Enablement for AMD znver2 architecture - skeleton patch

[GGanesh]

Summary

This patch enables the following

1. AMD family 17h "znver2" tune flag (-march, -mcpu).
2. ISAs that are enabled for "znver2" architecture.
3. For the time being, it uses the znver1 scheduler model.
4. Tests are updated.

[craig.topper]

include/llvm/Support/X86TargetParser.def

93

This list must be kept in the same order as libgcc's implementation for __builtin_cpu_is. New entries can't be added in the middle. If libgcc doesn't support znver2 yet, then you should use X86_CPU_SUBTYPE instead of X86_CPU_SUBTYPE_COMPAT.

lib/Target/X86/X86.td

1188

Why is CLZERO not in ZNFeatures? Looks common to both znver1 and znver2

[craig.topper]

include/llvm/Support/X86TargetParser.def

93

Looks like libgcc does support it. It should be just below ICELAKE_SERVER. https://github.com/gcc-mirror/gcc/blob/master/libgcc/config/i386/cpuinfo.h

[GGanesh]

Addressed the comments from Craig Topper

[craig.topper]

I think you uploaded the clang patch into the llvm review?

[craig.topper]

LGTM

include/llvm/Support/X86TargetParser.def

 X86_CPU_SUBTYPE_COMPAT("amdfam10",       AMDFAM10H_BARCELONA,         "barcelona")
 X86_CPU_SUBTYPE_COMPAT("amdfam10",       AMDFAM10H_SHANGHAI,          "shanghai")
 X86_CPU_SUBTYPE_COMPAT("amdfam10",       AMDFAM10H_ISTANBUL,          "istanbul")
 X86_CPU_SUBTYPE_COMPAT("bdver1",         AMDFAM15H_BDVER1,            "bdver1")
 X86_CPU_SUBTYPE_COMPAT("bdver2",         AMDFAM15H_BDVER2,            "bdver2")
 X86_CPU_SUBTYPE_COMPAT("bdver3",         AMDFAM15H_BDVER3,            "bdver3")
 X86_CPU_SUBTYPE_COMPAT("bdver4",         AMDFAM15H_BDVER4,            "bdver4")
 X86_CPU_SUBTYPE_COMPAT("znver1",         AMDFAM17H_ZNVER1,            "znver1")
+X86_CPU_SUBTYPE_COMPAT("znver2",         AMDFAM17H_ZNVER2,            "znver2")
 X86_CPU_SUBTYPE_COMPAT("ivybridge",      INTEL_COREI7_IVYBRIDGE,      "ivybridge")
 X86_CPU_SUBTYPE_COMPAT("haswell",        INTEL_COREI7_HASWELL,        "haswell")
 X86_CPU_SUBTYPE_COMPAT("broadwell",      INTEL_COREI7_BROADWELL,      "broadwell")
 X86_CPU_SUBTYPE_COMPAT("skylake",        INTEL_COREI7_SKYLAKE,        "skylake")
 X86_CPU_SUBTYPE_COMPAT("skylake-avx512", INTEL_COREI7_SKYLAKE_AVX512, "skylake-avx512")
 X86_CPU_SUBTYPE_COMPAT("cannonlake",     INTEL_COREI7_CANNONLAKE,     "cannonlake")
 X86_CPU_SUBTYPE_COMPAT("icelake-client", INTEL_COREI7_ICELAKE_CLIENT, "icelake-client")
 X86_CPU_SUBTYPE_COMPAT("icelake-server", INTEL_COREI7_ICELAKE_SERVER, "icelake-server")

