This is an archive of the discontinued LLVM Phabricator instance.

Differential D18897

[X86] Restrict max long nop length for Lakemont
ClosedPublic

Authored by aturetsk on Apr 8 2016, 4:00 AM.

Details

Reviewers
bruno
nadav
echristo
Commits
rG9df334c28e35: [X86] Restrict max long nop length for Lakemont.
rL265924: [X86] Restrict max long nop length for Lakemont.
Summary

Restrict the max length of long nops for Lakemont to 7. Experiments on MCU benchmarks (Dhrystone, Coremark) show that this is the most optimal length.

Diff Detail

Event Timeline

aturetsk updated this revision to Diff 53014.Apr 8 2016, 4:00 AM
aturetsk retitled this revision from to [X86] Restrict max long nop length for Lakemont.
aturetsk updated this object.
aturetsk added reviewers: nadav, bruno, echristo.
aturetsk added subscribers: llvm-commits, zinovy.nis.
bruno accepted this revision.Apr 8 2016, 9:32 AM
bruno edited edge metadata.

LGTM, one comment below

lib/Target/X86/MCTargetDesc/X86AsmBackend.cpp
75

clang-format this one.

This revision is now accepted and ready to land.Apr 8 2016, 9:32 AM
Closed by commit rL265924: [X86] Restrict max long nop length for Lakemont. (authored by aturetsk). · Explain WhyApr 11 2016, 3:13 AM
This revision was automatically updated to reflect the committed changes.
aturetsk added a comment.Apr 11 2016, 3:14 AM

Thanks for the review. The formatting was fixed.

Revision Contents

PathSize
lib/
Target/
X86/
MCTargetDesc/
2 lines
test/
MC/
X86/
20 lines

Diff 53014

lib/Target/X86/MCTargetDesc/X86AsmBackend.cpp

Show First 20 Lines • Show All 66 Lines • ▼ Show 20 Lines
}; };
class X86AsmBackend : public MCAsmBackend { class X86AsmBackend : public MCAsmBackend {
const StringRef CPU; const StringRef CPU;
bool HasNopl; bool HasNopl;
const uint64_t MaxNopLength; const uint64_t MaxNopLength;
public: public:
X86AsmBackend(const Target &T, StringRef CPU) X86AsmBackend(const Target &T, StringRef CPU)
: MCAsmBackend(), CPU(CPU), MaxNopLength(CPU == "slm" ? 7 : 15) { : MCAsmBackend(), CPU(CPU), MaxNopLength((CPU == "slm" || CPU == "lakemont") ? 7 : 15) {
brunoUnsubmitted
Not Done
ReplyInline Actions

clang-format this one.

bruno: clang-format this one.
HasNopl = CPU != "generic" && CPU != "i386" && CPU != "i486" && HasNopl = CPU != "generic" && CPU != "i386" && CPU != "i486" &&
CPU != "i586" && CPU != "pentium" && CPU != "pentium-mmx" && CPU != "i586" && CPU != "pentium" && CPU != "pentium-mmx" &&
CPU != "i686" && CPU != "k6" && CPU != "k6-2" && CPU != "k6-3" && CPU != "i686" && CPU != "k6" && CPU != "k6-2" && CPU != "k6-3" &&
CPU != "geode" && CPU != "winchip-c6" && CPU != "winchip2" && CPU != "geode" && CPU != "winchip-c6" && CPU != "winchip2" &&
CPU != "c3" && CPU != "c3-2"; CPU != "c3" && CPU != "c3-2";
} }
unsigned getNumFixupKinds() const override { unsigned getNumFixupKinds() const override {
▲ Show 20 Lines • Show All 763 Lines • Show Last 20 Lines

test/MC/X86/x86_long_nop.s

# RUN: llvm-mc -filetype=obj -arch=x86 -triple=x86_64-pc-linux-gnu %s | llvm-objdump -d -no-show-raw-insn - | FileCheck %s # RUN: llvm-mc -filetype=obj -arch=x86 -triple=x86_64-pc-linux-gnu %s | llvm-objdump -d -no-show-raw-insn - | FileCheck %s
# RUN: llvm-mc -filetype=obj -arch=x86 -triple=i686-pc-linux-gnu %s | llvm-objdump -d -no-show-raw-insn - | FileCheck %s # RUN: llvm-mc -filetype=obj -arch=x86 -triple=i686-pc-linux-gnu %s | llvm-objdump -d -no-show-raw-insn - | FileCheck %s
# RUN: llvm-mc -filetype=obj -arch=x86 -triple=x86_64-apple-darwin10.0 %s | llvm-objdump -d -no-show-raw-insn - | FileCheck %s # RUN: llvm-mc -filetype=obj -arch=x86 -triple=x86_64-apple-darwin10.0 %s | llvm-objdump -d -no-show-raw-insn - | FileCheck %s
# RUN: llvm-mc -filetype=obj -arch=x86 -triple=i686-apple-darwin8 %s | llvm-objdump -d -no-show-raw-insn - | FileCheck %s # RUN: llvm-mc -filetype=obj -arch=x86 -triple=i686-apple-darwin8 %s | llvm-objdump -d -no-show-raw-insn - | FileCheck %s
# RUN: llvm-mc -filetype=obj -arch=x86 -triple=i686-pc-linux-gnu -mcpu=slm %s | llvm-objdump -d -no-show-raw-insn - | FileCheck --check-prefix=SLM %s # RUN: llvm-mc -filetype=obj -arch=x86 -triple=i686-pc-linux-gnu -mcpu=slm %s | llvm-objdump -d -no-show-raw-insn - | FileCheck --check-prefix=LNOP7 %s
# RUN: llvm-mc -filetype=obj -arch=x86 -triple=i686-pc-linux-gnu -mcpu=lakemont %s | llvm-objdump -d -no-show-raw-insn - | FileCheck --check-prefix=LNOP7 %s
# Ensure alignment directives also emit sequences of 15-byte NOPs on processors # Ensure alignment directives also emit sequences of 15-byte NOPs on processors
# capable of using long NOPs. # capable of using long NOPs.
inc %eax inc %eax
.p2align 5 .p2align 5
inc %eax inc %eax
# CHECK: 0: inc # CHECK: 0: inc
# CHECK-NEXT: 1: nop # CHECK-NEXT: 1: nop
# CHECK-NEXT: 10: nop # CHECK-NEXT: 10: nop
# CHECK-NEXT: 1f: nop # CHECK-NEXT: 1f: nop
# CHECK-NEXT: 20: inc # CHECK-NEXT: 20: inc
# On Silvermont we emit only 7 byte NOPs since longer NOPs are not profitable # On Silvermont and Lakemont we emit only 7 byte NOPs since longer NOPs
# SLM: 0: inc # are not profitable.
# SLM-NEXT: 1: nop # LNOP7: 0: inc
# SLM-NEXT: 8: nop # LNOP7-NEXT: 1: nop
# SLM-NEXT: f: nop # LNOP7-NEXT: 8: nop
# SLM-NEXT: 16: nop # LNOP7-NEXT: f: nop
# SLM-NEXT: 1d: nop # LNOP7-NEXT: 16: nop
# SLM-NEXT: 20: inc # LNOP7-NEXT: 1d: nop
# LNOP7-NEXT: 20: inc