 %a32, i32 %b32, i32 5)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V2I32 = call <4 x i32> @llvm.fshr.v4i32(<4 x i32> %a128, <4 x i32> %b128, <4 x i32> <i32 5, i32 5, i32 5, i32 5>)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V4I32 = call <8 x i32> @llvm.fshr.v8i32(<8 x i32> %a256, <8 x i32> %b256, <8 x i32> <i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5>)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V8I32 = call <16 x i32> @llvm.fshr.v16i32(<16 x i32> %a512, <16 x i32> %b512, <16 x i32> <i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5>)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret void
 ;
   %I32   = call i32 @llvm.fshr.i32(i32 %a32, i32 %b32, i32 5)
   %V2I32 = call <4 x i32> @llvm.fshr.v4i32(<4 x i32> %a128, <4 x i32> %b128, <4 x i32> <i32 5, i32 5, i32 5, i32 5>)
@@ -1347,6 +1454,13 @@ define void @splatconstant_funnel_i16(i16 %a16, <8 x i16> %a128, <16 x i16> %a25
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 18 for instruction: %V16I16 = call <16 x i16> @llvm.fshr.v16i16(<16 x i16> %a256, <16 x i16> %b256, <16 x i16> <i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3>)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 36 for instruction: %V32I16 = call <32 x i16> @llvm.fshr.v32i16(<32 x i16> %a512, <32 x i16> %b512, <32 x i16> <i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3>)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret void
+;
+; AVX512GFNI-LABEL: 'splatconstant_funnel_i16'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %I16 = call i16 @llvm.fshr.i16(i16 %a16, i16 %b16, i16 3)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V8I16 = call <8 x i16> @llvm.fshr.v8i16(<8 x i16> %a128, <8 x i16> %b128, <8 x i16> <i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3>)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V16I16 = call <16 x i16> @llvm.fshr.v16i16(<16 x i16> %a256, <16 x i16> %b256, <16 x i16> <i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3>)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V32I16 = call <32 x i16> @llvm.fshr.v32i16(<32 x i16> %a512, <32 x i16> %b512, <32 x i16> <i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3>)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret void
 ;
   %I16    = call i16 @llvm.fshr.i16(i16 %a16, i16 %b16, i16 3)
   %V8I16  = call <8 x i16> @llvm.fshr.v8i16(<8 x i16> %a128, <8 x i16> %b128, <8 x i16> <i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3>)
@@ -1425,6 +1539,13 @@ define void @splatconstant_funnel_i8(i8 %a8, <16 x i8> %a128, <32 x i8> %a256, <
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 20 for instruction: %V32I8 = call <32 x i8> @llvm.fshr.v32i8(<32 x i8> %a256, <32 x i8> %b256, <32 x i8> <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, 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>)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 40 for instruction: %V64I8 = call <64 x i8> @llvm.fshr.v64i8(<64 x i8> %a512, <64 x i8> %b512, <64 x i8> <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, 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, 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, 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>)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret void
+;
+; AVX512GFNI-LABEL: 'splatconstant_funnel_i8'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %I8 = call i8 @llvm.fshr.i8(i8 %a8, i8 %b8, i8 3)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 6 for instruction: %V16I8 = call <16 x i8> @llvm.fshr.v16i8(<16 x i8> %a128, <16 x i8> %b128, <16 x i8> <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>)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 6 for instruction: %V32I8 = call <32 x i8> @llvm.fshr.v32i8(<32 x i8> %a256, <32 x i8> %b256, <32 x i8> <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, 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>)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 6 for instruction: %V64I8 = call <64 x i8> @llvm.fshr.v64i8(<64 x i8> %a512, <64 x i8> %b512, <64 x i8> <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, 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, 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, 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>)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret void
 ;
   %I8    = call i8 @llvm.fshr.i8(i8 %a8, i8 %b8, i8 3)
   %V16I8 = call <16 x i8> @llvm.fshr.v16i8(<16 x i8> %a128, <16 x i8> %b128, <16 x i8> <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>)
@@ -1635,6 +1756,13 @@ define void @var_rotate_i16(i16 %a16, <8 x i16> %a128, <16 x i16> %a256, <32 x i
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V16I16 = call <16 x i16> @llvm.fshr.v16i16(<16 x i16> %a256, <16 x i16> %a256, <16 x i16> %c256)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %V32I16 = call <32 x i16> @llvm.fshr.v32i16(<32 x i16> %a512, <32 x i16> %a512, <32 x i16> %c512)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret void
+;
+; AVX512GFNI-LABEL: 'var_rotate_i16'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %I16 = call i16 @llvm.fshr.i16(i16 %a16, i16 %a16, i16 %c16)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V8I16 = call <8 x i16> @llvm.fshr.v8i16(<8 x i16> %a128, <8 x i16> %a128, <8 x i16> %c128)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V16I16 = call <16 x i16> @llvm.fshr.v16i16(<16 x i16> %a256, <16 x i16> %a256, <16 x i16> %c256)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V32I16 = call <32 x i16> @llvm.fshr.v32i16(<32 x i16> %a512, <32 x i16> %a512, <32 x i16> %c512)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret void
 ;
   %I16    = call i16 @llvm.fshr.i16(i16 %a16, i16 %a16, i16 %c16)
   %V8I16  = call <8 x i16> @llvm.fshr.v8i16(<8 x i16> %a128, <8 x i16> %a128, <8 x i16> %c128)
@@ -1720,6 +1848,13 @@ define void @var_rotate_i8(i8 %a8, <16 x i8> %a128, <32 x i8> %a256, <64 x i8> %
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V32I8 = call <32 x i8> @llvm.fshr.v32i8(<32 x i8> %a256, <32 x i8> %a256, <32 x i8> %c256)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %V64I8 = call <64 x i8> @llvm.fshr.v64i8(<64 x i8> %a512, <64 x i8> %a512, <64 x i8> %c512)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret void
+;
+; AVX512GFNI-LABEL: 'var_rotate_i8'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %I8 = call i8 @llvm.fshr.i8(i8 %a8, i8 %a8, i8 %c8)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 5 for instruction: %V16I8 = call <16 x i8> @llvm.fshr.v16i8(<16 x i8> %a128, <16 x i8> %a128, <16 x i8> %c128)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 5 for instruction: %V32I8 = call <32 x i8> @llvm.fshr.v32i8(<32 x i8> %a256, <32 x i8> %a256, <32 x i8> %c256)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 5 for instruction: %V64I8 = call <64 x i8> @llvm.fshr.v64i8(<64 x i8> %a512, <64 x i8> %a512, <64 x i8> %c512)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret void
 ;
   %I8    = call i8 @llvm.fshr.i8(i8 %a8, i8 %a8, i8 %c8)
   %V16I8 = call <16 x i8> @llvm.fshr.v16i8(<16 x i8> %a128, <16 x i8> %a128, <16 x i8> %c128)
@@ -1968,6 +2103,15 @@ define void @splatvar_rotate_i16(i16 %a16, <8 x i16> %a128, <16 x i16> %a256, <3
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V16I16 = call <16 x i16> @llvm.fshr.v16i16(<16 x i16> %a256, <16 x i16> %a256, <16 x i16> %u256)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %V32I16 = call <32 x i16> @llvm.fshr.v32i16(<32 x i16> %a512, <32 x i16> %a512, <32 x i16> %u512)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret void
+;
+; AVX512GFNI-LABEL: 'splatvar_rotate_i16'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %u128 = shufflevector <8 x i16> %c128, <8 x i16> undef, <8 x i32> zeroinitializer
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %u256 = shufflevector <16 x i16> %c256, <16 x i16> undef, <16 x i32> zeroinitializer
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %u512 = shufflevector <32 x i16> %c512, <32 x i16> undef, <32 x i32> zeroinitializer
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V8I16 = call <8 x i16> @llvm.fshr.v8i16(<8 x i16> %a128, <8 x i16> %a128, <8 x i16> %u128)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V16I16 = call <16 x i16> @llvm.fshr.v16i16(<16 x i16> %a256, <16 x i16> %a256, <16 x i16> %u256)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V32I16 = call <32 x i16> @llvm.fshr.v32i16(<32 x i16> %a512, <32 x i16> %a512, <32 x i16> %u512)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret void
 ;
   %u128 = shufflevector <8 x i16> %c128, <8 x i16> undef, <8 x i32> zeroinitializer
   %u256 = shufflevector <16 x i16> %c256, <16 x i16> undef, <16 x i32> zeroinitializer
@@ -2068,6 +2212,15 @@ define void @splatvar_rotate_i8(i8 %a8, <16 x i8> %a128, <32 x i8> %a256, <64 x
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V32I8 = call <32 x i8> @llvm.fshr.v32i8(<32 x i8> %a256, <32 x i8> %a256, <32 x i8> %u256)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %V64I8 = call <64 x i8> @llvm.fshr.v64i8(<64 x i8> %a512, <64 x i8> %a512, <64 x i8> %u512)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret void
+;
+; AVX512GFNI-LABEL: 'splatvar_rotate_i8'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %u128 = shufflevector <16 x i8> %c128, <16 x i8> undef, <16 x i32> zeroinitializer
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %u256 = shufflevector <32 x i8> %c256, <32 x i8> undef, <32 x i32> zeroinitializer
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %u512 = shufflevector <64 x i8> %c512, <64 x i8> undef, <64 x i32> zeroinitializer
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 5 for instruction: %V16I8 = call <16 x i8> @llvm.fshr.v16i8(<16 x i8> %a128, <16 x i8> %a128, <16 x i8> %u128)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 5 for instruction: %V32I8 = call <32 x i8> @llvm.fshr.v32i8(<32 x i8> %a256, <32 x i8> %a256, <32 x i8> %u256)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 5 for instruction: %V64I8 = call <64 x i8> @llvm.fshr.v64i8(<64 x i8> %a512, <64 x i8> %a512, <64 x i8> %u512)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret void
 ;
   %u128 = shufflevector <16 x i8> %c128, <16 x i8> undef, <16 x i32> zeroinitializer
   %u256 = shufflevector <32 x i8> %c256, <32 x i8> undef, <32 x i32> zeroinitializer
@@ -2273,6 +2426,13 @@ define void @constant_rotate_i16(i16 %a16, <8 x i16> %a128, <16 x i16> %a256, <3
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V16I16 = call <16 x i16> @llvm.fshr.v16i16(<16 x i16> %a256, <16 x i16> %a256, <16 x i16> <i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7>)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %V32I16 = call <32 x i16> @llvm.fshr.v32i16(<32 x i16> %a512, <32 x i16> %a512, <32 x i16> <i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7>)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret void
+;
+; AVX512GFNI-LABEL: 'constant_rotate_i16'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %I16 = call i16 @llvm.fshr.i16(i16 %a16, i16 %a16, i16 7)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V8I16 = call <8 x i16> @llvm.fshr.v8i16(<8 x i16> %a128, <8 x i16> %a128, <8 x i16> <i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7>)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V16I16 = call <16 x i16> @llvm.fshr.v16i16(<16 x i16> %a256, <16 x i16> %a256, <16 x i16> <i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7>)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V32I16 = call <32 x i16> @llvm.fshr.v32i16(<32 x i16> %a512, <32 x i16> %a512, <32 x i16> <i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7>)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret void
 ;
   %I16    = call i16 @llvm.fshr.i16(i16 %a16, i16 %a16, i16 7)
   %V8I16  = call <8 x i16> @llvm.fshr.v8i16(<8 x i16> %a128, <8 x i16> %a128, <8 x i16> <i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7>)
@@ -2358,6 +2518,13 @@ define void @constant_rotate_i8(i8 %a8, <16 x i8> %a128, <32 x i8> %a256, <64 x
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V32I8 = call <32 x i8> @llvm.fshr.v32i8(<32 x i8> %a256, <32 x i8> %a256, <32 x i8> <i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0>)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %V64I8 = call <64 x i8> @llvm.fshr.v64i8(<64 x i8> %a512, <64 x i8> %a512, <64 x i8> <i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0>)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret void
+;
+; AVX512GFNI-LABEL: 'constant_rotate_i8'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %I8 = call i8 @llvm.fshr.i8(i8 %a8, i8 %a8, i8 7)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 5 for instruction: %V16I8 = call <16 x i8> @llvm.fshr.v16i8(<16 x i8> %a128, <16 x i8> %a128, <16 x i8> <i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0>)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 5 for instruction: %V32I8 = call <32 x i8> @llvm.fshr.v32i8(<32 x i8> %a256, <32 x i8> %a256, <32 x i8> <i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0>)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 5 for instruction: %V64I8 = call <64 x i8> @llvm.fshr.v64i8(<64 x i8> %a512, <64 x i8> %a512, <64 x i8> <i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0>)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret void
 ;
   %I8    = call i8 @llvm.fshr.i8(i8 %a8, i8 %a8, i8 7)
   %V16I8 = call <16 x i8> @llvm.fshr.v16i8(<16 x i8> %a128, <16 x i8> %a128, <16 x i8> <i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0>)
@@ -2554,6 +2721,13 @@ define void @splatconstant_rotate_i16(i16 %a16, <8 x i16> %a128, <16 x i16> %a25
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V16I16 = call <16 x i16> @llvm.fshr.v16i16(<16 x i16> %a256, <16 x i16> %a256, <16 x i16> <i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3>)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %V32I16 = call <32 x i16> @llvm.fshr.v32i16(<32 x i16> %a512, <32 x i16> %a512, <32 x i16> <i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3>)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret void
+;
+; AVX512GFNI-LABEL: 'splatconstant_rotate_i16'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %I16 = call i16 @llvm.fshr.i16(i16 %a16, i16 %a16, i16 3)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V8I16 = call <8 x i16> @llvm.fshr.v8i16(<8 x i16> %a128, <8 x i16> %a128, <8 x i16> <i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3>)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V16I16 = call <16 x i16> @llvm.fshr.v16i16(<16 x i16> %a256, <16 x i16> %a256, <16 x i16> <i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3>)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V32I16 = call <32 x i16> @llvm.fshr.v32i16(<32 x i16> %a512, <32 x i16> %a512, <32 x i16> <i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3>)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret void
 ;
   %I16    = call i16 @llvm.fshr.i16(i16 %a16, i16 %a16, i16 3)
   %V8I16  = call <8 x i16> @llvm.fshr.v8i16(<8 x i16> %a128, <8 x i16> %a128, <8 x i16> <i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3>)
@@ -2632,6 +2806,13 @@ define void @splatconstant_rotate_i8(i8 %a8, <16 x i8> %a128, <32 x i8> %a256, <
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V32I8 = call <32 x i8> @llvm.fshr.v32i8(<32 x i8> %a256, <32 x i8> %a256, <32 x i8> <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, 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>)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %V64I8 = call <64 x i8> @llvm.fshr.v64i8(<64 x i8> %a512, <64 x i8> %a512, <64 x i8> <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, 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, 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, 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>)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret void
+;
+; AVX512GFNI-LABEL: 'splatconstant_rotate_i8'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %I8 = call i8 @llvm.fshr.i8(i8 %a8, i8 %a8, i8 3)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 5 for instruction: %V16I8 = call <16 x i8> @llvm.fshr.v16i8(<16 x i8> %a128, <16 x i8> %a128, <16 x i8> <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>)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 5 for instruction: %V32I8 = call <32 x i8> @llvm.fshr.v32i8(<32 x i8> %a256, <32 x i8> %a256, <32 x i8> <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, 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>)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 5 for instruction: %V64I8 = call <64 x i8> @llvm.fshr.v64i8(<64 x i8> %a512, <64 x i8> %a512, <64 x i8> <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, 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, 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, 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>)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret void
 ;
   %I8    = call i8 @llvm.fshr.i8(i8 %a8, i8 %a8, i8 3)
   %V16I8 = call <16 x i8> @llvm.fshr.v16i8(<16 x i8> %a128, <16 x i8> %a128, <16 x i8> <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>)

diff  --git a/llvm/test/Analysis/CostModel/X86/vshift-ashr-codesize.ll b/llvm/test/Analysis/CostModel/X86/vshift-ashr-codesize.ll
index 8e3f38c796e6e5..a3c24bdd1a8820 100644
--- a/llvm/test/Analysis/CostModel/X86/vshift-ashr-codesize.ll
+++ b/llvm/test/Analysis/CostModel/X86/vshift-ashr-codesize.ll
@@ -15,6 +15,7 @@
 ; RUN: opt < %s -mtriple=x86_64-unknown-linux-gnu -passes="print<cost-model>" 2>&1 -disable-output -cost-kind=code-size -mcpu=slm | FileCheck %s --check-prefixes=CHECK,SSE42
 ; RUN: opt < %s -mtriple=x86_64-unknown-linux-gnu -passes="print<cost-model>" 2>&1 -disable-output -cost-kind=code-size -mcpu=goldmont | FileCheck %s --check-prefixes=CHECK,SSE42
 ; RUN: opt < %s -mtriple=x86_64-unknown-linux-gnu -passes="print<cost-model>" 2>&1 -disable-output -cost-kind=code-size -mcpu=btver2 | FileCheck %s --check-prefixes=CHECK,AVX,AVX1
+; RUN: opt < %s -mtriple=x86_64-unknown-linux-gnu -passes="print<cost-model>" 2>&1 -disable-output -cost-kind=code-size -mcpu=tigerlake | FileCheck %s --check-prefixes=CHECK,AVX512,AVX512GFNI
 
 ; Verify the cost of vector logical shift right instructions.
 
@@ -232,6 +233,10 @@ define <8 x i16> @var_shift_v8i16(<8 x i16> %a, <8 x i16> %b) {
 ; AVX512BW-LABEL: 'var_shift_v8i16'
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shift = ashr <8 x i16> %a, %b
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <8 x i16> %shift
+;
+; AVX512GFNI-LABEL: 'var_shift_v8i16'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shift = ashr <8 x i16> %a, %b
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <8 x i16> %shift
 ;
   %shift = ashr <8 x i16> %a, %b
   ret <8 x i16> %shift
@@ -265,6 +270,10 @@ define <16 x i16> @var_shift_v16i16(<16 x i16> %a, <16 x i16> %b) {
 ; AVX512BW-LABEL: 'var_shift_v16i16'
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shift = ashr <16 x i16> %a, %b
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <16 x i16> %shift
+;
+; AVX512GFNI-LABEL: 'var_shift_v16i16'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shift = ashr <16 x i16> %a, %b
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <16 x i16> %shift
 ;
   %shift = ashr <16 x i16> %a, %b
   ret <16 x i16> %shift
@@ -298,6 +307,10 @@ define <32 x i16> @var_shift_v32i16(<32 x i16> %a, <32 x i16> %b) {
 ; AVX512BW-LABEL: 'var_shift_v32i16'
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shift = ashr <32 x i16> %a, %b
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <32 x i16> %shift
+;
+; AVX512GFNI-LABEL: 'var_shift_v32i16'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shift = ashr <32 x i16> %a, %b
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <32 x i16> %shift
 ;
   %shift = ashr <32 x i16> %a, %b
   ret <32 x i16> %shift
@@ -327,6 +340,10 @@ define <16 x i8> @var_shift_v16i8(<16 x i8> %a, <16 x i8> %b) {
 ; AVX512BW-LABEL: 'var_shift_v16i8'
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %shift = ashr <16 x i8> %a, %b
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <16 x i8> %shift
+;
+; AVX512GFNI-LABEL: 'var_shift_v16i8'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %shift = ashr <16 x i8> %a, %b
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <16 x i8> %shift
 ;
   %shift = ashr <16 x i8> %a, %b
   ret <16 x i8> %shift
@@ -389,6 +406,10 @@ define <64 x i8> @var_shift_v64i8(<64 x i8> %a, <64 x i8> %b) {
 ; AVX512BW-LABEL: 'var_shift_v64i8'
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 30 for instruction: %shift = ashr <64 x i8> %a, %b
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <64 x i8> %shift
+;
+; AVX512GFNI-LABEL: 'var_shift_v64i8'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 30 for instruction: %shift = ashr <64 x i8> %a, %b
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <64 x i8> %shift
 ;
   %shift = ashr <64 x i8> %a, %b
   ret <64 x i8> %shift
@@ -778,6 +799,12 @@ define <32 x i16> @splatvar_shift_v32i16(<32 x i16> %a, i16 %b) {
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %splat = shufflevector <32 x i16> %insert, <32 x i16> undef, <32 x i32> zeroinitializer
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %shift = ashr <32 x i16> %a, %splat
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <32 x i16> %shift
+;
+; AVX512GFNI-LABEL: 'splatvar_shift_v32i16'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %insert = insertelement <32 x i16> undef, i16 %b, i32 0
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %splat = shufflevector <32 x i16> %insert, <32 x i16> undef, <32 x i32> zeroinitializer
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %shift = ashr <32 x i16> %a, %splat
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <32 x i16> %shift
 ;
   %insert = insertelement <32 x i16> undef, i16 %b, i32 0
   %splat = shufflevector <32 x i16> %insert, <32 x i16> undef, <32 x i32> zeroinitializer
@@ -827,6 +854,12 @@ define <16 x i8> @splatvar_shift_v16i8(<16 x i8> %a, i8 %b) {
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %splat = shufflevector <16 x i8> %insert, <16 x i8> undef, <16 x i32> zeroinitializer
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %shift = ashr <16 x i8> %a, %splat
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <16 x i8> %shift
+;
+; AVX512GFNI-LABEL: 'splatvar_shift_v16i8'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %insert = insertelement <16 x i8> undef, i8 %b, i32 0
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %splat = shufflevector <16 x i8> %insert, <16 x i8> undef, <16 x i32> zeroinitializer
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %shift = ashr <16 x i8> %a, %splat
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <16 x i8> %shift
 ;
   %insert = insertelement <16 x i8> undef, i8 %b, i32 0
   %splat = shufflevector <16 x i8> %insert, <16 x i8> undef, <16 x i32> zeroinitializer
@@ -882,6 +915,12 @@ define <32 x i8> @splatvar_shift_v32i8(<32 x i8> %a, i8 %b) {
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %splat = shufflevector <32 x i8> %insert, <32 x i8> undef, <32 x i32> zeroinitializer
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 10 for instruction: %shift = ashr <32 x i8> %a, %splat
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <32 x i8> %shift
+;
+; AVX512GFNI-LABEL: 'splatvar_shift_v32i8'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %insert = insertelement <32 x i8> undef, i8 %b, i32 0
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %splat = shufflevector <32 x i8> %insert, <32 x i8> undef, <32 x i32> zeroinitializer
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 10 for instruction: %shift = ashr <32 x i8> %a, %splat
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <32 x i8> %shift
 ;
   %insert = insertelement <32 x i8> undef, i8 %b, i32 0
   %splat = shufflevector <32 x i8> %insert, <32 x i8> undef, <32 x i32> zeroinitializer
@@ -937,6 +976,12 @@ define <64 x i8> @splatvar_shift_v64i8(<64 x i8> %a, i8 %b) {
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %splat = shufflevector <64 x i8> %insert, <64 x i8> undef, <64 x i32> zeroinitializer
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 10 for instruction: %shift = ashr <64 x i8> %a, %splat
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <64 x i8> %shift
+;
+; AVX512GFNI-LABEL: 'splatvar_shift_v64i8'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %insert = insertelement <64 x i8> undef, i8 %b, i32 0
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %splat = shufflevector <64 x i8> %insert, <64 x i8> undef, <64 x i32> zeroinitializer
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 10 for instruction: %shift = ashr <64 x i8> %a, %splat
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <64 x i8> %shift
 ;
   %insert = insertelement <64 x i8> undef, i8 %b, i32 0
   %splat = shufflevector <64 x i8> %insert, <64 x i8> undef, <64 x i32> zeroinitializer
@@ -1158,6 +1203,10 @@ define <8 x i16> @constant_shift_v8i16(<8 x i16> %a) {
 ; AVX512BW-LABEL: 'constant_shift_v8i16'
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shift = ashr <8 x i16> %a, <i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7>
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <8 x i16> %shift
+;
+; AVX512GFNI-LABEL: 'constant_shift_v8i16'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shift = ashr <8 x i16> %a, <i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7>
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <8 x i16> %shift
 ;
   %shift = ashr <8 x i16> %a, <i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7>
   ret <8 x i16> %shift
@@ -1191,6 +1240,10 @@ define <16 x i16> @constant_shift_v16i16(<16 x i16> %a) {
 ; AVX512BW-LABEL: 'constant_shift_v16i16'
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shift = ashr <16 x i16> %a, <i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7>
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <16 x i16> %shift
+;
+; AVX512GFNI-LABEL: 'constant_shift_v16i16'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shift = ashr <16 x i16> %a, <i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7>
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <16 x i16> %shift
 ;
   %shift = ashr <16 x i16> %a, <i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7>
   ret <16 x i16> %shift
@@ -1224,6 +1277,10 @@ define <32 x i16> @constant_shift_v32i16(<32 x i16> %a) {
 ; AVX512BW-LABEL: 'constant_shift_v32i16'
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shift = ashr <32 x i16> %a, <i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7>
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <32 x i16> %shift
+;
+; AVX512GFNI-LABEL: 'constant_shift_v32i16'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shift = ashr <32 x i16> %a, <i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7>
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <32 x i16> %shift
 ;
   %shift = ashr <32 x i16> %a, <i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7>
   ret <32 x i16> %shift
@@ -1253,6 +1310,10 @@ define <16 x i8> @constant_shift_v16i8(<16 x i8> %a) {
 ; AVX512BW-LABEL: 'constant_shift_v16i8'
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %shift = ashr <16 x i8> %a, <i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0>
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <16 x i8> %shift
+;
+; AVX512GFNI-LABEL: 'constant_shift_v16i8'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %shift = ashr <16 x i8> %a, <i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0>
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <16 x i8> %shift
 ;
   %shift = ashr <16 x i8> %a, <i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0>
   ret <16 x i8> %shift
@@ -1315,6 +1376,10 @@ define <64 x i8> @constant_shift_v64i8(<64 x i8> %a) {
 ; AVX512BW-LABEL: 'constant_shift_v64i8'
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 30 for instruction: %shift = ashr <64 x i8> %a, <i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0>
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <64 x i8> %shift
+;
+; AVX512GFNI-LABEL: 'constant_shift_v64i8'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 30 for instruction: %shift = ashr <64 x i8> %a, <i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0>
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <64 x i8> %shift
 ;
   %shift = ashr <64 x i8> %a, <i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0>
   ret <64 x i8> %shift
@@ -1531,6 +1596,10 @@ define <16 x i16> @splatconstant_shift_v16i16(<16 x i16> %a) {
 ; AVX512BW-LABEL: 'splatconstant_shift_v16i16'
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shift = ashr <16 x i16> %a, <i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3>
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <16 x i16> %shift
+;
+; AVX512GFNI-LABEL: 'splatconstant_shift_v16i16'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shift = ashr <16 x i16> %a, <i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3>
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <16 x i16> %shift
 ;
   %shift = ashr <16 x i16> %a, <i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3>
   ret <16 x i16> %shift
@@ -1568,6 +1637,10 @@ define <32 x i16> @splatconstant_shift_v32i16(<32 x i16> %a) {
 ; AVX512BW-LABEL: 'splatconstant_shift_v32i16'
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shift = ashr <32 x i16> %a, <i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3>
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <32 x i16> %shift
+;
+; AVX512GFNI-LABEL: 'splatconstant_shift_v32i16'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shift = ashr <32 x i16> %a, <i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3>
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <32 x i16> %shift
 ;
   %shift = ashr <32 x i16> %a, <i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3>
   ret <32 x i16> %shift
@@ -1601,6 +1674,10 @@ define <16 x i8> @splatconstant_shift_v16i8(<16 x i8> %a) {
 ; AVX512BW-LABEL: 'splatconstant_shift_v16i8'
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %shift = ashr <16 x i8> %a, <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>
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <16 x i8> %shift
+;
+; AVX512GFNI-LABEL: 'splatconstant_shift_v16i8'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %shift = ashr <16 x i8> %a, <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>
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <16 x i8> %shift
 ;
   %shift = ashr <16 x i8> %a, <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>
   ret <16 x i8> %shift
@@ -1634,6 +1711,10 @@ define <32 x i8> @splatconstant_shift_v32i8(<32 x i8> %a) {
 ; AVX512BW-LABEL: 'splatconstant_shift_v32i8'
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %shift = ashr <32 x i8> %a, <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, 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>
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <32 x i8> %shift
+;
+; AVX512GFNI-LABEL: 'splatconstant_shift_v32i8'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %shift = ashr <32 x i8> %a, <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, 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>
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <32 x i8> %shift
 ;
   %shift = ashr <32 x i8> %a, <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, 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>
   ret <32 x i8> %shift
@@ -1667,6 +1748,10 @@ define <64 x i8> @splatconstant_shift_v64i8(<64 x i8> %a) {
 ; AVX512BW-LABEL: 'splatconstant_shift_v64i8'
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %shift = ashr <64 x i8> %a, <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, 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, 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, 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>
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <64 x i8> %shift
+;
+; AVX512GFNI-LABEL: 'splatconstant_shift_v64i8'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %shift = ashr <64 x i8> %a, <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, 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, 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, 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>
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <64 x i8> %shift
 ;
   %shift = ashr <64 x i8> %a, <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, 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, 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, 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>
   ret <64 x i8> %shift

diff  --git a/llvm/test/Analysis/CostModel/X86/vshift-ashr-cost-inseltpoison.ll b/llvm/test/Analysis/CostModel/X86/vshift-ashr-cost-inseltpoison.ll
index d4ece3b2a1134e..67679f2cb85666 100644
--- a/llvm/test/Analysis/CostModel/X86/vshift-ashr-cost-inseltpoison.ll
+++ b/llvm/test/Analysis/CostModel/X86/vshift-ashr-cost-inseltpoison.ll
@@ -15,6 +15,7 @@
 ; RUN: opt < %s -mtriple=x86_64-unknown-linux-gnu -passes="print<cost-model>" 2>&1 -disable-output -mcpu=slm | FileCheck %s --check-prefixes=CHECK,SSE,SSE42
 ; RUN: opt < %s -mtriple=x86_64-unknown-linux-gnu -passes="print<cost-model>" 2>&1 -disable-output -mcpu=goldmont | FileCheck %s --check-prefixes=CHECK,SSE,SSE42
 ; RUN: opt < %s -mtriple=x86_64-unknown-linux-gnu -passes="print<cost-model>" 2>&1 -disable-output -mcpu=btver2 | FileCheck %s --check-prefixes=CHECK,AVX,AVX1
+; RUN: opt < %s -mtriple=x86_64-unknown-linux-gnu -passes="print<cost-model>" 2>&1 -disable-output -mcpu=tigerlake | FileCheck %s --check-prefixes=CHECK,AVX512,AVX512GFNI
 
 ; Verify the cost of vector arithmetic shift right instructions.
 
