[clang] [llvm] [LoongArch] Support amcas[_db].{b/h/w/d} instructions. (PR #114189)
via cfe-commits
cfe-commits at lists.llvm.org
Tue Nov 26 00:35:55 PST 2024
https://github.com/tangaac updated https://github.com/llvm/llvm-project/pull/114189
>From 88daebae45de7ac225629066f10e8b9765f54e9b Mon Sep 17 00:00:00 2001
From: tangaac <tangyan01 at loongson.cn>
Date: Wed, 23 Oct 2024 17:22:25 +0800
Subject: [PATCH 1/3] Support amcas[_db].{b/h/w/d} instructions.
---
clang/include/clang/Driver/Options.td | 4 +
clang/lib/Basic/Targets/LoongArch.cpp | 7 +-
clang/lib/Basic/Targets/LoongArch.h | 2 +
.../lib/Driver/ToolChains/Arch/LoongArch.cpp | 9 +
clang/test/Driver/loongarch-march.c | 8 +-
clang/test/Driver/loongarch-mlamcas.c | 30 +
clang/test/Preprocessor/init-loongarch.c | 25 +-
.../TargetParser/LoongArchTargetParser.def | 3 +-
.../llvm/TargetParser/LoongArchTargetParser.h | 4 +
llvm/lib/Target/LoongArch/LoongArch.td | 9 +-
.../LoongArch/LoongArchISelLowering.cpp | 16 +-
.../Target/LoongArch/LoongArchInstrInfo.td | 24 +-
.../TargetParser/LoongArchTargetParser.cpp | 1 +
.../ir-instruction/atomic-cmpxchg.ll | 79 +-
.../ir-instruction/atomicrmw-lamcas.ll | 5341 +++++++++++++++++
15 files changed, 5542 insertions(+), 20 deletions(-)
create mode 100644 clang/test/Driver/loongarch-mlamcas.c
create mode 100644 llvm/test/CodeGen/LoongArch/ir-instruction/atomicrmw-lamcas.ll
diff --git a/clang/include/clang/Driver/Options.td b/clang/include/clang/Driver/Options.td
index 40fd48761928b3..3be446e12b30ea 100644
--- a/clang/include/clang/Driver/Options.td
+++ b/clang/include/clang/Driver/Options.td
@@ -5413,6 +5413,10 @@ def mlam_bh : Flag<["-"], "mlam-bh">, Group<m_loongarch_Features_Group>,
HelpText<"Enable amswap[_db].{b/h} and amadd[_db].{b/h}">;
def mno_lam_bh : Flag<["-"], "mno-lam-bh">, Group<m_loongarch_Features_Group>,
HelpText<"Disable amswap[_db].{b/h} and amadd[_db].{b/h}">;
+def mlamcas : Flag<["-"], "mlamcas">, Group<m_loongarch_Features_Group>,
+ HelpText<"Enable amcas[_db].{b/h/w/d}">;
+def mno_lamcas : Flag<["-"], "mno-lamcas">, Group<m_loongarch_Features_Group>,
+ HelpText<"Disable amcas[_db].{b/h/w/d}">;
def mld_seq_sa : Flag<["-"], "mld-seq-sa">, Group<m_loongarch_Features_Group>,
HelpText<"Do not generate load-load barrier instructions (dbar 0x700)">;
def mno_ld_seq_sa : Flag<["-"], "mno-ld-seq-sa">, Group<m_loongarch_Features_Group>,
diff --git a/clang/lib/Basic/Targets/LoongArch.cpp b/clang/lib/Basic/Targets/LoongArch.cpp
index 3f2d7317532aaf..145c32bd27525a 100644
--- a/clang/lib/Basic/Targets/LoongArch.cpp
+++ b/clang/lib/Basic/Targets/LoongArch.cpp
@@ -205,7 +205,7 @@ void LoongArchTargetInfo::getTargetDefines(const LangOptions &Opts,
// TODO: As more features of the V1.1 ISA are supported, a unified "v1.1"
// arch feature set will be used to include all sub-features belonging to
// the V1.1 ISA version.
- if (HasFeatureFrecipe && HasFeatureLAM_BH && HasFeatureLD_SEQ_SA)
+ if (HasFeatureFrecipe && HasFeatureLAM_BH && HasFeatureLAMCAS && HasFeatureLD_SEQ_SA)
Builder.defineMacro("__loongarch_arch",
Twine('"') + "la64v1.1" + Twine('"'));
else
@@ -239,6 +239,9 @@ void LoongArchTargetInfo::getTargetDefines(const LangOptions &Opts,
if (HasFeatureLAM_BH)
Builder.defineMacro("__loongarch_lam_bh", Twine(1));
+ if (HasFeatureLAMCAS)
+ Builder.defineMacro("__loongarch_lamcas", Twine(1));
+
if (HasFeatureLD_SEQ_SA)
Builder.defineMacro("__loongarch_ld_seq_sa", Twine(1));
@@ -320,6 +323,8 @@ bool LoongArchTargetInfo::handleTargetFeatures(
HasFeatureFrecipe = true;
else if (Feature == "+lam-bh")
HasFeatureLAM_BH = true;
+ else if (Feature == "+lamcas")
+ HasFeatureLAMCAS = true;
else if (Feature == "+ld-seq-sa")
HasFeatureLD_SEQ_SA = true;
}
diff --git a/clang/lib/Basic/Targets/LoongArch.h b/clang/lib/Basic/Targets/LoongArch.h
index e5eae7a8fcf677..59e2bdeff7de89 100644
--- a/clang/lib/Basic/Targets/LoongArch.h
+++ b/clang/lib/Basic/Targets/LoongArch.h
@@ -31,6 +31,7 @@ class LLVM_LIBRARY_VISIBILITY LoongArchTargetInfo : public TargetInfo {
bool HasFeatureLASX;
bool HasFeatureFrecipe;
bool HasFeatureLAM_BH;
+ bool HasFeatureLAMCAS;
bool HasFeatureLD_SEQ_SA;
public:
@@ -42,6 +43,7 @@ class LLVM_LIBRARY_VISIBILITY LoongArchTargetInfo : public TargetInfo {
HasFeatureLASX = false;
HasFeatureFrecipe = false;
HasFeatureLAM_BH = false;
+ HasFeatureLAMCAS = false;
HasFeatureLD_SEQ_SA = false;
LongDoubleWidth = 128;
LongDoubleAlign = 128;
diff --git a/clang/lib/Driver/ToolChains/Arch/LoongArch.cpp b/clang/lib/Driver/ToolChains/Arch/LoongArch.cpp
index 67b71a3ec623e4..caa645d0ff46be 100644
--- a/clang/lib/Driver/ToolChains/Arch/LoongArch.cpp
+++ b/clang/lib/Driver/ToolChains/Arch/LoongArch.cpp
@@ -275,6 +275,15 @@ void loongarch::getLoongArchTargetFeatures(const Driver &D,
Features.push_back("-lam-bh");
}
+ // Select lamcas feature determined by -m[no-]lamcas.
+ if (const Arg *A =
+ Args.getLastArg(options::OPT_mlamcas, options::OPT_mno_lamcas)) {
+ if (A->getOption().matches(options::OPT_mlamcas))
+ Features.push_back("+lamcas");
+ else
+ Features.push_back("-lamcas");
+ }
+
// Select ld-seq-sa feature determined by -m[no-]ld-seq-sa.
if (const Arg *A = Args.getLastArg(options::OPT_mld_seq_sa,
options::OPT_mno_ld_seq_sa)) {
diff --git a/clang/test/Driver/loongarch-march.c b/clang/test/Driver/loongarch-march.c
index c7091336f3bc80..b3a6557f6f231b 100644
--- a/clang/test/Driver/loongarch-march.c
+++ b/clang/test/Driver/loongarch-march.c
@@ -39,21 +39,21 @@
// CC1-LA64V1P1: "-target-cpu" "loongarch64"
// CC1-LA64V1P1-NOT: "-target-feature"
-// CC1-LA64V1P1: "-target-feature" "+64bit" "-target-feature" "+d" "-target-feature" "+lsx" "-target-feature" "+ual" "-target-feature" "+frecipe" "-target-feature" "+lam-bh" "-target-feature" "+ld-seq-sa"
+// CC1-LA64V1P1: "-target-feature" "+64bit" "-target-feature" "+d" "-target-feature" "+lsx" "-target-feature" "+ual" "-target-feature" "+frecipe" "-target-feature" "+lam-bh" "-target-feature" "+lamcas" "-target-feature" "+ld-seq-sa"
// CC1-LA64V1P1-NOT: "-target-feature"
// CC1-LA64V1P1: "-target-abi" "lp64d"
// CC1-LA664: "-target-cpu" "la664"
// CC1-LA664-NOT: "-target-feature"
-// CC1-LA664: "-target-feature" "+64bit" "-target-feature" "+f" "-target-feature" "+d" "-target-feature" "+lsx" "-target-feature" "+lasx" "-target-feature" "+ual" "-target-feature" "+frecipe" "-target-feature" "+lam-bh" "-target-feature" "+ld-seq-sa"
+// CC1-LA664: "-target-feature" "+64bit" "-target-feature" "+f" "-target-feature" "+d" "-target-feature" "+lsx" "-target-feature" "+lasx" "-target-feature" "+ual" "-target-feature" "+frecipe" "-target-feature" "+lam-bh" "-target-feature" "+lamcas" "-target-feature" "+ld-seq-sa"
// CC1-LA664-NOT: "-target-feature"
// CC1-LA664: "-target-abi" "lp64d"
// IR-LOONGARCH64: attributes #[[#]] ={{.*}}"target-cpu"="loongarch64" {{.*}}"target-features"="+64bit,+d,+f,+ual"
// IR-LA464: attributes #[[#]] ={{.*}}"target-cpu"="la464" {{.*}}"target-features"="+64bit,+d,+f,+lasx,+lsx,+ual"
// IR-LA64V1P0: attributes #[[#]] ={{.*}}"target-cpu"="loongarch64" {{.*}}"target-features"="+64bit,+d,+lsx,+ual"
-// IR-LA64V1P1: attributes #[[#]] ={{.*}}"target-cpu"="loongarch64" {{.*}}"target-features"="+64bit,+d,+frecipe,+lam-bh,+ld-seq-sa,+lsx,+ual"
-// IR-LA664: attributes #[[#]] ={{.*}}"target-cpu"="la664" {{.*}}"target-features"="+64bit,+d,+f,+frecipe,+lam-bh,+lasx,+ld-seq-sa,+lsx,+ual"
+// IR-LA64V1P1: attributes #[[#]] ={{.*}}"target-cpu"="loongarch64" {{.*}}"target-features"="+64bit,+d,+frecipe,+lam-bh,+lamcas,+ld-seq-sa,+lsx,+ual"
+// IR-LA664: attributes #[[#]] ={{.*}}"target-cpu"="la664" {{.*}}"target-features"="+64bit,+d,+f,+frecipe,+lam-bh,+lamcas,+ld-seq-sa,+lasx,+lsx,+ual"
int foo(void) {
return 3;
diff --git a/clang/test/Driver/loongarch-mlamcas.c b/clang/test/Driver/loongarch-mlamcas.c
new file mode 100644
index 00000000000000..2185a1a8115d66
--- /dev/null
+++ b/clang/test/Driver/loongarch-mlamcas.c
@@ -0,0 +1,30 @@
+/// Test -m[no]lamcas options.
+
+// RUN: %clang --target=loongarch64 -mlamcas -fsyntax-only %s -### 2>&1 | \
+// RUN: FileCheck %s --check-prefix=CC1-LAMCAS
+// RUN: %clang --target=loongarch64 -mno-lamcas -fsyntax-only %s -### 2>&1 | \
+// RUN: FileCheck %s --check-prefix=CC1-NO-LAMCAS
+// RUN: %clang --target=loongarch64 -mno-lamcas -mlamcas -fsyntax-only %s -### 2>&1 | \
+// RUN: FileCheck %s --check-prefix=CC1-LAMCAS
+// RUN: %clang --target=loongarch64 -mlamcas -mno-lamcas -fsyntax-only %s -### 2>&1 | \
+// RUN: FileCheck %s --check-prefix=CC1-NO-LAMCAS
+
+// RUN: %clang --target=loongarch64 -mlamcas -S -emit-llvm %s -o - | \
+// RUN: FileCheck %s --check-prefix=IR-LAMCAS
+// RUN: %clang --target=loongarch64 -mno-lamcas -S -emit-llvm %s -o - | \
+// RUN: FileCheck %s --check-prefix=IR-NO-LAMCAS
+// RUN: %clang --target=loongarch64 -mno-lamcas -mlamcas -S -emit-llvm %s -o - | \
+// RUN: FileCheck %s --check-prefix=IR-LAMCAS
+// RUN: %clang --target=loongarch64 -mlamcas -mno-lamcas -S -emit-llvm %s -o - | \
+// RUN: FileCheck %s --check-prefix=IR-NO-LAMCAS
+
+
+// CC1-LAMCAS: "-target-feature" "+lamcas"
+// CC1-NO-LAMCAS: "-target-feature" "-lamcas"
+
+// IR-LAMCAS: attributes #[[#]] ={{.*}}"target-features"="{{(.*,)?}}+lamcas{{(,.*)?}}"
+// IR-NO-LAMCAS: attributes #[[#]] ={{.*}}"target-features"="{{(.*,)?}}-lamcas{{(,.*)?}}"
+
+int foo(void) {
+ return 42;
+}
diff --git a/clang/test/Preprocessor/init-loongarch.c b/clang/test/Preprocessor/init-loongarch.c
index 0e3320f01b328c..aa8b1bbc1e7aea 100644
--- a/clang/test/Preprocessor/init-loongarch.c
+++ b/clang/test/Preprocessor/init-loongarch.c
@@ -798,7 +798,7 @@
// LA64-FPU0-LP64S-NOT: #define __loongarch_single_float
// LA64-FPU0-LP64S: #define __loongarch_soft_float 1
-/// Check __loongarch_arch{_tune/_frecipe/_lam_bh/_ld_seq_sa}.
+/// Check __loongarch_arch{_tune/_frecipe/_lam_bh/_lamcas/_ld_seq_sa}.
// RUN: %clang --target=loongarch64 -x c -E -dM %s -o - | \
// RUN: FileCheck --match-full-lines --check-prefix=ARCH-TUNE -DARCH=la64v1.0 -DTUNE=loongarch64 %s
@@ -823,11 +823,11 @@
// RUN: %clang --target=loongarch64 -x c -E -dM %s -o - -march=loongarch64 -Xclang -target-feature -Xclang +lsx | \
// RUN: FileCheck --match-full-lines --check-prefix=ARCH-TUNE -DARCH=la64v1.0 -DTUNE=loongarch64 %s
// RUN: %clang --target=loongarch64 -x c -E -dM %s -o - -march=la64v1.1 | \
-// RUN: FileCheck --match-full-lines --check-prefixes=ARCH-TUNE,FRECIPE,LAM-BH,LD-SEQ-SA -DARCH=la64v1.1 -DTUNE=loongarch64 %s
+// RUN: FileCheck --match-full-lines --check-prefixes=ARCH-TUNE,FRECIPE,LAM-BH,LAMCAS,LD-SEQ-SA -DARCH=la64v1.1 -DTUNE=loongarch64 %s
// RUN: %clang --target=loongarch64 -x c -E -dM %s -o - -march=la64v1.1 -Xclang -target-feature -Xclang -frecipe | \
-// RUN: FileCheck --match-full-lines --check-prefixes=ARCH-TUNE,LAM-BH,LD-SEQ-SA -DARCH=la64v1.0 -DTUNE=loongarch64 %s
+// RUN: FileCheck --match-full-lines --check-prefixes=ARCH-TUNE,LAM-BH,LAMCAS,lD-SEQ-SA -DARCH=la64v1.0 -DTUNE=loongarch64 %s
// RUN: %clang --target=loongarch64 -x c -E -dM %s -o - -march=la64v1.1 -Xclang -target-feature -Xclang -lsx | \
-// RUN: FileCheck --match-full-lines --check-prefixes=ARCH-TUNE,FRECIPE,LAM-BH,LD-SEQ-SA -DARCH=loongarch64 -DTUNE=loongarch64 %s
+// RUN: FileCheck --match-full-lines --check-prefixes=ARCH-TUNE,FRECIPE,LAM-BH,LAMCAS,LD-SEQ-SA -DARCH=loongarch64 -DTUNE=loongarch64 %s
// RUN: %clang --target=loongarch64 -x c -E -dM %s -o - -march=loongarch64 -Xclang -target-feature -Xclang +frecipe | \
// RUN: FileCheck --match-full-lines --check-prefixes=ARCH-TUNE,FRECIPE -DARCH=loongarch64 -DTUNE=loongarch64 %s
// RUN: %clang --target=loongarch64 -x c -E -dM %s -o - -march=loongarch64 -Xclang -target-feature -Xclang +lsx -Xclang -target-feature -Xclang +frecipe | \
@@ -835,11 +835,19 @@
// RUN: %clang --target=loongarch64 -x c -E -dM %s -o - -march=la64v1.0 -Xclang -target-feature -Xclang +lam-bh | \
// RUN: FileCheck --match-full-lines --check-prefixes=ARCH-TUNE,LAM-BH -DARCH=la64v1.0 -DTUNE=loongarch64 %s
// RUN: %clang --target=loongarch64 -x c -E -dM %s -o - -march=la64v1.1 -Xclang -target-feature -Xclang -lam-bh | \
-// RUN: FileCheck --match-full-lines --check-prefixes=ARCH-TUNE,FRECIPE,LD-SEQ-SA -DARCH=la64v1.0 -DTUNE=loongarch64 %s
+// RUN: FileCheck --match-full-lines --check-prefixes=ARCH-TUNE,FRECIPE,LAMCAS,LD-SEQ-SA -DARCH=la64v1.0 -DTUNE=loongarch64 %s
// RUN: %clang --target=loongarch64 -x c -E -dM %s -o - -march=loongarch64 -Xclang -target-feature -Xclang +lam-bh | \
// RUN: FileCheck --match-full-lines --check-prefixes=ARCH-TUNE,LAM-BH -DARCH=loongarch64 -DTUNE=loongarch64 %s
// RUN: %clang --target=loongarch64 -x c -E -dM %s -o - -march=loongarch64 -Xclang -target-feature -Xclang +lsx -Xclang -target-feature -Xclang +lam-bh | \
// RUN: FileCheck --match-full-lines --check-prefixes=ARCH-TUNE,LAM-BH -DARCH=la64v1.0 -DTUNE=loongarch64 %s
+// RUN: %clang --target=loongarch64 -x c -E -dM %s -o - -march=la64v1.0 -Xclang -target-feature -Xclang +lamcas | \
+// RUN: FileCheck --match-full-lines --check-prefixes=ARCH-TUNE,LAMCAS -DARCH=la64v1.0 -DTUNE=loongarch64 %s
+// RUN: %clang --target=loongarch64 -x c -E -dM %s -o - -march=la64v1.1 -Xclang -target-feature -Xclang -lamcas | \
+// RUN: FileCheck --match-full-lines --check-prefixes=ARCH-TUNE,FRECIPE,LAM-BH,LD-SEQ-SA -DARCH=la64v1.0 -DTUNE=loongarch64 %s
+// RUN: %clang --target=loongarch64 -x c -E -dM %s -o - -march=loongarch64 -Xclang -target-feature -Xclang +lamcas | \
+// RUN: FileCheck --match-full-lines --check-prefixes=ARCH-TUNE,LAMCAS -DARCH=loongarch64 -DTUNE=loongarch64 %s
+// RUN: %clang --target=loongarch64 -x c -E -dM %s -o - -march=loongarch64 -Xclang -target-feature -Xclang +lsx -Xclang -target-feature -Xclang +lamcas | \
+// RUN: FileCheck --match-full-lines --check-prefixes=ARCH-TUNE,LAMCAS -DARCH=la64v1.0 -DTUNE=loongarch64 %s
// RUN: %clang --target=loongarch64 -x c -E -dM %s -o - -march=la64v1.0 -Xclang -target-feature -Xclang +ld-seq-sa | \
// RUN: FileCheck --match-full-lines --check-prefixes=ARCH-TUNE,LD-SEQ-SA -DARCH=la64v1.0 -DTUNE=loongarch64 %s
// RUN: %clang --target=loongarch64 -x c -E -dM %s -o - -march=la64v1.1 -Xclang -target-feature -Xclang -ld-seq-sa | \
@@ -848,20 +856,21 @@
// RUN: FileCheck --match-full-lines --check-prefixes=ARCH-TUNE,LD-SEQ-SA -DARCH=loongarch64 -DTUNE=loongarch64 %s
// RUN: %clang --target=loongarch64 -x c -E -dM %s -o - -march=loongarch64 -Xclang -target-feature -Xclang +lsx -Xclang -target-feature -Xclang +ld-seq-sa | \
// RUN: FileCheck --match-full-lines --check-prefixes=ARCH-TUNE,LD-SEQ-SA -DARCH=la64v1.0 -DTUNE=loongarch64 %s
-// RUN: %clang --target=loongarch64 -x c -E -dM %s -o - -march=la64v1.0 -Xclang -target-feature -Xclang +frecipe -Xclang -target-feature -Xclang +lam-bh -Xclang -target-feature -Xclang +ld-seq-sa | \
+// RUN: %clang --target=loongarch64 -x c -E -dM %s -o - -march=la64v1.0 -Xclang -target-feature -Xclang +frecipe -Xclang -target-feature -Xclang +lam-bh -Xclang -target-feature -Xclang +lamcas -Xclang -target-feature -Xclang +ld-seq-sa | \
// RUN: FileCheck --match-full-lines --check-prefixes=ARCH-TUNE -DARCH=la64v1.1 -DTUNE=loongarch64 %s
// RUN: %clang --target=loongarch64 -x c -E -dM %s -o - -march=la664 | \
-// RUN: FileCheck --match-full-lines --check-prefixes=ARCH-TUNE,FRECIPE,LAM-BH,LD-SEQ-SA -DARCH=la664 -DTUNE=la664 %s
+// RUN: FileCheck --match-full-lines --check-prefixes=ARCH-TUNE,FRECIPE,LAM-BH,LAMCAS,LD-SEQ-SA -DARCH=la664 -DTUNE=la664 %s
// RUN: %clang --target=loongarch64 -x c -E -dM %s -o - -mtune=la664 | \
// RUN: FileCheck --match-full-lines --check-prefix=ARCH-TUNE -DARCH=la64v1.0 -DTUNE=la664 %s
// RUN: %clang --target=loongarch64 -x c -E -dM %s -o - -march=loongarch64 -mtune=la664 | \
// RUN: FileCheck --match-full-lines --check-prefix=ARCH-TUNE -DARCH=loongarch64 -DTUNE=la664 %s
// RUN: %clang --target=loongarch64 -x c -E -dM %s -o - -march=la664 -mtune=loongarch64 | \
-// RUN: FileCheck --match-full-lines --check-prefixes=ARCH-TUNE,FRECIPE,LAM-BH,LD-SEQ-SA -DARCH=la664 -DTUNE=loongarch64 %s
+// RUN: FileCheck --match-full-lines --check-prefixes=ARCH-TUNE,FRECIPE,LAM-BH,LAMCAS,LD-SEQ-SA -DARCH=la664 -DTUNE=loongarch64 %s
// ARCH-TUNE: #define __loongarch_arch "[[ARCH]]"
// FRECIPE: #define __loongarch_frecipe 1
// LAM-BH: #define __loongarch_lam_bh 1
+// LAMCAS: #define __loongarch_lamcas 1
// LD-SEQ-SA: #define __loongarch_ld_seq_sa 1
// ARCH-TUNE: #define __loongarch_tune "[[TUNE]]"
diff --git a/llvm/include/llvm/TargetParser/LoongArchTargetParser.def b/llvm/include/llvm/TargetParser/LoongArchTargetParser.def
index 324d5c18e6dea3..78f3d0bee85e83 100644
--- a/llvm/include/llvm/TargetParser/LoongArchTargetParser.def
+++ b/llvm/include/llvm/TargetParser/LoongArchTargetParser.def
@@ -12,6 +12,7 @@ LOONGARCH_FEATURE("+lvz", FK_LVZ)
LOONGARCH_FEATURE("+ual", FK_UAL)
LOONGARCH_FEATURE("+frecipe", FK_FRECIPE)
LOONGARCH_FEATURE("+lam-bh", FK_LAM_BH)
+LOONGARCH_FEATURE("+lamcas", FK_LAMCAS)
LOONGARCH_FEATURE("+ld-seq-sa", FK_LD_SEQ_SA)
#undef LOONGARCH_FEATURE
@@ -22,6 +23,6 @@ LOONGARCH_FEATURE("+ld-seq-sa", FK_LD_SEQ_SA)
LOONGARCH_ARCH("loongarch64", AK_LOONGARCH64, FK_64BIT | FK_FP32 | FK_FP64 | FK_UAL)
LOONGARCH_ARCH("la464", AK_LA464, FK_64BIT | FK_FP32 | FK_FP64 | FK_LSX | FK_LASX | FK_UAL)
-LOONGARCH_ARCH("la664", AK_LA664, FK_64BIT | FK_FP32 | FK_FP64 | FK_LSX | FK_LASX | FK_UAL | FK_FRECIPE | FK_LAM_BH | FK_LD_SEQ_SA)
+LOONGARCH_ARCH("la664", AK_LA664, FK_64BIT | FK_FP32 | FK_FP64 | FK_LSX | FK_LASX | FK_UAL | FK_FRECIPE | FK_LAM_BH | FK_LAMCAS | FK_LD_SEQ_SA)
#undef LOONGARCH_ARCH
diff --git a/llvm/include/llvm/TargetParser/LoongArchTargetParser.h b/llvm/include/llvm/TargetParser/LoongArchTargetParser.h
index 00957b84ab576c..c0012c68fb7dbd 100644
--- a/llvm/include/llvm/TargetParser/LoongArchTargetParser.h
+++ b/llvm/include/llvm/TargetParser/LoongArchTargetParser.h
@@ -54,6 +54,10 @@ enum FeatureKind : uint32_t {
// available.
FK_LAM_BH = 1 << 10,
+ // Atomic memory compare and swap instructions for byte, half word, word and
+ // double word are available.
+ FK_LAMCAS = 1 << 11,
+
// Do not generate load-load barrier instructions (dbar 0x700).
FK_LD_SEQ_SA = 1 << 12,
diff --git a/llvm/lib/Target/LoongArch/LoongArch.td b/llvm/lib/Target/LoongArch/LoongArch.td
index 100bdba36c440c..e5ae61da5a6699 100644
--- a/llvm/lib/Target/LoongArch/LoongArch.td
+++ b/llvm/lib/Target/LoongArch/LoongArch.td
@@ -118,6 +118,12 @@ def FeatureLAM_BH
"Support amswap[_db].{b/h} and amadd[_db].{b/h} instructions.">;
def HasLAM_BH : Predicate<"Subtarget->hasLAM_BH()">;
+// Atomic memory compare and swap instructions for byte, half word, word and double word
+def FeatureLAMCAS
+ : SubtargetFeature<"lamcas", "HasLAMCAS", "true",
+ "Support amcas[_db].{b/h/w/d}.">;
+def HasLAMCAS : Predicate<"Subtarget->hasLAMCAS()">;
+
def FeatureLD_SEQ_SA
: SubtargetFeature<"ld-seq-sa", "HasLD_SEQ_SA", "true",
"Don't use load-load barrier (dbar 0x700).">;
@@ -165,7 +171,8 @@ def : ProcessorModel<"la664", NoSchedModel, [Feature64Bit,
FeatureExtLVZ,
FeatureExtLBT,
FeatureFrecipe,
- FeatureLAM_BH]>;
+ FeatureLAM_BH,
+ FeatureLAMCAS]>;
//===----------------------------------------------------------------------===//
// Define the LoongArch target.
diff --git a/llvm/lib/Target/LoongArch/LoongArchISelLowering.cpp b/llvm/lib/Target/LoongArch/LoongArchISelLowering.cpp
index 1abb428175eea7..b64c6dc2165577 100644
--- a/llvm/lib/Target/LoongArch/LoongArchISelLowering.cpp
+++ b/llvm/lib/Target/LoongArch/LoongArchISelLowering.cpp
@@ -371,6 +371,10 @@ LoongArchTargetLowering::LoongArchTargetLowering(const TargetMachine &TM,
setPrefFunctionAlignment(Subtarget.getPrefFunctionAlignment());
setPrefLoopAlignment(Subtarget.getPrefLoopAlignment());
setMaxBytesForAlignment(Subtarget.getMaxBytesForAlignment());
+
+ // cmpxchg sizes down to 8 bits become legal if LAMCAS is available.
+ if (Subtarget.hasLAMCAS())
+ setMinCmpXchgSizeInBits(8);
}
bool LoongArchTargetLowering::isOffsetFoldingLegal(
@@ -5765,6 +5769,10 @@ LoongArchTargetLowering::shouldExpandAtomicRMWInIR(AtomicRMWInst *AI) const {
}
unsigned Size = AI->getType()->getPrimitiveSizeInBits();
+ if (Subtarget.hasLAMCAS() &&
+ (AI->getOperation() == AtomicRMWInst::Nand || Size < 32))
+ return AtomicExpansionKind::CmpXChg;
+
if (Size == 8 || Size == 16)
return AtomicExpansionKind::MaskedIntrinsic;
return AtomicExpansionKind::None;
@@ -5819,6 +5827,10 @@ getIntrinsicForMaskedAtomicRMWBinOp(unsigned GRLen,
TargetLowering::AtomicExpansionKind
LoongArchTargetLowering::shouldExpandAtomicCmpXchgInIR(
AtomicCmpXchgInst *CI) const {
+
+ if (Subtarget.hasLAMCAS())
+ return AtomicExpansionKind::None;
+
unsigned Size = CI->getCompareOperand()->getType()->getPrimitiveSizeInBits();
if (Size == 8 || Size == 16)
return AtomicExpansionKind::MaskedIntrinsic;
@@ -6314,8 +6326,8 @@ bool LoongArchTargetLowering::hasAndNotCompare(SDValue Y) const {
}
ISD::NodeType LoongArchTargetLowering::getExtendForAtomicCmpSwapArg() const {
- // TODO: LAMCAS will use amcas{_DB,}.[bhwd] which does not require extension.
- return ISD::SIGN_EXTEND;
+ // LAMCAS will use amcas[_DB].{b/h/w/d} which does not require extension.
+ return Subtarget.hasLAMCAS() ? ISD::ANY_EXTEND : ISD::SIGN_EXTEND;
}
bool LoongArchTargetLowering::shouldSignExtendTypeInLibCall(
diff --git a/llvm/lib/Target/LoongArch/LoongArchInstrInfo.td b/llvm/lib/Target/LoongArch/LoongArchInstrInfo.td
index 7993f4f1326937..dd41327abc8422 100644
--- a/llvm/lib/Target/LoongArch/LoongArchInstrInfo.td
+++ b/llvm/lib/Target/LoongArch/LoongArchInstrInfo.td
@@ -719,9 +719,9 @@ class AM_3R<bits<32> op>
}
class AMCAS_3R<bits<32> op>
- : Fmt3R<op, (outs GPR:$rd_wb), (ins GPR:$rd, GPR:$rk, GPRMemAtomic:$rj),
+ : Fmt3R<op, (outs GPR:$dst), (ins GPR:$rd, GPR:$rk, GPRMemAtomic:$rj),
"$rd, $rk, $rj"> {
- let Constraints = "@earlyclobber $rd_wb, $rd_wb = $rd";
+ let Constraints = "@earlyclobber $dst, $dst = $rd";
let IsAMCAS = 1;
}
} // hasSideEffects = 0, mayLoad = 1, mayStore = 1,
@@ -2119,6 +2119,26 @@ def : Pat<(atomic_load_sub_i16 GPR:$rj, GPR:$rk),
(AMADD__DB_H (SUB_W R0, GPR:$rk), GPR:$rj)>;
} // Predicates = [ IsLA64, HasLAM_BH ]
+let Predicates = [ HasLAMCAS, IsLA64 ] in {
+
+def : Pat<(atomic_cmp_swap_i8_monotonic GPR:$addr, GPR:$cmp, GPR:$new),
+ (AMCAS_B GPR:$cmp, GPR:$new, GPR:$addr)>;
+def : Pat<(atomic_cmp_swap_i16_monotonic GPR:$addr, GPR:$cmp, GPR:$new),
+ (AMCAS_H GPR:$cmp, GPR:$new, GPR:$addr)>;
+def : Pat<(atomic_cmp_swap_i32_monotonic GPR:$addr, GPR:$cmp, GPR:$new),
+ (AMCAS_W GPR:$cmp, GPR:$new, GPR:$addr)>;
+def : Pat<(atomic_cmp_swap_i64_monotonic GPR:$addr, GPR:$cmp, GPR:$new),
+ (AMCAS_D GPR:$cmp, GPR:$new, GPR:$addr)>;
+
+def : Pat<(atomic_cmp_swap_i8 GPR:$addr, GPR:$cmp, GPR:$new),
+ (AMCAS__DB_B GPR:$cmp, GPR:$new, GPR:$addr)>;
+def : Pat<(atomic_cmp_swap_i16 GPR:$addr, GPR:$cmp, GPR:$new),
+ (AMCAS__DB_H GPR:$cmp, GPR:$new, GPR:$addr)>;
+def : Pat<(atomic_cmp_swap_i32 GPR:$addr, GPR:$cmp, GPR:$new),
+ (AMCAS__DB_W GPR:$cmp, GPR:$new, GPR:$addr)>;
+def : Pat<(atomic_cmp_swap_i64 GPR:$addr, GPR:$cmp, GPR:$new),
+ (AMCAS__DB_D GPR:$cmp, GPR:$new, GPR:$addr)>;
+}
let Predicates = [IsLA64] in {
diff --git a/llvm/lib/TargetParser/LoongArchTargetParser.cpp b/llvm/lib/TargetParser/LoongArchTargetParser.cpp
index 9b8407a73bea3f..9b42498b3d8953 100644
--- a/llvm/lib/TargetParser/LoongArchTargetParser.cpp
+++ b/llvm/lib/TargetParser/LoongArchTargetParser.cpp
@@ -53,6 +53,7 @@ bool LoongArch::getArchFeatures(StringRef Arch,
if (Arch == "la64v1.1") {
Features.push_back("+frecipe");
Features.push_back("+lam-bh");
+ Features.push_back("+lamcas");
Features.push_back("+ld-seq-sa");
}
return true;
diff --git a/llvm/test/CodeGen/LoongArch/ir-instruction/atomic-cmpxchg.ll b/llvm/test/CodeGen/LoongArch/ir-instruction/atomic-cmpxchg.ll
index 4ff15f2b7e4481..bbc9052549117a 100644
--- a/llvm/test/CodeGen/LoongArch/ir-instruction/atomic-cmpxchg.ll
+++ b/llvm/test/CodeGen/LoongArch/ir-instruction/atomic-cmpxchg.ll
@@ -1,6 +1,7 @@
; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py
-; RUN: llc --mtriple=loongarch64 -mattr=+d,-ld-seq-sa < %s | FileCheck %s --check-prefixes=LA64,NO-LD-SEQ-SA
+; RUN: llc --mtriple=loongarch64 -mattr=+d,-lamcas,-ld-seq-sa < %s | FileCheck %s --check-prefixes=LA64,NO-LD-SEQ-SA
; RUN: llc --mtriple=loongarch64 -mattr=+d,+ld-seq-sa < %s | FileCheck %s --check-prefixes=LA64,LD-SEQ-SA
+; RUN: llc --mtriple=loongarch64 -mattr=+d,+lamcas < %s | FileCheck %s --check-prefix=LA64-LAMCAS
define void @cmpxchg_i8_acquire_acquire(ptr %ptr, i8 %cmp, i8 %val) nounwind {
; LA64-LABEL: cmpxchg_i8_acquire_acquire:
@@ -27,6 +28,11 @@ define void @cmpxchg_i8_acquire_acquire(ptr %ptr, i8 %cmp, i8 %val) nounwind {
; LA64-NEXT: dbar 20
; LA64-NEXT: .LBB0_4:
; LA64-NEXT: ret
+;
+; LA64-LAMCAS-LABEL: cmpxchg_i8_acquire_acquire:
+; LA64-LAMCAS: # %bb.0:
+; LA64-LAMCAS-NEXT: amcas_db.b $a1, $a2, $a0
+; LA64-LAMCAS-NEXT: ret
%res = cmpxchg ptr %ptr, i8 %cmp, i8 %val acquire acquire
ret void
}
@@ -57,6 +63,11 @@ define void @cmpxchg_i16_acquire_acquire(ptr %ptr, i16 %cmp, i16 %val) nounwind
; LA64-NEXT: dbar 20
; LA64-NEXT: .LBB1_4:
; LA64-NEXT: ret
+;
+; LA64-LAMCAS-LABEL: cmpxchg_i16_acquire_acquire:
+; LA64-LAMCAS: # %bb.0:
+; LA64-LAMCAS-NEXT: amcas_db.h $a1, $a2, $a0
+; LA64-LAMCAS-NEXT: ret
%res = cmpxchg ptr %ptr, i16 %cmp, i16 %val acquire acquire
ret void
}
@@ -77,6 +88,11 @@ define void @cmpxchg_i32_acquire_acquire(ptr %ptr, i32 %cmp, i32 %val) nounwind
; LA64-NEXT: dbar 20
; LA64-NEXT: .LBB2_4:
; LA64-NEXT: ret
+;
+; LA64-LAMCAS-LABEL: cmpxchg_i32_acquire_acquire:
+; LA64-LAMCAS: # %bb.0:
+; LA64-LAMCAS-NEXT: amcas_db.w $a1, $a2, $a0
+; LA64-LAMCAS-NEXT: ret
%res = cmpxchg ptr %ptr, i32 %cmp, i32 %val acquire acquire
ret void
}
@@ -96,6 +112,11 @@ define void @cmpxchg_i64_acquire_acquire(ptr %ptr, i64 %cmp, i64 %val) nounwind
; LA64-NEXT: dbar 20
; LA64-NEXT: .LBB3_4:
; LA64-NEXT: ret
+;
+; LA64-LAMCAS-LABEL: cmpxchg_i64_acquire_acquire:
+; LA64-LAMCAS: # %bb.0:
+; LA64-LAMCAS-NEXT: amcas_db.d $a1, $a2, $a0
+; LA64-LAMCAS-NEXT: ret
%res = cmpxchg ptr %ptr, i64 %cmp, i64 %val acquire acquire
ret void
}
@@ -302,6 +323,12 @@ define i8 @cmpxchg_i8_acquire_acquire_reti8(ptr %ptr, i8 %cmp, i8 %val) nounwind
; LA64-NEXT: .LBB8_4:
; LA64-NEXT: srl.w $a0, $a5, $a3
; LA64-NEXT: ret
+;
+; LA64-LAMCAS-LABEL: cmpxchg_i8_acquire_acquire_reti8:
+; LA64-LAMCAS: # %bb.0:
+; LA64-LAMCAS-NEXT: amcas_db.b $a1, $a2, $a0
+; LA64-LAMCAS-NEXT: move $a0, $a1
+; LA64-LAMCAS-NEXT: ret
%tmp = cmpxchg ptr %ptr, i8 %cmp, i8 %val acquire acquire
%res = extractvalue { i8, i1 } %tmp, 0
ret i8 %res
@@ -334,6 +361,12 @@ define i16 @cmpxchg_i16_acquire_acquire_reti16(ptr %ptr, i16 %cmp, i16 %val) nou
; LA64-NEXT: .LBB9_4:
; LA64-NEXT: srl.w $a0, $a5, $a3
; LA64-NEXT: ret
+;
+; LA64-LAMCAS-LABEL: cmpxchg_i16_acquire_acquire_reti16:
+; LA64-LAMCAS: # %bb.0:
+; LA64-LAMCAS-NEXT: amcas_db.h $a1, $a2, $a0
+; LA64-LAMCAS-NEXT: move $a0, $a1
+; LA64-LAMCAS-NEXT: ret
%tmp = cmpxchg ptr %ptr, i16 %cmp, i16 %val acquire acquire
%res = extractvalue { i16, i1 } %tmp, 0
ret i16 %res
@@ -356,6 +389,12 @@ define i32 @cmpxchg_i32_acquire_acquire_reti32(ptr %ptr, i32 %cmp, i32 %val) nou
; LA64-NEXT: .LBB10_4:
; LA64-NEXT: move $a0, $a1
; LA64-NEXT: ret
+;
+; LA64-LAMCAS-LABEL: cmpxchg_i32_acquire_acquire_reti32:
+; LA64-LAMCAS: # %bb.0:
+; LA64-LAMCAS-NEXT: amcas_db.w $a1, $a2, $a0
+; LA64-LAMCAS-NEXT: move $a0, $a1
+; LA64-LAMCAS-NEXT: ret
%tmp = cmpxchg ptr %ptr, i32 %cmp, i32 %val acquire acquire
%res = extractvalue { i32, i1 } %tmp, 0
ret i32 %res
@@ -377,6 +416,12 @@ define i64 @cmpxchg_i64_acquire_acquire_reti64(ptr %ptr, i64 %cmp, i64 %val) nou
; LA64-NEXT: .LBB11_4:
; LA64-NEXT: move $a0, $a3
; LA64-NEXT: ret
+;
+; LA64-LAMCAS-LABEL: cmpxchg_i64_acquire_acquire_reti64:
+; LA64-LAMCAS: # %bb.0:
+; LA64-LAMCAS-NEXT: amcas_db.d $a1, $a2, $a0
+; LA64-LAMCAS-NEXT: move $a0, $a1
+; LA64-LAMCAS-NEXT: ret
%tmp = cmpxchg ptr %ptr, i64 %cmp, i64 %val acquire acquire
%res = extractvalue { i64, i1 } %tmp, 0
ret i64 %res
@@ -410,6 +455,14 @@ define i1 @cmpxchg_i8_acquire_acquire_reti1(ptr %ptr, i8 %cmp, i8 %val) nounwind
; LA64-NEXT: xor $a0, $a1, $a0
; LA64-NEXT: sltui $a0, $a0, 1
; LA64-NEXT: ret
+;
+; LA64-LAMCAS-LABEL: cmpxchg_i8_acquire_acquire_reti1:
+; LA64-LAMCAS: # %bb.0:
+; LA64-LAMCAS-NEXT: ext.w.b $a3, $a1
+; LA64-LAMCAS-NEXT: amcas_db.b $a1, $a2, $a0
+; LA64-LAMCAS-NEXT: xor $a0, $a1, $a3
+; LA64-LAMCAS-NEXT: sltui $a0, $a0, 1
+; LA64-LAMCAS-NEXT: ret
%tmp = cmpxchg ptr %ptr, i8 %cmp, i8 %val acquire acquire
%res = extractvalue { i8, i1 } %tmp, 1
ret i1 %res
@@ -444,6 +497,14 @@ define i1 @cmpxchg_i16_acquire_acquire_reti1(ptr %ptr, i16 %cmp, i16 %val) nounw
; LA64-NEXT: xor $a0, $a1, $a0
; LA64-NEXT: sltui $a0, $a0, 1
; LA64-NEXT: ret
+;
+; LA64-LAMCAS-LABEL: cmpxchg_i16_acquire_acquire_reti1:
+; LA64-LAMCAS: # %bb.0:
+; LA64-LAMCAS-NEXT: ext.w.h $a3, $a1
+; LA64-LAMCAS-NEXT: amcas_db.h $a1, $a2, $a0
+; LA64-LAMCAS-NEXT: xor $a0, $a1, $a3
+; LA64-LAMCAS-NEXT: sltui $a0, $a0, 1
+; LA64-LAMCAS-NEXT: ret
%tmp = cmpxchg ptr %ptr, i16 %cmp, i16 %val acquire acquire
%res = extractvalue { i16, i1 } %tmp, 1
ret i1 %res
@@ -467,6 +528,14 @@ define i1 @cmpxchg_i32_acquire_acquire_reti1(ptr %ptr, i32 %cmp, i32 %val) nounw
; LA64-NEXT: xor $a0, $a3, $a1
; LA64-NEXT: sltui $a0, $a0, 1
; LA64-NEXT: ret
+;
+; LA64-LAMCAS-LABEL: cmpxchg_i32_acquire_acquire_reti1:
+; LA64-LAMCAS: # %bb.0:
+; LA64-LAMCAS-NEXT: addi.w $a3, $a1, 0
+; LA64-LAMCAS-NEXT: amcas_db.w $a1, $a2, $a0
+; LA64-LAMCAS-NEXT: xor $a0, $a1, $a3
+; LA64-LAMCAS-NEXT: sltui $a0, $a0, 1
+; LA64-LAMCAS-NEXT: ret
%tmp = cmpxchg ptr %ptr, i32 %cmp, i32 %val acquire acquire
%res = extractvalue { i32, i1 } %tmp, 1
ret i1 %res
@@ -489,6 +558,14 @@ define i1 @cmpxchg_i64_acquire_acquire_reti1(ptr %ptr, i64 %cmp, i64 %val) nounw
; LA64-NEXT: xor $a0, $a3, $a1
; LA64-NEXT: sltui $a0, $a0, 1
; LA64-NEXT: ret
+;
+; LA64-LAMCAS-LABEL: cmpxchg_i64_acquire_acquire_reti1:
+; LA64-LAMCAS: # %bb.0:
+; LA64-LAMCAS-NEXT: move $a3, $a1
+; LA64-LAMCAS-NEXT: amcas_db.d $a3, $a2, $a0
+; LA64-LAMCAS-NEXT: xor $a0, $a3, $a1
+; LA64-LAMCAS-NEXT: sltui $a0, $a0, 1
+; LA64-LAMCAS-NEXT: ret
%tmp = cmpxchg ptr %ptr, i64 %cmp, i64 %val acquire acquire
%res = extractvalue { i64, i1 } %tmp, 1
ret i1 %res
diff --git a/llvm/test/CodeGen/LoongArch/ir-instruction/atomicrmw-lamcas.ll b/llvm/test/CodeGen/LoongArch/ir-instruction/atomicrmw-lamcas.ll
new file mode 100644
index 00000000000000..2f677038e1db37
--- /dev/null
+++ b/llvm/test/CodeGen/LoongArch/ir-instruction/atomicrmw-lamcas.ll
@@ -0,0 +1,5341 @@
+; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py
+; RUN: llc --mtriple=loongarch64 -mattr=+d,-lamcas < %s | FileCheck %s --check-prefix=LA64
+; RUN: llc --mtriple=loongarch64 -mattr=+d,+lamcas < %s | FileCheck %s --check-prefix=LA64-LAMCAS
+
+define i8 @atomicrmw_xchg_i8_acquire(ptr %a, i8 %b) nounwind {
+; LA64-LABEL: atomicrmw_xchg_i8_acquire:
+; LA64: # %bb.0:
+; LA64-NEXT: slli.d $a2, $a0, 3
+; LA64-NEXT: bstrins.d $a0, $zero, 1, 0
+; LA64-NEXT: ori $a3, $zero, 255
+; LA64-NEXT: sll.w $a3, $a3, $a2
+; LA64-NEXT: andi $a1, $a1, 255
+; LA64-NEXT: sll.w $a1, $a1, $a2
+; LA64-NEXT: .LBB0_1: # =>This Inner Loop Header: Depth=1