lib/Support/Host.cpp

if (Model <= 0x0f) {
       break; // "bdver1"; 00h-0Fh: Bulldozer
     }
     break;
   case 22:
     *Type = X86::AMD_BTVER2;
     break; // "btver2"
   case 23:
     *Type = X86::AMDFAM17H;
+    if (Model >= 0x30 && Model <= 0x3f) {
+      *Subtype = X86::AMDFAM17H_ZNVER2;
+      break; // "znver2"; 30h-3fh: Zen2
+    }
+    if (Model <= 0x0f) {
       *Subtype = X86::AMDFAM17H_ZNVER1;
+      break; // "znver1"; 00h-0Fh: Zen1
+    }
     break;
   default:
     break; // "generic"
   }
 }
 
 static void getAvailableFeatures(unsigned ECX, unsigned EDX, unsigned MaxLeaf,
                                  unsigned *FeaturesOut, unsigned *Features2Out,

lib/Target/X86/X86.td

def : Proc<"bdver4", [
   FeatureFSGSBase,
   FeatureLAHFSAHF,
   FeatureFastBEXTR,
   FeatureFast11ByteNOP,
   FeatureMWAITX,
   FeatureMacroFusion
 ]>;
 
-// Znver1
-def: ProcessorModel<"znver1", Znver1Model, [
+// AMD Zen Processors common ISAs
+def ZNFeatures : ProcessorFeatures<[], [
   FeatureADX,
   FeatureAES,
   FeatureAVX2,
   FeatureBMI,
   FeatureBMI2,
   FeatureCLFLUSHOPT,
-  FeatureCLZERO,
   FeatureCMOV,
   Feature64Bit,
   FeatureCMPXCHG16B,
   FeatureF16C,
   FeatureFMA,
   FeatureFSGSBase,
   FeatureFXSR,
   FeatureNOPL,
def ZNFeatures : ProcessorFeatures<[], [
   FeatureSSE4A,
   FeatureSlowSHLD,
   FeatureX87,
   FeatureXSAVE,
   FeatureXSAVEC,
   FeatureXSAVEOPT,
   FeatureXSAVES]>;
 
+class Znver1Proc<string Name> : ProcModel<Name, Znver1Model,
+                                         ZNFeatures.Value, [
+  FeatureCLZERO
+]>;
+def : Znver1Proc<"znver1">;
+
+class Znver2Proc<string Name> : ProcModel<Name, Znver1Model,
+                                         ZNFeatures.Value, [
+  FeatureCLZERO,
+  FeatureCLWB,
+  FeatureRDPID,
+  FeatureWBNOINVD
+]>;
+def : Znver2Proc<"znver2">;
+
 def : Proc<"geode",           [FeatureX87, FeatureSlowUAMem16, Feature3DNowA]>;
 
 def : Proc<"winchip-c6",      [FeatureX87, FeatureSlowUAMem16, FeatureMMX]>;
 def : Proc<"winchip2",        [FeatureX87, FeatureSlowUAMem16, Feature3DNow]>;
 def : Proc<"c3",              [FeatureX87, FeatureSlowUAMem16, Feature3DNow]>;
 def : Proc<"c3-2",            [FeatureX87, FeatureSlowUAMem16, FeatureMMX,
                                FeatureSSE1, FeatureFXSR, FeatureCMOV]>;
 