@@ -228,6 +229,10 @@ define <8 x i16> @var_shift_v8i16(<8 x i16> %a, <8 x i16> %b) {
 ; AVX512BWVL-LABEL: 'var_shift_v8i16'
 ; AVX512BWVL-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shift = ashr <8 x i16> %a, %b
 ; AVX512BWVL-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <8 x i16> %shift
+;
+; AVX512GFNI-LABEL: 'var_shift_v8i16'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shift = ashr <8 x i16> %a, %b
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <8 x i16> %shift
 ;
   %shift = ashr <8 x i16> %a, %b
   ret <8 x i16> %shift
@@ -269,6 +274,10 @@ define <16 x i16> @var_shift_v16i16(<16 x i16> %a, <16 x i16> %b) {
 ; AVX512BWVL-LABEL: 'var_shift_v16i16'
 ; AVX512BWVL-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shift = ashr <16 x i16> %a, %b
 ; AVX512BWVL-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <16 x i16> %shift
+;
+; AVX512GFNI-LABEL: 'var_shift_v16i16'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shift = ashr <16 x i16> %a, %b
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <16 x i16> %shift
 ;
   %shift = ashr <16 x i16> %a, %b
   ret <16 x i16> %shift
@@ -310,6 +319,10 @@ define <32 x i16> @var_shift_v32i16(<32 x i16> %a, <32 x i16> %b) {
 ; AVX512BWVL-LABEL: 'var_shift_v32i16'
 ; AVX512BWVL-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shift = ashr <32 x i16> %a, %b
 ; AVX512BWVL-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <32 x i16> %shift
+;
+; AVX512GFNI-LABEL: 'var_shift_v32i16'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shift = ashr <32 x i16> %a, %b
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <32 x i16> %shift
 ;
   %shift = ashr <32 x i16> %a, %b
   ret <32 x i16> %shift
@@ -351,6 +364,10 @@ define <16 x i8> @var_shift_v16i8(<16 x i8> %a, <16 x i8> %b) {
 ; AVX512BWVL-LABEL: 'var_shift_v16i8'
 ; AVX512BWVL-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %shift = ashr <16 x i8> %a, %b
 ; AVX512BWVL-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <16 x i8> %shift
+;
+; AVX512GFNI-LABEL: 'var_shift_v16i8'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %shift = ashr <16 x i8> %a, %b
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <16 x i8> %shift
 ;
   %shift = ashr <16 x i8> %a, %b
   ret <16 x i8> %shift
@@ -392,6 +409,10 @@ define <32 x i8> @var_shift_v32i8(<32 x i8> %a, <32 x i8> %b) {
 ; AVX512BWVL-LABEL: 'var_shift_v32i8'
 ; AVX512BWVL-NEXT:  Cost Model: Found an estimated cost of 6 for instruction: %shift = ashr <32 x i8> %a, %b
 ; AVX512BWVL-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <32 x i8> %shift
+;
+; AVX512GFNI-LABEL: 'var_shift_v32i8'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 6 for instruction: %shift = ashr <32 x i8> %a, %b
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <32 x i8> %shift
 ;
   %shift = ashr <32 x i8> %a, %b
   ret <32 x i8> %shift
@@ -433,6 +454,10 @@ define <64 x i8> @var_shift_v64i8(<64 x i8> %a, <64 x i8> %b) {
 ; AVX512BWVL-LABEL: 'var_shift_v64i8'
 ; AVX512BWVL-NEXT:  Cost Model: Found an estimated cost of 15 for instruction: %shift = ashr <64 x i8> %a, %b
 ; AVX512BWVL-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <64 x i8> %shift
+;
+; AVX512GFNI-LABEL: 'var_shift_v64i8'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 15 for instruction: %shift = ashr <64 x i8> %a, %b
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <64 x i8> %shift
 ;
   %shift = ashr <64 x i8> %a, %b
   ret <64 x i8> %shift
@@ -834,6 +859,12 @@ define <32 x i16> @splatvar_shift_v32i16(<32 x i16> %a, i16 %b) {
 ; AVX512BWVL-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %splat = shufflevector <32 x i16> %insert, <32 x i16> poison, <32 x i32> zeroinitializer
 ; AVX512BWVL-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %shift = ashr <32 x i16> %a, %splat
 ; AVX512BWVL-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <32 x i16> %shift
+;
+; AVX512GFNI-LABEL: 'splatvar_shift_v32i16'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %insert = insertelement <32 x i16> poison, i16 %b, i32 0
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %splat = shufflevector <32 x i16> %insert, <32 x i16> poison, <32 x i32> zeroinitializer
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %shift = ashr <32 x i16> %a, %splat
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <32 x i16> %shift
 ;
   %insert = insertelement <32 x i16> poison, i16 %b, i32 0
   %splat = shufflevector <32 x i16> %insert, <32 x i16> poison, <32 x i32> zeroinitializer
@@ -950,6 +981,12 @@ define <32 x i8> @splatvar_shift_v32i8(<32 x i8> %a, i8 %b) {
 ; AVX512BWVL-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %splat = shufflevector <32 x i8> %insert, <32 x i8> poison, <32 x i32> zeroinitializer
 ; AVX512BWVL-NEXT:  Cost Model: Found an estimated cost of 5 for instruction: %shift = ashr <32 x i8> %a, %splat
 ; AVX512BWVL-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <32 x i8> %shift
+;
+; AVX512GFNI-LABEL: 'splatvar_shift_v32i8'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %insert = insertelement <32 x i8> poison, i8 %b, i32 0
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %splat = shufflevector <32 x i8> %insert, <32 x i8> poison, <32 x i32> zeroinitializer
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 5 for instruction: %shift = ashr <32 x i8> %a, %splat
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <32 x i8> %shift
 ;
   %insert = insertelement <32 x i8> poison, i8 %b, i32 0
   %splat = shufflevector <32 x i8> %insert, <32 x i8> poison, <32 x i32> zeroinitializer
@@ -1017,6 +1054,12 @@ define <64 x i8> @splatvar_shift_v64i8(<64 x i8> %a, i8 %b) {
 ; AVX512BWVL-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %splat = shufflevector <64 x i8> %insert, <64 x i8> poison, <64 x i32> zeroinitializer
 ; AVX512BWVL-NEXT:  Cost Model: Found an estimated cost of 5 for instruction: %shift = ashr <64 x i8> %a, %splat
 ; AVX512BWVL-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <64 x i8> %shift
+;
+; AVX512GFNI-LABEL: 'splatvar_shift_v64i8'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %insert = insertelement <64 x i8> poison, i8 %b, i32 0
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %splat = shufflevector <64 x i8> %insert, <64 x i8> poison, <64 x i32> zeroinitializer
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 5 for instruction: %shift = ashr <64 x i8> %a, %splat
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <64 x i8> %shift
 ;
   %insert = insertelement <64 x i8> poison, i8 %b, i32 0
   %splat = shufflevector <64 x i8> %insert, <64 x i8> poison, <64 x i32> zeroinitializer
@@ -1226,6 +1269,10 @@ define <8 x i16> @constant_shift_v8i16(<8 x i16> %a) {
 ; AVX512BWVL-LABEL: 'constant_shift_v8i16'
 ; AVX512BWVL-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shift = ashr <8 x i16> %a, <i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7>
 ; AVX512BWVL-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <8 x i16> %shift
+;
+; AVX512GFNI-LABEL: 'constant_shift_v8i16'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shift = ashr <8 x i16> %a, <i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7>
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <8 x i16> %shift
 ;
   %shift = ashr <8 x i16> %a, <i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7>
   ret <8 x i16> %shift
@@ -1267,6 +1314,10 @@ define <16 x i16> @constant_shift_v16i16(<16 x i16> %a) {
 ; AVX512BWVL-LABEL: 'constant_shift_v16i16'
 ; AVX512BWVL-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shift = ashr <16 x i16> %a, <i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7>
 ; AVX512BWVL-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <16 x i16> %shift
+;
+; AVX512GFNI-LABEL: 'constant_shift_v16i16'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shift = ashr <16 x i16> %a, <i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7>
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <16 x i16> %shift
 ;
   %shift = ashr <16 x i16> %a, <i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7>
   ret <16 x i16> %shift
@@ -1308,6 +1359,10 @@ define <32 x i16> @constant_shift_v32i16(<32 x i16> %a) {
 ; AVX512BWVL-LABEL: 'constant_shift_v32i16'
 ; AVX512BWVL-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shift = ashr <32 x i16> %a, <i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7>
 ; AVX512BWVL-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <32 x i16> %shift
+;
+; AVX512GFNI-LABEL: 'constant_shift_v32i16'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shift = ashr <32 x i16> %a, <i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7>
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <32 x i16> %shift
 ;
   %shift = ashr <32 x i16> %a, <i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7>
   ret <32 x i16> %shift
@@ -1349,6 +1404,10 @@ define <16 x i8> @constant_shift_v16i8(<16 x i8> %a) {
 ; AVX512BWVL-LABEL: 'constant_shift_v16i8'
 ; AVX512BWVL-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %shift = ashr <16 x i8> %a, <i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0>
 ; AVX512BWVL-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <16 x i8> %shift
+;
+; AVX512GFNI-LABEL: 'constant_shift_v16i8'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %shift = ashr <16 x i8> %a, <i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0>
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <16 x i8> %shift
 ;
   %shift = ashr <16 x i8> %a, <i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0>
   ret <16 x i8> %shift
@@ -1390,6 +1449,10 @@ define <32 x i8> @constant_shift_v32i8(<32 x i8> %a) {
 ; AVX512BWVL-LABEL: 'constant_shift_v32i8'
 ; AVX512BWVL-NEXT:  Cost Model: Found an estimated cost of 6 for instruction: %shift = ashr <32 x i8> %a, <i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0>
 ; AVX512BWVL-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <32 x i8> %shift
+;
+; AVX512GFNI-LABEL: 'constant_shift_v32i8'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 6 for instruction: %shift = ashr <32 x i8> %a, <i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0>
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <32 x i8> %shift
 ;
   %shift = ashr <32 x i8> %a, <i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0>
   ret <32 x i8> %shift
@@ -1431,6 +1494,10 @@ define <64 x i8> @constant_shift_v64i8(<64 x i8> %a) {
 ; AVX512BWVL-LABEL: 'constant_shift_v64i8'
 ; AVX512BWVL-NEXT:  Cost Model: Found an estimated cost of 15 for instruction: %shift = ashr <64 x i8> %a, <i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0>
 ; AVX512BWVL-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <64 x i8> %shift
+;
+; AVX512GFNI-LABEL: 'constant_shift_v64i8'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 15 for instruction: %shift = ashr <64 x i8> %a, <i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0>
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <64 x i8> %shift
 ;
   %shift = ashr <64 x i8> %a, <i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0>
   ret <64 x i8> %shift
@@ -1631,6 +1698,10 @@ define <16 x i16> @splatconstant_shift_v16i16(<16 x i16> %a) {
 ; AVX512BWVL-LABEL: 'splatconstant_shift_v16i16'
 ; AVX512BWVL-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shift = ashr <16 x i16> %a, <i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3>
 ; AVX512BWVL-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <16 x i16> %shift
+;
+; AVX512GFNI-LABEL: 'splatconstant_shift_v16i16'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shift = ashr <16 x i16> %a, <i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3>
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <16 x i16> %shift
 ;
   %shift = ashr <16 x i16> %a, <i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3>
   ret <16 x i16> %shift
@@ -1672,6 +1743,10 @@ define <32 x i16> @splatconstant_shift_v32i16(<32 x i16> %a) {
 ; AVX512BWVL-LABEL: 'splatconstant_shift_v32i16'
 ; AVX512BWVL-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shift = ashr <32 x i16> %a, <i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3>
 ; AVX512BWVL-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <32 x i16> %shift
+;
+; AVX512GFNI-LABEL: 'splatconstant_shift_v32i16'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shift = ashr <32 x i16> %a, <i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3>
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <32 x i16> %shift
 ;
   %shift = ashr <32 x i16> %a, <i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3>
   ret <32 x i16> %shift
@@ -1713,6 +1788,10 @@ define <16 x i8> @splatconstant_shift_v16i8(<16 x i8> %a) {
 ; AVX512BWVL-LABEL: 'splatconstant_shift_v16i8'
 ; AVX512BWVL-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shift = ashr <16 x i8> %a, <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>
 ; AVX512BWVL-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <16 x i8> %shift
+;
+; AVX512GFNI-LABEL: 'splatconstant_shift_v16i8'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shift = ashr <16 x i8> %a, <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>
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <16 x i8> %shift
 ;
   %shift = ashr <16 x i8> %a, <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>
   ret <16 x i8> %shift
@@ -1754,6 +1833,10 @@ define <32 x i8> @splatconstant_shift_v32i8(<32 x i8> %a) {
 ; AVX512BWVL-LABEL: 'splatconstant_shift_v32i8'
 ; AVX512BWVL-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shift = ashr <32 x i8> %a, <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, 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>
 ; AVX512BWVL-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <32 x i8> %shift
+;
+; AVX512GFNI-LABEL: 'splatconstant_shift_v32i8'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shift = ashr <32 x i8> %a, <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, 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>
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <32 x i8> %shift
 ;
   %shift = ashr <32 x i8> %a, <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, 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>
   ret <32 x i8> %shift
@@ -1795,6 +1878,10 @@ define <64 x i8> @splatconstant_shift_v64i8(<64 x i8> %a) {
 ; AVX512BWVL-LABEL: 'splatconstant_shift_v64i8'
 ; AVX512BWVL-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shift = ashr <64 x i8> %a, <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, 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, 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, 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>
 ; AVX512BWVL-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <64 x i8> %shift
+;
+; AVX512GFNI-LABEL: 'splatconstant_shift_v64i8'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shift = ashr <64 x i8> %a, <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, 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, 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, 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>
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <64 x i8> %shift
 ;
   %shift = ashr <64 x i8> %a, <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, 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, 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, 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>
   ret <64 x i8> %shift

diff  --git a/llvm/test/Analysis/CostModel/X86/vshift-ashr-cost.ll b/llvm/test/Analysis/CostModel/X86/vshift-ashr-cost.ll
index f8fb2d76a778fc..efd378ee0b5a61 100644
--- a/llvm/test/Analysis/CostModel/X86/vshift-ashr-cost.ll
+++ b/llvm/test/Analysis/CostModel/X86/vshift-ashr-cost.ll
@@ -15,6 +15,7 @@
 ; RUN: opt < %s -mtriple=x86_64-unknown-linux-gnu -passes="print<cost-model>" 2>&1 -disable-output -mcpu=slm | FileCheck %s --check-prefixes=CHECK,SSE,SSE42
 ; RUN: opt < %s -mtriple=x86_64-unknown-linux-gnu -passes="print<cost-model>" 2>&1 -disable-output -mcpu=goldmont | FileCheck %s --check-prefixes=CHECK,SSE,SSE42
 ; RUN: opt < %s -mtriple=x86_64-unknown-linux-gnu -passes="print<cost-model>" 2>&1 -disable-output -mcpu=btver2 | FileCheck %s --check-prefixes=CHECK,AVX,AVX1
+; RUN: opt < %s -mtriple=x86_64-unknown-linux-gnu -passes="print<cost-model>" 2>&1 -disable-output -mcpu=tigerlake | FileCheck %s --check-prefixes=CHECK,AVX512,AVX512GFNI
 
 ; Verify the cost of vector arithmetic shift right instructions.
 
@@ -228,6 +229,10 @@ define <8 x i16> @var_shift_v8i16(<8 x i16> %a, <8 x i16> %b) {
 ; AVX512BWVL-LABEL: 'var_shift_v8i16'
 ; AVX512BWVL-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shift = ashr <8 x i16> %a, %b
 ; AVX512BWVL-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <8 x i16> %shift
+;
+; AVX512GFNI-LABEL: 'var_shift_v8i16'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shift = ashr <8 x i16> %a, %b
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <8 x i16> %shift
 ;
   %shift = ashr <8 x i16> %a, %b
   ret <8 x i16> %shift
@@ -269,6 +274,10 @@ define <16 x i16> @var_shift_v16i16(<16 x i16> %a, <16 x i16> %b) {
 ; AVX512BWVL-LABEL: 'var_shift_v16i16'
 ; AVX512BWVL-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shift = ashr <16 x i16> %a, %b
 ; AVX512BWVL-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <16 x i16> %shift
+;
+; AVX512GFNI-LABEL: 'var_shift_v16i16'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shift = ashr <16 x i16> %a, %b
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <16 x i16> %shift
 ;
   %shift = ashr <16 x i16> %a, %b
   ret <16 x i16> %shift
@@ -310,6 +319,10 @@ define <32 x i16> @var_shift_v32i16(<32 x i16> %a, <32 x i16> %b) {
 ; AVX512BWVL-LABEL: 'var_shift_v32i16'
 ; AVX512BWVL-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shift = ashr <32 x i16> %a, %b
 ; AVX512BWVL-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <32 x i16> %shift
+;
+; AVX512GFNI-LABEL: 'var_shift_v32i16'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shift = ashr <32 x i16> %a, %b
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <32 x i16> %shift
 ;
   %shift = ashr <32 x i16> %a, %b
   ret <32 x i16> %shift
@@ -351,6 +364,10 @@ define <16 x i8> @var_shift_v16i8(<16 x i8> %a, <16 x i8> %b) {
 ; AVX512BWVL-LABEL: 'var_shift_v16i8'
 ; AVX512BWVL-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %shift = ashr <16 x i8> %a, %b
 ; AVX512BWVL-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <16 x i8> %shift
+;
+; AVX512GFNI-LABEL: 'var_shift_v16i8'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %shift = ashr <16 x i8> %a, %b
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <16 x i8> %shift
 ;
   %shift = ashr <16 x i8> %a, %b
   ret <16 x i8> %shift
@@ -392,6 +409,10 @@ define <32 x i8> @var_shift_v32i8(<32 x i8> %a, <32 x i8> %b) {
 ; AVX512BWVL-LABEL: 'var_shift_v32i8'
 ; AVX512BWVL-NEXT:  Cost Model: Found an estimated cost of 6 for instruction: %shift = ashr <32 x i8> %a, %b
 ; AVX512BWVL-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <32 x i8> %shift
+;
+; AVX512GFNI-LABEL: 'var_shift_v32i8'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 6 for instruction: %shift = ashr <32 x i8> %a, %b
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <32 x i8> %shift
 ;
   %shift = ashr <32 x i8> %a, %b
   ret <32 x i8> %shift
@@ -433,6 +454,10 @@ define <64 x i8> @var_shift_v64i8(<64 x i8> %a, <64 x i8> %b) {
 ; AVX512BWVL-LABEL: 'var_shift_v64i8'
 ; AVX512BWVL-NEXT:  Cost Model: Found an estimated cost of 15 for instruction: %shift = ashr <64 x i8> %a, %b
 ; AVX512BWVL-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <64 x i8> %shift
+;
+; AVX512GFNI-LABEL: 'var_shift_v64i8'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 15 for instruction: %shift = ashr <64 x i8> %a, %b
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <64 x i8> %shift
 ;
   %shift = ashr <64 x i8> %a, %b
   ret <64 x i8> %shift
@@ -834,6 +859,12 @@ define <32 x i16> @splatvar_shift_v32i16(<32 x i16> %a, i16 %b) {
 ; AVX512BWVL-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %splat = shufflevector <32 x i16> %insert, <32 x i16> undef, <32 x i32> zeroinitializer
 ; AVX512BWVL-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %shift = ashr <32 x i16> %a, %splat
 ; AVX512BWVL-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <32 x i16> %shift
+;
+; AVX512GFNI-LABEL: 'splatvar_shift_v32i16'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %insert = insertelement <32 x i16> undef, i16 %b, i32 0
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %splat = shufflevector <32 x i16> %insert, <32 x i16> undef, <32 x i32> zeroinitializer
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %shift = ashr <32 x i16> %a, %splat
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <32 x i16> %shift
 ;
   %insert = insertelement <32 x i16> undef, i16 %b, i32 0
   %splat = shufflevector <32 x i16> %insert, <32 x i16> undef, <32 x i32> zeroinitializer
@@ -950,6 +981,12 @@ define <32 x i8> @splatvar_shift_v32i8(<32 x i8> %a, i8 %b) {
 ; AVX512BWVL-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %splat = shufflevector <32 x i8> %insert, <32 x i8> undef, <32 x i32> zeroinitializer
 ; AVX512BWVL-NEXT:  Cost Model: Found an estimated cost of 5 for instruction: %shift = ashr <32 x i8> %a, %splat
 ; AVX512BWVL-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <32 x i8> %shift
+;
+; AVX512GFNI-LABEL: 'splatvar_shift_v32i8'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %insert = insertelement <32 x i8> undef, i8 %b, i32 0
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %splat = shufflevector <32 x i8> %insert, <32 x i8> undef, <32 x i32> zeroinitializer
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 5 for instruction: %shift = ashr <32 x i8> %a, %splat
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <32 x i8> %shift
 ;
   %insert = insertelement <32 x i8> undef, i8 %b, i32 0
   %splat = shufflevector <32 x i8> %insert, <32 x i8> undef, <32 x i32> zeroinitializer
@@ -1017,6 +1054,12 @@ define <64 x i8> @splatvar_shift_v64i8(<64 x i8> %a, i8 %b) {
 ; AVX512BWVL-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %splat = shufflevector <64 x i8> %insert, <64 x i8> undef, <64 x i32> zeroinitializer
 ; AVX512BWVL-NEXT:  Cost Model: Found an estimated cost of 5 for instruction: %shift = ashr <64 x i8> %a, %splat
 ; AVX512BWVL-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <64 x i8> %shift
+;
+; AVX512GFNI-LABEL: 'splatvar_shift_v64i8'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %insert = insertelement <64 x i8> undef, i8 %b, i32 0
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %splat = shufflevector <64 x i8> %insert, <64 x i8> undef, <64 x i32> zeroinitializer
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 5 for instruction: %shift = ashr <64 x i8> %a, %splat
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <64 x i8> %shift
 ;
   %insert = insertelement <64 x i8> undef, i8 %b, i32 0
   %splat = shufflevector <64 x i8> %insert, <64 x i8> undef, <64 x i32> zeroinitializer
@@ -1226,6 +1269,10 @@ define <8 x i16> @constant_shift_v8i16(<8 x i16> %a) {
 ; AVX512BWVL-LABEL: 'constant_shift_v8i16'
 ; AVX512BWVL-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shift = ashr <8 x i16> %a, <i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7>
 ; AVX512BWVL-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <8 x i16> %shift
+;
+; AVX512GFNI-LABEL: 'constant_shift_v8i16'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shift = ashr <8 x i16> %a, <i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7>
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <8 x i16> %shift
 ;
   %shift = ashr <8 x i16> %a, <i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7>
   ret <8 x i16> %shift
@@ -1267,6 +1314,10 @@ define <16 x i16> @constant_shift_v16i16(<16 x i16> %a) {
 ; AVX512BWVL-LABEL: 'constant_shift_v16i16'
 ; AVX512BWVL-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shift = ashr <16 x i16> %a, <i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7>
 ; AVX512BWVL-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <16 x i16> %shift
+;
+; AVX512GFNI-LABEL: 'constant_shift_v16i16'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shift = ashr <16 x i16> %a, <i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7>
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <16 x i16> %shift
 ;
   %shift = ashr <16 x i16> %a, <i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7>
   ret <16 x i16> %shift
@@ -1308,6 +1359,10 @@ define <32 x i16> @constant_shift_v32i16(<32 x i16> %a) {
 ; AVX512BWVL-LABEL: 'constant_shift_v32i16'
 ; AVX512BWVL-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shift = ashr <32 x i16> %a, <i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7>
 ; AVX512BWVL-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <32 x i16> %shift
+;
+; AVX512GFNI-LABEL: 'constant_shift_v32i16'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shift = ashr <32 x i16> %a, <i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7>
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <32 x i16> %shift
 ;
   %shift = ashr <32 x i16> %a, <i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7>
   ret <32 x i16> %shift
@@ -1349,6 +1404,10 @@ define <16 x i8> @constant_shift_v16i8(<16 x i8> %a) {
 ; AVX512BWVL-LABEL: 'constant_shift_v16i8'
 ; AVX512BWVL-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %shift = ashr <16 x i8> %a, <i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0>
 ; AVX512BWVL-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <16 x i8> %shift
+;
+; AVX512GFNI-LABEL: 'constant_shift_v16i8'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %shift = ashr <16 x i8> %a, <i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0>
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <16 x i8> %shift
 ;
   %shift = ashr <16 x i8> %a, <i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0>
   ret <16 x i8> %shift
@@ -1390,6 +1449,10 @@ define <32 x i8> @constant_shift_v32i8(<32 x i8> %a) {
 ; AVX512BWVL-LABEL: 'constant_shift_v32i8'
 ; AVX512BWVL-NEXT:  Cost Model: Found an estimated cost of 6 for instruction: %shift = ashr <32 x i8> %a, <i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0>
 ; AVX512BWVL-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <32 x i8> %shift
+;
+; AVX512GFNI-LABEL: 'constant_shift_v32i8'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 6 for instruction: %shift = ashr <32 x i8> %a, <i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0>
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <32 x i8> %shift
 ;
   %shift = ashr <32 x i8> %a, <i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0>
   ret <32 x i8> %shift
@@ -1431,6 +1494,10 @@ define <64 x i8> @constant_shift_v64i8(<64 x i8> %a) {
 ; AVX512BWVL-LABEL: 'constant_shift_v64i8'
 ; AVX512BWVL-NEXT:  Cost Model: Found an estimated cost of 15 for instruction: %shift = ashr <64 x i8> %a, <i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0>
 ; AVX512BWVL-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <64 x i8> %shift
+;
+; AVX512GFNI-LABEL: 'constant_shift_v64i8'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 15 for instruction: %shift = ashr <64 x i8> %a, <i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0>
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <64 x i8> %shift
 ;
   %shift = ashr <64 x i8> %a, <i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0>
   ret <64 x i8> %shift
@@ -1631,6 +1698,10 @@ define <16 x i16> @splatconstant_shift_v16i16(<16 x i16> %a) {
 ; AVX512BWVL-LABEL: 'splatconstant_shift_v16i16'
 ; AVX512BWVL-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shift = ashr <16 x i16> %a, <i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3>
 ; AVX512BWVL-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <16 x i16> %shift
+;
+; AVX512GFNI-LABEL: 'splatconstant_shift_v16i16'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shift = ashr <16 x i16> %a, <i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3>
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <16 x i16> %shift
 ;
   %shift = ashr <16 x i16> %a, <i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3>
   ret <16 x i16> %shift
@@ -1672,6 +1743,10 @@ define <32 x i16> @splatconstant_shift_v32i16(<32 x i16> %a) {
 ; AVX512BWVL-LABEL: 'splatconstant_shift_v32i16'
 ; AVX512BWVL-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shift = ashr <32 x i16> %a, <i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3>
 ; AVX512BWVL-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <32 x i16> %shift
+;
+; AVX512GFNI-LABEL: 'splatconstant_shift_v32i16'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shift = ashr <32 x i16> %a, <i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3>
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <32 x i16> %shift
 ;
   %shift = ashr <32 x i16> %a, <i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3>
   ret <32 x i16> %shift
@@ -1713,6 +1788,10 @@ define <16 x i8> @splatconstant_shift_v16i8(<16 x i8> %a) {
 ; AVX512BWVL-LABEL: 'splatconstant_shift_v16i8'
 ; AVX512BWVL-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shift = ashr <16 x i8> %a, <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>
 ; AVX512BWVL-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <16 x i8> %shift
+;
+; AVX512GFNI-LABEL: 'splatconstant_shift_v16i8'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shift = ashr <16 x i8> %a, <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>
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <16 x i8> %shift
 ;
   %shift = ashr <16 x i8> %a, <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>
   ret <16 x i8> %shift
@@ -1754,6 +1833,10 @@ define <32 x i8> @splatconstant_shift_v32i8(<32 x i8> %a) {
 ; AVX512BWVL-LABEL: 'splatconstant_shift_v32i8'
 ; AVX512BWVL-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shift = ashr <32 x i8> %a, <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, 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>
 ; AVX512BWVL-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <32 x i8> %shift
+;
+; AVX512GFNI-LABEL: 'splatconstant_shift_v32i8'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shift = ashr <32 x i8> %a, <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, 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>
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <32 x i8> %shift
 ;
   %shift = ashr <32 x i8> %a, <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, 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>
   ret <32 x i8> %shift
@@ -1795,6 +1878,10 @@ define <64 x i8> @splatconstant_shift_v64i8(<64 x i8> %a) {
 ; AVX512BWVL-LABEL: 'splatconstant_shift_v64i8'
 ; AVX512BWVL-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shift = ashr <64 x i8> %a, <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, 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, 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, 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>
 ; AVX512BWVL-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <64 x i8> %shift
+;
+; AVX512GFNI-LABEL: 'splatconstant_shift_v64i8'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shift = ashr <64 x i8> %a, <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, 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, 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, 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>
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <64 x i8> %shift
 ;
   %shift = ashr <64 x i8> %a, <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, 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, 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, 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>
   ret <64 x i8> %shift