+; LA64-NEXT: ll.w $a4, $a0, 0
+; LA64-NEXT: addi.w $a5, $a1, 0
+; LA64-NEXT: xor $a5, $a4, $a5
+; LA64-NEXT: and $a5, $a5, $a3
+; LA64-NEXT: xor $a5, $a4, $a5
+; LA64-NEXT: sc.w $a5, $a0, 0
+; LA64-NEXT: beqz $a5, .LBB0_1
+; LA64-NEXT: # %bb.2:
+; LA64-NEXT: srl.w $a0, $a4, $a2
+; LA64-NEXT: ret
+;
+; LA64-LAMCAS-LABEL: atomicrmw_xchg_i8_acquire:
+; LA64-LAMCAS: # %bb.0:
+; LA64-LAMCAS-NEXT: ld.bu $a2, $a0, 0
+; LA64-LAMCAS-NEXT: .p2align 4, , 16
+; LA64-LAMCAS-NEXT: .LBB0_1: # %atomicrmw.start
+; LA64-LAMCAS-NEXT: # =>This Inner Loop Header: Depth=1
+; LA64-LAMCAS-NEXT: ext.w.b $a3, $a2
+; LA64-LAMCAS-NEXT: amcas_db.b $a2, $a1, $a0
+; LA64-LAMCAS-NEXT: bne $a2, $a3, .LBB0_1
+; LA64-LAMCAS-NEXT: # %bb.2: # %atomicrmw.end
+; LA64-LAMCAS-NEXT: move $a0, $a2
+; LA64-LAMCAS-NEXT: ret
+ %1 = atomicrmw xchg ptr %a, i8 %b acquire
+ ret i8 %1
+}
+
+define i8 @atomicrmw_xchg_0_i8_acquire(ptr %a) nounwind {
+; LA64-LABEL: atomicrmw_xchg_0_i8_acquire:
+; LA64: # %bb.0:
+; LA64-NEXT: slli.d $a1, $a0, 3
+; LA64-NEXT: bstrins.d $a0, $zero, 1, 0
+; LA64-NEXT: ori $a2, $zero, 255
+; LA64-NEXT: sll.w $a2, $a2, $a1
+; LA64-NEXT: nor $a2, $a2, $zero
+; LA64-NEXT: amand_db.w $a3, $a2, $a0
+; LA64-NEXT: srl.w $a0, $a3, $a1
+; LA64-NEXT: ret
+;
+; LA64-LAMCAS-LABEL: atomicrmw_xchg_0_i8_acquire:
+; LA64-LAMCAS: # %bb.0:
+; LA64-LAMCAS-NEXT: move $a1, $a0
+; LA64-LAMCAS-NEXT: ld.bu $a0, $a0, 0
+; LA64-LAMCAS-NEXT: .p2align 4, , 16
+; LA64-LAMCAS-NEXT: .LBB1_1: # %atomicrmw.start
+; LA64-LAMCAS-NEXT: # =>This Inner Loop Header: Depth=1
+; LA64-LAMCAS-NEXT: ext.w.b $a2, $a0
+; LA64-LAMCAS-NEXT: amcas_db.b $a0, $zero, $a1
+; LA64-LAMCAS-NEXT: bne $a0, $a2, .LBB1_1
+; LA64-LAMCAS-NEXT: # %bb.2: # %atomicrmw.end
+; LA64-LAMCAS-NEXT: ret
+ %1 = atomicrmw xchg ptr %a, i8 0 acquire
+ ret i8 %1
+}
+
+define i8 @atomicrmw_xchg_minus_1_i8_acquire(ptr %a) nounwind {
+; LA64-LABEL: atomicrmw_xchg_minus_1_i8_acquire:
+; LA64: # %bb.0:
+; LA64-NEXT: slli.d $a1, $a0, 3
+; LA64-NEXT: bstrins.d $a0, $zero, 1, 0
+; LA64-NEXT: ori $a2, $zero, 255
+; LA64-NEXT: sll.w $a2, $a2, $a1
+; LA64-NEXT: amor_db.w $a3, $a2, $a0
+; LA64-NEXT: srl.w $a0, $a3, $a1
+; LA64-NEXT: ret
+;
+; LA64-LAMCAS-LABEL: atomicrmw_xchg_minus_1_i8_acquire:
+; LA64-LAMCAS: # %bb.0:
+; LA64-LAMCAS-NEXT: move $a1, $a0
+; LA64-LAMCAS-NEXT: ld.bu $a0, $a0, 0
+; LA64-LAMCAS-NEXT: addi.w $a2, $zero, -1
+; LA64-LAMCAS-NEXT: .p2align 4, , 16
+; LA64-LAMCAS-NEXT: .LBB2_1: # %atomicrmw.start
+; LA64-LAMCAS-NEXT: # =>This Inner Loop Header: Depth=1
+; LA64-LAMCAS-NEXT: ext.w.b $a3, $a0
+; LA64-LAMCAS-NEXT: amcas_db.b $a0, $a2, $a1
+; LA64-LAMCAS-NEXT: bne $a0, $a3, .LBB2_1
+; LA64-LAMCAS-NEXT: # %bb.2: # %atomicrmw.end
+; LA64-LAMCAS-NEXT: ret
+ %1 = atomicrmw xchg ptr %a, i8 -1 acquire
+ ret i8 %1
+}
+
+define i16 @atomicrmw_xchg_i16_acquire(ptr %a, i16 %b) nounwind {
+; LA64-LABEL: atomicrmw_xchg_i16_acquire:
+; LA64: # %bb.0:
+; LA64-NEXT: slli.d $a2, $a0, 3
+; LA64-NEXT: bstrins.d $a0, $zero, 1, 0
+; LA64-NEXT: lu12i.w $a3, 15
+; LA64-NEXT: ori $a3, $a3, 4095
+; LA64-NEXT: sll.w $a3, $a3, $a2
+; LA64-NEXT: bstrpick.d $a1, $a1, 15, 0
+; LA64-NEXT: sll.w $a1, $a1, $a2
+; LA64-NEXT: .LBB3_1: # =>This Inner Loop Header: Depth=1
+; LA64-NEXT: ll.w $a4, $a0, 0
+; LA64-NEXT: addi.w $a5, $a1, 0
+; LA64-NEXT: xor $a5, $a4, $a5
+; LA64-NEXT: and $a5, $a5, $a3
+; LA64-NEXT: xor $a5, $a4, $a5
+; LA64-NEXT: sc.w $a5, $a0, 0
+; LA64-NEXT: beqz $a5, .LBB3_1
+; LA64-NEXT: # %bb.2:
+; LA64-NEXT: srl.w $a0, $a4, $a2
+; LA64-NEXT: ret
+;
+; LA64-LAMCAS-LABEL: atomicrmw_xchg_i16_acquire:
+; LA64-LAMCAS: # %bb.0:
+; LA64-LAMCAS-NEXT: ld.hu $a2, $a0, 0
+; LA64-LAMCAS-NEXT: .p2align 4, , 16
+; LA64-LAMCAS-NEXT: .LBB3_1: # %atomicrmw.start
+; LA64-LAMCAS-NEXT: # =>This Inner Loop Header: Depth=1
+; LA64-LAMCAS-NEXT: ext.w.h $a3, $a2
+; LA64-LAMCAS-NEXT: amcas_db.h $a2, $a1, $a0
+; LA64-LAMCAS-NEXT: bne $a2, $a3, .LBB3_1
+; LA64-LAMCAS-NEXT: # %bb.2: # %atomicrmw.end
+; LA64-LAMCAS-NEXT: move $a0, $a2
+; LA64-LAMCAS-NEXT: ret
+ %1 = atomicrmw xchg ptr %a, i16 %b acquire
+ ret i16 %1
+}
+
+define i16 @atomicrmw_xchg_0_i16_acquire(ptr %a) nounwind {
+; LA64-LABEL: atomicrmw_xchg_0_i16_acquire:
+; LA64: # %bb.0:
+; LA64-NEXT: slli.d $a1, $a0, 3
+; LA64-NEXT: bstrins.d $a0, $zero, 1, 0
+; LA64-NEXT: lu12i.w $a2, 15
+; LA64-NEXT: ori $a2, $a2, 4095
+; LA64-NEXT: sll.w $a2, $a2, $a1
+; LA64-NEXT: nor $a2, $a2, $zero
+; LA64-NEXT: amand_db.w $a3, $a2, $a0
+; LA64-NEXT: srl.w $a0, $a3, $a1
+; LA64-NEXT: ret
+;
+; LA64-LAMCAS-LABEL: atomicrmw_xchg_0_i16_acquire:
+; LA64-LAMCAS: # %bb.0:
+; LA64-LAMCAS-NEXT: move $a1, $a0
+; LA64-LAMCAS-NEXT: ld.hu $a0, $a0, 0
+; LA64-LAMCAS-NEXT: .p2align 4, , 16
+; LA64-LAMCAS-NEXT: .LBB4_1: # %atomicrmw.start
+; LA64-LAMCAS-NEXT: # =>This Inner Loop Header: Depth=1
+; LA64-LAMCAS-NEXT: ext.w.h $a2, $a0
+; LA64-LAMCAS-NEXT: amcas_db.h $a0, $zero, $a1
+; LA64-LAMCAS-NEXT: bne $a0, $a2, .LBB4_1
+; LA64-LAMCAS-NEXT: # %bb.2: # %atomicrmw.end
+; LA64-LAMCAS-NEXT: ret
+ %1 = atomicrmw xchg ptr %a, i16 0 acquire
+ ret i16 %1
+}
+
+define i16 @atomicrmw_xchg_minus_1_i16_acquire(ptr %a) nounwind {
+; LA64-LABEL: atomicrmw_xchg_minus_1_i16_acquire:
+; LA64: # %bb.0:
+; LA64-NEXT: slli.d $a1, $a0, 3
+; LA64-NEXT: bstrins.d $a0, $zero, 1, 0
+; LA64-NEXT: lu12i.w $a2, 15
+; LA64-NEXT: ori $a2, $a2, 4095
+; LA64-NEXT: sll.w $a2, $a2, $a1
+; LA64-NEXT: amor_db.w $a3, $a2, $a0
+; LA64-NEXT: srl.w $a0, $a3, $a1
+; LA64-NEXT: ret
+;
+; LA64-LAMCAS-LABEL: atomicrmw_xchg_minus_1_i16_acquire:
+; LA64-LAMCAS: # %bb.0:
+; LA64-LAMCAS-NEXT: move $a1, $a0
+; LA64-LAMCAS-NEXT: ld.hu $a0, $a0, 0
+; LA64-LAMCAS-NEXT: addi.w $a2, $zero, -1
+; LA64-LAMCAS-NEXT: .p2align 4, , 16
+; LA64-LAMCAS-NEXT: .LBB5_1: # %atomicrmw.start
+; LA64-LAMCAS-NEXT: # =>This Inner Loop Header: Depth=1
+; LA64-LAMCAS-NEXT: ext.w.h $a3, $a0
+; LA64-LAMCAS-NEXT: amcas_db.h $a0, $a2, $a1
+; LA64-LAMCAS-NEXT: bne $a0, $a3, .LBB5_1
+; LA64-LAMCAS-NEXT: # %bb.2: # %atomicrmw.end
+; LA64-LAMCAS-NEXT: ret
+ %1 = atomicrmw xchg ptr %a, i16 -1 acquire
+ ret i16 %1
+
+}
+
+define i8 @atomicrmw_add_i8_acquire(ptr %a, i8 %b) nounwind {
+; LA64-LABEL: atomicrmw_add_i8_acquire:
+; LA64: # %bb.0:
+; LA64-NEXT: slli.d $a2, $a0, 3
+; LA64-NEXT: bstrins.d $a0, $zero, 1, 0
+; LA64-NEXT: ori $a3, $zero, 255
+; LA64-NEXT: sll.w $a3, $a3, $a2
+; LA64-NEXT: andi $a1, $a1, 255
+; LA64-NEXT: sll.w $a1, $a1, $a2
+; LA64-NEXT: .LBB6_1: # =>This Inner Loop Header: Depth=1
+; LA64-NEXT: ll.w $a4, $a0, 0
+; LA64-NEXT: add.w $a5, $a4, $a1
+; LA64-NEXT: xor $a5, $a4, $a5
+; LA64-NEXT: and $a5, $a5, $a3
+; LA64-NEXT: xor $a5, $a4, $a5
+; LA64-NEXT: sc.w $a5, $a0, 0
+; LA64-NEXT: beqz $a5, .LBB6_1
+; LA64-NEXT: # %bb.2:
+; LA64-NEXT: srl.w $a0, $a4, $a2
+; LA64-NEXT: ret
+;
+; LA64-LAMCAS-LABEL: atomicrmw_add_i8_acquire:
+; LA64-LAMCAS: # %bb.0:
+; LA64-LAMCAS-NEXT: move $a2, $a0
+; LA64-LAMCAS-NEXT: ld.bu $a0, $a0, 0
+; LA64-LAMCAS-NEXT: .p2align 4, , 16
+; LA64-LAMCAS-NEXT: .LBB6_1: # %atomicrmw.start
+; LA64-LAMCAS-NEXT: # =>This Inner Loop Header: Depth=1
+; LA64-LAMCAS-NEXT: add.d $a3, $a0, $a1
+; LA64-LAMCAS-NEXT: ext.w.b $a4, $a0
+; LA64-LAMCAS-NEXT: amcas_db.b $a0, $a3, $a2
+; LA64-LAMCAS-NEXT: bne $a0, $a4, .LBB6_1
+; LA64-LAMCAS-NEXT: # %bb.2: # %atomicrmw.end
+; LA64-LAMCAS-NEXT: ret
+ %1 = atomicrmw add ptr %a, i8 %b acquire
+ ret i8 %1
+}
+
+define i16 @atomicrmw_add_i16_acquire(ptr %a, i16 %b) nounwind {
+; LA64-LABEL: atomicrmw_add_i16_acquire:
+; LA64: # %bb.0:
+; LA64-NEXT: slli.d $a2, $a0, 3
+; LA64-NEXT: bstrins.d $a0, $zero, 1, 0
+; LA64-NEXT: lu12i.w $a3, 15
+; LA64-NEXT: ori $a3, $a3, 4095
+; LA64-NEXT: sll.w $a3, $a3, $a2
+; LA64-NEXT: bstrpick.d $a1, $a1, 15, 0
+; LA64-NEXT: sll.w $a1, $a1, $a2
+; LA64-NEXT: .LBB7_1: # =>This Inner Loop Header: Depth=1
+; LA64-NEXT: ll.w $a4, $a0, 0
+; LA64-NEXT: add.w $a5, $a4, $a1
+; LA64-NEXT: xor $a5, $a4, $a5
+; LA64-NEXT: and $a5, $a5, $a3
+; LA64-NEXT: xor $a5, $a4, $a5
+; LA64-NEXT: sc.w $a5, $a0, 0
+; LA64-NEXT: beqz $a5, .LBB7_1
+; LA64-NEXT: # %bb.2:
+; LA64-NEXT: srl.w $a0, $a4, $a2
+; LA64-NEXT: ret
+;
+; LA64-LAMCAS-LABEL: atomicrmw_add_i16_acquire:
+; LA64-LAMCAS: # %bb.0:
+; LA64-LAMCAS-NEXT: move $a2, $a0
+; LA64-LAMCAS-NEXT: ld.hu $a0, $a0, 0
+; LA64-LAMCAS-NEXT: .p2align 4, , 16
+; LA64-LAMCAS-NEXT: .LBB7_1: # %atomicrmw.start
+; LA64-LAMCAS-NEXT: # =>This Inner Loop Header: Depth=1
+; LA64-LAMCAS-NEXT: add.d $a3, $a0, $a1
+; LA64-LAMCAS-NEXT: ext.w.h $a4, $a0
+; LA64-LAMCAS-NEXT: amcas_db.h $a0, $a3, $a2
+; LA64-LAMCAS-NEXT: bne $a0, $a4, .LBB7_1
+; LA64-LAMCAS-NEXT: # %bb.2: # %atomicrmw.end
+; LA64-LAMCAS-NEXT: ret
+ %1 = atomicrmw add ptr %a, i16 %b acquire
+ ret i16 %1
+
+}
+
+define i8 @atomicrmw_sub_i8_acquire(ptr %a, i8 %b) nounwind {
+; LA64-LABEL: atomicrmw_sub_i8_acquire:
+; LA64: # %bb.0:
+; LA64-NEXT: slli.d $a2, $a0, 3
+; LA64-NEXT: bstrins.d $a0, $zero, 1, 0
+; LA64-NEXT: ori $a3, $zero, 255
+; LA64-NEXT: sll.w $a3, $a3, $a2
+; LA64-NEXT: andi $a1, $a1, 255
+; LA64-NEXT: sll.w $a1, $a1, $a2
+; LA64-NEXT: .LBB8_1: # =>This Inner Loop Header: Depth=1
+; LA64-NEXT: ll.w $a4, $a0, 0
+; LA64-NEXT: sub.w $a5, $a4, $a1
+; LA64-NEXT: xor $a5, $a4, $a5
+; LA64-NEXT: and $a5, $a5, $a3
+; LA64-NEXT: xor $a5, $a4, $a5
+; LA64-NEXT: sc.w $a5, $a0, 0
+; LA64-NEXT: beqz $a5, .LBB8_1
+; LA64-NEXT: # %bb.2:
+; LA64-NEXT: srl.w $a0, $a4, $a2
+; LA64-NEXT: ret
+;
+; LA64-LAMCAS-LABEL: atomicrmw_sub_i8_acquire:
+; LA64-LAMCAS: # %bb.0:
+; LA64-LAMCAS-NEXT: move $a2, $a0
+; LA64-LAMCAS-NEXT: ld.bu $a0, $a0, 0
+; LA64-LAMCAS-NEXT: .p2align 4, , 16
+; LA64-LAMCAS-NEXT: .LBB8_1: # %atomicrmw.start
+; LA64-LAMCAS-NEXT: # =>This Inner Loop Header: Depth=1
+; LA64-LAMCAS-NEXT: sub.d $a3, $a0, $a1
+; LA64-LAMCAS-NEXT: ext.w.b $a4, $a0
+; LA64-LAMCAS-NEXT: amcas_db.b $a0, $a3, $a2
+; LA64-LAMCAS-NEXT: bne $a0, $a4, .LBB8_1
+; LA64-LAMCAS-NEXT: # %bb.2: # %atomicrmw.end
+; LA64-LAMCAS-NEXT: ret
+ %1 = atomicrmw sub ptr %a, i8 %b acquire
+ ret i8 %1
+}
+
+define i16 @atomicrmw_sub_i16_acquire(ptr %a, i16 %b) nounwind {
+; LA64-LABEL: atomicrmw_sub_i16_acquire:
+; LA64: # %bb.0:
+; LA64-NEXT: slli.d $a2, $a0, 3
+; LA64-NEXT: bstrins.d $a0, $zero, 1, 0
+; LA64-NEXT: lu12i.w $a3, 15
+; LA64-NEXT: ori $a3, $a3, 4095
+; LA64-NEXT: sll.w $a3, $a3, $a2
+; LA64-NEXT: bstrpick.d $a1, $a1, 15, 0
+; LA64-NEXT: sll.w $a1, $a1, $a2
+; LA64-NEXT: .LBB9_1: # =>This Inner Loop Header: Depth=1
+; LA64-NEXT: ll.w $a4, $a0, 0
+; LA64-NEXT: sub.w $a5, $a4, $a1
+; LA64-NEXT: xor $a5, $a4, $a5
+; LA64-NEXT: and $a5, $a5, $a3
+; LA64-NEXT: xor $a5, $a4, $a5
+; LA64-NEXT: sc.w $a5, $a0, 0
+; LA64-NEXT: beqz $a5, .LBB9_1
+; LA64-NEXT: # %bb.2:
+; LA64-NEXT: srl.w $a0, $a4, $a2
+; LA64-NEXT: ret
+;
+; LA64-LAMCAS-LABEL: atomicrmw_sub_i16_acquire:
+; LA64-LAMCAS: # %bb.0:
+; LA64-LAMCAS-NEXT: move $a2, $a0
+; LA64-LAMCAS-NEXT: ld.hu $a0, $a0, 0
+; LA64-LAMCAS-NEXT: .p2align 4, , 16
+; LA64-LAMCAS-NEXT: .LBB9_1: # %atomicrmw.start
+; LA64-LAMCAS-NEXT: # =>This Inner Loop Header: Depth=1
+; LA64-LAMCAS-NEXT: sub.d $a3, $a0, $a1
+; LA64-LAMCAS-NEXT: ext.w.h $a4, $a0
+; LA64-LAMCAS-NEXT: amcas_db.h $a0, $a3, $a2
+; LA64-LAMCAS-NEXT: bne $a0, $a4, .LBB9_1
+; LA64-LAMCAS-NEXT: # %bb.2: # %atomicrmw.end
+; LA64-LAMCAS-NEXT: ret
+ %1 = atomicrmw sub ptr %a, i16 %b acquire
+ ret i16 %1
+
+}
+
+define i8 @atomicrmw_umax_i8_acquire(ptr %a, i8 %b) nounwind {
+; LA64-LABEL: atomicrmw_umax_i8_acquire:
+; LA64: # %bb.0:
+; LA64-NEXT: slli.d $a2, $a0, 3
+; LA64-NEXT: bstrins.d $a0, $zero, 1, 0
+; LA64-NEXT: ori $a3, $zero, 255
+; LA64-NEXT: sll.w $a3, $a3, $a2
+; LA64-NEXT: andi $a1, $a1, 255
+; LA64-NEXT: sll.w $a1, $a1, $a2
+; LA64-NEXT: .LBB10_1: # =>This Inner Loop Header: Depth=1
+; LA64-NEXT: ll.w $a4, $a0, 0
+; LA64-NEXT: and $a6, $a4, $a3
+; LA64-NEXT: move $a5, $a4
+; LA64-NEXT: bgeu $a6, $a1, .LBB10_3
+; LA64-NEXT: # %bb.2: # in Loop: Header=BB10_1 Depth=1
+; LA64-NEXT: xor $a5, $a4, $a1
+; LA64-NEXT: and $a5, $a5, $a3
+; LA64-NEXT: xor $a5, $a4, $a5
+; LA64-NEXT: .LBB10_3: # in Loop: Header=BB10_1 Depth=1
+; LA64-NEXT: sc.w $a5, $a0, 0
+; LA64-NEXT: beqz $a5, .LBB10_1
+; LA64-NEXT: # %bb.4:
+; LA64-NEXT: srl.w $a0, $a4, $a2
+; LA64-NEXT: ret
+;
+; LA64-LAMCAS-LABEL: atomicrmw_umax_i8_acquire:
+; LA64-LAMCAS: # %bb.0:
+; LA64-LAMCAS-NEXT: move $a2, $a0
+; LA64-LAMCAS-NEXT: ld.bu $a0, $a0, 0
+; LA64-LAMCAS-NEXT: andi $a3, $a1, 255
+; LA64-LAMCAS-NEXT: .p2align 4, , 16
+; LA64-LAMCAS-NEXT: .LBB10_1: # %atomicrmw.start
+; LA64-LAMCAS-NEXT: # =>This Inner Loop Header: Depth=1
+; LA64-LAMCAS-NEXT: andi $a4, $a0, 255
+; LA64-LAMCAS-NEXT: sltu $a4, $a3, $a4
+; LA64-LAMCAS-NEXT: masknez $a5, $a1, $a4
+; LA64-LAMCAS-NEXT: maskeqz $a4, $a0, $a4
+; LA64-LAMCAS-NEXT: or $a4, $a4, $a5
+; LA64-LAMCAS-NEXT: ext.w.b $a5, $a0
+; LA64-LAMCAS-NEXT: amcas_db.b $a0, $a4, $a2
+; LA64-LAMCAS-NEXT: bne $a0, $a5, .LBB10_1
+; LA64-LAMCAS-NEXT: # %bb.2: # %atomicrmw.end
+; LA64-LAMCAS-NEXT: ret
+ %1 = atomicrmw umax ptr %a, i8 %b acquire
+ ret i8 %1
+}
+
+define i16 @atomicrmw_umax_i16_acquire(ptr %a, i16 %b) nounwind {
+; LA64-LABEL: atomicrmw_umax_i16_acquire:
+; LA64: # %bb.0:
+; LA64-NEXT: slli.d $a2, $a0, 3
+; LA64-NEXT: bstrins.d $a0, $zero, 1, 0
+; LA64-NEXT: lu12i.w $a3, 15
+; LA64-NEXT: ori $a3, $a3, 4095
+; LA64-NEXT: sll.w $a3, $a3, $a2
+; LA64-NEXT: bstrpick.d $a1, $a1, 15, 0
+; LA64-NEXT: sll.w $a1, $a1, $a2
+; LA64-NEXT: .LBB11_1: # =>This Inner Loop Header: Depth=1
+; LA64-NEXT: ll.w $a4, $a0, 0
+; LA64-NEXT: and $a6, $a4, $a3
+; LA64-NEXT: move $a5, $a4
+; LA64-NEXT: bgeu $a6, $a1, .LBB11_3
+; LA64-NEXT: # %bb.2: # in Loop: Header=BB11_1 Depth=1
+; LA64-NEXT: xor $a5, $a4, $a1
+; LA64-NEXT: and $a5, $a5, $a3
+; LA64-NEXT: xor $a5, $a4, $a5
+; LA64-NEXT: .LBB11_3: # in Loop: Header=BB11_1 Depth=1
+; LA64-NEXT: sc.w $a5, $a0, 0
+; LA64-NEXT: beqz $a5, .LBB11_1
+; LA64-NEXT: # %bb.4:
+; LA64-NEXT: srl.w $a0, $a4, $a2
+; LA64-NEXT: ret
+;
+; LA64-LAMCAS-LABEL: atomicrmw_umax_i16_acquire:
+; LA64-LAMCAS: # %bb.0:
+; LA64-LAMCAS-NEXT: move $a2, $a0
+; LA64-LAMCAS-NEXT: ld.hu $a0, $a0, 0
+; LA64-LAMCAS-NEXT: bstrpick.d $a3, $a1, 15, 0
+; LA64-LAMCAS-NEXT: .p2align 4, , 16
+; LA64-LAMCAS-NEXT: .LBB11_1: # %atomicrmw.start
+; LA64-LAMCAS-NEXT: # =>This Inner Loop Header: Depth=1
+; LA64-LAMCAS-NEXT: bstrpick.d $a4, $a0, 15, 0
+; LA64-LAMCAS-NEXT: sltu $a4, $a3, $a4
+; LA64-LAMCAS-NEXT: masknez $a5, $a1, $a4
+; LA64-LAMCAS-NEXT: maskeqz $a4, $a0, $a4
+; LA64-LAMCAS-NEXT: or $a4, $a4, $a5
+; LA64-LAMCAS-NEXT: ext.w.h $a5, $a0
+; LA64-LAMCAS-NEXT: amcas_db.h $a0, $a4, $a2
+; LA64-LAMCAS-NEXT: bne $a0, $a5, .LBB11_1
+; LA64-LAMCAS-NEXT: # %bb.2: # %atomicrmw.end
+; LA64-LAMCAS-NEXT: ret
+ %1 = atomicrmw umax ptr %a, i16 %b acquire
+ ret i16 %1
+}
+
+define i8 @atomicrmw_umin_i8_acquire(ptr %a, i8 %b) nounwind {
+; LA64-LABEL: atomicrmw_umin_i8_acquire:
+; LA64: # %bb.0:
+; LA64-NEXT: slli.d $a2, $a0, 3
+; LA64-NEXT: bstrins.d $a0, $zero, 1, 0
+; LA64-NEXT: ori $a3, $zero, 255
+; LA64-NEXT: sll.w $a3, $a3, $a2
+; LA64-NEXT: andi $a1, $a1, 255
+; LA64-NEXT: sll.w $a1, $a1, $a2
+; LA64-NEXT: .LBB12_1: # =>This Inner Loop Header: Depth=1
+; LA64-NEXT: ll.w $a4, $a0, 0
+; LA64-NEXT: and $a6, $a4, $a3
+; LA64-NEXT: move $a5, $a4
+; LA64-NEXT: bgeu $a1, $a6, .LBB12_3
+; LA64-NEXT: # %bb.2: # in Loop: Header=BB12_1 Depth=1
+; LA64-NEXT: xor $a5, $a4, $a1
+; LA64-NEXT: and $a5, $a5, $a3
+; LA64-NEXT: xor $a5, $a4, $a5
+; LA64-NEXT: .LBB12_3: # in Loop: Header=BB12_1 Depth=1
+; LA64-NEXT: sc.w $a5, $a0, 0
+; LA64-NEXT: beqz $a5, .LBB12_1
+; LA64-NEXT: # %bb.4:
+; LA64-NEXT: srl.w $a0, $a4, $a2
+; LA64-NEXT: ret
+;
+; LA64-LAMCAS-LABEL: atomicrmw_umin_i8_acquire:
+; LA64-LAMCAS: # %bb.0:
+; LA64-LAMCAS-NEXT: move $a2, $a0
+; LA64-LAMCAS-NEXT: ld.bu $a0, $a0, 0
+; LA64-LAMCAS-NEXT: andi $a3, $a1, 255
+; LA64-LAMCAS-NEXT: .p2align 4, , 16
+; LA64-LAMCAS-NEXT: .LBB12_1: # %atomicrmw.start
+; LA64-LAMCAS-NEXT: # =>This Inner Loop Header: Depth=1
+; LA64-LAMCAS-NEXT: andi $a4, $a0, 255
+; LA64-LAMCAS-NEXT: sltu $a4, $a3, $a4
+; LA64-LAMCAS-NEXT: xori $a4, $a4, 1
+; LA64-LAMCAS-NEXT: masknez $a5, $a1, $a4
+; LA64-LAMCAS-NEXT: maskeqz $a4, $a0, $a4
+; LA64-LAMCAS-NEXT: or $a4, $a4, $a5
+; LA64-LAMCAS-NEXT: ext.w.b $a5, $a0
+; LA64-LAMCAS-NEXT: amcas_db.b $a0, $a4, $a2
+; LA64-LAMCAS-NEXT: bne $a0, $a5, .LBB12_1
+; LA64-LAMCAS-NEXT: # %bb.2: # %atomicrmw.end
+; LA64-LAMCAS-NEXT: ret
+ %1 = atomicrmw umin ptr %a, i8 %b acquire
+ ret i8 %1
+}
+
+define i16 @atomicrmw_umin_i16_acquire(ptr %a, i16 %b) nounwind {
+; LA64-LABEL: atomicrmw_umin_i16_acquire:
+; LA64: # %bb.0:
+; LA64-NEXT: slli.d $a2, $a0, 3
+; LA64-NEXT: bstrins.d $a0, $zero, 1, 0
+; LA64-NEXT: lu12i.w $a3, 15
+; LA64-NEXT: ori $a3, $a3, 4095
+; LA64-NEXT: sll.w $a3, $a3, $a2
+; LA64-NEXT: bstrpick.d $a1, $a1, 15, 0
+; LA64-NEXT: sll.w $a1, $a1, $a2
+; LA64-NEXT: .LBB13_1: # =>This Inner Loop Header: Depth=1
+; LA64-NEXT: ll.w $a4, $a0, 0
+; LA64-NEXT: and $a6, $a4, $a3
+; LA64-NEXT: move $a5, $a4
+; LA64-NEXT: bgeu $a1, $a6, .LBB13_3
+; LA64-NEXT: # %bb.2: # in Loop: Header=BB13_1 Depth=1
+; LA64-NEXT: xor $a5, $a4, $a1
+; LA64-NEXT: and $a5, $a5, $a3
+; LA64-NEXT: xor $a5, $a4, $a5
+; LA64-NEXT: .LBB13_3: # in Loop: Header=BB13_1 Depth=1
+; LA64-NEXT: sc.w $a5, $a0, 0
+; LA64-NEXT: beqz $a5, .LBB13_1
+; LA64-NEXT: # %bb.4:
+; LA64-NEXT: srl.w $a0, $a4, $a2
+; LA64-NEXT: ret
+;
+; LA64-LAMCAS-LABEL: atomicrmw_umin_i16_acquire:
+; LA64-LAMCAS: # %bb.0:
+; LA64-LAMCAS-NEXT: move $a2, $a0
+; LA64-LAMCAS-NEXT: ld.hu $a0, $a0, 0
+; LA64-LAMCAS-NEXT: bstrpick.d $a3, $a1, 15, 0
+; LA64-LAMCAS-NEXT: .p2align 4, , 16
+; LA64-LAMCAS-NEXT: .LBB13_1: # %atomicrmw.start
+; LA64-LAMCAS-NEXT: # =>This Inner Loop Header: Depth=1
+; LA64-LAMCAS-NEXT: bstrpick.d $a4, $a0, 15, 0
+; LA64-LAMCAS-NEXT: sltu $a4, $a3, $a4
+; LA64-LAMCAS-NEXT: xori $a4, $a4, 1
+; LA64-LAMCAS-NEXT: masknez $a5, $a1, $a4
+; LA64-LAMCAS-NEXT: maskeqz $a4, $a0, $a4
+; LA64-LAMCAS-NEXT: or $a4, $a4, $a5
+; LA64-LAMCAS-NEXT: ext.w.h $a5, $a0
+; LA64-LAMCAS-NEXT: amcas_db.h $a0, $a4, $a2
+; LA64-LAMCAS-NEXT: bne $a0, $a5, .LBB13_1
+; LA64-LAMCAS-NEXT: # %bb.2: # %atomicrmw.end
+; LA64-LAMCAS-NEXT: ret
+ %1 = atomicrmw umin ptr %a, i16 %b acquire
+ ret i16 %1
+}
+
+define i8 @atomicrmw_max_i8_acquire(ptr %a, i8 %b) nounwind {
+; LA64-LABEL: atomicrmw_max_i8_acquire:
+; LA64: # %bb.0:
+; LA64-NEXT: slli.d $a2, $a0, 3
+; LA64-NEXT: bstrins.d $a0, $zero, 1, 0
+; LA64-NEXT: andi $a3, $a2, 24
+; LA64-NEXT: ori $a4, $zero, 255
+; LA64-NEXT: sll.w $a4, $a4, $a2
+; LA64-NEXT: ext.w.b $a1, $a1
+; LA64-NEXT: sll.w $a1, $a1, $a2
+; LA64-NEXT: xori $a3, $a3, 56
+; LA64-NEXT: .LBB14_1: # =>This Inner Loop Header: Depth=1
+; LA64-NEXT: ll.w $a5, $a0, 0
+; LA64-NEXT: and $a7, $a5, $a4
+; LA64-NEXT: move $a6, $a5
+; LA64-NEXT: sll.w $a7, $a7, $a3
+; LA64-NEXT: sra.w $a7, $a7, $a3
+; LA64-NEXT: bge $a7, $a1, .LBB14_3
+; LA64-NEXT: # %bb.2: # in Loop: Header=BB14_1 Depth=1
+; LA64-NEXT: xor $a6, $a5, $a1
+; LA64-NEXT: and $a6, $a6, $a4
+; LA64-NEXT: xor $a6, $a5, $a6
+; LA64-NEXT: .LBB14_3: # in Loop: Header=BB14_1 Depth=1
+; LA64-NEXT: sc.w $a6, $a0, 0
+; LA64-NEXT: beqz $a6, .LBB14_1
+; LA64-NEXT: # %bb.4:
+; LA64-NEXT: srl.w $a0, $a5, $a2
+; LA64-NEXT: ret
+;
+; LA64-LAMCAS-LABEL: atomicrmw_max_i8_acquire:
+; LA64-LAMCAS: # %bb.0:
+; LA64-LAMCAS-NEXT: move $a2, $a0
+; LA64-LAMCAS-NEXT: ld.bu $a0, $a0, 0
+; LA64-LAMCAS-NEXT: ext.w.b $a3, $a1
+; LA64-LAMCAS-NEXT: .p2align 4, , 16
+; LA64-LAMCAS-NEXT: .LBB14_1: # %atomicrmw.start
+; LA64-LAMCAS-NEXT: # =>This Inner Loop Header: Depth=1
+; LA64-LAMCAS-NEXT: ext.w.b $a4, $a0
+; LA64-LAMCAS-NEXT: slt $a5, $a3, $a4
+; LA64-LAMCAS-NEXT: masknez $a6, $a1, $a5
+; LA64-LAMCAS-NEXT: maskeqz $a5, $a0, $a5
+; LA64-LAMCAS-NEXT: or $a5, $a5, $a6
+; LA64-LAMCAS-NEXT: amcas_db.b $a0, $a5, $a2
+; LA64-LAMCAS-NEXT: bne $a0, $a4, .LBB14_1
+; LA64-LAMCAS-NEXT: # %bb.2: # %atomicrmw.end
+; LA64-LAMCAS-NEXT: ret
+ %1 = atomicrmw max ptr %a, i8 %b acquire
+ ret i8 %1
+}
+
+define i16 @atomicrmw_max_i16_acquire(ptr %a, i16 %b) nounwind {
+; LA64-LABEL: atomicrmw_max_i16_acquire:
+; LA64: # %bb.0:
+; LA64-NEXT: slli.d $a2, $a0, 3
+; LA64-NEXT: bstrins.d $a0, $zero, 1, 0
+; LA64-NEXT: andi $a3, $a2, 24
+; LA64-NEXT: lu12i.w $a4, 15
+; LA64-NEXT: ori $a4, $a4, 4095
+; LA64-NEXT: sll.w $a4, $a4, $a2
+; LA64-NEXT: ext.w.h $a1, $a1
+; LA64-NEXT: sll.w $a1, $a1, $a2
+; LA64-NEXT: ori $a5, $zero, 48
+; LA64-NEXT: sub.d $a3, $a5, $a3
+; LA64-NEXT: .LBB15_1: # =>This Inner Loop Header: Depth=1
+; LA64-NEXT: ll.w $a5, $a0, 0
+; LA64-NEXT: and $a7, $a5, $a4
+; LA64-NEXT: move $a6, $a5
+; LA64-NEXT: sll.w $a7, $a7, $a3
+; LA64-NEXT: sra.w $a7, $a7, $a3
+; LA64-NEXT: bge $a7, $a1, .LBB15_3
+; LA64-NEXT: # %bb.2: # in Loop: Header=BB15_1 Depth=1
+; LA64-NEXT: xor $a6, $a5, $a1
+; LA64-NEXT: and $a6, $a6, $a4
+; LA64-NEXT: xor $a6, $a5, $a6
+; LA64-NEXT: .LBB15_3: # in Loop: Header=BB15_1 Depth=1
+; LA64-NEXT: sc.w $a6, $a0, 0
+; LA64-NEXT: beqz $a6, .LBB15_1
+; LA64-NEXT: # %bb.4:
+; LA64-NEXT: srl.w $a0, $a5, $a2
+; LA64-NEXT: ret
+;
+; LA64-LAMCAS-LABEL: atomicrmw_max_i16_acquire:
+; LA64-LAMCAS: # %bb.0:
+; LA64-LAMCAS-NEXT: move $a2, $a0
+; LA64-LAMCAS-NEXT: ld.hu $a0, $a0, 0
+; LA64-LAMCAS-NEXT: ext.w.h $a3, $a1
+; LA64-LAMCAS-NEXT: .p2align 4, , 16
+; LA64-LAMCAS-NEXT: .LBB15_1: # %atomicrmw.start
+; LA64-LAMCAS-NEXT: # =>This Inner Loop Header: Depth=1
+; LA64-LAMCAS-NEXT: ext.w.h $a4, $a0
+; LA64-LAMCAS-NEXT: slt $a5, $a3, $a4
+; LA64-LAMCAS-NEXT: masknez $a6, $a1, $a5
+; LA64-LAMCAS-NEXT: maskeqz $a5, $a0, $a5
+; LA64-LAMCAS-NEXT: or $a5, $a5, $a6
+; LA64-LAMCAS-NEXT: amcas_db.h $a0, $a5, $a2
+; LA64-LAMCAS-NEXT: bne $a0, $a4, .LBB15_1
+; LA64-LAMCAS-NEXT: # %bb.2: # %atomicrmw.end
+; LA64-LAMCAS-NEXT: ret
+ %1 = atomicrmw max ptr %a, i16 %b acquire
+ ret i16 %1
+}
+
+define i8 @atomicrmw_min_i8_acquire(ptr %a, i8 %b) nounwind {
+; LA64-LABEL: atomicrmw_min_i8_acquire:
+; LA64: # %bb.0:
+; LA64-NEXT: slli.d $a2, $a0, 3
+; LA64-NEXT: bstrins.d $a0, $zero, 1, 0
+; LA64-NEXT: andi $a3, $a2, 24
+; LA64-NEXT: ori $a4, $zero, 255
+; LA64-NEXT: sll.w $a4, $a4, $a2
+; LA64-NEXT: ext.w.b $a1, $a1
+; LA64-NEXT: sll.w $a1, $a1, $a2
+; LA64-NEXT: xori $a3, $a3, 56
+; LA64-NEXT: .LBB16_1: # =>This Inner Loop Header: Depth=1
+; LA64-NEXT: ll.w $a5, $a0, 0
+; LA64-NEXT: and $a7, $a5, $a4
+; LA64-NEXT: move $a6, $a5
+; LA64-NEXT: sll.w $a7, $a7, $a3
+; LA64-NEXT: sra.w $a7, $a7, $a3
+; LA64-NEXT: bge $a1, $a7, .LBB16_3
+; LA64-NEXT: # %bb.2: # in Loop: Header=BB16_1 Depth=1
+; LA64-NEXT: xor $a6, $a5, $a1
+; LA64-NEXT: and $a6, $a6, $a4
+; LA64-NEXT: xor $a6, $a5, $a6
+; LA64-NEXT: .LBB16_3: # in Loop: Header=BB16_1 Depth=1
+; LA64-NEXT: sc.w $a6, $a0, 0
+; LA64-NEXT: beqz $a6, .LBB16_1
+; LA64-NEXT: # %bb.4:
+; LA64-NEXT: srl.w $a0, $a5, $a2
+; LA64-NEXT: ret
+;
+; LA64-LAMCAS-LABEL: atomicrmw_min_i8_acquire:
+; LA64-LAMCAS: # %bb.0:
+; LA64-LAMCAS-NEXT: move $a2, $a0
+; LA64-LAMCAS-NEXT: ld.bu $a0, $a0, 0
+; LA64-LAMCAS-NEXT: ext.w.b $a3, $a1
+; LA64-LAMCAS-NEXT: .p2align 4, , 16
+; LA64-LAMCAS-NEXT: .LBB16_1: # %atomicrmw.start
+; LA64-LAMCAS-NEXT: # =>This Inner Loop Header: Depth=1
+; LA64-LAMCAS-NEXT: ext.w.b $a4, $a0
+; LA64-LAMCAS-NEXT: slt $a5, $a3, $a4
+; LA64-LAMCAS-NEXT: xori $a5, $a5, 1
+; LA64-LAMCAS-NEXT: masknez $a6, $a1, $a5
+; LA64-LAMCAS-NEXT: maskeqz $a5, $a0, $a5
+; LA64-LAMCAS-NEXT: or $a5, $a5, $a6
+; LA64-LAMCAS-NEXT: amcas_db.b $a0, $a5, $a2
+; LA64-LAMCAS-NEXT: bne $a0, $a4, .LBB16_1
+; LA64-LAMCAS-NEXT: # %bb.2: # %atomicrmw.end
+; LA64-LAMCAS-NEXT: ret
+ %1 = atomicrmw min ptr %a, i8 %b acquire
+ ret i8 %1
+}
+
+define i16 @atomicrmw_min_i16_acquire(ptr %a, i16 %b) nounwind {
+; LA64-LABEL: atomicrmw_min_i16_acquire:
+; LA64: # %bb.0:
+; LA64-NEXT: slli.d $a2, $a0, 3
+; LA64-NEXT: bstrins.d $a0, $zero, 1, 0
+; LA64-NEXT: andi $a3, $a2, 24
+; LA64-NEXT: lu12i.w $a4, 15
+; LA64-NEXT: ori $a4, $a4, 4095
+; LA64-NEXT: sll.w $a4, $a4, $a2
+; LA64-NEXT: ext.w.h $a1, $a1
+; LA64-NEXT: sll.w $a1, $a1, $a2
+; LA64-NEXT: ori $a5, $zero, 48
+; LA64-NEXT: sub.d $a3, $a5, $a3
+; LA64-NEXT: .LBB17_1: # =>This Inner Loop Header: Depth=1
+; LA64-NEXT: ll.w $a5, $a0, 0
+; LA64-NEXT: and $a7, $a5, $a4
+; LA64-NEXT: move $a6, $a5
+; LA64-NEXT: sll.w $a7, $a7, $a3
+; LA64-NEXT: sra.w $a7, $a7, $a3
+; LA64-NEXT: bge $a1, $a7, .LBB17_3
+; LA64-NEXT: # %bb.2: # in Loop: Header=BB17_1 Depth=1
+; LA64-NEXT: xor $a6, $a5, $a1
+; LA64-NEXT: and $a6, $a6, $a4
+; LA64-NEXT: xor $a6, $a5, $a6
+; LA64-NEXT: .LBB17_3: # in Loop: Header=BB17_1 Depth=1
+; LA64-NEXT: sc.w $a6, $a0, 0
+; LA64-NEXT: beqz $a6, .LBB17_1
+; LA64-NEXT: # %bb.4:
+; LA64-NEXT: srl.w $a0, $a5, $a2
+; LA64-NEXT: ret
+;
+; LA64-LAMCAS-LABEL: atomicrmw_min_i16_acquire:
+; LA64-LAMCAS: # %bb.0:
+; LA64-LAMCAS-NEXT: move $a2, $a0
+; LA64-LAMCAS-NEXT: ld.hu $a0, $a0, 0
+; LA64-LAMCAS-NEXT: ext.w.h $a3, $a1
+; LA64-LAMCAS-NEXT: .p2align 4, , 16
+; LA64-LAMCAS-NEXT: .LBB17_1: # %atomicrmw.start
+; LA64-LAMCAS-NEXT: # =>This Inner Loop Header: Depth=1
+; LA64-LAMCAS-NEXT: ext.w.h $a4, $a0
+; LA64-LAMCAS-NEXT: slt $a5, $a3, $a4
+; LA64-LAMCAS-NEXT: xori $a5, $a5, 1
+; LA64-LAMCAS-NEXT: masknez $a6, $a1, $a5
+; LA64-LAMCAS-NEXT: maskeqz $a5, $a0, $a5
+; LA64-LAMCAS-NEXT: or $a5, $a5, $a6
+; LA64-LAMCAS-NEXT: amcas_db.h $a0, $a5, $a2
+; LA64-LAMCAS-NEXT: bne $a0, $a4, .LBB17_1
+; LA64-LAMCAS-NEXT: # %bb.2: # %atomicrmw.end
+; LA64-LAMCAS-NEXT: ret
+ %1 = atomicrmw min ptr %a, i16 %b acquire
+ ret i16 %1
+}
+
+define i8 @atomicrmw_nand_i8_acquire(ptr %a, i8 %b) nounwind {
+; LA64-LABEL: atomicrmw_nand_i8_acquire:
+; LA64: # %bb.0:
+; LA64-NEXT: slli.d $a2, $a0, 3
+; LA64-NEXT: bstrins.d $a0, $zero, 1, 0
+; LA64-NEXT: ori $a3, $zero, 255
+; LA64-NEXT: sll.w $a3, $a3, $a2
+; LA64-NEXT: andi $a1, $a1, 255
+; LA64-NEXT: sll.w $a1, $a1, $a2
+; LA64-NEXT: .LBB18_1: # =>This Inner Loop Header: Depth=1
+; LA64-NEXT: ll.w $a4, $a0, 0
+; LA64-NEXT: and $a5, $a4, $a1
+; LA64-NEXT: nor $a5, $a5, $zero
+; LA64-NEXT: xor $a5, $a4, $a5
+; LA64-NEXT: and $a5, $a5, $a3
+; LA64-NEXT: xor $a5, $a4, $a5
+; LA64-NEXT: sc.w $a5, $a0, 0
+; LA64-NEXT: beqz $a5, .LBB18_1
+; LA64-NEXT: # %bb.2:
+; LA64-NEXT: srl.w $a0, $a4, $a2
+; LA64-NEXT: ret
+;
+; LA64-LAMCAS-LABEL: atomicrmw_nand_i8_acquire:
+; LA64-LAMCAS: # %bb.0:
+; LA64-LAMCAS-NEXT: move $a2, $a0
+; LA64-LAMCAS-NEXT: ld.bu $a0, $a0, 0
+; LA64-LAMCAS-NEXT: .p2align 4, , 16
+; LA64-LAMCAS-NEXT: .LBB18_1: # %atomicrmw.start
+; LA64-LAMCAS-NEXT: # =>This Inner Loop Header: Depth=1
+; LA64-LAMCAS-NEXT: and $a3, $a0, $a1
+; LA64-LAMCAS-NEXT: nor $a3, $a3, $zero
+; LA64-LAMCAS-NEXT: ext.w.b $a4, $a0
+; LA64-LAMCAS-NEXT: amcas_db.b $a0, $a3, $a2
+; LA64-LAMCAS-NEXT: bne $a0, $a4, .LBB18_1
+; LA64-LAMCAS-NEXT: # %bb.2: # %atomicrmw.end
+; LA64-LAMCAS-NEXT: ret
+ %1 = atomicrmw nand ptr %a, i8 %b acquire
+ ret i8 %1
+}
+
+define i16 @atomicrmw_nand_i16_acquire(ptr %a, i16 %b) nounwind {
+; LA64-LABEL: atomicrmw_nand_i16_acquire:
+; LA64: # %bb.0:
+; LA64-NEXT: slli.d $a2, $a0, 3
+; LA64-NEXT: bstrins.d $a0, $zero, 1, 0
+; LA64-NEXT: lu12i.w $a3, 15
+; LA64-NEXT: ori $a3, $a3, 4095
+; LA64-NEXT: sll.w $a3, $a3, $a2
+; LA64-NEXT: bstrpick.d $a1, $a1, 15, 0
+; LA64-NEXT: sll.w $a1, $a1, $a2
+; LA64-NEXT: .LBB19_1: # =>This Inner Loop Header: Depth=1
+; LA64-NEXT: ll.w $a4, $a0, 0
+; LA64-NEXT: and $a5, $a4, $a1
+; LA64-NEXT: nor $a5, $a5, $zero
+; LA64-NEXT: xor $a5, $a4, $a5
+; LA64-NEXT: and $a5, $a5, $a3
+; LA64-NEXT: xor $a5, $a4, $a5
+; LA64-NEXT: sc.w $a5, $a0, 0
+; LA64-NEXT: beqz $a5, .LBB19_1
+; LA64-NEXT: # %bb.2:
+; LA64-NEXT: srl.w $a0, $a4, $a2
+; LA64-NEXT: ret
+;
+; LA64-LAMCAS-LABEL: atomicrmw_nand_i16_acquire:
+; LA64-LAMCAS: # %bb.0:
+; LA64-LAMCAS-NEXT: move $a2, $a0
+; LA64-LAMCAS-NEXT: ld.hu $a0, $a0, 0
+; LA64-LAMCAS-NEXT: .p2align 4, , 16
+; LA64-LAMCAS-NEXT: .LBB19_1: # %atomicrmw.start
+; LA64-LAMCAS-NEXT: # =>This Inner Loop Header: Depth=1
+; LA64-LAMCAS-NEXT: and $a3, $a0, $a1
+; LA64-LAMCAS-NEXT: nor $a3, $a3, $zero
+; LA64-LAMCAS-NEXT: ext.w.h $a4, $a0
+; LA64-LAMCAS-NEXT: amcas_db.h $a0, $a3, $a2
+; LA64-LAMCAS-NEXT: bne $a0, $a4, .LBB19_1
+; LA64-LAMCAS-NEXT: # %bb.2: # %atomicrmw.end
+; LA64-LAMCAS-NEXT: ret
+ %1 = atomicrmw nand ptr %a, i16 %b acquire
+ ret i16 %1
+
+}
+
+define i32 @atomicrmw_nand_i32_acquire(ptr %a, i32 %b) nounwind {
+; LA64-LABEL: atomicrmw_nand_i32_acquire:
+; LA64: # %bb.0:
+; LA64-NEXT: .LBB20_1: # =>This Inner Loop Header: Depth=1
+; LA64-NEXT: ll.w $a2, $a0, 0
+; LA64-NEXT: and $a3, $a2, $a1
+; LA64-NEXT: nor $a3, $a3, $zero
+; LA64-NEXT: sc.w $a3, $a0, 0
+; LA64-NEXT: beqz $a3, .LBB20_1
+; LA64-NEXT: # %bb.2:
+; LA64-NEXT: move $a0, $a2
+; LA64-NEXT: ret
+;
+; LA64-LAMCAS-LABEL: atomicrmw_nand_i32_acquire:
+; LA64-LAMCAS: # %bb.0:
+; LA64-LAMCAS-NEXT: move $a2, $a0
+; LA64-LAMCAS-NEXT: ld.w $a0, $a0, 0
+; LA64-LAMCAS-NEXT: .p2align 4, , 16
+; LA64-LAMCAS-NEXT: .LBB20_1: # %atomicrmw.start
+; LA64-LAMCAS-NEXT: # =>This Inner Loop Header: Depth=1
+; LA64-LAMCAS-NEXT: move $a3, $a0
+; LA64-LAMCAS-NEXT: and $a4, $a0, $a1
+; LA64-LAMCAS-NEXT: nor $a4, $a4, $zero
+; LA64-LAMCAS-NEXT: amcas_db.w $a0, $a4, $a2
+; LA64-LAMCAS-NEXT: bne $a0, $a3, .LBB20_1
+; LA64-LAMCAS-NEXT: # %bb.2: # %atomicrmw.end
+; LA64-LAMCAS-NEXT: ret
+ %1 = atomicrmw nand ptr %a, i32 %b acquire
+ ret i32 %1
+}
+
+define i64 @atomicrmw_nand_i64_acquire(ptr %a, i64 %b) nounwind {
+; LA64-LABEL: atomicrmw_nand_i64_acquire:
+; LA64: # %bb.0:
+; LA64-NEXT: .LBB21_1: # =>This Inner Loop Header: Depth=1
+; LA64-NEXT: ll.d $a2, $a0, 0
+; LA64-NEXT: and $a3, $a2, $a1
+; LA64-NEXT: nor $a3, $a3, $zero
+; LA64-NEXT: sc.d $a3, $a0, 0
+; LA64-NEXT: beqz $a3, .LBB21_1
+; LA64-NEXT: # %bb.2:
+; LA64-NEXT: move $a0, $a2
+; LA64-NEXT: ret
+;
+; LA64-LAMCAS-LABEL: atomicrmw_nand_i64_acquire:
+; LA64-LAMCAS: # %bb.0:
+; LA64-LAMCAS-NEXT: move $a2, $a0
+; LA64-LAMCAS-NEXT: ld.d $a0, $a0, 0
+; LA64-LAMCAS-NEXT: .p2align 4, , 16
+; LA64-LAMCAS-NEXT: .LBB21_1: # %atomicrmw.start
+; LA64-LAMCAS-NEXT: # =>This Inner Loop Header: Depth=1
+; LA64-LAMCAS-NEXT: move $a3, $a0
+; LA64-LAMCAS-NEXT: and $a4, $a0, $a1
+; LA64-LAMCAS-NEXT: nor $a4, $a4, $zero
+; LA64-LAMCAS-NEXT: amcas_db.d $a0, $a4, $a2
+; LA64-LAMCAS-NEXT: bne $a0, $a3, .LBB21_1
+; LA64-LAMCAS-NEXT: # %bb.2: # %atomicrmw.end
+; LA64-LAMCAS-NEXT: ret
+ %1 = atomicrmw nand ptr %a, i64 %b acquire
+ ret i64 %1
+}
+
+define i8 @atomicrmw_and_i8_acquire(ptr %a, i8 %b) nounwind {
+; LA64-LABEL: atomicrmw_and_i8_acquire:
+; LA64: # %bb.0:
+; LA64-NEXT: slli.d $a2, $a0, 3
+; LA64-NEXT: bstrins.d $a0, $zero, 1, 0
+; LA64-NEXT: ori $a3, $zero, 255
+; LA64-NEXT: sll.w $a3, $a3, $a2
+; LA64-NEXT: andi $a1, $a1, 255
+; LA64-NEXT: sll.w $a1, $a1, $a2
+; LA64-NEXT: orn $a1, $a1, $a3
+; LA64-NEXT: amand_db.w $a3, $a1, $a0
+; LA64-NEXT: srl.w $a0, $a3, $a2
+; LA64-NEXT: ret
+;
+; LA64-LAMCAS-LABEL: atomicrmw_and_i8_acquire:
+; LA64-LAMCAS: # %bb.0:
+; LA64-LAMCAS-NEXT: move $a2, $a0
+; LA64-LAMCAS-NEXT: ld.bu $a0, $a0, 0
+; LA64-LAMCAS-NEXT: .p2align 4, , 16
+; LA64-LAMCAS-NEXT: .LBB22_1: # %atomicrmw.start
+; LA64-LAMCAS-NEXT: # =>This Inner Loop Header: Depth=1
