[LLVM] Please could you review my patches to add Cortex-A72 support to LLVM.

[rs]

Summary

The patches add support to LLVM for targeting the Cortex-A72 processor. Currently, Cortex-A72 is modelled as an Cortex-A57 except the fp load balancing pass isn't enabled for Cortex-A72 as it's not profitable to have it enabled for this core.

Tests for Cortex-A72 are also added.

[rengolin]

LGTM, thanks! Please commit this one before the Clang patch.

cheers,
--renato

lib/Target/AArch64/AArch64.td

94

I think this is ok until we know more about the core.

[rs]

In D7403#118268, @rengolin wrote:

LGTM, thanks! Please commit this one before the Clang patch.

cheers,
--renato

Again thanks for the reviews.

I don't have commit access so could you please commit the llvm and the clang patches?

[rengolin]

Sure

[rengolin]

r228140

lib/Target/AArch64/AArch64.td

def ProcCyclone : SubtargetFeature<"cyclone", "ARMProcFamily", "Cyclone",
                                    FeatureZCRegMove, FeatureZCZeroing]>;
 
 def : ProcessorModel<"generic", NoSchedModel, [FeatureFPARMv8,
                                               FeatureNEON,
                                               FeatureCRC]>;
 
 def : ProcessorModel<"cortex-a53", CortexA53Model, [ProcA53]>;
 def : ProcessorModel<"cortex-a57", CortexA57Model, [ProcA57]>;
+// FIXME: Cortex-A72 is currently modelled as an Cortex-A57.
+def : ProcessorModel<"cortex-a72", CortexA57Model, [ProcA57]>;
 def : ProcessorModel<"cyclone", CycloneModel, [ProcCyclone]>;
 
 //===----------------------------------------------------------------------===//
 // Assembly parser
 //===----------------------------------------------------------------------===//
 
 def GenericAsmParserVariant : AsmParserVariant {
   int Variant = 0;

lib/Target/ARM/ARM.td

 
 // V8 Processors
 def : ProcNoItin<"cortex-a53",      [ProcA53, HasV8Ops, FeatureAClass,
                                     FeatureDB, FeatureFPARMv8,
                                     FeatureNEON, FeatureDSPThumb2]>;
 def : ProcNoItin<"cortex-a57",      [ProcA57, HasV8Ops, FeatureAClass,
                                     FeatureDB, FeatureFPARMv8,
                                     FeatureNEON, FeatureDSPThumb2]>;
+// FIXME: Cortex-A72 is currently modelled as an Cortex-A57.
+def : ProcNoItin<"cortex-a72",      [ProcA57, HasV8Ops, FeatureAClass,
+                                    FeatureDB, FeatureFPARMv8,
+                                    FeatureNEON, FeatureDSPThumb2]>;
 
 // Cyclone is very similar to swift
 def : ProcessorModel<"cyclone",     SwiftModel,
                                     [ProcSwift, HasV8Ops, HasV7Ops,
                                      FeatureCrypto, FeatureFPARMv8,
                                      FeatureDB,FeatureDSPThumb2,
                                      FeatureHasRAS, FeatureZCZeroing]>;
 

test/CodeGen/AArch64/cpus.ll

 ; This tests that llc accepts all valid AArch64 CPUs
 
 
 ; RUN: llc < %s -mtriple=arm64-unknown-unknown -mcpu=generic 2>&1 | FileCheck %s
 ; RUN: llc < %s -mtriple=arm64-unknown-unknown -mcpu=cortex-a53 2>&1 | FileCheck %s
 ; RUN: llc < %s -mtriple=arm64-unknown-unknown -mcpu=cortex-a57 2>&1 | FileCheck %s
+; RUN: llc < %s -mtriple=arm64-unknown-unknown -mcpu=cortex-a72 2>&1 | FileCheck %s
 ; RUN: llc < %s -mtriple=arm64-unknown-unknown -mcpu=invalidcpu 2>&1 | FileCheck %s --check-prefix=INVALID
 