diff  --git a/llvm/test/Analysis/CostModel/X86/vshift-ashr-latency.ll b/llvm/test/Analysis/CostModel/X86/vshift-ashr-latency.ll
index af5278e355cfb8..cd4189d4a7f84d 100644
--- a/llvm/test/Analysis/CostModel/X86/vshift-ashr-latency.ll
+++ b/llvm/test/Analysis/CostModel/X86/vshift-ashr-latency.ll
@@ -15,6 +15,7 @@
 ; RUN: opt < %s -mtriple=x86_64-unknown-linux-gnu -passes="print<cost-model>" 2>&1 -disable-output -cost-kind=latency -mcpu=slm | FileCheck %s --check-prefixes=SSE42
 ; RUN: opt < %s -mtriple=x86_64-unknown-linux-gnu -passes="print<cost-model>" 2>&1 -disable-output -cost-kind=latency -mcpu=goldmont | FileCheck %s --check-prefixes=SSE42
 ; RUN: opt < %s -mtriple=x86_64-unknown-linux-gnu -passes="print<cost-model>" 2>&1 -disable-output -cost-kind=latency -mcpu=btver2 | FileCheck %s --check-prefixes=AVX,AVX1
+; RUN: opt < %s -mtriple=x86_64-unknown-linux-gnu -passes="print<cost-model>" 2>&1 -disable-output -cost-kind=latency -mcpu=tigerlake | FileCheck %s --check-prefixes=AVX512,AVX512GFNI
 
 ; Verify the cost of vector logical shift right instructions.
 
@@ -232,6 +233,10 @@ define <8 x i16> @var_shift_v8i16(<8 x i16> %a, <8 x i16> %b) {
 ; AVX512BW-LABEL: 'var_shift_v8i16'
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shift = ashr <8 x i16> %a, %b
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <8 x i16> %shift
+;
+; AVX512GFNI-LABEL: 'var_shift_v8i16'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shift = ashr <8 x i16> %a, %b
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <8 x i16> %shift
 ;
   %shift = ashr <8 x i16> %a, %b
   ret <8 x i16> %shift
@@ -265,6 +270,10 @@ define <16 x i16> @var_shift_v16i16(<16 x i16> %a, <16 x i16> %b) {
 ; AVX512BW-LABEL: 'var_shift_v16i16'
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shift = ashr <16 x i16> %a, %b
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <16 x i16> %shift
+;
+; AVX512GFNI-LABEL: 'var_shift_v16i16'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shift = ashr <16 x i16> %a, %b
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <16 x i16> %shift
 ;
   %shift = ashr <16 x i16> %a, %b
   ret <16 x i16> %shift
@@ -298,6 +307,10 @@ define <32 x i16> @var_shift_v32i16(<32 x i16> %a, <32 x i16> %b) {
 ; AVX512BW-LABEL: 'var_shift_v32i16'
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shift = ashr <32 x i16> %a, %b
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <32 x i16> %shift
+;
+; AVX512GFNI-LABEL: 'var_shift_v32i16'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shift = ashr <32 x i16> %a, %b
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <32 x i16> %shift
 ;
   %shift = ashr <32 x i16> %a, %b
   ret <32 x i16> %shift
@@ -331,6 +344,10 @@ define <16 x i8> @var_shift_v16i8(<16 x i8> %a, <16 x i8> %b) {
 ; AVX512BW-LABEL: 'var_shift_v16i8'
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %shift = ashr <16 x i8> %a, %b
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <16 x i8> %shift
+;
+; AVX512GFNI-LABEL: 'var_shift_v16i8'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %shift = ashr <16 x i8> %a, %b
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <16 x i8> %shift
 ;
   %shift = ashr <16 x i8> %a, %b
   ret <16 x i8> %shift
@@ -364,6 +381,10 @@ define <32 x i8> @var_shift_v32i8(<32 x i8> %a, <32 x i8> %b) {
 ; AVX512BW-LABEL: 'var_shift_v32i8'
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 13 for instruction: %shift = ashr <32 x i8> %a, %b
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <32 x i8> %shift
+;
+; AVX512GFNI-LABEL: 'var_shift_v32i8'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 13 for instruction: %shift = ashr <32 x i8> %a, %b
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <32 x i8> %shift
 ;
   %shift = ashr <32 x i8> %a, %b
   ret <32 x i8> %shift
@@ -397,6 +418,10 @@ define <64 x i8> @var_shift_v64i8(<64 x i8> %a, <64 x i8> %b) {
 ; AVX512BW-LABEL: 'var_shift_v64i8'
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 15 for instruction: %shift = ashr <64 x i8> %a, %b
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <64 x i8> %shift
+;
+; AVX512GFNI-LABEL: 'var_shift_v64i8'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 15 for instruction: %shift = ashr <64 x i8> %a, %b
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <64 x i8> %shift
 ;
   %shift = ashr <64 x i8> %a, %b
   ret <64 x i8> %shift
@@ -834,6 +859,12 @@ define <32 x i16> @splatvar_shift_v32i16(<32 x i16> %a, i16 %b) {
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %splat = shufflevector <32 x i16> %insert, <32 x i16> undef, <32 x i32> zeroinitializer
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %shift = ashr <32 x i16> %a, %splat
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <32 x i16> %shift
+;
+; AVX512GFNI-LABEL: 'splatvar_shift_v32i16'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %insert = insertelement <32 x i16> undef, i16 %b, i32 0
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %splat = shufflevector <32 x i16> %insert, <32 x i16> undef, <32 x i32> zeroinitializer
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %shift = ashr <32 x i16> %a, %splat
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <32 x i16> %shift
 ;
   %insert = insertelement <32 x i16> undef, i16 %b, i32 0
   %splat = shufflevector <32 x i16> %insert, <32 x i16> undef, <32 x i32> zeroinitializer
@@ -889,6 +920,12 @@ define <16 x i8> @splatvar_shift_v16i8(<16 x i8> %a, i8 %b) {
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %splat = shufflevector <16 x i8> %insert, <16 x i8> undef, <16 x i32> zeroinitializer
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 12 for instruction: %shift = ashr <16 x i8> %a, %splat
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <16 x i8> %shift
+;
+; AVX512GFNI-LABEL: 'splatvar_shift_v16i8'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %insert = insertelement <16 x i8> undef, i8 %b, i32 0
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %splat = shufflevector <16 x i8> %insert, <16 x i8> undef, <16 x i32> zeroinitializer
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 12 for instruction: %shift = ashr <16 x i8> %a, %splat
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <16 x i8> %shift
 ;
   %insert = insertelement <16 x i8> undef, i8 %b, i32 0
   %splat = shufflevector <16 x i8> %insert, <16 x i8> undef, <16 x i32> zeroinitializer
@@ -944,6 +981,12 @@ define <32 x i8> @splatvar_shift_v32i8(<32 x i8> %a, i8 %b) {
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %splat = shufflevector <32 x i8> %insert, <32 x i8> undef, <32 x i32> zeroinitializer
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 10 for instruction: %shift = ashr <32 x i8> %a, %splat
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <32 x i8> %shift
+;
+; AVX512GFNI-LABEL: 'splatvar_shift_v32i8'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %insert = insertelement <32 x i8> undef, i8 %b, i32 0
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %splat = shufflevector <32 x i8> %insert, <32 x i8> undef, <32 x i32> zeroinitializer
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 10 for instruction: %shift = ashr <32 x i8> %a, %splat
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <32 x i8> %shift
 ;
   %insert = insertelement <32 x i8> undef, i8 %b, i32 0
   %splat = shufflevector <32 x i8> %insert, <32 x i8> undef, <32 x i32> zeroinitializer
@@ -999,6 +1042,12 @@ define <64 x i8> @splatvar_shift_v64i8(<64 x i8> %a, i8 %b) {
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %splat = shufflevector <64 x i8> %insert, <64 x i8> undef, <64 x i32> zeroinitializer
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 10 for instruction: %shift = ashr <64 x i8> %a, %splat
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <64 x i8> %shift
+;
+; AVX512GFNI-LABEL: 'splatvar_shift_v64i8'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %insert = insertelement <64 x i8> undef, i8 %b, i32 0
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %splat = shufflevector <64 x i8> %insert, <64 x i8> undef, <64 x i32> zeroinitializer
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 10 for instruction: %shift = ashr <64 x i8> %a, %splat
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <64 x i8> %shift
 ;
   %insert = insertelement <64 x i8> undef, i8 %b, i32 0
   %splat = shufflevector <64 x i8> %insert, <64 x i8> undef, <64 x i32> zeroinitializer
@@ -1220,6 +1269,10 @@ define <8 x i16> @constant_shift_v8i16(<8 x i16> %a) {
 ; AVX512BW-LABEL: 'constant_shift_v8i16'
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shift = ashr <8 x i16> %a, <i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7>
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <8 x i16> %shift
+;
+; AVX512GFNI-LABEL: 'constant_shift_v8i16'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shift = ashr <8 x i16> %a, <i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7>
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <8 x i16> %shift
 ;
   %shift = ashr <8 x i16> %a, <i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7>
   ret <8 x i16> %shift
@@ -1253,6 +1306,10 @@ define <16 x i16> @constant_shift_v16i16(<16 x i16> %a) {
 ; AVX512BW-LABEL: 'constant_shift_v16i16'
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shift = ashr <16 x i16> %a, <i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7>
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <16 x i16> %shift
+;
+; AVX512GFNI-LABEL: 'constant_shift_v16i16'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shift = ashr <16 x i16> %a, <i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7>
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <16 x i16> %shift
 ;
   %shift = ashr <16 x i16> %a, <i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7>
   ret <16 x i16> %shift
@@ -1286,6 +1343,10 @@ define <32 x i16> @constant_shift_v32i16(<32 x i16> %a) {
 ; AVX512BW-LABEL: 'constant_shift_v32i16'
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shift = ashr <32 x i16> %a, <i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7>
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <32 x i16> %shift
+;
+; AVX512GFNI-LABEL: 'constant_shift_v32i16'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shift = ashr <32 x i16> %a, <i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7>
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <32 x i16> %shift
 ;
   %shift = ashr <32 x i16> %a, <i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7>
   ret <32 x i16> %shift
@@ -1319,6 +1380,10 @@ define <16 x i8> @constant_shift_v16i8(<16 x i8> %a) {
 ; AVX512BW-LABEL: 'constant_shift_v16i8'
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %shift = ashr <16 x i8> %a, <i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0>
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <16 x i8> %shift
+;
+; AVX512GFNI-LABEL: 'constant_shift_v16i8'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %shift = ashr <16 x i8> %a, <i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0>
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <16 x i8> %shift
 ;
   %shift = ashr <16 x i8> %a, <i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0>
   ret <16 x i8> %shift
@@ -1352,6 +1417,10 @@ define <32 x i8> @constant_shift_v32i8(<32 x i8> %a) {
 ; AVX512BW-LABEL: 'constant_shift_v32i8'
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 13 for instruction: %shift = ashr <32 x i8> %a, <i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0>
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <32 x i8> %shift
+;
+; AVX512GFNI-LABEL: 'constant_shift_v32i8'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 13 for instruction: %shift = ashr <32 x i8> %a, <i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0>
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <32 x i8> %shift
 ;
   %shift = ashr <32 x i8> %a, <i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0>
   ret <32 x i8> %shift
@@ -1385,6 +1454,10 @@ define <64 x i8> @constant_shift_v64i8(<64 x i8> %a) {
 ; AVX512BW-LABEL: 'constant_shift_v64i8'
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 15 for instruction: %shift = ashr <64 x i8> %a, <i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0>
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <64 x i8> %shift
+;
+; AVX512GFNI-LABEL: 'constant_shift_v64i8'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 15 for instruction: %shift = ashr <64 x i8> %a, <i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0>
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <64 x i8> %shift
 ;
   %shift = ashr <64 x i8> %a, <i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0>
   ret <64 x i8> %shift
@@ -1649,6 +1722,10 @@ define <16 x i16> @splatconstant_shift_v16i16(<16 x i16> %a) {
 ; AVX512BW-LABEL: 'splatconstant_shift_v16i16'
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shift = ashr <16 x i16> %a, <i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3>
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <16 x i16> %shift
+;
+; AVX512GFNI-LABEL: 'splatconstant_shift_v16i16'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shift = ashr <16 x i16> %a, <i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3>
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <16 x i16> %shift
 ;
   %shift = ashr <16 x i16> %a, <i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3>
   ret <16 x i16> %shift
@@ -1686,6 +1763,10 @@ define <32 x i16> @splatconstant_shift_v32i16(<32 x i16> %a) {
 ; AVX512BW-LABEL: 'splatconstant_shift_v32i16'
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shift = ashr <32 x i16> %a, <i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3>
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <32 x i16> %shift
+;
+; AVX512GFNI-LABEL: 'splatconstant_shift_v32i16'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shift = ashr <32 x i16> %a, <i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3>
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <32 x i16> %shift
 ;
   %shift = ashr <32 x i16> %a, <i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3>
   ret <32 x i16> %shift
@@ -1723,6 +1804,10 @@ define <16 x i8> @splatconstant_shift_v16i8(<16 x i8> %a) {
 ; AVX512BW-LABEL: 'splatconstant_shift_v16i8'
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %shift = ashr <16 x i8> %a, <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>
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <16 x i8> %shift
+;
+; AVX512GFNI-LABEL: 'splatconstant_shift_v16i8'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %shift = ashr <16 x i8> %a, <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>
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <16 x i8> %shift
 ;
   %shift = ashr <16 x i8> %a, <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>
   ret <16 x i8> %shift
@@ -1760,6 +1845,10 @@ define <32 x i8> @splatconstant_shift_v32i8(<32 x i8> %a) {
 ; AVX512BW-LABEL: 'splatconstant_shift_v32i8'
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 9 for instruction: %shift = ashr <32 x i8> %a, <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, 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>
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <32 x i8> %shift
+;
+; AVX512GFNI-LABEL: 'splatconstant_shift_v32i8'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 9 for instruction: %shift = ashr <32 x i8> %a, <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, 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>
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <32 x i8> %shift
 ;
   %shift = ashr <32 x i8> %a, <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, 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>
   ret <32 x i8> %shift
@@ -1797,6 +1886,10 @@ define <64 x i8> @splatconstant_shift_v64i8(<64 x i8> %a) {
 ; AVX512BW-LABEL: 'splatconstant_shift_v64i8'
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 9 for instruction: %shift = ashr <64 x i8> %a, <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, 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, 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, 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>
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <64 x i8> %shift
+;
+; AVX512GFNI-LABEL: 'splatconstant_shift_v64i8'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 9 for instruction: %shift = ashr <64 x i8> %a, <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, 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, 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, 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>
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <64 x i8> %shift
 ;
   %shift = ashr <64 x i8> %a, <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, 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, 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, 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>
   ret <64 x i8> %shift

diff  --git a/llvm/test/Analysis/CostModel/X86/vshift-ashr-sizelatency.ll b/llvm/test/Analysis/CostModel/X86/vshift-ashr-sizelatency.ll
index 8286b3967af606..84ccad0294155f 100644
--- a/llvm/test/Analysis/CostModel/X86/vshift-ashr-sizelatency.ll
+++ b/llvm/test/Analysis/CostModel/X86/vshift-ashr-sizelatency.ll
@@ -15,6 +15,7 @@
 ; RUN: opt < %s -mtriple=x86_64-unknown-linux-gnu -passes="print<cost-model>" 2>&1 -disable-output -cost-kind=size-latency -mcpu=slm | FileCheck %s --check-prefixes=CHECK,SSE42
 ; RUN: opt < %s -mtriple=x86_64-unknown-linux-gnu -passes="print<cost-model>" 2>&1 -disable-output -cost-kind=size-latency -mcpu=goldmont | FileCheck %s --check-prefixes=CHECK,SSE42
 ; RUN: opt < %s -mtriple=x86_64-unknown-linux-gnu -passes="print<cost-model>" 2>&1 -disable-output -cost-kind=size-latency -mcpu=btver2 | FileCheck %s --check-prefixes=CHECK,AVX,AVX1
+; RUN: opt < %s -mtriple=x86_64-unknown-linux-gnu -passes="print<cost-model>" 2>&1 -disable-output -cost-kind=size-latency -mcpu=tigerlake | FileCheck %s --check-prefixes=CHECK,AVX512,AVX512GFNI
 
 ; Verify the cost of vector logical shift right instructions.
 