+; LA64-LAMCAS-NEXT: and $a3, $a0, $a1
+; LA64-LAMCAS-NEXT: ext.w.b $a4, $a0
+; LA64-LAMCAS-NEXT: amcas_db.b $a0, $a3, $a2
+; LA64-LAMCAS-NEXT: bne $a0, $a4, .LBB22_1
+; LA64-LAMCAS-NEXT: # %bb.2: # %atomicrmw.end
+; LA64-LAMCAS-NEXT: ret
+ %1 = atomicrmw and ptr %a, i8 %b acquire
+ ret i8 %1
+}
+
+define i16 @atomicrmw_and_i16_acquire(ptr %a, i16 %b) nounwind {
+; LA64-LABEL: atomicrmw_and_i16_acquire:
+; LA64: # %bb.0:
+; LA64-NEXT: slli.d $a2, $a0, 3
+; LA64-NEXT: bstrins.d $a0, $zero, 1, 0
+; LA64-NEXT: lu12i.w $a3, 15
+; LA64-NEXT: ori $a3, $a3, 4095
+; LA64-NEXT: sll.w $a3, $a3, $a2
+; LA64-NEXT: bstrpick.d $a1, $a1, 15, 0
+; LA64-NEXT: sll.w $a1, $a1, $a2
+; LA64-NEXT: orn $a1, $a1, $a3
+; LA64-NEXT: amand_db.w $a3, $a1, $a0
+; LA64-NEXT: srl.w $a0, $a3, $a2
+; LA64-NEXT: ret
+;
+; LA64-LAMCAS-LABEL: atomicrmw_and_i16_acquire:
+; LA64-LAMCAS: # %bb.0:
+; LA64-LAMCAS-NEXT: move $a2, $a0
+; LA64-LAMCAS-NEXT: ld.hu $a0, $a0, 0
+; LA64-LAMCAS-NEXT: .p2align 4, , 16
+; LA64-LAMCAS-NEXT: .LBB23_1: # %atomicrmw.start
+; LA64-LAMCAS-NEXT: # =>This Inner Loop Header: Depth=1
+; LA64-LAMCAS-NEXT: and $a3, $a0, $a1
+; LA64-LAMCAS-NEXT: ext.w.h $a4, $a0
+; LA64-LAMCAS-NEXT: amcas_db.h $a0, $a3, $a2
+; LA64-LAMCAS-NEXT: bne $a0, $a4, .LBB23_1
+; LA64-LAMCAS-NEXT: # %bb.2: # %atomicrmw.end
+; LA64-LAMCAS-NEXT: ret
+ %1 = atomicrmw and ptr %a, i16 %b acquire
+ ret i16 %1
+
+}
+
+define i8 @atomicrmw_or_i8_acquire(ptr %a, i8 %b) nounwind {
+; LA64-LABEL: atomicrmw_or_i8_acquire:
+; LA64: # %bb.0:
+; LA64-NEXT: slli.d $a2, $a0, 3
+; LA64-NEXT: bstrins.d $a0, $zero, 1, 0
+; LA64-NEXT: andi $a1, $a1, 255
+; LA64-NEXT: sll.w $a1, $a1, $a2
+; LA64-NEXT: amor_db.w $a3, $a1, $a0
+; LA64-NEXT: srl.w $a0, $a3, $a2
+; LA64-NEXT: ret
+;
+; LA64-LAMCAS-LABEL: atomicrmw_or_i8_acquire:
+; LA64-LAMCAS: # %bb.0:
+; LA64-LAMCAS-NEXT: move $a2, $a0
+; LA64-LAMCAS-NEXT: ld.bu $a0, $a0, 0
+; LA64-LAMCAS-NEXT: .p2align 4, , 16
+; LA64-LAMCAS-NEXT: .LBB24_1: # %atomicrmw.start
+; LA64-LAMCAS-NEXT: # =>This Inner Loop Header: Depth=1
+; LA64-LAMCAS-NEXT: or $a3, $a0, $a1
+; LA64-LAMCAS-NEXT: ext.w.b $a4, $a0
+; LA64-LAMCAS-NEXT: amcas_db.b $a0, $a3, $a2
+; LA64-LAMCAS-NEXT: bne $a0, $a4, .LBB24_1
+; LA64-LAMCAS-NEXT: # %bb.2: # %atomicrmw.end
+; LA64-LAMCAS-NEXT: ret
+ %1 = atomicrmw or ptr %a, i8 %b acquire
+ ret i8 %1
+}
+
+define i16 @atomicrmw_or_i16_acquire(ptr %a, i16 %b) nounwind {
+; LA64-LABEL: atomicrmw_or_i16_acquire:
+; LA64: # %bb.0:
+; LA64-NEXT: slli.d $a2, $a0, 3
+; LA64-NEXT: bstrins.d $a0, $zero, 1, 0
+; LA64-NEXT: bstrpick.d $a1, $a1, 15, 0
+; LA64-NEXT: sll.w $a1, $a1, $a2
+; LA64-NEXT: amor_db.w $a3, $a1, $a0
+; LA64-NEXT: srl.w $a0, $a3, $a2
+; LA64-NEXT: ret
+;
+; LA64-LAMCAS-LABEL: atomicrmw_or_i16_acquire:
+; LA64-LAMCAS: # %bb.0:
+; LA64-LAMCAS-NEXT: move $a2, $a0
+; LA64-LAMCAS-NEXT: ld.hu $a0, $a0, 0
+; LA64-LAMCAS-NEXT: .p2align 4, , 16
+; LA64-LAMCAS-NEXT: .LBB25_1: # %atomicrmw.start
+; LA64-LAMCAS-NEXT: # =>This Inner Loop Header: Depth=1
+; LA64-LAMCAS-NEXT: or $a3, $a0, $a1
+; LA64-LAMCAS-NEXT: ext.w.h $a4, $a0
+; LA64-LAMCAS-NEXT: amcas_db.h $a0, $a3, $a2
+; LA64-LAMCAS-NEXT: bne $a0, $a4, .LBB25_1
+; LA64-LAMCAS-NEXT: # %bb.2: # %atomicrmw.end
+; LA64-LAMCAS-NEXT: ret
+ %1 = atomicrmw or ptr %a, i16 %b acquire
+ ret i16 %1
+
+}
+
+define i8 @atomicrmw_xor_i8_acquire(ptr %a, i8 %b) nounwind {
+; LA64-LABEL: atomicrmw_xor_i8_acquire:
+; LA64: # %bb.0:
+; LA64-NEXT: slli.d $a2, $a0, 3
+; LA64-NEXT: bstrins.d $a0, $zero, 1, 0
+; LA64-NEXT: andi $a1, $a1, 255
+; LA64-NEXT: sll.w $a1, $a1, $a2
+; LA64-NEXT: amxor_db.w $a3, $a1, $a0
+; LA64-NEXT: srl.w $a0, $a3, $a2
+; LA64-NEXT: ret
+;
+; LA64-LAMCAS-LABEL: atomicrmw_xor_i8_acquire:
+; LA64-LAMCAS: # %bb.0:
+; LA64-LAMCAS-NEXT: move $a2, $a0
+; LA64-LAMCAS-NEXT: ld.bu $a0, $a0, 0
+; LA64-LAMCAS-NEXT: .p2align 4, , 16
+; LA64-LAMCAS-NEXT: .LBB26_1: # %atomicrmw.start
+; LA64-LAMCAS-NEXT: # =>This Inner Loop Header: Depth=1
+; LA64-LAMCAS-NEXT: xor $a3, $a0, $a1
+; LA64-LAMCAS-NEXT: ext.w.b $a4, $a0
+; LA64-LAMCAS-NEXT: amcas_db.b $a0, $a3, $a2
+; LA64-LAMCAS-NEXT: bne $a0, $a4, .LBB26_1
+; LA64-LAMCAS-NEXT: # %bb.2: # %atomicrmw.end
+; LA64-LAMCAS-NEXT: ret
+ %1 = atomicrmw xor ptr %a, i8 %b acquire
+ ret i8 %1
+}
+
+define i16 @atomicrmw_xor_i16_acquire(ptr %a, i16 %b) nounwind {
+; LA64-LABEL: atomicrmw_xor_i16_acquire:
+; LA64: # %bb.0:
+; LA64-NEXT: slli.d $a2, $a0, 3
+; LA64-NEXT: bstrins.d $a0, $zero, 1, 0
+; LA64-NEXT: bstrpick.d $a1, $a1, 15, 0
+; LA64-NEXT: sll.w $a1, $a1, $a2
+; LA64-NEXT: amxor_db.w $a3, $a1, $a0
+; LA64-NEXT: srl.w $a0, $a3, $a2
+; LA64-NEXT: ret
+;
+; LA64-LAMCAS-LABEL: atomicrmw_xor_i16_acquire:
+; LA64-LAMCAS: # %bb.0:
+; LA64-LAMCAS-NEXT: move $a2, $a0
+; LA64-LAMCAS-NEXT: ld.hu $a0, $a0, 0
+; LA64-LAMCAS-NEXT: .p2align 4, , 16
+; LA64-LAMCAS-NEXT: .LBB27_1: # %atomicrmw.start
+; LA64-LAMCAS-NEXT: # =>This Inner Loop Header: Depth=1
+; LA64-LAMCAS-NEXT: xor $a3, $a0, $a1
+; LA64-LAMCAS-NEXT: ext.w.h $a4, $a0
+; LA64-LAMCAS-NEXT: amcas_db.h $a0, $a3, $a2
+; LA64-LAMCAS-NEXT: bne $a0, $a4, .LBB27_1
+; LA64-LAMCAS-NEXT: # %bb.2: # %atomicrmw.end
+; LA64-LAMCAS-NEXT: ret
+ %1 = atomicrmw xor ptr %a, i16 %b acquire
+ ret i16 %1
+
+}
+
+define i8 @atomicrmw_xchg_i8_release(ptr %a, i8 %b) nounwind {
+; LA64-LABEL: atomicrmw_xchg_i8_release:
+; LA64: # %bb.0:
+; LA64-NEXT: slli.d $a2, $a0, 3
+; LA64-NEXT: bstrins.d $a0, $zero, 1, 0
+; LA64-NEXT: ori $a3, $zero, 255
+; LA64-NEXT: sll.w $a3, $a3, $a2
+; LA64-NEXT: andi $a1, $a1, 255
+; LA64-NEXT: sll.w $a1, $a1, $a2
+; LA64-NEXT: .LBB28_1: # =>This Inner Loop Header: Depth=1
+; LA64-NEXT: ll.w $a4, $a0, 0
+; LA64-NEXT: addi.w $a5, $a1, 0
+; LA64-NEXT: xor $a5, $a4, $a5
+; LA64-NEXT: and $a5, $a5, $a3
+; LA64-NEXT: xor $a5, $a4, $a5
+; LA64-NEXT: sc.w $a5, $a0, 0
+; LA64-NEXT: beqz $a5, .LBB28_1
+; LA64-NEXT: # %bb.2:
+; LA64-NEXT: srl.w $a0, $a4, $a2
+; LA64-NEXT: ret
+;
+; LA64-LAMCAS-LABEL: atomicrmw_xchg_i8_release:
+; LA64-LAMCAS: # %bb.0:
+; LA64-LAMCAS-NEXT: ld.bu $a2, $a0, 0
+; LA64-LAMCAS-NEXT: .p2align 4, , 16
+; LA64-LAMCAS-NEXT: .LBB28_1: # %atomicrmw.start
+; LA64-LAMCAS-NEXT: # =>This Inner Loop Header: Depth=1
+; LA64-LAMCAS-NEXT: ext.w.b $a3, $a2
+; LA64-LAMCAS-NEXT: amcas_db.b $a2, $a1, $a0
+; LA64-LAMCAS-NEXT: bne $a2, $a3, .LBB28_1
+; LA64-LAMCAS-NEXT: # %bb.2: # %atomicrmw.end
+; LA64-LAMCAS-NEXT: move $a0, $a2
+; LA64-LAMCAS-NEXT: ret
+ %1 = atomicrmw xchg ptr %a, i8 %b release
+ ret i8 %1
+}
+
+define i8 @atomicrmw_xchg_0_i8_release(ptr %a) nounwind {
+; LA64-LABEL: atomicrmw_xchg_0_i8_release:
+; LA64: # %bb.0:
+; LA64-NEXT: slli.d $a1, $a0, 3
+; LA64-NEXT: bstrins.d $a0, $zero, 1, 0
+; LA64-NEXT: ori $a2, $zero, 255
+; LA64-NEXT: sll.w $a2, $a2, $a1
+; LA64-NEXT: nor $a2, $a2, $zero
+; LA64-NEXT: amand_db.w $a3, $a2, $a0
+; LA64-NEXT: srl.w $a0, $a3, $a1
+; LA64-NEXT: ret
+;
+; LA64-LAMCAS-LABEL: atomicrmw_xchg_0_i8_release:
+; LA64-LAMCAS: # %bb.0:
+; LA64-LAMCAS-NEXT: move $a1, $a0
+; LA64-LAMCAS-NEXT: ld.bu $a0, $a0, 0
+; LA64-LAMCAS-NEXT: .p2align 4, , 16
+; LA64-LAMCAS-NEXT: .LBB29_1: # %atomicrmw.start
+; LA64-LAMCAS-NEXT: # =>This Inner Loop Header: Depth=1
+; LA64-LAMCAS-NEXT: ext.w.b $a2, $a0
+; LA64-LAMCAS-NEXT: amcas_db.b $a0, $zero, $a1
+; LA64-LAMCAS-NEXT: bne $a0, $a2, .LBB29_1
+; LA64-LAMCAS-NEXT: # %bb.2: # %atomicrmw.end
+; LA64-LAMCAS-NEXT: ret
+ %1 = atomicrmw xchg ptr %a, i8 0 release
+ ret i8 %1
+}
+
+define i8 @atomicrmw_xchg_minus_1_i8_release(ptr %a) nounwind {
+; LA64-LABEL: atomicrmw_xchg_minus_1_i8_release:
+; LA64: # %bb.0:
+; LA64-NEXT: slli.d $a1, $a0, 3
+; LA64-NEXT: bstrins.d $a0, $zero, 1, 0
+; LA64-NEXT: ori $a2, $zero, 255
+; LA64-NEXT: sll.w $a2, $a2, $a1
+; LA64-NEXT: amor_db.w $a3, $a2, $a0
+; LA64-NEXT: srl.w $a0, $a3, $a1
+; LA64-NEXT: ret
+;
+; LA64-LAMCAS-LABEL: atomicrmw_xchg_minus_1_i8_release:
+; LA64-LAMCAS: # %bb.0:
+; LA64-LAMCAS-NEXT: move $a1, $a0
+; LA64-LAMCAS-NEXT: ld.bu $a0, $a0, 0
+; LA64-LAMCAS-NEXT: addi.w $a2, $zero, -1
+; LA64-LAMCAS-NEXT: .p2align 4, , 16
+; LA64-LAMCAS-NEXT: .LBB30_1: # %atomicrmw.start
+; LA64-LAMCAS-NEXT: # =>This Inner Loop Header: Depth=1
+; LA64-LAMCAS-NEXT: ext.w.b $a3, $a0
+; LA64-LAMCAS-NEXT: amcas_db.b $a0, $a2, $a1
+; LA64-LAMCAS-NEXT: bne $a0, $a3, .LBB30_1
+; LA64-LAMCAS-NEXT: # %bb.2: # %atomicrmw.end
+; LA64-LAMCAS-NEXT: ret
+ %1 = atomicrmw xchg ptr %a, i8 -1 release
+ ret i8 %1
+}
+
+define i16 @atomicrmw_xchg_i16_release(ptr %a, i16 %b) nounwind {
+; LA64-LABEL: atomicrmw_xchg_i16_release:
+; LA64: # %bb.0:
+; LA64-NEXT: slli.d $a2, $a0, 3
+; LA64-NEXT: bstrins.d $a0, $zero, 1, 0
+; LA64-NEXT: lu12i.w $a3, 15
+; LA64-NEXT: ori $a3, $a3, 4095
+; LA64-NEXT: sll.w $a3, $a3, $a2
+; LA64-NEXT: bstrpick.d $a1, $a1, 15, 0
+; LA64-NEXT: sll.w $a1, $a1, $a2
+; LA64-NEXT: .LBB31_1: # =>This Inner Loop Header: Depth=1
+; LA64-NEXT: ll.w $a4, $a0, 0
+; LA64-NEXT: addi.w $a5, $a1, 0
+; LA64-NEXT: xor $a5, $a4, $a5
+; LA64-NEXT: and $a5, $a5, $a3
+; LA64-NEXT: xor $a5, $a4, $a5
+; LA64-NEXT: sc.w $a5, $a0, 0
+; LA64-NEXT: beqz $a5, .LBB31_1
+; LA64-NEXT: # %bb.2:
+; LA64-NEXT: srl.w $a0, $a4, $a2
+; LA64-NEXT: ret
+;
+; LA64-LAMCAS-LABEL: atomicrmw_xchg_i16_release:
+; LA64-LAMCAS: # %bb.0:
+; LA64-LAMCAS-NEXT: ld.hu $a2, $a0, 0
+; LA64-LAMCAS-NEXT: .p2align 4, , 16
+; LA64-LAMCAS-NEXT: .LBB31_1: # %atomicrmw.start
+; LA64-LAMCAS-NEXT: # =>This Inner Loop Header: Depth=1
+; LA64-LAMCAS-NEXT: ext.w.h $a3, $a2
+; LA64-LAMCAS-NEXT: amcas_db.h $a2, $a1, $a0
+; LA64-LAMCAS-NEXT: bne $a2, $a3, .LBB31_1
+; LA64-LAMCAS-NEXT: # %bb.2: # %atomicrmw.end
+; LA64-LAMCAS-NEXT: move $a0, $a2
+; LA64-LAMCAS-NEXT: ret
+ %1 = atomicrmw xchg ptr %a, i16 %b release
+ ret i16 %1
+}
+
+define i16 @atomicrmw_xchg_0_i16_release(ptr %a) nounwind {
+; LA64-LABEL: atomicrmw_xchg_0_i16_release:
+; LA64: # %bb.0:
+; LA64-NEXT: slli.d $a1, $a0, 3
+; LA64-NEXT: bstrins.d $a0, $zero, 1, 0
+; LA64-NEXT: lu12i.w $a2, 15
+; LA64-NEXT: ori $a2, $a2, 4095
+; LA64-NEXT: sll.w $a2, $a2, $a1
+; LA64-NEXT: nor $a2, $a2, $zero
+; LA64-NEXT: amand_db.w $a3, $a2, $a0
+; LA64-NEXT: srl.w $a0, $a3, $a1
+; LA64-NEXT: ret
+;
+; LA64-LAMCAS-LABEL: atomicrmw_xchg_0_i16_release:
+; LA64-LAMCAS: # %bb.0:
+; LA64-LAMCAS-NEXT: move $a1, $a0
+; LA64-LAMCAS-NEXT: ld.hu $a0, $a0, 0
+; LA64-LAMCAS-NEXT: .p2align 4, , 16
+; LA64-LAMCAS-NEXT: .LBB32_1: # %atomicrmw.start
+; LA64-LAMCAS-NEXT: # =>This Inner Loop Header: Depth=1
+; LA64-LAMCAS-NEXT: ext.w.h $a2, $a0
+; LA64-LAMCAS-NEXT: amcas_db.h $a0, $zero, $a1
+; LA64-LAMCAS-NEXT: bne $a0, $a2, .LBB32_1
+; LA64-LAMCAS-NEXT: # %bb.2: # %atomicrmw.end
+; LA64-LAMCAS-NEXT: ret
+ %1 = atomicrmw xchg ptr %a, i16 0 release
+ ret i16 %1
+}
+
+define i16 @atomicrmw_xchg_minus_1_i16_release(ptr %a) nounwind {
+; LA64-LABEL: atomicrmw_xchg_minus_1_i16_release:
+; LA64: # %bb.0:
+; LA64-NEXT: slli.d $a1, $a0, 3
+; LA64-NEXT: bstrins.d $a0, $zero, 1, 0
+; LA64-NEXT: lu12i.w $a2, 15
+; LA64-NEXT: ori $a2, $a2, 4095
+; LA64-NEXT: sll.w $a2, $a2, $a1
+; LA64-NEXT: amor_db.w $a3, $a2, $a0
+; LA64-NEXT: srl.w $a0, $a3, $a1
+; LA64-NEXT: ret
+;
+; LA64-LAMCAS-LABEL: atomicrmw_xchg_minus_1_i16_release:
+; LA64-LAMCAS: # %bb.0:
+; LA64-LAMCAS-NEXT: move $a1, $a0
+; LA64-LAMCAS-NEXT: ld.hu $a0, $a0, 0
+; LA64-LAMCAS-NEXT: addi.w $a2, $zero, -1
+; LA64-LAMCAS-NEXT: .p2align 4, , 16
+; LA64-LAMCAS-NEXT: .LBB33_1: # %atomicrmw.start
+; LA64-LAMCAS-NEXT: # =>This Inner Loop Header: Depth=1
+; LA64-LAMCAS-NEXT: ext.w.h $a3, $a0
+; LA64-LAMCAS-NEXT: amcas_db.h $a0, $a2, $a1
+; LA64-LAMCAS-NEXT: bne $a0, $a3, .LBB33_1
+; LA64-LAMCAS-NEXT: # %bb.2: # %atomicrmw.end
+; LA64-LAMCAS-NEXT: ret
+ %1 = atomicrmw xchg ptr %a, i16 -1 release
+ ret i16 %1
+
+}
+
+define i8 @atomicrmw_add_i8_release(ptr %a, i8 %b) nounwind {
+; LA64-LABEL: atomicrmw_add_i8_release:
+; LA64: # %bb.0:
+; LA64-NEXT: slli.d $a2, $a0, 3
+; LA64-NEXT: bstrins.d $a0, $zero, 1, 0
+; LA64-NEXT: ori $a3, $zero, 255
+; LA64-NEXT: sll.w $a3, $a3, $a2
+; LA64-NEXT: andi $a1, $a1, 255
+; LA64-NEXT: sll.w $a1, $a1, $a2
+; LA64-NEXT: .LBB34_1: # =>This Inner Loop Header: Depth=1
+; LA64-NEXT: ll.w $a4, $a0, 0
+; LA64-NEXT: add.w $a5, $a4, $a1
+; LA64-NEXT: xor $a5, $a4, $a5
+; LA64-NEXT: and $a5, $a5, $a3
+; LA64-NEXT: xor $a5, $a4, $a5
+; LA64-NEXT: sc.w $a5, $a0, 0
+; LA64-NEXT: beqz $a5, .LBB34_1
+; LA64-NEXT: # %bb.2:
+; LA64-NEXT: srl.w $a0, $a4, $a2
+; LA64-NEXT: ret
+;
+; LA64-LAMCAS-LABEL: atomicrmw_add_i8_release:
+; LA64-LAMCAS: # %bb.0:
+; LA64-LAMCAS-NEXT: move $a2, $a0
+; LA64-LAMCAS-NEXT: ld.bu $a0, $a0, 0
+; LA64-LAMCAS-NEXT: .p2align 4, , 16
+; LA64-LAMCAS-NEXT: .LBB34_1: # %atomicrmw.start
+; LA64-LAMCAS-NEXT: # =>This Inner Loop Header: Depth=1
+; LA64-LAMCAS-NEXT: add.d $a3, $a0, $a1
+; LA64-LAMCAS-NEXT: ext.w.b $a4, $a0
+; LA64-LAMCAS-NEXT: amcas_db.b $a0, $a3, $a2
+; LA64-LAMCAS-NEXT: bne $a0, $a4, .LBB34_1
+; LA64-LAMCAS-NEXT: # %bb.2: # %atomicrmw.end
+; LA64-LAMCAS-NEXT: ret
+ %1 = atomicrmw add ptr %a, i8 %b release
+ ret i8 %1
+}
+
+define i16 @atomicrmw_add_i16_release(ptr %a, i16 %b) nounwind {
+; LA64-LABEL: atomicrmw_add_i16_release:
+; LA64: # %bb.0:
+; LA64-NEXT: slli.d $a2, $a0, 3
+; LA64-NEXT: bstrins.d $a0, $zero, 1, 0
+; LA64-NEXT: lu12i.w $a3, 15
+; LA64-NEXT: ori $a3, $a3, 4095
+; LA64-NEXT: sll.w $a3, $a3, $a2
+; LA64-NEXT: bstrpick.d $a1, $a1, 15, 0
+; LA64-NEXT: sll.w $a1, $a1, $a2
+; LA64-NEXT: .LBB35_1: # =>This Inner Loop Header: Depth=1
+; LA64-NEXT: ll.w $a4, $a0, 0
+; LA64-NEXT: add.w $a5, $a4, $a1
+; LA64-NEXT: xor $a5, $a4, $a5
+; LA64-NEXT: and $a5, $a5, $a3
+; LA64-NEXT: xor $a5, $a4, $a5
+; LA64-NEXT: sc.w $a5, $a0, 0
+; LA64-NEXT: beqz $a5, .LBB35_1
+; LA64-NEXT: # %bb.2:
+; LA64-NEXT: srl.w $a0, $a4, $a2
+; LA64-NEXT: ret
+;
+; LA64-LAMCAS-LABEL: atomicrmw_add_i16_release:
+; LA64-LAMCAS: # %bb.0:
+; LA64-LAMCAS-NEXT: move $a2, $a0
+; LA64-LAMCAS-NEXT: ld.hu $a0, $a0, 0
+; LA64-LAMCAS-NEXT: .p2align 4, , 16
+; LA64-LAMCAS-NEXT: .LBB35_1: # %atomicrmw.start
+; LA64-LAMCAS-NEXT: # =>This Inner Loop Header: Depth=1
+; LA64-LAMCAS-NEXT: add.d $a3, $a0, $a1
+; LA64-LAMCAS-NEXT: ext.w.h $a4, $a0
+; LA64-LAMCAS-NEXT: amcas_db.h $a0, $a3, $a2
+; LA64-LAMCAS-NEXT: bne $a0, $a4, .LBB35_1
+; LA64-LAMCAS-NEXT: # %bb.2: # %atomicrmw.end
+; LA64-LAMCAS-NEXT: ret
+ %1 = atomicrmw add ptr %a, i16 %b release
+ ret i16 %1
+
+}
+
+define i8 @atomicrmw_sub_i8_release(ptr %a, i8 %b) nounwind {
+; LA64-LABEL: atomicrmw_sub_i8_release:
+; LA64: # %bb.0:
+; LA64-NEXT: slli.d $a2, $a0, 3
+; LA64-NEXT: bstrins.d $a0, $zero, 1, 0
+; LA64-NEXT: ori $a3, $zero, 255
+; LA64-NEXT: sll.w $a3, $a3, $a2
+; LA64-NEXT: andi $a1, $a1, 255
+; LA64-NEXT: sll.w $a1, $a1, $a2
+; LA64-NEXT: .LBB36_1: # =>This Inner Loop Header: Depth=1
+; LA64-NEXT: ll.w $a4, $a0, 0
+; LA64-NEXT: sub.w $a5, $a4, $a1
+; LA64-NEXT: xor $a5, $a4, $a5
+; LA64-NEXT: and $a5, $a5, $a3
+; LA64-NEXT: xor $a5, $a4, $a5
+; LA64-NEXT: sc.w $a5, $a0, 0
+; LA64-NEXT: beqz $a5, .LBB36_1
+; LA64-NEXT: # %bb.2:
+; LA64-NEXT: srl.w $a0, $a4, $a2
+; LA64-NEXT: ret
+;
+; LA64-LAMCAS-LABEL: atomicrmw_sub_i8_release:
+; LA64-LAMCAS: # %bb.0:
+; LA64-LAMCAS-NEXT: move $a2, $a0
+; LA64-LAMCAS-NEXT: ld.bu $a0, $a0, 0
+; LA64-LAMCAS-NEXT: .p2align 4, , 16
+; LA64-LAMCAS-NEXT: .LBB36_1: # %atomicrmw.start
+; LA64-LAMCAS-NEXT: # =>This Inner Loop Header: Depth=1
+; LA64-LAMCAS-NEXT: sub.d $a3, $a0, $a1
+; LA64-LAMCAS-NEXT: ext.w.b $a4, $a0
+; LA64-LAMCAS-NEXT: amcas_db.b $a0, $a3, $a2
+; LA64-LAMCAS-NEXT: bne $a0, $a4, .LBB36_1
+; LA64-LAMCAS-NEXT: # %bb.2: # %atomicrmw.end
+; LA64-LAMCAS-NEXT: ret
+ %1 = atomicrmw sub ptr %a, i8 %b release
+ ret i8 %1
+}
+
+define i16 @atomicrmw_sub_i16_release(ptr %a, i16 %b) nounwind {
+; LA64-LABEL: atomicrmw_sub_i16_release:
+; LA64: # %bb.0:
+; LA64-NEXT: slli.d $a2, $a0, 3
+; LA64-NEXT: bstrins.d $a0, $zero, 1, 0
+; LA64-NEXT: lu12i.w $a3, 15
+; LA64-NEXT: ori $a3, $a3, 4095
+; LA64-NEXT: sll.w $a3, $a3, $a2
+; LA64-NEXT: bstrpick.d $a1, $a1, 15, 0
+; LA64-NEXT: sll.w $a1, $a1, $a2
+; LA64-NEXT: .LBB37_1: # =>This Inner Loop Header: Depth=1
+; LA64-NEXT: ll.w $a4, $a0, 0
+; LA64-NEXT: sub.w $a5, $a4, $a1
+; LA64-NEXT: xor $a5, $a4, $a5
+; LA64-NEXT: and $a5, $a5, $a3
+; LA64-NEXT: xor $a5, $a4, $a5
+; LA64-NEXT: sc.w $a5, $a0, 0
+; LA64-NEXT: beqz $a5, .LBB37_1
+; LA64-NEXT: # %bb.2:
+; LA64-NEXT: srl.w $a0, $a4, $a2
+; LA64-NEXT: ret
+;
+; LA64-LAMCAS-LABEL: atomicrmw_sub_i16_release:
+; LA64-LAMCAS: # %bb.0:
+; LA64-LAMCAS-NEXT: move $a2, $a0
+; LA64-LAMCAS-NEXT: ld.hu $a0, $a0, 0
+; LA64-LAMCAS-NEXT: .p2align 4, , 16
+; LA64-LAMCAS-NEXT: .LBB37_1: # %atomicrmw.start
+; LA64-LAMCAS-NEXT: # =>This Inner Loop Header: Depth=1
+; LA64-LAMCAS-NEXT: sub.d $a3, $a0, $a1
+; LA64-LAMCAS-NEXT: ext.w.h $a4, $a0
+; LA64-LAMCAS-NEXT: amcas_db.h $a0, $a3, $a2
+; LA64-LAMCAS-NEXT: bne $a0, $a4, .LBB37_1
+; LA64-LAMCAS-NEXT: # %bb.2: # %atomicrmw.end
+; LA64-LAMCAS-NEXT: ret
+ %1 = atomicrmw sub ptr %a, i16 %b release
+ ret i16 %1
+
+}
+
+define i8 @atomicrmw_umax_i8_release(ptr %a, i8 %b) nounwind {
+; LA64-LABEL: atomicrmw_umax_i8_release:
+; LA64: # %bb.0:
+; LA64-NEXT: slli.d $a2, $a0, 3
+; LA64-NEXT: bstrins.d $a0, $zero, 1, 0
+; LA64-NEXT: ori $a3, $zero, 255
+; LA64-NEXT: sll.w $a3, $a3, $a2
+; LA64-NEXT: andi $a1, $a1, 255
+; LA64-NEXT: sll.w $a1, $a1, $a2
+; LA64-NEXT: .LBB38_1: # =>This Inner Loop Header: Depth=1
+; LA64-NEXT: ll.w $a4, $a0, 0
+; LA64-NEXT: and $a6, $a4, $a3
+; LA64-NEXT: move $a5, $a4
+; LA64-NEXT: bgeu $a6, $a1, .LBB38_3
+; LA64-NEXT: # %bb.2: # in Loop: Header=BB38_1 Depth=1
+; LA64-NEXT: xor $a5, $a4, $a1
+; LA64-NEXT: and $a5, $a5, $a3
+; LA64-NEXT: xor $a5, $a4, $a5
+; LA64-NEXT: .LBB38_3: # in Loop: Header=BB38_1 Depth=1
+; LA64-NEXT: sc.w $a5, $a0, 0
+; LA64-NEXT: beqz $a5, .LBB38_1
+; LA64-NEXT: # %bb.4:
+; LA64-NEXT: srl.w $a0, $a4, $a2
+; LA64-NEXT: ret
+;
+; LA64-LAMCAS-LABEL: atomicrmw_umax_i8_release:
+; LA64-LAMCAS: # %bb.0:
+; LA64-LAMCAS-NEXT: move $a2, $a0
+; LA64-LAMCAS-NEXT: ld.bu $a0, $a0, 0
+; LA64-LAMCAS-NEXT: andi $a3, $a1, 255
+; LA64-LAMCAS-NEXT: .p2align 4, , 16
+; LA64-LAMCAS-NEXT: .LBB38_1: # %atomicrmw.start
+; LA64-LAMCAS-NEXT: # =>This Inner Loop Header: Depth=1
+; LA64-LAMCAS-NEXT: andi $a4, $a0, 255
+; LA64-LAMCAS-NEXT: sltu $a4, $a3, $a4
+; LA64-LAMCAS-NEXT: masknez $a5, $a1, $a4
+; LA64-LAMCAS-NEXT: maskeqz $a4, $a0, $a4
+; LA64-LAMCAS-NEXT: or $a4, $a4, $a5
+; LA64-LAMCAS-NEXT: ext.w.b $a5, $a0
+; LA64-LAMCAS-NEXT: amcas_db.b $a0, $a4, $a2
+; LA64-LAMCAS-NEXT: bne $a0, $a5, .LBB38_1
+; LA64-LAMCAS-NEXT: # %bb.2: # %atomicrmw.end
+; LA64-LAMCAS-NEXT: ret
+ %1 = atomicrmw umax ptr %a, i8 %b release
+ ret i8 %1
+}
+
+define i16 @atomicrmw_umax_i16_release(ptr %a, i16 %b) nounwind {
+; LA64-LABEL: atomicrmw_umax_i16_release:
+; LA64: # %bb.0:
+; LA64-NEXT: slli.d $a2, $a0, 3
+; LA64-NEXT: bstrins.d $a0, $zero, 1, 0
+; LA64-NEXT: lu12i.w $a3, 15
+; LA64-NEXT: ori $a3, $a3, 4095
+; LA64-NEXT: sll.w $a3, $a3, $a2
+; LA64-NEXT: bstrpick.d $a1, $a1, 15, 0
+; LA64-NEXT: sll.w $a1, $a1, $a2
+; LA64-NEXT: .LBB39_1: # =>This Inner Loop Header: Depth=1
+; LA64-NEXT: ll.w $a4, $a0, 0
+; LA64-NEXT: and $a6, $a4, $a3
+; LA64-NEXT: move $a5, $a4
+; LA64-NEXT: bgeu $a6, $a1, .LBB39_3
+; LA64-NEXT: # %bb.2: # in Loop: Header=BB39_1 Depth=1
+; LA64-NEXT: xor $a5, $a4, $a1
+; LA64-NEXT: and $a5, $a5, $a3
+; LA64-NEXT: xor $a5, $a4, $a5
+; LA64-NEXT: .LBB39_3: # in Loop: Header=BB39_1 Depth=1
+; LA64-NEXT: sc.w $a5, $a0, 0
+; LA64-NEXT: beqz $a5, .LBB39_1
+; LA64-NEXT: # %bb.4:
+; LA64-NEXT: srl.w $a0, $a4, $a2
+; LA64-NEXT: ret
+;
+; LA64-LAMCAS-LABEL: atomicrmw_umax_i16_release:
+; LA64-LAMCAS: # %bb.0:
+; LA64-LAMCAS-NEXT: move $a2, $a0
+; LA64-LAMCAS-NEXT: ld.hu $a0, $a0, 0
+; LA64-LAMCAS-NEXT: bstrpick.d $a3, $a1, 15, 0
+; LA64-LAMCAS-NEXT: .p2align 4, , 16
+; LA64-LAMCAS-NEXT: .LBB39_1: # %atomicrmw.start
+; LA64-LAMCAS-NEXT: # =>This Inner Loop Header: Depth=1
+; LA64-LAMCAS-NEXT: bstrpick.d $a4, $a0, 15, 0
+; LA64-LAMCAS-NEXT: sltu $a4, $a3, $a4
+; LA64-LAMCAS-NEXT: masknez $a5, $a1, $a4
+; LA64-LAMCAS-NEXT: maskeqz $a4, $a0, $a4
+; LA64-LAMCAS-NEXT: or $a4, $a4, $a5
+; LA64-LAMCAS-NEXT: ext.w.h $a5, $a0
+; LA64-LAMCAS-NEXT: amcas_db.h $a0, $a4, $a2
+; LA64-LAMCAS-NEXT: bne $a0, $a5, .LBB39_1
+; LA64-LAMCAS-NEXT: # %bb.2: # %atomicrmw.end
+; LA64-LAMCAS-NEXT: ret
+ %1 = atomicrmw umax ptr %a, i16 %b release
+ ret i16 %1
+}
+
+define i8 @atomicrmw_umin_i8_release(ptr %a, i8 %b) nounwind {
+; LA64-LABEL: atomicrmw_umin_i8_release:
+; LA64: # %bb.0:
+; LA64-NEXT: slli.d $a2, $a0, 3
+; LA64-NEXT: bstrins.d $a0, $zero, 1, 0
+; LA64-NEXT: ori $a3, $zero, 255
+; LA64-NEXT: sll.w $a3, $a3, $a2
+; LA64-NEXT: andi $a1, $a1, 255
+; LA64-NEXT: sll.w $a1, $a1, $a2
+; LA64-NEXT: .LBB40_1: # =>This Inner Loop Header: Depth=1
+; LA64-NEXT: ll.w $a4, $a0, 0
+; LA64-NEXT: and $a6, $a4, $a3
+; LA64-NEXT: move $a5, $a4
+; LA64-NEXT: bgeu $a1, $a6, .LBB40_3
+; LA64-NEXT: # %bb.2: # in Loop: Header=BB40_1 Depth=1
+; LA64-NEXT: xor $a5, $a4, $a1
+; LA64-NEXT: and $a5, $a5, $a3
+; LA64-NEXT: xor $a5, $a4, $a5
+; LA64-NEXT: .LBB40_3: # in Loop: Header=BB40_1 Depth=1
+; LA64-NEXT: sc.w $a5, $a0, 0
+; LA64-NEXT: beqz $a5, .LBB40_1
+; LA64-NEXT: # %bb.4:
+; LA64-NEXT: srl.w $a0, $a4, $a2
+; LA64-NEXT: ret
+;
+; LA64-LAMCAS-LABEL: atomicrmw_umin_i8_release:
+; LA64-LAMCAS: # %bb.0:
+; LA64-LAMCAS-NEXT: move $a2, $a0
+; LA64-LAMCAS-NEXT: ld.bu $a0, $a0, 0
+; LA64-LAMCAS-NEXT: andi $a3, $a1, 255
+; LA64-LAMCAS-NEXT: .p2align 4, , 16
+; LA64-LAMCAS-NEXT: .LBB40_1: # %atomicrmw.start
+; LA64-LAMCAS-NEXT: # =>This Inner Loop Header: Depth=1
+; LA64-LAMCAS-NEXT: andi $a4, $a0, 255
+; LA64-LAMCAS-NEXT: sltu $a4, $a3, $a4
+; LA64-LAMCAS-NEXT: xori $a4, $a4, 1
+; LA64-LAMCAS-NEXT: masknez $a5, $a1, $a4
+; LA64-LAMCAS-NEXT: maskeqz $a4, $a0, $a4
+; LA64-LAMCAS-NEXT: or $a4, $a4, $a5
+; LA64-LAMCAS-NEXT: ext.w.b $a5, $a0
+; LA64-LAMCAS-NEXT: amcas_db.b $a0, $a4, $a2
+; LA64-LAMCAS-NEXT: bne $a0, $a5, .LBB40_1
+; LA64-LAMCAS-NEXT: # %bb.2: # %atomicrmw.end
+; LA64-LAMCAS-NEXT: ret
+ %1 = atomicrmw umin ptr %a, i8 %b release
+ ret i8 %1
+}
+
+define i16 @atomicrmw_umin_i16_release(ptr %a, i16 %b) nounwind {
+; LA64-LABEL: atomicrmw_umin_i16_release:
+; LA64: # %bb.0:
+; LA64-NEXT: slli.d $a2, $a0, 3
+; LA64-NEXT: bstrins.d $a0, $zero, 1, 0
+; LA64-NEXT: lu12i.w $a3, 15
+; LA64-NEXT: ori $a3, $a3, 4095
+; LA64-NEXT: sll.w $a3, $a3, $a2
+; LA64-NEXT: bstrpick.d $a1, $a1, 15, 0
+; LA64-NEXT: sll.w $a1, $a1, $a2
+; LA64-NEXT: .LBB41_1: # =>This Inner Loop Header: Depth=1
+; LA64-NEXT: ll.w $a4, $a0, 0
+; LA64-NEXT: and $a6, $a4, $a3
+; LA64-NEXT: move $a5, $a4
+; LA64-NEXT: bgeu $a1, $a6, .LBB41_3
+; LA64-NEXT: # %bb.2: # in Loop: Header=BB41_1 Depth=1
+; LA64-NEXT: xor $a5, $a4, $a1
+; LA64-NEXT: and $a5, $a5, $a3
+; LA64-NEXT: xor $a5, $a4, $a5
+; LA64-NEXT: .LBB41_3: # in Loop: Header=BB41_1 Depth=1
+; LA64-NEXT: sc.w $a5, $a0, 0
+; LA64-NEXT: beqz $a5, .LBB41_1
+; LA64-NEXT: # %bb.4:
+; LA64-NEXT: srl.w $a0, $a4, $a2
+; LA64-NEXT: ret
+;
+; LA64-LAMCAS-LABEL: atomicrmw_umin_i16_release:
+; LA64-LAMCAS: # %bb.0:
+; LA64-LAMCAS-NEXT: move $a2, $a0
+; LA64-LAMCAS-NEXT: ld.hu $a0, $a0, 0
+; LA64-LAMCAS-NEXT: bstrpick.d $a3, $a1, 15, 0
+; LA64-LAMCAS-NEXT: .p2align 4, , 16
+; LA64-LAMCAS-NEXT: .LBB41_1: # %atomicrmw.start
+; LA64-LAMCAS-NEXT: # =>This Inner Loop Header: Depth=1
+; LA64-LAMCAS-NEXT: bstrpick.d $a4, $a0, 15, 0
+; LA64-LAMCAS-NEXT: sltu $a4, $a3, $a4
+; LA64-LAMCAS-NEXT: xori $a4, $a4, 1
+; LA64-LAMCAS-NEXT: masknez $a5, $a1, $a4
+; LA64-LAMCAS-NEXT: maskeqz $a4, $a0, $a4
+; LA64-LAMCAS-NEXT: or $a4, $a4, $a5
+; LA64-LAMCAS-NEXT: ext.w.h $a5, $a0
+; LA64-LAMCAS-NEXT: amcas_db.h $a0, $a4, $a2
+; LA64-LAMCAS-NEXT: bne $a0, $a5, .LBB41_1
+; LA64-LAMCAS-NEXT: # %bb.2: # %atomicrmw.end
+; LA64-LAMCAS-NEXT: ret
+ %1 = atomicrmw umin ptr %a, i16 %b release
+ ret i16 %1
+}
+
+define i8 @atomicrmw_max_i8_release(ptr %a, i8 %b) nounwind {
+; LA64-LABEL: atomicrmw_max_i8_release:
+; LA64: # %bb.0:
+; LA64-NEXT: slli.d $a2, $a0, 3
+; LA64-NEXT: bstrins.d $a0, $zero, 1, 0
+; LA64-NEXT: andi $a3, $a2, 24
+; LA64-NEXT: ori $a4, $zero, 255
+; LA64-NEXT: sll.w $a4, $a4, $a2
+; LA64-NEXT: ext.w.b $a1, $a1
+; LA64-NEXT: sll.w $a1, $a1, $a2
+; LA64-NEXT: xori $a3, $a3, 56
+; LA64-NEXT: .LBB42_1: # =>This Inner Loop Header: Depth=1
+; LA64-NEXT: ll.w $a5, $a0, 0
+; LA64-NEXT: and $a7, $a5, $a4
+; LA64-NEXT: move $a6, $a5
+; LA64-NEXT: sll.w $a7, $a7, $a3
+; LA64-NEXT: sra.w $a7, $a7, $a3
+; LA64-NEXT: bge $a7, $a1, .LBB42_3
+; LA64-NEXT: # %bb.2: # in Loop: Header=BB42_1 Depth=1
+; LA64-NEXT: xor $a6, $a5, $a1
+; LA64-NEXT: and $a6, $a6, $a4
+; LA64-NEXT: xor $a6, $a5, $a6
+; LA64-NEXT: .LBB42_3: # in Loop: Header=BB42_1 Depth=1
+; LA64-NEXT: sc.w $a6, $a0, 0
+; LA64-NEXT: beqz $a6, .LBB42_1
+; LA64-NEXT: # %bb.4:
+; LA64-NEXT: srl.w $a0, $a5, $a2
+; LA64-NEXT: ret
+;
+; LA64-LAMCAS-LABEL: atomicrmw_max_i8_release:
+; LA64-LAMCAS: # %bb.0:
+; LA64-LAMCAS-NEXT: move $a2, $a0
+; LA64-LAMCAS-NEXT: ld.bu $a0, $a0, 0
+; LA64-LAMCAS-NEXT: ext.w.b $a3, $a1
+; LA64-LAMCAS-NEXT: .p2align 4, , 16
+; LA64-LAMCAS-NEXT: .LBB42_1: # %atomicrmw.start
+; LA64-LAMCAS-NEXT: # =>This Inner Loop Header: Depth=1
+; LA64-LAMCAS-NEXT: ext.w.b $a4, $a0
+; LA64-LAMCAS-NEXT: slt $a5, $a3, $a4
+; LA64-LAMCAS-NEXT: masknez $a6, $a1, $a5
+; LA64-LAMCAS-NEXT: maskeqz $a5, $a0, $a5
+; LA64-LAMCAS-NEXT: or $a5, $a5, $a6
+; LA64-LAMCAS-NEXT: amcas_db.b $a0, $a5, $a2
+; LA64-LAMCAS-NEXT: bne $a0, $a4, .LBB42_1
+; LA64-LAMCAS-NEXT: # %bb.2: # %atomicrmw.end
+; LA64-LAMCAS-NEXT: ret
+ %1 = atomicrmw max ptr %a, i8 %b release
+ ret i8 %1
+}
+
+define i16 @atomicrmw_max_i16_release(ptr %a, i16 %b) nounwind {
+; LA64-LABEL: atomicrmw_max_i16_release:
+; LA64: # %bb.0:
+; LA64-NEXT: slli.d $a2, $a0, 3
+; LA64-NEXT: bstrins.d $a0, $zero, 1, 0
+; LA64-NEXT: andi $a3, $a2, 24
+; LA64-NEXT: lu12i.w $a4, 15
+; LA64-NEXT: ori $a4, $a4, 4095
+; LA64-NEXT: sll.w $a4, $a4, $a2
+; LA64-NEXT: ext.w.h $a1, $a1
+; LA64-NEXT: sll.w $a1, $a1, $a2
+; LA64-NEXT: ori $a5, $zero, 48
+; LA64-NEXT: sub.d $a3, $a5, $a3
+; LA64-NEXT: .LBB43_1: # =>This Inner Loop Header: Depth=1
+; LA64-NEXT: ll.w $a5, $a0, 0
+; LA64-NEXT: and $a7, $a5, $a4
+; LA64-NEXT: move $a6, $a5
+; LA64-NEXT: sll.w $a7, $a7, $a3
+; LA64-NEXT: sra.w $a7, $a7, $a3
+; LA64-NEXT: bge $a7, $a1, .LBB43_3
+; LA64-NEXT: # %bb.2: # in Loop: Header=BB43_1 Depth=1
+; LA64-NEXT: xor $a6, $a5, $a1
+; LA64-NEXT: and $a6, $a6, $a4
+; LA64-NEXT: xor $a6, $a5, $a6
+; LA64-NEXT: .LBB43_3: # in Loop: Header=BB43_1 Depth=1
+; LA64-NEXT: sc.w $a6, $a0, 0
+; LA64-NEXT: beqz $a6, .LBB43_1
+; LA64-NEXT: # %bb.4:
+; LA64-NEXT: srl.w $a0, $a5, $a2
+; LA64-NEXT: ret
+;
+; LA64-LAMCAS-LABEL: atomicrmw_max_i16_release:
+; LA64-LAMCAS: # %bb.0:
+; LA64-LAMCAS-NEXT: move $a2, $a0
+; LA64-LAMCAS-NEXT: ld.hu $a0, $a0, 0
+; LA64-LAMCAS-NEXT: ext.w.h $a3, $a1
+; LA64-LAMCAS-NEXT: .p2align 4, , 16
+; LA64-LAMCAS-NEXT: .LBB43_1: # %atomicrmw.start
+; LA64-LAMCAS-NEXT: # =>This Inner Loop Header: Depth=1
+; LA64-LAMCAS-NEXT: ext.w.h $a4, $a0
+; LA64-LAMCAS-NEXT: slt $a5, $a3, $a4
+; LA64-LAMCAS-NEXT: masknez $a6, $a1, $a5
+; LA64-LAMCAS-NEXT: maskeqz $a5, $a0, $a5
+; LA64-LAMCAS-NEXT: or $a5, $a5, $a6
+; LA64-LAMCAS-NEXT: amcas_db.h $a0, $a5, $a2
+; LA64-LAMCAS-NEXT: bne $a0, $a4, .LBB43_1
+; LA64-LAMCAS-NEXT: # %bb.2: # %atomicrmw.end
+; LA64-LAMCAS-NEXT: ret
+ %1 = atomicrmw max ptr %a, i16 %b release
+ ret i16 %1
+}
+
+define i8 @atomicrmw_min_i8_release(ptr %a, i8 %b) nounwind {
+; LA64-LABEL: atomicrmw_min_i8_release:
+; LA64: # %bb.0:
+; LA64-NEXT: slli.d $a2, $a0, 3
+; LA64-NEXT: bstrins.d $a0, $zero, 1, 0
+; LA64-NEXT: andi $a3, $a2, 24
+; LA64-NEXT: ori $a4, $zero, 255
+; LA64-NEXT: sll.w $a4, $a4, $a2
+; LA64-NEXT: ext.w.b $a1, $a1
+; LA64-NEXT: sll.w $a1, $a1, $a2
+; LA64-NEXT: xori $a3, $a3, 56
+; LA64-NEXT: .LBB44_1: # =>This Inner Loop Header: Depth=1
+; LA64-NEXT: ll.w $a5, $a0, 0
+; LA64-NEXT: and $a7, $a5, $a4
+; LA64-NEXT: move $a6, $a5
+; LA64-NEXT: sll.w $a7, $a7, $a3
+; LA64-NEXT: sra.w $a7, $a7, $a3
+; LA64-NEXT: bge $a1, $a7, .LBB44_3
+; LA64-NEXT: # %bb.2: # in Loop: Header=BB44_1 Depth=1
+; LA64-NEXT: xor $a6, $a5, $a1
+; LA64-NEXT: and $a6, $a6, $a4
+; LA64-NEXT: xor $a6, $a5, $a6
+; LA64-NEXT: .LBB44_3: # in Loop: Header=BB44_1 Depth=1
+; LA64-NEXT: sc.w $a6, $a0, 0
+; LA64-NEXT: beqz $a6, .LBB44_1
+; LA64-NEXT: # %bb.4:
+; LA64-NEXT: srl.w $a0, $a5, $a2
+; LA64-NEXT: ret
+;
+; LA64-LAMCAS-LABEL: atomicrmw_min_i8_release:
+; LA64-LAMCAS: # %bb.0:
+; LA64-LAMCAS-NEXT: move $a2, $a0
+; LA64-LAMCAS-NEXT: ld.bu $a0, $a0, 0
+; LA64-LAMCAS-NEXT: ext.w.b $a3, $a1
+; LA64-LAMCAS-NEXT: .p2align 4, , 16
+; LA64-LAMCAS-NEXT: .LBB44_1: # %atomicrmw.start
+; LA64-LAMCAS-NEXT: # =>This Inner Loop Header: Depth=1
+; LA64-LAMCAS-NEXT: ext.w.b $a4, $a0
+; LA64-LAMCAS-NEXT: slt $a5, $a3, $a4
+; LA64-LAMCAS-NEXT: xori $a5, $a5, 1
+; LA64-LAMCAS-NEXT: masknez $a6, $a1, $a5
+; LA64-LAMCAS-NEXT: maskeqz $a5, $a0, $a5
+; LA64-LAMCAS-NEXT: or $a5, $a5, $a6
+; LA64-LAMCAS-NEXT: amcas_db.b $a0, $a5, $a2
+; LA64-LAMCAS-NEXT: bne $a0, $a4, .LBB44_1
+; LA64-LAMCAS-NEXT: # %bb.2: # %atomicrmw.end
+; LA64-LAMCAS-NEXT: ret
+ %1 = atomicrmw min ptr %a, i8 %b release
+ ret i8 %1
+}
+
+define i16 @atomicrmw_min_i16_release(ptr %a, i16 %b) nounwind {
+; LA64-LABEL: atomicrmw_min_i16_release:
+; LA64: # %bb.0:
+; LA64-NEXT: slli.d $a2, $a0, 3
+; LA64-NEXT: bstrins.d $a0, $zero, 1, 0
+; LA64-NEXT: andi $a3, $a2, 24
+; LA64-NEXT: lu12i.w $a4, 15
+; LA64-NEXT: ori $a4, $a4, 4095
+; LA64-NEXT: sll.w $a4, $a4, $a2
+; LA64-NEXT: ext.w.h $a1, $a1
+; LA64-NEXT: sll.w $a1, $a1, $a2
+; LA64-NEXT: ori $a5, $zero, 48
+; LA64-NEXT: sub.d $a3, $a5, $a3
+; LA64-NEXT: .LBB45_1: # =>This Inner Loop Header: Depth=1
+; LA64-NEXT: ll.w $a5, $a0, 0
+; LA64-NEXT: and $a7, $a5, $a4
+; LA64-NEXT: move $a6, $a5
+; LA64-NEXT: sll.w $a7, $a7, $a3
+; LA64-NEXT: sra.w $a7, $a7, $a3
+; LA64-NEXT: bge $a1, $a7, .LBB45_3
+; LA64-NEXT: # %bb.2: # in Loop: Header=BB45_1 Depth=1
+; LA64-NEXT: xor $a6, $a5, $a1
+; LA64-NEXT: and $a6, $a6, $a4
+; LA64-NEXT: xor $a6, $a5, $a6
+; LA64-NEXT: .LBB45_3: # in Loop: Header=BB45_1 Depth=1
+; LA64-NEXT: sc.w $a6, $a0, 0
+; LA64-NEXT: beqz $a6, .LBB45_1
+; LA64-NEXT: # %bb.4:
+; LA64-NEXT: srl.w $a0, $a5, $a2
+; LA64-NEXT: ret
+;
+; LA64-LAMCAS-LABEL: atomicrmw_min_i16_release:
+; LA64-LAMCAS: # %bb.0:
+; LA64-LAMCAS-NEXT: move $a2, $a0
+; LA64-LAMCAS-NEXT: ld.hu $a0, $a0, 0
+; LA64-LAMCAS-NEXT: ext.w.h $a3, $a1
+; LA64-LAMCAS-NEXT: .p2align 4, , 16
+; LA64-LAMCAS-NEXT: .LBB45_1: # %atomicrmw.start
+; LA64-LAMCAS-NEXT: # =>This Inner Loop Header: Depth=1
+; LA64-LAMCAS-NEXT: ext.w.h $a4, $a0
+; LA64-LAMCAS-NEXT: slt $a5, $a3, $a4
+; LA64-LAMCAS-NEXT: xori $a5, $a5, 1
+; LA64-LAMCAS-NEXT: masknez $a6, $a1, $a5
+; LA64-LAMCAS-NEXT: maskeqz $a5, $a0, $a5
+; LA64-LAMCAS-NEXT: or $a5, $a5, $a6
+; LA64-LAMCAS-NEXT: amcas_db.h $a0, $a5, $a2
+; LA64-LAMCAS-NEXT: bne $a0, $a4, .LBB45_1
+; LA64-LAMCAS-NEXT: # %bb.2: # %atomicrmw.end
+; LA64-LAMCAS-NEXT: ret
+ %1 = atomicrmw min ptr %a, i16 %b release
+ ret i16 %1