 ; CHECK-NOT: {{.*}}  is not a recognized processor for this target
 ; INVALID: {{.*}}  is not a recognized processor for this target
 
 define i32 @f(i64 %z) {
   ret i32 0
 }

test/CodeGen/AArch64/remat.ll

 ; RUN: llc -mtriple=aarch64-linux-gnuabi -mcpu=cortex-a57 -o - %s | FileCheck %s
 ; RUN: llc -mtriple=aarch64-linux-gnuabi -mcpu=cortex-a53 -o - %s | FileCheck %s
+; RUN: llc -mtriple=aarch64-linux-gnuabi -mcpu=cortex-a72 -o - %s | FileCheck %s
 
 %X = type { i64, i64, i64 }
 declare void @f(%X*)
 define void @t() {
 entry:
   %tmp = alloca %X
   call void @f(%X* %tmp)
 ; CHECK: add x0, sp, #8
 ; CHECK-NEXT-NOT: mov
   call void @f(%X* %tmp)               
 ; CHECK: add x0, sp, #8
 ; CHECK-NEXT-NOT: mov
   ret void 
 }

test/CodeGen/ARM/build-attributes.ll

 ; RUN: llc < %s -mtriple=armv7r-linux-gnueabi -mcpu=cortex-r5  -enable-unsafe-fp-math -disable-fp-elim -enable-no-infs-fp-math -enable-no-nans-fp-math -fp-contract=fast | FileCheck %s --check-prefix=CORTEX-R5-FAST
 ; RUN: llc < %s -mtriple=armv7r-linux-gnueabi -mcpu=cortex-r5 -enable-sign-dependent-rounding-fp-math | FileCheck %s --check-prefix=DYN-ROUNDING
 ; RUN: llc < %s -mtriple=armv8-linux-gnueabi -mcpu=cortex-a53 | FileCheck %s --check-prefix=CORTEX-A53
 ; RUN: llc < %s -mtriple=armv8-linux-gnueabi -mcpu=cortex-a53  -enable-unsafe-fp-math -disable-fp-elim -enable-no-infs-fp-math -enable-no-nans-fp-math -fp-contract=fast | FileCheck %s --check-prefix=CORTEX-A53-FAST
 ; RUN: llc < %s -mtriple=armv8-linux-gnueabi -mcpu=cortex-a53 -enable-sign-dependent-rounding-fp-math | FileCheck %s --check-prefix=DYN-ROUNDING
 ; RUN: llc < %s -mtriple=armv8-linux-gnueabi -mcpu=cortex-a57 | FileCheck %s --check-prefix=CORTEX-A57
 ; RUN: llc < %s -mtriple=armv8-linux-gnueabi -mcpu=cortex-a57  -enable-unsafe-fp-math -disable-fp-elim -enable-no-infs-fp-math -enable-no-nans-fp-math -fp-contract=fast | FileCheck %s --check-prefix=CORTEX-A57-FAST
 ; RUN: llc < %s -mtriple=armv8-linux-gnueabi -mcpu=cortex-a57 -enable-sign-dependent-rounding-fp-math | FileCheck %s --check-prefix=DYN-ROUNDING
+; RUN: llc < %s -mtriple=armv8-linux-gnueabi -mcpu=cortex-a72 | FileCheck %s --check-prefix=CORTEX-A72
+; RUN: llc < %s -mtriple=armv8-linux-gnueabi -mcpu=cortex-a72  -enable-unsafe-fp-math -disable-fp-elim -enable-no-infs-fp-math -enable-no-nans-fp-math -fp-contract=fast | FileCheck %s --check-prefix=CORTEX-A72-FAST
+; RUN: llc < %s -mtriple=armv8-linux-gnueabi -mcpu=cortex-a72 -enable-sign-dependent-rounding-fp-math | FileCheck %s --check-prefix=DYN-ROUNDING
 ; RUN: llc < %s -mtriple=armv7-none-linux-gnueabi -mcpu=cortex-a7 | FileCheck %s  --check-prefix=CORTEX-A7-CHECK
 ; RUN: llc < %s -mtriple=armv7-none-linux-gnueabi -mcpu=cortex-a7  -enable-unsafe-fp-math -disable-fp-elim -enable-no-infs-fp-math -enable-no-nans-fp-math -fp-contract=fast | FileCheck %s  --check-prefix=CORTEX-A7-CHECK-FAST
 ; RUN: llc < %s -mtriple=armv7-none-linux-gnueabi -mcpu=cortex-a7 -mattr=-vfp2,-vfp3,-vfp4,-neon | FileCheck %s --check-prefix=CORTEX-A7-NOFPU
 ; RUN: llc < %s -mtriple=armv7-none-linux-gnueabi -mcpu=cortex-a7 -mattr=-vfp2,-vfp3,-vfp4,-neon  -enable-unsafe-fp-math -disable-fp-elim -enable-no-infs-fp-math -enable-no-nans-fp-math -fp-contract=fast | FileCheck %s --check-prefix=CORTEX-A7-NOFPU-FAST
 ; RUN: llc < %s -mtriple=armv7-none-linux-gnueabi -mcpu=cortex-a7 -mattr=+vfp4,-neon | FileCheck %s --check-prefix=CORTEX-A7-FPUV4
 ; RUN: llc < %s -mtriple=armv7-none-linux-gnueabi -mcpu=cortex-a7 -enable-sign-dependent-rounding-fp-math | FileCheck %s --check-prefix=DYN-ROUNDING
 ; RUN: llc < %s -mtriple=armv7-none-linux-gnueabi -mcpu=cortex-a7 -mattr=+vfp4,-neon  -enable-unsafe-fp-math -disable-fp-elim -enable-no-infs-fp-math -enable-no-nans-fp-math -fp-contract=fast | FileCheck %s --check-prefix=CORTEX-A7-FPUV4-FAST
 ; RUN: llc < %s -mtriple=armv7-none-linux-gnueabi -mcpu=cortex-a7 -mattr=+vfp4,,+d16,-neon | FileCheck %s --check-prefix=CORTEX-A7-FPUV4
 ; RUN: llc < %s -mtriple=arm-none-linux-gnueabi -relocation-model=pic | FileCheck %s --check-prefix=RELOC-PIC
 ; RUN: llc < %s -mtriple=arm-none-linux-gnueabi -relocation-model=static | FileCheck %s --check-prefix=RELOC-OTHER
 ; RUN: llc < %s -mtriple=arm-none-linux-gnueabi -relocation-model=default | FileCheck %s --check-prefix=RELOC-OTHER
 ; RUN: llc < %s -mtriple=arm-none-linux-gnueabi -relocation-model=dynamic-no-pic | FileCheck %s --check-prefix=RELOC-OTHER
 ; RUN: llc < %s -mtriple=arm-none-linux-gnueabi | FileCheck %s --check-prefix=RELOC-OTHER
 ; RUN: llc < %s -mtriple=arm-none-linux-gnueabi | FileCheck %s --check-prefix=PCS-R9-USE
 ; RUN: llc < %s -mtriple=arm-none-linux-gnueabi -arm-reserve-r9 | FileCheck %s --check-prefix=PCS-R9-RESERVE
 