test/CodeGen/X86/cpus-amd.ll

 ; RUN: llc < %s -o /dev/null -mtriple=x86_64-unknown-unknown -mcpu=barcelona 2>&1 | FileCheck %s --check-prefix=CHECK-NO-ERROR --allow-empty
 ; RUN: llc < %s -o /dev/null -mtriple=x86_64-unknown-unknown -mcpu=bdver1 2>&1 | FileCheck %s --check-prefix=CHECK-NO-ERROR --allow-empty
 ; RUN: llc < %s -o /dev/null -mtriple=x86_64-unknown-unknown -mcpu=bdver2 2>&1 | FileCheck %s --check-prefix=CHECK-NO-ERROR --allow-empty
 ; RUN: llc < %s -o /dev/null -mtriple=x86_64-unknown-unknown -mcpu=bdver3 2>&1 | FileCheck %s --check-prefix=CHECK-NO-ERROR --allow-empty
 ; RUN: llc < %s -o /dev/null -mtriple=x86_64-unknown-unknown -mcpu=bdver4 2>&1 | FileCheck %s --check-prefix=CHECK-NO-ERROR --allow-empty
 ; RUN: llc < %s -o /dev/null -mtriple=x86_64-unknown-unknown -mcpu=btver1 2>&1 | FileCheck %s --check-prefix=CHECK-NO-ERROR --allow-empty
 ; RUN: llc < %s -o /dev/null -mtriple=x86_64-unknown-unknown -mcpu=btver2 2>&1 | FileCheck %s --check-prefix=CHECK-NO-ERROR --allow-empty
 ; RUN: llc < %s -o /dev/null -mtriple=x86_64-unknown-unknown -mcpu=znver1 2>&1 | FileCheck %s --check-prefix=CHECK-NO-ERROR --allow-empty
+; RUN: llc < %s -o /dev/null -mtriple=x86_64-unknown-unknown -mcpu=znver2 2>&1 | FileCheck %s --check-prefix=CHECK-NO-ERROR --allow-empty
 
 define void @foo() {
   ret void
 }

test/CodeGen/X86/lzcnt-zext-cmp.ll

 ; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py
 ; Test patterns which generates lzcnt instructions.
 ; Eg: zext(or(setcc(cmp), setcc(cmp))) -> shr(or(lzcnt, lzcnt))
 ; RUN: llc < %s -mtriple=x86_64-pc-linux -mcpu=btver2 | FileCheck --check-prefix=ALL --check-prefix=FASTLZCNT %s
 ; RUN: llc < %s -mtriple=x86_64-pc-linux -mcpu=btver2 -mattr=-fast-lzcnt | FileCheck --check-prefix=ALL --check-prefix=NOFASTLZCNT %s
 ; RUN: llc < %s -mtriple=x86_64-pc-linux -mcpu=znver1 | FileCheck --check-prefix=ALL --check-prefix=FASTLZCNT %s
 ; RUN: llc < %s -mtriple=x86_64-pc-linux -mcpu=znver1 -mattr=-fast-lzcnt | FileCheck --check-prefix=ALL --check-prefix=NOFASTLZCNT %s
+; RUN: llc < %s -mtriple=x86_64-pc-linux -mcpu=znver2 | FileCheck --check-prefix=ALL --check-prefix=FASTLZCNT %s
+; RUN: llc < %s -mtriple=x86_64-pc-linux -mcpu=znver2 -mattr=-fast-lzcnt | FileCheck --check-prefix=ALL --check-prefix=NOFASTLZCNT %s
 
 ; Test one 32-bit input, output is 32-bit, no transformations expected.
 define i32 @test_zext_cmp0(i32 %a) {
 ; ALL-LABEL: test_zext_cmp0:
 ; ALL:       # %bb.0: # %entry
 ; ALL-NEXT:    xorl %eax, %eax
 ; ALL-NEXT:    testl %edi, %edi
 ; ALL-NEXT:    sete %al

test/CodeGen/X86/slow-unaligned-mem.ll

 ; RUN: llc < %s -mtriple=i386-unknown-unknown -mcpu=barcelona     2>&1 | FileCheck %s --check-prefix=FAST
 ; RUN: llc < %s -mtriple=i386-unknown-unknown -mcpu=btver1        2>&1 | FileCheck %s --check-prefix=FAST
 ; RUN: llc < %s -mtriple=i386-unknown-unknown -mcpu=btver2        2>&1 | FileCheck %s --check-prefix=FAST
 ; RUN: llc < %s -mtriple=i386-unknown-unknown -mcpu=bdver1        2>&1 | FileCheck %s --check-prefix=FAST
 ; RUN: llc < %s -mtriple=i386-unknown-unknown -mcpu=bdver2        2>&1 | FileCheck %s --check-prefix=FAST
 ; RUN: llc < %s -mtriple=i386-unknown-unknown -mcpu=bdver3        2>&1 | FileCheck %s --check-prefix=FAST
 ; RUN: llc < %s -mtriple=i386-unknown-unknown -mcpu=bdver4        2>&1 | FileCheck %s --check-prefix=FAST
 ; RUN: llc < %s -mtriple=i386-unknown-unknown -mcpu=znver1        2>&1 | FileCheck %s --check-prefix=FAST
+; RUN: llc < %s -mtriple=i386-unknown-unknown -mcpu=znver2        2>&1 | FileCheck %s --check-prefix=FAST
 