+}
+
+define i8 @atomicrmw_nand_i8_release(ptr %a, i8 %b) nounwind {
+; LA64-LABEL: atomicrmw_nand_i8_release:
+; LA64: # %bb.0:
+; LA64-NEXT: slli.d $a2, $a0, 3
+; LA64-NEXT: bstrins.d $a0, $zero, 1, 0
+; LA64-NEXT: ori $a3, $zero, 255
+; LA64-NEXT: sll.w $a3, $a3, $a2
+; LA64-NEXT: andi $a1, $a1, 255
+; LA64-NEXT: sll.w $a1, $a1, $a2
+; LA64-NEXT: .LBB46_1: # =>This Inner Loop Header: Depth=1
+; LA64-NEXT: ll.w $a4, $a0, 0
+; LA64-NEXT: and $a5, $a4, $a1
+; LA64-NEXT: nor $a5, $a5, $zero
+; LA64-NEXT: xor $a5, $a4, $a5
+; LA64-NEXT: and $a5, $a5, $a3
+; LA64-NEXT: xor $a5, $a4, $a5
+; LA64-NEXT: sc.w $a5, $a0, 0
+; LA64-NEXT: beqz $a5, .LBB46_1
+; LA64-NEXT: # %bb.2:
+; LA64-NEXT: srl.w $a0, $a4, $a2
+; LA64-NEXT: ret
+;
+; LA64-LAMCAS-LABEL: atomicrmw_nand_i8_release:
+; LA64-LAMCAS: # %bb.0:
+; LA64-LAMCAS-NEXT: move $a2, $a0
+; LA64-LAMCAS-NEXT: ld.bu $a0, $a0, 0
+; LA64-LAMCAS-NEXT: .p2align 4, , 16
+; LA64-LAMCAS-NEXT: .LBB46_1: # %atomicrmw.start
+; LA64-LAMCAS-NEXT: # =>This Inner Loop Header: Depth=1
+; LA64-LAMCAS-NEXT: and $a3, $a0, $a1
+; LA64-LAMCAS-NEXT: nor $a3, $a3, $zero
+; LA64-LAMCAS-NEXT: ext.w.b $a4, $a0
+; LA64-LAMCAS-NEXT: amcas_db.b $a0, $a3, $a2
+; LA64-LAMCAS-NEXT: bne $a0, $a4, .LBB46_1
+; LA64-LAMCAS-NEXT: # %bb.2: # %atomicrmw.end
+; LA64-LAMCAS-NEXT: ret
+ %1 = atomicrmw nand ptr %a, i8 %b release
+ ret i8 %1
+}
+
+define i16 @atomicrmw_nand_i16_release(ptr %a, i16 %b) nounwind {
+; LA64-LABEL: atomicrmw_nand_i16_release:
+; LA64: # %bb.0:
+; LA64-NEXT: slli.d $a2, $a0, 3
+; LA64-NEXT: bstrins.d $a0, $zero, 1, 0
+; LA64-NEXT: lu12i.w $a3, 15
+; LA64-NEXT: ori $a3, $a3, 4095
+; LA64-NEXT: sll.w $a3, $a3, $a2
+; LA64-NEXT: bstrpick.d $a1, $a1, 15, 0
+; LA64-NEXT: sll.w $a1, $a1, $a2
+; LA64-NEXT: .LBB47_1: # =>This Inner Loop Header: Depth=1
+; LA64-NEXT: ll.w $a4, $a0, 0
+; LA64-NEXT: and $a5, $a4, $a1
+; LA64-NEXT: nor $a5, $a5, $zero
+; LA64-NEXT: xor $a5, $a4, $a5
+; LA64-NEXT: and $a5, $a5, $a3
+; LA64-NEXT: xor $a5, $a4, $a5
+; LA64-NEXT: sc.w $a5, $a0, 0
+; LA64-NEXT: beqz $a5, .LBB47_1
+; LA64-NEXT: # %bb.2:
+; LA64-NEXT: srl.w $a0, $a4, $a2
+; LA64-NEXT: ret
+;
+; LA64-LAMCAS-LABEL: atomicrmw_nand_i16_release:
+; LA64-LAMCAS: # %bb.0:
+; LA64-LAMCAS-NEXT: move $a2, $a0
+; LA64-LAMCAS-NEXT: ld.hu $a0, $a0, 0
+; LA64-LAMCAS-NEXT: .p2align 4, , 16
+; LA64-LAMCAS-NEXT: .LBB47_1: # %atomicrmw.start
+; LA64-LAMCAS-NEXT: # =>This Inner Loop Header: Depth=1
+; LA64-LAMCAS-NEXT: and $a3, $a0, $a1
+; LA64-LAMCAS-NEXT: nor $a3, $a3, $zero
+; LA64-LAMCAS-NEXT: ext.w.h $a4, $a0
+; LA64-LAMCAS-NEXT: amcas_db.h $a0, $a3, $a2
+; LA64-LAMCAS-NEXT: bne $a0, $a4, .LBB47_1
+; LA64-LAMCAS-NEXT: # %bb.2: # %atomicrmw.end
+; LA64-LAMCAS-NEXT: ret
+ %1 = atomicrmw nand ptr %a, i16 %b release
+ ret i16 %1
+
+}
+
+define i32 @atomicrmw_nand_i32_release(ptr %a, i32 %b) nounwind {
+; LA64-LABEL: atomicrmw_nand_i32_release:
+; LA64: # %bb.0:
+; LA64-NEXT: .LBB48_1: # =>This Inner Loop Header: Depth=1
+; LA64-NEXT: ll.w $a2, $a0, 0
+; LA64-NEXT: and $a3, $a2, $a1
+; LA64-NEXT: nor $a3, $a3, $zero
+; LA64-NEXT: sc.w $a3, $a0, 0
+; LA64-NEXT: beqz $a3, .LBB48_1
+; LA64-NEXT: # %bb.2:
+; LA64-NEXT: move $a0, $a2
+; LA64-NEXT: ret
+;
+; LA64-LAMCAS-LABEL: atomicrmw_nand_i32_release:
+; LA64-LAMCAS: # %bb.0:
+; LA64-LAMCAS-NEXT: move $a2, $a0
+; LA64-LAMCAS-NEXT: ld.w $a0, $a0, 0
+; LA64-LAMCAS-NEXT: .p2align 4, , 16
+; LA64-LAMCAS-NEXT: .LBB48_1: # %atomicrmw.start
+; LA64-LAMCAS-NEXT: # =>This Inner Loop Header: Depth=1
+; LA64-LAMCAS-NEXT: move $a3, $a0
+; LA64-LAMCAS-NEXT: and $a4, $a0, $a1
+; LA64-LAMCAS-NEXT: nor $a4, $a4, $zero
+; LA64-LAMCAS-NEXT: amcas_db.w $a0, $a4, $a2
+; LA64-LAMCAS-NEXT: bne $a0, $a3, .LBB48_1
+; LA64-LAMCAS-NEXT: # %bb.2: # %atomicrmw.end
+; LA64-LAMCAS-NEXT: ret
+ %1 = atomicrmw nand ptr %a, i32 %b release
+ ret i32 %1
+}
+
+define i64 @atomicrmw_nand_i64_release(ptr %a, i64 %b) nounwind {
+; LA64-LABEL: atomicrmw_nand_i64_release:
+; LA64: # %bb.0:
+; LA64-NEXT: .LBB49_1: # =>This Inner Loop Header: Depth=1
+; LA64-NEXT: ll.d $a2, $a0, 0
+; LA64-NEXT: and $a3, $a2, $a1
+; LA64-NEXT: nor $a3, $a3, $zero
+; LA64-NEXT: sc.d $a3, $a0, 0
+; LA64-NEXT: beqz $a3, .LBB49_1
+; LA64-NEXT: # %bb.2:
+; LA64-NEXT: move $a0, $a2
+; LA64-NEXT: ret
+;
+; LA64-LAMCAS-LABEL: atomicrmw_nand_i64_release:
+; LA64-LAMCAS: # %bb.0:
+; LA64-LAMCAS-NEXT: move $a2, $a0
+; LA64-LAMCAS-NEXT: ld.d $a0, $a0, 0
+; LA64-LAMCAS-NEXT: .p2align 4, , 16
+; LA64-LAMCAS-NEXT: .LBB49_1: # %atomicrmw.start
+; LA64-LAMCAS-NEXT: # =>This Inner Loop Header: Depth=1
+; LA64-LAMCAS-NEXT: move $a3, $a0
+; LA64-LAMCAS-NEXT: and $a4, $a0, $a1
+; LA64-LAMCAS-NEXT: nor $a4, $a4, $zero
+; LA64-LAMCAS-NEXT: amcas_db.d $a0, $a4, $a2
+; LA64-LAMCAS-NEXT: bne $a0, $a3, .LBB49_1
+; LA64-LAMCAS-NEXT: # %bb.2: # %atomicrmw.end
+; LA64-LAMCAS-NEXT: ret
+ %1 = atomicrmw nand ptr %a, i64 %b release
+ ret i64 %1
+}
+
+define i8 @atomicrmw_and_i8_release(ptr %a, i8 %b) nounwind {
+; LA64-LABEL: atomicrmw_and_i8_release:
+; LA64: # %bb.0:
+; LA64-NEXT: slli.d $a2, $a0, 3
+; LA64-NEXT: bstrins.d $a0, $zero, 1, 0
+; LA64-NEXT: ori $a3, $zero, 255
+; LA64-NEXT: sll.w $a3, $a3, $a2
+; LA64-NEXT: andi $a1, $a1, 255
+; LA64-NEXT: sll.w $a1, $a1, $a2
+; LA64-NEXT: orn $a1, $a1, $a3
+; LA64-NEXT: amand_db.w $a3, $a1, $a0
+; LA64-NEXT: srl.w $a0, $a3, $a2
+; LA64-NEXT: ret
+;
+; LA64-LAMCAS-LABEL: atomicrmw_and_i8_release:
+; LA64-LAMCAS: # %bb.0:
+; LA64-LAMCAS-NEXT: move $a2, $a0
+; LA64-LAMCAS-NEXT: ld.bu $a0, $a0, 0
+; LA64-LAMCAS-NEXT: .p2align 4, , 16
+; LA64-LAMCAS-NEXT: .LBB50_1: # %atomicrmw.start
+; LA64-LAMCAS-NEXT: # =>This Inner Loop Header: Depth=1
+; LA64-LAMCAS-NEXT: and $a3, $a0, $a1
+; LA64-LAMCAS-NEXT: ext.w.b $a4, $a0
+; LA64-LAMCAS-NEXT: amcas_db.b $a0, $a3, $a2
+; LA64-LAMCAS-NEXT: bne $a0, $a4, .LBB50_1
+; LA64-LAMCAS-NEXT: # %bb.2: # %atomicrmw.end
+; LA64-LAMCAS-NEXT: ret
+ %1 = atomicrmw and ptr %a, i8 %b release
+ ret i8 %1
+}
+
+define i16 @atomicrmw_and_i16_release(ptr %a, i16 %b) nounwind {
+; LA64-LABEL: atomicrmw_and_i16_release:
+; LA64: # %bb.0:
+; LA64-NEXT: slli.d $a2, $a0, 3
+; LA64-NEXT: bstrins.d $a0, $zero, 1, 0
+; LA64-NEXT: lu12i.w $a3, 15
+; LA64-NEXT: ori $a3, $a3, 4095
+; LA64-NEXT: sll.w $a3, $a3, $a2
+; LA64-NEXT: bstrpick.d $a1, $a1, 15, 0
+; LA64-NEXT: sll.w $a1, $a1, $a2
+; LA64-NEXT: orn $a1, $a1, $a3
+; LA64-NEXT: amand_db.w $a3, $a1, $a0
+; LA64-NEXT: srl.w $a0, $a3, $a2
+; LA64-NEXT: ret
+;
+; LA64-LAMCAS-LABEL: atomicrmw_and_i16_release:
+; LA64-LAMCAS: # %bb.0:
+; LA64-LAMCAS-NEXT: move $a2, $a0
+; LA64-LAMCAS-NEXT: ld.hu $a0, $a0, 0
+; LA64-LAMCAS-NEXT: .p2align 4, , 16
+; LA64-LAMCAS-NEXT: .LBB51_1: # %atomicrmw.start
+; LA64-LAMCAS-NEXT: # =>This Inner Loop Header: Depth=1
+; LA64-LAMCAS-NEXT: and $a3, $a0, $a1
+; LA64-LAMCAS-NEXT: ext.w.h $a4, $a0
+; LA64-LAMCAS-NEXT: amcas_db.h $a0, $a3, $a2
+; LA64-LAMCAS-NEXT: bne $a0, $a4, .LBB51_1
+; LA64-LAMCAS-NEXT: # %bb.2: # %atomicrmw.end
+; LA64-LAMCAS-NEXT: ret
+ %1 = atomicrmw and ptr %a, i16 %b release
+ ret i16 %1
+
+}
+
+define i8 @atomicrmw_or_i8_release(ptr %a, i8 %b) nounwind {
+; LA64-LABEL: atomicrmw_or_i8_release:
+; LA64: # %bb.0:
+; LA64-NEXT: slli.d $a2, $a0, 3
+; LA64-NEXT: bstrins.d $a0, $zero, 1, 0
+; LA64-NEXT: andi $a1, $a1, 255
+; LA64-NEXT: sll.w $a1, $a1, $a2
+; LA64-NEXT: amor_db.w $a3, $a1, $a0
+; LA64-NEXT: srl.w $a0, $a3, $a2
+; LA64-NEXT: ret
+;
+; LA64-LAMCAS-LABEL: atomicrmw_or_i8_release:
+; LA64-LAMCAS: # %bb.0:
+; LA64-LAMCAS-NEXT: move $a2, $a0
+; LA64-LAMCAS-NEXT: ld.bu $a0, $a0, 0
+; LA64-LAMCAS-NEXT: .p2align 4, , 16
+; LA64-LAMCAS-NEXT: .LBB52_1: # %atomicrmw.start
+; LA64-LAMCAS-NEXT: # =>This Inner Loop Header: Depth=1
+; LA64-LAMCAS-NEXT: or $a3, $a0, $a1
+; LA64-LAMCAS-NEXT: ext.w.b $a4, $a0
+; LA64-LAMCAS-NEXT: amcas_db.b $a0, $a3, $a2
+; LA64-LAMCAS-NEXT: bne $a0, $a4, .LBB52_1
+; LA64-LAMCAS-NEXT: # %bb.2: # %atomicrmw.end
+; LA64-LAMCAS-NEXT: ret
+ %1 = atomicrmw or ptr %a, i8 %b release
+ ret i8 %1
+}
+
+define i16 @atomicrmw_or_i16_release(ptr %a, i16 %b) nounwind {
+; LA64-LABEL: atomicrmw_or_i16_release:
+; LA64: # %bb.0:
+; LA64-NEXT: slli.d $a2, $a0, 3
+; LA64-NEXT: bstrins.d $a0, $zero, 1, 0
+; LA64-NEXT: bstrpick.d $a1, $a1, 15, 0
+; LA64-NEXT: sll.w $a1, $a1, $a2
+; LA64-NEXT: amor_db.w $a3, $a1, $a0
+; LA64-NEXT: srl.w $a0, $a3, $a2
+; LA64-NEXT: ret
+;
+; LA64-LAMCAS-LABEL: atomicrmw_or_i16_release:
+; LA64-LAMCAS: # %bb.0:
+; LA64-LAMCAS-NEXT: move $a2, $a0
+; LA64-LAMCAS-NEXT: ld.hu $a0, $a0, 0
+; LA64-LAMCAS-NEXT: .p2align 4, , 16
+; LA64-LAMCAS-NEXT: .LBB53_1: # %atomicrmw.start
+; LA64-LAMCAS-NEXT: # =>This Inner Loop Header: Depth=1
+; LA64-LAMCAS-NEXT: or $a3, $a0, $a1
+; LA64-LAMCAS-NEXT: ext.w.h $a4, $a0
+; LA64-LAMCAS-NEXT: amcas_db.h $a0, $a3, $a2
+; LA64-LAMCAS-NEXT: bne $a0, $a4, .LBB53_1
+; LA64-LAMCAS-NEXT: # %bb.2: # %atomicrmw.end
+; LA64-LAMCAS-NEXT: ret
+ %1 = atomicrmw or ptr %a, i16 %b release
+ ret i16 %1
+
+}
+
+define i8 @atomicrmw_xor_i8_release(ptr %a, i8 %b) nounwind {
+; LA64-LABEL: atomicrmw_xor_i8_release:
+; LA64: # %bb.0:
+; LA64-NEXT: slli.d $a2, $a0, 3
+; LA64-NEXT: bstrins.d $a0, $zero, 1, 0
+; LA64-NEXT: andi $a1, $a1, 255
+; LA64-NEXT: sll.w $a1, $a1, $a2
+; LA64-NEXT: amxor_db.w $a3, $a1, $a0
+; LA64-NEXT: srl.w $a0, $a3, $a2
+; LA64-NEXT: ret
+;
+; LA64-LAMCAS-LABEL: atomicrmw_xor_i8_release:
+; LA64-LAMCAS: # %bb.0:
+; LA64-LAMCAS-NEXT: move $a2, $a0
+; LA64-LAMCAS-NEXT: ld.bu $a0, $a0, 0
+; LA64-LAMCAS-NEXT: .p2align 4, , 16
+; LA64-LAMCAS-NEXT: .LBB54_1: # %atomicrmw.start
+; LA64-LAMCAS-NEXT: # =>This Inner Loop Header: Depth=1
+; LA64-LAMCAS-NEXT: xor $a3, $a0, $a1
+; LA64-LAMCAS-NEXT: ext.w.b $a4, $a0
+; LA64-LAMCAS-NEXT: amcas_db.b $a0, $a3, $a2
+; LA64-LAMCAS-NEXT: bne $a0, $a4, .LBB54_1
+; LA64-LAMCAS-NEXT: # %bb.2: # %atomicrmw.end
+; LA64-LAMCAS-NEXT: ret
+ %1 = atomicrmw xor ptr %a, i8 %b release
+ ret i8 %1
+}
+
+define i16 @atomicrmw_xor_i16_release(ptr %a, i16 %b) nounwind {
+; LA64-LABEL: atomicrmw_xor_i16_release:
+; LA64: # %bb.0:
+; LA64-NEXT: slli.d $a2, $a0, 3
+; LA64-NEXT: bstrins.d $a0, $zero, 1, 0
+; LA64-NEXT: bstrpick.d $a1, $a1, 15, 0
+; LA64-NEXT: sll.w $a1, $a1, $a2
+; LA64-NEXT: amxor_db.w $a3, $a1, $a0
+; LA64-NEXT: srl.w $a0, $a3, $a2
+; LA64-NEXT: ret
+;
+; LA64-LAMCAS-LABEL: atomicrmw_xor_i16_release:
+; LA64-LAMCAS: # %bb.0:
+; LA64-LAMCAS-NEXT: move $a2, $a0
+; LA64-LAMCAS-NEXT: ld.hu $a0, $a0, 0
+; LA64-LAMCAS-NEXT: .p2align 4, , 16
+; LA64-LAMCAS-NEXT: .LBB55_1: # %atomicrmw.start
+; LA64-LAMCAS-NEXT: # =>This Inner Loop Header: Depth=1
+; LA64-LAMCAS-NEXT: xor $a3, $a0, $a1
+; LA64-LAMCAS-NEXT: ext.w.h $a4, $a0
+; LA64-LAMCAS-NEXT: amcas_db.h $a0, $a3, $a2
+; LA64-LAMCAS-NEXT: bne $a0, $a4, .LBB55_1
+; LA64-LAMCAS-NEXT: # %bb.2: # %atomicrmw.end
+; LA64-LAMCAS-NEXT: ret
+ %1 = atomicrmw xor ptr %a, i16 %b release
+ ret i16 %1
+
+}
+
+define i8 @atomicrmw_xchg_i8_acq_rel(ptr %a, i8 %b) nounwind {
+; LA64-LABEL: atomicrmw_xchg_i8_acq_rel:
+; LA64: # %bb.0:
+; LA64-NEXT: slli.d $a2, $a0, 3
+; LA64-NEXT: bstrins.d $a0, $zero, 1, 0
+; LA64-NEXT: ori $a3, $zero, 255
+; LA64-NEXT: sll.w $a3, $a3, $a2
+; LA64-NEXT: andi $a1, $a1, 255
+; LA64-NEXT: sll.w $a1, $a1, $a2
+; LA64-NEXT: .LBB56_1: # =>This Inner Loop Header: Depth=1
+; LA64-NEXT: ll.w $a4, $a0, 0
+; LA64-NEXT: addi.w $a5, $a1, 0
+; LA64-NEXT: xor $a5, $a4, $a5
+; LA64-NEXT: and $a5, $a5, $a3
+; LA64-NEXT: xor $a5, $a4, $a5
+; LA64-NEXT: sc.w $a5, $a0, 0
+; LA64-NEXT: beqz $a5, .LBB56_1
+; LA64-NEXT: # %bb.2:
+; LA64-NEXT: srl.w $a0, $a4, $a2
+; LA64-NEXT: ret
+;
+; LA64-LAMCAS-LABEL: atomicrmw_xchg_i8_acq_rel:
+; LA64-LAMCAS: # %bb.0:
+; LA64-LAMCAS-NEXT: ld.bu $a2, $a0, 0
+; LA64-LAMCAS-NEXT: .p2align 4, , 16
+; LA64-LAMCAS-NEXT: .LBB56_1: # %atomicrmw.start
+; LA64-LAMCAS-NEXT: # =>This Inner Loop Header: Depth=1
+; LA64-LAMCAS-NEXT: ext.w.b $a3, $a2
+; LA64-LAMCAS-NEXT: amcas_db.b $a2, $a1, $a0
+; LA64-LAMCAS-NEXT: bne $a2, $a3, .LBB56_1
+; LA64-LAMCAS-NEXT: # %bb.2: # %atomicrmw.end
+; LA64-LAMCAS-NEXT: move $a0, $a2
+; LA64-LAMCAS-NEXT: ret
+ %1 = atomicrmw xchg ptr %a, i8 %b acq_rel
+ ret i8 %1
+}
+
+define i8 @atomicrmw_xchg_0_i8_acq_rel(ptr %a) nounwind {
+; LA64-LABEL: atomicrmw_xchg_0_i8_acq_rel:
+; LA64: # %bb.0:
+; LA64-NEXT: slli.d $a1, $a0, 3
+; LA64-NEXT: bstrins.d $a0, $zero, 1, 0
+; LA64-NEXT: ori $a2, $zero, 255
+; LA64-NEXT: sll.w $a2, $a2, $a1
+; LA64-NEXT: nor $a2, $a2, $zero
+; LA64-NEXT: amand_db.w $a3, $a2, $a0
+; LA64-NEXT: srl.w $a0, $a3, $a1
+; LA64-NEXT: ret
+;
+; LA64-LAMCAS-LABEL: atomicrmw_xchg_0_i8_acq_rel:
+; LA64-LAMCAS: # %bb.0:
+; LA64-LAMCAS-NEXT: move $a1, $a0
+; LA64-LAMCAS-NEXT: ld.bu $a0, $a0, 0
+; LA64-LAMCAS-NEXT: .p2align 4, , 16
+; LA64-LAMCAS-NEXT: .LBB57_1: # %atomicrmw.start
+; LA64-LAMCAS-NEXT: # =>This Inner Loop Header: Depth=1
+; LA64-LAMCAS-NEXT: ext.w.b $a2, $a0
+; LA64-LAMCAS-NEXT: amcas_db.b $a0, $zero, $a1
+; LA64-LAMCAS-NEXT: bne $a0, $a2, .LBB57_1
+; LA64-LAMCAS-NEXT: # %bb.2: # %atomicrmw.end
+; LA64-LAMCAS-NEXT: ret
+ %1 = atomicrmw xchg ptr %a, i8 0 acq_rel
+ ret i8 %1
+}
+
+define i8 @atomicrmw_xchg_minus_1_i8_acq_rel(ptr %a) nounwind {
+; LA64-LABEL: atomicrmw_xchg_minus_1_i8_acq_rel:
+; LA64: # %bb.0:
+; LA64-NEXT: slli.d $a1, $a0, 3
+; LA64-NEXT: bstrins.d $a0, $zero, 1, 0
+; LA64-NEXT: ori $a2, $zero, 255
+; LA64-NEXT: sll.w $a2, $a2, $a1
+; LA64-NEXT: amor_db.w $a3, $a2, $a0
+; LA64-NEXT: srl.w $a0, $a3, $a1
+; LA64-NEXT: ret
+;
+; LA64-LAMCAS-LABEL: atomicrmw_xchg_minus_1_i8_acq_rel:
+; LA64-LAMCAS: # %bb.0:
+; LA64-LAMCAS-NEXT: move $a1, $a0
+; LA64-LAMCAS-NEXT: ld.bu $a0, $a0, 0
+; LA64-LAMCAS-NEXT: addi.w $a2, $zero, -1
+; LA64-LAMCAS-NEXT: .p2align 4, , 16
+; LA64-LAMCAS-NEXT: .LBB58_1: # %atomicrmw.start
+; LA64-LAMCAS-NEXT: # =>This Inner Loop Header: Depth=1
+; LA64-LAMCAS-NEXT: ext.w.b $a3, $a0
+; LA64-LAMCAS-NEXT: amcas_db.b $a0, $a2, $a1
+; LA64-LAMCAS-NEXT: bne $a0, $a3, .LBB58_1
+; LA64-LAMCAS-NEXT: # %bb.2: # %atomicrmw.end
+; LA64-LAMCAS-NEXT: ret
+ %1 = atomicrmw xchg ptr %a, i8 -1 acq_rel
+ ret i8 %1
+}
+
+define i16 @atomicrmw_xchg_i16_acq_rel(ptr %a, i16 %b) nounwind {
+; LA64-LABEL: atomicrmw_xchg_i16_acq_rel:
+; LA64: # %bb.0:
+; LA64-NEXT: slli.d $a2, $a0, 3
+; LA64-NEXT: bstrins.d $a0, $zero, 1, 0
+; LA64-NEXT: lu12i.w $a3, 15
+; LA64-NEXT: ori $a3, $a3, 4095
+; LA64-NEXT: sll.w $a3, $a3, $a2
+; LA64-NEXT: bstrpick.d $a1, $a1, 15, 0
+; LA64-NEXT: sll.w $a1, $a1, $a2
+; LA64-NEXT: .LBB59_1: # =>This Inner Loop Header: Depth=1
+; LA64-NEXT: ll.w $a4, $a0, 0
+; LA64-NEXT: addi.w $a5, $a1, 0
+; LA64-NEXT: xor $a5, $a4, $a5
+; LA64-NEXT: and $a5, $a5, $a3
+; LA64-NEXT: xor $a5, $a4, $a5
+; LA64-NEXT: sc.w $a5, $a0, 0
+; LA64-NEXT: beqz $a5, .LBB59_1
+; LA64-NEXT: # %bb.2:
+; LA64-NEXT: srl.w $a0, $a4, $a2
+; LA64-NEXT: ret
+;
+; LA64-LAMCAS-LABEL: atomicrmw_xchg_i16_acq_rel:
+; LA64-LAMCAS: # %bb.0:
+; LA64-LAMCAS-NEXT: ld.hu $a2, $a0, 0
+; LA64-LAMCAS-NEXT: .p2align 4, , 16
+; LA64-LAMCAS-NEXT: .LBB59_1: # %atomicrmw.start
+; LA64-LAMCAS-NEXT: # =>This Inner Loop Header: Depth=1
+; LA64-LAMCAS-NEXT: ext.w.h $a3, $a2
+; LA64-LAMCAS-NEXT: amcas_db.h $a2, $a1, $a0
+; LA64-LAMCAS-NEXT: bne $a2, $a3, .LBB59_1
+; LA64-LAMCAS-NEXT: # %bb.2: # %atomicrmw.end
+; LA64-LAMCAS-NEXT: move $a0, $a2
+; LA64-LAMCAS-NEXT: ret
+ %1 = atomicrmw xchg ptr %a, i16 %b acq_rel
+ ret i16 %1
+}
+
+define i16 @atomicrmw_xchg_0_i16_acq_rel(ptr %a) nounwind {
+; LA64-LABEL: atomicrmw_xchg_0_i16_acq_rel:
+; LA64: # %bb.0:
+; LA64-NEXT: slli.d $a1, $a0, 3
+; LA64-NEXT: bstrins.d $a0, $zero, 1, 0
+; LA64-NEXT: lu12i.w $a2, 15
+; LA64-NEXT: ori $a2, $a2, 4095
+; LA64-NEXT: sll.w $a2, $a2, $a1
+; LA64-NEXT: nor $a2, $a2, $zero
+; LA64-NEXT: amand_db.w $a3, $a2, $a0
+; LA64-NEXT: srl.w $a0, $a3, $a1
+; LA64-NEXT: ret
+;
+; LA64-LAMCAS-LABEL: atomicrmw_xchg_0_i16_acq_rel:
+; LA64-LAMCAS: # %bb.0:
+; LA64-LAMCAS-NEXT: move $a1, $a0
+; LA64-LAMCAS-NEXT: ld.hu $a0, $a0, 0
+; LA64-LAMCAS-NEXT: .p2align 4, , 16
+; LA64-LAMCAS-NEXT: .LBB60_1: # %atomicrmw.start
+; LA64-LAMCAS-NEXT: # =>This Inner Loop Header: Depth=1
+; LA64-LAMCAS-NEXT: ext.w.h $a2, $a0
+; LA64-LAMCAS-NEXT: amcas_db.h $a0, $zero, $a1
+; LA64-LAMCAS-NEXT: bne $a0, $a2, .LBB60_1
+; LA64-LAMCAS-NEXT: # %bb.2: # %atomicrmw.end
+; LA64-LAMCAS-NEXT: ret
+ %1 = atomicrmw xchg ptr %a, i16 0 acq_rel
+ ret i16 %1
+}
+
+define i16 @atomicrmw_xchg_minus_1_i16_acq_rel(ptr %a) nounwind {
+; LA64-LABEL: atomicrmw_xchg_minus_1_i16_acq_rel:
+; LA64: # %bb.0:
+; LA64-NEXT: slli.d $a1, $a0, 3
+; LA64-NEXT: bstrins.d $a0, $zero, 1, 0
+; LA64-NEXT: lu12i.w $a2, 15
+; LA64-NEXT: ori $a2, $a2, 4095
+; LA64-NEXT: sll.w $a2, $a2, $a1
+; LA64-NEXT: amor_db.w $a3, $a2, $a0
+; LA64-NEXT: srl.w $a0, $a3, $a1
+; LA64-NEXT: ret
+;
+; LA64-LAMCAS-LABEL: atomicrmw_xchg_minus_1_i16_acq_rel:
+; LA64-LAMCAS: # %bb.0:
+; LA64-LAMCAS-NEXT: move $a1, $a0
+; LA64-LAMCAS-NEXT: ld.hu $a0, $a0, 0
+; LA64-LAMCAS-NEXT: addi.w $a2, $zero, -1
+; LA64-LAMCAS-NEXT: .p2align 4, , 16
+; LA64-LAMCAS-NEXT: .LBB61_1: # %atomicrmw.start
+; LA64-LAMCAS-NEXT: # =>This Inner Loop Header: Depth=1
+; LA64-LAMCAS-NEXT: ext.w.h $a3, $a0
+; LA64-LAMCAS-NEXT: amcas_db.h $a0, $a2, $a1
+; LA64-LAMCAS-NEXT: bne $a0, $a3, .LBB61_1
+; LA64-LAMCAS-NEXT: # %bb.2: # %atomicrmw.end
+; LA64-LAMCAS-NEXT: ret
+ %1 = atomicrmw xchg ptr %a, i16 -1 acq_rel
+ ret i16 %1
+
+}
+
+define i8 @atomicrmw_add_i8_acq_rel(ptr %a, i8 %b) nounwind {
+; LA64-LABEL: atomicrmw_add_i8_acq_rel:
+; LA64: # %bb.0:
+; LA64-NEXT: slli.d $a2, $a0, 3
+; LA64-NEXT: bstrins.d $a0, $zero, 1, 0
+; LA64-NEXT: ori $a3, $zero, 255
+; LA64-NEXT: sll.w $a3, $a3, $a2
+; LA64-NEXT: andi $a1, $a1, 255
+; LA64-NEXT: sll.w $a1, $a1, $a2
+; LA64-NEXT: .LBB62_1: # =>This Inner Loop Header: Depth=1
+; LA64-NEXT: ll.w $a4, $a0, 0
+; LA64-NEXT: add.w $a5, $a4, $a1
+; LA64-NEXT: xor $a5, $a4, $a5
+; LA64-NEXT: and $a5, $a5, $a3
+; LA64-NEXT: xor $a5, $a4, $a5
+; LA64-NEXT: sc.w $a5, $a0, 0
+; LA64-NEXT: beqz $a5, .LBB62_1
+; LA64-NEXT: # %bb.2:
+; LA64-NEXT: srl.w $a0, $a4, $a2
+; LA64-NEXT: ret
+;
+; LA64-LAMCAS-LABEL: atomicrmw_add_i8_acq_rel:
+; LA64-LAMCAS: # %bb.0:
+; LA64-LAMCAS-NEXT: move $a2, $a0
+; LA64-LAMCAS-NEXT: ld.bu $a0, $a0, 0
+; LA64-LAMCAS-NEXT: .p2align 4, , 16
+; LA64-LAMCAS-NEXT: .LBB62_1: # %atomicrmw.start
+; LA64-LAMCAS-NEXT: # =>This Inner Loop Header: Depth=1
+; LA64-LAMCAS-NEXT: add.d $a3, $a0, $a1
+; LA64-LAMCAS-NEXT: ext.w.b $a4, $a0
+; LA64-LAMCAS-NEXT: amcas_db.b $a0, $a3, $a2
+; LA64-LAMCAS-NEXT: bne $a0, $a4, .LBB62_1
+; LA64-LAMCAS-NEXT: # %bb.2: # %atomicrmw.end
+; LA64-LAMCAS-NEXT: ret
+ %1 = atomicrmw add ptr %a, i8 %b acq_rel
+ ret i8 %1
+}
+
+define i16 @atomicrmw_add_i16_acq_rel(ptr %a, i16 %b) nounwind {
+; LA64-LABEL: atomicrmw_add_i16_acq_rel:
+; LA64: # %bb.0:
+; LA64-NEXT: slli.d $a2, $a0, 3
+; LA64-NEXT: bstrins.d $a0, $zero, 1, 0
+; LA64-NEXT: lu12i.w $a3, 15
+; LA64-NEXT: ori $a3, $a3, 4095
+; LA64-NEXT: sll.w $a3, $a3, $a2
+; LA64-NEXT: bstrpick.d $a1, $a1, 15, 0
+; LA64-NEXT: sll.w $a1, $a1, $a2
+; LA64-NEXT: .LBB63_1: # =>This Inner Loop Header: Depth=1
+; LA64-NEXT: ll.w $a4, $a0, 0
+; LA64-NEXT: add.w $a5, $a4, $a1
+; LA64-NEXT: xor $a5, $a4, $a5
+; LA64-NEXT: and $a5, $a5, $a3
+; LA64-NEXT: xor $a5, $a4, $a5
+; LA64-NEXT: sc.w $a5, $a0, 0
+; LA64-NEXT: beqz $a5, .LBB63_1
+; LA64-NEXT: # %bb.2:
+; LA64-NEXT: srl.w $a0, $a4, $a2
+; LA64-NEXT: ret
+;
+; LA64-LAMCAS-LABEL: atomicrmw_add_i16_acq_rel:
+; LA64-LAMCAS: # %bb.0:
+; LA64-LAMCAS-NEXT: move $a2, $a0
+; LA64-LAMCAS-NEXT: ld.hu $a0, $a0, 0
+; LA64-LAMCAS-NEXT: .p2align 4, , 16
+; LA64-LAMCAS-NEXT: .LBB63_1: # %atomicrmw.start
+; LA64-LAMCAS-NEXT: # =>This Inner Loop Header: Depth=1
+; LA64-LAMCAS-NEXT: add.d $a3, $a0, $a1
+; LA64-LAMCAS-NEXT: ext.w.h $a4, $a0
+; LA64-LAMCAS-NEXT: amcas_db.h $a0, $a3, $a2
+; LA64-LAMCAS-NEXT: bne $a0, $a4, .LBB63_1
+; LA64-LAMCAS-NEXT: # %bb.2: # %atomicrmw.end
+; LA64-LAMCAS-NEXT: ret
+ %1 = atomicrmw add ptr %a, i16 %b acq_rel
+ ret i16 %1
+
+}
+
+define i8 @atomicrmw_sub_i8_acq_rel(ptr %a, i8 %b) nounwind {
+; LA64-LABEL: atomicrmw_sub_i8_acq_rel:
+; LA64: # %bb.0:
+; LA64-NEXT: slli.d $a2, $a0, 3
+; LA64-NEXT: bstrins.d $a0, $zero, 1, 0
+; LA64-NEXT: ori $a3, $zero, 255
+; LA64-NEXT: sll.w $a3, $a3, $a2
+; LA64-NEXT: andi $a1, $a1, 255
+; LA64-NEXT: sll.w $a1, $a1, $a2
+; LA64-NEXT: .LBB64_1: # =>This Inner Loop Header: Depth=1
+; LA64-NEXT: ll.w $a4, $a0, 0
+; LA64-NEXT: sub.w $a5, $a4, $a1
+; LA64-NEXT: xor $a5, $a4, $a5
+; LA64-NEXT: and $a5, $a5, $a3
+; LA64-NEXT: xor $a5, $a4, $a5
+; LA64-NEXT: sc.w $a5, $a0, 0
+; LA64-NEXT: beqz $a5, .LBB64_1
+; LA64-NEXT: # %bb.2:
+; LA64-NEXT: srl.w $a0, $a4, $a2
+; LA64-NEXT: ret
+;
+; LA64-LAMCAS-LABEL: atomicrmw_sub_i8_acq_rel:
+; LA64-LAMCAS: # %bb.0:
+; LA64-LAMCAS-NEXT: move $a2, $a0
+; LA64-LAMCAS-NEXT: ld.bu $a0, $a0, 0
+; LA64-LAMCAS-NEXT: .p2align 4, , 16
+; LA64-LAMCAS-NEXT: .LBB64_1: # %atomicrmw.start
+; LA64-LAMCAS-NEXT: # =>This Inner Loop Header: Depth=1
+; LA64-LAMCAS-NEXT: sub.d $a3, $a0, $a1
+; LA64-LAMCAS-NEXT: ext.w.b $a4, $a0
+; LA64-LAMCAS-NEXT: amcas_db.b $a0, $a3, $a2
+; LA64-LAMCAS-NEXT: bne $a0, $a4, .LBB64_1
+; LA64-LAMCAS-NEXT: # %bb.2: # %atomicrmw.end
+; LA64-LAMCAS-NEXT: ret
+ %1 = atomicrmw sub ptr %a, i8 %b acq_rel
+ ret i8 %1
+}
+
+define i16 @atomicrmw_sub_i16_acq_rel(ptr %a, i16 %b) nounwind {
+; LA64-LABEL: atomicrmw_sub_i16_acq_rel:
+; LA64: # %bb.0:
+; LA64-NEXT: slli.d $a2, $a0, 3
+; LA64-NEXT: bstrins.d $a0, $zero, 1, 0
+; LA64-NEXT: lu12i.w $a3, 15
+; LA64-NEXT: ori $a3, $a3, 4095
+; LA64-NEXT: sll.w $a3, $a3, $a2
+; LA64-NEXT: bstrpick.d $a1, $a1, 15, 0
+; LA64-NEXT: sll.w $a1, $a1, $a2
+; LA64-NEXT: .LBB65_1: # =>This Inner Loop Header: Depth=1
+; LA64-NEXT: ll.w $a4, $a0, 0
+; LA64-NEXT: sub.w $a5, $a4, $a1
+; LA64-NEXT: xor $a5, $a4, $a5
+; LA64-NEXT: and $a5, $a5, $a3
+; LA64-NEXT: xor $a5, $a4, $a5
+; LA64-NEXT: sc.w $a5, $a0, 0
+; LA64-NEXT: beqz $a5, .LBB65_1
+; LA64-NEXT: # %bb.2:
+; LA64-NEXT: srl.w $a0, $a4, $a2
+; LA64-NEXT: ret
+;
+; LA64-LAMCAS-LABEL: atomicrmw_sub_i16_acq_rel:
+; LA64-LAMCAS: # %bb.0:
+; LA64-LAMCAS-NEXT: move $a2, $a0
+; LA64-LAMCAS-NEXT: ld.hu $a0, $a0, 0
+; LA64-LAMCAS-NEXT: .p2align 4, , 16
+; LA64-LAMCAS-NEXT: .LBB65_1: # %atomicrmw.start
+; LA64-LAMCAS-NEXT: # =>This Inner Loop Header: Depth=1
+; LA64-LAMCAS-NEXT: sub.d $a3, $a0, $a1
+; LA64-LAMCAS-NEXT: ext.w.h $a4, $a0
+; LA64-LAMCAS-NEXT: amcas_db.h $a0, $a3, $a2
+; LA64-LAMCAS-NEXT: bne $a0, $a4, .LBB65_1
+; LA64-LAMCAS-NEXT: # %bb.2: # %atomicrmw.end
+; LA64-LAMCAS-NEXT: ret
+ %1 = atomicrmw sub ptr %a, i16 %b acq_rel
+ ret i16 %1
+
+}
+
+define i8 @atomicrmw_umax_i8_acq_rel(ptr %a, i8 %b) nounwind {
+; LA64-LABEL: atomicrmw_umax_i8_acq_rel:
+; LA64: # %bb.0:
+; LA64-NEXT: slli.d $a2, $a0, 3
+; LA64-NEXT: bstrins.d $a0, $zero, 1, 0
+; LA64-NEXT: ori $a3, $zero, 255
+; LA64-NEXT: sll.w $a3, $a3, $a2
+; LA64-NEXT: andi $a1, $a1, 255
+; LA64-NEXT: sll.w $a1, $a1, $a2
+; LA64-NEXT: .LBB66_1: # =>This Inner Loop Header: Depth=1
+; LA64-NEXT: ll.w $a4, $a0, 0
+; LA64-NEXT: and $a6, $a4, $a3
+; LA64-NEXT: move $a5, $a4
+; LA64-NEXT: bgeu $a6, $a1, .LBB66_3
+; LA64-NEXT: # %bb.2: # in Loop: Header=BB66_1 Depth=1
+; LA64-NEXT: xor $a5, $a4, $a1
+; LA64-NEXT: and $a5, $a5, $a3
+; LA64-NEXT: xor $a5, $a4, $a5
+; LA64-NEXT: .LBB66_3: # in Loop: Header=BB66_1 Depth=1
+; LA64-NEXT: sc.w $a5, $a0, 0
+; LA64-NEXT: beqz $a5, .LBB66_1
+; LA64-NEXT: # %bb.4:
+; LA64-NEXT: srl.w $a0, $a4, $a2
+; LA64-NEXT: ret
+;
+; LA64-LAMCAS-LABEL: atomicrmw_umax_i8_acq_rel:
+; LA64-LAMCAS: # %bb.0:
+; LA64-LAMCAS-NEXT: move $a2, $a0
+; LA64-LAMCAS-NEXT: ld.bu $a0, $a0, 0
+; LA64-LAMCAS-NEXT: andi $a3, $a1, 255
+; LA64-LAMCAS-NEXT: .p2align 4, , 16
+; LA64-LAMCAS-NEXT: .LBB66_1: # %atomicrmw.start
+; LA64-LAMCAS-NEXT: # =>This Inner Loop Header: Depth=1
+; LA64-LAMCAS-NEXT: andi $a4, $a0, 255
+; LA64-LAMCAS-NEXT: sltu $a4, $a3, $a4
+; LA64-LAMCAS-NEXT: masknez $a5, $a1, $a4
+; LA64-LAMCAS-NEXT: maskeqz $a4, $a0, $a4
+; LA64-LAMCAS-NEXT: or $a4, $a4, $a5
+; LA64-LAMCAS-NEXT: ext.w.b $a5, $a0
+; LA64-LAMCAS-NEXT: amcas_db.b $a0, $a4, $a2
+; LA64-LAMCAS-NEXT: bne $a0, $a5, .LBB66_1
+; LA64-LAMCAS-NEXT: # %bb.2: # %atomicrmw.end
+; LA64-LAMCAS-NEXT: ret
+ %1 = atomicrmw umax ptr %a, i8 %b acq_rel
+ ret i8 %1
+}
+
+define i16 @atomicrmw_umax_i16_acq_rel(ptr %a, i16 %b) nounwind {
+; LA64-LABEL: atomicrmw_umax_i16_acq_rel:
+; LA64: # %bb.0:
+; LA64-NEXT: slli.d $a2, $a0, 3
+; LA64-NEXT: bstrins.d $a0, $zero, 1, 0
+; LA64-NEXT: lu12i.w $a3, 15
+; LA64-NEXT: ori $a3, $a3, 4095
+; LA64-NEXT: sll.w $a3, $a3, $a2
+; LA64-NEXT: bstrpick.d $a1, $a1, 15, 0
+; LA64-NEXT: sll.w $a1, $a1, $a2
+; LA64-NEXT: .LBB67_1: # =>This Inner Loop Header: Depth=1
+; LA64-NEXT: ll.w $a4, $a0, 0
+; LA64-NEXT: and $a6, $a4, $a3
+; LA64-NEXT: move $a5, $a4
+; LA64-NEXT: bgeu $a6, $a1, .LBB67_3
+; LA64-NEXT: # %bb.2: # in Loop: Header=BB67_1 Depth=1
+; LA64-NEXT: xor $a5, $a4, $a1
+; LA64-NEXT: and $a5, $a5, $a3
+; LA64-NEXT: xor $a5, $a4, $a5
+; LA64-NEXT: .LBB67_3: # in Loop: Header=BB67_1 Depth=1
+; LA64-NEXT: sc.w $a5, $a0, 0
+; LA64-NEXT: beqz $a5, .LBB67_1
+; LA64-NEXT: # %bb.4:
+; LA64-NEXT: srl.w $a0, $a4, $a2
+; LA64-NEXT: ret
+;
+; LA64-LAMCAS-LABEL: atomicrmw_umax_i16_acq_rel:
+; LA64-LAMCAS: # %bb.0:
+; LA64-LAMCAS-NEXT: move $a2, $a0
+; LA64-LAMCAS-NEXT: ld.hu $a0, $a0, 0
+; LA64-LAMCAS-NEXT: bstrpick.d $a3, $a1, 15, 0
+; LA64-LAMCAS-NEXT: .p2align 4, , 16
+; LA64-LAMCAS-NEXT: .LBB67_1: # %atomicrmw.start
+; LA64-LAMCAS-NEXT: # =>This Inner Loop Header: Depth=1
+; LA64-LAMCAS-NEXT: bstrpick.d $a4, $a0, 15, 0
+; LA64-LAMCAS-NEXT: sltu $a4, $a3, $a4
+; LA64-LAMCAS-NEXT: masknez $a5, $a1, $a4
+; LA64-LAMCAS-NEXT: maskeqz $a4, $a0, $a4
+; LA64-LAMCAS-NEXT: or $a4, $a4, $a5
+; LA64-LAMCAS-NEXT: ext.w.h $a5, $a0
+; LA64-LAMCAS-NEXT: amcas_db.h $a0, $a4, $a2
+; LA64-LAMCAS-NEXT: bne $a0, $a5, .LBB67_1
+; LA64-LAMCAS-NEXT: # %bb.2: # %atomicrmw.end
+; LA64-LAMCAS-NEXT: ret
+ %1 = atomicrmw umax ptr %a, i16 %b acq_rel
+ ret i16 %1
+}
+
+define i8 @atomicrmw_umin_i8_acq_rel(ptr %a, i8 %b) nounwind {
+; LA64-LABEL: atomicrmw_umin_i8_acq_rel:
+; LA64: # %bb.0:
+; LA64-NEXT: slli.d $a2, $a0, 3
+; LA64-NEXT: bstrins.d $a0, $zero, 1, 0
+; LA64-NEXT: ori $a3, $zero, 255
+; LA64-NEXT: sll.w $a3, $a3, $a2
+; LA64-NEXT: andi $a1, $a1, 255
+; LA64-NEXT: sll.w $a1, $a1, $a2
+; LA64-NEXT: .LBB68_1: # =>This Inner Loop Header: Depth=1
+; LA64-NEXT: ll.w $a4, $a0, 0
+; LA64-NEXT: and $a6, $a4, $a3
+; LA64-NEXT: move $a5, $a4
+; LA64-NEXT: bgeu $a1, $a6, .LBB68_3
+; LA64-NEXT: # %bb.2: # in Loop: Header=BB68_1 Depth=1
+; LA64-NEXT: xor $a5, $a4, $a1
+; LA64-NEXT: and $a5, $a5, $a3
+; LA64-NEXT: xor $a5, $a4, $a5
+; LA64-NEXT: .LBB68_3: # in Loop: Header=BB68_1 Depth=1
+; LA64-NEXT: sc.w $a5, $a0, 0
+; LA64-NEXT: beqz $a5, .LBB68_1
+; LA64-NEXT: # %bb.4:
+; LA64-NEXT: srl.w $a0, $a4, $a2
+; LA64-NEXT: ret
+;
+; LA64-LAMCAS-LABEL: atomicrmw_umin_i8_acq_rel:
+; LA64-LAMCAS: # %bb.0:
+; LA64-LAMCAS-NEXT: move $a2, $a0
+; LA64-LAMCAS-NEXT: ld.bu $a0, $a0, 0
+; LA64-LAMCAS-NEXT: andi $a3, $a1, 255
+; LA64-LAMCAS-NEXT: .p2align 4, , 16
+; LA64-LAMCAS-NEXT: .LBB68_1: # %atomicrmw.start
+; LA64-LAMCAS-NEXT: # =>This Inner Loop Header: Depth=1
+; LA64-LAMCAS-NEXT: andi $a4, $a0, 255
+; LA64-LAMCAS-NEXT: sltu $a4, $a3, $a4
+; LA64-LAMCAS-NEXT: xori $a4, $a4, 1
+; LA64-LAMCAS-NEXT: masknez $a5, $a1, $a4
+; LA64-LAMCAS-NEXT: maskeqz $a4, $a0, $a4
+; LA64-LAMCAS-NEXT: or $a4, $a4, $a5
+; LA64-LAMCAS-NEXT: ext.w.b $a5, $a0
+; LA64-LAMCAS-NEXT: amcas_db.b $a0, $a4, $a2
+; LA64-LAMCAS-NEXT: bne $a0, $a5, .LBB68_1
+; LA64-LAMCAS-NEXT: # %bb.2: # %atomicrmw.end
+; LA64-LAMCAS-NEXT: ret
+ %1 = atomicrmw umin ptr %a, i8 %b acq_rel
+ ret i8 %1
+}
+
+define i16 @atomicrmw_umin_i16_acq_rel(ptr %a, i16 %b) nounwind {
+; LA64-LABEL: atomicrmw_umin_i16_acq_rel:
+; LA64: # %bb.0:
+; LA64-NEXT: slli.d $a2, $a0, 3
+; LA64-NEXT: bstrins.d $a0, $zero, 1, 0
+; LA64-NEXT: lu12i.w $a3, 15
+; LA64-NEXT: ori $a3, $a3, 4095
+; LA64-NEXT: sll.w $a3, $a3, $a2
+; LA64-NEXT: bstrpick.d $a1, $a1, 15, 0
+; LA64-NEXT: sll.w $a1, $a1, $a2
+; LA64-NEXT: .LBB69_1: # =>This Inner Loop Header: Depth=1
+; LA64-NEXT: ll.w $a4, $a0, 0
+; LA64-NEXT: and $a6, $a4, $a3
+; LA64-NEXT: move $a5, $a4
+; LA64-NEXT: bgeu $a1, $a6, .LBB69_3
+; LA64-NEXT: # %bb.2: # in Loop: Header=BB69_1 Depth=1
+; LA64-NEXT: xor $a5, $a4, $a1
+; LA64-NEXT: and $a5, $a5, $a3
+; LA64-NEXT: xor $a5, $a4, $a5
+; LA64-NEXT: .LBB69_3: # in Loop: Header=BB69_1 Depth=1
+; LA64-NEXT: sc.w $a5, $a0, 0
+; LA64-NEXT: beqz $a5, .LBB69_1
+; LA64-NEXT: # %bb.4:
+; LA64-NEXT: srl.w $a0, $a4, $a2
+; LA64-NEXT: ret
+;
+; LA64-LAMCAS-LABEL: atomicrmw_umin_i16_acq_rel:
+; LA64-LAMCAS: # %bb.0:
+; LA64-LAMCAS-NEXT: move $a2, $a0
+; LA64-LAMCAS-NEXT: ld.hu $a0, $a0, 0
+; LA64-LAMCAS-NEXT: bstrpick.d $a3, $a1, 15, 0
+; LA64-LAMCAS-NEXT: .p2align 4, , 16
+; LA64-LAMCAS-NEXT: .LBB69_1: # %atomicrmw.start
+; LA64-LAMCAS-NEXT: # =>This Inner Loop Header: Depth=1
+; LA64-LAMCAS-NEXT: bstrpick.d $a4, $a0, 15, 0
+; LA64-LAMCAS-NEXT: sltu $a4, $a3, $a4
+; LA64-LAMCAS-NEXT: xori $a4, $a4, 1
+; LA64-LAMCAS-NEXT: masknez $a5, $a1, $a4
+; LA64-LAMCAS-NEXT: maskeqz $a4, $a0, $a4
+; LA64-LAMCAS-NEXT: or $a4, $a4, $a5
+; LA64-LAMCAS-NEXT: ext.w.h $a5, $a0
+; LA64-LAMCAS-NEXT: amcas_db.h $a0, $a4, $a2
+; LA64-LAMCAS-NEXT: bne $a0, $a5, .LBB69_1
+; LA64-LAMCAS-NEXT: # %bb.2: # %atomicrmw.end
+; LA64-LAMCAS-NEXT: ret
+ %1 = atomicrmw umin ptr %a, i16 %b acq_rel
+ ret i16 %1
+}
+
+define i8 @atomicrmw_max_i8_acq_rel(ptr %a, i8 %b) nounwind {
+; LA64-LABEL: atomicrmw_max_i8_acq_rel:
+; LA64: # %bb.0:
+; LA64-NEXT: slli.d $a2, $a0, 3
+; LA64-NEXT: bstrins.d $a0, $zero, 1, 0
+; LA64-NEXT: andi $a3, $a2, 24
+; LA64-NEXT: ori $a4, $zero, 255
+; LA64-NEXT: sll.w $a4, $a4, $a2
+; LA64-NEXT: ext.w.b $a1, $a1
+; LA64-NEXT: sll.w $a1, $a1, $a2
+; LA64-NEXT: xori $a3, $a3, 56
+; LA64-NEXT: .LBB70_1: # =>This Inner Loop Header: Depth=1
+; LA64-NEXT: ll.w $a5, $a0, 0
+; LA64-NEXT: and $a7, $a5, $a4
+; LA64-NEXT: move $a6, $a5
+; LA64-NEXT: sll.w $a7, $a7, $a3
+; LA64-NEXT: sra.w $a7, $a7, $a3
+; LA64-NEXT: bge $a7, $a1, .LBB70_3
+; LA64-NEXT: # %bb.2: # in Loop: Header=BB70_1 Depth=1
+; LA64-NEXT: xor $a6, $a5, $a1
+; LA64-NEXT: and $a6, $a6, $a4
+; LA64-NEXT: xor $a6, $a5, $a6
+; LA64-NEXT: .LBB70_3: # in Loop: Header=BB70_1 Depth=1
+; LA64-NEXT: sc.w $a6, $a0, 0
+; LA64-NEXT: beqz $a6, .LBB70_1
+; LA64-NEXT: # %bb.4:
+; LA64-NEXT: srl.w $a0, $a5, $a2
+; LA64-NEXT: ret
+;
+; LA64-LAMCAS-LABEL: atomicrmw_max_i8_acq_rel:
+; LA64-LAMCAS: # %bb.0:
+; LA64-LAMCAS-NEXT: move $a2, $a0
+; LA64-LAMCAS-NEXT: ld.bu $a0, $a0, 0
+; LA64-LAMCAS-NEXT: ext.w.b $a3, $a1
+; LA64-LAMCAS-NEXT: .p2align 4, , 16
+; LA64-LAMCAS-NEXT: .LBB70_1: # %atomicrmw.start
+; LA64-LAMCAS-NEXT: # =>This Inner Loop Header: Depth=1
+; LA64-LAMCAS-NEXT: ext.w.b $a4, $a0
+; LA64-LAMCAS-NEXT: slt $a5, $a3, $a4
+; LA64-LAMCAS-NEXT: masknez $a6, $a1, $a5
+; LA64-LAMCAS-NEXT: maskeqz $a5, $a0, $a5
+; LA64-LAMCAS-NEXT: or $a5, $a5, $a6
+; LA64-LAMCAS-NEXT: amcas_db.b $a0, $a5, $a2
+; LA64-LAMCAS-NEXT: bne $a0, $a4, .LBB70_1