@@ -232,6 +233,10 @@ define <8 x i16> @var_shift_v8i16(<8 x i16> %a, <8 x i16> %b) {
 ; AVX512BW-LABEL: 'var_shift_v8i16'
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shift = ashr <8 x i16> %a, %b
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <8 x i16> %shift
+;
+; AVX512GFNI-LABEL: 'var_shift_v8i16'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shift = ashr <8 x i16> %a, %b
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <8 x i16> %shift
 ;
   %shift = ashr <8 x i16> %a, %b
   ret <8 x i16> %shift
@@ -265,6 +270,10 @@ define <16 x i16> @var_shift_v16i16(<16 x i16> %a, <16 x i16> %b) {
 ; AVX512BW-LABEL: 'var_shift_v16i16'
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shift = ashr <16 x i16> %a, %b
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <16 x i16> %shift
+;
+; AVX512GFNI-LABEL: 'var_shift_v16i16'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shift = ashr <16 x i16> %a, %b
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <16 x i16> %shift
 ;
   %shift = ashr <16 x i16> %a, %b
   ret <16 x i16> %shift
@@ -298,6 +307,10 @@ define <32 x i16> @var_shift_v32i16(<32 x i16> %a, <32 x i16> %b) {
 ; AVX512BW-LABEL: 'var_shift_v32i16'
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shift = ashr <32 x i16> %a, %b
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <32 x i16> %shift
+;
+; AVX512GFNI-LABEL: 'var_shift_v32i16'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shift = ashr <32 x i16> %a, %b
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <32 x i16> %shift
 ;
   %shift = ashr <32 x i16> %a, %b
   ret <32 x i16> %shift
@@ -331,6 +344,10 @@ define <16 x i8> @var_shift_v16i8(<16 x i8> %a, <16 x i8> %b) {
 ; AVX512BW-LABEL: 'var_shift_v16i8'
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 5 for instruction: %shift = ashr <16 x i8> %a, %b
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <16 x i8> %shift
+;
+; AVX512GFNI-LABEL: 'var_shift_v16i8'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 5 for instruction: %shift = ashr <16 x i8> %a, %b
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <16 x i8> %shift
 ;
   %shift = ashr <16 x i8> %a, %b
   ret <16 x i8> %shift
@@ -364,6 +381,10 @@ define <32 x i8> @var_shift_v32i8(<32 x i8> %a, <32 x i8> %b) {
 ; AVX512BW-LABEL: 'var_shift_v32i8'
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 30 for instruction: %shift = ashr <32 x i8> %a, %b
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <32 x i8> %shift
+;
+; AVX512GFNI-LABEL: 'var_shift_v32i8'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 30 for instruction: %shift = ashr <32 x i8> %a, %b
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <32 x i8> %shift
 ;
   %shift = ashr <32 x i8> %a, %b
   ret <32 x i8> %shift
@@ -397,6 +418,10 @@ define <64 x i8> @var_shift_v64i8(<64 x i8> %a, <64 x i8> %b) {
 ; AVX512BW-LABEL: 'var_shift_v64i8'
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 30 for instruction: %shift = ashr <64 x i8> %a, %b
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <64 x i8> %shift
+;
+; AVX512GFNI-LABEL: 'var_shift_v64i8'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 30 for instruction: %shift = ashr <64 x i8> %a, %b
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <64 x i8> %shift
 ;
   %shift = ashr <64 x i8> %a, %b
   ret <64 x i8> %shift
@@ -786,6 +811,12 @@ define <32 x i16> @splatvar_shift_v32i16(<32 x i16> %a, i16 %b) {
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %splat = shufflevector <32 x i16> %insert, <32 x i16> undef, <32 x i32> zeroinitializer
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %shift = ashr <32 x i16> %a, %splat
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <32 x i16> %shift
+;
+; AVX512GFNI-LABEL: 'splatvar_shift_v32i16'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %insert = insertelement <32 x i16> undef, i16 %b, i32 0
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %splat = shufflevector <32 x i16> %insert, <32 x i16> undef, <32 x i32> zeroinitializer
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %shift = ashr <32 x i16> %a, %splat
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <32 x i16> %shift
 ;
   %insert = insertelement <32 x i16> undef, i16 %b, i32 0
   %splat = shufflevector <32 x i16> %insert, <32 x i16> undef, <32 x i32> zeroinitializer
@@ -835,6 +866,12 @@ define <16 x i8> @splatvar_shift_v16i8(<16 x i8> %a, i8 %b) {
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %splat = shufflevector <16 x i8> %insert, <16 x i8> undef, <16 x i32> zeroinitializer
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 12 for instruction: %shift = ashr <16 x i8> %a, %splat
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <16 x i8> %shift
+;
+; AVX512GFNI-LABEL: 'splatvar_shift_v16i8'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %insert = insertelement <16 x i8> undef, i8 %b, i32 0
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %splat = shufflevector <16 x i8> %insert, <16 x i8> undef, <16 x i32> zeroinitializer
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 12 for instruction: %shift = ashr <16 x i8> %a, %splat
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <16 x i8> %shift
 ;
   %insert = insertelement <16 x i8> undef, i8 %b, i32 0
   %splat = shufflevector <16 x i8> %insert, <16 x i8> undef, <16 x i32> zeroinitializer
@@ -890,6 +927,12 @@ define <32 x i8> @splatvar_shift_v32i8(<32 x i8> %a, i8 %b) {
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %splat = shufflevector <32 x i8> %insert, <32 x i8> undef, <32 x i32> zeroinitializer
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 13 for instruction: %shift = ashr <32 x i8> %a, %splat
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <32 x i8> %shift
+;
+; AVX512GFNI-LABEL: 'splatvar_shift_v32i8'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %insert = insertelement <32 x i8> undef, i8 %b, i32 0
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %splat = shufflevector <32 x i8> %insert, <32 x i8> undef, <32 x i32> zeroinitializer
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 13 for instruction: %shift = ashr <32 x i8> %a, %splat
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <32 x i8> %shift
 ;
   %insert = insertelement <32 x i8> undef, i8 %b, i32 0
   %splat = shufflevector <32 x i8> %insert, <32 x i8> undef, <32 x i32> zeroinitializer
@@ -945,6 +988,12 @@ define <64 x i8> @splatvar_shift_v64i8(<64 x i8> %a, i8 %b) {
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %splat = shufflevector <64 x i8> %insert, <64 x i8> undef, <64 x i32> zeroinitializer
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 15 for instruction: %shift = ashr <64 x i8> %a, %splat
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <64 x i8> %shift
+;
+; AVX512GFNI-LABEL: 'splatvar_shift_v64i8'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %insert = insertelement <64 x i8> undef, i8 %b, i32 0
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %splat = shufflevector <64 x i8> %insert, <64 x i8> undef, <64 x i32> zeroinitializer
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 15 for instruction: %shift = ashr <64 x i8> %a, %splat
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <64 x i8> %shift
 ;
   %insert = insertelement <64 x i8> undef, i8 %b, i32 0
   %splat = shufflevector <64 x i8> %insert, <64 x i8> undef, <64 x i32> zeroinitializer
@@ -1166,6 +1215,10 @@ define <8 x i16> @constant_shift_v8i16(<8 x i16> %a) {
 ; AVX512BW-LABEL: 'constant_shift_v8i16'
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shift = ashr <8 x i16> %a, <i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7>
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <8 x i16> %shift
+;
+; AVX512GFNI-LABEL: 'constant_shift_v8i16'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shift = ashr <8 x i16> %a, <i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7>
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <8 x i16> %shift
 ;
   %shift = ashr <8 x i16> %a, <i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7>
   ret <8 x i16> %shift
@@ -1199,6 +1252,10 @@ define <16 x i16> @constant_shift_v16i16(<16 x i16> %a) {
 ; AVX512BW-LABEL: 'constant_shift_v16i16'
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shift = ashr <16 x i16> %a, <i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7>
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <16 x i16> %shift
+;
+; AVX512GFNI-LABEL: 'constant_shift_v16i16'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shift = ashr <16 x i16> %a, <i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7>
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <16 x i16> %shift
 ;
   %shift = ashr <16 x i16> %a, <i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7>
   ret <16 x i16> %shift
@@ -1232,6 +1289,10 @@ define <32 x i16> @constant_shift_v32i16(<32 x i16> %a) {
 ; AVX512BW-LABEL: 'constant_shift_v32i16'
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shift = ashr <32 x i16> %a, <i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7>
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <32 x i16> %shift
+;
+; AVX512GFNI-LABEL: 'constant_shift_v32i16'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shift = ashr <32 x i16> %a, <i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7>
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <32 x i16> %shift
 ;
   %shift = ashr <32 x i16> %a, <i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7>
   ret <32 x i16> %shift
@@ -1265,6 +1326,10 @@ define <16 x i8> @constant_shift_v16i8(<16 x i8> %a) {
 ; AVX512BW-LABEL: 'constant_shift_v16i8'
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 5 for instruction: %shift = ashr <16 x i8> %a, <i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0>
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <16 x i8> %shift
+;
+; AVX512GFNI-LABEL: 'constant_shift_v16i8'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 5 for instruction: %shift = ashr <16 x i8> %a, <i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0>
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <16 x i8> %shift
 ;
   %shift = ashr <16 x i8> %a, <i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0>
   ret <16 x i8> %shift
@@ -1298,6 +1363,10 @@ define <32 x i8> @constant_shift_v32i8(<32 x i8> %a) {
 ; AVX512BW-LABEL: 'constant_shift_v32i8'
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 30 for instruction: %shift = ashr <32 x i8> %a, <i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0>
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <32 x i8> %shift
+;
+; AVX512GFNI-LABEL: 'constant_shift_v32i8'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 30 for instruction: %shift = ashr <32 x i8> %a, <i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0>
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <32 x i8> %shift
 ;
   %shift = ashr <32 x i8> %a, <i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0>
   ret <32 x i8> %shift
@@ -1331,6 +1400,10 @@ define <64 x i8> @constant_shift_v64i8(<64 x i8> %a) {
 ; AVX512BW-LABEL: 'constant_shift_v64i8'
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 30 for instruction: %shift = ashr <64 x i8> %a, <i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0>
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <64 x i8> %shift
+;
+; AVX512GFNI-LABEL: 'constant_shift_v64i8'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 30 for instruction: %shift = ashr <64 x i8> %a, <i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0>
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <64 x i8> %shift
 ;
   %shift = ashr <64 x i8> %a, <i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0>
   ret <64 x i8> %shift
@@ -1547,6 +1620,10 @@ define <16 x i16> @splatconstant_shift_v16i16(<16 x i16> %a) {
 ; AVX512BW-LABEL: 'splatconstant_shift_v16i16'
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shift = ashr <16 x i16> %a, <i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3>
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <16 x i16> %shift
+;
+; AVX512GFNI-LABEL: 'splatconstant_shift_v16i16'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shift = ashr <16 x i16> %a, <i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3>
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <16 x i16> %shift
 ;
   %shift = ashr <16 x i16> %a, <i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3>
   ret <16 x i16> %shift
@@ -1584,6 +1661,10 @@ define <32 x i16> @splatconstant_shift_v32i16(<32 x i16> %a) {
 ; AVX512BW-LABEL: 'splatconstant_shift_v32i16'
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shift = ashr <32 x i16> %a, <i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3>
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <32 x i16> %shift
+;
+; AVX512GFNI-LABEL: 'splatconstant_shift_v32i16'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shift = ashr <32 x i16> %a, <i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3>
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <32 x i16> %shift
 ;
   %shift = ashr <32 x i16> %a, <i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3>
   ret <32 x i16> %shift
@@ -1617,6 +1698,10 @@ define <16 x i8> @splatconstant_shift_v16i8(<16 x i8> %a) {
 ; AVX512BW-LABEL: 'splatconstant_shift_v16i8'
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 5 for instruction: %shift = ashr <16 x i8> %a, <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>
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <16 x i8> %shift
+;
+; AVX512GFNI-LABEL: 'splatconstant_shift_v16i8'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 5 for instruction: %shift = ashr <16 x i8> %a, <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>
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <16 x i8> %shift
 ;
   %shift = ashr <16 x i8> %a, <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>
   ret <16 x i8> %shift
@@ -1654,6 +1739,10 @@ define <32 x i8> @splatconstant_shift_v32i8(<32 x i8> %a) {
 ; AVX512BW-LABEL: 'splatconstant_shift_v32i8'
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 5 for instruction: %shift = ashr <32 x i8> %a, <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, 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>
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <32 x i8> %shift
+;
+; AVX512GFNI-LABEL: 'splatconstant_shift_v32i8'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 5 for instruction: %shift = ashr <32 x i8> %a, <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, 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>
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <32 x i8> %shift
 ;
   %shift = ashr <32 x i8> %a, <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, 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>
   ret <32 x i8> %shift
@@ -1691,6 +1780,10 @@ define <64 x i8> @splatconstant_shift_v64i8(<64 x i8> %a) {
 ; AVX512BW-LABEL: 'splatconstant_shift_v64i8'
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 6 for instruction: %shift = ashr <64 x i8> %a, <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, 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, 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, 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>
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <64 x i8> %shift
+;
+; AVX512GFNI-LABEL: 'splatconstant_shift_v64i8'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 6 for instruction: %shift = ashr <64 x i8> %a, <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, 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, 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, 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>
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <64 x i8> %shift
 ;
   %shift = ashr <64 x i8> %a, <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, 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, 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, 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>
   ret <64 x i8> %shift

diff  --git a/llvm/test/Analysis/CostModel/X86/vshift-lshr-codesize.ll b/llvm/test/Analysis/CostModel/X86/vshift-lshr-codesize.ll
index 7c506af62e8178..a0e15bb8ff738e 100644
--- a/llvm/test/Analysis/CostModel/X86/vshift-lshr-codesize.ll
+++ b/llvm/test/Analysis/CostModel/X86/vshift-lshr-codesize.ll
@@ -15,6 +15,7 @@
 ; RUN: opt < %s -mtriple=x86_64-unknown-linux-gnu -passes="print<cost-model>" 2>&1 -disable-output -cost-kind=code-size -mcpu=slm | FileCheck %s --check-prefixes=CHECK,SSE42
 ; RUN: opt < %s -mtriple=x86_64-unknown-linux-gnu -passes="print<cost-model>" 2>&1 -disable-output -cost-kind=code-size -mcpu=goldmont | FileCheck %s --check-prefixes=CHECK,SSE42
 ; RUN: opt < %s -mtriple=x86_64-unknown-linux-gnu -passes="print<cost-model>" 2>&1 -disable-output -cost-kind=code-size -mcpu=btver2 | FileCheck %s --check-prefixes=CHECK,AVX,AVX1
+; RUN: opt < %s -mtriple=x86_64-unknown-linux-gnu -passes="print<cost-model>" 2>&1 -disable-output -cost-kind=code-size -mcpu=tigerlake | FileCheck %s --check-prefixes=CHECK,AVX512,AVX512GFNI
 
 ; Verify the cost of vector logical shift right instructions.
 
@@ -240,6 +241,10 @@ define <8 x i16> @var_shift_v8i16(<8 x i16> %a, <8 x i16> %b) {
 ; AVX512BW-LABEL: 'var_shift_v8i16'
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shift = lshr <8 x i16> %a, %b
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <8 x i16> %shift
+;
+; AVX512GFNI-LABEL: 'var_shift_v8i16'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shift = lshr <8 x i16> %a, %b
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <8 x i16> %shift
 ;
   %shift = lshr <8 x i16> %a, %b
   ret <8 x i16> %shift
@@ -273,6 +278,10 @@ define <16 x i16> @var_shift_v16i16(<16 x i16> %a, <16 x i16> %b) {
 ; AVX512BW-LABEL: 'var_shift_v16i16'
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shift = lshr <16 x i16> %a, %b
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <16 x i16> %shift
+;
+; AVX512GFNI-LABEL: 'var_shift_v16i16'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shift = lshr <16 x i16> %a, %b
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <16 x i16> %shift
 ;
   %shift = lshr <16 x i16> %a, %b
   ret <16 x i16> %shift
@@ -306,6 +315,10 @@ define <32 x i16> @var_shift_v32i16(<32 x i16> %a, <32 x i16> %b) {
 ; AVX512BW-LABEL: 'var_shift_v32i16'
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shift = lshr <32 x i16> %a, %b
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <32 x i16> %shift
+;
+; AVX512GFNI-LABEL: 'var_shift_v32i16'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shift = lshr <32 x i16> %a, %b
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <32 x i16> %shift
 ;
   %shift = lshr <32 x i16> %a, %b
   ret <32 x i16> %shift
@@ -335,6 +348,10 @@ define <16 x i8> @var_shift_v16i8(<16 x i8> %a, <16 x i8> %b) {
 ; AVX512BW-LABEL: 'var_shift_v16i8'
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %shift = lshr <16 x i8> %a, %b
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <16 x i8> %shift
+;
+; AVX512GFNI-LABEL: 'var_shift_v16i8'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %shift = lshr <16 x i8> %a, %b
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <16 x i8> %shift
 ;
   %shift = lshr <16 x i8> %a, %b
   ret <16 x i8> %shift
@@ -397,6 +414,10 @@ define <64 x i8> @var_shift_v64i8(<64 x i8> %a, <64 x i8> %b) {
 ; AVX512BW-LABEL: 'var_shift_v64i8'
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 15 for instruction: %shift = lshr <64 x i8> %a, %b
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <64 x i8> %shift
+;
+; AVX512GFNI-LABEL: 'var_shift_v64i8'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 15 for instruction: %shift = lshr <64 x i8> %a, %b
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <64 x i8> %shift
 ;
   %shift = lshr <64 x i8> %a, %b
   ret <64 x i8> %shift
@@ -762,6 +783,12 @@ define <32 x i16> @splatvar_shift_v32i16(<32 x i16> %a, i16 %b) {
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %splat = shufflevector <32 x i16> %insert, <32 x i16> undef, <32 x i32> zeroinitializer
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %shift = lshr <32 x i16> %a, %splat
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <32 x i16> %shift
+;
+; AVX512GFNI-LABEL: 'splatvar_shift_v32i16'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %insert = insertelement <32 x i16> undef, i16 %b, i32 0
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %splat = shufflevector <32 x i16> %insert, <32 x i16> undef, <32 x i32> zeroinitializer
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %shift = lshr <32 x i16> %a, %splat
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <32 x i16> %shift
 ;
   %insert = insertelement <32 x i16> undef, i16 %b, i32 0
   %splat = shufflevector <32 x i16> %insert, <32 x i16> undef, <32 x i32> zeroinitializer
@@ -909,6 +936,12 @@ define <64 x i8> @splatvar_shift_v64i8(<64 x i8> %a, i8 %b) {
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %splat = shufflevector <64 x i8> %insert, <64 x i8> undef, <64 x i32> zeroinitializer
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 7 for instruction: %shift = lshr <64 x i8> %a, %splat
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <64 x i8> %shift
+;
+; AVX512GFNI-LABEL: 'splatvar_shift_v64i8'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %insert = insertelement <64 x i8> undef, i8 %b, i32 0
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %splat = shufflevector <64 x i8> %insert, <64 x i8> undef, <64 x i32> zeroinitializer
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 7 for instruction: %shift = lshr <64 x i8> %a, %splat
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <64 x i8> %shift
 ;
   %insert = insertelement <64 x i8> undef, i8 %b, i32 0
   %splat = shufflevector <64 x i8> %insert, <64 x i8> undef, <64 x i32> zeroinitializer
@@ -1138,6 +1171,10 @@ define <8 x i16> @constant_shift_v8i16(<8 x i16> %a) {
 ; AVX512BW-LABEL: 'constant_shift_v8i16'
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shift = lshr <8 x i16> %a, <i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7>
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <8 x i16> %shift
+;
+; AVX512GFNI-LABEL: 'constant_shift_v8i16'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shift = lshr <8 x i16> %a, <i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7>
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <8 x i16> %shift
 ;
   %shift = lshr <8 x i16> %a, <i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7>
   ret <8 x i16> %shift
@@ -1171,6 +1208,10 @@ define <16 x i16> @constant_shift_v16i16(<16 x i16> %a) {
 ; AVX512BW-LABEL: 'constant_shift_v16i16'
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shift = lshr <16 x i16> %a, <i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7>
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <16 x i16> %shift
+;
+; AVX512GFNI-LABEL: 'constant_shift_v16i16'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shift = lshr <16 x i16> %a, <i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7>
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <16 x i16> %shift
 ;
   %shift = lshr <16 x i16> %a, <i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7>
   ret <16 x i16> %shift
@@ -1204,6 +1245,10 @@ define <32 x i16> @constant_shift_v32i16(<32 x i16> %a) {
 ; AVX512BW-LABEL: 'constant_shift_v32i16'
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shift = lshr <32 x i16> %a, <i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7>
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <32 x i16> %shift
+;
+; AVX512GFNI-LABEL: 'constant_shift_v32i16'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shift = lshr <32 x i16> %a, <i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7>
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <32 x i16> %shift
 ;
   %shift = lshr <32 x i16> %a, <i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7>
   ret <32 x i16> %shift
@@ -1233,6 +1278,10 @@ define <16 x i8> @constant_shift_v16i8(<16 x i8> %a) {
 ; AVX512BW-LABEL: 'constant_shift_v16i8'
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %shift = lshr <16 x i8> %a, <i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0>
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <16 x i8> %shift
+;
+; AVX512GFNI-LABEL: 'constant_shift_v16i8'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %shift = lshr <16 x i8> %a, <i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0>
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <16 x i8> %shift
 ;
   %shift = lshr <16 x i8> %a, <i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0>
   ret <16 x i8> %shift
@@ -1295,6 +1344,10 @@ define <64 x i8> @constant_shift_v64i8(<64 x i8> %a) {
 ; AVX512BW-LABEL: 'constant_shift_v64i8'
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 15 for instruction: %shift = lshr <64 x i8> %a, <i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0>
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <64 x i8> %shift
+;
+; AVX512GFNI-LABEL: 'constant_shift_v64i8'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 15 for instruction: %shift = lshr <64 x i8> %a, <i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0>
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <64 x i8> %shift
 ;
   %shift = lshr <64 x i8> %a, <i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0>
   ret <64 x i8> %shift
@@ -1495,6 +1548,10 @@ define <16 x i16> @splatconstant_shift_v16i16(<16 x i16> %a) {
 ; AVX512BW-LABEL: 'splatconstant_shift_v16i16'
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shift = lshr <16 x i16> %a, <i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3>
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <16 x i16> %shift
+;
+; AVX512GFNI-LABEL: 'splatconstant_shift_v16i16'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shift = lshr <16 x i16> %a, <i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3>
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <16 x i16> %shift
 ;
   %shift = lshr <16 x i16> %a, <i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3>
   ret <16 x i16> %shift
@@ -1532,6 +1589,10 @@ define <32 x i16> @splatconstant_shift_v32i16(<32 x i16> %a) {
 ; AVX512BW-LABEL: 'splatconstant_shift_v32i16'
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shift = lshr <32 x i16> %a, <i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3>
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <32 x i16> %shift
+;
+; AVX512GFNI-LABEL: 'splatconstant_shift_v32i16'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shift = lshr <32 x i16> %a, <i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3>
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <32 x i16> %shift
 ;
   %shift = lshr <32 x i16> %a, <i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3>
   ret <32 x i16> %shift
@@ -1631,6 +1692,10 @@ define <64 x i8> @splatconstant_shift_v64i8(<64 x i8> %a) {
 ; AVX512BW-LABEL: 'splatconstant_shift_v64i8'
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %shift = lshr <64 x i8> %a, <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, 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, 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, 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>
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <64 x i8> %shift
+;
+; AVX512GFNI-LABEL: 'splatconstant_shift_v64i8'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %shift = lshr <64 x i8> %a, <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, 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, 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, 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>
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <64 x i8> %shift
 ;
   %shift = lshr <64 x i8> %a, <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, 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, 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, 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>
   ret <64 x i8> %shift

diff  --git a/llvm/test/Analysis/CostModel/X86/vshift-lshr-cost-inseltpoison.ll b/llvm/test/Analysis/CostModel/X86/vshift-lshr-cost-inseltpoison.ll
index 855d2b8e88e3a5..34ea1a644f6cca 100644
--- a/llvm/test/Analysis/CostModel/X86/vshift-lshr-cost-inseltpoison.ll
+++ b/llvm/test/Analysis/CostModel/X86/vshift-lshr-cost-inseltpoison.ll
@@ -15,6 +15,7 @@
 ; RUN: opt < %s -mtriple=x86_64-unknown-linux-gnu -passes="print<cost-model>" 2>&1 -disable-output -mcpu=slm | FileCheck %s --check-prefixes=CHECK,SSE,SSE42
 ; RUN: opt < %s -mtriple=x86_64-unknown-linux-gnu -passes="print<cost-model>" 2>&1 -disable-output -mcpu=goldmont | FileCheck %s --check-prefixes=CHECK,SSE,SSE42
 ; RUN: opt < %s -mtriple=x86_64-unknown-linux-gnu -passes="print<cost-model>" 2>&1 -disable-output -mcpu=btver2 | FileCheck %s --check-prefixes=CHECK,AVX,AVX1
+; RUN: opt < %s -mtriple=x86_64-unknown-linux-gnu -passes="print<cost-model>" 2>&1 -disable-output -mcpu=tigerlake | FileCheck %s --check-prefixes=CHECK,AVX512,AVX512GFNI
 
 ; Verify the cost of vector logical shift right instructions.
 