 ; Other chips with slow unaligned memory accesses
 
 ; RUN: llc < %s -mtriple=i386-unknown-unknown -mcpu=c3-2          2>&1 | FileCheck %s --check-prefix=SLOW
 
 ; Verify that the slow/fast unaligned memory attribute is set correctly for each CPU model.
 ; Slow chips use 4-byte stores. Fast chips with SSE or later use something other than 4-byte stores.
 ; Chips that don't have SSE use 4-byte stores either way, so they're not tested.

test/CodeGen/X86/x86-64-double-shifts-var.ll

 ; RUN: llc < %s -mtriple=x86_64-- -mcpu=k8 | FileCheck %s
 ; RUN: llc < %s -mtriple=x86_64-- -mcpu=opteron | FileCheck %s
 ; RUN: llc < %s -mtriple=x86_64-- -mcpu=athlon64 | FileCheck %s
 ; RUN: llc < %s -mtriple=x86_64-- -mcpu=athlon-fx | FileCheck %s
 ; RUN: llc < %s -mtriple=x86_64-- -mcpu=k8-sse3 | FileCheck %s
 ; RUN: llc < %s -mtriple=x86_64-- -mcpu=opteron-sse3 | FileCheck %s
 ; RUN: llc < %s -mtriple=x86_64-- -mcpu=athlon64-sse3 | FileCheck %s
 ; RUN: llc < %s -mtriple=x86_64-- -mcpu=amdfam10 | FileCheck %s
 ; RUN: llc < %s -mtriple=x86_64-- -mcpu=btver1 | FileCheck %s
 ; RUN: llc < %s -mtriple=x86_64-- -mcpu=btver2 | FileCheck %s
 ; RUN: llc < %s -mtriple=x86_64-- -mcpu=bdver1 | FileCheck %s
 ; RUN: llc < %s -mtriple=x86_64-- -mcpu=bdver2 | FileCheck %s
 ; RUN: llc < %s -mtriple=x86_64-- -mcpu=bdver3 | FileCheck %s
 ; RUN: llc < %s -mtriple=x86_64-- -mcpu=bdver4 | FileCheck %s
 ; RUN: llc < %s -mtriple=x86_64-- -mcpu=znver1 | FileCheck %s
+; RUN: llc < %s -mtriple=x86_64-- -mcpu=znver2 | FileCheck %s
 
-; Verify that for the X86_64 processors that are known to have poor latency 
+; Verify that for the X86_64 processors that are known to have poor latency
 ; double precision shift instructions we do not generate 'shld' or 'shrd'
 ; instructions.
 
 ;uint64_t lshift(uint64_t a, uint64_t b, int c)
 ;{
 ;    return (a << c) | (b >> (64-c));
 ;}
 
 define i64 @lshift(i64 %a, i64 %b, i32 %c) nounwind readnone {
 entry:
-; CHECK-NOT: shld 
+; CHECK-NOT: shld
   %sh_prom = zext i32 %c to i64
   %shl = shl i64 %a, %sh_prom
   %sub = sub nsw i32 64, %c
   %sh_prom1 = zext i32 %sub to i64
   %shr = lshr i64 %b, %sh_prom1
   %or = or i64 %shr, %shl
   ret i64 %or
 }