+; LA64-LAMCAS-NEXT: # %bb.2: # %atomicrmw.end
+; LA64-LAMCAS-NEXT: ret
+ %1 = atomicrmw max ptr %a, i8 %b acq_rel
+ ret i8 %1
+}
+
+define i16 @atomicrmw_max_i16_acq_rel(ptr %a, i16 %b) nounwind {
+; LA64-LABEL: atomicrmw_max_i16_acq_rel:
+; LA64: # %bb.0:
+; LA64-NEXT: slli.d $a2, $a0, 3
+; LA64-NEXT: bstrins.d $a0, $zero, 1, 0
+; LA64-NEXT: andi $a3, $a2, 24
+; LA64-NEXT: lu12i.w $a4, 15
+; LA64-NEXT: ori $a4, $a4, 4095
+; LA64-NEXT: sll.w $a4, $a4, $a2
+; LA64-NEXT: ext.w.h $a1, $a1
+; LA64-NEXT: sll.w $a1, $a1, $a2
+; LA64-NEXT: ori $a5, $zero, 48
+; LA64-NEXT: sub.d $a3, $a5, $a3
+; LA64-NEXT: .LBB71_1: # =>This Inner Loop Header: Depth=1
+; LA64-NEXT: ll.w $a5, $a0, 0
+; LA64-NEXT: and $a7, $a5, $a4
+; LA64-NEXT: move $a6, $a5
+; LA64-NEXT: sll.w $a7, $a7, $a3
+; LA64-NEXT: sra.w $a7, $a7, $a3
+; LA64-NEXT: bge $a7, $a1, .LBB71_3
+; LA64-NEXT: # %bb.2: # in Loop: Header=BB71_1 Depth=1
+; LA64-NEXT: xor $a6, $a5, $a1
+; LA64-NEXT: and $a6, $a6, $a4
+; LA64-NEXT: xor $a6, $a5, $a6
+; LA64-NEXT: .LBB71_3: # in Loop: Header=BB71_1 Depth=1
+; LA64-NEXT: sc.w $a6, $a0, 0
+; LA64-NEXT: beqz $a6, .LBB71_1
+; LA64-NEXT: # %bb.4:
+; LA64-NEXT: srl.w $a0, $a5, $a2
+; LA64-NEXT: ret
+;
+; LA64-LAMCAS-LABEL: atomicrmw_max_i16_acq_rel:
+; LA64-LAMCAS: # %bb.0:
+; LA64-LAMCAS-NEXT: move $a2, $a0
+; LA64-LAMCAS-NEXT: ld.hu $a0, $a0, 0
+; LA64-LAMCAS-NEXT: ext.w.h $a3, $a1
+; LA64-LAMCAS-NEXT: .p2align 4, , 16
+; LA64-LAMCAS-NEXT: .LBB71_1: # %atomicrmw.start
+; LA64-LAMCAS-NEXT: # =>This Inner Loop Header: Depth=1
+; LA64-LAMCAS-NEXT: ext.w.h $a4, $a0
+; LA64-LAMCAS-NEXT: slt $a5, $a3, $a4
+; LA64-LAMCAS-NEXT: masknez $a6, $a1, $a5
+; LA64-LAMCAS-NEXT: maskeqz $a5, $a0, $a5
+; LA64-LAMCAS-NEXT: or $a5, $a5, $a6
+; LA64-LAMCAS-NEXT: amcas_db.h $a0, $a5, $a2
+; LA64-LAMCAS-NEXT: bne $a0, $a4, .LBB71_1
+; LA64-LAMCAS-NEXT: # %bb.2: # %atomicrmw.end
+; LA64-LAMCAS-NEXT: ret
+ %1 = atomicrmw max ptr %a, i16 %b acq_rel
+ ret i16 %1
+}
+
+define i8 @atomicrmw_min_i8_acq_rel(ptr %a, i8 %b) nounwind {
+; LA64-LABEL: atomicrmw_min_i8_acq_rel:
+; LA64: # %bb.0:
+; LA64-NEXT: slli.d $a2, $a0, 3
+; LA64-NEXT: bstrins.d $a0, $zero, 1, 0
+; LA64-NEXT: andi $a3, $a2, 24
+; LA64-NEXT: ori $a4, $zero, 255
+; LA64-NEXT: sll.w $a4, $a4, $a2
+; LA64-NEXT: ext.w.b $a1, $a1
+; LA64-NEXT: sll.w $a1, $a1, $a2
+; LA64-NEXT: xori $a3, $a3, 56
+; LA64-NEXT: .LBB72_1: # =>This Inner Loop Header: Depth=1
+; LA64-NEXT: ll.w $a5, $a0, 0
+; LA64-NEXT: and $a7, $a5, $a4
+; LA64-NEXT: move $a6, $a5
+; LA64-NEXT: sll.w $a7, $a7, $a3
+; LA64-NEXT: sra.w $a7, $a7, $a3
+; LA64-NEXT: bge $a1, $a7, .LBB72_3
+; LA64-NEXT: # %bb.2: # in Loop: Header=BB72_1 Depth=1
+; LA64-NEXT: xor $a6, $a5, $a1
+; LA64-NEXT: and $a6, $a6, $a4
+; LA64-NEXT: xor $a6, $a5, $a6
+; LA64-NEXT: .LBB72_3: # in Loop: Header=BB72_1 Depth=1
+; LA64-NEXT: sc.w $a6, $a0, 0
+; LA64-NEXT: beqz $a6, .LBB72_1
+; LA64-NEXT: # %bb.4:
+; LA64-NEXT: srl.w $a0, $a5, $a2
+; LA64-NEXT: ret
+;
+; LA64-LAMCAS-LABEL: atomicrmw_min_i8_acq_rel:
+; LA64-LAMCAS: # %bb.0:
+; LA64-LAMCAS-NEXT: move $a2, $a0
+; LA64-LAMCAS-NEXT: ld.bu $a0, $a0, 0
+; LA64-LAMCAS-NEXT: ext.w.b $a3, $a1
+; LA64-LAMCAS-NEXT: .p2align 4, , 16
+; LA64-LAMCAS-NEXT: .LBB72_1: # %atomicrmw.start
+; LA64-LAMCAS-NEXT: # =>This Inner Loop Header: Depth=1
+; LA64-LAMCAS-NEXT: ext.w.b $a4, $a0
+; LA64-LAMCAS-NEXT: slt $a5, $a3, $a4
+; LA64-LAMCAS-NEXT: xori $a5, $a5, 1
+; LA64-LAMCAS-NEXT: masknez $a6, $a1, $a5
+; LA64-LAMCAS-NEXT: maskeqz $a5, $a0, $a5
+; LA64-LAMCAS-NEXT: or $a5, $a5, $a6
+; LA64-LAMCAS-NEXT: amcas_db.b $a0, $a5, $a2
+; LA64-LAMCAS-NEXT: bne $a0, $a4, .LBB72_1
+; LA64-LAMCAS-NEXT: # %bb.2: # %atomicrmw.end
+; LA64-LAMCAS-NEXT: ret
+ %1 = atomicrmw min ptr %a, i8 %b acq_rel
+ ret i8 %1
+}
+
+define i16 @atomicrmw_min_i16_acq_rel(ptr %a, i16 %b) nounwind {
+; LA64-LABEL: atomicrmw_min_i16_acq_rel:
+; LA64: # %bb.0:
+; LA64-NEXT: slli.d $a2, $a0, 3
+; LA64-NEXT: bstrins.d $a0, $zero, 1, 0
+; LA64-NEXT: andi $a3, $a2, 24
+; LA64-NEXT: lu12i.w $a4, 15
+; LA64-NEXT: ori $a4, $a4, 4095
+; LA64-NEXT: sll.w $a4, $a4, $a2
+; LA64-NEXT: ext.w.h $a1, $a1
+; LA64-NEXT: sll.w $a1, $a1, $a2
+; LA64-NEXT: ori $a5, $zero, 48
+; LA64-NEXT: sub.d $a3, $a5, $a3
+; LA64-NEXT: .LBB73_1: # =>This Inner Loop Header: Depth=1
+; LA64-NEXT: ll.w $a5, $a0, 0
+; LA64-NEXT: and $a7, $a5, $a4
+; LA64-NEXT: move $a6, $a5
+; LA64-NEXT: sll.w $a7, $a7, $a3
+; LA64-NEXT: sra.w $a7, $a7, $a3
+; LA64-NEXT: bge $a1, $a7, .LBB73_3
+; LA64-NEXT: # %bb.2: # in Loop: Header=BB73_1 Depth=1
+; LA64-NEXT: xor $a6, $a5, $a1
+; LA64-NEXT: and $a6, $a6, $a4
+; LA64-NEXT: xor $a6, $a5, $a6
+; LA64-NEXT: .LBB73_3: # in Loop: Header=BB73_1 Depth=1
+; LA64-NEXT: sc.w $a6, $a0, 0
+; LA64-NEXT: beqz $a6, .LBB73_1
+; LA64-NEXT: # %bb.4:
+; LA64-NEXT: srl.w $a0, $a5, $a2
+; LA64-NEXT: ret
+;
+; LA64-LAMCAS-LABEL: atomicrmw_min_i16_acq_rel:
+; LA64-LAMCAS: # %bb.0:
+; LA64-LAMCAS-NEXT: move $a2, $a0
+; LA64-LAMCAS-NEXT: ld.hu $a0, $a0, 0
+; LA64-LAMCAS-NEXT: ext.w.h $a3, $a1
+; LA64-LAMCAS-NEXT: .p2align 4, , 16
+; LA64-LAMCAS-NEXT: .LBB73_1: # %atomicrmw.start
+; LA64-LAMCAS-NEXT: # =>This Inner Loop Header: Depth=1
+; LA64-LAMCAS-NEXT: ext.w.h $a4, $a0
+; LA64-LAMCAS-NEXT: slt $a5, $a3, $a4
+; LA64-LAMCAS-NEXT: xori $a5, $a5, 1
+; LA64-LAMCAS-NEXT: masknez $a6, $a1, $a5
+; LA64-LAMCAS-NEXT: maskeqz $a5, $a0, $a5
+; LA64-LAMCAS-NEXT: or $a5, $a5, $a6
+; LA64-LAMCAS-NEXT: amcas_db.h $a0, $a5, $a2
+; LA64-LAMCAS-NEXT: bne $a0, $a4, .LBB73_1
+; LA64-LAMCAS-NEXT: # %bb.2: # %atomicrmw.end
+; LA64-LAMCAS-NEXT: ret
+ %1 = atomicrmw min ptr %a, i16 %b acq_rel
+ ret i16 %1
+}
+
+define i8 @atomicrmw_nand_i8_acq_rel(ptr %a, i8 %b) nounwind {
+; LA64-LABEL: atomicrmw_nand_i8_acq_rel:
+; LA64: # %bb.0:
+; LA64-NEXT: slli.d $a2, $a0, 3
+; LA64-NEXT: bstrins.d $a0, $zero, 1, 0
+; LA64-NEXT: ori $a3, $zero, 255
+; LA64-NEXT: sll.w $a3, $a3, $a2
+; LA64-NEXT: andi $a1, $a1, 255
+; LA64-NEXT: sll.w $a1, $a1, $a2
+; LA64-NEXT: .LBB74_1: # =>This Inner Loop Header: Depth=1
+; LA64-NEXT: ll.w $a4, $a0, 0
+; LA64-NEXT: and $a5, $a4, $a1
+; LA64-NEXT: nor $a5, $a5, $zero
+; LA64-NEXT: xor $a5, $a4, $a5
+; LA64-NEXT: and $a5, $a5, $a3
+; LA64-NEXT: xor $a5, $a4, $a5
+; LA64-NEXT: sc.w $a5, $a0, 0
+; LA64-NEXT: beqz $a5, .LBB74_1
+; LA64-NEXT: # %bb.2:
+; LA64-NEXT: srl.w $a0, $a4, $a2
+; LA64-NEXT: ret
+;
+; LA64-LAMCAS-LABEL: atomicrmw_nand_i8_acq_rel:
+; LA64-LAMCAS: # %bb.0:
+; LA64-LAMCAS-NEXT: move $a2, $a0
+; LA64-LAMCAS-NEXT: ld.bu $a0, $a0, 0
+; LA64-LAMCAS-NEXT: .p2align 4, , 16
+; LA64-LAMCAS-NEXT: .LBB74_1: # %atomicrmw.start
+; LA64-LAMCAS-NEXT: # =>This Inner Loop Header: Depth=1
+; LA64-LAMCAS-NEXT: and $a3, $a0, $a1
+; LA64-LAMCAS-NEXT: nor $a3, $a3, $zero
+; LA64-LAMCAS-NEXT: ext.w.b $a4, $a0
+; LA64-LAMCAS-NEXT: amcas_db.b $a0, $a3, $a2
+; LA64-LAMCAS-NEXT: bne $a0, $a4, .LBB74_1
+; LA64-LAMCAS-NEXT: # %bb.2: # %atomicrmw.end
+; LA64-LAMCAS-NEXT: ret
+ %1 = atomicrmw nand ptr %a, i8 %b acq_rel
+ ret i8 %1
+}
+
+define i16 @atomicrmw_nand_i16_acq_rel(ptr %a, i16 %b) nounwind {
+; LA64-LABEL: atomicrmw_nand_i16_acq_rel:
+; LA64: # %bb.0:
+; LA64-NEXT: slli.d $a2, $a0, 3
+; LA64-NEXT: bstrins.d $a0, $zero, 1, 0
+; LA64-NEXT: lu12i.w $a3, 15
+; LA64-NEXT: ori $a3, $a3, 4095
+; LA64-NEXT: sll.w $a3, $a3, $a2
+; LA64-NEXT: bstrpick.d $a1, $a1, 15, 0
+; LA64-NEXT: sll.w $a1, $a1, $a2
+; LA64-NEXT: .LBB75_1: # =>This Inner Loop Header: Depth=1
+; LA64-NEXT: ll.w $a4, $a0, 0
+; LA64-NEXT: and $a5, $a4, $a1
+; LA64-NEXT: nor $a5, $a5, $zero
+; LA64-NEXT: xor $a5, $a4, $a5
+; LA64-NEXT: and $a5, $a5, $a3
+; LA64-NEXT: xor $a5, $a4, $a5
+; LA64-NEXT: sc.w $a5, $a0, 0
+; LA64-NEXT: beqz $a5, .LBB75_1
+; LA64-NEXT: # %bb.2:
+; LA64-NEXT: srl.w $a0, $a4, $a2
+; LA64-NEXT: ret
+;
+; LA64-LAMCAS-LABEL: atomicrmw_nand_i16_acq_rel:
+; LA64-LAMCAS: # %bb.0:
+; LA64-LAMCAS-NEXT: move $a2, $a0
+; LA64-LAMCAS-NEXT: ld.hu $a0, $a0, 0
+; LA64-LAMCAS-NEXT: .p2align 4, , 16
+; LA64-LAMCAS-NEXT: .LBB75_1: # %atomicrmw.start
+; LA64-LAMCAS-NEXT: # =>This Inner Loop Header: Depth=1
+; LA64-LAMCAS-NEXT: and $a3, $a0, $a1
+; LA64-LAMCAS-NEXT: nor $a3, $a3, $zero
+; LA64-LAMCAS-NEXT: ext.w.h $a4, $a0
+; LA64-LAMCAS-NEXT: amcas_db.h $a0, $a3, $a2
+; LA64-LAMCAS-NEXT: bne $a0, $a4, .LBB75_1
+; LA64-LAMCAS-NEXT: # %bb.2: # %atomicrmw.end
+; LA64-LAMCAS-NEXT: ret
+ %1 = atomicrmw nand ptr %a, i16 %b acq_rel
+ ret i16 %1
+
+}
+
+define i32 @atomicrmw_nand_i32_acq_rel(ptr %a, i32 %b) nounwind {
+; LA64-LABEL: atomicrmw_nand_i32_acq_rel:
+; LA64: # %bb.0:
+; LA64-NEXT: .LBB76_1: # =>This Inner Loop Header: Depth=1
+; LA64-NEXT: ll.w $a2, $a0, 0
+; LA64-NEXT: and $a3, $a2, $a1
+; LA64-NEXT: nor $a3, $a3, $zero
+; LA64-NEXT: sc.w $a3, $a0, 0
+; LA64-NEXT: beqz $a3, .LBB76_1
+; LA64-NEXT: # %bb.2:
+; LA64-NEXT: move $a0, $a2
+; LA64-NEXT: ret
+;
+; LA64-LAMCAS-LABEL: atomicrmw_nand_i32_acq_rel:
+; LA64-LAMCAS: # %bb.0:
+; LA64-LAMCAS-NEXT: move $a2, $a0
+; LA64-LAMCAS-NEXT: ld.w $a0, $a0, 0
+; LA64-LAMCAS-NEXT: .p2align 4, , 16
+; LA64-LAMCAS-NEXT: .LBB76_1: # %atomicrmw.start
+; LA64-LAMCAS-NEXT: # =>This Inner Loop Header: Depth=1
+; LA64-LAMCAS-NEXT: move $a3, $a0
+; LA64-LAMCAS-NEXT: and $a4, $a0, $a1
+; LA64-LAMCAS-NEXT: nor $a4, $a4, $zero
+; LA64-LAMCAS-NEXT: amcas_db.w $a0, $a4, $a2
+; LA64-LAMCAS-NEXT: bne $a0, $a3, .LBB76_1
+; LA64-LAMCAS-NEXT: # %bb.2: # %atomicrmw.end
+; LA64-LAMCAS-NEXT: ret
+ %1 = atomicrmw nand ptr %a, i32 %b acq_rel
+ ret i32 %1
+}
+
+define i64 @atomicrmw_nand_i64_acq_rel(ptr %a, i64 %b) nounwind {
+; LA64-LABEL: atomicrmw_nand_i64_acq_rel:
+; LA64: # %bb.0:
+; LA64-NEXT: .LBB77_1: # =>This Inner Loop Header: Depth=1
+; LA64-NEXT: ll.d $a2, $a0, 0
+; LA64-NEXT: and $a3, $a2, $a1
+; LA64-NEXT: nor $a3, $a3, $zero
+; LA64-NEXT: sc.d $a3, $a0, 0
+; LA64-NEXT: beqz $a3, .LBB77_1
+; LA64-NEXT: # %bb.2:
+; LA64-NEXT: move $a0, $a2
+; LA64-NEXT: ret
+;
+; LA64-LAMCAS-LABEL: atomicrmw_nand_i64_acq_rel:
+; LA64-LAMCAS: # %bb.0:
+; LA64-LAMCAS-NEXT: move $a2, $a0
+; LA64-LAMCAS-NEXT: ld.d $a0, $a0, 0
+; LA64-LAMCAS-NEXT: .p2align 4, , 16
+; LA64-LAMCAS-NEXT: .LBB77_1: # %atomicrmw.start
+; LA64-LAMCAS-NEXT: # =>This Inner Loop Header: Depth=1
+; LA64-LAMCAS-NEXT: move $a3, $a0
+; LA64-LAMCAS-NEXT: and $a4, $a0, $a1
+; LA64-LAMCAS-NEXT: nor $a4, $a4, $zero
+; LA64-LAMCAS-NEXT: amcas_db.d $a0, $a4, $a2
+; LA64-LAMCAS-NEXT: bne $a0, $a3, .LBB77_1
+; LA64-LAMCAS-NEXT: # %bb.2: # %atomicrmw.end
+; LA64-LAMCAS-NEXT: ret
+ %1 = atomicrmw nand ptr %a, i64 %b acq_rel
+ ret i64 %1
+}
+
+
+
+define i8 @atomicrmw_and_i8_acq_rel(ptr %a, i8 %b) nounwind {
+; LA64-LABEL: atomicrmw_and_i8_acq_rel:
+; LA64: # %bb.0:
+; LA64-NEXT: slli.d $a2, $a0, 3
+; LA64-NEXT: bstrins.d $a0, $zero, 1, 0
+; LA64-NEXT: ori $a3, $zero, 255
+; LA64-NEXT: sll.w $a3, $a3, $a2
+; LA64-NEXT: andi $a1, $a1, 255
+; LA64-NEXT: sll.w $a1, $a1, $a2
+; LA64-NEXT: orn $a1, $a1, $a3
+; LA64-NEXT: amand_db.w $a3, $a1, $a0
+; LA64-NEXT: srl.w $a0, $a3, $a2
+; LA64-NEXT: ret
+;
+; LA64-LAMCAS-LABEL: atomicrmw_and_i8_acq_rel:
+; LA64-LAMCAS: # %bb.0:
+; LA64-LAMCAS-NEXT: move $a2, $a0
+; LA64-LAMCAS-NEXT: ld.bu $a0, $a0, 0
+; LA64-LAMCAS-NEXT: .p2align 4, , 16
+; LA64-LAMCAS-NEXT: .LBB78_1: # %atomicrmw.start
+; LA64-LAMCAS-NEXT: # =>This Inner Loop Header: Depth=1
+; LA64-LAMCAS-NEXT: and $a3, $a0, $a1
+; LA64-LAMCAS-NEXT: ext.w.b $a4, $a0
+; LA64-LAMCAS-NEXT: amcas_db.b $a0, $a3, $a2
+; LA64-LAMCAS-NEXT: bne $a0, $a4, .LBB78_1
+; LA64-LAMCAS-NEXT: # %bb.2: # %atomicrmw.end
+; LA64-LAMCAS-NEXT: ret
+ %1 = atomicrmw and ptr %a, i8 %b acq_rel
+ ret i8 %1
+}
+
+define i16 @atomicrmw_and_i16_acq_rel(ptr %a, i16 %b) nounwind {
+; LA64-LABEL: atomicrmw_and_i16_acq_rel:
+; LA64: # %bb.0:
+; LA64-NEXT: slli.d $a2, $a0, 3
+; LA64-NEXT: bstrins.d $a0, $zero, 1, 0
+; LA64-NEXT: lu12i.w $a3, 15
+; LA64-NEXT: ori $a3, $a3, 4095
+; LA64-NEXT: sll.w $a3, $a3, $a2
+; LA64-NEXT: bstrpick.d $a1, $a1, 15, 0
+; LA64-NEXT: sll.w $a1, $a1, $a2
+; LA64-NEXT: orn $a1, $a1, $a3
+; LA64-NEXT: amand_db.w $a3, $a1, $a0
+; LA64-NEXT: srl.w $a0, $a3, $a2
+; LA64-NEXT: ret
+;
+; LA64-LAMCAS-LABEL: atomicrmw_and_i16_acq_rel:
+; LA64-LAMCAS: # %bb.0:
+; LA64-LAMCAS-NEXT: move $a2, $a0
+; LA64-LAMCAS-NEXT: ld.hu $a0, $a0, 0
+; LA64-LAMCAS-NEXT: .p2align 4, , 16
+; LA64-LAMCAS-NEXT: .LBB79_1: # %atomicrmw.start
+; LA64-LAMCAS-NEXT: # =>This Inner Loop Header: Depth=1
+; LA64-LAMCAS-NEXT: and $a3, $a0, $a1
+; LA64-LAMCAS-NEXT: ext.w.h $a4, $a0
+; LA64-LAMCAS-NEXT: amcas_db.h $a0, $a3, $a2
+; LA64-LAMCAS-NEXT: bne $a0, $a4, .LBB79_1
+; LA64-LAMCAS-NEXT: # %bb.2: # %atomicrmw.end
+; LA64-LAMCAS-NEXT: ret
+ %1 = atomicrmw and ptr %a, i16 %b acq_rel
+ ret i16 %1
+
+}
+
+define i8 @atomicrmw_or_i8_acq_rel(ptr %a, i8 %b) nounwind {
+; LA64-LABEL: atomicrmw_or_i8_acq_rel:
+; LA64: # %bb.0:
+; LA64-NEXT: slli.d $a2, $a0, 3
+; LA64-NEXT: bstrins.d $a0, $zero, 1, 0
+; LA64-NEXT: andi $a1, $a1, 255
+; LA64-NEXT: sll.w $a1, $a1, $a2
+; LA64-NEXT: amor_db.w $a3, $a1, $a0
+; LA64-NEXT: srl.w $a0, $a3, $a2
+; LA64-NEXT: ret
+;
+; LA64-LAMCAS-LABEL: atomicrmw_or_i8_acq_rel:
+; LA64-LAMCAS: # %bb.0:
+; LA64-LAMCAS-NEXT: move $a2, $a0
+; LA64-LAMCAS-NEXT: ld.bu $a0, $a0, 0
+; LA64-LAMCAS-NEXT: .p2align 4, , 16
+; LA64-LAMCAS-NEXT: .LBB80_1: # %atomicrmw.start
+; LA64-LAMCAS-NEXT: # =>This Inner Loop Header: Depth=1
+; LA64-LAMCAS-NEXT: or $a3, $a0, $a1
+; LA64-LAMCAS-NEXT: ext.w.b $a4, $a0
+; LA64-LAMCAS-NEXT: amcas_db.b $a0, $a3, $a2
+; LA64-LAMCAS-NEXT: bne $a0, $a4, .LBB80_1
+; LA64-LAMCAS-NEXT: # %bb.2: # %atomicrmw.end
+; LA64-LAMCAS-NEXT: ret
+ %1 = atomicrmw or ptr %a, i8 %b acq_rel
+ ret i8 %1
+}
+
+define i16 @atomicrmw_or_i16_acq_rel(ptr %a, i16 %b) nounwind {
+; LA64-LABEL: atomicrmw_or_i16_acq_rel:
+; LA64: # %bb.0:
+; LA64-NEXT: slli.d $a2, $a0, 3
+; LA64-NEXT: bstrins.d $a0, $zero, 1, 0
+; LA64-NEXT: bstrpick.d $a1, $a1, 15, 0
+; LA64-NEXT: sll.w $a1, $a1, $a2
+; LA64-NEXT: amor_db.w $a3, $a1, $a0
+; LA64-NEXT: srl.w $a0, $a3, $a2
+; LA64-NEXT: ret
+;
+; LA64-LAMCAS-LABEL: atomicrmw_or_i16_acq_rel:
+; LA64-LAMCAS: # %bb.0:
+; LA64-LAMCAS-NEXT: move $a2, $a0
+; LA64-LAMCAS-NEXT: ld.hu $a0, $a0, 0
+; LA64-LAMCAS-NEXT: .p2align 4, , 16
+; LA64-LAMCAS-NEXT: .LBB81_1: # %atomicrmw.start
+; LA64-LAMCAS-NEXT: # =>This Inner Loop Header: Depth=1
+; LA64-LAMCAS-NEXT: or $a3, $a0, $a1
+; LA64-LAMCAS-NEXT: ext.w.h $a4, $a0
+; LA64-LAMCAS-NEXT: amcas_db.h $a0, $a3, $a2
+; LA64-LAMCAS-NEXT: bne $a0, $a4, .LBB81_1
+; LA64-LAMCAS-NEXT: # %bb.2: # %atomicrmw.end
+; LA64-LAMCAS-NEXT: ret
+ %1 = atomicrmw or ptr %a, i16 %b acq_rel
+ ret i16 %1
+
+}
+
+define i8 @atomicrmw_xor_i8_acq_rel(ptr %a, i8 %b) nounwind {
+; LA64-LABEL: atomicrmw_xor_i8_acq_rel:
+; LA64: # %bb.0:
+; LA64-NEXT: slli.d $a2, $a0, 3
+; LA64-NEXT: bstrins.d $a0, $zero, 1, 0
+; LA64-NEXT: andi $a1, $a1, 255
+; LA64-NEXT: sll.w $a1, $a1, $a2
+; LA64-NEXT: amxor_db.w $a3, $a1, $a0
+; LA64-NEXT: srl.w $a0, $a3, $a2
+; LA64-NEXT: ret
+;
+; LA64-LAMCAS-LABEL: atomicrmw_xor_i8_acq_rel:
+; LA64-LAMCAS: # %bb.0:
+; LA64-LAMCAS-NEXT: move $a2, $a0
+; LA64-LAMCAS-NEXT: ld.bu $a0, $a0, 0
+; LA64-LAMCAS-NEXT: .p2align 4, , 16
+; LA64-LAMCAS-NEXT: .LBB82_1: # %atomicrmw.start
+; LA64-LAMCAS-NEXT: # =>This Inner Loop Header: Depth=1
+; LA64-LAMCAS-NEXT: xor $a3, $a0, $a1
+; LA64-LAMCAS-NEXT: ext.w.b $a4, $a0
+; LA64-LAMCAS-NEXT: amcas_db.b $a0, $a3, $a2
+; LA64-LAMCAS-NEXT: bne $a0, $a4, .LBB82_1
+; LA64-LAMCAS-NEXT: # %bb.2: # %atomicrmw.end
+; LA64-LAMCAS-NEXT: ret
+ %1 = atomicrmw xor ptr %a, i8 %b acq_rel
+ ret i8 %1
+}
+
+define i16 @atomicrmw_xor_i16_acq_rel(ptr %a, i16 %b) nounwind {
+; LA64-LABEL: atomicrmw_xor_i16_acq_rel:
+; LA64: # %bb.0:
+; LA64-NEXT: slli.d $a2, $a0, 3
+; LA64-NEXT: bstrins.d $a0, $zero, 1, 0
+; LA64-NEXT: bstrpick.d $a1, $a1, 15, 0
+; LA64-NEXT: sll.w $a1, $a1, $a2
+; LA64-NEXT: amxor_db.w $a3, $a1, $a0
+; LA64-NEXT: srl.w $a0, $a3, $a2
+; LA64-NEXT: ret
+;
+; LA64-LAMCAS-LABEL: atomicrmw_xor_i16_acq_rel:
+; LA64-LAMCAS: # %bb.0:
+; LA64-LAMCAS-NEXT: move $a2, $a0
+; LA64-LAMCAS-NEXT: ld.hu $a0, $a0, 0
+; LA64-LAMCAS-NEXT: .p2align 4, , 16
+; LA64-LAMCAS-NEXT: .LBB83_1: # %atomicrmw.start
+; LA64-LAMCAS-NEXT: # =>This Inner Loop Header: Depth=1
+; LA64-LAMCAS-NEXT: xor $a3, $a0, $a1
+; LA64-LAMCAS-NEXT: ext.w.h $a4, $a0
+; LA64-LAMCAS-NEXT: amcas_db.h $a0, $a3, $a2
+; LA64-LAMCAS-NEXT: bne $a0, $a4, .LBB83_1
+; LA64-LAMCAS-NEXT: # %bb.2: # %atomicrmw.end
+; LA64-LAMCAS-NEXT: ret
+ %1 = atomicrmw xor ptr %a, i16 %b acq_rel
+ ret i16 %1
+
+}
+
+define i8 @atomicrmw_xchg_i8_seq_cst(ptr %a, i8 %b) nounwind {
+; LA64-LABEL: atomicrmw_xchg_i8_seq_cst:
+; LA64: # %bb.0:
+; LA64-NEXT: slli.d $a2, $a0, 3
+; LA64-NEXT: bstrins.d $a0, $zero, 1, 0
+; LA64-NEXT: ori $a3, $zero, 255
+; LA64-NEXT: sll.w $a3, $a3, $a2
+; LA64-NEXT: andi $a1, $a1, 255
+; LA64-NEXT: sll.w $a1, $a1, $a2
+; LA64-NEXT: .LBB84_1: # =>This Inner Loop Header: Depth=1
+; LA64-NEXT: ll.w $a4, $a0, 0
+; LA64-NEXT: addi.w $a5, $a1, 0
+; LA64-NEXT: xor $a5, $a4, $a5
+; LA64-NEXT: and $a5, $a5, $a3
+; LA64-NEXT: xor $a5, $a4, $a5
+; LA64-NEXT: sc.w $a5, $a0, 0
+; LA64-NEXT: beqz $a5, .LBB84_1
+; LA64-NEXT: # %bb.2:
+; LA64-NEXT: srl.w $a0, $a4, $a2
+; LA64-NEXT: ret
+;
+; LA64-LAMCAS-LABEL: atomicrmw_xchg_i8_seq_cst:
+; LA64-LAMCAS: # %bb.0:
+; LA64-LAMCAS-NEXT: ld.bu $a2, $a0, 0
+; LA64-LAMCAS-NEXT: .p2align 4, , 16
+; LA64-LAMCAS-NEXT: .LBB84_1: # %atomicrmw.start
+; LA64-LAMCAS-NEXT: # =>This Inner Loop Header: Depth=1
+; LA64-LAMCAS-NEXT: ext.w.b $a3, $a2
+; LA64-LAMCAS-NEXT: amcas_db.b $a2, $a1, $a0
+; LA64-LAMCAS-NEXT: bne $a2, $a3, .LBB84_1
+; LA64-LAMCAS-NEXT: # %bb.2: # %atomicrmw.end
+; LA64-LAMCAS-NEXT: move $a0, $a2
+; LA64-LAMCAS-NEXT: ret
+ %1 = atomicrmw xchg ptr %a, i8 %b seq_cst
+ ret i8 %1
+}
+
+define i8 @atomicrmw_xchg_0_i8_seq_cst(ptr %a) nounwind {
+; LA64-LABEL: atomicrmw_xchg_0_i8_seq_cst:
+; LA64: # %bb.0:
+; LA64-NEXT: slli.d $a1, $a0, 3
+; LA64-NEXT: bstrins.d $a0, $zero, 1, 0
+; LA64-NEXT: ori $a2, $zero, 255
+; LA64-NEXT: sll.w $a2, $a2, $a1
+; LA64-NEXT: nor $a2, $a2, $zero
+; LA64-NEXT: amand_db.w $a3, $a2, $a0
+; LA64-NEXT: srl.w $a0, $a3, $a1
+; LA64-NEXT: ret
+;
+; LA64-LAMCAS-LABEL: atomicrmw_xchg_0_i8_seq_cst:
+; LA64-LAMCAS: # %bb.0:
+; LA64-LAMCAS-NEXT: move $a1, $a0
+; LA64-LAMCAS-NEXT: ld.bu $a0, $a0, 0
+; LA64-LAMCAS-NEXT: .p2align 4, , 16
+; LA64-LAMCAS-NEXT: .LBB85_1: # %atomicrmw.start
+; LA64-LAMCAS-NEXT: # =>This Inner Loop Header: Depth=1
+; LA64-LAMCAS-NEXT: ext.w.b $a2, $a0
+; LA64-LAMCAS-NEXT: amcas_db.b $a0, $zero, $a1
+; LA64-LAMCAS-NEXT: bne $a0, $a2, .LBB85_1
+; LA64-LAMCAS-NEXT: # %bb.2: # %atomicrmw.end
+; LA64-LAMCAS-NEXT: ret
+ %1 = atomicrmw xchg ptr %a, i8 0 seq_cst
+ ret i8 %1
+}
+
+define i8 @atomicrmw_xchg_minus_1_i8_seq_cst(ptr %a) nounwind {
+; LA64-LABEL: atomicrmw_xchg_minus_1_i8_seq_cst:
+; LA64: # %bb.0:
+; LA64-NEXT: slli.d $a1, $a0, 3
+; LA64-NEXT: bstrins.d $a0, $zero, 1, 0
+; LA64-NEXT: ori $a2, $zero, 255
+; LA64-NEXT: sll.w $a2, $a2, $a1
+; LA64-NEXT: amor_db.w $a3, $a2, $a0
+; LA64-NEXT: srl.w $a0, $a3, $a1
+; LA64-NEXT: ret
+;
+; LA64-LAMCAS-LABEL: atomicrmw_xchg_minus_1_i8_seq_cst:
+; LA64-LAMCAS: # %bb.0:
+; LA64-LAMCAS-NEXT: move $a1, $a0
+; LA64-LAMCAS-NEXT: ld.bu $a0, $a0, 0
+; LA64-LAMCAS-NEXT: addi.w $a2, $zero, -1
+; LA64-LAMCAS-NEXT: .p2align 4, , 16
+; LA64-LAMCAS-NEXT: .LBB86_1: # %atomicrmw.start
+; LA64-LAMCAS-NEXT: # =>This Inner Loop Header: Depth=1
+; LA64-LAMCAS-NEXT: ext.w.b $a3, $a0
+; LA64-LAMCAS-NEXT: amcas_db.b $a0, $a2, $a1
+; LA64-LAMCAS-NEXT: bne $a0, $a3, .LBB86_1
+; LA64-LAMCAS-NEXT: # %bb.2: # %atomicrmw.end
+; LA64-LAMCAS-NEXT: ret
+ %1 = atomicrmw xchg ptr %a, i8 -1 seq_cst
+ ret i8 %1
+}
+
+define i16 @atomicrmw_xchg_i16_seq_cst(ptr %a, i16 %b) nounwind {
+; LA64-LABEL: atomicrmw_xchg_i16_seq_cst:
+; LA64: # %bb.0:
+; LA64-NEXT: slli.d $a2, $a0, 3
+; LA64-NEXT: bstrins.d $a0, $zero, 1, 0
+; LA64-NEXT: lu12i.w $a3, 15
+; LA64-NEXT: ori $a3, $a3, 4095
+; LA64-NEXT: sll.w $a3, $a3, $a2
+; LA64-NEXT: bstrpick.d $a1, $a1, 15, 0
+; LA64-NEXT: sll.w $a1, $a1, $a2
+; LA64-NEXT: .LBB87_1: # =>This Inner Loop Header: Depth=1
+; LA64-NEXT: ll.w $a4, $a0, 0
+; LA64-NEXT: addi.w $a5, $a1, 0
+; LA64-NEXT: xor $a5, $a4, $a5
+; LA64-NEXT: and $a5, $a5, $a3
+; LA64-NEXT: xor $a5, $a4, $a5
+; LA64-NEXT: sc.w $a5, $a0, 0
+; LA64-NEXT: beqz $a5, .LBB87_1
+; LA64-NEXT: # %bb.2:
+; LA64-NEXT: srl.w $a0, $a4, $a2
+; LA64-NEXT: ret
+;
+; LA64-LAMCAS-LABEL: atomicrmw_xchg_i16_seq_cst:
+; LA64-LAMCAS: # %bb.0:
+; LA64-LAMCAS-NEXT: ld.hu $a2, $a0, 0
+; LA64-LAMCAS-NEXT: .p2align 4, , 16
+; LA64-LAMCAS-NEXT: .LBB87_1: # %atomicrmw.start
+; LA64-LAMCAS-NEXT: # =>This Inner Loop Header: Depth=1
+; LA64-LAMCAS-NEXT: ext.w.h $a3, $a2
+; LA64-LAMCAS-NEXT: amcas_db.h $a2, $a1, $a0
+; LA64-LAMCAS-NEXT: bne $a2, $a3, .LBB87_1
+; LA64-LAMCAS-NEXT: # %bb.2: # %atomicrmw.end
+; LA64-LAMCAS-NEXT: move $a0, $a2
+; LA64-LAMCAS-NEXT: ret
+ %1 = atomicrmw xchg ptr %a, i16 %b seq_cst
+ ret i16 %1
+}
+
+define i16 @atomicrmw_xchg_0_i16_seq_cst(ptr %a) nounwind {
+; LA64-LABEL: atomicrmw_xchg_0_i16_seq_cst:
+; LA64: # %bb.0:
+; LA64-NEXT: slli.d $a1, $a0, 3
+; LA64-NEXT: bstrins.d $a0, $zero, 1, 0
+; LA64-NEXT: lu12i.w $a2, 15
+; LA64-NEXT: ori $a2, $a2, 4095
+; LA64-NEXT: sll.w $a2, $a2, $a1
+; LA64-NEXT: nor $a2, $a2, $zero
+; LA64-NEXT: amand_db.w $a3, $a2, $a0
+; LA64-NEXT: srl.w $a0, $a3, $a1
+; LA64-NEXT: ret
+;
+; LA64-LAMCAS-LABEL: atomicrmw_xchg_0_i16_seq_cst:
+; LA64-LAMCAS: # %bb.0:
+; LA64-LAMCAS-NEXT: move $a1, $a0
+; LA64-LAMCAS-NEXT: ld.hu $a0, $a0, 0
+; LA64-LAMCAS-NEXT: .p2align 4, , 16
+; LA64-LAMCAS-NEXT: .LBB88_1: # %atomicrmw.start
+; LA64-LAMCAS-NEXT: # =>This Inner Loop Header: Depth=1
+; LA64-LAMCAS-NEXT: ext.w.h $a2, $a0
+; LA64-LAMCAS-NEXT: amcas_db.h $a0, $zero, $a1
+; LA64-LAMCAS-NEXT: bne $a0, $a2, .LBB88_1
+; LA64-LAMCAS-NEXT: # %bb.2: # %atomicrmw.end
+; LA64-LAMCAS-NEXT: ret
+ %1 = atomicrmw xchg ptr %a, i16 0 seq_cst
+ ret i16 %1
+}
+
+define i16 @atomicrmw_xchg_minus_1_i16_seq_cst(ptr %a) nounwind {
+; LA64-LABEL: atomicrmw_xchg_minus_1_i16_seq_cst:
+; LA64: # %bb.0:
+; LA64-NEXT: slli.d $a1, $a0, 3
+; LA64-NEXT: bstrins.d $a0, $zero, 1, 0
+; LA64-NEXT: lu12i.w $a2, 15
+; LA64-NEXT: ori $a2, $a2, 4095
+; LA64-NEXT: sll.w $a2, $a2, $a1
+; LA64-NEXT: amor_db.w $a3, $a2, $a0
+; LA64-NEXT: srl.w $a0, $a3, $a1
+; LA64-NEXT: ret
+;
+; LA64-LAMCAS-LABEL: atomicrmw_xchg_minus_1_i16_seq_cst:
+; LA64-LAMCAS: # %bb.0:
+; LA64-LAMCAS-NEXT: move $a1, $a0
+; LA64-LAMCAS-NEXT: ld.hu $a0, $a0, 0
+; LA64-LAMCAS-NEXT: addi.w $a2, $zero, -1
+; LA64-LAMCAS-NEXT: .p2align 4, , 16
+; LA64-LAMCAS-NEXT: .LBB89_1: # %atomicrmw.start
+; LA64-LAMCAS-NEXT: # =>This Inner Loop Header: Depth=1
+; LA64-LAMCAS-NEXT: ext.w.h $a3, $a0
+; LA64-LAMCAS-NEXT: amcas_db.h $a0, $a2, $a1
+; LA64-LAMCAS-NEXT: bne $a0, $a3, .LBB89_1
+; LA64-LAMCAS-NEXT: # %bb.2: # %atomicrmw.end
+; LA64-LAMCAS-NEXT: ret
+ %1 = atomicrmw xchg ptr %a, i16 -1 seq_cst
+ ret i16 %1
+
+}
+
+define i8 @atomicrmw_add_i8_seq_cst(ptr %a, i8 %b) nounwind {
+; LA64-LABEL: atomicrmw_add_i8_seq_cst:
+; LA64: # %bb.0:
+; LA64-NEXT: slli.d $a2, $a0, 3
+; LA64-NEXT: bstrins.d $a0, $zero, 1, 0
+; LA64-NEXT: ori $a3, $zero, 255
+; LA64-NEXT: sll.w $a3, $a3, $a2
+; LA64-NEXT: andi $a1, $a1, 255
+; LA64-NEXT: sll.w $a1, $a1, $a2
+; LA64-NEXT: .LBB90_1: # =>This Inner Loop Header: Depth=1
+; LA64-NEXT: ll.w $a4, $a0, 0
+; LA64-NEXT: add.w $a5, $a4, $a1
+; LA64-NEXT: xor $a5, $a4, $a5
+; LA64-NEXT: and $a5, $a5, $a3
+; LA64-NEXT: xor $a5, $a4, $a5
+; LA64-NEXT: sc.w $a5, $a0, 0
+; LA64-NEXT: beqz $a5, .LBB90_1
+; LA64-NEXT: # %bb.2:
+; LA64-NEXT: srl.w $a0, $a4, $a2
+; LA64-NEXT: ret
+;
+; LA64-LAMCAS-LABEL: atomicrmw_add_i8_seq_cst:
+; LA64-LAMCAS: # %bb.0:
+; LA64-LAMCAS-NEXT: move $a2, $a0
+; LA64-LAMCAS-NEXT: ld.bu $a0, $a0, 0
+; LA64-LAMCAS-NEXT: .p2align 4, , 16
+; LA64-LAMCAS-NEXT: .LBB90_1: # %atomicrmw.start
+; LA64-LAMCAS-NEXT: # =>This Inner Loop Header: Depth=1
+; LA64-LAMCAS-NEXT: add.d $a3, $a0, $a1
+; LA64-LAMCAS-NEXT: ext.w.b $a4, $a0
+; LA64-LAMCAS-NEXT: amcas_db.b $a0, $a3, $a2
+; LA64-LAMCAS-NEXT: bne $a0, $a4, .LBB90_1
+; LA64-LAMCAS-NEXT: # %bb.2: # %atomicrmw.end
+; LA64-LAMCAS-NEXT: ret
+ %1 = atomicrmw add ptr %a, i8 %b seq_cst
+ ret i8 %1
+}
+
+define i16 @atomicrmw_add_i16_seq_cst(ptr %a, i16 %b) nounwind {
+; LA64-LABEL: atomicrmw_add_i16_seq_cst:
+; LA64: # %bb.0:
+; LA64-NEXT: slli.d $a2, $a0, 3
+; LA64-NEXT: bstrins.d $a0, $zero, 1, 0
+; LA64-NEXT: lu12i.w $a3, 15
+; LA64-NEXT: ori $a3, $a3, 4095
+; LA64-NEXT: sll.w $a3, $a3, $a2
+; LA64-NEXT: bstrpick.d $a1, $a1, 15, 0
+; LA64-NEXT: sll.w $a1, $a1, $a2
+; LA64-NEXT: .LBB91_1: # =>This Inner Loop Header: Depth=1
+; LA64-NEXT: ll.w $a4, $a0, 0
+; LA64-NEXT: add.w $a5, $a4, $a1
+; LA64-NEXT: xor $a5, $a4, $a5
+; LA64-NEXT: and $a5, $a5, $a3
+; LA64-NEXT: xor $a5, $a4, $a5
+; LA64-NEXT: sc.w $a5, $a0, 0
+; LA64-NEXT: beqz $a5, .LBB91_1
+; LA64-NEXT: # %bb.2:
+; LA64-NEXT: srl.w $a0, $a4, $a2
+; LA64-NEXT: ret
+;
+; LA64-LAMCAS-LABEL: atomicrmw_add_i16_seq_cst:
+; LA64-LAMCAS: # %bb.0:
+; LA64-LAMCAS-NEXT: move $a2, $a0
+; LA64-LAMCAS-NEXT: ld.hu $a0, $a0, 0
+; LA64-LAMCAS-NEXT: .p2align 4, , 16
+; LA64-LAMCAS-NEXT: .LBB91_1: # %atomicrmw.start
+; LA64-LAMCAS-NEXT: # =>This Inner Loop Header: Depth=1
+; LA64-LAMCAS-NEXT: add.d $a3, $a0, $a1
+; LA64-LAMCAS-NEXT: ext.w.h $a4, $a0
+; LA64-LAMCAS-NEXT: amcas_db.h $a0, $a3, $a2
+; LA64-LAMCAS-NEXT: bne $a0, $a4, .LBB91_1
+; LA64-LAMCAS-NEXT: # %bb.2: # %atomicrmw.end
+; LA64-LAMCAS-NEXT: ret
+ %1 = atomicrmw add ptr %a, i16 %b seq_cst
+ ret i16 %1
+
+}
+
+define i8 @atomicrmw_sub_i8_seq_cst(ptr %a, i8 %b) nounwind {
+; LA64-LABEL: atomicrmw_sub_i8_seq_cst:
+; LA64: # %bb.0:
+; LA64-NEXT: slli.d $a2, $a0, 3
+; LA64-NEXT: bstrins.d $a0, $zero, 1, 0
+; LA64-NEXT: ori $a3, $zero, 255
+; LA64-NEXT: sll.w $a3, $a3, $a2
+; LA64-NEXT: andi $a1, $a1, 255
+; LA64-NEXT: sll.w $a1, $a1, $a2
+; LA64-NEXT: .LBB92_1: # =>This Inner Loop Header: Depth=1
+; LA64-NEXT: ll.w $a4, $a0, 0
+; LA64-NEXT: sub.w $a5, $a4, $a1
+; LA64-NEXT: xor $a5, $a4, $a5
+; LA64-NEXT: and $a5, $a5, $a3
+; LA64-NEXT: xor $a5, $a4, $a5
+; LA64-NEXT: sc.w $a5, $a0, 0
+; LA64-NEXT: beqz $a5, .LBB92_1
+; LA64-NEXT: # %bb.2:
+; LA64-NEXT: srl.w $a0, $a4, $a2
+; LA64-NEXT: ret
+;
+; LA64-LAMCAS-LABEL: atomicrmw_sub_i8_seq_cst:
+; LA64-LAMCAS: # %bb.0:
+; LA64-LAMCAS-NEXT: move $a2, $a0
+; LA64-LAMCAS-NEXT: ld.bu $a0, $a0, 0
+; LA64-LAMCAS-NEXT: .p2align 4, , 16
+; LA64-LAMCAS-NEXT: .LBB92_1: # %atomicrmw.start
+; LA64-LAMCAS-NEXT: # =>This Inner Loop Header: Depth=1
+; LA64-LAMCAS-NEXT: sub.d $a3, $a0, $a1
+; LA64-LAMCAS-NEXT: ext.w.b $a4, $a0
+; LA64-LAMCAS-NEXT: amcas_db.b $a0, $a3, $a2
+; LA64-LAMCAS-NEXT: bne $a0, $a4, .LBB92_1
+; LA64-LAMCAS-NEXT: # %bb.2: # %atomicrmw.end
+; LA64-LAMCAS-NEXT: ret
+ %1 = atomicrmw sub ptr %a, i8 %b seq_cst
+ ret i8 %1
+}
+
+define i16 @atomicrmw_sub_i16_seq_cst(ptr %a, i16 %b) nounwind {
+; LA64-LABEL: atomicrmw_sub_i16_seq_cst:
+; LA64: # %bb.0:
+; LA64-NEXT: slli.d $a2, $a0, 3
+; LA64-NEXT: bstrins.d $a0, $zero, 1, 0
+; LA64-NEXT: lu12i.w $a3, 15
+; LA64-NEXT: ori $a3, $a3, 4095
+; LA64-NEXT: sll.w $a3, $a3, $a2
+; LA64-NEXT: bstrpick.d $a1, $a1, 15, 0
+; LA64-NEXT: sll.w $a1, $a1, $a2
+; LA64-NEXT: .LBB93_1: # =>This Inner Loop Header: Depth=1
+; LA64-NEXT: ll.w $a4, $a0, 0
+; LA64-NEXT: sub.w $a5, $a4, $a1
+; LA64-NEXT: xor $a5, $a4, $a5
+; LA64-NEXT: and $a5, $a5, $a3
+; LA64-NEXT: xor $a5, $a4, $a5
+; LA64-NEXT: sc.w $a5, $a0, 0
+; LA64-NEXT: beqz $a5, .LBB93_1
+; LA64-NEXT: # %bb.2:
+; LA64-NEXT: srl.w $a0, $a4, $a2
+; LA64-NEXT: ret
+;
+; LA64-LAMCAS-LABEL: atomicrmw_sub_i16_seq_cst:
+; LA64-LAMCAS: # %bb.0:
+; LA64-LAMCAS-NEXT: move $a2, $a0
+; LA64-LAMCAS-NEXT: ld.hu $a0, $a0, 0
+; LA64-LAMCAS-NEXT: .p2align 4, , 16
+; LA64-LAMCAS-NEXT: .LBB93_1: # %atomicrmw.start
+; LA64-LAMCAS-NEXT: # =>This Inner Loop Header: Depth=1
+; LA64-LAMCAS-NEXT: sub.d $a3, $a0, $a1
+; LA64-LAMCAS-NEXT: ext.w.h $a4, $a0
+; LA64-LAMCAS-NEXT: amcas_db.h $a0, $a3, $a2
+; LA64-LAMCAS-NEXT: bne $a0, $a4, .LBB93_1
+; LA64-LAMCAS-NEXT: # %bb.2: # %atomicrmw.end
+; LA64-LAMCAS-NEXT: ret
+ %1 = atomicrmw sub ptr %a, i16 %b seq_cst
+ ret i16 %1
+
+}
+
+define i8 @atomicrmw_umax_i8_seq_cst(ptr %a, i8 %b) nounwind {
+; LA64-LABEL: atomicrmw_umax_i8_seq_cst:
+; LA64: # %bb.0:
+; LA64-NEXT: slli.d $a2, $a0, 3
+; LA64-NEXT: bstrins.d $a0, $zero, 1, 0
+; LA64-NEXT: ori $a3, $zero, 255
+; LA64-NEXT: sll.w $a3, $a3, $a2
+; LA64-NEXT: andi $a1, $a1, 255
+; LA64-NEXT: sll.w $a1, $a1, $a2
+; LA64-NEXT: .LBB94_1: # =>This Inner Loop Header: Depth=1
+; LA64-NEXT: ll.w $a4, $a0, 0
+; LA64-NEXT: and $a6, $a4, $a3
+; LA64-NEXT: move $a5, $a4
+; LA64-NEXT: bgeu $a6, $a1, .LBB94_3
+; LA64-NEXT: # %bb.2: # in Loop: Header=BB94_1 Depth=1
+; LA64-NEXT: xor $a5, $a4, $a1
+; LA64-NEXT: and $a5, $a5, $a3
+; LA64-NEXT: xor $a5, $a4, $a5
+; LA64-NEXT: .LBB94_3: # in Loop: Header=BB94_1 Depth=1
+; LA64-NEXT: sc.w $a5, $a0, 0
+; LA64-NEXT: beqz $a5, .LBB94_1
+; LA64-NEXT: # %bb.4:
+; LA64-NEXT: srl.w $a0, $a4, $a2
+; LA64-NEXT: ret
+;
+; LA64-LAMCAS-LABEL: atomicrmw_umax_i8_seq_cst:
+; LA64-LAMCAS: # %bb.0:
+; LA64-LAMCAS-NEXT: move $a2, $a0
+; LA64-LAMCAS-NEXT: ld.bu $a0, $a0, 0
+; LA64-LAMCAS-NEXT: andi $a3, $a1, 255
+; LA64-LAMCAS-NEXT: .p2align 4, , 16
+; LA64-LAMCAS-NEXT: .LBB94_1: # %atomicrmw.start
+; LA64-LAMCAS-NEXT: # =>This Inner Loop Header: Depth=1
+; LA64-LAMCAS-NEXT: andi $a4, $a0, 255
+; LA64-LAMCAS-NEXT: sltu $a4, $a3, $a4
+; LA64-LAMCAS-NEXT: masknez $a5, $a1, $a4
+; LA64-LAMCAS-NEXT: maskeqz $a4, $a0, $a4
+; LA64-LAMCAS-NEXT: or $a4, $a4, $a5
+; LA64-LAMCAS-NEXT: ext.w.b $a5, $a0
+; LA64-LAMCAS-NEXT: amcas_db.b $a0, $a4, $a2
+; LA64-LAMCAS-NEXT: bne $a0, $a5, .LBB94_1
+; LA64-LAMCAS-NEXT: # %bb.2: # %atomicrmw.end
+; LA64-LAMCAS-NEXT: ret
+ %1 = atomicrmw umax ptr %a, i8 %b seq_cst
+ ret i8 %1
+}
+
+define i16 @atomicrmw_umax_i16_seq_cst(ptr %a, i16 %b) nounwind {
+; LA64-LABEL: atomicrmw_umax_i16_seq_cst:
+; LA64: # %bb.0:
+; LA64-NEXT: slli.d $a2, $a0, 3
+; LA64-NEXT: bstrins.d $a0, $zero, 1, 0
+; LA64-NEXT: lu12i.w $a3, 15
+; LA64-NEXT: ori $a3, $a3, 4095
+; LA64-NEXT: sll.w $a3, $a3, $a2
+; LA64-NEXT: bstrpick.d $a1, $a1, 15, 0
+; LA64-NEXT: sll.w $a1, $a1, $a2
+; LA64-NEXT: .LBB95_1: # =>This Inner Loop Header: Depth=1
+; LA64-NEXT: ll.w $a4, $a0, 0
+; LA64-NEXT: and $a6, $a4, $a3
+; LA64-NEXT: move $a5, $a4
+; LA64-NEXT: bgeu $a6, $a1, .LBB95_3
+; LA64-NEXT: # %bb.2: # in Loop: Header=BB95_1 Depth=1
+; LA64-NEXT: xor $a5, $a4, $a1
+; LA64-NEXT: and $a5, $a5, $a3
+; LA64-NEXT: xor $a5, $a4, $a5
+; LA64-NEXT: .LBB95_3: # in Loop: Header=BB95_1 Depth=1
+; LA64-NEXT: sc.w $a5, $a0, 0
+; LA64-NEXT: beqz $a5, .LBB95_1
+; LA64-NEXT: # %bb.4:
+; LA64-NEXT: srl.w $a0, $a4, $a2
+; LA64-NEXT: ret
+;
+; LA64-LAMCAS-LABEL: atomicrmw_umax_i16_seq_cst:
+; LA64-LAMCAS: # %bb.0:
+; LA64-LAMCAS-NEXT: move $a2, $a0
+; LA64-LAMCAS-NEXT: ld.hu $a0, $a0, 0
+; LA64-LAMCAS-NEXT: bstrpick.d $a3, $a1, 15, 0
+; LA64-LAMCAS-NEXT: .p2align 4, , 16
+; LA64-LAMCAS-NEXT: .LBB95_1: # %atomicrmw.start