 ; ARMv8a (AArch32)
 ; RUN: llc < %s -mtriple=armv8-none-linux-gnueabi -mcpu=cortex-a57 -arm-no-strict-align | FileCheck %s --check-prefix=NO-STRICT-ALIGN
 ; RUN: llc < %s -mtriple=armv8-none-linux-gnueabi -mcpu=cortex-a57 -arm-strict-align | FileCheck %s --check-prefix=STRICT-ALIGN
 ; RUN: llc < %s -mtriple=armv8-none-linux-gnueabi -mcpu=cortex-a57 | FileCheck %s --check-prefix=NO-STRICT-ALIGN
+; RUN: llc < %s -mtriple=armv8-none-linux-gnueabi -mcpu=cortex-a72 -arm-no-strict-align | FileCheck %s --check-prefix=NO-STRICT-ALIGN
+; RUN: llc < %s -mtriple=armv8-none-linux-gnueabi -mcpu=cortex-a72 -arm-strict-align | FileCheck %s --check-prefix=STRICT-ALIGN
+; RUN: llc < %s -mtriple=armv8-none-linux-gnueabi -mcpu=cortex-a72 | FileCheck %s --check-prefix=NO-STRICT-ALIGN
 ; ARMv7a
 ; RUN: llc < %s -mtriple=armv7-none-linux-gnueabi -mcpu=cortex-a7 -arm-no-strict-align | FileCheck %s --check-prefix=NO-STRICT-ALIGN
 ; RUN: llc < %s -mtriple=armv7-none-linux-gnueabi -mcpu=cortex-a7 -arm-strict-align | FileCheck %s --check-prefix=STRICT-ALIGN
 ; RUN: llc < %s -mtriple=armv7-none-linux-gnueabi -mcpu=cortex-a7 | FileCheck %s --check-prefix=NO-STRICT-ALIGN
 ; ARMv7r
 ; RUN: llc < %s -mtriple=armv7r-none-linux-gnueabi -mcpu=cortex-r5 -arm-no-strict-align | FileCheck %s --check-prefix=NO-STRICT-ALIGN
 ; RUN: llc < %s -mtriple=armv7r-none-linux-gnueabi -mcpu=cortex-r5 -arm-strict-align | FileCheck %s --check-prefix=STRICT-ALIGN
 ; RUN: llc < %s -mtriple=armv7r-none-linux-gnueabi -mcpu=cortex-r5 | FileCheck %s --check-prefix=NO-STRICT-ALIGN