@@ -232,6 +233,10 @@ define <8 x i16> @var_shift_v8i16(<8 x i16> %a, <8 x i16> %b) {
 ; AVX512BWVL-LABEL: 'var_shift_v8i16'
 ; AVX512BWVL-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shift = lshr <8 x i16> %a, %b
 ; AVX512BWVL-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <8 x i16> %shift
+;
+; AVX512GFNI-LABEL: 'var_shift_v8i16'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shift = lshr <8 x i16> %a, %b
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <8 x i16> %shift
 ;
   %shift = lshr <8 x i16> %a, %b
   ret <8 x i16> %shift
@@ -273,6 +278,10 @@ define <16 x i16> @var_shift_v16i16(<16 x i16> %a, <16 x i16> %b) {
 ; AVX512BWVL-LABEL: 'var_shift_v16i16'
 ; AVX512BWVL-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shift = lshr <16 x i16> %a, %b
 ; AVX512BWVL-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <16 x i16> %shift
+;
+; AVX512GFNI-LABEL: 'var_shift_v16i16'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shift = lshr <16 x i16> %a, %b
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <16 x i16> %shift
 ;
   %shift = lshr <16 x i16> %a, %b
   ret <16 x i16> %shift
@@ -314,6 +323,10 @@ define <32 x i16> @var_shift_v32i16(<32 x i16> %a, <32 x i16> %b) {
 ; AVX512BWVL-LABEL: 'var_shift_v32i16'
 ; AVX512BWVL-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shift = lshr <32 x i16> %a, %b
 ; AVX512BWVL-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <32 x i16> %shift
+;
+; AVX512GFNI-LABEL: 'var_shift_v32i16'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shift = lshr <32 x i16> %a, %b
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <32 x i16> %shift
 ;
   %shift = lshr <32 x i16> %a, %b
   ret <32 x i16> %shift
@@ -355,6 +368,10 @@ define <16 x i8> @var_shift_v16i8(<16 x i8> %a, <16 x i8> %b) {
 ; AVX512BWVL-LABEL: 'var_shift_v16i8'
 ; AVX512BWVL-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %shift = lshr <16 x i8> %a, %b
 ; AVX512BWVL-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <16 x i8> %shift
+;
+; AVX512GFNI-LABEL: 'var_shift_v16i8'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %shift = lshr <16 x i8> %a, %b
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <16 x i8> %shift
 ;
   %shift = lshr <16 x i8> %a, %b
   ret <16 x i8> %shift
@@ -396,6 +413,10 @@ define <32 x i8> @var_shift_v32i8(<32 x i8> %a, <32 x i8> %b) {
 ; AVX512BWVL-LABEL: 'var_shift_v32i8'
 ; AVX512BWVL-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %shift = lshr <32 x i8> %a, %b
 ; AVX512BWVL-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <32 x i8> %shift
+;
+; AVX512GFNI-LABEL: 'var_shift_v32i8'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %shift = lshr <32 x i8> %a, %b
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <32 x i8> %shift
 ;
   %shift = lshr <32 x i8> %a, %b
   ret <32 x i8> %shift
@@ -437,6 +458,10 @@ define <64 x i8> @var_shift_v64i8(<64 x i8> %a, <64 x i8> %b) {
 ; AVX512BWVL-LABEL: 'var_shift_v64i8'
 ; AVX512BWVL-NEXT:  Cost Model: Found an estimated cost of 7 for instruction: %shift = lshr <64 x i8> %a, %b
 ; AVX512BWVL-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <64 x i8> %shift
+;
+; AVX512GFNI-LABEL: 'var_shift_v64i8'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 7 for instruction: %shift = lshr <64 x i8> %a, %b
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <64 x i8> %shift
 ;
   %shift = lshr <64 x i8> %a, %b
   ret <64 x i8> %shift
@@ -826,6 +851,12 @@ define <32 x i16> @splatvar_shift_v32i16(<32 x i16> %a, i16 %b) {
 ; AVX512BWVL-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %splat = shufflevector <32 x i16> %insert, <32 x i16> poison, <32 x i32> zeroinitializer
 ; AVX512BWVL-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %shift = lshr <32 x i16> %a, %splat
 ; AVX512BWVL-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <32 x i16> %shift
+;
+; AVX512GFNI-LABEL: 'splatvar_shift_v32i16'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %insert = insertelement <32 x i16> poison, i16 %b, i32 0
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %splat = shufflevector <32 x i16> %insert, <32 x i16> poison, <32 x i32> zeroinitializer
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %shift = lshr <32 x i16> %a, %splat
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <32 x i16> %shift
 ;
   %insert = insertelement <32 x i16> poison, i16 %b, i32 0
   %splat = shufflevector <32 x i16> %insert, <32 x i16> poison, <32 x i32> zeroinitializer
@@ -991,6 +1022,12 @@ define <64 x i8> @splatvar_shift_v64i8(<64 x i8> %a, i8 %b) {
 ; AVX512BWVL-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %splat = shufflevector <64 x i8> %insert, <64 x i8> poison, <64 x i32> zeroinitializer
 ; AVX512BWVL-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %shift = lshr <64 x i8> %a, %splat
 ; AVX512BWVL-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <64 x i8> %shift
+;
+; AVX512GFNI-LABEL: 'splatvar_shift_v64i8'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %insert = insertelement <64 x i8> poison, i8 %b, i32 0
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %splat = shufflevector <64 x i8> %insert, <64 x i8> poison, <64 x i32> zeroinitializer
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %shift = lshr <64 x i8> %a, %splat
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <64 x i8> %shift
 ;
   %insert = insertelement <64 x i8> poison, i8 %b, i32 0
   %splat = shufflevector <64 x i8> %insert, <64 x i8> poison, <64 x i32> zeroinitializer
@@ -1200,6 +1237,10 @@ define <8 x i16> @constant_shift_v8i16(<8 x i16> %a) {
 ; AVX512BWVL-LABEL: 'constant_shift_v8i16'
 ; AVX512BWVL-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shift = lshr <8 x i16> %a, <i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7>
 ; AVX512BWVL-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <8 x i16> %shift
+;
+; AVX512GFNI-LABEL: 'constant_shift_v8i16'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shift = lshr <8 x i16> %a, <i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7>
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <8 x i16> %shift
 ;
   %shift = lshr <8 x i16> %a, <i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7>
   ret <8 x i16> %shift
@@ -1241,6 +1282,10 @@ define <16 x i16> @constant_shift_v16i16(<16 x i16> %a) {
 ; AVX512BWVL-LABEL: 'constant_shift_v16i16'
 ; AVX512BWVL-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shift = lshr <16 x i16> %a, <i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7>
 ; AVX512BWVL-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <16 x i16> %shift
+;
+; AVX512GFNI-LABEL: 'constant_shift_v16i16'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shift = lshr <16 x i16> %a, <i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7>
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <16 x i16> %shift
 ;
   %shift = lshr <16 x i16> %a, <i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7>
   ret <16 x i16> %shift
@@ -1282,6 +1327,10 @@ define <32 x i16> @constant_shift_v32i16(<32 x i16> %a) {
 ; AVX512BWVL-LABEL: 'constant_shift_v32i16'
 ; AVX512BWVL-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shift = lshr <32 x i16> %a, <i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7>
 ; AVX512BWVL-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <32 x i16> %shift
+;
+; AVX512GFNI-LABEL: 'constant_shift_v32i16'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shift = lshr <32 x i16> %a, <i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7>
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <32 x i16> %shift
 ;
   %shift = lshr <32 x i16> %a, <i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7>
   ret <32 x i16> %shift
@@ -1323,6 +1372,10 @@ define <16 x i8> @constant_shift_v16i8(<16 x i8> %a) {
 ; AVX512BWVL-LABEL: 'constant_shift_v16i8'
 ; AVX512BWVL-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %shift = lshr <16 x i8> %a, <i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0>
 ; AVX512BWVL-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <16 x i8> %shift
+;
+; AVX512GFNI-LABEL: 'constant_shift_v16i8'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %shift = lshr <16 x i8> %a, <i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0>
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <16 x i8> %shift
 ;
   %shift = lshr <16 x i8> %a, <i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0>
   ret <16 x i8> %shift
@@ -1364,6 +1417,10 @@ define <32 x i8> @constant_shift_v32i8(<32 x i8> %a) {
 ; AVX512BWVL-LABEL: 'constant_shift_v32i8'
 ; AVX512BWVL-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %shift = lshr <32 x i8> %a, <i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0>
 ; AVX512BWVL-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <32 x i8> %shift
+;
+; AVX512GFNI-LABEL: 'constant_shift_v32i8'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %shift = lshr <32 x i8> %a, <i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0>
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <32 x i8> %shift
 ;
   %shift = lshr <32 x i8> %a, <i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0>
   ret <32 x i8> %shift
@@ -1405,6 +1462,10 @@ define <64 x i8> @constant_shift_v64i8(<64 x i8> %a) {
 ; AVX512BWVL-LABEL: 'constant_shift_v64i8'
 ; AVX512BWVL-NEXT:  Cost Model: Found an estimated cost of 7 for instruction: %shift = lshr <64 x i8> %a, <i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0>
 ; AVX512BWVL-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <64 x i8> %shift
+;
+; AVX512GFNI-LABEL: 'constant_shift_v64i8'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 7 for instruction: %shift = lshr <64 x i8> %a, <i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0>
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <64 x i8> %shift
 ;
   %shift = lshr <64 x i8> %a, <i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0>
   ret <64 x i8> %shift
@@ -1593,6 +1654,10 @@ define <16 x i16> @splatconstant_shift_v16i16(<16 x i16> %a) {
 ; AVX512BWVL-LABEL: 'splatconstant_shift_v16i16'
 ; AVX512BWVL-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shift = lshr <16 x i16> %a, <i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3>
 ; AVX512BWVL-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <16 x i16> %shift
+;
+; AVX512GFNI-LABEL: 'splatconstant_shift_v16i16'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shift = lshr <16 x i16> %a, <i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3>
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <16 x i16> %shift
 ;
   %shift = lshr <16 x i16> %a, <i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3>
   ret <16 x i16> %shift
@@ -1634,6 +1699,10 @@ define <32 x i16> @splatconstant_shift_v32i16(<32 x i16> %a) {
 ; AVX512BWVL-LABEL: 'splatconstant_shift_v32i16'
 ; AVX512BWVL-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shift = lshr <32 x i16> %a, <i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3>
 ; AVX512BWVL-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <32 x i16> %shift
+;
+; AVX512GFNI-LABEL: 'splatconstant_shift_v32i16'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shift = lshr <32 x i16> %a, <i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3>
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <32 x i16> %shift
 ;
   %shift = lshr <32 x i16> %a, <i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3>
   ret <32 x i16> %shift
@@ -1700,6 +1769,10 @@ define <32 x i8> @splatconstant_shift_v32i8(<32 x i8> %a) {
 ; AVX512BWVL-LABEL: 'splatconstant_shift_v32i8'
 ; AVX512BWVL-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shift = lshr <32 x i8> %a, <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, 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>
 ; AVX512BWVL-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <32 x i8> %shift
+;
+; AVX512GFNI-LABEL: 'splatconstant_shift_v32i8'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shift = lshr <32 x i8> %a, <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, 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>
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <32 x i8> %shift
 ;
   %shift = lshr <32 x i8> %a, <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, 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>
   ret <32 x i8> %shift
@@ -1741,6 +1814,10 @@ define <64 x i8> @splatconstant_shift_v64i8(<64 x i8> %a) {
 ; AVX512BWVL-LABEL: 'splatconstant_shift_v64i8'
 ; AVX512BWVL-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shift = lshr <64 x i8> %a, <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, 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, 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, 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>
 ; AVX512BWVL-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <64 x i8> %shift
+;
+; AVX512GFNI-LABEL: 'splatconstant_shift_v64i8'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shift = lshr <64 x i8> %a, <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, 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, 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, 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>
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <64 x i8> %shift
 ;
   %shift = lshr <64 x i8> %a, <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, 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, 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, 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>
   ret <64 x i8> %shift

diff  --git a/llvm/test/Analysis/CostModel/X86/vshift-lshr-cost.ll b/llvm/test/Analysis/CostModel/X86/vshift-lshr-cost.ll
index 55894dc9329578..93b844eddf1824 100644
--- a/llvm/test/Analysis/CostModel/X86/vshift-lshr-cost.ll
+++ b/llvm/test/Analysis/CostModel/X86/vshift-lshr-cost.ll
@@ -15,6 +15,7 @@
 ; RUN: opt < %s -mtriple=x86_64-unknown-linux-gnu -passes="print<cost-model>" 2>&1 -disable-output -mcpu=slm | FileCheck %s --check-prefixes=CHECK,SSE,SSE42
 ; RUN: opt < %s -mtriple=x86_64-unknown-linux-gnu -passes="print<cost-model>" 2>&1 -disable-output -mcpu=goldmont | FileCheck %s --check-prefixes=CHECK,SSE,SSE42
 ; RUN: opt < %s -mtriple=x86_64-unknown-linux-gnu -passes="print<cost-model>" 2>&1 -disable-output -mcpu=btver2 | FileCheck %s --check-prefixes=CHECK,AVX,AVX1
+; RUN: opt < %s -mtriple=x86_64-unknown-linux-gnu -passes="print<cost-model>" 2>&1 -disable-output -mcpu=tigerlake | FileCheck %s --check-prefixes=CHECK,AVX512,AVX512GFNI
 
 ; Verify the cost of vector logical shift right instructions.
 
@@ -232,6 +233,10 @@ define <8 x i16> @var_shift_v8i16(<8 x i16> %a, <8 x i16> %b) {
 ; AVX512BWVL-LABEL: 'var_shift_v8i16'
 ; AVX512BWVL-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shift = lshr <8 x i16> %a, %b
 ; AVX512BWVL-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <8 x i16> %shift
+;
+; AVX512GFNI-LABEL: 'var_shift_v8i16'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shift = lshr <8 x i16> %a, %b
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <8 x i16> %shift
 ;
   %shift = lshr <8 x i16> %a, %b
   ret <8 x i16> %shift
@@ -273,6 +278,10 @@ define <16 x i16> @var_shift_v16i16(<16 x i16> %a, <16 x i16> %b) {
 ; AVX512BWVL-LABEL: 'var_shift_v16i16'
 ; AVX512BWVL-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shift = lshr <16 x i16> %a, %b
 ; AVX512BWVL-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <16 x i16> %shift
+;
+; AVX512GFNI-LABEL: 'var_shift_v16i16'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shift = lshr <16 x i16> %a, %b
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <16 x i16> %shift
 ;
   %shift = lshr <16 x i16> %a, %b
   ret <16 x i16> %shift
@@ -314,6 +323,10 @@ define <32 x i16> @var_shift_v32i16(<32 x i16> %a, <32 x i16> %b) {
 ; AVX512BWVL-LABEL: 'var_shift_v32i16'
 ; AVX512BWVL-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shift = lshr <32 x i16> %a, %b
 ; AVX512BWVL-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <32 x i16> %shift
+;
+; AVX512GFNI-LABEL: 'var_shift_v32i16'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shift = lshr <32 x i16> %a, %b
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <32 x i16> %shift
 ;
   %shift = lshr <32 x i16> %a, %b
   ret <32 x i16> %shift
@@ -355,6 +368,10 @@ define <16 x i8> @var_shift_v16i8(<16 x i8> %a, <16 x i8> %b) {
 ; AVX512BWVL-LABEL: 'var_shift_v16i8'
 ; AVX512BWVL-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %shift = lshr <16 x i8> %a, %b
 ; AVX512BWVL-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <16 x i8> %shift
+;
+; AVX512GFNI-LABEL: 'var_shift_v16i8'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %shift = lshr <16 x i8> %a, %b
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <16 x i8> %shift
 ;
   %shift = lshr <16 x i8> %a, %b
   ret <16 x i8> %shift
@@ -396,6 +413,10 @@ define <32 x i8> @var_shift_v32i8(<32 x i8> %a, <32 x i8> %b) {
 ; AVX512BWVL-LABEL: 'var_shift_v32i8'
 ; AVX512BWVL-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %shift = lshr <32 x i8> %a, %b
 ; AVX512BWVL-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <32 x i8> %shift
+;
+; AVX512GFNI-LABEL: 'var_shift_v32i8'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %shift = lshr <32 x i8> %a, %b
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <32 x i8> %shift
 ;
   %shift = lshr <32 x i8> %a, %b
   ret <32 x i8> %shift
@@ -437,6 +458,10 @@ define <64 x i8> @var_shift_v64i8(<64 x i8> %a, <64 x i8> %b) {
 ; AVX512BWVL-LABEL: 'var_shift_v64i8'
 ; AVX512BWVL-NEXT:  Cost Model: Found an estimated cost of 7 for instruction: %shift = lshr <64 x i8> %a, %b
 ; AVX512BWVL-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <64 x i8> %shift
+;
+; AVX512GFNI-LABEL: 'var_shift_v64i8'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 7 for instruction: %shift = lshr <64 x i8> %a, %b
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <64 x i8> %shift
 ;
   %shift = lshr <64 x i8> %a, %b
   ret <64 x i8> %shift
@@ -826,6 +851,12 @@ define <32 x i16> @splatvar_shift_v32i16(<32 x i16> %a, i16 %b) {
 ; AVX512BWVL-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %splat = shufflevector <32 x i16> %insert, <32 x i16> undef, <32 x i32> zeroinitializer
 ; AVX512BWVL-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %shift = lshr <32 x i16> %a, %splat
 ; AVX512BWVL-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <32 x i16> %shift
+;
+; AVX512GFNI-LABEL: 'splatvar_shift_v32i16'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %insert = insertelement <32 x i16> undef, i16 %b, i32 0
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %splat = shufflevector <32 x i16> %insert, <32 x i16> undef, <32 x i32> zeroinitializer
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %shift = lshr <32 x i16> %a, %splat
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <32 x i16> %shift
 ;
   %insert = insertelement <32 x i16> undef, i16 %b, i32 0
   %splat = shufflevector <32 x i16> %insert, <32 x i16> undef, <32 x i32> zeroinitializer
@@ -991,6 +1022,12 @@ define <64 x i8> @splatvar_shift_v64i8(<64 x i8> %a, i8 %b) {
 ; AVX512BWVL-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %splat = shufflevector <64 x i8> %insert, <64 x i8> undef, <64 x i32> zeroinitializer
 ; AVX512BWVL-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %shift = lshr <64 x i8> %a, %splat
 ; AVX512BWVL-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <64 x i8> %shift
+;
+; AVX512GFNI-LABEL: 'splatvar_shift_v64i8'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %insert = insertelement <64 x i8> undef, i8 %b, i32 0
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %splat = shufflevector <64 x i8> %insert, <64 x i8> undef, <64 x i32> zeroinitializer
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %shift = lshr <64 x i8> %a, %splat
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <64 x i8> %shift
 ;
   %insert = insertelement <64 x i8> undef, i8 %b, i32 0
   %splat = shufflevector <64 x i8> %insert, <64 x i8> undef, <64 x i32> zeroinitializer
@@ -1200,6 +1237,10 @@ define <8 x i16> @constant_shift_v8i16(<8 x i16> %a) {
 ; AVX512BWVL-LABEL: 'constant_shift_v8i16'
 ; AVX512BWVL-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shift = lshr <8 x i16> %a, <i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7>
 ; AVX512BWVL-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <8 x i16> %shift
+;
+; AVX512GFNI-LABEL: 'constant_shift_v8i16'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shift = lshr <8 x i16> %a, <i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7>
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <8 x i16> %shift
 ;
   %shift = lshr <8 x i16> %a, <i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7>
   ret <8 x i16> %shift
@@ -1241,6 +1282,10 @@ define <16 x i16> @constant_shift_v16i16(<16 x i16> %a) {
 ; AVX512BWVL-LABEL: 'constant_shift_v16i16'
 ; AVX512BWVL-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shift = lshr <16 x i16> %a, <i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7>
 ; AVX512BWVL-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <16 x i16> %shift
+;
+; AVX512GFNI-LABEL: 'constant_shift_v16i16'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shift = lshr <16 x i16> %a, <i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7>
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <16 x i16> %shift
 ;
   %shift = lshr <16 x i16> %a, <i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7>
   ret <16 x i16> %shift
@@ -1282,6 +1327,10 @@ define <32 x i16> @constant_shift_v32i16(<32 x i16> %a) {
 ; AVX512BWVL-LABEL: 'constant_shift_v32i16'
 ; AVX512BWVL-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shift = lshr <32 x i16> %a, <i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7>
 ; AVX512BWVL-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <32 x i16> %shift
+;
+; AVX512GFNI-LABEL: 'constant_shift_v32i16'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shift = lshr <32 x i16> %a, <i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7>
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <32 x i16> %shift
 ;
   %shift = lshr <32 x i16> %a, <i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7>
   ret <32 x i16> %shift
@@ -1323,6 +1372,10 @@ define <16 x i8> @constant_shift_v16i8(<16 x i8> %a) {
 ; AVX512BWVL-LABEL: 'constant_shift_v16i8'
 ; AVX512BWVL-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %shift = lshr <16 x i8> %a, <i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0>
 ; AVX512BWVL-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <16 x i8> %shift
+;
+; AVX512GFNI-LABEL: 'constant_shift_v16i8'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %shift = lshr <16 x i8> %a, <i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0>
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <16 x i8> %shift
 ;
   %shift = lshr <16 x i8> %a, <i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0>
   ret <16 x i8> %shift
@@ -1364,6 +1417,10 @@ define <32 x i8> @constant_shift_v32i8(<32 x i8> %a) {
 ; AVX512BWVL-LABEL: 'constant_shift_v32i8'
 ; AVX512BWVL-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %shift = lshr <32 x i8> %a, <i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0>
 ; AVX512BWVL-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <32 x i8> %shift
+;
+; AVX512GFNI-LABEL: 'constant_shift_v32i8'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %shift = lshr <32 x i8> %a, <i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0>
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <32 x i8> %shift
 ;
   %shift = lshr <32 x i8> %a, <i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0>
   ret <32 x i8> %shift
@@ -1405,6 +1462,10 @@ define <64 x i8> @constant_shift_v64i8(<64 x i8> %a) {
 ; AVX512BWVL-LABEL: 'constant_shift_v64i8'
 ; AVX512BWVL-NEXT:  Cost Model: Found an estimated cost of 7 for instruction: %shift = lshr <64 x i8> %a, <i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0>
 ; AVX512BWVL-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <64 x i8> %shift
+;
+; AVX512GFNI-LABEL: 'constant_shift_v64i8'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 7 for instruction: %shift = lshr <64 x i8> %a, <i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0>
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <64 x i8> %shift
 ;
   %shift = lshr <64 x i8> %a, <i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0>
   ret <64 x i8> %shift
@@ -1593,6 +1654,10 @@ define <16 x i16> @splatconstant_shift_v16i16(<16 x i16> %a) {
 ; AVX512BWVL-LABEL: 'splatconstant_shift_v16i16'
 ; AVX512BWVL-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shift = lshr <16 x i16> %a, <i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3>
 ; AVX512BWVL-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <16 x i16> %shift
+;
+; AVX512GFNI-LABEL: 'splatconstant_shift_v16i16'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shift = lshr <16 x i16> %a, <i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3>
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <16 x i16> %shift
 ;
   %shift = lshr <16 x i16> %a, <i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3>
   ret <16 x i16> %shift
@@ -1634,6 +1699,10 @@ define <32 x i16> @splatconstant_shift_v32i16(<32 x i16> %a) {
 ; AVX512BWVL-LABEL: 'splatconstant_shift_v32i16'
 ; AVX512BWVL-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shift = lshr <32 x i16> %a, <i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3>
 ; AVX512BWVL-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <32 x i16> %shift
+;
+; AVX512GFNI-LABEL: 'splatconstant_shift_v32i16'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shift = lshr <32 x i16> %a, <i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3>
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <32 x i16> %shift
 ;
   %shift = lshr <32 x i16> %a, <i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3>
   ret <32 x i16> %shift
@@ -1700,6 +1769,10 @@ define <32 x i8> @splatconstant_shift_v32i8(<32 x i8> %a) {
 ; AVX512BWVL-LABEL: 'splatconstant_shift_v32i8'
 ; AVX512BWVL-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shift = lshr <32 x i8> %a, <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, 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>
 ; AVX512BWVL-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <32 x i8> %shift
+;
+; AVX512GFNI-LABEL: 'splatconstant_shift_v32i8'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shift = lshr <32 x i8> %a, <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, 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>
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <32 x i8> %shift
 ;
   %shift = lshr <32 x i8> %a, <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, 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>
   ret <32 x i8> %shift
@@ -1741,6 +1814,10 @@ define <64 x i8> @splatconstant_shift_v64i8(<64 x i8> %a) {
 ; AVX512BWVL-LABEL: 'splatconstant_shift_v64i8'
 ; AVX512BWVL-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shift = lshr <64 x i8> %a, <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, 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, 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, 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>
 ; AVX512BWVL-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <64 x i8> %shift
+;
+; AVX512GFNI-LABEL: 'splatconstant_shift_v64i8'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shift = lshr <64 x i8> %a, <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, 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, 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, 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>
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <64 x i8> %shift
 ;
   %shift = lshr <64 x i8> %a, <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, 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, 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, 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>
   ret <64 x i8> %shift

diff  --git a/llvm/test/Analysis/CostModel/X86/vshift-lshr-latency.ll b/llvm/test/Analysis/CostModel/X86/vshift-lshr-latency.ll
index 92fe253d38e738..61620e2cc97ed8 100644
--- a/llvm/test/Analysis/CostModel/X86/vshift-lshr-latency.ll
+++ b/llvm/test/Analysis/CostModel/X86/vshift-lshr-latency.ll
@@ -15,6 +15,7 @@
 ; RUN: opt < %s -mtriple=x86_64-unknown-linux-gnu -passes="print<cost-model>" 2>&1 -disable-output -cost-kind=latency -mcpu=slm | FileCheck %s --check-prefixes=SSE,SSE42
 ; RUN: opt < %s -mtriple=x86_64-unknown-linux-gnu -passes="print<cost-model>" 2>&1 -disable-output -cost-kind=latency -mcpu=goldmont | FileCheck %s --check-prefixes=SSE,SSE42
 ; RUN: opt < %s -mtriple=x86_64-unknown-linux-gnu -passes="print<cost-model>" 2>&1 -disable-output -cost-kind=latency -mcpu=btver2 | FileCheck %s --check-prefixes=AVX1
+; RUN: opt < %s -mtriple=x86_64-unknown-linux-gnu -passes="print<cost-model>" 2>&1 -disable-output -cost-kind=latency -mcpu=tigerlake | FileCheck %s --check-prefixes=AVX512,AVX512GFNI
 
 ; Verify the cost of vector logical shift right instructions.
 
@@ -240,6 +241,10 @@ define <8 x i16> @var_shift_v8i16(<8 x i16> %a, <8 x i16> %b) {
 ; AVX512BW-LABEL: 'var_shift_v8i16'
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shift = lshr <8 x i16> %a, %b
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <8 x i16> %shift
+;
+; AVX512GFNI-LABEL: 'var_shift_v8i16'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shift = lshr <8 x i16> %a, %b
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <8 x i16> %shift
 ;
   %shift = lshr <8 x i16> %a, %b
   ret <8 x i16> %shift
@@ -277,6 +282,10 @@ define <16 x i16> @var_shift_v16i16(<16 x i16> %a, <16 x i16> %b) {
 ; AVX512BW-LABEL: 'var_shift_v16i16'
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shift = lshr <16 x i16> %a, %b
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <16 x i16> %shift
+;
+; AVX512GFNI-LABEL: 'var_shift_v16i16'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shift = lshr <16 x i16> %a, %b
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <16 x i16> %shift
 ;
   %shift = lshr <16 x i16> %a, %b
   ret <16 x i16> %shift
@@ -314,6 +323,10 @@ define <32 x i16> @var_shift_v32i16(<32 x i16> %a, <32 x i16> %b) {
 ; AVX512BW-LABEL: 'var_shift_v32i16'
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shift = lshr <32 x i16> %a, %b
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <32 x i16> %shift
+;
+; AVX512GFNI-LABEL: 'var_shift_v32i16'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shift = lshr <32 x i16> %a, %b
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <32 x i16> %shift
 ;
   %shift = lshr <32 x i16> %a, %b
   ret <32 x i16> %shift
@@ -351,6 +364,10 @@ define <16 x i8> @var_shift_v16i8(<16 x i8> %a, <16 x i8> %b) {
 ; AVX512BW-LABEL: 'var_shift_v16i8'
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %shift = lshr <16 x i8> %a, %b
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <16 x i8> %shift
+;
+; AVX512GFNI-LABEL: 'var_shift_v16i8'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %shift = lshr <16 x i8> %a, %b
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <16 x i8> %shift
 ;
   %shift = lshr <16 x i8> %a, %b
   ret <16 x i8> %shift
@@ -421,6 +438,10 @@ define <64 x i8> @var_shift_v64i8(<64 x i8> %a, <64 x i8> %b) {
 ; AVX512BW-LABEL: 'var_shift_v64i8'
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 27 for instruction: %shift = lshr <64 x i8> %a, %b
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <64 x i8> %shift
+;
+; AVX512GFNI-LABEL: 'var_shift_v64i8'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 27 for instruction: %shift = lshr <64 x i8> %a, %b
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <64 x i8> %shift
 ;
   %shift = lshr <64 x i8> %a, %b
   ret <64 x i8> %shift
@@ -834,6 +855,12 @@ define <32 x i16> @splatvar_shift_v32i16(<32 x i16> %a, i16 %b) {
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %splat = shufflevector <32 x i16> %insert, <32 x i16> undef, <32 x i32> zeroinitializer
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %shift = lshr <32 x i16> %a, %splat
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <32 x i16> %shift
+;
+; AVX512GFNI-LABEL: 'splatvar_shift_v32i16'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %insert = insertelement <32 x i16> undef, i16 %b, i32 0
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %splat = shufflevector <32 x i16> %insert, <32 x i16> undef, <32 x i32> zeroinitializer
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %shift = lshr <32 x i16> %a, %splat
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <32 x i16> %shift
 ;
   %insert = insertelement <32 x i16> undef, i16 %b, i32 0
   %splat = shufflevector <32 x i16> %insert, <32 x i16> undef, <32 x i32> zeroinitializer
@@ -889,6 +916,12 @@ define <16 x i8> @splatvar_shift_v16i8(<16 x i8> %a, i8 %b) {
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %splat = shufflevector <16 x i8> %insert, <16 x i8> undef, <16 x i32> zeroinitializer
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 10 for instruction: %shift = lshr <16 x i8> %a, %splat
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <16 x i8> %shift
+;
+; AVX512GFNI-LABEL: 'splatvar_shift_v16i8'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %insert = insertelement <16 x i8> undef, i8 %b, i32 0
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %splat = shufflevector <16 x i8> %insert, <16 x i8> undef, <16 x i32> zeroinitializer
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 10 for instruction: %shift = lshr <16 x i8> %a, %splat
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <16 x i8> %shift
 ;
   %insert = insertelement <16 x i8> undef, i8 %b, i32 0
   %splat = shufflevector <16 x i8> %insert, <16 x i8> undef, <16 x i32> zeroinitializer
@@ -993,6 +1026,12 @@ define <64 x i8> @splatvar_shift_v64i8(<64 x i8> %a, i8 %b) {
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %splat = shufflevector <64 x i8> %insert, <64 x i8> undef, <64 x i32> zeroinitializer
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %shift = lshr <64 x i8> %a, %splat
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <64 x i8> %shift
+;
+; AVX512GFNI-LABEL: 'splatvar_shift_v64i8'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %insert = insertelement <64 x i8> undef, i8 %b, i32 0
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %splat = shufflevector <64 x i8> %insert, <64 x i8> undef, <64 x i32> zeroinitializer
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %shift = lshr <64 x i8> %a, %splat
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <64 x i8> %shift
 ;
   %insert = insertelement <64 x i8> undef, i8 %b, i32 0
   %splat = shufflevector <64 x i8> %insert, <64 x i8> undef, <64 x i32> zeroinitializer
@@ -1222,6 +1261,10 @@ define <8 x i16> @constant_shift_v8i16(<8 x i16> %a) {
 ; AVX512BW-LABEL: 'constant_shift_v8i16'
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shift = lshr <8 x i16> %a, <i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7>
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <8 x i16> %shift
+;
+; AVX512GFNI-LABEL: 'constant_shift_v8i16'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shift = lshr <8 x i16> %a, <i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7>
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <8 x i16> %shift
 ;
   %shift = lshr <8 x i16> %a, <i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7>
   ret <8 x i16> %shift
@@ -1259,6 +1302,10 @@ define <16 x i16> @constant_shift_v16i16(<16 x i16> %a) {
 ; AVX512BW-LABEL: 'constant_shift_v16i16'
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shift = lshr <16 x i16> %a, <i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7>
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <16 x i16> %shift
+;
+; AVX512GFNI-LABEL: 'constant_shift_v16i16'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shift = lshr <16 x i16> %a, <i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7>
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <16 x i16> %shift
 ;
   %shift = lshr <16 x i16> %a, <i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7>
   ret <16 x i16> %shift
@@ -1296,6 +1343,10 @@ define <32 x i16> @constant_shift_v32i16(<32 x i16> %a) {
 ; AVX512BW-LABEL: 'constant_shift_v32i16'
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shift = lshr <32 x i16> %a, <i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7>
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <32 x i16> %shift
+;
+; AVX512GFNI-LABEL: 'constant_shift_v32i16'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shift = lshr <32 x i16> %a, <i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7>
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <32 x i16> %shift
 ;
   %shift = lshr <32 x i16> %a, <i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7>
   ret <32 x i16> %shift
@@ -1333,6 +1384,10 @@ define <16 x i8> @constant_shift_v16i8(<16 x i8> %a) {
 ; AVX512BW-LABEL: 'constant_shift_v16i8'
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %shift = lshr <16 x i8> %a, <i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0>
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <16 x i8> %shift
+;
+; AVX512GFNI-LABEL: 'constant_shift_v16i8'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %shift = lshr <16 x i8> %a, <i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0>
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <16 x i8> %shift
 ;
   %shift = lshr <16 x i8> %a, <i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0>
   ret <16 x i8> %shift
@@ -1403,6 +1458,10 @@ define <64 x i8> @constant_shift_v64i8(<64 x i8> %a) {
 ; AVX512BW-LABEL: 'constant_shift_v64i8'
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 27 for instruction: %shift = lshr <64 x i8> %a, <i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0>
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <64 x i8> %shift
+;
+; AVX512GFNI-LABEL: 'constant_shift_v64i8'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 27 for instruction: %shift = lshr <64 x i8> %a, <i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0>
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <64 x i8> %shift
 ;
   %shift = lshr <64 x i8> %a, <i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0>
   ret <64 x i8> %shift
@@ -1643,6 +1702,10 @@ define <16 x i16> @splatconstant_shift_v16i16(<16 x i16> %a) {
 ; AVX512BW-LABEL: 'splatconstant_shift_v16i16'
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shift = lshr <16 x i16> %a, <i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3>
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <16 x i16> %shift
+;
+; AVX512GFNI-LABEL: 'splatconstant_shift_v16i16'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shift = lshr <16 x i16> %a, <i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3>
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <16 x i16> %shift
 ;
   %shift = lshr <16 x i16> %a, <i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3>
   ret <16 x i16> %shift
@@ -1676,6 +1739,10 @@ define <32 x i16> @splatconstant_shift_v32i16(<32 x i16> %a) {
 ; AVX512BW-LABEL: 'splatconstant_shift_v32i16'
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shift = lshr <32 x i16> %a, <i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3>
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <32 x i16> %shift
+;
+; AVX512GFNI-LABEL: 'splatconstant_shift_v32i16'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shift = lshr <32 x i16> %a, <i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3>
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <32 x i16> %shift
 ;
   %shift = lshr <32 x i16> %a, <i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3>
   ret <32 x i16> %shift
@@ -1709,6 +1776,10 @@ define <16 x i8> @splatconstant_shift_v16i8(<16 x i8> %a) {
 ; AVX512BW-LABEL: 'splatconstant_shift_v16i8'
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 7 for instruction: %shift = lshr <16 x i8> %a, <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>
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <16 x i8> %shift
+;
+; AVX512GFNI-LABEL: 'splatconstant_shift_v16i8'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 7 for instruction: %shift = lshr <16 x i8> %a, <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>
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <16 x i8> %shift
 ;
   %shift = lshr <16 x i8> %a, <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>
   ret <16 x i8> %shift
@@ -1771,6 +1842,10 @@ define <64 x i8> @splatconstant_shift_v64i8(<64 x i8> %a) {
 ; AVX512BW-LABEL: 'splatconstant_shift_v64i8'
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %shift = lshr <64 x i8> %a, <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, 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, 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, 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>
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <64 x i8> %shift
+;
+; AVX512GFNI-LABEL: 'splatconstant_shift_v64i8'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %shift = lshr <64 x i8> %a, <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, 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, 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, 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>
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <64 x i8> %shift
 ;
   %shift = lshr <64 x i8> %a, <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, 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, 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, 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>
   ret <64 x i8> %shift