+; LA64-LAMCAS-NEXT: # =>This Inner Loop Header: Depth=1
+; LA64-LAMCAS-NEXT: bstrpick.d $a4, $a0, 15, 0
+; LA64-LAMCAS-NEXT: sltu $a4, $a3, $a4
+; LA64-LAMCAS-NEXT: masknez $a5, $a1, $a4
+; LA64-LAMCAS-NEXT: maskeqz $a4, $a0, $a4
+; LA64-LAMCAS-NEXT: or $a4, $a4, $a5
+; LA64-LAMCAS-NEXT: ext.w.h $a5, $a0
+; LA64-LAMCAS-NEXT: amcas_db.h $a0, $a4, $a2
+; LA64-LAMCAS-NEXT: bne $a0, $a5, .LBB95_1
+; LA64-LAMCAS-NEXT: # %bb.2: # %atomicrmw.end
+; LA64-LAMCAS-NEXT: ret
+ %1 = atomicrmw umax ptr %a, i16 %b seq_cst
+ ret i16 %1
+}
+
+define i8 @atomicrmw_umin_i8_seq_cst(ptr %a, i8 %b) nounwind {
+; LA64-LABEL: atomicrmw_umin_i8_seq_cst:
+; LA64: # %bb.0:
+; LA64-NEXT: slli.d $a2, $a0, 3
+; LA64-NEXT: bstrins.d $a0, $zero, 1, 0
+; LA64-NEXT: ori $a3, $zero, 255
+; LA64-NEXT: sll.w $a3, $a3, $a2
+; LA64-NEXT: andi $a1, $a1, 255
+; LA64-NEXT: sll.w $a1, $a1, $a2
+; LA64-NEXT: .LBB96_1: # =>This Inner Loop Header: Depth=1
+; LA64-NEXT: ll.w $a4, $a0, 0
+; LA64-NEXT: and $a6, $a4, $a3
+; LA64-NEXT: move $a5, $a4
+; LA64-NEXT: bgeu $a1, $a6, .LBB96_3
+; LA64-NEXT: # %bb.2: # in Loop: Header=BB96_1 Depth=1
+; LA64-NEXT: xor $a5, $a4, $a1
+; LA64-NEXT: and $a5, $a5, $a3
+; LA64-NEXT: xor $a5, $a4, $a5
+; LA64-NEXT: .LBB96_3: # in Loop: Header=BB96_1 Depth=1
+; LA64-NEXT: sc.w $a5, $a0, 0
+; LA64-NEXT: beqz $a5, .LBB96_1
+; LA64-NEXT: # %bb.4:
+; LA64-NEXT: srl.w $a0, $a4, $a2
+; LA64-NEXT: ret
+;
+; LA64-LAMCAS-LABEL: atomicrmw_umin_i8_seq_cst:
+; LA64-LAMCAS: # %bb.0:
+; LA64-LAMCAS-NEXT: move $a2, $a0
+; LA64-LAMCAS-NEXT: ld.bu $a0, $a0, 0
+; LA64-LAMCAS-NEXT: andi $a3, $a1, 255
+; LA64-LAMCAS-NEXT: .p2align 4, , 16
+; LA64-LAMCAS-NEXT: .LBB96_1: # %atomicrmw.start
+; LA64-LAMCAS-NEXT: # =>This Inner Loop Header: Depth=1
+; LA64-LAMCAS-NEXT: andi $a4, $a0, 255
+; LA64-LAMCAS-NEXT: sltu $a4, $a3, $a4
+; LA64-LAMCAS-NEXT: xori $a4, $a4, 1
+; LA64-LAMCAS-NEXT: masknez $a5, $a1, $a4
+; LA64-LAMCAS-NEXT: maskeqz $a4, $a0, $a4
+; LA64-LAMCAS-NEXT: or $a4, $a4, $a5
+; LA64-LAMCAS-NEXT: ext.w.b $a5, $a0
+; LA64-LAMCAS-NEXT: amcas_db.b $a0, $a4, $a2
+; LA64-LAMCAS-NEXT: bne $a0, $a5, .LBB96_1
+; LA64-LAMCAS-NEXT: # %bb.2: # %atomicrmw.end
+; LA64-LAMCAS-NEXT: ret
+ %1 = atomicrmw umin ptr %a, i8 %b seq_cst
+ ret i8 %1
+}
+
+define i16 @atomicrmw_umin_i16_seq_cst(ptr %a, i16 %b) nounwind {
+; LA64-LABEL: atomicrmw_umin_i16_seq_cst:
+; LA64: # %bb.0:
+; LA64-NEXT: slli.d $a2, $a0, 3
+; LA64-NEXT: bstrins.d $a0, $zero, 1, 0
+; LA64-NEXT: lu12i.w $a3, 15
+; LA64-NEXT: ori $a3, $a3, 4095
+; LA64-NEXT: sll.w $a3, $a3, $a2
+; LA64-NEXT: bstrpick.d $a1, $a1, 15, 0
+; LA64-NEXT: sll.w $a1, $a1, $a2
+; LA64-NEXT: .LBB97_1: # =>This Inner Loop Header: Depth=1
+; LA64-NEXT: ll.w $a4, $a0, 0
+; LA64-NEXT: and $a6, $a4, $a3
+; LA64-NEXT: move $a5, $a4
+; LA64-NEXT: bgeu $a1, $a6, .LBB97_3
+; LA64-NEXT: # %bb.2: # in Loop: Header=BB97_1 Depth=1
+; LA64-NEXT: xor $a5, $a4, $a1
+; LA64-NEXT: and $a5, $a5, $a3
+; LA64-NEXT: xor $a5, $a4, $a5
+; LA64-NEXT: .LBB97_3: # in Loop: Header=BB97_1 Depth=1
+; LA64-NEXT: sc.w $a5, $a0, 0
+; LA64-NEXT: beqz $a5, .LBB97_1
+; LA64-NEXT: # %bb.4:
+; LA64-NEXT: srl.w $a0, $a4, $a2
+; LA64-NEXT: ret
+;
+; LA64-LAMCAS-LABEL: atomicrmw_umin_i16_seq_cst:
+; LA64-LAMCAS: # %bb.0:
+; LA64-LAMCAS-NEXT: move $a2, $a0
+; LA64-LAMCAS-NEXT: ld.hu $a0, $a0, 0
+; LA64-LAMCAS-NEXT: bstrpick.d $a3, $a1, 15, 0
+; LA64-LAMCAS-NEXT: .p2align 4, , 16
+; LA64-LAMCAS-NEXT: .LBB97_1: # %atomicrmw.start
+; LA64-LAMCAS-NEXT: # =>This Inner Loop Header: Depth=1
+; LA64-LAMCAS-NEXT: bstrpick.d $a4, $a0, 15, 0
+; LA64-LAMCAS-NEXT: sltu $a4, $a3, $a4
+; LA64-LAMCAS-NEXT: xori $a4, $a4, 1
+; LA64-LAMCAS-NEXT: masknez $a5, $a1, $a4
+; LA64-LAMCAS-NEXT: maskeqz $a4, $a0, $a4
+; LA64-LAMCAS-NEXT: or $a4, $a4, $a5
+; LA64-LAMCAS-NEXT: ext.w.h $a5, $a0
+; LA64-LAMCAS-NEXT: amcas_db.h $a0, $a4, $a2
+; LA64-LAMCAS-NEXT: bne $a0, $a5, .LBB97_1
+; LA64-LAMCAS-NEXT: # %bb.2: # %atomicrmw.end
+; LA64-LAMCAS-NEXT: ret
+ %1 = atomicrmw umin ptr %a, i16 %b seq_cst
+ ret i16 %1
+}
+
+define i8 @atomicrmw_max_i8_seq_cst(ptr %a, i8 %b) nounwind {
+; LA64-LABEL: atomicrmw_max_i8_seq_cst:
+; LA64: # %bb.0:
+; LA64-NEXT: slli.d $a2, $a0, 3
+; LA64-NEXT: bstrins.d $a0, $zero, 1, 0
+; LA64-NEXT: andi $a3, $a2, 24
+; LA64-NEXT: ori $a4, $zero, 255
+; LA64-NEXT: sll.w $a4, $a4, $a2
+; LA64-NEXT: ext.w.b $a1, $a1
+; LA64-NEXT: sll.w $a1, $a1, $a2
+; LA64-NEXT: xori $a3, $a3, 56
+; LA64-NEXT: .LBB98_1: # =>This Inner Loop Header: Depth=1
+; LA64-NEXT: ll.w $a5, $a0, 0
+; LA64-NEXT: and $a7, $a5, $a4
+; LA64-NEXT: move $a6, $a5
+; LA64-NEXT: sll.w $a7, $a7, $a3
+; LA64-NEXT: sra.w $a7, $a7, $a3
+; LA64-NEXT: bge $a7, $a1, .LBB98_3
+; LA64-NEXT: # %bb.2: # in Loop: Header=BB98_1 Depth=1
+; LA64-NEXT: xor $a6, $a5, $a1
+; LA64-NEXT: and $a6, $a6, $a4
+; LA64-NEXT: xor $a6, $a5, $a6
+; LA64-NEXT: .LBB98_3: # in Loop: Header=BB98_1 Depth=1
+; LA64-NEXT: sc.w $a6, $a0, 0
+; LA64-NEXT: beqz $a6, .LBB98_1
+; LA64-NEXT: # %bb.4:
+; LA64-NEXT: srl.w $a0, $a5, $a2
+; LA64-NEXT: ret
+;
+; LA64-LAMCAS-LABEL: atomicrmw_max_i8_seq_cst:
+; LA64-LAMCAS: # %bb.0:
+; LA64-LAMCAS-NEXT: move $a2, $a0
+; LA64-LAMCAS-NEXT: ld.bu $a0, $a0, 0
+; LA64-LAMCAS-NEXT: ext.w.b $a3, $a1
+; LA64-LAMCAS-NEXT: .p2align 4, , 16
+; LA64-LAMCAS-NEXT: .LBB98_1: # %atomicrmw.start
+; LA64-LAMCAS-NEXT: # =>This Inner Loop Header: Depth=1
+; LA64-LAMCAS-NEXT: ext.w.b $a4, $a0
+; LA64-LAMCAS-NEXT: slt $a5, $a3, $a4
+; LA64-LAMCAS-NEXT: masknez $a6, $a1, $a5
+; LA64-LAMCAS-NEXT: maskeqz $a5, $a0, $a5
+; LA64-LAMCAS-NEXT: or $a5, $a5, $a6
+; LA64-LAMCAS-NEXT: amcas_db.b $a0, $a5, $a2
+; LA64-LAMCAS-NEXT: bne $a0, $a4, .LBB98_1
+; LA64-LAMCAS-NEXT: # %bb.2: # %atomicrmw.end
+; LA64-LAMCAS-NEXT: ret
+ %1 = atomicrmw max ptr %a, i8 %b seq_cst
+ ret i8 %1
+}
+
+define i16 @atomicrmw_max_i16_seq_cst(ptr %a, i16 %b) nounwind {
+; LA64-LABEL: atomicrmw_max_i16_seq_cst:
+; LA64: # %bb.0:
+; LA64-NEXT: slli.d $a2, $a0, 3
+; LA64-NEXT: bstrins.d $a0, $zero, 1, 0
+; LA64-NEXT: andi $a3, $a2, 24
+; LA64-NEXT: lu12i.w $a4, 15
+; LA64-NEXT: ori $a4, $a4, 4095
+; LA64-NEXT: sll.w $a4, $a4, $a2
+; LA64-NEXT: ext.w.h $a1, $a1
+; LA64-NEXT: sll.w $a1, $a1, $a2
+; LA64-NEXT: ori $a5, $zero, 48
+; LA64-NEXT: sub.d $a3, $a5, $a3
+; LA64-NEXT: .LBB99_1: # =>This Inner Loop Header: Depth=1
+; LA64-NEXT: ll.w $a5, $a0, 0
+; LA64-NEXT: and $a7, $a5, $a4
+; LA64-NEXT: move $a6, $a5
+; LA64-NEXT: sll.w $a7, $a7, $a3
+; LA64-NEXT: sra.w $a7, $a7, $a3
+; LA64-NEXT: bge $a7, $a1, .LBB99_3
+; LA64-NEXT: # %bb.2: # in Loop: Header=BB99_1 Depth=1
+; LA64-NEXT: xor $a6, $a5, $a1
+; LA64-NEXT: and $a6, $a6, $a4
+; LA64-NEXT: xor $a6, $a5, $a6
+; LA64-NEXT: .LBB99_3: # in Loop: Header=BB99_1 Depth=1
+; LA64-NEXT: sc.w $a6, $a0, 0
+; LA64-NEXT: beqz $a6, .LBB99_1
+; LA64-NEXT: # %bb.4:
+; LA64-NEXT: srl.w $a0, $a5, $a2
+; LA64-NEXT: ret
+;
+; LA64-LAMCAS-LABEL: atomicrmw_max_i16_seq_cst:
+; LA64-LAMCAS: # %bb.0:
+; LA64-LAMCAS-NEXT: move $a2, $a0
+; LA64-LAMCAS-NEXT: ld.hu $a0, $a0, 0
+; LA64-LAMCAS-NEXT: ext.w.h $a3, $a1
+; LA64-LAMCAS-NEXT: .p2align 4, , 16
+; LA64-LAMCAS-NEXT: .LBB99_1: # %atomicrmw.start
+; LA64-LAMCAS-NEXT: # =>This Inner Loop Header: Depth=1
+; LA64-LAMCAS-NEXT: ext.w.h $a4, $a0
+; LA64-LAMCAS-NEXT: slt $a5, $a3, $a4
+; LA64-LAMCAS-NEXT: masknez $a6, $a1, $a5
+; LA64-LAMCAS-NEXT: maskeqz $a5, $a0, $a5
+; LA64-LAMCAS-NEXT: or $a5, $a5, $a6
+; LA64-LAMCAS-NEXT: amcas_db.h $a0, $a5, $a2
+; LA64-LAMCAS-NEXT: bne $a0, $a4, .LBB99_1
+; LA64-LAMCAS-NEXT: # %bb.2: # %atomicrmw.end
+; LA64-LAMCAS-NEXT: ret
+ %1 = atomicrmw max ptr %a, i16 %b seq_cst
+ ret i16 %1
+}
+
+define i8 @atomicrmw_min_i8_seq_cst(ptr %a, i8 %b) nounwind {
+; LA64-LABEL: atomicrmw_min_i8_seq_cst:
+; LA64: # %bb.0:
+; LA64-NEXT: slli.d $a2, $a0, 3
+; LA64-NEXT: bstrins.d $a0, $zero, 1, 0
+; LA64-NEXT: andi $a3, $a2, 24
+; LA64-NEXT: ori $a4, $zero, 255
+; LA64-NEXT: sll.w $a4, $a4, $a2
+; LA64-NEXT: ext.w.b $a1, $a1
+; LA64-NEXT: sll.w $a1, $a1, $a2
+; LA64-NEXT: xori $a3, $a3, 56
+; LA64-NEXT: .LBB100_1: # =>This Inner Loop Header: Depth=1
+; LA64-NEXT: ll.w $a5, $a0, 0
+; LA64-NEXT: and $a7, $a5, $a4
+; LA64-NEXT: move $a6, $a5
+; LA64-NEXT: sll.w $a7, $a7, $a3
+; LA64-NEXT: sra.w $a7, $a7, $a3
+; LA64-NEXT: bge $a1, $a7, .LBB100_3
+; LA64-NEXT: # %bb.2: # in Loop: Header=BB100_1 Depth=1
+; LA64-NEXT: xor $a6, $a5, $a1
+; LA64-NEXT: and $a6, $a6, $a4
+; LA64-NEXT: xor $a6, $a5, $a6
+; LA64-NEXT: .LBB100_3: # in Loop: Header=BB100_1 Depth=1
+; LA64-NEXT: sc.w $a6, $a0, 0
+; LA64-NEXT: beqz $a6, .LBB100_1
+; LA64-NEXT: # %bb.4:
+; LA64-NEXT: srl.w $a0, $a5, $a2
+; LA64-NEXT: ret
+;
+; LA64-LAMCAS-LABEL: atomicrmw_min_i8_seq_cst:
+; LA64-LAMCAS: # %bb.0:
+; LA64-LAMCAS-NEXT: move $a2, $a0
+; LA64-LAMCAS-NEXT: ld.bu $a0, $a0, 0
+; LA64-LAMCAS-NEXT: ext.w.b $a3, $a1
+; LA64-LAMCAS-NEXT: .p2align 4, , 16
+; LA64-LAMCAS-NEXT: .LBB100_1: # %atomicrmw.start
+; LA64-LAMCAS-NEXT: # =>This Inner Loop Header: Depth=1
+; LA64-LAMCAS-NEXT: ext.w.b $a4, $a0
+; LA64-LAMCAS-NEXT: slt $a5, $a3, $a4
+; LA64-LAMCAS-NEXT: xori $a5, $a5, 1
+; LA64-LAMCAS-NEXT: masknez $a6, $a1, $a5
+; LA64-LAMCAS-NEXT: maskeqz $a5, $a0, $a5
+; LA64-LAMCAS-NEXT: or $a5, $a5, $a6
+; LA64-LAMCAS-NEXT: amcas_db.b $a0, $a5, $a2
+; LA64-LAMCAS-NEXT: bne $a0, $a4, .LBB100_1
+; LA64-LAMCAS-NEXT: # %bb.2: # %atomicrmw.end
+; LA64-LAMCAS-NEXT: ret
+ %1 = atomicrmw min ptr %a, i8 %b seq_cst
+ ret i8 %1
+}
+
+define i16 @atomicrmw_min_i16_seq_cst(ptr %a, i16 %b) nounwind {
+; LA64-LABEL: atomicrmw_min_i16_seq_cst:
+; LA64: # %bb.0:
+; LA64-NEXT: slli.d $a2, $a0, 3
+; LA64-NEXT: bstrins.d $a0, $zero, 1, 0
+; LA64-NEXT: andi $a3, $a2, 24
+; LA64-NEXT: lu12i.w $a4, 15
+; LA64-NEXT: ori $a4, $a4, 4095
+; LA64-NEXT: sll.w $a4, $a4, $a2
+; LA64-NEXT: ext.w.h $a1, $a1
+; LA64-NEXT: sll.w $a1, $a1, $a2
+; LA64-NEXT: ori $a5, $zero, 48
+; LA64-NEXT: sub.d $a3, $a5, $a3
+; LA64-NEXT: .LBB101_1: # =>This Inner Loop Header: Depth=1
+; LA64-NEXT: ll.w $a5, $a0, 0
+; LA64-NEXT: and $a7, $a5, $a4
+; LA64-NEXT: move $a6, $a5
+; LA64-NEXT: sll.w $a7, $a7, $a3
+; LA64-NEXT: sra.w $a7, $a7, $a3
+; LA64-NEXT: bge $a1, $a7, .LBB101_3
+; LA64-NEXT: # %bb.2: # in Loop: Header=BB101_1 Depth=1
+; LA64-NEXT: xor $a6, $a5, $a1
+; LA64-NEXT: and $a6, $a6, $a4
+; LA64-NEXT: xor $a6, $a5, $a6
+; LA64-NEXT: .LBB101_3: # in Loop: Header=BB101_1 Depth=1
+; LA64-NEXT: sc.w $a6, $a0, 0
+; LA64-NEXT: beqz $a6, .LBB101_1
+; LA64-NEXT: # %bb.4:
+; LA64-NEXT: srl.w $a0, $a5, $a2
+; LA64-NEXT: ret
+;
+; LA64-LAMCAS-LABEL: atomicrmw_min_i16_seq_cst:
+; LA64-LAMCAS: # %bb.0:
+; LA64-LAMCAS-NEXT: move $a2, $a0
+; LA64-LAMCAS-NEXT: ld.hu $a0, $a0, 0
+; LA64-LAMCAS-NEXT: ext.w.h $a3, $a1
+; LA64-LAMCAS-NEXT: .p2align 4, , 16
+; LA64-LAMCAS-NEXT: .LBB101_1: # %atomicrmw.start
+; LA64-LAMCAS-NEXT: # =>This Inner Loop Header: Depth=1
+; LA64-LAMCAS-NEXT: ext.w.h $a4, $a0
+; LA64-LAMCAS-NEXT: slt $a5, $a3, $a4
+; LA64-LAMCAS-NEXT: xori $a5, $a5, 1
+; LA64-LAMCAS-NEXT: masknez $a6, $a1, $a5
+; LA64-LAMCAS-NEXT: maskeqz $a5, $a0, $a5
+; LA64-LAMCAS-NEXT: or $a5, $a5, $a6
+; LA64-LAMCAS-NEXT: amcas_db.h $a0, $a5, $a2
+; LA64-LAMCAS-NEXT: bne $a0, $a4, .LBB101_1
+; LA64-LAMCAS-NEXT: # %bb.2: # %atomicrmw.end
+; LA64-LAMCAS-NEXT: ret
+ %1 = atomicrmw min ptr %a, i16 %b seq_cst
+ ret i16 %1
+}
+
+define i8 @atomicrmw_nand_i8_seq_cst(ptr %a, i8 %b) nounwind {
+; LA64-LABEL: atomicrmw_nand_i8_seq_cst:
+; LA64: # %bb.0:
+; LA64-NEXT: slli.d $a2, $a0, 3
+; LA64-NEXT: bstrins.d $a0, $zero, 1, 0
+; LA64-NEXT: ori $a3, $zero, 255
+; LA64-NEXT: sll.w $a3, $a3, $a2
+; LA64-NEXT: andi $a1, $a1, 255
+; LA64-NEXT: sll.w $a1, $a1, $a2
+; LA64-NEXT: .LBB102_1: # =>This Inner Loop Header: Depth=1
+; LA64-NEXT: ll.w $a4, $a0, 0
+; LA64-NEXT: and $a5, $a4, $a1
+; LA64-NEXT: nor $a5, $a5, $zero
+; LA64-NEXT: xor $a5, $a4, $a5
+; LA64-NEXT: and $a5, $a5, $a3
+; LA64-NEXT: xor $a5, $a4, $a5
+; LA64-NEXT: sc.w $a5, $a0, 0
+; LA64-NEXT: beqz $a5, .LBB102_1
+; LA64-NEXT: # %bb.2:
+; LA64-NEXT: srl.w $a0, $a4, $a2
+; LA64-NEXT: ret
+;
+; LA64-LAMCAS-LABEL: atomicrmw_nand_i8_seq_cst:
+; LA64-LAMCAS: # %bb.0:
+; LA64-LAMCAS-NEXT: move $a2, $a0
+; LA64-LAMCAS-NEXT: ld.bu $a0, $a0, 0
+; LA64-LAMCAS-NEXT: .p2align 4, , 16
+; LA64-LAMCAS-NEXT: .LBB102_1: # %atomicrmw.start
+; LA64-LAMCAS-NEXT: # =>This Inner Loop Header: Depth=1
+; LA64-LAMCAS-NEXT: and $a3, $a0, $a1
+; LA64-LAMCAS-NEXT: nor $a3, $a3, $zero
+; LA64-LAMCAS-NEXT: ext.w.b $a4, $a0
+; LA64-LAMCAS-NEXT: amcas_db.b $a0, $a3, $a2
+; LA64-LAMCAS-NEXT: bne $a0, $a4, .LBB102_1
+; LA64-LAMCAS-NEXT: # %bb.2: # %atomicrmw.end
+; LA64-LAMCAS-NEXT: ret
+ %1 = atomicrmw nand ptr %a, i8 %b seq_cst
+ ret i8 %1
+}
+
+define i16 @atomicrmw_nand_i16_seq_cst(ptr %a, i16 %b) nounwind {
+; LA64-LABEL: atomicrmw_nand_i16_seq_cst:
+; LA64: # %bb.0:
+; LA64-NEXT: slli.d $a2, $a0, 3
+; LA64-NEXT: bstrins.d $a0, $zero, 1, 0
+; LA64-NEXT: lu12i.w $a3, 15
+; LA64-NEXT: ori $a3, $a3, 4095
+; LA64-NEXT: sll.w $a3, $a3, $a2
+; LA64-NEXT: bstrpick.d $a1, $a1, 15, 0
+; LA64-NEXT: sll.w $a1, $a1, $a2
+; LA64-NEXT: .LBB103_1: # =>This Inner Loop Header: Depth=1
+; LA64-NEXT: ll.w $a4, $a0, 0
+; LA64-NEXT: and $a5, $a4, $a1
+; LA64-NEXT: nor $a5, $a5, $zero
+; LA64-NEXT: xor $a5, $a4, $a5
+; LA64-NEXT: and $a5, $a5, $a3
+; LA64-NEXT: xor $a5, $a4, $a5
+; LA64-NEXT: sc.w $a5, $a0, 0
+; LA64-NEXT: beqz $a5, .LBB103_1
+; LA64-NEXT: # %bb.2:
+; LA64-NEXT: srl.w $a0, $a4, $a2
+; LA64-NEXT: ret
+;
+; LA64-LAMCAS-LABEL: atomicrmw_nand_i16_seq_cst:
+; LA64-LAMCAS: # %bb.0:
+; LA64-LAMCAS-NEXT: move $a2, $a0
+; LA64-LAMCAS-NEXT: ld.hu $a0, $a0, 0
+; LA64-LAMCAS-NEXT: .p2align 4, , 16
+; LA64-LAMCAS-NEXT: .LBB103_1: # %atomicrmw.start
+; LA64-LAMCAS-NEXT: # =>This Inner Loop Header: Depth=1
+; LA64-LAMCAS-NEXT: and $a3, $a0, $a1
+; LA64-LAMCAS-NEXT: nor $a3, $a3, $zero
+; LA64-LAMCAS-NEXT: ext.w.h $a4, $a0
+; LA64-LAMCAS-NEXT: amcas_db.h $a0, $a3, $a2
+; LA64-LAMCAS-NEXT: bne $a0, $a4, .LBB103_1
+; LA64-LAMCAS-NEXT: # %bb.2: # %atomicrmw.end
+; LA64-LAMCAS-NEXT: ret
+ %1 = atomicrmw nand ptr %a, i16 %b seq_cst
+ ret i16 %1
+
+}
+
+define i32 @atomicrmw_nand_i32_seq_cst(ptr %a, i32 %b) nounwind {
+; LA64-LABEL: atomicrmw_nand_i32_seq_cst:
+; LA64: # %bb.0:
+; LA64-NEXT: .LBB104_1: # =>This Inner Loop Header: Depth=1
+; LA64-NEXT: ll.w $a2, $a0, 0
+; LA64-NEXT: and $a3, $a2, $a1
+; LA64-NEXT: nor $a3, $a3, $zero
+; LA64-NEXT: sc.w $a3, $a0, 0
+; LA64-NEXT: beqz $a3, .LBB104_1
+; LA64-NEXT: # %bb.2:
+; LA64-NEXT: move $a0, $a2
+; LA64-NEXT: ret
+;
+; LA64-LAMCAS-LABEL: atomicrmw_nand_i32_seq_cst:
+; LA64-LAMCAS: # %bb.0:
+; LA64-LAMCAS-NEXT: move $a2, $a0
+; LA64-LAMCAS-NEXT: ld.w $a0, $a0, 0
+; LA64-LAMCAS-NEXT: .p2align 4, , 16
+; LA64-LAMCAS-NEXT: .LBB104_1: # %atomicrmw.start
+; LA64-LAMCAS-NEXT: # =>This Inner Loop Header: Depth=1
+; LA64-LAMCAS-NEXT: move $a3, $a0
+; LA64-LAMCAS-NEXT: and $a4, $a0, $a1
+; LA64-LAMCAS-NEXT: nor $a4, $a4, $zero
+; LA64-LAMCAS-NEXT: amcas_db.w $a0, $a4, $a2
+; LA64-LAMCAS-NEXT: bne $a0, $a3, .LBB104_1
+; LA64-LAMCAS-NEXT: # %bb.2: # %atomicrmw.end
+; LA64-LAMCAS-NEXT: ret
+ %1 = atomicrmw nand ptr %a, i32 %b seq_cst
+ ret i32 %1
+}
+
+define i64 @atomicrmw_nand_i64_seq_cst(ptr %a, i64 %b) nounwind {
+; LA64-LABEL: atomicrmw_nand_i64_seq_cst:
+; LA64: # %bb.0:
+; LA64-NEXT: .LBB105_1: # =>This Inner Loop Header: Depth=1
+; LA64-NEXT: ll.d $a2, $a0, 0
+; LA64-NEXT: and $a3, $a2, $a1
+; LA64-NEXT: nor $a3, $a3, $zero
+; LA64-NEXT: sc.d $a3, $a0, 0
+; LA64-NEXT: beqz $a3, .LBB105_1
+; LA64-NEXT: # %bb.2:
+; LA64-NEXT: move $a0, $a2
+; LA64-NEXT: ret
+;
+; LA64-LAMCAS-LABEL: atomicrmw_nand_i64_seq_cst:
+; LA64-LAMCAS: # %bb.0:
+; LA64-LAMCAS-NEXT: move $a2, $a0
+; LA64-LAMCAS-NEXT: ld.d $a0, $a0, 0
+; LA64-LAMCAS-NEXT: .p2align 4, , 16
+; LA64-LAMCAS-NEXT: .LBB105_1: # %atomicrmw.start
+; LA64-LAMCAS-NEXT: # =>This Inner Loop Header: Depth=1
+; LA64-LAMCAS-NEXT: move $a3, $a0
+; LA64-LAMCAS-NEXT: and $a4, $a0, $a1
+; LA64-LAMCAS-NEXT: nor $a4, $a4, $zero
+; LA64-LAMCAS-NEXT: amcas_db.d $a0, $a4, $a2
+; LA64-LAMCAS-NEXT: bne $a0, $a3, .LBB105_1
+; LA64-LAMCAS-NEXT: # %bb.2: # %atomicrmw.end
+; LA64-LAMCAS-NEXT: ret
+ %1 = atomicrmw nand ptr %a, i64 %b seq_cst
+ ret i64 %1
+}
+
+
+
+define i8 @atomicrmw_and_i8_seq_cst(ptr %a, i8 %b) nounwind {
+; LA64-LABEL: atomicrmw_and_i8_seq_cst:
+; LA64: # %bb.0:
+; LA64-NEXT: slli.d $a2, $a0, 3
+; LA64-NEXT: bstrins.d $a0, $zero, 1, 0
+; LA64-NEXT: ori $a3, $zero, 255
+; LA64-NEXT: sll.w $a3, $a3, $a2
+; LA64-NEXT: andi $a1, $a1, 255
+; LA64-NEXT: sll.w $a1, $a1, $a2
+; LA64-NEXT: orn $a1, $a1, $a3
+; LA64-NEXT: amand_db.w $a3, $a1, $a0
+; LA64-NEXT: srl.w $a0, $a3, $a2
+; LA64-NEXT: ret
+;
+; LA64-LAMCAS-LABEL: atomicrmw_and_i8_seq_cst:
+; LA64-LAMCAS: # %bb.0:
+; LA64-LAMCAS-NEXT: move $a2, $a0
+; LA64-LAMCAS-NEXT: ld.bu $a0, $a0, 0
+; LA64-LAMCAS-NEXT: .p2align 4, , 16
+; LA64-LAMCAS-NEXT: .LBB106_1: # %atomicrmw.start
+; LA64-LAMCAS-NEXT: # =>This Inner Loop Header: Depth=1
+; LA64-LAMCAS-NEXT: and $a3, $a0, $a1
+; LA64-LAMCAS-NEXT: ext.w.b $a4, $a0
+; LA64-LAMCAS-NEXT: amcas_db.b $a0, $a3, $a2
+; LA64-LAMCAS-NEXT: bne $a0, $a4, .LBB106_1
+; LA64-LAMCAS-NEXT: # %bb.2: # %atomicrmw.end
+; LA64-LAMCAS-NEXT: ret
+ %1 = atomicrmw and ptr %a, i8 %b seq_cst
+ ret i8 %1
+}
+
+define i16 @atomicrmw_and_i16_seq_cst(ptr %a, i16 %b) nounwind {
+; LA64-LABEL: atomicrmw_and_i16_seq_cst:
+; LA64: # %bb.0:
+; LA64-NEXT: slli.d $a2, $a0, 3
+; LA64-NEXT: bstrins.d $a0, $zero, 1, 0
+; LA64-NEXT: lu12i.w $a3, 15
+; LA64-NEXT: ori $a3, $a3, 4095
+; LA64-NEXT: sll.w $a3, $a3, $a2
+; LA64-NEXT: bstrpick.d $a1, $a1, 15, 0
+; LA64-NEXT: sll.w $a1, $a1, $a2
+; LA64-NEXT: orn $a1, $a1, $a3
+; LA64-NEXT: amand_db.w $a3, $a1, $a0
+; LA64-NEXT: srl.w $a0, $a3, $a2
+; LA64-NEXT: ret
+;
+; LA64-LAMCAS-LABEL: atomicrmw_and_i16_seq_cst:
+; LA64-LAMCAS: # %bb.0:
+; LA64-LAMCAS-NEXT: move $a2, $a0
+; LA64-LAMCAS-NEXT: ld.hu $a0, $a0, 0
+; LA64-LAMCAS-NEXT: .p2align 4, , 16
+; LA64-LAMCAS-NEXT: .LBB107_1: # %atomicrmw.start
+; LA64-LAMCAS-NEXT: # =>This Inner Loop Header: Depth=1
+; LA64-LAMCAS-NEXT: and $a3, $a0, $a1
+; LA64-LAMCAS-NEXT: ext.w.h $a4, $a0
+; LA64-LAMCAS-NEXT: amcas_db.h $a0, $a3, $a2
+; LA64-LAMCAS-NEXT: bne $a0, $a4, .LBB107_1
+; LA64-LAMCAS-NEXT: # %bb.2: # %atomicrmw.end
+; LA64-LAMCAS-NEXT: ret
+ %1 = atomicrmw and ptr %a, i16 %b seq_cst
+ ret i16 %1
+
+}
+
+define i8 @atomicrmw_or_i8_seq_cst(ptr %a, i8 %b) nounwind {
+; LA64-LABEL: atomicrmw_or_i8_seq_cst:
+; LA64: # %bb.0:
+; LA64-NEXT: slli.d $a2, $a0, 3
+; LA64-NEXT: bstrins.d $a0, $zero, 1, 0
+; LA64-NEXT: andi $a1, $a1, 255
+; LA64-NEXT: sll.w $a1, $a1, $a2
+; LA64-NEXT: amor_db.w $a3, $a1, $a0
+; LA64-NEXT: srl.w $a0, $a3, $a2
+; LA64-NEXT: ret
+;
+; LA64-LAMCAS-LABEL: atomicrmw_or_i8_seq_cst:
+; LA64-LAMCAS: # %bb.0:
+; LA64-LAMCAS-NEXT: move $a2, $a0
+; LA64-LAMCAS-NEXT: ld.bu $a0, $a0, 0
+; LA64-LAMCAS-NEXT: .p2align 4, , 16
+; LA64-LAMCAS-NEXT: .LBB108_1: # %atomicrmw.start
+; LA64-LAMCAS-NEXT: # =>This Inner Loop Header: Depth=1
+; LA64-LAMCAS-NEXT: or $a3, $a0, $a1
+; LA64-LAMCAS-NEXT: ext.w.b $a4, $a0
+; LA64-LAMCAS-NEXT: amcas_db.b $a0, $a3, $a2
+; LA64-LAMCAS-NEXT: bne $a0, $a4, .LBB108_1
+; LA64-LAMCAS-NEXT: # %bb.2: # %atomicrmw.end
+; LA64-LAMCAS-NEXT: ret
+ %1 = atomicrmw or ptr %a, i8 %b seq_cst
+ ret i8 %1
+}
+
+define i16 @atomicrmw_or_i16_seq_cst(ptr %a, i16 %b) nounwind {
+; LA64-LABEL: atomicrmw_or_i16_seq_cst:
+; LA64: # %bb.0:
+; LA64-NEXT: slli.d $a2, $a0, 3
+; LA64-NEXT: bstrins.d $a0, $zero, 1, 0
+; LA64-NEXT: bstrpick.d $a1, $a1, 15, 0
+; LA64-NEXT: sll.w $a1, $a1, $a2
+; LA64-NEXT: amor_db.w $a3, $a1, $a0
+; LA64-NEXT: srl.w $a0, $a3, $a2
+; LA64-NEXT: ret
+;
+; LA64-LAMCAS-LABEL: atomicrmw_or_i16_seq_cst:
+; LA64-LAMCAS: # %bb.0:
+; LA64-LAMCAS-NEXT: move $a2, $a0
+; LA64-LAMCAS-NEXT: ld.hu $a0, $a0, 0
+; LA64-LAMCAS-NEXT: .p2align 4, , 16
+; LA64-LAMCAS-NEXT: .LBB109_1: # %atomicrmw.start
+; LA64-LAMCAS-NEXT: # =>This Inner Loop Header: Depth=1
+; LA64-LAMCAS-NEXT: or $a3, $a0, $a1
+; LA64-LAMCAS-NEXT: ext.w.h $a4, $a0
+; LA64-LAMCAS-NEXT: amcas_db.h $a0, $a3, $a2
+; LA64-LAMCAS-NEXT: bne $a0, $a4, .LBB109_1
+; LA64-LAMCAS-NEXT: # %bb.2: # %atomicrmw.end
+; LA64-LAMCAS-NEXT: ret
+ %1 = atomicrmw or ptr %a, i16 %b seq_cst
+ ret i16 %1
+
+}
+
+define i8 @atomicrmw_xor_i8_seq_cst(ptr %a, i8 %b) nounwind {
+; LA64-LABEL: atomicrmw_xor_i8_seq_cst:
+; LA64: # %bb.0:
+; LA64-NEXT: slli.d $a2, $a0, 3
+; LA64-NEXT: bstrins.d $a0, $zero, 1, 0
+; LA64-NEXT: andi $a1, $a1, 255
+; LA64-NEXT: sll.w $a1, $a1, $a2
+; LA64-NEXT: amxor_db.w $a3, $a1, $a0
+; LA64-NEXT: srl.w $a0, $a3, $a2
+; LA64-NEXT: ret
+;
+; LA64-LAMCAS-LABEL: atomicrmw_xor_i8_seq_cst:
+; LA64-LAMCAS: # %bb.0:
+; LA64-LAMCAS-NEXT: move $a2, $a0
+; LA64-LAMCAS-NEXT: ld.bu $a0, $a0, 0
+; LA64-LAMCAS-NEXT: .p2align 4, , 16
+; LA64-LAMCAS-NEXT: .LBB110_1: # %atomicrmw.start
+; LA64-LAMCAS-NEXT: # =>This Inner Loop Header: Depth=1
+; LA64-LAMCAS-NEXT: xor $a3, $a0, $a1
+; LA64-LAMCAS-NEXT: ext.w.b $a4, $a0
+; LA64-LAMCAS-NEXT: amcas_db.b $a0, $a3, $a2
+; LA64-LAMCAS-NEXT: bne $a0, $a4, .LBB110_1
+; LA64-LAMCAS-NEXT: # %bb.2: # %atomicrmw.end
+; LA64-LAMCAS-NEXT: ret
+ %1 = atomicrmw xor ptr %a, i8 %b seq_cst
+ ret i8 %1
+}
+
+define i16 @atomicrmw_xor_i16_seq_cst(ptr %a, i16 %b) nounwind {
+; LA64-LABEL: atomicrmw_xor_i16_seq_cst:
+; LA64: # %bb.0:
+; LA64-NEXT: slli.d $a2, $a0, 3
+; LA64-NEXT: bstrins.d $a0, $zero, 1, 0
+; LA64-NEXT: bstrpick.d $a1, $a1, 15, 0
+; LA64-NEXT: sll.w $a1, $a1, $a2
+; LA64-NEXT: amxor_db.w $a3, $a1, $a0
+; LA64-NEXT: srl.w $a0, $a3, $a2
+; LA64-NEXT: ret
+;
+; LA64-LAMCAS-LABEL: atomicrmw_xor_i16_seq_cst:
+; LA64-LAMCAS: # %bb.0:
+; LA64-LAMCAS-NEXT: move $a2, $a0
+; LA64-LAMCAS-NEXT: ld.hu $a0, $a0, 0
+; LA64-LAMCAS-NEXT: .p2align 4, , 16
+; LA64-LAMCAS-NEXT: .LBB111_1: # %atomicrmw.start
+; LA64-LAMCAS-NEXT: # =>This Inner Loop Header: Depth=1
+; LA64-LAMCAS-NEXT: xor $a3, $a0, $a1
+; LA64-LAMCAS-NEXT: ext.w.h $a4, $a0
+; LA64-LAMCAS-NEXT: amcas_db.h $a0, $a3, $a2
+; LA64-LAMCAS-NEXT: bne $a0, $a4, .LBB111_1
+; LA64-LAMCAS-NEXT: # %bb.2: # %atomicrmw.end
+; LA64-LAMCAS-NEXT: ret
+ %1 = atomicrmw xor ptr %a, i16 %b seq_cst
+ ret i16 %1
+
+}
+
+define i8 @atomicrmw_xchg_i8_monotonic(ptr %a, i8 %b) nounwind {
+; LA64-LABEL: atomicrmw_xchg_i8_monotonic:
+; LA64: # %bb.0:
+; LA64-NEXT: slli.d $a2, $a0, 3
+; LA64-NEXT: bstrins.d $a0, $zero, 1, 0
+; LA64-NEXT: ori $a3, $zero, 255
+; LA64-NEXT: sll.w $a3, $a3, $a2
+; LA64-NEXT: andi $a1, $a1, 255
+; LA64-NEXT: sll.w $a1, $a1, $a2
+; LA64-NEXT: .LBB112_1: # =>This Inner Loop Header: Depth=1
+; LA64-NEXT: ll.w $a4, $a0, 0
+; LA64-NEXT: addi.w $a5, $a1, 0
+; LA64-NEXT: xor $a5, $a4, $a5
+; LA64-NEXT: and $a5, $a5, $a3
+; LA64-NEXT: xor $a5, $a4, $a5
+; LA64-NEXT: sc.w $a5, $a0, 0
+; LA64-NEXT: beqz $a5, .LBB112_1
+; LA64-NEXT: # %bb.2:
+; LA64-NEXT: srl.w $a0, $a4, $a2
+; LA64-NEXT: ret
+;
+; LA64-LAMCAS-LABEL: atomicrmw_xchg_i8_monotonic:
+; LA64-LAMCAS: # %bb.0:
+; LA64-LAMCAS-NEXT: ld.bu $a2, $a0, 0
+; LA64-LAMCAS-NEXT: .p2align 4, , 16
+; LA64-LAMCAS-NEXT: .LBB112_1: # %atomicrmw.start
+; LA64-LAMCAS-NEXT: # =>This Inner Loop Header: Depth=1
+; LA64-LAMCAS-NEXT: ext.w.b $a3, $a2
+; LA64-LAMCAS-NEXT: amcas.b $a2, $a1, $a0
+; LA64-LAMCAS-NEXT: bne $a2, $a3, .LBB112_1
+; LA64-LAMCAS-NEXT: # %bb.2: # %atomicrmw.end
+; LA64-LAMCAS-NEXT: move $a0, $a2
+; LA64-LAMCAS-NEXT: ret
+ %1 = atomicrmw xchg ptr %a, i8 %b monotonic
+ ret i8 %1
+}
+
+define i8 @atomicrmw_xchg_0_i8_monotonic(ptr %a) nounwind {
+; LA64-LABEL: atomicrmw_xchg_0_i8_monotonic:
+; LA64: # %bb.0:
+; LA64-NEXT: slli.d $a1, $a0, 3
+; LA64-NEXT: bstrins.d $a0, $zero, 1, 0
+; LA64-NEXT: ori $a2, $zero, 255
+; LA64-NEXT: sll.w $a2, $a2, $a1
+; LA64-NEXT: nor $a2, $a2, $zero
+; LA64-NEXT: amand.w $a3, $a2, $a0
+; LA64-NEXT: srl.w $a0, $a3, $a1
+; LA64-NEXT: ret
+;
+; LA64-LAMCAS-LABEL: atomicrmw_xchg_0_i8_monotonic:
+; LA64-LAMCAS: # %bb.0:
+; LA64-LAMCAS-NEXT: move $a1, $a0
+; LA64-LAMCAS-NEXT: ld.bu $a0, $a0, 0
+; LA64-LAMCAS-NEXT: .p2align 4, , 16
+; LA64-LAMCAS-NEXT: .LBB113_1: # %atomicrmw.start
+; LA64-LAMCAS-NEXT: # =>This Inner Loop Header: Depth=1
+; LA64-LAMCAS-NEXT: ext.w.b $a2, $a0
+; LA64-LAMCAS-NEXT: amcas.b $a0, $zero, $a1
+; LA64-LAMCAS-NEXT: bne $a0, $a2, .LBB113_1
+; LA64-LAMCAS-NEXT: # %bb.2: # %atomicrmw.end
+; LA64-LAMCAS-NEXT: ret
+ %1 = atomicrmw xchg ptr %a, i8 0 monotonic
+ ret i8 %1
+}
+
+define i8 @atomicrmw_xchg_minus_1_i8_monotonic(ptr %a) nounwind {
+; LA64-LABEL: atomicrmw_xchg_minus_1_i8_monotonic:
+; LA64: # %bb.0:
+; LA64-NEXT: slli.d $a1, $a0, 3
+; LA64-NEXT: bstrins.d $a0, $zero, 1, 0
+; LA64-NEXT: ori $a2, $zero, 255
+; LA64-NEXT: sll.w $a2, $a2, $a1
+; LA64-NEXT: amor.w $a3, $a2, $a0
+; LA64-NEXT: srl.w $a0, $a3, $a1
+; LA64-NEXT: ret
+;
+; LA64-LAMCAS-LABEL: atomicrmw_xchg_minus_1_i8_monotonic:
+; LA64-LAMCAS: # %bb.0:
+; LA64-LAMCAS-NEXT: move $a1, $a0
+; LA64-LAMCAS-NEXT: ld.bu $a0, $a0, 0
+; LA64-LAMCAS-NEXT: addi.w $a2, $zero, -1
+; LA64-LAMCAS-NEXT: .p2align 4, , 16
+; LA64-LAMCAS-NEXT: .LBB114_1: # %atomicrmw.start
+; LA64-LAMCAS-NEXT: # =>This Inner Loop Header: Depth=1
+; LA64-LAMCAS-NEXT: ext.w.b $a3, $a0
+; LA64-LAMCAS-NEXT: amcas.b $a0, $a2, $a1
+; LA64-LAMCAS-NEXT: bne $a0, $a3, .LBB114_1
+; LA64-LAMCAS-NEXT: # %bb.2: # %atomicrmw.end
+; LA64-LAMCAS-NEXT: ret
+ %1 = atomicrmw xchg ptr %a, i8 -1 monotonic
+ ret i8 %1
+}
+
+define i16 @atomicrmw_xchg_i16_monotonic(ptr %a, i16 %b) nounwind {
+; LA64-LABEL: atomicrmw_xchg_i16_monotonic:
+; LA64: # %bb.0:
+; LA64-NEXT: slli.d $a2, $a0, 3
+; LA64-NEXT: bstrins.d $a0, $zero, 1, 0
+; LA64-NEXT: lu12i.w $a3, 15
+; LA64-NEXT: ori $a3, $a3, 4095
+; LA64-NEXT: sll.w $a3, $a3, $a2
+; LA64-NEXT: bstrpick.d $a1, $a1, 15, 0
+; LA64-NEXT: sll.w $a1, $a1, $a2
+; LA64-NEXT: .LBB115_1: # =>This Inner Loop Header: Depth=1
+; LA64-NEXT: ll.w $a4, $a0, 0
+; LA64-NEXT: addi.w $a5, $a1, 0
+; LA64-NEXT: xor $a5, $a4, $a5
+; LA64-NEXT: and $a5, $a5, $a3
+; LA64-NEXT: xor $a5, $a4, $a5
+; LA64-NEXT: sc.w $a5, $a0, 0
+; LA64-NEXT: beqz $a5, .LBB115_1
+; LA64-NEXT: # %bb.2:
+; LA64-NEXT: srl.w $a0, $a4, $a2
+; LA64-NEXT: ret
+;
+; LA64-LAMCAS-LABEL: atomicrmw_xchg_i16_monotonic:
+; LA64-LAMCAS: # %bb.0:
+; LA64-LAMCAS-NEXT: ld.hu $a2, $a0, 0
+; LA64-LAMCAS-NEXT: .p2align 4, , 16
+; LA64-LAMCAS-NEXT: .LBB115_1: # %atomicrmw.start
+; LA64-LAMCAS-NEXT: # =>This Inner Loop Header: Depth=1
+; LA64-LAMCAS-NEXT: ext.w.h $a3, $a2
+; LA64-LAMCAS-NEXT: amcas.h $a2, $a1, $a0
+; LA64-LAMCAS-NEXT: bne $a2, $a3, .LBB115_1
+; LA64-LAMCAS-NEXT: # %bb.2: # %atomicrmw.end
+; LA64-LAMCAS-NEXT: move $a0, $a2
+; LA64-LAMCAS-NEXT: ret
+ %1 = atomicrmw xchg ptr %a, i16 %b monotonic
+ ret i16 %1
+}
+
+define i16 @atomicrmw_xchg_0_i16_monotonic(ptr %a) nounwind {
+; LA64-LABEL: atomicrmw_xchg_0_i16_monotonic:
+; LA64: # %bb.0:
+; LA64-NEXT: slli.d $a1, $a0, 3
+; LA64-NEXT: bstrins.d $a0, $zero, 1, 0
+; LA64-NEXT: lu12i.w $a2, 15
+; LA64-NEXT: ori $a2, $a2, 4095
+; LA64-NEXT: sll.w $a2, $a2, $a1
+; LA64-NEXT: nor $a2, $a2, $zero
+; LA64-NEXT: amand.w $a3, $a2, $a0
+; LA64-NEXT: srl.w $a0, $a3, $a1
+; LA64-NEXT: ret
+;
+; LA64-LAMCAS-LABEL: atomicrmw_xchg_0_i16_monotonic:
+; LA64-LAMCAS: # %bb.0:
+; LA64-LAMCAS-NEXT: move $a1, $a0
+; LA64-LAMCAS-NEXT: ld.hu $a0, $a0, 0
+; LA64-LAMCAS-NEXT: .p2align 4, , 16
+; LA64-LAMCAS-NEXT: .LBB116_1: # %atomicrmw.start
+; LA64-LAMCAS-NEXT: # =>This Inner Loop Header: Depth=1
+; LA64-LAMCAS-NEXT: ext.w.h $a2, $a0
+; LA64-LAMCAS-NEXT: amcas.h $a0, $zero, $a1
+; LA64-LAMCAS-NEXT: bne $a0, $a2, .LBB116_1
+; LA64-LAMCAS-NEXT: # %bb.2: # %atomicrmw.end
+; LA64-LAMCAS-NEXT: ret
+ %1 = atomicrmw xchg ptr %a, i16 0 monotonic
+ ret i16 %1
+}
+
+define i16 @atomicrmw_xchg_minus_1_i16_monotonic(ptr %a) nounwind {
+; LA64-LABEL: atomicrmw_xchg_minus_1_i16_monotonic:
+; LA64: # %bb.0:
+; LA64-NEXT: slli.d $a1, $a0, 3
+; LA64-NEXT: bstrins.d $a0, $zero, 1, 0
+; LA64-NEXT: lu12i.w $a2, 15
+; LA64-NEXT: ori $a2, $a2, 4095
+; LA64-NEXT: sll.w $a2, $a2, $a1
+; LA64-NEXT: amor.w $a3, $a2, $a0
+; LA64-NEXT: srl.w $a0, $a3, $a1
+; LA64-NEXT: ret
+;
+; LA64-LAMCAS-LABEL: atomicrmw_xchg_minus_1_i16_monotonic:
+; LA64-LAMCAS: # %bb.0:
+; LA64-LAMCAS-NEXT: move $a1, $a0
+; LA64-LAMCAS-NEXT: ld.hu $a0, $a0, 0
+; LA64-LAMCAS-NEXT: addi.w $a2, $zero, -1
+; LA64-LAMCAS-NEXT: .p2align 4, , 16
+; LA64-LAMCAS-NEXT: .LBB117_1: # %atomicrmw.start
+; LA64-LAMCAS-NEXT: # =>This Inner Loop Header: Depth=1
+; LA64-LAMCAS-NEXT: ext.w.h $a3, $a0
+; LA64-LAMCAS-NEXT: amcas.h $a0, $a2, $a1
+; LA64-LAMCAS-NEXT: bne $a0, $a3, .LBB117_1
+; LA64-LAMCAS-NEXT: # %bb.2: # %atomicrmw.end
+; LA64-LAMCAS-NEXT: ret
+ %1 = atomicrmw xchg ptr %a, i16 -1 monotonic
+ ret i16 %1
+
+}
+
+define i8 @atomicrmw_add_i8_monotonic(ptr %a, i8 %b) nounwind {
+; LA64-LABEL: atomicrmw_add_i8_monotonic:
+; LA64: # %bb.0:
+; LA64-NEXT: slli.d $a2, $a0, 3
+; LA64-NEXT: bstrins.d $a0, $zero, 1, 0
+; LA64-NEXT: ori $a3, $zero, 255
+; LA64-NEXT: sll.w $a3, $a3, $a2
+; LA64-NEXT: andi $a1, $a1, 255
+; LA64-NEXT: sll.w $a1, $a1, $a2
+; LA64-NEXT: .LBB118_1: # =>This Inner Loop Header: Depth=1
+; LA64-NEXT: ll.w $a4, $a0, 0
+; LA64-NEXT: add.w $a5, $a4, $a1
+; LA64-NEXT: xor $a5, $a4, $a5
+; LA64-NEXT: and $a5, $a5, $a3
+; LA64-NEXT: xor $a5, $a4, $a5
+; LA64-NEXT: sc.w $a5, $a0, 0
+; LA64-NEXT: beqz $a5, .LBB118_1
+; LA64-NEXT: # %bb.2:
+; LA64-NEXT: srl.w $a0, $a4, $a2
+; LA64-NEXT: ret
+;
+; LA64-LAMCAS-LABEL: atomicrmw_add_i8_monotonic:
+; LA64-LAMCAS: # %bb.0:
+; LA64-LAMCAS-NEXT: move $a2, $a0
+; LA64-LAMCAS-NEXT: ld.bu $a0, $a0, 0
+; LA64-LAMCAS-NEXT: .p2align 4, , 16
+; LA64-LAMCAS-NEXT: .LBB118_1: # %atomicrmw.start
+; LA64-LAMCAS-NEXT: # =>This Inner Loop Header: Depth=1
+; LA64-LAMCAS-NEXT: add.d $a3, $a0, $a1
+; LA64-LAMCAS-NEXT: ext.w.b $a4, $a0
+; LA64-LAMCAS-NEXT: amcas.b $a0, $a3, $a2
+; LA64-LAMCAS-NEXT: bne $a0, $a4, .LBB118_1
+; LA64-LAMCAS-NEXT: # %bb.2: # %atomicrmw.end
+; LA64-LAMCAS-NEXT: ret
+ %1 = atomicrmw add ptr %a, i8 %b monotonic
+ ret i8 %1
+}
+
+define i16 @atomicrmw_add_i16_monotonic(ptr %a, i16 %b) nounwind {
+; LA64-LABEL: atomicrmw_add_i16_monotonic:
+; LA64: # %bb.0:
+; LA64-NEXT: slli.d $a2, $a0, 3
+; LA64-NEXT: bstrins.d $a0, $zero, 1, 0
+; LA64-NEXT: lu12i.w $a3, 15
+; LA64-NEXT: ori $a3, $a3, 4095
+; LA64-NEXT: sll.w $a3, $a3, $a2
+; LA64-NEXT: bstrpick.d $a1, $a1, 15, 0
+; LA64-NEXT: sll.w $a1, $a1, $a2
+; LA64-NEXT: .LBB119_1: # =>This Inner Loop Header: Depth=1
+; LA64-NEXT: ll.w $a4, $a0, 0
+; LA64-NEXT: add.w $a5, $a4, $a1
+; LA64-NEXT: xor $a5, $a4, $a5
+; LA64-NEXT: and $a5, $a5, $a3
+; LA64-NEXT: xor $a5, $a4, $a5
+; LA64-NEXT: sc.w $a5, $a0, 0
+; LA64-NEXT: beqz $a5, .LBB119_1
+; LA64-NEXT: # %bb.2:
+; LA64-NEXT: srl.w $a0, $a4, $a2
+; LA64-NEXT: ret
+;
+; LA64-LAMCAS-LABEL: atomicrmw_add_i16_monotonic:
+; LA64-LAMCAS: # %bb.0:
+; LA64-LAMCAS-NEXT: move $a2, $a0
+; LA64-LAMCAS-NEXT: ld.hu $a0, $a0, 0
+; LA64-LAMCAS-NEXT: .p2align 4, , 16
+; LA64-LAMCAS-NEXT: .LBB119_1: # %atomicrmw.start
+; LA64-LAMCAS-NEXT: # =>This Inner Loop Header: Depth=1
+; LA64-LAMCAS-NEXT: add.d $a3, $a0, $a1
+; LA64-LAMCAS-NEXT: ext.w.h $a4, $a0