 
 ; CORTEX-A57-FAST-NOT:   .eabi_attribute 19
 ;; The A57 has the ARMv8 FP unit, which always flushes preserving sign.
 ; CORTEX-A57-FAST:  .eabi_attribute 20, 2
 ; CORTEX-A57-FAST-NOT:  .eabi_attribute 21
 ; CORTEX-A57-FAST-NOT:  .eabi_attribute 22
 ; CORTEX-A57-FAST:  .eabi_attribute 23, 1
 
+; CORTEX-A72:  .cpu cortex-a72
+; CORTEX-A72:  .eabi_attribute 6, 14
+; CORTEX-A72:  .eabi_attribute 7, 65
+; CORTEX-A72:  .eabi_attribute 8, 1
+; CORTEX-A72:  .eabi_attribute 9, 2
+; CORTEX-A72:  .fpu crypto-neon-fp-armv8
+; CORTEX-A72:  .eabi_attribute 12, 3
+; CORTEX-A72-NOT:   .eabi_attribute 19
+;; We default to IEEE 754 compliance
+; CORTEX-A72:  .eabi_attribute 20, 1
+; CORTEX-A72:  .eabi_attribute 21, 1
+; CORTEX-A72-NOT:  .eabi_attribute 22
+; CORTEX-A72:  .eabi_attribute 23, 3
+; CORTEX-A72:  .eabi_attribute 24, 1
+; CORTEX-A72:  .eabi_attribute 25, 1
+; CORTEX-A72-NOT:  .eabi_attribute 27
+; CORTEX-A72-NOT:  .eabi_attribute 28
+; CORTEX-A72:  .eabi_attribute 36, 1
+; CORTEX-A72:  .eabi_attribute 38, 1
+; CORTEX-A72:  .eabi_attribute 42, 1
+; CORTEX-A72-NOT:  .eabi_attribute 44
+; CORTEX-A72:  .eabi_attribute 68, 3
+
+; CORTEX-A72-FAST-NOT:   .eabi_attribute 19
+;; The A72 has the ARMv8 FP unit, which always flushes preserving sign.
+; CORTEX-A72-FAST:  .eabi_attribute 20, 2
+; CORTEX-A72-FAST-NOT:  .eabi_attribute 21
+; CORTEX-A72-FAST-NOT:  .eabi_attribute 22
+; CORTEX-A72-FAST:  .eabi_attribute 23, 1
+
 ; RELOC-PIC:  .eabi_attribute 15, 1
 ; RELOC-PIC:  .eabi_attribute 16, 1
 ; RELOC-PIC:  .eabi_attribute 17, 2
 ; RELOC-OTHER:  .eabi_attribute 17, 1
 
 ; PCS-R9-USE:  .eabi_attribute 14, 0
 ; PCS-R9-RESERVE:  .eabi_attribute 14, 3
 
 define i32 @f(i64 %z) {
     ret i32 0
 }