diff  --git a/llvm/test/Analysis/CostModel/X86/vshift-lshr-sizelatency.ll b/llvm/test/Analysis/CostModel/X86/vshift-lshr-sizelatency.ll
index fe8a7d91cc3752..e6b6ac75b65d58 100644
--- a/llvm/test/Analysis/CostModel/X86/vshift-lshr-sizelatency.ll
+++ b/llvm/test/Analysis/CostModel/X86/vshift-lshr-sizelatency.ll
@@ -15,6 +15,7 @@
 ; RUN: opt < %s -mtriple=x86_64-unknown-linux-gnu -passes="print<cost-model>" 2>&1 -disable-output -cost-kind=size-latency -mcpu=slm | FileCheck %s --check-prefixes=CHECK,SSE42
 ; RUN: opt < %s -mtriple=x86_64-unknown-linux-gnu -passes="print<cost-model>" 2>&1 -disable-output -cost-kind=size-latency -mcpu=goldmont | FileCheck %s --check-prefixes=CHECK,SSE42
 ; RUN: opt < %s -mtriple=x86_64-unknown-linux-gnu -passes="print<cost-model>" 2>&1 -disable-output -cost-kind=size-latency -mcpu=btver2 | FileCheck %s --check-prefixes=CHECK,AVX,AVX1
+; RUN: opt < %s -mtriple=x86_64-unknown-linux-gnu -passes="print<cost-model>" 2>&1 -disable-output -cost-kind=size-latency -mcpu=tigerlake | FileCheck %s --check-prefixes=CHECK,AVX512,AVX512GFNI
 
 ; Verify the cost of vector logical shift right instructions.
 
@@ -240,6 +241,10 @@ define <8 x i16> @var_shift_v8i16(<8 x i16> %a, <8 x i16> %b) {
 ; AVX512BW-LABEL: 'var_shift_v8i16'
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shift = lshr <8 x i16> %a, %b
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <8 x i16> %shift
+;
+; AVX512GFNI-LABEL: 'var_shift_v8i16'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shift = lshr <8 x i16> %a, %b
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <8 x i16> %shift
 ;
   %shift = lshr <8 x i16> %a, %b
   ret <8 x i16> %shift
@@ -273,6 +278,10 @@ define <16 x i16> @var_shift_v16i16(<16 x i16> %a, <16 x i16> %b) {
 ; AVX512BW-LABEL: 'var_shift_v16i16'
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shift = lshr <16 x i16> %a, %b
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <16 x i16> %shift
+;
+; AVX512GFNI-LABEL: 'var_shift_v16i16'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shift = lshr <16 x i16> %a, %b
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <16 x i16> %shift
 ;
   %shift = lshr <16 x i16> %a, %b
   ret <16 x i16> %shift
@@ -306,6 +315,10 @@ define <32 x i16> @var_shift_v32i16(<32 x i16> %a, <32 x i16> %b) {
 ; AVX512BW-LABEL: 'var_shift_v32i16'
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shift = lshr <32 x i16> %a, %b
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <32 x i16> %shift
+;
+; AVX512GFNI-LABEL: 'var_shift_v32i16'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shift = lshr <32 x i16> %a, %b
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <32 x i16> %shift
 ;
   %shift = lshr <32 x i16> %a, %b
   ret <32 x i16> %shift
@@ -335,6 +348,10 @@ define <16 x i8> @var_shift_v16i8(<16 x i8> %a, <16 x i8> %b) {
 ; AVX512BW-LABEL: 'var_shift_v16i8'
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 5 for instruction: %shift = lshr <16 x i8> %a, %b
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <16 x i8> %shift
+;
+; AVX512GFNI-LABEL: 'var_shift_v16i8'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 5 for instruction: %shift = lshr <16 x i8> %a, %b
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <16 x i8> %shift
 ;
   %shift = lshr <16 x i8> %a, %b
   ret <16 x i8> %shift
@@ -368,6 +385,10 @@ define <32 x i8> @var_shift_v32i8(<32 x i8> %a, <32 x i8> %b) {
 ; AVX512BW-LABEL: 'var_shift_v32i8'
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 18 for instruction: %shift = lshr <32 x i8> %a, %b
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <32 x i8> %shift
+;
+; AVX512GFNI-LABEL: 'var_shift_v32i8'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 18 for instruction: %shift = lshr <32 x i8> %a, %b
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <32 x i8> %shift
 ;
   %shift = lshr <32 x i8> %a, %b
   ret <32 x i8> %shift
@@ -401,6 +422,10 @@ define <64 x i8> @var_shift_v64i8(<64 x i8> %a, <64 x i8> %b) {
 ; AVX512BW-LABEL: 'var_shift_v64i8'
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 18 for instruction: %shift = lshr <64 x i8> %a, %b
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <64 x i8> %shift
+;
+; AVX512GFNI-LABEL: 'var_shift_v64i8'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 18 for instruction: %shift = lshr <64 x i8> %a, %b
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <64 x i8> %shift
 ;
   %shift = lshr <64 x i8> %a, %b
   ret <64 x i8> %shift
@@ -766,6 +791,12 @@ define <32 x i16> @splatvar_shift_v32i16(<32 x i16> %a, i16 %b) {
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %splat = shufflevector <32 x i16> %insert, <32 x i16> undef, <32 x i32> zeroinitializer
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %shift = lshr <32 x i16> %a, %splat
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <32 x i16> %shift
+;
+; AVX512GFNI-LABEL: 'splatvar_shift_v32i16'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %insert = insertelement <32 x i16> undef, i16 %b, i32 0
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %splat = shufflevector <32 x i16> %insert, <32 x i16> undef, <32 x i32> zeroinitializer
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %shift = lshr <32 x i16> %a, %splat
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <32 x i16> %shift
 ;
   %insert = insertelement <32 x i16> undef, i16 %b, i32 0
   %splat = shufflevector <32 x i16> %insert, <32 x i16> undef, <32 x i32> zeroinitializer
@@ -913,6 +944,12 @@ define <64 x i8> @splatvar_shift_v64i8(<64 x i8> %a, i8 %b) {
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %splat = shufflevector <64 x i8> %insert, <64 x i8> undef, <64 x i32> zeroinitializer
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 10 for instruction: %shift = lshr <64 x i8> %a, %splat
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <64 x i8> %shift
+;
+; AVX512GFNI-LABEL: 'splatvar_shift_v64i8'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %insert = insertelement <64 x i8> undef, i8 %b, i32 0
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %splat = shufflevector <64 x i8> %insert, <64 x i8> undef, <64 x i32> zeroinitializer
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 10 for instruction: %shift = lshr <64 x i8> %a, %splat
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <64 x i8> %shift
 ;
   %insert = insertelement <64 x i8> undef, i8 %b, i32 0
   %splat = shufflevector <64 x i8> %insert, <64 x i8> undef, <64 x i32> zeroinitializer
@@ -1142,6 +1179,10 @@ define <8 x i16> @constant_shift_v8i16(<8 x i16> %a) {
 ; AVX512BW-LABEL: 'constant_shift_v8i16'
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shift = lshr <8 x i16> %a, <i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7>
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <8 x i16> %shift
+;
+; AVX512GFNI-LABEL: 'constant_shift_v8i16'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shift = lshr <8 x i16> %a, <i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7>
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <8 x i16> %shift
 ;
   %shift = lshr <8 x i16> %a, <i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7>
   ret <8 x i16> %shift
@@ -1175,6 +1216,10 @@ define <16 x i16> @constant_shift_v16i16(<16 x i16> %a) {
 ; AVX512BW-LABEL: 'constant_shift_v16i16'
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shift = lshr <16 x i16> %a, <i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7>
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <16 x i16> %shift
+;
+; AVX512GFNI-LABEL: 'constant_shift_v16i16'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shift = lshr <16 x i16> %a, <i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7>
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <16 x i16> %shift
 ;
   %shift = lshr <16 x i16> %a, <i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7>
   ret <16 x i16> %shift
@@ -1208,6 +1253,10 @@ define <32 x i16> @constant_shift_v32i16(<32 x i16> %a) {
 ; AVX512BW-LABEL: 'constant_shift_v32i16'
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shift = lshr <32 x i16> %a, <i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7>
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <32 x i16> %shift
+;
+; AVX512GFNI-LABEL: 'constant_shift_v32i16'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shift = lshr <32 x i16> %a, <i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7>
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <32 x i16> %shift
 ;
   %shift = lshr <32 x i16> %a, <i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7>
   ret <32 x i16> %shift
@@ -1237,6 +1286,10 @@ define <16 x i8> @constant_shift_v16i8(<16 x i8> %a) {
 ; AVX512BW-LABEL: 'constant_shift_v16i8'
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 5 for instruction: %shift = lshr <16 x i8> %a, <i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0>
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <16 x i8> %shift
+;
+; AVX512GFNI-LABEL: 'constant_shift_v16i8'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 5 for instruction: %shift = lshr <16 x i8> %a, <i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0>
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <16 x i8> %shift
 ;
   %shift = lshr <16 x i8> %a, <i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0>
   ret <16 x i8> %shift
@@ -1270,6 +1323,10 @@ define <32 x i8> @constant_shift_v32i8(<32 x i8> %a) {
 ; AVX512BW-LABEL: 'constant_shift_v32i8'
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 18 for instruction: %shift = lshr <32 x i8> %a, <i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0>
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <32 x i8> %shift
+;
+; AVX512GFNI-LABEL: 'constant_shift_v32i8'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 18 for instruction: %shift = lshr <32 x i8> %a, <i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0>
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <32 x i8> %shift
 ;
   %shift = lshr <32 x i8> %a, <i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0>
   ret <32 x i8> %shift
@@ -1303,6 +1360,10 @@ define <64 x i8> @constant_shift_v64i8(<64 x i8> %a) {
 ; AVX512BW-LABEL: 'constant_shift_v64i8'
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 18 for instruction: %shift = lshr <64 x i8> %a, <i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0>
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <64 x i8> %shift
+;
+; AVX512GFNI-LABEL: 'constant_shift_v64i8'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 18 for instruction: %shift = lshr <64 x i8> %a, <i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0>
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <64 x i8> %shift
 ;
   %shift = lshr <64 x i8> %a, <i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0>
   ret <64 x i8> %shift
@@ -1503,6 +1564,10 @@ define <16 x i16> @splatconstant_shift_v16i16(<16 x i16> %a) {
 ; AVX512BW-LABEL: 'splatconstant_shift_v16i16'
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shift = lshr <16 x i16> %a, <i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3>
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <16 x i16> %shift
+;
+; AVX512GFNI-LABEL: 'splatconstant_shift_v16i16'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shift = lshr <16 x i16> %a, <i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3>
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <16 x i16> %shift
 ;
   %shift = lshr <16 x i16> %a, <i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3>
   ret <16 x i16> %shift
@@ -1540,6 +1605,10 @@ define <32 x i16> @splatconstant_shift_v32i16(<32 x i16> %a) {
 ; AVX512BW-LABEL: 'splatconstant_shift_v32i16'
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shift = lshr <32 x i16> %a, <i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3>
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <32 x i16> %shift
+;
+; AVX512GFNI-LABEL: 'splatconstant_shift_v32i16'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shift = lshr <32 x i16> %a, <i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3>
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <32 x i16> %shift
 ;
   %shift = lshr <32 x i16> %a, <i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3>
   ret <32 x i16> %shift
@@ -1606,6 +1675,10 @@ define <32 x i8> @splatconstant_shift_v32i8(<32 x i8> %a) {
 ; AVX512BW-LABEL: 'splatconstant_shift_v32i8'
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %shift = lshr <32 x i8> %a, <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, 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>
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <32 x i8> %shift
+;
+; AVX512GFNI-LABEL: 'splatconstant_shift_v32i8'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %shift = lshr <32 x i8> %a, <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, 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>
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <32 x i8> %shift
 ;
   %shift = lshr <32 x i8> %a, <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, 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>
   ret <32 x i8> %shift
@@ -1643,6 +1716,10 @@ define <64 x i8> @splatconstant_shift_v64i8(<64 x i8> %a) {
 ; AVX512BW-LABEL: 'splatconstant_shift_v64i8'
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %shift = lshr <64 x i8> %a, <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, 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, 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, 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>
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <64 x i8> %shift
+;
+; AVX512GFNI-LABEL: 'splatconstant_shift_v64i8'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %shift = lshr <64 x i8> %a, <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, 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, 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, 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>
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <64 x i8> %shift
 ;
   %shift = lshr <64 x i8> %a, <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, 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, 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, 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>
   ret <64 x i8> %shift

diff  --git a/llvm/test/Analysis/CostModel/X86/vshift-shl-codesize.ll b/llvm/test/Analysis/CostModel/X86/vshift-shl-codesize.ll
index 07229e22873c43..265658b1e3a2da 100644
--- a/llvm/test/Analysis/CostModel/X86/vshift-shl-codesize.ll
+++ b/llvm/test/Analysis/CostModel/X86/vshift-shl-codesize.ll
@@ -15,6 +15,7 @@
 ; RUN: opt < %s -mtriple=x86_64-unknown-linux-gnu -passes="print<cost-model>" 2>&1 -disable-output -cost-kind=code-size -mcpu=slm | FileCheck %s --check-prefixes=CHECK,SSE42
 ; RUN: opt < %s -mtriple=x86_64-unknown-linux-gnu -passes="print<cost-model>" 2>&1 -disable-output -cost-kind=code-size -mcpu=goldmont | FileCheck %s --check-prefixes=CHECK,SSE42
 ; RUN: opt < %s -mtriple=x86_64-unknown-linux-gnu -passes="print<cost-model>" 2>&1 -disable-output -cost-kind=code-size -mcpu=btver2 | FileCheck %s --check-prefixes=CHECK,AVX,AVX1
+; RUN: opt < %s -mtriple=x86_64-unknown-linux-gnu -passes="print<cost-model>" 2>&1 -disable-output -cost-kind=code-size -mcpu=tigerlake | FileCheck %s --check-prefixes=CHECK,AVX512,AVX512GFNI
 
 ; Verify the cost of vector logical shift right instructions.
 
@@ -240,6 +241,10 @@ define <8 x i16> @var_shift_v8i16(<8 x i16> %a, <8 x i16> %b) {
 ; AVX512BW-LABEL: 'var_shift_v8i16'
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shift = shl <8 x i16> %a, %b
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <8 x i16> %shift
+;
+; AVX512GFNI-LABEL: 'var_shift_v8i16'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shift = shl <8 x i16> %a, %b
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <8 x i16> %shift
 ;
   %shift = shl <8 x i16> %a, %b
   ret <8 x i16> %shift
@@ -273,6 +278,10 @@ define <16 x i16> @var_shift_v16i16(<16 x i16> %a, <16 x i16> %b) {
 ; AVX512BW-LABEL: 'var_shift_v16i16'
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shift = shl <16 x i16> %a, %b
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <16 x i16> %shift
+;
+; AVX512GFNI-LABEL: 'var_shift_v16i16'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shift = shl <16 x i16> %a, %b
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <16 x i16> %shift
 ;
   %shift = shl <16 x i16> %a, %b
   ret <16 x i16> %shift
@@ -306,6 +315,10 @@ define <32 x i16> @var_shift_v32i16(<32 x i16> %a, <32 x i16> %b) {
 ; AVX512BW-LABEL: 'var_shift_v32i16'
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shift = shl <32 x i16> %a, %b
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <32 x i16> %shift
+;
+; AVX512GFNI-LABEL: 'var_shift_v32i16'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shift = shl <32 x i16> %a, %b
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <32 x i16> %shift
 ;
   %shift = shl <32 x i16> %a, %b
   ret <32 x i16> %shift
@@ -335,6 +348,10 @@ define <16 x i8> @var_shift_v16i8(<16 x i8> %a, <16 x i8> %b) {
 ; AVX512BW-LABEL: 'var_shift_v16i8'
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %shift = shl <16 x i8> %a, %b
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <16 x i8> %shift
+;
+; AVX512GFNI-LABEL: 'var_shift_v16i8'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %shift = shl <16 x i8> %a, %b
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <16 x i8> %shift
 ;
   %shift = shl <16 x i8> %a, %b
   ret <16 x i8> %shift
@@ -397,6 +414,10 @@ define <64 x i8> @var_shift_v64i8(<64 x i8> %a, <64 x i8> %b) {
 ; AVX512BW-LABEL: 'var_shift_v64i8'
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 13 for instruction: %shift = shl <64 x i8> %a, %b
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <64 x i8> %shift
+;
+; AVX512GFNI-LABEL: 'var_shift_v64i8'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 13 for instruction: %shift = shl <64 x i8> %a, %b
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <64 x i8> %shift
 ;
   %shift = shl <64 x i8> %a, %b
   ret <64 x i8> %shift
@@ -762,6 +783,12 @@ define <32 x i16> @splatvar_shift_v32i16(<32 x i16> %a, i16 %b) {
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %splat = shufflevector <32 x i16> %insert, <32 x i16> undef, <32 x i32> zeroinitializer
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %shift = shl <32 x i16> %a, %splat
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <32 x i16> %shift
+;
+; AVX512GFNI-LABEL: 'splatvar_shift_v32i16'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %insert = insertelement <32 x i16> undef, i16 %b, i32 0
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %splat = shufflevector <32 x i16> %insert, <32 x i16> undef, <32 x i32> zeroinitializer
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %shift = shl <32 x i16> %a, %splat
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <32 x i16> %shift
 ;
   %insert = insertelement <32 x i16> undef, i16 %b, i32 0
   %splat = shufflevector <32 x i16> %insert, <32 x i16> undef, <32 x i32> zeroinitializer
@@ -915,6 +942,12 @@ define <64 x i8> @splatvar_shift_v64i8(<64 x i8> %a, i8 %b) {
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %splat = shufflevector <64 x i8> %insert, <64 x i8> undef, <64 x i32> zeroinitializer
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 6 for instruction: %shift = shl <64 x i8> %a, %splat
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <64 x i8> %shift
+;
+; AVX512GFNI-LABEL: 'splatvar_shift_v64i8'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %insert = insertelement <64 x i8> undef, i8 %b, i32 0
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %splat = shufflevector <64 x i8> %insert, <64 x i8> undef, <64 x i32> zeroinitializer
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 6 for instruction: %shift = shl <64 x i8> %a, %splat
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <64 x i8> %shift
 ;
   %insert = insertelement <64 x i8> undef, i8 %b, i32 0
   %splat = shufflevector <64 x i8> %insert, <64 x i8> undef, <64 x i32> zeroinitializer
@@ -1186,6 +1219,10 @@ define <32 x i16> @constant_shift_v32i16(<32 x i16> %a) {
 ; AVX512BW-LABEL: 'constant_shift_v32i16'
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shift = shl <32 x i16> %a, <i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7>
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <32 x i16> %shift
+;
+; AVX512GFNI-LABEL: 'constant_shift_v32i16'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shift = shl <32 x i16> %a, <i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7>
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <32 x i16> %shift
 ;
   %shift = shl <32 x i16> %a, <i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7>
   ret <32 x i16> %shift
@@ -1215,6 +1252,10 @@ define <16 x i8> @constant_shift_v16i8(<16 x i8> %a) {
 ; AVX512BW-LABEL: 'constant_shift_v16i8'
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %shift = shl <16 x i8> %a, <i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0>
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <16 x i8> %shift
+;
+; AVX512GFNI-LABEL: 'constant_shift_v16i8'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %shift = shl <16 x i8> %a, <i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0>
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <16 x i8> %shift
 ;
   %shift = shl <16 x i8> %a, <i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0>
   ret <16 x i8> %shift
@@ -1277,6 +1318,10 @@ define <64 x i8> @constant_shift_v64i8(<64 x i8> %a) {
 ; AVX512BW-LABEL: 'constant_shift_v64i8'
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 13 for instruction: %shift = shl <64 x i8> %a, <i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0>
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <64 x i8> %shift
+;
+; AVX512GFNI-LABEL: 'constant_shift_v64i8'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 13 for instruction: %shift = shl <64 x i8> %a, <i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0>
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <64 x i8> %shift
 ;
   %shift = shl <64 x i8> %a, <i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0>
   ret <64 x i8> %shift
@@ -1477,6 +1522,10 @@ define <16 x i16> @splatconstant_shift_v16i16(<16 x i16> %a) {
 ; AVX512BW-LABEL: 'splatconstant_shift_v16i16'
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shift = shl <16 x i16> %a, <i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3>
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <16 x i16> %shift
+;
+; AVX512GFNI-LABEL: 'splatconstant_shift_v16i16'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shift = shl <16 x i16> %a, <i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3>
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <16 x i16> %shift
 ;
   %shift = shl <16 x i16> %a, <i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3>
   ret <16 x i16> %shift
@@ -1514,6 +1563,10 @@ define <32 x i16> @splatconstant_shift_v32i16(<32 x i16> %a) {
 ; AVX512BW-LABEL: 'splatconstant_shift_v32i16'
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shift = shl <32 x i16> %a, <i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3>
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <32 x i16> %shift
+;
+; AVX512GFNI-LABEL: 'splatconstant_shift_v32i16'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shift = shl <32 x i16> %a, <i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3>
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <32 x i16> %shift
 ;
   %shift = shl <32 x i16> %a, <i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3>
   ret <32 x i16> %shift
@@ -1613,6 +1666,10 @@ define <64 x i8> @splatconstant_shift_v64i8(<64 x i8> %a) {
 ; AVX512BW-LABEL: 'splatconstant_shift_v64i8'
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %shift = shl <64 x i8> %a, <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, 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, 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, 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>
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <64 x i8> %shift
+;
+; AVX512GFNI-LABEL: 'splatconstant_shift_v64i8'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %shift = shl <64 x i8> %a, <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, 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, 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, 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>
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <64 x i8> %shift
 ;
   %shift = shl <64 x i8> %a, <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, 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, 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, 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>
   ret <64 x i8> %shift

diff  --git a/llvm/test/Analysis/CostModel/X86/vshift-shl-cost-inseltpoison.ll b/llvm/test/Analysis/CostModel/X86/vshift-shl-cost-inseltpoison.ll
index d8737d2365649e..f093d1c8ca358a 100644
--- a/llvm/test/Analysis/CostModel/X86/vshift-shl-cost-inseltpoison.ll
+++ b/llvm/test/Analysis/CostModel/X86/vshift-shl-cost-inseltpoison.ll
@@ -15,6 +15,7 @@
 ; RUN: opt < %s -mtriple=x86_64-unknown-linux-gnu -passes="print<cost-model>" 2>&1 -disable-output -mcpu=slm | FileCheck %s --check-prefixes=CHECK,SSE,SLM
 ; RUN: opt < %s -mtriple=x86_64-unknown-linux-gnu -passes="print<cost-model>" 2>&1 -disable-output -mcpu=goldmont | FileCheck %s --check-prefixes=CHECK,SSE,SSE42
 ; RUN: opt < %s -mtriple=x86_64-unknown-linux-gnu -passes="print<cost-model>" 2>&1 -disable-output -mcpu=btver2 | FileCheck %s --check-prefixes=CHECK,AVX,AVX1
+; RUN: opt < %s -mtriple=x86_64-unknown-linux-gnu -passes="print<cost-model>" 2>&1 -disable-output -mcpu=tigerlake | FileCheck %s --check-prefixes=CHECK,AVX512,AVX512GFNI
 
 ; Verify the cost of vector shift left instructions.
 