+; LA64-LAMCAS-NEXT: amcas.h $a0, $a3, $a2
+; LA64-LAMCAS-NEXT: bne $a0, $a4, .LBB119_1
+; LA64-LAMCAS-NEXT: # %bb.2: # %atomicrmw.end
+; LA64-LAMCAS-NEXT: ret
+ %1 = atomicrmw add ptr %a, i16 %b monotonic
+ ret i16 %1
+
+}
+
+define i8 @atomicrmw_sub_i8_monotonic(ptr %a, i8 %b) nounwind {
+; LA64-LABEL: atomicrmw_sub_i8_monotonic:
+; LA64: # %bb.0:
+; LA64-NEXT: slli.d $a2, $a0, 3
+; LA64-NEXT: bstrins.d $a0, $zero, 1, 0
+; LA64-NEXT: ori $a3, $zero, 255
+; LA64-NEXT: sll.w $a3, $a3, $a2
+; LA64-NEXT: andi $a1, $a1, 255
+; LA64-NEXT: sll.w $a1, $a1, $a2
+; LA64-NEXT: .LBB120_1: # =>This Inner Loop Header: Depth=1
+; LA64-NEXT: ll.w $a4, $a0, 0
+; LA64-NEXT: sub.w $a5, $a4, $a1
+; LA64-NEXT: xor $a5, $a4, $a5
+; LA64-NEXT: and $a5, $a5, $a3
+; LA64-NEXT: xor $a5, $a4, $a5
+; LA64-NEXT: sc.w $a5, $a0, 0
+; LA64-NEXT: beqz $a5, .LBB120_1
+; LA64-NEXT: # %bb.2:
+; LA64-NEXT: srl.w $a0, $a4, $a2
+; LA64-NEXT: ret
+;
+; LA64-LAMCAS-LABEL: atomicrmw_sub_i8_monotonic:
+; LA64-LAMCAS: # %bb.0:
+; LA64-LAMCAS-NEXT: move $a2, $a0
+; LA64-LAMCAS-NEXT: ld.bu $a0, $a0, 0
+; LA64-LAMCAS-NEXT: .p2align 4, , 16
+; LA64-LAMCAS-NEXT: .LBB120_1: # %atomicrmw.start
+; LA64-LAMCAS-NEXT: # =>This Inner Loop Header: Depth=1
+; LA64-LAMCAS-NEXT: sub.d $a3, $a0, $a1
+; LA64-LAMCAS-NEXT: ext.w.b $a4, $a0
+; LA64-LAMCAS-NEXT: amcas.b $a0, $a3, $a2
+; LA64-LAMCAS-NEXT: bne $a0, $a4, .LBB120_1
+; LA64-LAMCAS-NEXT: # %bb.2: # %atomicrmw.end
+; LA64-LAMCAS-NEXT: ret
+ %1 = atomicrmw sub ptr %a, i8 %b monotonic
+ ret i8 %1
+}
+
+define i16 @atomicrmw_sub_i16_monotonic(ptr %a, i16 %b) nounwind {
+; LA64-LABEL: atomicrmw_sub_i16_monotonic:
+; LA64: # %bb.0:
+; LA64-NEXT: slli.d $a2, $a0, 3
+; LA64-NEXT: bstrins.d $a0, $zero, 1, 0
+; LA64-NEXT: lu12i.w $a3, 15
+; LA64-NEXT: ori $a3, $a3, 4095
+; LA64-NEXT: sll.w $a3, $a3, $a2
+; LA64-NEXT: bstrpick.d $a1, $a1, 15, 0
+; LA64-NEXT: sll.w $a1, $a1, $a2
+; LA64-NEXT: .LBB121_1: # =>This Inner Loop Header: Depth=1
+; LA64-NEXT: ll.w $a4, $a0, 0
+; LA64-NEXT: sub.w $a5, $a4, $a1
+; LA64-NEXT: xor $a5, $a4, $a5
+; LA64-NEXT: and $a5, $a5, $a3
+; LA64-NEXT: xor $a5, $a4, $a5
+; LA64-NEXT: sc.w $a5, $a0, 0
+; LA64-NEXT: beqz $a5, .LBB121_1
+; LA64-NEXT: # %bb.2:
+; LA64-NEXT: srl.w $a0, $a4, $a2
+; LA64-NEXT: ret
+;
+; LA64-LAMCAS-LABEL: atomicrmw_sub_i16_monotonic:
+; LA64-LAMCAS: # %bb.0:
+; LA64-LAMCAS-NEXT: move $a2, $a0
+; LA64-LAMCAS-NEXT: ld.hu $a0, $a0, 0
+; LA64-LAMCAS-NEXT: .p2align 4, , 16
+; LA64-LAMCAS-NEXT: .LBB121_1: # %atomicrmw.start
+; LA64-LAMCAS-NEXT: # =>This Inner Loop Header: Depth=1
+; LA64-LAMCAS-NEXT: sub.d $a3, $a0, $a1
+; LA64-LAMCAS-NEXT: ext.w.h $a4, $a0
+; LA64-LAMCAS-NEXT: amcas.h $a0, $a3, $a2
+; LA64-LAMCAS-NEXT: bne $a0, $a4, .LBB121_1
+; LA64-LAMCAS-NEXT: # %bb.2: # %atomicrmw.end
+; LA64-LAMCAS-NEXT: ret
+ %1 = atomicrmw sub ptr %a, i16 %b monotonic
+ ret i16 %1
+
+}
+
+define i8 @atomicrmw_umax_i8_monotonic(ptr %a, i8 %b) nounwind {
+; LA64-LABEL: atomicrmw_umax_i8_monotonic:
+; LA64: # %bb.0:
+; LA64-NEXT: slli.d $a2, $a0, 3
+; LA64-NEXT: bstrins.d $a0, $zero, 1, 0
+; LA64-NEXT: ori $a3, $zero, 255
+; LA64-NEXT: sll.w $a3, $a3, $a2
+; LA64-NEXT: andi $a1, $a1, 255
+; LA64-NEXT: sll.w $a1, $a1, $a2
+; LA64-NEXT: .LBB122_1: # =>This Inner Loop Header: Depth=1
+; LA64-NEXT: ll.w $a4, $a0, 0
+; LA64-NEXT: and $a6, $a4, $a3
+; LA64-NEXT: move $a5, $a4
+; LA64-NEXT: bgeu $a6, $a1, .LBB122_3
+; LA64-NEXT: # %bb.2: # in Loop: Header=BB122_1 Depth=1
+; LA64-NEXT: xor $a5, $a4, $a1
+; LA64-NEXT: and $a5, $a5, $a3
+; LA64-NEXT: xor $a5, $a4, $a5
+; LA64-NEXT: .LBB122_3: # in Loop: Header=BB122_1 Depth=1
+; LA64-NEXT: sc.w $a5, $a0, 0
+; LA64-NEXT: beqz $a5, .LBB122_1
+; LA64-NEXT: # %bb.4:
+; LA64-NEXT: srl.w $a0, $a4, $a2
+; LA64-NEXT: ret
+;
+; LA64-LAMCAS-LABEL: atomicrmw_umax_i8_monotonic:
+; LA64-LAMCAS: # %bb.0:
+; LA64-LAMCAS-NEXT: move $a2, $a0
+; LA64-LAMCAS-NEXT: ld.bu $a0, $a0, 0
+; LA64-LAMCAS-NEXT: andi $a3, $a1, 255
+; LA64-LAMCAS-NEXT: .p2align 4, , 16
+; LA64-LAMCAS-NEXT: .LBB122_1: # %atomicrmw.start
+; LA64-LAMCAS-NEXT: # =>This Inner Loop Header: Depth=1
+; LA64-LAMCAS-NEXT: andi $a4, $a0, 255
+; LA64-LAMCAS-NEXT: sltu $a4, $a3, $a4
+; LA64-LAMCAS-NEXT: masknez $a5, $a1, $a4
+; LA64-LAMCAS-NEXT: maskeqz $a4, $a0, $a4
+; LA64-LAMCAS-NEXT: or $a4, $a4, $a5
+; LA64-LAMCAS-NEXT: ext.w.b $a5, $a0
+; LA64-LAMCAS-NEXT: amcas.b $a0, $a4, $a2
+; LA64-LAMCAS-NEXT: bne $a0, $a5, .LBB122_1
+; LA64-LAMCAS-NEXT: # %bb.2: # %atomicrmw.end
+; LA64-LAMCAS-NEXT: ret
+ %1 = atomicrmw umax ptr %a, i8 %b monotonic
+ ret i8 %1
+}
+
+define i16 @atomicrmw_umax_i16_monotonic(ptr %a, i16 %b) nounwind {
+; LA64-LABEL: atomicrmw_umax_i16_monotonic:
+; LA64: # %bb.0:
+; LA64-NEXT: slli.d $a2, $a0, 3
+; LA64-NEXT: bstrins.d $a0, $zero, 1, 0
+; LA64-NEXT: lu12i.w $a3, 15
+; LA64-NEXT: ori $a3, $a3, 4095
+; LA64-NEXT: sll.w $a3, $a3, $a2
+; LA64-NEXT: bstrpick.d $a1, $a1, 15, 0
+; LA64-NEXT: sll.w $a1, $a1, $a2
+; LA64-NEXT: .LBB123_1: # =>This Inner Loop Header: Depth=1
+; LA64-NEXT: ll.w $a4, $a0, 0
+; LA64-NEXT: and $a6, $a4, $a3
+; LA64-NEXT: move $a5, $a4
+; LA64-NEXT: bgeu $a6, $a1, .LBB123_3
+; LA64-NEXT: # %bb.2: # in Loop: Header=BB123_1 Depth=1
+; LA64-NEXT: xor $a5, $a4, $a1
+; LA64-NEXT: and $a5, $a5, $a3
+; LA64-NEXT: xor $a5, $a4, $a5
+; LA64-NEXT: .LBB123_3: # in Loop: Header=BB123_1 Depth=1
+; LA64-NEXT: sc.w $a5, $a0, 0
+; LA64-NEXT: beqz $a5, .LBB123_1
+; LA64-NEXT: # %bb.4:
+; LA64-NEXT: srl.w $a0, $a4, $a2
+; LA64-NEXT: ret
+;
+; LA64-LAMCAS-LABEL: atomicrmw_umax_i16_monotonic:
+; LA64-LAMCAS: # %bb.0:
+; LA64-LAMCAS-NEXT: move $a2, $a0
+; LA64-LAMCAS-NEXT: ld.hu $a0, $a0, 0
+; LA64-LAMCAS-NEXT: bstrpick.d $a3, $a1, 15, 0
+; LA64-LAMCAS-NEXT: .p2align 4, , 16
+; LA64-LAMCAS-NEXT: .LBB123_1: # %atomicrmw.start
+; LA64-LAMCAS-NEXT: # =>This Inner Loop Header: Depth=1
+; LA64-LAMCAS-NEXT: bstrpick.d $a4, $a0, 15, 0
+; LA64-LAMCAS-NEXT: sltu $a4, $a3, $a4
+; LA64-LAMCAS-NEXT: masknez $a5, $a1, $a4
+; LA64-LAMCAS-NEXT: maskeqz $a4, $a0, $a4
+; LA64-LAMCAS-NEXT: or $a4, $a4, $a5
+; LA64-LAMCAS-NEXT: ext.w.h $a5, $a0
+; LA64-LAMCAS-NEXT: amcas.h $a0, $a4, $a2
+; LA64-LAMCAS-NEXT: bne $a0, $a5, .LBB123_1
+; LA64-LAMCAS-NEXT: # %bb.2: # %atomicrmw.end
+; LA64-LAMCAS-NEXT: ret
+ %1 = atomicrmw umax ptr %a, i16 %b monotonic
+ ret i16 %1
+}
+
+define i8 @atomicrmw_umin_i8_monotonic(ptr %a, i8 %b) nounwind {
+; LA64-LABEL: atomicrmw_umin_i8_monotonic:
+; LA64: # %bb.0:
+; LA64-NEXT: slli.d $a2, $a0, 3
+; LA64-NEXT: bstrins.d $a0, $zero, 1, 0
+; LA64-NEXT: ori $a3, $zero, 255
+; LA64-NEXT: sll.w $a3, $a3, $a2
+; LA64-NEXT: andi $a1, $a1, 255
+; LA64-NEXT: sll.w $a1, $a1, $a2
+; LA64-NEXT: .LBB124_1: # =>This Inner Loop Header: Depth=1
+; LA64-NEXT: ll.w $a4, $a0, 0
+; LA64-NEXT: and $a6, $a4, $a3
+; LA64-NEXT: move $a5, $a4
+; LA64-NEXT: bgeu $a1, $a6, .LBB124_3
+; LA64-NEXT: # %bb.2: # in Loop: Header=BB124_1 Depth=1
+; LA64-NEXT: xor $a5, $a4, $a1
+; LA64-NEXT: and $a5, $a5, $a3
+; LA64-NEXT: xor $a5, $a4, $a5
+; LA64-NEXT: .LBB124_3: # in Loop: Header=BB124_1 Depth=1
+; LA64-NEXT: sc.w $a5, $a0, 0
+; LA64-NEXT: beqz $a5, .LBB124_1
+; LA64-NEXT: # %bb.4:
+; LA64-NEXT: srl.w $a0, $a4, $a2
+; LA64-NEXT: ret
+;
+; LA64-LAMCAS-LABEL: atomicrmw_umin_i8_monotonic:
+; LA64-LAMCAS: # %bb.0:
+; LA64-LAMCAS-NEXT: move $a2, $a0
+; LA64-LAMCAS-NEXT: ld.bu $a0, $a0, 0
+; LA64-LAMCAS-NEXT: andi $a3, $a1, 255
+; LA64-LAMCAS-NEXT: .p2align 4, , 16
+; LA64-LAMCAS-NEXT: .LBB124_1: # %atomicrmw.start
+; LA64-LAMCAS-NEXT: # =>This Inner Loop Header: Depth=1
+; LA64-LAMCAS-NEXT: andi $a4, $a0, 255
+; LA64-LAMCAS-NEXT: sltu $a4, $a3, $a4
+; LA64-LAMCAS-NEXT: xori $a4, $a4, 1
+; LA64-LAMCAS-NEXT: masknez $a5, $a1, $a4
+; LA64-LAMCAS-NEXT: maskeqz $a4, $a0, $a4
+; LA64-LAMCAS-NEXT: or $a4, $a4, $a5
+; LA64-LAMCAS-NEXT: ext.w.b $a5, $a0
+; LA64-LAMCAS-NEXT: amcas.b $a0, $a4, $a2
+; LA64-LAMCAS-NEXT: bne $a0, $a5, .LBB124_1
+; LA64-LAMCAS-NEXT: # %bb.2: # %atomicrmw.end
+; LA64-LAMCAS-NEXT: ret
+ %1 = atomicrmw umin ptr %a, i8 %b monotonic
+ ret i8 %1
+}
+
+define i16 @atomicrmw_umin_i16_monotonic(ptr %a, i16 %b) nounwind {
+; LA64-LABEL: atomicrmw_umin_i16_monotonic:
+; LA64: # %bb.0:
+; LA64-NEXT: slli.d $a2, $a0, 3
+; LA64-NEXT: bstrins.d $a0, $zero, 1, 0
+; LA64-NEXT: lu12i.w $a3, 15
+; LA64-NEXT: ori $a3, $a3, 4095
+; LA64-NEXT: sll.w $a3, $a3, $a2
+; LA64-NEXT: bstrpick.d $a1, $a1, 15, 0
+; LA64-NEXT: sll.w $a1, $a1, $a2
+; LA64-NEXT: .LBB125_1: # =>This Inner Loop Header: Depth=1
+; LA64-NEXT: ll.w $a4, $a0, 0
+; LA64-NEXT: and $a6, $a4, $a3
+; LA64-NEXT: move $a5, $a4
+; LA64-NEXT: bgeu $a1, $a6, .LBB125_3
+; LA64-NEXT: # %bb.2: # in Loop: Header=BB125_1 Depth=1
+; LA64-NEXT: xor $a5, $a4, $a1
+; LA64-NEXT: and $a5, $a5, $a3
+; LA64-NEXT: xor $a5, $a4, $a5
+; LA64-NEXT: .LBB125_3: # in Loop: Header=BB125_1 Depth=1
+; LA64-NEXT: sc.w $a5, $a0, 0
+; LA64-NEXT: beqz $a5, .LBB125_1
+; LA64-NEXT: # %bb.4:
+; LA64-NEXT: srl.w $a0, $a4, $a2
+; LA64-NEXT: ret
+;
+; LA64-LAMCAS-LABEL: atomicrmw_umin_i16_monotonic:
+; LA64-LAMCAS: # %bb.0:
+; LA64-LAMCAS-NEXT: move $a2, $a0
+; LA64-LAMCAS-NEXT: ld.hu $a0, $a0, 0
+; LA64-LAMCAS-NEXT: bstrpick.d $a3, $a1, 15, 0
+; LA64-LAMCAS-NEXT: .p2align 4, , 16
+; LA64-LAMCAS-NEXT: .LBB125_1: # %atomicrmw.start
+; LA64-LAMCAS-NEXT: # =>This Inner Loop Header: Depth=1
+; LA64-LAMCAS-NEXT: bstrpick.d $a4, $a0, 15, 0
+; LA64-LAMCAS-NEXT: sltu $a4, $a3, $a4
+; LA64-LAMCAS-NEXT: xori $a4, $a4, 1
+; LA64-LAMCAS-NEXT: masknez $a5, $a1, $a4
+; LA64-LAMCAS-NEXT: maskeqz $a4, $a0, $a4
+; LA64-LAMCAS-NEXT: or $a4, $a4, $a5
+; LA64-LAMCAS-NEXT: ext.w.h $a5, $a0
+; LA64-LAMCAS-NEXT: amcas.h $a0, $a4, $a2
+; LA64-LAMCAS-NEXT: bne $a0, $a5, .LBB125_1
+; LA64-LAMCAS-NEXT: # %bb.2: # %atomicrmw.end
+; LA64-LAMCAS-NEXT: ret
+ %1 = atomicrmw umin ptr %a, i16 %b monotonic
+ ret i16 %1
+}
+
+define i8 @atomicrmw_max_i8_monotonic(ptr %a, i8 %b) nounwind {
+; LA64-LABEL: atomicrmw_max_i8_monotonic:
+; LA64: # %bb.0:
+; LA64-NEXT: slli.d $a2, $a0, 3
+; LA64-NEXT: bstrins.d $a0, $zero, 1, 0
+; LA64-NEXT: andi $a3, $a2, 24
+; LA64-NEXT: ori $a4, $zero, 255
+; LA64-NEXT: sll.w $a4, $a4, $a2
+; LA64-NEXT: ext.w.b $a1, $a1
+; LA64-NEXT: sll.w $a1, $a1, $a2
+; LA64-NEXT: xori $a3, $a3, 56
+; LA64-NEXT: .LBB126_1: # =>This Inner Loop Header: Depth=1
+; LA64-NEXT: ll.w $a5, $a0, 0
+; LA64-NEXT: and $a7, $a5, $a4
+; LA64-NEXT: move $a6, $a5
+; LA64-NEXT: sll.w $a7, $a7, $a3
+; LA64-NEXT: sra.w $a7, $a7, $a3
+; LA64-NEXT: bge $a7, $a1, .LBB126_3
+; LA64-NEXT: # %bb.2: # in Loop: Header=BB126_1 Depth=1
+; LA64-NEXT: xor $a6, $a5, $a1
+; LA64-NEXT: and $a6, $a6, $a4
+; LA64-NEXT: xor $a6, $a5, $a6
+; LA64-NEXT: .LBB126_3: # in Loop: Header=BB126_1 Depth=1
+; LA64-NEXT: sc.w $a6, $a0, 0
+; LA64-NEXT: beqz $a6, .LBB126_1
+; LA64-NEXT: # %bb.4:
+; LA64-NEXT: srl.w $a0, $a5, $a2
+; LA64-NEXT: ret
+;
+; LA64-LAMCAS-LABEL: atomicrmw_max_i8_monotonic:
+; LA64-LAMCAS: # %bb.0:
+; LA64-LAMCAS-NEXT: move $a2, $a0
+; LA64-LAMCAS-NEXT: ld.bu $a0, $a0, 0
+; LA64-LAMCAS-NEXT: ext.w.b $a3, $a1
+; LA64-LAMCAS-NEXT: .p2align 4, , 16
+; LA64-LAMCAS-NEXT: .LBB126_1: # %atomicrmw.start
+; LA64-LAMCAS-NEXT: # =>This Inner Loop Header: Depth=1
+; LA64-LAMCAS-NEXT: ext.w.b $a4, $a0
+; LA64-LAMCAS-NEXT: slt $a5, $a3, $a4
+; LA64-LAMCAS-NEXT: masknez $a6, $a1, $a5
+; LA64-LAMCAS-NEXT: maskeqz $a5, $a0, $a5
+; LA64-LAMCAS-NEXT: or $a5, $a5, $a6
+; LA64-LAMCAS-NEXT: amcas.b $a0, $a5, $a2
+; LA64-LAMCAS-NEXT: bne $a0, $a4, .LBB126_1
+; LA64-LAMCAS-NEXT: # %bb.2: # %atomicrmw.end
+; LA64-LAMCAS-NEXT: ret
+ %1 = atomicrmw max ptr %a, i8 %b monotonic
+ ret i8 %1
+}
+
+define i16 @atomicrmw_max_i16_monotonic(ptr %a, i16 %b) nounwind {
+; LA64-LABEL: atomicrmw_max_i16_monotonic:
+; LA64: # %bb.0:
+; LA64-NEXT: slli.d $a2, $a0, 3
+; LA64-NEXT: bstrins.d $a0, $zero, 1, 0
+; LA64-NEXT: andi $a3, $a2, 24
+; LA64-NEXT: lu12i.w $a4, 15
+; LA64-NEXT: ori $a4, $a4, 4095
+; LA64-NEXT: sll.w $a4, $a4, $a2
+; LA64-NEXT: ext.w.h $a1, $a1
+; LA64-NEXT: sll.w $a1, $a1, $a2
+; LA64-NEXT: ori $a5, $zero, 48
+; LA64-NEXT: sub.d $a3, $a5, $a3
+; LA64-NEXT: .LBB127_1: # =>This Inner Loop Header: Depth=1
+; LA64-NEXT: ll.w $a5, $a0, 0
+; LA64-NEXT: and $a7, $a5, $a4
+; LA64-NEXT: move $a6, $a5
+; LA64-NEXT: sll.w $a7, $a7, $a3
+; LA64-NEXT: sra.w $a7, $a7, $a3
+; LA64-NEXT: bge $a7, $a1, .LBB127_3
+; LA64-NEXT: # %bb.2: # in Loop: Header=BB127_1 Depth=1
+; LA64-NEXT: xor $a6, $a5, $a1
+; LA64-NEXT: and $a6, $a6, $a4
+; LA64-NEXT: xor $a6, $a5, $a6
+; LA64-NEXT: .LBB127_3: # in Loop: Header=BB127_1 Depth=1
+; LA64-NEXT: sc.w $a6, $a0, 0
+; LA64-NEXT: beqz $a6, .LBB127_1
+; LA64-NEXT: # %bb.4:
+; LA64-NEXT: srl.w $a0, $a5, $a2
+; LA64-NEXT: ret
+;
+; LA64-LAMCAS-LABEL: atomicrmw_max_i16_monotonic:
+; LA64-LAMCAS: # %bb.0:
+; LA64-LAMCAS-NEXT: move $a2, $a0
+; LA64-LAMCAS-NEXT: ld.hu $a0, $a0, 0
+; LA64-LAMCAS-NEXT: ext.w.h $a3, $a1
+; LA64-LAMCAS-NEXT: .p2align 4, , 16
+; LA64-LAMCAS-NEXT: .LBB127_1: # %atomicrmw.start
+; LA64-LAMCAS-NEXT: # =>This Inner Loop Header: Depth=1
+; LA64-LAMCAS-NEXT: ext.w.h $a4, $a0
+; LA64-LAMCAS-NEXT: slt $a5, $a3, $a4
+; LA64-LAMCAS-NEXT: masknez $a6, $a1, $a5
+; LA64-LAMCAS-NEXT: maskeqz $a5, $a0, $a5
+; LA64-LAMCAS-NEXT: or $a5, $a5, $a6
+; LA64-LAMCAS-NEXT: amcas.h $a0, $a5, $a2
+; LA64-LAMCAS-NEXT: bne $a0, $a4, .LBB127_1
+; LA64-LAMCAS-NEXT: # %bb.2: # %atomicrmw.end
+; LA64-LAMCAS-NEXT: ret
+ %1 = atomicrmw max ptr %a, i16 %b monotonic
+ ret i16 %1
+}
+
+define i8 @atomicrmw_min_i8_monotonic(ptr %a, i8 %b) nounwind {
+; LA64-LABEL: atomicrmw_min_i8_monotonic:
+; LA64: # %bb.0:
+; LA64-NEXT: slli.d $a2, $a0, 3
+; LA64-NEXT: bstrins.d $a0, $zero, 1, 0
+; LA64-NEXT: andi $a3, $a2, 24
+; LA64-NEXT: ori $a4, $zero, 255
+; LA64-NEXT: sll.w $a4, $a4, $a2
+; LA64-NEXT: ext.w.b $a1, $a1
+; LA64-NEXT: sll.w $a1, $a1, $a2
+; LA64-NEXT: xori $a3, $a3, 56
+; LA64-NEXT: .LBB128_1: # =>This Inner Loop Header: Depth=1
+; LA64-NEXT: ll.w $a5, $a0, 0
+; LA64-NEXT: and $a7, $a5, $a4
+; LA64-NEXT: move $a6, $a5
+; LA64-NEXT: sll.w $a7, $a7, $a3
+; LA64-NEXT: sra.w $a7, $a7, $a3
+; LA64-NEXT: bge $a1, $a7, .LBB128_3
+; LA64-NEXT: # %bb.2: # in Loop: Header=BB128_1 Depth=1
+; LA64-NEXT: xor $a6, $a5, $a1
+; LA64-NEXT: and $a6, $a6, $a4
+; LA64-NEXT: xor $a6, $a5, $a6
+; LA64-NEXT: .LBB128_3: # in Loop: Header=BB128_1 Depth=1
+; LA64-NEXT: sc.w $a6, $a0, 0
+; LA64-NEXT: beqz $a6, .LBB128_1
+; LA64-NEXT: # %bb.4:
+; LA64-NEXT: srl.w $a0, $a5, $a2
+; LA64-NEXT: ret
+;
+; LA64-LAMCAS-LABEL: atomicrmw_min_i8_monotonic:
+; LA64-LAMCAS: # %bb.0:
+; LA64-LAMCAS-NEXT: move $a2, $a0
+; LA64-LAMCAS-NEXT: ld.bu $a0, $a0, 0
+; LA64-LAMCAS-NEXT: ext.w.b $a3, $a1
+; LA64-LAMCAS-NEXT: .p2align 4, , 16
+; LA64-LAMCAS-NEXT: .LBB128_1: # %atomicrmw.start
+; LA64-LAMCAS-NEXT: # =>This Inner Loop Header: Depth=1
+; LA64-LAMCAS-NEXT: ext.w.b $a4, $a0
+; LA64-LAMCAS-NEXT: slt $a5, $a3, $a4
+; LA64-LAMCAS-NEXT: xori $a5, $a5, 1
+; LA64-LAMCAS-NEXT: masknez $a6, $a1, $a5
+; LA64-LAMCAS-NEXT: maskeqz $a5, $a0, $a5
+; LA64-LAMCAS-NEXT: or $a5, $a5, $a6
+; LA64-LAMCAS-NEXT: amcas.b $a0, $a5, $a2
+; LA64-LAMCAS-NEXT: bne $a0, $a4, .LBB128_1
+; LA64-LAMCAS-NEXT: # %bb.2: # %atomicrmw.end
+; LA64-LAMCAS-NEXT: ret
+ %1 = atomicrmw min ptr %a, i8 %b monotonic
+ ret i8 %1
+}
+
+define i16 @atomicrmw_min_i16_monotonic(ptr %a, i16 %b) nounwind {
+; LA64-LABEL: atomicrmw_min_i16_monotonic:
+; LA64: # %bb.0:
+; LA64-NEXT: slli.d $a2, $a0, 3
+; LA64-NEXT: bstrins.d $a0, $zero, 1, 0
+; LA64-NEXT: andi $a3, $a2, 24
+; LA64-NEXT: lu12i.w $a4, 15
+; LA64-NEXT: ori $a4, $a4, 4095
+; LA64-NEXT: sll.w $a4, $a4, $a2
+; LA64-NEXT: ext.w.h $a1, $a1
+; LA64-NEXT: sll.w $a1, $a1, $a2
+; LA64-NEXT: ori $a5, $zero, 48
+; LA64-NEXT: sub.d $a3, $a5, $a3
+; LA64-NEXT: .LBB129_1: # =>This Inner Loop Header: Depth=1
+; LA64-NEXT: ll.w $a5, $a0, 0
+; LA64-NEXT: and $a7, $a5, $a4
+; LA64-NEXT: move $a6, $a5
+; LA64-NEXT: sll.w $a7, $a7, $a3
+; LA64-NEXT: sra.w $a7, $a7, $a3
+; LA64-NEXT: bge $a1, $a7, .LBB129_3
+; LA64-NEXT: # %bb.2: # in Loop: Header=BB129_1 Depth=1
+; LA64-NEXT: xor $a6, $a5, $a1
+; LA64-NEXT: and $a6, $a6, $a4
+; LA64-NEXT: xor $a6, $a5, $a6
+; LA64-NEXT: .LBB129_3: # in Loop: Header=BB129_1 Depth=1
+; LA64-NEXT: sc.w $a6, $a0, 0
+; LA64-NEXT: beqz $a6, .LBB129_1
+; LA64-NEXT: # %bb.4:
+; LA64-NEXT: srl.w $a0, $a5, $a2
+; LA64-NEXT: ret
+;
+; LA64-LAMCAS-LABEL: atomicrmw_min_i16_monotonic:
+; LA64-LAMCAS: # %bb.0:
+; LA64-LAMCAS-NEXT: move $a2, $a0
+; LA64-LAMCAS-NEXT: ld.hu $a0, $a0, 0
+; LA64-LAMCAS-NEXT: ext.w.h $a3, $a1
+; LA64-LAMCAS-NEXT: .p2align 4, , 16
+; LA64-LAMCAS-NEXT: .LBB129_1: # %atomicrmw.start
+; LA64-LAMCAS-NEXT: # =>This Inner Loop Header: Depth=1
+; LA64-LAMCAS-NEXT: ext.w.h $a4, $a0
+; LA64-LAMCAS-NEXT: slt $a5, $a3, $a4
+; LA64-LAMCAS-NEXT: xori $a5, $a5, 1
+; LA64-LAMCAS-NEXT: masknez $a6, $a1, $a5
+; LA64-LAMCAS-NEXT: maskeqz $a5, $a0, $a5
+; LA64-LAMCAS-NEXT: or $a5, $a5, $a6
+; LA64-LAMCAS-NEXT: amcas.h $a0, $a5, $a2
+; LA64-LAMCAS-NEXT: bne $a0, $a4, .LBB129_1
+; LA64-LAMCAS-NEXT: # %bb.2: # %atomicrmw.end
+; LA64-LAMCAS-NEXT: ret
+ %1 = atomicrmw min ptr %a, i16 %b monotonic
+ ret i16 %1
+}
+
+
+
+define i8 @atomicrmw_nand_i8_monotonic(ptr %a, i8 %b) nounwind {
+; LA64-LABEL: atomicrmw_nand_i8_monotonic:
+; LA64: # %bb.0:
+; LA64-NEXT: slli.d $a2, $a0, 3
+; LA64-NEXT: bstrins.d $a0, $zero, 1, 0
+; LA64-NEXT: ori $a3, $zero, 255
+; LA64-NEXT: sll.w $a3, $a3, $a2
+; LA64-NEXT: andi $a1, $a1, 255
+; LA64-NEXT: sll.w $a1, $a1, $a2
+; LA64-NEXT: .LBB130_1: # =>This Inner Loop Header: Depth=1
+; LA64-NEXT: ll.w $a4, $a0, 0
+; LA64-NEXT: and $a5, $a4, $a1
+; LA64-NEXT: nor $a5, $a5, $zero
+; LA64-NEXT: xor $a5, $a4, $a5
+; LA64-NEXT: and $a5, $a5, $a3
+; LA64-NEXT: xor $a5, $a4, $a5
+; LA64-NEXT: sc.w $a5, $a0, 0
+; LA64-NEXT: beqz $a5, .LBB130_1
+; LA64-NEXT: # %bb.2:
+; LA64-NEXT: srl.w $a0, $a4, $a2
+; LA64-NEXT: ret
+;
+; LA64-LAMCAS-LABEL: atomicrmw_nand_i8_monotonic:
+; LA64-LAMCAS: # %bb.0:
+; LA64-LAMCAS-NEXT: move $a2, $a0
+; LA64-LAMCAS-NEXT: ld.bu $a0, $a0, 0
+; LA64-LAMCAS-NEXT: .p2align 4, , 16
+; LA64-LAMCAS-NEXT: .LBB130_1: # %atomicrmw.start
+; LA64-LAMCAS-NEXT: # =>This Inner Loop Header: Depth=1
+; LA64-LAMCAS-NEXT: and $a3, $a0, $a1
+; LA64-LAMCAS-NEXT: nor $a3, $a3, $zero
+; LA64-LAMCAS-NEXT: ext.w.b $a4, $a0
+; LA64-LAMCAS-NEXT: amcas.b $a0, $a3, $a2
+; LA64-LAMCAS-NEXT: bne $a0, $a4, .LBB130_1
+; LA64-LAMCAS-NEXT: # %bb.2: # %atomicrmw.end
+; LA64-LAMCAS-NEXT: ret
+ %1 = atomicrmw nand ptr %a, i8 %b monotonic
+ ret i8 %1
+}
+
+define i16 @atomicrmw_nand_i16_monotonic(ptr %a, i16 %b) nounwind {
+; LA64-LABEL: atomicrmw_nand_i16_monotonic:
+; LA64: # %bb.0:
+; LA64-NEXT: slli.d $a2, $a0, 3
+; LA64-NEXT: bstrins.d $a0, $zero, 1, 0
+; LA64-NEXT: lu12i.w $a3, 15
+; LA64-NEXT: ori $a3, $a3, 4095
+; LA64-NEXT: sll.w $a3, $a3, $a2
+; LA64-NEXT: bstrpick.d $a1, $a1, 15, 0
+; LA64-NEXT: sll.w $a1, $a1, $a2
+; LA64-NEXT: .LBB131_1: # =>This Inner Loop Header: Depth=1
+; LA64-NEXT: ll.w $a4, $a0, 0
+; LA64-NEXT: and $a5, $a4, $a1
+; LA64-NEXT: nor $a5, $a5, $zero
+; LA64-NEXT: xor $a5, $a4, $a5
+; LA64-NEXT: and $a5, $a5, $a3
+; LA64-NEXT: xor $a5, $a4, $a5
+; LA64-NEXT: sc.w $a5, $a0, 0
+; LA64-NEXT: beqz $a5, .LBB131_1
+; LA64-NEXT: # %bb.2:
+; LA64-NEXT: srl.w $a0, $a4, $a2
+; LA64-NEXT: ret
+;
+; LA64-LAMCAS-LABEL: atomicrmw_nand_i16_monotonic:
+; LA64-LAMCAS: # %bb.0:
+; LA64-LAMCAS-NEXT: move $a2, $a0
+; LA64-LAMCAS-NEXT: ld.hu $a0, $a0, 0
+; LA64-LAMCAS-NEXT: .p2align 4, , 16
+; LA64-LAMCAS-NEXT: .LBB131_1: # %atomicrmw.start
+; LA64-LAMCAS-NEXT: # =>This Inner Loop Header: Depth=1
+; LA64-LAMCAS-NEXT: and $a3, $a0, $a1
+; LA64-LAMCAS-NEXT: nor $a3, $a3, $zero
+; LA64-LAMCAS-NEXT: ext.w.h $a4, $a0
+; LA64-LAMCAS-NEXT: amcas.h $a0, $a3, $a2
+; LA64-LAMCAS-NEXT: bne $a0, $a4, .LBB131_1
+; LA64-LAMCAS-NEXT: # %bb.2: # %atomicrmw.end
+; LA64-LAMCAS-NEXT: ret
+ %1 = atomicrmw nand ptr %a, i16 %b monotonic
+ ret i16 %1
+
+}
+
+define i32 @atomicrmw_nand_i32_monotonic(ptr %a, i32 %b) nounwind {
+; LA64-LABEL: atomicrmw_nand_i32_monotonic:
+; LA64: # %bb.0:
+; LA64-NEXT: .LBB132_1: # =>This Inner Loop Header: Depth=1
+; LA64-NEXT: ll.w $a2, $a0, 0
+; LA64-NEXT: and $a3, $a2, $a1
+; LA64-NEXT: nor $a3, $a3, $zero
+; LA64-NEXT: sc.w $a3, $a0, 0
+; LA64-NEXT: beqz $a3, .LBB132_1
+; LA64-NEXT: # %bb.2:
+; LA64-NEXT: move $a0, $a2
+; LA64-NEXT: ret
+;
+; LA64-LAMCAS-LABEL: atomicrmw_nand_i32_monotonic:
+; LA64-LAMCAS: # %bb.0:
+; LA64-LAMCAS-NEXT: move $a2, $a0
+; LA64-LAMCAS-NEXT: ld.w $a0, $a0, 0
+; LA64-LAMCAS-NEXT: .p2align 4, , 16
+; LA64-LAMCAS-NEXT: .LBB132_1: # %atomicrmw.start
+; LA64-LAMCAS-NEXT: # =>This Inner Loop Header: Depth=1
+; LA64-LAMCAS-NEXT: move $a3, $a0
+; LA64-LAMCAS-NEXT: and $a4, $a0, $a1
+; LA64-LAMCAS-NEXT: nor $a4, $a4, $zero
+; LA64-LAMCAS-NEXT: amcas.w $a0, $a4, $a2
+; LA64-LAMCAS-NEXT: bne $a0, $a3, .LBB132_1
+; LA64-LAMCAS-NEXT: # %bb.2: # %atomicrmw.end
+; LA64-LAMCAS-NEXT: ret
+ %1 = atomicrmw nand ptr %a, i32 %b monotonic
+ ret i32 %1
+}
+
+define i64 @atomicrmw_nand_i64_monotonic(ptr %a, i64 %b) nounwind {
+; LA64-LABEL: atomicrmw_nand_i64_monotonic:
+; LA64: # %bb.0:
+; LA64-NEXT: .LBB133_1: # =>This Inner Loop Header: Depth=1
+; LA64-NEXT: ll.d $a2, $a0, 0
+; LA64-NEXT: and $a3, $a2, $a1
+; LA64-NEXT: nor $a3, $a3, $zero
+; LA64-NEXT: sc.d $a3, $a0, 0
+; LA64-NEXT: beqz $a3, .LBB133_1
+; LA64-NEXT: # %bb.2:
+; LA64-NEXT: move $a0, $a2
+; LA64-NEXT: ret
+;
+; LA64-LAMCAS-LABEL: atomicrmw_nand_i64_monotonic:
+; LA64-LAMCAS: # %bb.0:
+; LA64-LAMCAS-NEXT: move $a2, $a0
+; LA64-LAMCAS-NEXT: ld.d $a0, $a0, 0
+; LA64-LAMCAS-NEXT: .p2align 4, , 16
+; LA64-LAMCAS-NEXT: .LBB133_1: # %atomicrmw.start
+; LA64-LAMCAS-NEXT: # =>This Inner Loop Header: Depth=1
+; LA64-LAMCAS-NEXT: move $a3, $a0
+; LA64-LAMCAS-NEXT: and $a4, $a0, $a1
+; LA64-LAMCAS-NEXT: nor $a4, $a4, $zero
+; LA64-LAMCAS-NEXT: amcas.d $a0, $a4, $a2
+; LA64-LAMCAS-NEXT: bne $a0, $a3, .LBB133_1
+; LA64-LAMCAS-NEXT: # %bb.2: # %atomicrmw.end
+; LA64-LAMCAS-NEXT: ret
+ %1 = atomicrmw nand ptr %a, i64 %b monotonic
+ ret i64 %1
+}
+
+
+
+define i8 @atomicrmw_and_i8_monotonic(ptr %a, i8 %b) nounwind {
+; LA64-LABEL: atomicrmw_and_i8_monotonic:
+; LA64: # %bb.0:
+; LA64-NEXT: slli.d $a2, $a0, 3
+; LA64-NEXT: bstrins.d $a0, $zero, 1, 0
+; LA64-NEXT: ori $a3, $zero, 255
+; LA64-NEXT: sll.w $a3, $a3, $a2
+; LA64-NEXT: andi $a1, $a1, 255
+; LA64-NEXT: sll.w $a1, $a1, $a2
+; LA64-NEXT: orn $a1, $a1, $a3
+; LA64-NEXT: amand.w $a3, $a1, $a0
+; LA64-NEXT: srl.w $a0, $a3, $a2
+; LA64-NEXT: ret
+;
+; LA64-LAMCAS-LABEL: atomicrmw_and_i8_monotonic:
+; LA64-LAMCAS: # %bb.0:
+; LA64-LAMCAS-NEXT: move $a2, $a0
+; LA64-LAMCAS-NEXT: ld.bu $a0, $a0, 0
+; LA64-LAMCAS-NEXT: .p2align 4, , 16
+; LA64-LAMCAS-NEXT: .LBB134_1: # %atomicrmw.start
+; LA64-LAMCAS-NEXT: # =>This Inner Loop Header: Depth=1
+; LA64-LAMCAS-NEXT: and $a3, $a0, $a1
+; LA64-LAMCAS-NEXT: ext.w.b $a4, $a0
+; LA64-LAMCAS-NEXT: amcas.b $a0, $a3, $a2
+; LA64-LAMCAS-NEXT: bne $a0, $a4, .LBB134_1
+; LA64-LAMCAS-NEXT: # %bb.2: # %atomicrmw.end
+; LA64-LAMCAS-NEXT: ret
+ %1 = atomicrmw and ptr %a, i8 %b monotonic
+ ret i8 %1
+}
+
+define i16 @atomicrmw_and_i16_monotonic(ptr %a, i16 %b) nounwind {
+; LA64-LABEL: atomicrmw_and_i16_monotonic:
+; LA64: # %bb.0:
+; LA64-NEXT: slli.d $a2, $a0, 3
+; LA64-NEXT: bstrins.d $a0, $zero, 1, 0
+; LA64-NEXT: lu12i.w $a3, 15
+; LA64-NEXT: ori $a3, $a3, 4095
+; LA64-NEXT: sll.w $a3, $a3, $a2
+; LA64-NEXT: bstrpick.d $a1, $a1, 15, 0
+; LA64-NEXT: sll.w $a1, $a1, $a2
+; LA64-NEXT: orn $a1, $a1, $a3
+; LA64-NEXT: amand.w $a3, $a1, $a0
+; LA64-NEXT: srl.w $a0, $a3, $a2
+; LA64-NEXT: ret
+;
+; LA64-LAMCAS-LABEL: atomicrmw_and_i16_monotonic:
+; LA64-LAMCAS: # %bb.0:
+; LA64-LAMCAS-NEXT: move $a2, $a0
+; LA64-LAMCAS-NEXT: ld.hu $a0, $a0, 0
+; LA64-LAMCAS-NEXT: .p2align 4, , 16
+; LA64-LAMCAS-NEXT: .LBB135_1: # %atomicrmw.start
+; LA64-LAMCAS-NEXT: # =>This Inner Loop Header: Depth=1
+; LA64-LAMCAS-NEXT: and $a3, $a0, $a1
+; LA64-LAMCAS-NEXT: ext.w.h $a4, $a0
+; LA64-LAMCAS-NEXT: amcas.h $a0, $a3, $a2
+; LA64-LAMCAS-NEXT: bne $a0, $a4, .LBB135_1
+; LA64-LAMCAS-NEXT: # %bb.2: # %atomicrmw.end
+; LA64-LAMCAS-NEXT: ret
+ %1 = atomicrmw and ptr %a, i16 %b monotonic
+ ret i16 %1
+
+}
+
+define i8 @atomicrmw_or_i8_monotonic(ptr %a, i8 %b) nounwind {
+; LA64-LABEL: atomicrmw_or_i8_monotonic:
+; LA64: # %bb.0:
+; LA64-NEXT: slli.d $a2, $a0, 3
+; LA64-NEXT: bstrins.d $a0, $zero, 1, 0
+; LA64-NEXT: andi $a1, $a1, 255
+; LA64-NEXT: sll.w $a1, $a1, $a2
+; LA64-NEXT: amor.w $a3, $a1, $a0
+; LA64-NEXT: srl.w $a0, $a3, $a2
+; LA64-NEXT: ret
+;
+; LA64-LAMCAS-LABEL: atomicrmw_or_i8_monotonic:
+; LA64-LAMCAS: # %bb.0:
+; LA64-LAMCAS-NEXT: move $a2, $a0
+; LA64-LAMCAS-NEXT: ld.bu $a0, $a0, 0
+; LA64-LAMCAS-NEXT: .p2align 4, , 16
+; LA64-LAMCAS-NEXT: .LBB136_1: # %atomicrmw.start
+; LA64-LAMCAS-NEXT: # =>This Inner Loop Header: Depth=1
+; LA64-LAMCAS-NEXT: or $a3, $a0, $a1
+; LA64-LAMCAS-NEXT: ext.w.b $a4, $a0
+; LA64-LAMCAS-NEXT: amcas.b $a0, $a3, $a2
+; LA64-LAMCAS-NEXT: bne $a0, $a4, .LBB136_1
+; LA64-LAMCAS-NEXT: # %bb.2: # %atomicrmw.end
+; LA64-LAMCAS-NEXT: ret
+ %1 = atomicrmw or ptr %a, i8 %b monotonic
+ ret i8 %1
+}
+
+define i16 @atomicrmw_or_i16_monotonic(ptr %a, i16 %b) nounwind {
+; LA64-LABEL: atomicrmw_or_i16_monotonic:
+; LA64: # %bb.0:
+; LA64-NEXT: slli.d $a2, $a0, 3
+; LA64-NEXT: bstrins.d $a0, $zero, 1, 0
+; LA64-NEXT: bstrpick.d $a1, $a1, 15, 0
+; LA64-NEXT: sll.w $a1, $a1, $a2
+; LA64-NEXT: amor.w $a3, $a1, $a0
+; LA64-NEXT: srl.w $a0, $a3, $a2
+; LA64-NEXT: ret
+;
+; LA64-LAMCAS-LABEL: atomicrmw_or_i16_monotonic:
+; LA64-LAMCAS: # %bb.0:
+; LA64-LAMCAS-NEXT: move $a2, $a0
+; LA64-LAMCAS-NEXT: ld.hu $a0, $a0, 0
+; LA64-LAMCAS-NEXT: .p2align 4, , 16
+; LA64-LAMCAS-NEXT: .LBB137_1: # %atomicrmw.start
+; LA64-LAMCAS-NEXT: # =>This Inner Loop Header: Depth=1
+; LA64-LAMCAS-NEXT: or $a3, $a0, $a1
+; LA64-LAMCAS-NEXT: ext.w.h $a4, $a0
+; LA64-LAMCAS-NEXT: amcas.h $a0, $a3, $a2
+; LA64-LAMCAS-NEXT: bne $a0, $a4, .LBB137_1
+; LA64-LAMCAS-NEXT: # %bb.2: # %atomicrmw.end
+; LA64-LAMCAS-NEXT: ret
+ %1 = atomicrmw or ptr %a, i16 %b monotonic
+ ret i16 %1
+
+}
+
+define i8 @atomicrmw_xor_i8_monotonic(ptr %a, i8 %b) nounwind {
+; LA64-LABEL: atomicrmw_xor_i8_monotonic:
+; LA64: # %bb.0:
+; LA64-NEXT: slli.d $a2, $a0, 3
+; LA64-NEXT: bstrins.d $a0, $zero, 1, 0
+; LA64-NEXT: andi $a1, $a1, 255
+; LA64-NEXT: sll.w $a1, $a1, $a2
+; LA64-NEXT: amxor.w $a3, $a1, $a0
+; LA64-NEXT: srl.w $a0, $a3, $a2
+; LA64-NEXT: ret
+;
+; LA64-LAMCAS-LABEL: atomicrmw_xor_i8_monotonic:
+; LA64-LAMCAS: # %bb.0:
+; LA64-LAMCAS-NEXT: move $a2, $a0
+; LA64-LAMCAS-NEXT: ld.bu $a0, $a0, 0
+; LA64-LAMCAS-NEXT: .p2align 4, , 16
+; LA64-LAMCAS-NEXT: .LBB138_1: # %atomicrmw.start
+; LA64-LAMCAS-NEXT: # =>This Inner Loop Header: Depth=1
+; LA64-LAMCAS-NEXT: xor $a3, $a0, $a1
+; LA64-LAMCAS-NEXT: ext.w.b $a4, $a0
+; LA64-LAMCAS-NEXT: amcas.b $a0, $a3, $a2
+; LA64-LAMCAS-NEXT: bne $a0, $a4, .LBB138_1
+; LA64-LAMCAS-NEXT: # %bb.2: # %atomicrmw.end
+; LA64-LAMCAS-NEXT: ret
+ %1 = atomicrmw xor ptr %a, i8 %b monotonic
+ ret i8 %1
+}
+
+define i16 @atomicrmw_xor_i16_monotonic(ptr %a, i16 %b) nounwind {
+; LA64-LABEL: atomicrmw_xor_i16_monotonic:
+; LA64: # %bb.0:
+; LA64-NEXT: slli.d $a2, $a0, 3
+; LA64-NEXT: bstrins.d $a0, $zero, 1, 0
+; LA64-NEXT: bstrpick.d $a1, $a1, 15, 0
+; LA64-NEXT: sll.w $a1, $a1, $a2
+; LA64-NEXT: amxor.w $a3, $a1, $a0
+; LA64-NEXT: srl.w $a0, $a3, $a2
+; LA64-NEXT: ret
+;
+; LA64-LAMCAS-LABEL: atomicrmw_xor_i16_monotonic:
+; LA64-LAMCAS: # %bb.0:
+; LA64-LAMCAS-NEXT: move $a2, $a0
+; LA64-LAMCAS-NEXT: ld.hu $a0, $a0, 0
+; LA64-LAMCAS-NEXT: .p2align 4, , 16
+; LA64-LAMCAS-NEXT: .LBB139_1: # %atomicrmw.start
+; LA64-LAMCAS-NEXT: # =>This Inner Loop Header: Depth=1
+; LA64-LAMCAS-NEXT: xor $a3, $a0, $a1
+; LA64-LAMCAS-NEXT: ext.w.h $a4, $a0
+; LA64-LAMCAS-NEXT: amcas.h $a0, $a3, $a2
+; LA64-LAMCAS-NEXT: bne $a0, $a4, .LBB139_1
+; LA64-LAMCAS-NEXT: # %bb.2: # %atomicrmw.end
+; LA64-LAMCAS-NEXT: ret
+ %1 = atomicrmw xor ptr %a, i16 %b monotonic
+ ret i16 %1
+
+}
>From 4e088ed61270de65acc53025b45ad0e4fb4c4aa5 Mon Sep 17 00:00:00 2001
From: tangaac <tangyan01 at loongson.cn>
Date: Mon, 18 Nov 2024 12:58:44 +0800
Subject: [PATCH 2/3] implement emitExpandAtomicRMW to custom expand i8/i16
atomicrmw and/or/xor
---
.../LoongArch/LoongArchISelLowering.cpp | 63 +-
.../Target/LoongArch/LoongArchISelLowering.h | 1 +
.../ir-instruction/atomicrmw-lamcas.ll | 2294 ++++++++++++++---
3 files changed, 2025 insertions(+), 333 deletions(-)
diff --git a/llvm/lib/Target/LoongArch/LoongArchISelLowering.cpp b/llvm/lib/Target/LoongArch/LoongArchISelLowering.cpp
index b64c6dc2165577..da4b8353ca2b01 100644
--- a/llvm/lib/Target/LoongArch/LoongArchISelLowering.cpp
+++ b/llvm/lib/Target/LoongArch/LoongArchISelLowering.cpp
@@ -5748,6 +5748,58 @@ bool LoongArchTargetLowering::getTgtMemIntrinsic(IntrinsicInfo &Info,
}
}
+void LoongArchTargetLowering::emitExpandAtomicRMW(AtomicRMWInst *AI) const {
+ AtomicRMWInst::BinOp Op = AI->getOperation();
+
+ assert((Op == AtomicRMWInst::Or || Op == AtomicRMWInst::Xor ||
+ Op == AtomicRMWInst::And) &&
+ "Unable to expand");
+ unsigned MinWordSize = 4;
+
+ IRBuilder<> Builder(AI);
+ LLVMContext &Ctx = Builder.getContext();
+ const DataLayout &DL = AI->getDataLayout();
+ Type *ValueType = AI->getType();
+ Type *WordType = Type::getIntNTy(Ctx, MinWordSize * 8);
+
+ Value *Addr = AI->getPointerOperand();
+ PointerType *PtrTy = cast<PointerType>(Addr->getType());
+ IntegerType *IntTy = DL.getIndexType(Ctx, PtrTy->getAddressSpace());
+
+ Value *AlignedAddr = Builder.CreateIntrinsic(
+ Intrinsic::ptrmask, {PtrTy, IntTy},
+ {Addr, ConstantInt::get(IntTy, ~(uint64_t)(MinWordSize - 1))}, nullptr,
+ "AlignedAddr");
+
+ Value *AddrInt = Builder.CreatePtrToInt(Addr, IntTy);
+ Value *PtrLSB = Builder.CreateAnd(AddrInt, MinWordSize - 1, "PtrLSB");
+ Value *ShiftAmt = Builder.CreateShl(PtrLSB, 3);
+ ShiftAmt = Builder.CreateTrunc(ShiftAmt, WordType, "ShiftAmt");
+ Value *Mask = Builder.CreateShl(
+ ConstantInt::get(WordType,
+ (1 << (DL.getTypeStoreSize(ValueType) * 8)) - 1),
+ ShiftAmt, "Mask");
+ Value *Inv_Mask = Builder.CreateNot(Mask, "Inv_Mask");
+ Value *ValOperand_Shifted =
+ Builder.CreateShl(Builder.CreateZExt(AI->getValOperand(), WordType),
+ ShiftAmt, "ValOperand_Shifted");
+ Value *NewOperand;
+ if (Op == AtomicRMWInst::And)
+ NewOperand = Builder.CreateOr(ValOperand_Shifted, Inv_Mask, "AndOperand");
+ else
+ NewOperand = ValOperand_Shifted;
+
+ AtomicRMWInst *NewAI =
+ Builder.CreateAtomicRMW(Op, AlignedAddr, NewOperand, Align(MinWordSize),
+ AI->getOrdering(), AI->getSyncScopeID());
+
+ Value *Shift = Builder.CreateLShr(NewAI, ShiftAmt, "shifted");
+ Value *Trunc = Builder.CreateTrunc(Shift, ValueType, "extracted");
+ Value *FinalOldResult = Builder.CreateBitCast(Trunc, ValueType);
+ AI->replaceAllUsesWith(FinalOldResult);
+ AI->eraseFromParent();
+}
+
TargetLowering::AtomicExpansionKind
LoongArchTargetLowering::shouldExpandAtomicRMWInIR(AtomicRMWInst *AI) const {
// TODO: Add more AtomicRMWInst that needs to be extended.