@@ -236,6 +237,10 @@ define <8 x i16> @var_shift_v8i16(<8 x i16> %a, <8 x i16> %b) {
 ; SLM-LABEL: 'var_shift_v8i16'
 ; SLM-NEXT:  Cost Model: Found an estimated cost of 11 for instruction: %shift = shl <8 x i16> %a, %b
 ; SLM-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <8 x i16> %shift
+;
+; AVX512GFNI-LABEL: 'var_shift_v8i16'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shift = shl <8 x i16> %a, %b
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <8 x i16> %shift
 ;
   %shift = shl <8 x i16> %a, %b
   ret <8 x i16> %shift
@@ -281,6 +286,10 @@ define <16 x i16> @var_shift_v16i16(<16 x i16> %a, <16 x i16> %b) {
 ; SLM-LABEL: 'var_shift_v16i16'
 ; SLM-NEXT:  Cost Model: Found an estimated cost of 22 for instruction: %shift = shl <16 x i16> %a, %b
 ; SLM-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <16 x i16> %shift
+;
+; AVX512GFNI-LABEL: 'var_shift_v16i16'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shift = shl <16 x i16> %a, %b
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <16 x i16> %shift
 ;
   %shift = shl <16 x i16> %a, %b
   ret <16 x i16> %shift
@@ -326,6 +335,10 @@ define <32 x i16> @var_shift_v32i16(<32 x i16> %a, <32 x i16> %b) {
 ; SLM-LABEL: 'var_shift_v32i16'
 ; SLM-NEXT:  Cost Model: Found an estimated cost of 44 for instruction: %shift = shl <32 x i16> %a, %b
 ; SLM-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <32 x i16> %shift
+;
+; AVX512GFNI-LABEL: 'var_shift_v32i16'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shift = shl <32 x i16> %a, %b
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <32 x i16> %shift
 ;
   %shift = shl <32 x i16> %a, %b
   ret <32 x i16> %shift
@@ -371,6 +384,10 @@ define <16 x i8> @var_shift_v16i8(<16 x i8> %a, <16 x i8> %b) {
 ; SLM-LABEL: 'var_shift_v16i8'
 ; SLM-NEXT:  Cost Model: Found an estimated cost of 15 for instruction: %shift = shl <16 x i8> %a, %b
 ; SLM-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <16 x i8> %shift
+;
+; AVX512GFNI-LABEL: 'var_shift_v16i8'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %shift = shl <16 x i8> %a, %b
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <16 x i8> %shift
 ;
   %shift = shl <16 x i8> %a, %b
   ret <16 x i8> %shift
@@ -416,6 +433,10 @@ define <32 x i8> @var_shift_v32i8(<32 x i8> %a, <32 x i8> %b) {
 ; SLM-LABEL: 'var_shift_v32i8'
 ; SLM-NEXT:  Cost Model: Found an estimated cost of 30 for instruction: %shift = shl <32 x i8> %a, %b
 ; SLM-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <32 x i8> %shift
+;
+; AVX512GFNI-LABEL: 'var_shift_v32i8'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %shift = shl <32 x i8> %a, %b
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <32 x i8> %shift
 ;
   %shift = shl <32 x i8> %a, %b
   ret <32 x i8> %shift
@@ -461,6 +482,10 @@ define <64 x i8> @var_shift_v64i8(<64 x i8> %a, <64 x i8> %b) {
 ; SLM-LABEL: 'var_shift_v64i8'
 ; SLM-NEXT:  Cost Model: Found an estimated cost of 60 for instruction: %shift = shl <64 x i8> %a, %b
 ; SLM-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <64 x i8> %shift
+;
+; AVX512GFNI-LABEL: 'var_shift_v64i8'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 6 for instruction: %shift = shl <64 x i8> %a, %b
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <64 x i8> %shift
 ;
   %shift = shl <64 x i8> %a, %b
   ret <64 x i8> %shift
@@ -868,6 +893,12 @@ define <32 x i16> @splatvar_shift_v32i16(<32 x i16> %a, i16 %b) {
 ; SLM-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %splat = shufflevector <32 x i16> %insert, <32 x i16> poison, <32 x i32> zeroinitializer
 ; SLM-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %shift = shl <32 x i16> %a, %splat
 ; SLM-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <32 x i16> %shift
+;
+; AVX512GFNI-LABEL: 'splatvar_shift_v32i16'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %insert = insertelement <32 x i16> poison, i16 %b, i32 0
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %splat = shufflevector <32 x i16> %insert, <32 x i16> poison, <32 x i32> zeroinitializer
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %shift = shl <32 x i16> %a, %splat
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <32 x i16> %shift
 ;
   %insert = insertelement <32 x i16> poison, i16 %b, i32 0
   %splat = shufflevector <32 x i16> %insert, <32 x i16> poison, <32 x i32> zeroinitializer
@@ -1051,6 +1082,12 @@ define <64 x i8> @splatvar_shift_v64i8(<64 x i8> %a, i8 %b) {
 ; SLM-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %splat = shufflevector <64 x i8> %insert, <64 x i8> poison, <64 x i32> zeroinitializer
 ; SLM-NEXT:  Cost Model: Found an estimated cost of 36 for instruction: %shift = shl <64 x i8> %a, %splat
 ; SLM-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <64 x i8> %shift
+;
+; AVX512GFNI-LABEL: 'splatvar_shift_v64i8'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %insert = insertelement <64 x i8> poison, i8 %b, i32 0
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %splat = shufflevector <64 x i8> %insert, <64 x i8> poison, <64 x i32> zeroinitializer
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %shift = shl <64 x i8> %a, %splat
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <64 x i8> %shift
 ;
   %insert = insertelement <64 x i8> poison, i8 %b, i32 0
   %splat = shufflevector <64 x i8> %insert, <64 x i8> poison, <64 x i32> zeroinitializer
@@ -1305,6 +1342,10 @@ define <16 x i16> @constant_shift_v16i16(<16 x i16> %a) {
 ; SLM-LABEL: 'constant_shift_v16i16'
 ; SLM-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %shift = shl <16 x i16> %a, <i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7>
 ; SLM-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <16 x i16> %shift
+;
+; AVX512GFNI-LABEL: 'constant_shift_v16i16'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shift = shl <16 x i16> %a, <i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7>
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <16 x i16> %shift
 ;
   %shift = shl <16 x i16> %a, <i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7>
   ret <16 x i16> %shift
@@ -1354,6 +1395,10 @@ define <32 x i16> @constant_shift_v32i16(<32 x i16> %a) {
 ; SLM-LABEL: 'constant_shift_v32i16'
 ; SLM-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %shift = shl <32 x i16> %a, <i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7>
 ; SLM-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <32 x i16> %shift
+;
+; AVX512GFNI-LABEL: 'constant_shift_v32i16'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shift = shl <32 x i16> %a, <i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7>
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <32 x i16> %shift
 ;
   %shift = shl <32 x i16> %a, <i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7>
   ret <32 x i16> %shift
@@ -1399,6 +1444,10 @@ define <16 x i8> @constant_shift_v16i8(<16 x i8> %a) {
 ; SLM-LABEL: 'constant_shift_v16i8'
 ; SLM-NEXT:  Cost Model: Found an estimated cost of 15 for instruction: %shift = shl <16 x i8> %a, <i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0>
 ; SLM-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <16 x i8> %shift
+;
+; AVX512GFNI-LABEL: 'constant_shift_v16i8'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %shift = shl <16 x i8> %a, <i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0>
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <16 x i8> %shift
 ;
   %shift = shl <16 x i8> %a, <i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0>
   ret <16 x i8> %shift
@@ -1444,6 +1493,10 @@ define <32 x i8> @constant_shift_v32i8(<32 x i8> %a) {
 ; SLM-LABEL: 'constant_shift_v32i8'
 ; SLM-NEXT:  Cost Model: Found an estimated cost of 30 for instruction: %shift = shl <32 x i8> %a, <i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0>
 ; SLM-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <32 x i8> %shift
+;
+; AVX512GFNI-LABEL: 'constant_shift_v32i8'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %shift = shl <32 x i8> %a, <i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0>
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <32 x i8> %shift
 ;
   %shift = shl <32 x i8> %a, <i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0>
   ret <32 x i8> %shift
@@ -1489,6 +1542,10 @@ define <64 x i8> @constant_shift_v64i8(<64 x i8> %a) {
 ; SLM-LABEL: 'constant_shift_v64i8'
 ; SLM-NEXT:  Cost Model: Found an estimated cost of 60 for instruction: %shift = shl <64 x i8> %a, <i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0>
 ; SLM-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <64 x i8> %shift
+;
+; AVX512GFNI-LABEL: 'constant_shift_v64i8'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 6 for instruction: %shift = shl <64 x i8> %a, <i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0>
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <64 x i8> %shift
 ;
   %shift = shl <64 x i8> %a, <i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0>
   ret <64 x i8> %shift
@@ -1677,6 +1734,10 @@ define <16 x i16> @splatconstant_shift_v16i16(<16 x i16> %a) {
 ; AVX512BWVL-LABEL: 'splatconstant_shift_v16i16'
 ; AVX512BWVL-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shift = shl <16 x i16> %a, <i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3>
 ; AVX512BWVL-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <16 x i16> %shift
+;
+; AVX512GFNI-LABEL: 'splatconstant_shift_v16i16'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shift = shl <16 x i16> %a, <i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3>
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <16 x i16> %shift
 ;
   %shift = shl <16 x i16> %a, <i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3>
   ret <16 x i16> %shift
@@ -1718,6 +1779,10 @@ define <32 x i16> @splatconstant_shift_v32i16(<32 x i16> %a) {
 ; AVX512BWVL-LABEL: 'splatconstant_shift_v32i16'
 ; AVX512BWVL-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shift = shl <32 x i16> %a, <i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3>
 ; AVX512BWVL-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <32 x i16> %shift
+;
+; AVX512GFNI-LABEL: 'splatconstant_shift_v32i16'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shift = shl <32 x i16> %a, <i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3>
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <32 x i16> %shift
 ;
   %shift = shl <32 x i16> %a, <i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3>
   ret <32 x i16> %shift
@@ -1784,6 +1849,10 @@ define <32 x i8> @splatconstant_shift_v32i8(<32 x i8> %a) {
 ; AVX512BWVL-LABEL: 'splatconstant_shift_v32i8'
 ; AVX512BWVL-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shift = shl <32 x i8> %a, <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, 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>
 ; AVX512BWVL-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <32 x i8> %shift
+;
+; AVX512GFNI-LABEL: 'splatconstant_shift_v32i8'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shift = shl <32 x i8> %a, <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, 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>
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <32 x i8> %shift
 ;
   %shift = shl <32 x i8> %a, <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, 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>
   ret <32 x i8> %shift
@@ -1825,6 +1894,10 @@ define <64 x i8> @splatconstant_shift_v64i8(<64 x i8> %a) {
 ; AVX512BWVL-LABEL: 'splatconstant_shift_v64i8'
 ; AVX512BWVL-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shift = shl <64 x i8> %a, <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, 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, 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, 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>
 ; AVX512BWVL-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <64 x i8> %shift
+;
+; AVX512GFNI-LABEL: 'splatconstant_shift_v64i8'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shift = shl <64 x i8> %a, <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, 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, 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, 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>
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <64 x i8> %shift
 ;
   %shift = shl <64 x i8> %a, <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, 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, 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, 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>
   ret <64 x i8> %shift
@@ -2044,6 +2117,10 @@ define <16 x i16> @test6(<16 x i16> %a) {
 ; SLM-LABEL: 'test6'
 ; SLM-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %shl = shl <16 x i16> %a, <i16 1, i16 1, i16 2, i16 3, i16 7, i16 0, i16 9, i16 11, i16 1, i16 1, i16 2, i16 3, i16 7, i16 0, i16 9, i16 11>
 ; SLM-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <16 x i16> %shl
+;
+; AVX512GFNI-LABEL: 'test6'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shl = shl <16 x i16> %a, <i16 1, i16 1, i16 2, i16 3, i16 7, i16 0, i16 9, i16 11, i16 1, i16 1, i16 2, i16 3, i16 7, i16 0, i16 9, i16 11>
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <16 x i16> %shl
 ;
   %shl = shl <16 x i16> %a, <i16 1, i16 1, i16 2, i16 3, i16 7, i16 0, i16 9, i16 11, i16 1, i16 1, i16 2, i16 3, i16 7, i16 0, i16 9, i16 11>
   ret <16 x i16> %shl
@@ -2162,6 +2239,10 @@ define <32 x i16> @test9(<32 x i16> %a) {
 ; SLM-LABEL: 'test9'
 ; SLM-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %shl = shl <32 x i16> %a, <i16 1, i16 1, i16 2, i16 3, i16 7, i16 0, i16 9, i16 11, i16 1, i16 1, i16 2, i16 3, i16 7, i16 0, i16 9, i16 11, i16 1, i16 1, i16 2, i16 3, i16 7, i16 0, i16 9, i16 11, i16 1, i16 1, i16 2, i16 3, i16 7, i16 0, i16 9, i16 11>
 ; SLM-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <32 x i16> %shl
+;
+; AVX512GFNI-LABEL: 'test9'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shl = shl <32 x i16> %a, <i16 1, i16 1, i16 2, i16 3, i16 7, i16 0, i16 9, i16 11, i16 1, i16 1, i16 2, i16 3, i16 7, i16 0, i16 9, i16 11, i16 1, i16 1, i16 2, i16 3, i16 7, i16 0, i16 9, i16 11, i16 1, i16 1, i16 2, i16 3, i16 7, i16 0, i16 9, i16 11>
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <32 x i16> %shl
 ;
   %shl = shl <32 x i16> %a, <i16 1, i16 1, i16 2, i16 3, i16 7, i16 0, i16 9, i16 11, i16 1, i16 1, i16 2, i16 3, i16 7, i16 0, i16 9, i16 11, i16 1, i16 1, i16 2, i16 3, i16 7, i16 0, i16 9, i16 11, i16 1, i16 1, i16 2, i16 3, i16 7, i16 0, i16 9, i16 11>
   ret <32 x i16> %shl

diff  --git a/llvm/test/Analysis/CostModel/X86/vshift-shl-cost.ll b/llvm/test/Analysis/CostModel/X86/vshift-shl-cost.ll
index b254ff27daccb9..09521cfa2d23d3 100644
--- a/llvm/test/Analysis/CostModel/X86/vshift-shl-cost.ll
+++ b/llvm/test/Analysis/CostModel/X86/vshift-shl-cost.ll
@@ -15,6 +15,7 @@
 ; RUN: opt < %s -mtriple=x86_64-unknown-linux-gnu -passes="print<cost-model>" 2>&1 -disable-output -mcpu=slm | FileCheck %s --check-prefixes=CHECK,SSE,SLM
 ; RUN: opt < %s -mtriple=x86_64-unknown-linux-gnu -passes="print<cost-model>" 2>&1 -disable-output -mcpu=goldmont | FileCheck %s --check-prefixes=CHECK,SSE,SSE42
 ; RUN: opt < %s -mtriple=x86_64-unknown-linux-gnu -passes="print<cost-model>" 2>&1 -disable-output -mcpu=btver2 | FileCheck %s --check-prefixes=CHECK,AVX,AVX1
+; RUN: opt < %s -mtriple=x86_64-unknown-linux-gnu -passes="print<cost-model>" 2>&1 -disable-output -mcpu=tigerlake | FileCheck %s --check-prefixes=CHECK,AVX512,AVX512GFNI
 
 ; Verify the cost of vector shift left instructions.
 
@@ -236,6 +237,10 @@ define <8 x i16> @var_shift_v8i16(<8 x i16> %a, <8 x i16> %b) {
 ; SLM-LABEL: 'var_shift_v8i16'
 ; SLM-NEXT:  Cost Model: Found an estimated cost of 11 for instruction: %shift = shl <8 x i16> %a, %b
 ; SLM-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <8 x i16> %shift
+;
+; AVX512GFNI-LABEL: 'var_shift_v8i16'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shift = shl <8 x i16> %a, %b
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <8 x i16> %shift
 ;
   %shift = shl <8 x i16> %a, %b
   ret <8 x i16> %shift
@@ -281,6 +286,10 @@ define <16 x i16> @var_shift_v16i16(<16 x i16> %a, <16 x i16> %b) {
 ; SLM-LABEL: 'var_shift_v16i16'
 ; SLM-NEXT:  Cost Model: Found an estimated cost of 22 for instruction: %shift = shl <16 x i16> %a, %b
 ; SLM-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <16 x i16> %shift
+;
+; AVX512GFNI-LABEL: 'var_shift_v16i16'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shift = shl <16 x i16> %a, %b
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <16 x i16> %shift
 ;
   %shift = shl <16 x i16> %a, %b
   ret <16 x i16> %shift
@@ -326,6 +335,10 @@ define <32 x i16> @var_shift_v32i16(<32 x i16> %a, <32 x i16> %b) {
 ; SLM-LABEL: 'var_shift_v32i16'
 ; SLM-NEXT:  Cost Model: Found an estimated cost of 44 for instruction: %shift = shl <32 x i16> %a, %b
 ; SLM-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <32 x i16> %shift
+;
+; AVX512GFNI-LABEL: 'var_shift_v32i16'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shift = shl <32 x i16> %a, %b
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <32 x i16> %shift
 ;
   %shift = shl <32 x i16> %a, %b
   ret <32 x i16> %shift
@@ -371,6 +384,10 @@ define <16 x i8> @var_shift_v16i8(<16 x i8> %a, <16 x i8> %b) {
 ; SLM-LABEL: 'var_shift_v16i8'
 ; SLM-NEXT:  Cost Model: Found an estimated cost of 15 for instruction: %shift = shl <16 x i8> %a, %b
 ; SLM-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <16 x i8> %shift
+;
+; AVX512GFNI-LABEL: 'var_shift_v16i8'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %shift = shl <16 x i8> %a, %b
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <16 x i8> %shift
 ;
   %shift = shl <16 x i8> %a, %b
   ret <16 x i8> %shift
@@ -416,6 +433,10 @@ define <32 x i8> @var_shift_v32i8(<32 x i8> %a, <32 x i8> %b) {
 ; SLM-LABEL: 'var_shift_v32i8'
 ; SLM-NEXT:  Cost Model: Found an estimated cost of 30 for instruction: %shift = shl <32 x i8> %a, %b
 ; SLM-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <32 x i8> %shift
+;
+; AVX512GFNI-LABEL: 'var_shift_v32i8'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %shift = shl <32 x i8> %a, %b
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <32 x i8> %shift
 ;
   %shift = shl <32 x i8> %a, %b
   ret <32 x i8> %shift
@@ -461,6 +482,10 @@ define <64 x i8> @var_shift_v64i8(<64 x i8> %a, <64 x i8> %b) {
 ; SLM-LABEL: 'var_shift_v64i8'
 ; SLM-NEXT:  Cost Model: Found an estimated cost of 60 for instruction: %shift = shl <64 x i8> %a, %b
 ; SLM-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <64 x i8> %shift
+;
+; AVX512GFNI-LABEL: 'var_shift_v64i8'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 6 for instruction: %shift = shl <64 x i8> %a, %b
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <64 x i8> %shift
 ;
   %shift = shl <64 x i8> %a, %b
   ret <64 x i8> %shift
@@ -868,6 +893,12 @@ define <32 x i16> @splatvar_shift_v32i16(<32 x i16> %a, i16 %b) {
 ; SLM-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %splat = shufflevector <32 x i16> %insert, <32 x i16> undef, <32 x i32> zeroinitializer
 ; SLM-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %shift = shl <32 x i16> %a, %splat
 ; SLM-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <32 x i16> %shift
+;
+; AVX512GFNI-LABEL: 'splatvar_shift_v32i16'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %insert = insertelement <32 x i16> undef, i16 %b, i32 0
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %splat = shufflevector <32 x i16> %insert, <32 x i16> undef, <32 x i32> zeroinitializer
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %shift = shl <32 x i16> %a, %splat
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <32 x i16> %shift
 ;
   %insert = insertelement <32 x i16> undef, i16 %b, i32 0
   %splat = shufflevector <32 x i16> %insert, <32 x i16> undef, <32 x i32> zeroinitializer
@@ -1051,6 +1082,12 @@ define <64 x i8> @splatvar_shift_v64i8(<64 x i8> %a, i8 %b) {
 ; SLM-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %splat = shufflevector <64 x i8> %insert, <64 x i8> undef, <64 x i32> zeroinitializer
 ; SLM-NEXT:  Cost Model: Found an estimated cost of 36 for instruction: %shift = shl <64 x i8> %a, %splat
 ; SLM-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <64 x i8> %shift
+;
+; AVX512GFNI-LABEL: 'splatvar_shift_v64i8'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %insert = insertelement <64 x i8> undef, i8 %b, i32 0
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %splat = shufflevector <64 x i8> %insert, <64 x i8> undef, <64 x i32> zeroinitializer
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %shift = shl <64 x i8> %a, %splat
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <64 x i8> %shift
 ;
   %insert = insertelement <64 x i8> undef, i8 %b, i32 0
   %splat = shufflevector <64 x i8> %insert, <64 x i8> undef, <64 x i32> zeroinitializer
@@ -1305,6 +1342,10 @@ define <16 x i16> @constant_shift_v16i16(<16 x i16> %a) {
 ; SLM-LABEL: 'constant_shift_v16i16'
 ; SLM-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %shift = shl <16 x i16> %a, <i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7>
 ; SLM-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <16 x i16> %shift
+;
+; AVX512GFNI-LABEL: 'constant_shift_v16i16'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shift = shl <16 x i16> %a, <i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7>
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <16 x i16> %shift
 ;
   %shift = shl <16 x i16> %a, <i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7>
   ret <16 x i16> %shift
@@ -1354,6 +1395,10 @@ define <32 x i16> @constant_shift_v32i16(<32 x i16> %a) {
 ; SLM-LABEL: 'constant_shift_v32i16'
 ; SLM-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %shift = shl <32 x i16> %a, <i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7>
 ; SLM-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <32 x i16> %shift
+;
+; AVX512GFNI-LABEL: 'constant_shift_v32i16'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shift = shl <32 x i16> %a, <i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7>
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <32 x i16> %shift
 ;
   %shift = shl <32 x i16> %a, <i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7>
   ret <32 x i16> %shift
@@ -1399,6 +1444,10 @@ define <16 x i8> @constant_shift_v16i8(<16 x i8> %a) {
 ; SLM-LABEL: 'constant_shift_v16i8'
 ; SLM-NEXT:  Cost Model: Found an estimated cost of 15 for instruction: %shift = shl <16 x i8> %a, <i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0>
 ; SLM-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <16 x i8> %shift
+;
+; AVX512GFNI-LABEL: 'constant_shift_v16i8'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %shift = shl <16 x i8> %a, <i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0>
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <16 x i8> %shift
 ;
   %shift = shl <16 x i8> %a, <i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0>
   ret <16 x i8> %shift
@@ -1444,6 +1493,10 @@ define <32 x i8> @constant_shift_v32i8(<32 x i8> %a) {
 ; SLM-LABEL: 'constant_shift_v32i8'
 ; SLM-NEXT:  Cost Model: Found an estimated cost of 30 for instruction: %shift = shl <32 x i8> %a, <i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0>
 ; SLM-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <32 x i8> %shift
+;
+; AVX512GFNI-LABEL: 'constant_shift_v32i8'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %shift = shl <32 x i8> %a, <i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0>
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <32 x i8> %shift
 ;
   %shift = shl <32 x i8> %a, <i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0>
   ret <32 x i8> %shift
@@ -1489,6 +1542,10 @@ define <64 x i8> @constant_shift_v64i8(<64 x i8> %a) {
 ; SLM-LABEL: 'constant_shift_v64i8'
 ; SLM-NEXT:  Cost Model: Found an estimated cost of 60 for instruction: %shift = shl <64 x i8> %a, <i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0>
 ; SLM-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <64 x i8> %shift
+;
+; AVX512GFNI-LABEL: 'constant_shift_v64i8'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 6 for instruction: %shift = shl <64 x i8> %a, <i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0>
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <64 x i8> %shift
 ;
   %shift = shl <64 x i8> %a, <i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0>
   ret <64 x i8> %shift
@@ -1677,6 +1734,10 @@ define <16 x i16> @splatconstant_shift_v16i16(<16 x i16> %a) {
 ; AVX512BWVL-LABEL: 'splatconstant_shift_v16i16'
 ; AVX512BWVL-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shift = shl <16 x i16> %a, <i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3>
 ; AVX512BWVL-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <16 x i16> %shift
+;
+; AVX512GFNI-LABEL: 'splatconstant_shift_v16i16'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shift = shl <16 x i16> %a, <i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3>
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <16 x i16> %shift
 ;
   %shift = shl <16 x i16> %a, <i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3>
   ret <16 x i16> %shift
@@ -1718,6 +1779,10 @@ define <32 x i16> @splatconstant_shift_v32i16(<32 x i16> %a) {
 ; AVX512BWVL-LABEL: 'splatconstant_shift_v32i16'
 ; AVX512BWVL-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shift = shl <32 x i16> %a, <i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3>
 ; AVX512BWVL-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <32 x i16> %shift
+;
+; AVX512GFNI-LABEL: 'splatconstant_shift_v32i16'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shift = shl <32 x i16> %a, <i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3>
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <32 x i16> %shift
 ;
   %shift = shl <32 x i16> %a, <i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3>
   ret <32 x i16> %shift
@@ -1784,6 +1849,10 @@ define <32 x i8> @splatconstant_shift_v32i8(<32 x i8> %a) {
 ; AVX512BWVL-LABEL: 'splatconstant_shift_v32i8'
 ; AVX512BWVL-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shift = shl <32 x i8> %a, <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, 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>
 ; AVX512BWVL-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <32 x i8> %shift
+;
+; AVX512GFNI-LABEL: 'splatconstant_shift_v32i8'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shift = shl <32 x i8> %a, <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, 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>
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <32 x i8> %shift
 ;
   %shift = shl <32 x i8> %a, <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, 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>
   ret <32 x i8> %shift
@@ -1825,6 +1894,10 @@ define <64 x i8> @splatconstant_shift_v64i8(<64 x i8> %a) {
 ; AVX512BWVL-LABEL: 'splatconstant_shift_v64i8'
 ; AVX512BWVL-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shift = shl <64 x i8> %a, <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, 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, 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, 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>
 ; AVX512BWVL-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <64 x i8> %shift
+;
+; AVX512GFNI-LABEL: 'splatconstant_shift_v64i8'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shift = shl <64 x i8> %a, <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, 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, 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, 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>
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <64 x i8> %shift
 ;
   %shift = shl <64 x i8> %a, <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, 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, 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, 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>
   ret <64 x i8> %shift
@@ -2044,6 +2117,10 @@ define <16 x i16> @test6(<16 x i16> %a) {
 ; SLM-LABEL: 'test6'
 ; SLM-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %shl = shl <16 x i16> %a, <i16 1, i16 1, i16 2, i16 3, i16 7, i16 0, i16 9, i16 11, i16 1, i16 1, i16 2, i16 3, i16 7, i16 0, i16 9, i16 11>
 ; SLM-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <16 x i16> %shl
+;
+; AVX512GFNI-LABEL: 'test6'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shl = shl <16 x i16> %a, <i16 1, i16 1, i16 2, i16 3, i16 7, i16 0, i16 9, i16 11, i16 1, i16 1, i16 2, i16 3, i16 7, i16 0, i16 9, i16 11>
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <16 x i16> %shl
 ;
   %shl = shl <16 x i16> %a, <i16 1, i16 1, i16 2, i16 3, i16 7, i16 0, i16 9, i16 11, i16 1, i16 1, i16 2, i16 3, i16 7, i16 0, i16 9, i16 11>
   ret <16 x i16> %shl
@@ -2162,6 +2239,10 @@ define <32 x i16> @test9(<32 x i16> %a) {
 ; SLM-LABEL: 'test9'
 ; SLM-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %shl = shl <32 x i16> %a, <i16 1, i16 1, i16 2, i16 3, i16 7, i16 0, i16 9, i16 11, i16 1, i16 1, i16 2, i16 3, i16 7, i16 0, i16 9, i16 11, i16 1, i16 1, i16 2, i16 3, i16 7, i16 0, i16 9, i16 11, i16 1, i16 1, i16 2, i16 3, i16 7, i16 0, i16 9, i16 11>
 ; SLM-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <32 x i16> %shl
+;
+; AVX512GFNI-LABEL: 'test9'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shl = shl <32 x i16> %a, <i16 1, i16 1, i16 2, i16 3, i16 7, i16 0, i16 9, i16 11, i16 1, i16 1, i16 2, i16 3, i16 7, i16 0, i16 9, i16 11, i16 1, i16 1, i16 2, i16 3, i16 7, i16 0, i16 9, i16 11, i16 1, i16 1, i16 2, i16 3, i16 7, i16 0, i16 9, i16 11>
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <32 x i16> %shl
 ;
   %shl = shl <32 x i16> %a, <i16 1, i16 1, i16 2, i16 3, i16 7, i16 0, i16 9, i16 11, i16 1, i16 1, i16 2, i16 3, i16 7, i16 0, i16 9, i16 11, i16 1, i16 1, i16 2, i16 3, i16 7, i16 0, i16 9, i16 11, i16 1, i16 1, i16 2, i16 3, i16 7, i16 0, i16 9, i16 11>
   ret <32 x i16> %shl

diff  --git a/llvm/test/Analysis/CostModel/X86/vshift-shl-latency.ll b/llvm/test/Analysis/CostModel/X86/vshift-shl-latency.ll
index c53dbc5f586aa4..42c91144ff6f99 100644
--- a/llvm/test/Analysis/CostModel/X86/vshift-shl-latency.ll
+++ b/llvm/test/Analysis/CostModel/X86/vshift-shl-latency.ll
@@ -15,6 +15,7 @@
 ; RUN: opt < %s -mtriple=x86_64-unknown-linux-gnu -passes="print<cost-model>" 2>&1 -disable-output -cost-kind=latency -mcpu=slm | FileCheck %s --check-prefixes=SSE,SSE42
 ; RUN: opt < %s -mtriple=x86_64-unknown-linux-gnu -passes="print<cost-model>" 2>&1 -disable-output -cost-kind=latency -mcpu=goldmont | FileCheck %s --check-prefixes=SSE,SSE42
 ; RUN: opt < %s -mtriple=x86_64-unknown-linux-gnu -passes="print<cost-model>" 2>&1 -disable-output -cost-kind=latency -mcpu=btver2 | FileCheck %s --check-prefixes=AVX1
+; RUN: opt < %s -mtriple=x86_64-unknown-linux-gnu -passes="print<cost-model>" 2>&1 -disable-output -cost-kind=latency -mcpu=tigerlake | FileCheck %s --check-prefixes=AVX512,AVX512GFNI
 
 ; Verify the cost of vector logical shift right instructions.
 
@@ -240,6 +241,10 @@ define <8 x i16> @var_shift_v8i16(<8 x i16> %a, <8 x i16> %b) {
 ; AVX512BW-LABEL: 'var_shift_v8i16'
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shift = shl <8 x i16> %a, %b
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <8 x i16> %shift
+;
+; AVX512GFNI-LABEL: 'var_shift_v8i16'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shift = shl <8 x i16> %a, %b
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <8 x i16> %shift
 ;
   %shift = shl <8 x i16> %a, %b
   ret <8 x i16> %shift
@@ -277,6 +282,10 @@ define <16 x i16> @var_shift_v16i16(<16 x i16> %a, <16 x i16> %b) {
 ; AVX512BW-LABEL: 'var_shift_v16i16'
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shift = shl <16 x i16> %a, %b
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <16 x i16> %shift
+;
+; AVX512GFNI-LABEL: 'var_shift_v16i16'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shift = shl <16 x i16> %a, %b
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <16 x i16> %shift
 ;
   %shift = shl <16 x i16> %a, %b
   ret <16 x i16> %shift
@@ -314,6 +323,10 @@ define <32 x i16> @var_shift_v32i16(<32 x i16> %a, <32 x i16> %b) {
 ; AVX512BW-LABEL: 'var_shift_v32i16'
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shift = shl <32 x i16> %a, %b
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <32 x i16> %shift
+;
+; AVX512GFNI-LABEL: 'var_shift_v32i16'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shift = shl <32 x i16> %a, %b
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <32 x i16> %shift
 ;
   %shift = shl <32 x i16> %a, %b
   ret <32 x i16> %shift
@@ -351,6 +364,10 @@ define <16 x i8> @var_shift_v16i8(<16 x i8> %a, <16 x i8> %b) {
 ; AVX512BW-LABEL: 'var_shift_v16i8'
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %shift = shl <16 x i8> %a, %b
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <16 x i8> %shift
+;
+; AVX512GFNI-LABEL: 'var_shift_v16i8'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %shift = shl <16 x i8> %a, %b
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <16 x i8> %shift
 ;
   %shift = shl <16 x i8> %a, %b
   ret <16 x i8> %shift
@@ -830,6 +847,12 @@ define <32 x i16> @splatvar_shift_v32i16(<32 x i16> %a, i16 %b) {
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %splat = shufflevector <32 x i16> %insert, <32 x i16> undef, <32 x i32> zeroinitializer
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %shift = shl <32 x i16> %a, %splat
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <32 x i16> %shift
+;
+; AVX512GFNI-LABEL: 'splatvar_shift_v32i16'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %insert = insertelement <32 x i16> undef, i16 %b, i32 0
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %splat = shufflevector <32 x i16> %insert, <32 x i16> undef, <32 x i32> zeroinitializer
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %shift = shl <32 x i16> %a, %splat
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <32 x i16> %shift
 ;
   %insert = insertelement <32 x i16> undef, i16 %b, i32 0
   %splat = shufflevector <32 x i16> %insert, <32 x i16> undef, <32 x i32> zeroinitializer
@@ -983,6 +1006,12 @@ define <64 x i8> @splatvar_shift_v64i8(<64 x i8> %a, i8 %b) {
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %splat = shufflevector <64 x i8> %insert, <64 x i8> undef, <64 x i32> zeroinitializer
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 7 for instruction: %shift = shl <64 x i8> %a, %splat
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <64 x i8> %shift
+;
+; AVX512GFNI-LABEL: 'splatvar_shift_v64i8'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %insert = insertelement <64 x i8> undef, i8 %b, i32 0
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %splat = shufflevector <64 x i8> %insert, <64 x i8> undef, <64 x i32> zeroinitializer
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 7 for instruction: %shift = shl <64 x i8> %a, %splat
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <64 x i8> %shift
 ;
   %insert = insertelement <64 x i8> undef, i8 %b, i32 0
   %splat = shufflevector <64 x i8> %insert, <64 x i8> undef, <64 x i32> zeroinitializer
@@ -1208,6 +1237,10 @@ define <8 x i16> @constant_shift_v8i16(<8 x i16> %a) {
 ; AVX512BW-LABEL: 'constant_shift_v8i16'
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shift = shl <8 x i16> %a, <i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7>
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <8 x i16> %shift
+;
+; AVX512GFNI-LABEL: 'constant_shift_v8i16'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shift = shl <8 x i16> %a, <i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7>
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <8 x i16> %shift
 ;
   %shift = shl <8 x i16> %a, <i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7>
   ret <8 x i16> %shift
@@ -1241,6 +1274,10 @@ define <16 x i16> @constant_shift_v16i16(<16 x i16> %a) {
 ; AVX512BW-LABEL: 'constant_shift_v16i16'
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shift = shl <16 x i16> %a, <i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7>
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <16 x i16> %shift
+;
+; AVX512GFNI-LABEL: 'constant_shift_v16i16'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shift = shl <16 x i16> %a, <i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7>
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <16 x i16> %shift
 ;
   %shift = shl <16 x i16> %a, <i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7>
   ret <16 x i16> %shift
@@ -1274,6 +1311,10 @@ define <32 x i16> @constant_shift_v32i16(<32 x i16> %a) {
 ; AVX512BW-LABEL: 'constant_shift_v32i16'
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shift = shl <32 x i16> %a, <i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7>
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <32 x i16> %shift
+;
+; AVX512GFNI-LABEL: 'constant_shift_v32i16'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shift = shl <32 x i16> %a, <i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7>
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <32 x i16> %shift
 ;
   %shift = shl <32 x i16> %a, <i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7>
   ret <32 x i16> %shift
@@ -1311,6 +1352,10 @@ define <16 x i8> @constant_shift_v16i8(<16 x i8> %a) {
 ; AVX512BW-LABEL: 'constant_shift_v16i8'
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %shift = shl <16 x i8> %a, <i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0>
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <16 x i8> %shift
+;
+; AVX512GFNI-LABEL: 'constant_shift_v16i8'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %shift = shl <16 x i8> %a, <i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0>
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <16 x i8> %shift
 ;
   %shift = shl <16 x i8> %a, <i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0>
   ret <16 x i8> %shift
@@ -1617,6 +1662,10 @@ define <16 x i16> @splatconstant_shift_v16i16(<16 x i16> %a) {
 ; AVX512BW-LABEL: 'splatconstant_shift_v16i16'
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shift = shl <16 x i16> %a, <i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3>
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <16 x i16> %shift
+;
+; AVX512GFNI-LABEL: 'splatconstant_shift_v16i16'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shift = shl <16 x i16> %a, <i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3>
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <16 x i16> %shift
 ;
   %shift = shl <16 x i16> %a, <i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3>
   ret <16 x i16> %shift
@@ -1650,6 +1699,10 @@ define <32 x i16> @splatconstant_shift_v32i16(<32 x i16> %a) {
 ; AVX512BW-LABEL: 'splatconstant_shift_v32i16'
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shift = shl <32 x i16> %a, <i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3>
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <32 x i16> %shift
+;
+; AVX512GFNI-LABEL: 'splatconstant_shift_v32i16'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shift = shl <32 x i16> %a, <i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3>
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <32 x i16> %shift
 ;
   %shift = shl <32 x i16> %a, <i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3>
   ret <32 x i16> %shift
@@ -1683,6 +1736,10 @@ define <16 x i8> @splatconstant_shift_v16i8(<16 x i8> %a) {
 ; AVX512BW-LABEL: 'splatconstant_shift_v16i8'
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 7 for instruction: %shift = shl <16 x i8> %a, <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>
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <16 x i8> %shift
+;
+; AVX512GFNI-LABEL: 'splatconstant_shift_v16i8'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 7 for instruction: %shift = shl <16 x i8> %a, <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>
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <16 x i8> %shift
 ;
   %shift = shl <16 x i8> %a, <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>
   ret <16 x i8> %shift
@@ -1745,6 +1802,10 @@ define <64 x i8> @splatconstant_shift_v64i8(<64 x i8> %a) {
 ; AVX512BW-LABEL: 'splatconstant_shift_v64i8'
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %shift = shl <64 x i8> %a, <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, 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, 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, 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>
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <64 x i8> %shift
+;
+; AVX512GFNI-LABEL: 'splatconstant_shift_v64i8'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %shift = shl <64 x i8> %a, <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, 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, 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, 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>
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <64 x i8> %shift
 ;
   %shift = shl <64 x i8> %a, <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, 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, 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, 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>
   ret <64 x i8> %shift

diff  --git a/llvm/test/Analysis/CostModel/X86/vshift-shl-sizelatency.ll b/llvm/test/Analysis/CostModel/X86/vshift-shl-sizelatency.ll
index cbf0f24b90fb89..47b24df063ef70 100644
--- a/llvm/test/Analysis/CostModel/X86/vshift-shl-sizelatency.ll
+++ b/llvm/test/Analysis/CostModel/X86/vshift-shl-sizelatency.ll
@@ -15,6 +15,7 @@
 ; RUN: opt < %s -mtriple=x86_64-unknown-linux-gnu -passes="print<cost-model>" 2>&1 -disable-output -cost-kind=size-latency -mcpu=slm | FileCheck %s --check-prefixes=CHECK,SSE,SLM
 ; RUN: opt < %s -mtriple=x86_64-unknown-linux-gnu -passes="print<cost-model>" 2>&1 -disable-output -cost-kind=size-latency -mcpu=goldmont | FileCheck %s --check-prefixes=CHECK,SSE,SSE42
 ; RUN: opt < %s -mtriple=x86_64-unknown-linux-gnu -passes="print<cost-model>" 2>&1 -disable-output -cost-kind=size-latency -mcpu=btver2 | FileCheck %s --check-prefixes=CHECK,AVX,AVX1
+; RUN: opt < %s -mtriple=x86_64-unknown-linux-gnu -passes="print<cost-model>" 2>&1 -disable-output -cost-kind=size-latency -mcpu=tigerlake | FileCheck %s --check-prefixes=CHECK,AVX512,AVX512GFNI
 
 ; Verify the cost of vector logical shift right instructions.
 
@@ -263,6 +264,10 @@ define <8 x i16> @var_shift_v8i16(<8 x i16> %a, <8 x i16> %b) {
 ; SLM-NEXT:  Cost Model: Found an estimated cost of 11 for instruction: %shift = shl <8 x i16> %a, %b
 ; SLM-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <8 x i16> %shift
 ;
+; AVX512GFNI-LABEL: 'var_shift_v8i16'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shift = shl <8 x i16> %a, %b
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <8 x i16> %shift
+;
 ; BTVER2-LABEL: 'var_shift_v8i16'
 ; BTVER2-NEXT:  Cost Model: Found an estimated cost of 11 for instruction: %shift = shl <8 x i16> %a, %b
 ; BTVER2-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <8 x i16> %shift
@@ -303,6 +308,10 @@ define <16 x i16> @var_shift_v16i16(<16 x i16> %a, <16 x i16> %b) {
 ; SLM-NEXT:  Cost Model: Found an estimated cost of 22 for instruction: %shift = shl <16 x i16> %a, %b
 ; SLM-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <16 x i16> %shift
 ;
+; AVX512GFNI-LABEL: 'var_shift_v16i16'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shift = shl <16 x i16> %a, %b
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <16 x i16> %shift
+;
 ; BTVER2-LABEL: 'var_shift_v16i16'
 ; BTVER2-NEXT:  Cost Model: Found an estimated cost of 25 for instruction: %shift = shl <16 x i16> %a, %b
 ; BTVER2-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <16 x i16> %shift
@@ -343,6 +352,10 @@ define <32 x i16> @var_shift_v32i16(<32 x i16> %a, <32 x i16> %b) {
 ; SLM-NEXT:  Cost Model: Found an estimated cost of 44 for instruction: %shift = shl <32 x i16> %a, %b
 ; SLM-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <32 x i16> %shift
 ;
+; AVX512GFNI-LABEL: 'var_shift_v32i16'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shift = shl <32 x i16> %a, %b
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <32 x i16> %shift
+;
 ; BTVER2-LABEL: 'var_shift_v32i16'
 ; BTVER2-NEXT:  Cost Model: Found an estimated cost of 50 for instruction: %shift = shl <32 x i16> %a, %b
 ; BTVER2-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <32 x i16> %shift
@@ -383,6 +396,10 @@ define <16 x i8> @var_shift_v16i8(<16 x i8> %a, <16 x i8> %b) {
 ; SLM-NEXT:  Cost Model: Found an estimated cost of 22 for instruction: %shift = shl <16 x i8> %a, %b
 ; SLM-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <16 x i8> %shift
 ;
+; AVX512GFNI-LABEL: 'var_shift_v16i8'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 5 for instruction: %shift = shl <16 x i8> %a, %b
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <16 x i8> %shift
+;
 ; BTVER2-LABEL: 'var_shift_v16i8'
 ; BTVER2-NEXT:  Cost Model: Found an estimated cost of 17 for instruction: %shift = shl <16 x i8> %a, %b
 ; BTVER2-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <16 x i8> %shift
@@ -423,6 +440,10 @@ define <32 x i8> @var_shift_v32i8(<32 x i8> %a, <32 x i8> %b) {
 ; SLM-NEXT:  Cost Model: Found an estimated cost of 44 for instruction: %shift = shl <32 x i8> %a, %b
 ; SLM-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <32 x i8> %shift
 ;
+; AVX512GFNI-LABEL: 'var_shift_v32i8'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 16 for instruction: %shift = shl <32 x i8> %a, %b
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <32 x i8> %shift
+;
 ; BTVER2-LABEL: 'var_shift_v32i8'
 ; BTVER2-NEXT:  Cost Model: Found an estimated cost of 40 for instruction: %shift = shl <32 x i8> %a, %b
 ; BTVER2-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <32 x i8> %shift
@@ -463,6 +484,10 @@ define <64 x i8> @var_shift_v64i8(<64 x i8> %a, <64 x i8> %b) {
 ; SLM-NEXT:  Cost Model: Found an estimated cost of 88 for instruction: %shift = shl <64 x i8> %a, %b
 ; SLM-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <64 x i8> %shift
 ;
+; AVX512GFNI-LABEL: 'var_shift_v64i8'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 15 for instruction: %shift = shl <64 x i8> %a, %b
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <64 x i8> %shift
+;
 ; BTVER2-LABEL: 'var_shift_v64i8'
 ; BTVER2-NEXT:  Cost Model: Found an estimated cost of 80 for instruction: %shift = shl <64 x i8> %a, %b
 ; BTVER2-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <64 x i8> %shift
@@ -824,6 +849,12 @@ define <32 x i16> @splatvar_shift_v32i16(<32 x i16> %a, i16 %b) {
 ; SLM-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %splat = shufflevector <32 x i16> %insert, <32 x i16> undef, <32 x i32> zeroinitializer
 ; SLM-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %shift = shl <32 x i16> %a, %splat
 ; SLM-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <32 x i16> %shift
+;
+; AVX512GFNI-LABEL: 'splatvar_shift_v32i16'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %insert = insertelement <32 x i16> undef, i16 %b, i32 0
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %splat = shufflevector <32 x i16> %insert, <32 x i16> undef, <32 x i32> zeroinitializer
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %shift = shl <32 x i16> %a, %splat
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <32 x i16> %shift
 ;
   %insert = insertelement <32 x i16> undef, i16 %b, i32 0
   %splat = shufflevector <32 x i16> %insert, <32 x i16> undef, <32 x i32> zeroinitializer
@@ -995,6 +1026,12 @@ define <64 x i8> @splatvar_shift_v64i8(<64 x i8> %a, i8 %b) {
 ; SLM-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %splat = shufflevector <64 x i8> %insert, <64 x i8> undef, <64 x i32> zeroinitializer
 ; SLM-NEXT:  Cost Model: Found an estimated cost of 36 for instruction: %shift = shl <64 x i8> %a, %splat
 ; SLM-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <64 x i8> %shift
+;
+; AVX512GFNI-LABEL: 'splatvar_shift_v64i8'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %insert = insertelement <64 x i8> undef, i8 %b, i32 0
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %splat = shufflevector <64 x i8> %insert, <64 x i8> undef, <64 x i32> zeroinitializer
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %shift = shl <64 x i8> %a, %splat
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <64 x i8> %shift
 ;
   %insert = insertelement <64 x i8> undef, i8 %b, i32 0
   %splat = shufflevector <64 x i8> %insert, <64 x i8> undef, <64 x i32> zeroinitializer
@@ -1238,6 +1275,10 @@ define <16 x i16> @constant_shift_v16i16(<16 x i16> %a) {
 ; AVX512BW-LABEL: 'constant_shift_v16i16'
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shift = shl <16 x i16> %a, <i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7>
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <16 x i16> %shift
+;
+; AVX512GFNI-LABEL: 'constant_shift_v16i16'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shift = shl <16 x i16> %a, <i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7>
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <16 x i16> %shift
 ;
   %shift = shl <16 x i16> %a, <i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7>
   ret <16 x i16> %shift
@@ -1271,6 +1312,10 @@ define <32 x i16> @constant_shift_v32i16(<32 x i16> %a) {
 ; AVX512BW-LABEL: 'constant_shift_v32i16'
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shift = shl <32 x i16> %a, <i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7>
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <32 x i16> %shift
+;
+; AVX512GFNI-LABEL: 'constant_shift_v32i16'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shift = shl <32 x i16> %a, <i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7>
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <32 x i16> %shift
 ;
   %shift = shl <32 x i16> %a, <i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7>
   ret <32 x i16> %shift
@@ -1309,6 +1354,10 @@ define <16 x i8> @constant_shift_v16i8(<16 x i8> %a) {
 ; SLM-NEXT:  Cost Model: Found an estimated cost of 22 for instruction: %shift = shl <16 x i8> %a, <i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0>
 ; SLM-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <16 x i8> %shift
 ;
+; AVX512GFNI-LABEL: 'constant_shift_v16i8'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 5 for instruction: %shift = shl <16 x i8> %a, <i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0>
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <16 x i8> %shift
+;
 ; BTVER2-LABEL: 'constant_shift_v16i8'
 ; BTVER2-NEXT:  Cost Model: Found an estimated cost of 17 for instruction: %shift = shl <16 x i8> %a, <i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0>
 ; BTVER2-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <16 x i8> %shift
@@ -1349,6 +1398,10 @@ define <32 x i8> @constant_shift_v32i8(<32 x i8> %a) {
 ; SLM-NEXT:  Cost Model: Found an estimated cost of 44 for instruction: %shift = shl <32 x i8> %a, <i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0>
 ; SLM-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <32 x i8> %shift
 ;
+; AVX512GFNI-LABEL: 'constant_shift_v32i8'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 16 for instruction: %shift = shl <32 x i8> %a, <i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0>
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <32 x i8> %shift
+;
 ; BTVER2-LABEL: 'constant_shift_v32i8'
 ; BTVER2-NEXT:  Cost Model: Found an estimated cost of 40 for instruction: %shift = shl <32 x i8> %a, <i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0>
 ; BTVER2-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <32 x i8> %shift
@@ -1389,6 +1442,10 @@ define <64 x i8> @constant_shift_v64i8(<64 x i8> %a) {
 ; SLM-NEXT:  Cost Model: Found an estimated cost of 88 for instruction: %shift = shl <64 x i8> %a, <i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0>
 ; SLM-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <64 x i8> %shift
 ;
+; AVX512GFNI-LABEL: 'constant_shift_v64i8'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 15 for instruction: %shift = shl <64 x i8> %a, <i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0>
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <64 x i8> %shift
+;
 ; BTVER2-LABEL: 'constant_shift_v64i8'
 ; BTVER2-NEXT:  Cost Model: Found an estimated cost of 80 for instruction: %shift = shl <64 x i8> %a, <i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0>
 ; BTVER2-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <64 x i8> %shift
@@ -1571,6 +1628,10 @@ define <16 x i16> @splatconstant_shift_v16i16(<16 x i16> %a) {
 ; AVX512BW-LABEL: 'splatconstant_shift_v16i16'
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shift = shl <16 x i16> %a, <i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3>
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <16 x i16> %shift
+;
+; AVX512GFNI-LABEL: 'splatconstant_shift_v16i16'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shift = shl <16 x i16> %a, <i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3>
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <16 x i16> %shift
 ;
   %shift = shl <16 x i16> %a, <i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3>
   ret <16 x i16> %shift
@@ -1604,6 +1665,10 @@ define <32 x i16> @splatconstant_shift_v32i16(<32 x i16> %a) {
 ; AVX512BW-LABEL: 'splatconstant_shift_v32i16'
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shift = shl <32 x i16> %a, <i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3>
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <32 x i16> %shift
+;
+; AVX512GFNI-LABEL: 'splatconstant_shift_v32i16'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shift = shl <32 x i16> %a, <i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3>
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <32 x i16> %shift
 ;
   %shift = shl <32 x i16> %a, <i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3>
   ret <32 x i16> %shift
@@ -1662,6 +1727,10 @@ define <32 x i8> @splatconstant_shift_v32i8(<32 x i8> %a) {
 ; AVX512BW-LABEL: 'splatconstant_shift_v32i8'
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %shift = shl <32 x i8> %a, <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, 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>
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <32 x i8> %shift
+;
+; AVX512GFNI-LABEL: 'splatconstant_shift_v32i8'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %shift = shl <32 x i8> %a, <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, 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>
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <32 x i8> %shift
 ;
   %shift = shl <32 x i8> %a, <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, 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>
   ret <32 x i8> %shift
@@ -1695,6 +1764,10 @@ define <64 x i8> @splatconstant_shift_v64i8(<64 x i8> %a) {
 ; AVX512BW-LABEL: 'splatconstant_shift_v64i8'
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %shift = shl <64 x i8> %a, <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, 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, 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, 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>
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <64 x i8> %shift
+;
+; AVX512GFNI-LABEL: 'splatconstant_shift_v64i8'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %shift = shl <64 x i8> %a, <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, 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, 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, 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>
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <64 x i8> %shift
 ;
   %shift = shl <64 x i8> %a, <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, 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, 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, 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>
   ret <64 x i8> %shift


        