@@ -5769,9 +5821,14 @@ LoongArchTargetLowering::shouldExpandAtomicRMWInIR(AtomicRMWInst *AI) const {
}
unsigned Size = AI->getType()->getPrimitiveSizeInBits();
- if (Subtarget.hasLAMCAS() &&
- (AI->getOperation() == AtomicRMWInst::Nand || Size < 32))
- return AtomicExpansionKind::CmpXChg;
+ if (Subtarget.hasLAMCAS()) {
+ if (Size < 32 && (AI->getOperation() == AtomicRMWInst::And ||
+ AI->getOperation() == AtomicRMWInst::Or ||
+ AI->getOperation() == AtomicRMWInst::Xor))
+ return AtomicExpansionKind::Expand;
+ if ((AI->getOperation() == AtomicRMWInst::Nand || Size < 32))
+ return AtomicExpansionKind::CmpXChg;
+ }
if (Size == 8 || Size == 16)
return AtomicExpansionKind::MaskedIntrinsic;
diff --git a/llvm/lib/Target/LoongArch/LoongArchISelLowering.h b/llvm/lib/Target/LoongArch/LoongArchISelLowering.h
index 1aa686695b49b8..da5892e1ab1252 100644
--- a/llvm/lib/Target/LoongArch/LoongArchISelLowering.h
+++ b/llvm/lib/Target/LoongArch/LoongArchISelLowering.h
@@ -192,6 +192,7 @@ class LoongArchTargetLowering : public TargetLowering {
bool hasAndNot(SDValue Y) const override;
TargetLowering::AtomicExpansionKind
shouldExpandAtomicRMWInIR(AtomicRMWInst *AI) const override;
+ void emitExpandAtomicRMW(AtomicRMWInst *AI) const override;
Value *emitMaskedAtomicRMWIntrinsic(IRBuilderBase &Builder, AtomicRMWInst *AI,
Value *AlignedAddr, Value *Incr,
diff --git a/llvm/test/CodeGen/LoongArch/ir-instruction/atomicrmw-lamcas.ll b/llvm/test/CodeGen/LoongArch/ir-instruction/atomicrmw-lamcas.ll
index 2f677038e1db37..728dd778bdf240 100644
--- a/llvm/test/CodeGen/LoongArch/ir-instruction/atomicrmw-lamcas.ll
+++ b/llvm/test/CodeGen/LoongArch/ir-instruction/atomicrmw-lamcas.ll
@@ -1,6 +1,10 @@
; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py
; RUN: llc --mtriple=loongarch64 -mattr=+d,-lamcas < %s | FileCheck %s --check-prefix=LA64
; RUN: llc --mtriple=loongarch64 -mattr=+d,+lamcas < %s | FileCheck %s --check-prefix=LA64-LAMCAS
+; RUN: llc --mtriple=loongarch64 -mattr=+d,+lamcas,+lam-bh < %s | FileCheck %s --check-prefix=LA64-LAMCAS-LAM-BH
+
+; i8/i16 atomicrmw and/or/xor shouldn't use amcas[_db].b/h to expand
+; lam-bh precedes lamcas
define i8 @atomicrmw_xchg_i8_acquire(ptr %a, i8 %b) nounwind {
; LA64-LABEL: atomicrmw_xchg_i8_acquire:
@@ -35,6 +39,12 @@ define i8 @atomicrmw_xchg_i8_acquire(ptr %a, i8 %b) nounwind {
; LA64-LAMCAS-NEXT: # %bb.2: # %atomicrmw.end
; LA64-LAMCAS-NEXT: move $a0, $a2
; LA64-LAMCAS-NEXT: ret
+;
+; LA64-LAMCAS-LAM-BH-LABEL: atomicrmw_xchg_i8_acquire:
+; LA64-LAMCAS-LAM-BH: # %bb.0:
+; LA64-LAMCAS-LAM-BH-NEXT: amswap_db.b $a2, $a1, $a0
+; LA64-LAMCAS-LAM-BH-NEXT: move $a0, $a2
+; LA64-LAMCAS-LAM-BH-NEXT: ret
%1 = atomicrmw xchg ptr %a, i8 %b acquire
ret i8 %1
}
@@ -63,6 +73,12 @@ define i8 @atomicrmw_xchg_0_i8_acquire(ptr %a) nounwind {
; LA64-LAMCAS-NEXT: bne $a0, $a2, .LBB1_1
; LA64-LAMCAS-NEXT: # %bb.2: # %atomicrmw.end
; LA64-LAMCAS-NEXT: ret
+;
+; LA64-LAMCAS-LAM-BH-LABEL: atomicrmw_xchg_0_i8_acquire:
+; LA64-LAMCAS-LAM-BH: # %bb.0:
+; LA64-LAMCAS-LAM-BH-NEXT: amswap_db.b $a1, $zero, $a0
+; LA64-LAMCAS-LAM-BH-NEXT: move $a0, $a1
+; LA64-LAMCAS-LAM-BH-NEXT: ret
%1 = atomicrmw xchg ptr %a, i8 0 acquire
ret i8 %1
}
@@ -91,6 +107,13 @@ define i8 @atomicrmw_xchg_minus_1_i8_acquire(ptr %a) nounwind {
; LA64-LAMCAS-NEXT: bne $a0, $a3, .LBB2_1
; LA64-LAMCAS-NEXT: # %bb.2: # %atomicrmw.end
; LA64-LAMCAS-NEXT: ret
+;
+; LA64-LAMCAS-LAM-BH-LABEL: atomicrmw_xchg_minus_1_i8_acquire:
+; LA64-LAMCAS-LAM-BH: # %bb.0:
+; LA64-LAMCAS-LAM-BH-NEXT: addi.w $a2, $zero, -1
+; LA64-LAMCAS-LAM-BH-NEXT: amswap_db.b $a1, $a2, $a0
+; LA64-LAMCAS-LAM-BH-NEXT: move $a0, $a1
+; LA64-LAMCAS-LAM-BH-NEXT: ret
%1 = atomicrmw xchg ptr %a, i8 -1 acquire
ret i8 %1
}
@@ -129,6 +152,12 @@ define i16 @atomicrmw_xchg_i16_acquire(ptr %a, i16 %b) nounwind {
; LA64-LAMCAS-NEXT: # %bb.2: # %atomicrmw.end
; LA64-LAMCAS-NEXT: move $a0, $a2
; LA64-LAMCAS-NEXT: ret
+;
+; LA64-LAMCAS-LAM-BH-LABEL: atomicrmw_xchg_i16_acquire:
+; LA64-LAMCAS-LAM-BH: # %bb.0:
+; LA64-LAMCAS-LAM-BH-NEXT: amswap_db.h $a2, $a1, $a0
+; LA64-LAMCAS-LAM-BH-NEXT: move $a0, $a2
+; LA64-LAMCAS-LAM-BH-NEXT: ret
%1 = atomicrmw xchg ptr %a, i16 %b acquire
ret i16 %1
}
@@ -158,6 +187,12 @@ define i16 @atomicrmw_xchg_0_i16_acquire(ptr %a) nounwind {
; LA64-LAMCAS-NEXT: bne $a0, $a2, .LBB4_1
; LA64-LAMCAS-NEXT: # %bb.2: # %atomicrmw.end
; LA64-LAMCAS-NEXT: ret
+;
+; LA64-LAMCAS-LAM-BH-LABEL: atomicrmw_xchg_0_i16_acquire:
+; LA64-LAMCAS-LAM-BH: # %bb.0:
+; LA64-LAMCAS-LAM-BH-NEXT: amswap_db.h $a1, $zero, $a0
+; LA64-LAMCAS-LAM-BH-NEXT: move $a0, $a1
+; LA64-LAMCAS-LAM-BH-NEXT: ret
%1 = atomicrmw xchg ptr %a, i16 0 acquire
ret i16 %1
}
@@ -187,6 +222,13 @@ define i16 @atomicrmw_xchg_minus_1_i16_acquire(ptr %a) nounwind {
; LA64-LAMCAS-NEXT: bne $a0, $a3, .LBB5_1
; LA64-LAMCAS-NEXT: # %bb.2: # %atomicrmw.end
; LA64-LAMCAS-NEXT: ret
+;
+; LA64-LAMCAS-LAM-BH-LABEL: atomicrmw_xchg_minus_1_i16_acquire:
+; LA64-LAMCAS-LAM-BH: # %bb.0:
+; LA64-LAMCAS-LAM-BH-NEXT: addi.w $a2, $zero, -1
+; LA64-LAMCAS-LAM-BH-NEXT: amswap_db.h $a1, $a2, $a0
+; LA64-LAMCAS-LAM-BH-NEXT: move $a0, $a1
+; LA64-LAMCAS-LAM-BH-NEXT: ret
%1 = atomicrmw xchg ptr %a, i16 -1 acquire
ret i16 %1
@@ -226,6 +268,12 @@ define i8 @atomicrmw_add_i8_acquire(ptr %a, i8 %b) nounwind {
; LA64-LAMCAS-NEXT: bne $a0, $a4, .LBB6_1
; LA64-LAMCAS-NEXT: # %bb.2: # %atomicrmw.end
; LA64-LAMCAS-NEXT: ret
+;
+; LA64-LAMCAS-LAM-BH-LABEL: atomicrmw_add_i8_acquire:
+; LA64-LAMCAS-LAM-BH: # %bb.0:
+; LA64-LAMCAS-LAM-BH-NEXT: amadd_db.b $a2, $a1, $a0
+; LA64-LAMCAS-LAM-BH-NEXT: move $a0, $a2
+; LA64-LAMCAS-LAM-BH-NEXT: ret
%1 = atomicrmw add ptr %a, i8 %b acquire
ret i8 %1
}
@@ -265,6 +313,12 @@ define i16 @atomicrmw_add_i16_acquire(ptr %a, i16 %b) nounwind {
; LA64-LAMCAS-NEXT: bne $a0, $a4, .LBB7_1
; LA64-LAMCAS-NEXT: # %bb.2: # %atomicrmw.end
; LA64-LAMCAS-NEXT: ret
+;
+; LA64-LAMCAS-LAM-BH-LABEL: atomicrmw_add_i16_acquire:
+; LA64-LAMCAS-LAM-BH: # %bb.0:
+; LA64-LAMCAS-LAM-BH-NEXT: amadd_db.h $a2, $a1, $a0
+; LA64-LAMCAS-LAM-BH-NEXT: move $a0, $a2
+; LA64-LAMCAS-LAM-BH-NEXT: ret
%1 = atomicrmw add ptr %a, i16 %b acquire
ret i16 %1
@@ -304,6 +358,13 @@ define i8 @atomicrmw_sub_i8_acquire(ptr %a, i8 %b) nounwind {
; LA64-LAMCAS-NEXT: bne $a0, $a4, .LBB8_1
; LA64-LAMCAS-NEXT: # %bb.2: # %atomicrmw.end
; LA64-LAMCAS-NEXT: ret
+;
+; LA64-LAMCAS-LAM-BH-LABEL: atomicrmw_sub_i8_acquire:
+; LA64-LAMCAS-LAM-BH: # %bb.0:
+; LA64-LAMCAS-LAM-BH-NEXT: sub.w $a2, $zero, $a1
+; LA64-LAMCAS-LAM-BH-NEXT: amadd_db.b $a1, $a2, $a0
+; LA64-LAMCAS-LAM-BH-NEXT: move $a0, $a1
+; LA64-LAMCAS-LAM-BH-NEXT: ret
%1 = atomicrmw sub ptr %a, i8 %b acquire
ret i8 %1
}
@@ -343,6 +404,13 @@ define i16 @atomicrmw_sub_i16_acquire(ptr %a, i16 %b) nounwind {
; LA64-LAMCAS-NEXT: bne $a0, $a4, .LBB9_1
; LA64-LAMCAS-NEXT: # %bb.2: # %atomicrmw.end
; LA64-LAMCAS-NEXT: ret
+;
+; LA64-LAMCAS-LAM-BH-LABEL: atomicrmw_sub_i16_acquire:
+; LA64-LAMCAS-LAM-BH: # %bb.0:
+; LA64-LAMCAS-LAM-BH-NEXT: sub.w $a2, $zero, $a1
+; LA64-LAMCAS-LAM-BH-NEXT: amadd_db.h $a1, $a2, $a0
+; LA64-LAMCAS-LAM-BH-NEXT: move $a0, $a1
+; LA64-LAMCAS-LAM-BH-NEXT: ret
%1 = atomicrmw sub ptr %a, i16 %b acquire
ret i16 %1
@@ -391,6 +459,25 @@ define i8 @atomicrmw_umax_i8_acquire(ptr %a, i8 %b) nounwind {
; LA64-LAMCAS-NEXT: bne $a0, $a5, .LBB10_1
; LA64-LAMCAS-NEXT: # %bb.2: # %atomicrmw.end
; LA64-LAMCAS-NEXT: ret
+;
+; LA64-LAMCAS-LAM-BH-LABEL: atomicrmw_umax_i8_acquire:
+; LA64-LAMCAS-LAM-BH: # %bb.0:
+; LA64-LAMCAS-LAM-BH-NEXT: move $a2, $a0
+; LA64-LAMCAS-LAM-BH-NEXT: ld.bu $a0, $a0, 0
+; LA64-LAMCAS-LAM-BH-NEXT: andi $a3, $a1, 255
+; LA64-LAMCAS-LAM-BH-NEXT: .p2align 4, , 16
+; LA64-LAMCAS-LAM-BH-NEXT: .LBB10_1: # %atomicrmw.start
+; LA64-LAMCAS-LAM-BH-NEXT: # =>This Inner Loop Header: Depth=1
+; LA64-LAMCAS-LAM-BH-NEXT: andi $a4, $a0, 255
+; LA64-LAMCAS-LAM-BH-NEXT: sltu $a4, $a3, $a4
+; LA64-LAMCAS-LAM-BH-NEXT: masknez $a5, $a1, $a4
+; LA64-LAMCAS-LAM-BH-NEXT: maskeqz $a4, $a0, $a4
+; LA64-LAMCAS-LAM-BH-NEXT: or $a4, $a4, $a5
+; LA64-LAMCAS-LAM-BH-NEXT: ext.w.b $a5, $a0
+; LA64-LAMCAS-LAM-BH-NEXT: amcas_db.b $a0, $a4, $a2
+; LA64-LAMCAS-LAM-BH-NEXT: bne $a0, $a5, .LBB10_1
+; LA64-LAMCAS-LAM-BH-NEXT: # %bb.2: # %atomicrmw.end
+; LA64-LAMCAS-LAM-BH-NEXT: ret
%1 = atomicrmw umax ptr %a, i8 %b acquire
ret i8 %1
}
@@ -439,6 +526,25 @@ define i16 @atomicrmw_umax_i16_acquire(ptr %a, i16 %b) nounwind {
; LA64-LAMCAS-NEXT: bne $a0, $a5, .LBB11_1
; LA64-LAMCAS-NEXT: # %bb.2: # %atomicrmw.end
; LA64-LAMCAS-NEXT: ret
+;
+; LA64-LAMCAS-LAM-BH-LABEL: atomicrmw_umax_i16_acquire:
+; LA64-LAMCAS-LAM-BH: # %bb.0:
+; LA64-LAMCAS-LAM-BH-NEXT: move $a2, $a0
+; LA64-LAMCAS-LAM-BH-NEXT: ld.hu $a0, $a0, 0
+; LA64-LAMCAS-LAM-BH-NEXT: bstrpick.d $a3, $a1, 15, 0
+; LA64-LAMCAS-LAM-BH-NEXT: .p2align 4, , 16
+; LA64-LAMCAS-LAM-BH-NEXT: .LBB11_1: # %atomicrmw.start
+; LA64-LAMCAS-LAM-BH-NEXT: # =>This Inner Loop Header: Depth=1
+; LA64-LAMCAS-LAM-BH-NEXT: bstrpick.d $a4, $a0, 15, 0
+; LA64-LAMCAS-LAM-BH-NEXT: sltu $a4, $a3, $a4
+; LA64-LAMCAS-LAM-BH-NEXT: masknez $a5, $a1, $a4
+; LA64-LAMCAS-LAM-BH-NEXT: maskeqz $a4, $a0, $a4
+; LA64-LAMCAS-LAM-BH-NEXT: or $a4, $a4, $a5
+; LA64-LAMCAS-LAM-BH-NEXT: ext.w.h $a5, $a0
+; LA64-LAMCAS-LAM-BH-NEXT: amcas_db.h $a0, $a4, $a2
+; LA64-LAMCAS-LAM-BH-NEXT: bne $a0, $a5, .LBB11_1
+; LA64-LAMCAS-LAM-BH-NEXT: # %bb.2: # %atomicrmw.end
+; LA64-LAMCAS-LAM-BH-NEXT: ret
%1 = atomicrmw umax ptr %a, i16 %b acquire
ret i16 %1
}
@@ -487,6 +593,26 @@ define i8 @atomicrmw_umin_i8_acquire(ptr %a, i8 %b) nounwind {
; LA64-LAMCAS-NEXT: bne $a0, $a5, .LBB12_1
; LA64-LAMCAS-NEXT: # %bb.2: # %atomicrmw.end
; LA64-LAMCAS-NEXT: ret
+;
+; LA64-LAMCAS-LAM-BH-LABEL: atomicrmw_umin_i8_acquire:
+; LA64-LAMCAS-LAM-BH: # %bb.0:
+; LA64-LAMCAS-LAM-BH-NEXT: move $a2, $a0
+; LA64-LAMCAS-LAM-BH-NEXT: ld.bu $a0, $a0, 0
+; LA64-LAMCAS-LAM-BH-NEXT: andi $a3, $a1, 255
+; LA64-LAMCAS-LAM-BH-NEXT: .p2align 4, , 16
+; LA64-LAMCAS-LAM-BH-NEXT: .LBB12_1: # %atomicrmw.start
+; LA64-LAMCAS-LAM-BH-NEXT: # =>This Inner Loop Header: Depth=1
+; LA64-LAMCAS-LAM-BH-NEXT: andi $a4, $a0, 255
+; LA64-LAMCAS-LAM-BH-NEXT: sltu $a4, $a3, $a4
+; LA64-LAMCAS-LAM-BH-NEXT: xori $a4, $a4, 1
+; LA64-LAMCAS-LAM-BH-NEXT: masknez $a5, $a1, $a4
+; LA64-LAMCAS-LAM-BH-NEXT: maskeqz $a4, $a0, $a4
+; LA64-LAMCAS-LAM-BH-NEXT: or $a4, $a4, $a5
+; LA64-LAMCAS-LAM-BH-NEXT: ext.w.b $a5, $a0
+; LA64-LAMCAS-LAM-BH-NEXT: amcas_db.b $a0, $a4, $a2
+; LA64-LAMCAS-LAM-BH-NEXT: bne $a0, $a5, .LBB12_1
+; LA64-LAMCAS-LAM-BH-NEXT: # %bb.2: # %atomicrmw.end
+; LA64-LAMCAS-LAM-BH-NEXT: ret
%1 = atomicrmw umin ptr %a, i8 %b acquire
ret i8 %1
}
@@ -536,6 +662,26 @@ define i16 @atomicrmw_umin_i16_acquire(ptr %a, i16 %b) nounwind {
; LA64-LAMCAS-NEXT: bne $a0, $a5, .LBB13_1
; LA64-LAMCAS-NEXT: # %bb.2: # %atomicrmw.end
; LA64-LAMCAS-NEXT: ret
+;
+; LA64-LAMCAS-LAM-BH-LABEL: atomicrmw_umin_i16_acquire:
+; LA64-LAMCAS-LAM-BH: # %bb.0:
+; LA64-LAMCAS-LAM-BH-NEXT: move $a2, $a0
+; LA64-LAMCAS-LAM-BH-NEXT: ld.hu $a0, $a0, 0
+; LA64-LAMCAS-LAM-BH-NEXT: bstrpick.d $a3, $a1, 15, 0
+; LA64-LAMCAS-LAM-BH-NEXT: .p2align 4, , 16
+; LA64-LAMCAS-LAM-BH-NEXT: .LBB13_1: # %atomicrmw.start
+; LA64-LAMCAS-LAM-BH-NEXT: # =>This Inner Loop Header: Depth=1
+; LA64-LAMCAS-LAM-BH-NEXT: bstrpick.d $a4, $a0, 15, 0
+; LA64-LAMCAS-LAM-BH-NEXT: sltu $a4, $a3, $a4
+; LA64-LAMCAS-LAM-BH-NEXT: xori $a4, $a4, 1
+; LA64-LAMCAS-LAM-BH-NEXT: masknez $a5, $a1, $a4
+; LA64-LAMCAS-LAM-BH-NEXT: maskeqz $a4, $a0, $a4
+; LA64-LAMCAS-LAM-BH-NEXT: or $a4, $a4, $a5
+; LA64-LAMCAS-LAM-BH-NEXT: ext.w.h $a5, $a0
+; LA64-LAMCAS-LAM-BH-NEXT: amcas_db.h $a0, $a4, $a2
+; LA64-LAMCAS-LAM-BH-NEXT: bne $a0, $a5, .LBB13_1
+; LA64-LAMCAS-LAM-BH-NEXT: # %bb.2: # %atomicrmw.end
+; LA64-LAMCAS-LAM-BH-NEXT: ret
%1 = atomicrmw umin ptr %a, i16 %b acquire
ret i16 %1
}
@@ -586,6 +732,24 @@ define i8 @atomicrmw_max_i8_acquire(ptr %a, i8 %b) nounwind {
; LA64-LAMCAS-NEXT: bne $a0, $a4, .LBB14_1
; LA64-LAMCAS-NEXT: # %bb.2: # %atomicrmw.end
; LA64-LAMCAS-NEXT: ret
+;
+; LA64-LAMCAS-LAM-BH-LABEL: atomicrmw_max_i8_acquire:
+; LA64-LAMCAS-LAM-BH: # %bb.0:
+; LA64-LAMCAS-LAM-BH-NEXT: move $a2, $a0
+; LA64-LAMCAS-LAM-BH-NEXT: ld.bu $a0, $a0, 0
+; LA64-LAMCAS-LAM-BH-NEXT: ext.w.b $a3, $a1
+; LA64-LAMCAS-LAM-BH-NEXT: .p2align 4, , 16
+; LA64-LAMCAS-LAM-BH-NEXT: .LBB14_1: # %atomicrmw.start
+; LA64-LAMCAS-LAM-BH-NEXT: # =>This Inner Loop Header: Depth=1
+; LA64-LAMCAS-LAM-BH-NEXT: ext.w.b $a4, $a0
+; LA64-LAMCAS-LAM-BH-NEXT: slt $a5, $a3, $a4
+; LA64-LAMCAS-LAM-BH-NEXT: masknez $a6, $a1, $a5
+; LA64-LAMCAS-LAM-BH-NEXT: maskeqz $a5, $a0, $a5
+; LA64-LAMCAS-LAM-BH-NEXT: or $a5, $a5, $a6
+; LA64-LAMCAS-LAM-BH-NEXT: amcas_db.b $a0, $a5, $a2
+; LA64-LAMCAS-LAM-BH-NEXT: bne $a0, $a4, .LBB14_1
+; LA64-LAMCAS-LAM-BH-NEXT: # %bb.2: # %atomicrmw.end
+; LA64-LAMCAS-LAM-BH-NEXT: ret
%1 = atomicrmw max ptr %a, i8 %b acquire
ret i8 %1
}
@@ -638,6 +802,24 @@ define i16 @atomicrmw_max_i16_acquire(ptr %a, i16 %b) nounwind {
; LA64-LAMCAS-NEXT: bne $a0, $a4, .LBB15_1
; LA64-LAMCAS-NEXT: # %bb.2: # %atomicrmw.end
; LA64-LAMCAS-NEXT: ret
+;
+; LA64-LAMCAS-LAM-BH-LABEL: atomicrmw_max_i16_acquire:
+; LA64-LAMCAS-LAM-BH: # %bb.0:
+; LA64-LAMCAS-LAM-BH-NEXT: move $a2, $a0
+; LA64-LAMCAS-LAM-BH-NEXT: ld.hu $a0, $a0, 0
+; LA64-LAMCAS-LAM-BH-NEXT: ext.w.h $a3, $a1
+; LA64-LAMCAS-LAM-BH-NEXT: .p2align 4, , 16
+; LA64-LAMCAS-LAM-BH-NEXT: .LBB15_1: # %atomicrmw.start
+; LA64-LAMCAS-LAM-BH-NEXT: # =>This Inner Loop Header: Depth=1
+; LA64-LAMCAS-LAM-BH-NEXT: ext.w.h $a4, $a0
+; LA64-LAMCAS-LAM-BH-NEXT: slt $a5, $a3, $a4
+; LA64-LAMCAS-LAM-BH-NEXT: masknez $a6, $a1, $a5
+; LA64-LAMCAS-LAM-BH-NEXT: maskeqz $a5, $a0, $a5
+; LA64-LAMCAS-LAM-BH-NEXT: or $a5, $a5, $a6
+; LA64-LAMCAS-LAM-BH-NEXT: amcas_db.h $a0, $a5, $a2
+; LA64-LAMCAS-LAM-BH-NEXT: bne $a0, $a4, .LBB15_1
+; LA64-LAMCAS-LAM-BH-NEXT: # %bb.2: # %atomicrmw.end
+; LA64-LAMCAS-LAM-BH-NEXT: ret
%1 = atomicrmw max ptr %a, i16 %b acquire
ret i16 %1
}
@@ -689,6 +871,25 @@ define i8 @atomicrmw_min_i8_acquire(ptr %a, i8 %b) nounwind {
; LA64-LAMCAS-NEXT: bne $a0, $a4, .LBB16_1
; LA64-LAMCAS-NEXT: # %bb.2: # %atomicrmw.end
; LA64-LAMCAS-NEXT: ret
+;
+; LA64-LAMCAS-LAM-BH-LABEL: atomicrmw_min_i8_acquire:
+; LA64-LAMCAS-LAM-BH: # %bb.0:
+; LA64-LAMCAS-LAM-BH-NEXT: move $a2, $a0
+; LA64-LAMCAS-LAM-BH-NEXT: ld.bu $a0, $a0, 0
+; LA64-LAMCAS-LAM-BH-NEXT: ext.w.b $a3, $a1
+; LA64-LAMCAS-LAM-BH-NEXT: .p2align 4, , 16
+; LA64-LAMCAS-LAM-BH-NEXT: .LBB16_1: # %atomicrmw.start
+; LA64-LAMCAS-LAM-BH-NEXT: # =>This Inner Loop Header: Depth=1
+; LA64-LAMCAS-LAM-BH-NEXT: ext.w.b $a4, $a0
+; LA64-LAMCAS-LAM-BH-NEXT: slt $a5, $a3, $a4
+; LA64-LAMCAS-LAM-BH-NEXT: xori $a5, $a5, 1
+; LA64-LAMCAS-LAM-BH-NEXT: masknez $a6, $a1, $a5
+; LA64-LAMCAS-LAM-BH-NEXT: maskeqz $a5, $a0, $a5
+; LA64-LAMCAS-LAM-BH-NEXT: or $a5, $a5, $a6
+; LA64-LAMCAS-LAM-BH-NEXT: amcas_db.b $a0, $a5, $a2
+; LA64-LAMCAS-LAM-BH-NEXT: bne $a0, $a4, .LBB16_1
+; LA64-LAMCAS-LAM-BH-NEXT: # %bb.2: # %atomicrmw.end
+; LA64-LAMCAS-LAM-BH-NEXT: ret
%1 = atomicrmw min ptr %a, i8 %b acquire
ret i8 %1
}
@@ -742,6 +943,25 @@ define i16 @atomicrmw_min_i16_acquire(ptr %a, i16 %b) nounwind {
; LA64-LAMCAS-NEXT: bne $a0, $a4, .LBB17_1
; LA64-LAMCAS-NEXT: # %bb.2: # %atomicrmw.end
; LA64-LAMCAS-NEXT: ret
+;
+; LA64-LAMCAS-LAM-BH-LABEL: atomicrmw_min_i16_acquire:
+; LA64-LAMCAS-LAM-BH: # %bb.0:
+; LA64-LAMCAS-LAM-BH-NEXT: move $a2, $a0
+; LA64-LAMCAS-LAM-BH-NEXT: ld.hu $a0, $a0, 0
+; LA64-LAMCAS-LAM-BH-NEXT: ext.w.h $a3, $a1
+; LA64-LAMCAS-LAM-BH-NEXT: .p2align 4, , 16
+; LA64-LAMCAS-LAM-BH-NEXT: .LBB17_1: # %atomicrmw.start
+; LA64-LAMCAS-LAM-BH-NEXT: # =>This Inner Loop Header: Depth=1
+; LA64-LAMCAS-LAM-BH-NEXT: ext.w.h $a4, $a0
+; LA64-LAMCAS-LAM-BH-NEXT: slt $a5, $a3, $a4
+; LA64-LAMCAS-LAM-BH-NEXT: xori $a5, $a5, 1
+; LA64-LAMCAS-LAM-BH-NEXT: masknez $a6, $a1, $a5
+; LA64-LAMCAS-LAM-BH-NEXT: maskeqz $a5, $a0, $a5
+; LA64-LAMCAS-LAM-BH-NEXT: or $a5, $a5, $a6
+; LA64-LAMCAS-LAM-BH-NEXT: amcas_db.h $a0, $a5, $a2
+; LA64-LAMCAS-LAM-BH-NEXT: bne $a0, $a4, .LBB17_1
+; LA64-LAMCAS-LAM-BH-NEXT: # %bb.2: # %atomicrmw.end
+; LA64-LAMCAS-LAM-BH-NEXT: ret
%1 = atomicrmw min ptr %a, i16 %b acquire
ret i16 %1
}
@@ -782,6 +1002,21 @@ define i8 @atomicrmw_nand_i8_acquire(ptr %a, i8 %b) nounwind {
; LA64-LAMCAS-NEXT: bne $a0, $a4, .LBB18_1
; LA64-LAMCAS-NEXT: # %bb.2: # %atomicrmw.end
; LA64-LAMCAS-NEXT: ret
+;
+; LA64-LAMCAS-LAM-BH-LABEL: atomicrmw_nand_i8_acquire:
+; LA64-LAMCAS-LAM-BH: # %bb.0:
+; LA64-LAMCAS-LAM-BH-NEXT: move $a2, $a0
+; LA64-LAMCAS-LAM-BH-NEXT: ld.bu $a0, $a0, 0
+; LA64-LAMCAS-LAM-BH-NEXT: .p2align 4, , 16
+; LA64-LAMCAS-LAM-BH-NEXT: .LBB18_1: # %atomicrmw.start
+; LA64-LAMCAS-LAM-BH-NEXT: # =>This Inner Loop Header: Depth=1
+; LA64-LAMCAS-LAM-BH-NEXT: and $a3, $a0, $a1
+; LA64-LAMCAS-LAM-BH-NEXT: nor $a3, $a3, $zero
+; LA64-LAMCAS-LAM-BH-NEXT: ext.w.b $a4, $a0
+; LA64-LAMCAS-LAM-BH-NEXT: amcas_db.b $a0, $a3, $a2
+; LA64-LAMCAS-LAM-BH-NEXT: bne $a0, $a4, .LBB18_1
+; LA64-LAMCAS-LAM-BH-NEXT: # %bb.2: # %atomicrmw.end
+; LA64-LAMCAS-LAM-BH-NEXT: ret
%1 = atomicrmw nand ptr %a, i8 %b acquire
ret i8 %1
}
@@ -823,6 +1058,21 @@ define i16 @atomicrmw_nand_i16_acquire(ptr %a, i16 %b) nounwind {
; LA64-LAMCAS-NEXT: bne $a0, $a4, .LBB19_1
; LA64-LAMCAS-NEXT: # %bb.2: # %atomicrmw.end
; LA64-LAMCAS-NEXT: ret
+;
+; LA64-LAMCAS-LAM-BH-LABEL: atomicrmw_nand_i16_acquire:
+; LA64-LAMCAS-LAM-BH: # %bb.0:
+; LA64-LAMCAS-LAM-BH-NEXT: move $a2, $a0
+; LA64-LAMCAS-LAM-BH-NEXT: ld.hu $a0, $a0, 0
+; LA64-LAMCAS-LAM-BH-NEXT: .p2align 4, , 16
+; LA64-LAMCAS-LAM-BH-NEXT: .LBB19_1: # %atomicrmw.start
+; LA64-LAMCAS-LAM-BH-NEXT: # =>This Inner Loop Header: Depth=1
+; LA64-LAMCAS-LAM-BH-NEXT: and $a3, $a0, $a1
+; LA64-LAMCAS-LAM-BH-NEXT: nor $a3, $a3, $zero
+; LA64-LAMCAS-LAM-BH-NEXT: ext.w.h $a4, $a0
+; LA64-LAMCAS-LAM-BH-NEXT: amcas_db.h $a0, $a3, $a2
+; LA64-LAMCAS-LAM-BH-NEXT: bne $a0, $a4, .LBB19_1
+; LA64-LAMCAS-LAM-BH-NEXT: # %bb.2: # %atomicrmw.end
+; LA64-LAMCAS-LAM-BH-NEXT: ret
%1 = atomicrmw nand ptr %a, i16 %b acquire
ret i16 %1
@@ -855,6 +1105,21 @@ define i32 @atomicrmw_nand_i32_acquire(ptr %a, i32 %b) nounwind {
; LA64-LAMCAS-NEXT: bne $a0, $a3, .LBB20_1
; LA64-LAMCAS-NEXT: # %bb.2: # %atomicrmw.end
; LA64-LAMCAS-NEXT: ret
+;
+; LA64-LAMCAS-LAM-BH-LABEL: atomicrmw_nand_i32_acquire:
+; LA64-LAMCAS-LAM-BH: # %bb.0:
+; LA64-LAMCAS-LAM-BH-NEXT: move $a2, $a0
+; LA64-LAMCAS-LAM-BH-NEXT: ld.w $a0, $a0, 0
+; LA64-LAMCAS-LAM-BH-NEXT: .p2align 4, , 16
+; LA64-LAMCAS-LAM-BH-NEXT: .LBB20_1: # %atomicrmw.start
+; LA64-LAMCAS-LAM-BH-NEXT: # =>This Inner Loop Header: Depth=1
+; LA64-LAMCAS-LAM-BH-NEXT: move $a3, $a0
+; LA64-LAMCAS-LAM-BH-NEXT: and $a4, $a0, $a1
+; LA64-LAMCAS-LAM-BH-NEXT: nor $a4, $a4, $zero
+; LA64-LAMCAS-LAM-BH-NEXT: amcas_db.w $a0, $a4, $a2
+; LA64-LAMCAS-LAM-BH-NEXT: bne $a0, $a3, .LBB20_1
+; LA64-LAMCAS-LAM-BH-NEXT: # %bb.2: # %atomicrmw.end
+; LA64-LAMCAS-LAM-BH-NEXT: ret
%1 = atomicrmw nand ptr %a, i32 %b acquire
ret i32 %1
}
@@ -886,6 +1151,21 @@ define i64 @atomicrmw_nand_i64_acquire(ptr %a, i64 %b) nounwind {
; LA64-LAMCAS-NEXT: bne $a0, $a3, .LBB21_1
; LA64-LAMCAS-NEXT: # %bb.2: # %atomicrmw.end
; LA64-LAMCAS-NEXT: ret
+;
+; LA64-LAMCAS-LAM-BH-LABEL: atomicrmw_nand_i64_acquire:
+; LA64-LAMCAS-LAM-BH: # %bb.0:
+; LA64-LAMCAS-LAM-BH-NEXT: move $a2, $a0
+; LA64-LAMCAS-LAM-BH-NEXT: ld.d $a0, $a0, 0
+; LA64-LAMCAS-LAM-BH-NEXT: .p2align 4, , 16
+; LA64-LAMCAS-LAM-BH-NEXT: .LBB21_1: # %atomicrmw.start
+; LA64-LAMCAS-LAM-BH-NEXT: # =>This Inner Loop Header: Depth=1
+; LA64-LAMCAS-LAM-BH-NEXT: move $a3, $a0
+; LA64-LAMCAS-LAM-BH-NEXT: and $a4, $a0, $a1
+; LA64-LAMCAS-LAM-BH-NEXT: nor $a4, $a4, $zero
+; LA64-LAMCAS-LAM-BH-NEXT: amcas_db.d $a0, $a4, $a2
+; LA64-LAMCAS-LAM-BH-NEXT: bne $a0, $a3, .LBB21_1
+; LA64-LAMCAS-LAM-BH-NEXT: # %bb.2: # %atomicrmw.end
+; LA64-LAMCAS-LAM-BH-NEXT: ret
%1 = atomicrmw nand ptr %a, i64 %b acquire
ret i64 %1
}
@@ -906,17 +1186,29 @@ define i8 @atomicrmw_and_i8_acquire(ptr %a, i8 %b) nounwind {
;
; LA64-LAMCAS-LABEL: atomicrmw_and_i8_acquire:
; LA64-LAMCAS: # %bb.0:
-; LA64-LAMCAS-NEXT: move $a2, $a0
-; LA64-LAMCAS-NEXT: ld.bu $a0, $a0, 0
-; LA64-LAMCAS-NEXT: .p2align 4, , 16
-; LA64-LAMCAS-NEXT: .LBB22_1: # %atomicrmw.start
-; LA64-LAMCAS-NEXT: # =>This Inner Loop Header: Depth=1
-; LA64-LAMCAS-NEXT: and $a3, $a0, $a1
-; LA64-LAMCAS-NEXT: ext.w.b $a4, $a0
-; LA64-LAMCAS-NEXT: amcas_db.b $a0, $a3, $a2
-; LA64-LAMCAS-NEXT: bne $a0, $a4, .LBB22_1
-; LA64-LAMCAS-NEXT: # %bb.2: # %atomicrmw.end
-; LA64-LAMCAS-NEXT: ret
+; LA64-LAMCAS-NEXT: slli.d $a2, $a0, 3
+; LA64-LAMCAS-NEXT: bstrins.d $a0, $zero, 1, 0
+; LA64-LAMCAS-NEXT: ori $a3, $zero, 255
+; LA64-LAMCAS-NEXT: sll.w $a3, $a3, $a2
+; LA64-LAMCAS-NEXT: andi $a1, $a1, 255
+; LA64-LAMCAS-NEXT: sll.w $a1, $a1, $a2
+; LA64-LAMCAS-NEXT: orn $a1, $a1, $a3
+; LA64-LAMCAS-NEXT: amand_db.w $a3, $a1, $a0
+; LA64-LAMCAS-NEXT: srl.w $a0, $a3, $a2
+; LA64-LAMCAS-NEXT: ret
+;
+; LA64-LAMCAS-LAM-BH-LABEL: atomicrmw_and_i8_acquire:
+; LA64-LAMCAS-LAM-BH: # %bb.0:
+; LA64-LAMCAS-LAM-BH-NEXT: slli.d $a2, $a0, 3
+; LA64-LAMCAS-LAM-BH-NEXT: bstrins.d $a0, $zero, 1, 0
+; LA64-LAMCAS-LAM-BH-NEXT: ori $a3, $zero, 255
+; LA64-LAMCAS-LAM-BH-NEXT: sll.w $a3, $a3, $a2
+; LA64-LAMCAS-LAM-BH-NEXT: andi $a1, $a1, 255
+; LA64-LAMCAS-LAM-BH-NEXT: sll.w $a1, $a1, $a2
+; LA64-LAMCAS-LAM-BH-NEXT: orn $a1, $a1, $a3
+; LA64-LAMCAS-LAM-BH-NEXT: amand_db.w $a3, $a1, $a0
+; LA64-LAMCAS-LAM-BH-NEXT: srl.w $a0, $a3, $a2
+; LA64-LAMCAS-LAM-BH-NEXT: ret
%1 = atomicrmw and ptr %a, i8 %b acquire
ret i8 %1
}
@@ -938,17 +1230,31 @@ define i16 @atomicrmw_and_i16_acquire(ptr %a, i16 %b) nounwind {
;
; LA64-LAMCAS-LABEL: atomicrmw_and_i16_acquire:
; LA64-LAMCAS: # %bb.0:
-; LA64-LAMCAS-NEXT: move $a2, $a0
-; LA64-LAMCAS-NEXT: ld.hu $a0, $a0, 0
-; LA64-LAMCAS-NEXT: .p2align 4, , 16
-; LA64-LAMCAS-NEXT: .LBB23_1: # %atomicrmw.start
-; LA64-LAMCAS-NEXT: # =>This Inner Loop Header: Depth=1
-; LA64-LAMCAS-NEXT: and $a3, $a0, $a1
-; LA64-LAMCAS-NEXT: ext.w.h $a4, $a0
-; LA64-LAMCAS-NEXT: amcas_db.h $a0, $a3, $a2
-; LA64-LAMCAS-NEXT: bne $a0, $a4, .LBB23_1
-; LA64-LAMCAS-NEXT: # %bb.2: # %atomicrmw.end
-; LA64-LAMCAS-NEXT: ret
+; LA64-LAMCAS-NEXT: slli.d $a2, $a0, 3
+; LA64-LAMCAS-NEXT: bstrins.d $a0, $zero, 1, 0
+; LA64-LAMCAS-NEXT: lu12i.w $a3, 15
+; LA64-LAMCAS-NEXT: ori $a3, $a3, 4095
+; LA64-LAMCAS-NEXT: sll.w $a3, $a3, $a2
+; LA64-LAMCAS-NEXT: bstrpick.d $a1, $a1, 15, 0
+; LA64-LAMCAS-NEXT: sll.w $a1, $a1, $a2
+; LA64-LAMCAS-NEXT: orn $a1, $a1, $a3
+; LA64-LAMCAS-NEXT: amand_db.w $a3, $a1, $a0
+; LA64-LAMCAS-NEXT: srl.w $a0, $a3, $a2
+; LA64-LAMCAS-NEXT: ret
+;
+; LA64-LAMCAS-LAM-BH-LABEL: atomicrmw_and_i16_acquire:
+; LA64-LAMCAS-LAM-BH: # %bb.0:
+; LA64-LAMCAS-LAM-BH-NEXT: slli.d $a2, $a0, 3
+; LA64-LAMCAS-LAM-BH-NEXT: bstrins.d $a0, $zero, 1, 0
+; LA64-LAMCAS-LAM-BH-NEXT: lu12i.w $a3, 15
+; LA64-LAMCAS-LAM-BH-NEXT: ori $a3, $a3, 4095
+; LA64-LAMCAS-LAM-BH-NEXT: sll.w $a3, $a3, $a2
+; LA64-LAMCAS-LAM-BH-NEXT: bstrpick.d $a1, $a1, 15, 0
+; LA64-LAMCAS-LAM-BH-NEXT: sll.w $a1, $a1, $a2
+; LA64-LAMCAS-LAM-BH-NEXT: orn $a1, $a1, $a3
+; LA64-LAMCAS-LAM-BH-NEXT: amand_db.w $a3, $a1, $a0
+; LA64-LAMCAS-LAM-BH-NEXT: srl.w $a0, $a3, $a2
+; LA64-LAMCAS-LAM-BH-NEXT: ret
%1 = atomicrmw and ptr %a, i16 %b acquire
ret i16 %1
@@ -967,17 +1273,23 @@ define i8 @atomicrmw_or_i8_acquire(ptr %a, i8 %b) nounwind {
;
; LA64-LAMCAS-LABEL: atomicrmw_or_i8_acquire:
; LA64-LAMCAS: # %bb.0:
-; LA64-LAMCAS-NEXT: move $a2, $a0
-; LA64-LAMCAS-NEXT: ld.bu $a0, $a0, 0
-; LA64-LAMCAS-NEXT: .p2align 4, , 16
-; LA64-LAMCAS-NEXT: .LBB24_1: # %atomicrmw.start
-; LA64-LAMCAS-NEXT: # =>This Inner Loop Header: Depth=1
-; LA64-LAMCAS-NEXT: or $a3, $a0, $a1
-; LA64-LAMCAS-NEXT: ext.w.b $a4, $a0
-; LA64-LAMCAS-NEXT: amcas_db.b $a0, $a3, $a2
-; LA64-LAMCAS-NEXT: bne $a0, $a4, .LBB24_1
-; LA64-LAMCAS-NEXT: # %bb.2: # %atomicrmw.end
-; LA64-LAMCAS-NEXT: ret
+; LA64-LAMCAS-NEXT: slli.d $a2, $a0, 3
+; LA64-LAMCAS-NEXT: bstrins.d $a0, $zero, 1, 0
+; LA64-LAMCAS-NEXT: andi $a1, $a1, 255
+; LA64-LAMCAS-NEXT: sll.w $a1, $a1, $a2
+; LA64-LAMCAS-NEXT: amor_db.w $a3, $a1, $a0
+; LA64-LAMCAS-NEXT: srl.w $a0, $a3, $a2
+; LA64-LAMCAS-NEXT: ret
+;
+; LA64-LAMCAS-LAM-BH-LABEL: atomicrmw_or_i8_acquire:
+; LA64-LAMCAS-LAM-BH: # %bb.0:
+; LA64-LAMCAS-LAM-BH-NEXT: slli.d $a2, $a0, 3
+; LA64-LAMCAS-LAM-BH-NEXT: bstrins.d $a0, $zero, 1, 0
+; LA64-LAMCAS-LAM-BH-NEXT: andi $a1, $a1, 255
+; LA64-LAMCAS-LAM-BH-NEXT: sll.w $a1, $a1, $a2
+; LA64-LAMCAS-LAM-BH-NEXT: amor_db.w $a3, $a1, $a0
+; LA64-LAMCAS-LAM-BH-NEXT: srl.w $a0, $a3, $a2
+; LA64-LAMCAS-LAM-BH-NEXT: ret
%1 = atomicrmw or ptr %a, i8 %b acquire
ret i8 %1
}
@@ -995,17 +1307,23 @@ define i16 @atomicrmw_or_i16_acquire(ptr %a, i16 %b) nounwind {
;
; LA64-LAMCAS-LABEL: atomicrmw_or_i16_acquire:
; LA64-LAMCAS: # %bb.0:
-; LA64-LAMCAS-NEXT: move $a2, $a0
-; LA64-LAMCAS-NEXT: ld.hu $a0, $a0, 0
-; LA64-LAMCAS-NEXT: .p2align 4, , 16
-; LA64-LAMCAS-NEXT: .LBB25_1: # %atomicrmw.start
-; LA64-LAMCAS-NEXT: # =>This Inner Loop Header: Depth=1
-; LA64-LAMCAS-NEXT: or $a3, $a0, $a1
-; LA64-LAMCAS-NEXT: ext.w.h $a4, $a0
-; LA64-LAMCAS-NEXT: amcas_db.h $a0, $a3, $a2
-; LA64-LAMCAS-NEXT: bne $a0, $a4, .LBB25_1
-; LA64-LAMCAS-NEXT: # %bb.2: # %atomicrmw.end
-; LA64-LAMCAS-NEXT: ret
+; LA64-LAMCAS-NEXT: slli.d $a2, $a0, 3
+; LA64-LAMCAS-NEXT: bstrins.d $a0, $zero, 1, 0
+; LA64-LAMCAS-NEXT: bstrpick.d $a1, $a1, 15, 0
+; LA64-LAMCAS-NEXT: sll.w $a1, $a1, $a2
+; LA64-LAMCAS-NEXT: amor_db.w $a3, $a1, $a0
+; LA64-LAMCAS-NEXT: srl.w $a0, $a3, $a2
+; LA64-LAMCAS-NEXT: ret
+;
+; LA64-LAMCAS-LAM-BH-LABEL: atomicrmw_or_i16_acquire:
+; LA64-LAMCAS-LAM-BH: # %bb.0:
+; LA64-LAMCAS-LAM-BH-NEXT: slli.d $a2, $a0, 3
+; LA64-LAMCAS-LAM-BH-NEXT: bstrins.d $a0, $zero, 1, 0
+; LA64-LAMCAS-LAM-BH-NEXT: bstrpick.d $a1, $a1, 15, 0
+; LA64-LAMCAS-LAM-BH-NEXT: sll.w $a1, $a1, $a2
+; LA64-LAMCAS-LAM-BH-NEXT: amor_db.w $a3, $a1, $a0
+; LA64-LAMCAS-LAM-BH-NEXT: srl.w $a0, $a3, $a2
+; LA64-LAMCAS-LAM-BH-NEXT: ret
%1 = atomicrmw or ptr %a, i16 %b acquire
ret i16 %1
@@ -1024,17 +1342,23 @@ define i8 @atomicrmw_xor_i8_acquire(ptr %a, i8 %b) nounwind {
;
; LA64-LAMCAS-LABEL: atomicrmw_xor_i8_acquire:
; LA64-LAMCAS: # %bb.0:
-; LA64-LAMCAS-NEXT: move $a2, $a0
-; LA64-LAMCAS-NEXT: ld.bu $a0, $a0, 0
-; LA64-LAMCAS-NEXT: .p2align 4, , 16
-; LA64-LAMCAS-NEXT: .LBB26_1: # %atomicrmw.start
-; LA64-LAMCAS-NEXT: # =>This Inner Loop Header: Depth=1
-; LA64-LAMCAS-NEXT: xor $a3, $a0, $a1
-; LA64-LAMCAS-NEXT: ext.w.b $a4, $a0
-; LA64-LAMCAS-NEXT: amcas_db.b $a0, $a3, $a2
-; LA64-LAMCAS-NEXT: bne $a0, $a4, .LBB26_1
-; LA64-LAMCAS-NEXT: # %bb.2: # %atomicrmw.end
-; LA64-LAMCAS-NEXT: ret
+; LA64-LAMCAS-NEXT: slli.d $a2, $a0, 3
+; LA64-LAMCAS-NEXT: bstrins.d $a0, $zero, 1, 0
+; LA64-LAMCAS-NEXT: andi $a1, $a1, 255
+; LA64-LAMCAS-NEXT: sll.w $a1, $a1, $a2
+; LA64-LAMCAS-NEXT: amxor_db.w $a3, $a1, $a0
+; LA64-LAMCAS-NEXT: srl.w $a0, $a3, $a2
+; LA64-LAMCAS-NEXT: ret
+;
+; LA64-LAMCAS-LAM-BH-LABEL: atomicrmw_xor_i8_acquire:
+; LA64-LAMCAS-LAM-BH: # %bb.0:
+; LA64-LAMCAS-LAM-BH-NEXT: slli.d $a2, $a0, 3
+; LA64-LAMCAS-LAM-BH-NEXT: bstrins.d $a0, $zero, 1, 0
+; LA64-LAMCAS-LAM-BH-NEXT: andi $a1, $a1, 255
+; LA64-LAMCAS-LAM-BH-NEXT: sll.w $a1, $a1, $a2
+; LA64-LAMCAS-LAM-BH-NEXT: amxor_db.w $a3, $a1, $a0
+; LA64-LAMCAS-LAM-BH-NEXT: srl.w $a0, $a3, $a2
+; LA64-LAMCAS-LAM-BH-NEXT: ret
%1 = atomicrmw xor ptr %a, i8 %b acquire
ret i8 %1
}
@@ -1052,17 +1376,23 @@ define i16 @atomicrmw_xor_i16_acquire(ptr %a, i16 %b) nounwind {
;
; LA64-LAMCAS-LABEL: atomicrmw_xor_i16_acquire:
; LA64-LAMCAS: # %bb.0:
-; LA64-LAMCAS-NEXT: move $a2, $a0
-; LA64-LAMCAS-NEXT: ld.hu $a0, $a0, 0
-; LA64-LAMCAS-NEXT: .p2align 4, , 16
-; LA64-LAMCAS-NEXT: .LBB27_1: # %atomicrmw.start
-; LA64-LAMCAS-NEXT: # =>This Inner Loop Header: Depth=1
-; LA64-LAMCAS-NEXT: xor $a3, $a0, $a1
-; LA64-LAMCAS-NEXT: ext.w.h $a4, $a0
-; LA64-LAMCAS-NEXT: amcas_db.h $a0, $a3, $a2
-; LA64-LAMCAS-NEXT: bne $a0, $a4, .LBB27_1
-; LA64-LAMCAS-NEXT: # %bb.2: # %atomicrmw.end
-; LA64-LAMCAS-NEXT: ret
+; LA64-LAMCAS-NEXT: slli.d $a2, $a0, 3
+; LA64-LAMCAS-NEXT: bstrins.d $a0, $zero, 1, 0
+; LA64-LAMCAS-NEXT: bstrpick.d $a1, $a1, 15, 0
+; LA64-LAMCAS-NEXT: sll.w $a1, $a1, $a2
+; LA64-LAMCAS-NEXT: amxor_db.w $a3, $a1, $a0
+; LA64-LAMCAS-NEXT: srl.w $a0, $a3, $a2
+; LA64-LAMCAS-NEXT: ret
+;
+; LA64-LAMCAS-LAM-BH-LABEL: atomicrmw_xor_i16_acquire:
+; LA64-LAMCAS-LAM-BH: # %bb.0:
+; LA64-LAMCAS-LAM-BH-NEXT: slli.d $a2, $a0, 3
+; LA64-LAMCAS-LAM-BH-NEXT: bstrins.d $a0, $zero, 1, 0
+; LA64-LAMCAS-LAM-BH-NEXT: bstrpick.d $a1, $a1, 15, 0
+; LA64-LAMCAS-LAM-BH-NEXT: sll.w $a1, $a1, $a2
+; LA64-LAMCAS-LAM-BH-NEXT: amxor_db.w $a3, $a1, $a0
+; LA64-LAMCAS-LAM-BH-NEXT: srl.w $a0, $a3, $a2
+; LA64-LAMCAS-LAM-BH-NEXT: ret
%1 = atomicrmw xor ptr %a, i16 %b acquire
ret i16 %1
@@ -1101,6 +1431,12 @@ define i8 @atomicrmw_xchg_i8_release(ptr %a, i8 %b) nounwind {
; LA64-LAMCAS-NEXT: # %bb.2: # %atomicrmw.end
; LA64-LAMCAS-NEXT: move $a0, $a2
; LA64-LAMCAS-NEXT: ret
+;
+; LA64-LAMCAS-LAM-BH-LABEL: atomicrmw_xchg_i8_release:
+; LA64-LAMCAS-LAM-BH: # %bb.0:
+; LA64-LAMCAS-LAM-BH-NEXT: amswap_db.b $a2, $a1, $a0
+; LA64-LAMCAS-LAM-BH-NEXT: move $a0, $a2
+; LA64-LAMCAS-LAM-BH-NEXT: ret
%1 = atomicrmw xchg ptr %a, i8 %b release
ret i8 %1
}
@@ -1129,6 +1465,12 @@ define i8 @atomicrmw_xchg_0_i8_release(ptr %a) nounwind {
; LA64-LAMCAS-NEXT: bne $a0, $a2, .LBB29_1
; LA64-LAMCAS-NEXT: # %bb.2: # %atomicrmw.end
; LA64-LAMCAS-NEXT: ret
+;
+; LA64-LAMCAS-LAM-BH-LABEL: atomicrmw_xchg_0_i8_release:
+; LA64-LAMCAS-LAM-BH: # %bb.0:
+; LA64-LAMCAS-LAM-BH-NEXT: amswap_db.b $a1, $zero, $a0
+; LA64-LAMCAS-LAM-BH-NEXT: move $a0, $a1
+; LA64-LAMCAS-LAM-BH-NEXT: ret
%1 = atomicrmw xchg ptr %a, i8 0 release
ret i8 %1
}
@@ -1157,6 +1499,13 @@ define i8 @atomicrmw_xchg_minus_1_i8_release(ptr %a) nounwind {
; LA64-LAMCAS-NEXT: bne $a0, $a3, .LBB30_1
; LA64-LAMCAS-NEXT: # %bb.2: # %atomicrmw.end
; LA64-LAMCAS-NEXT: ret
+;
+; LA64-LAMCAS-LAM-BH-LABEL: atomicrmw_xchg_minus_1_i8_release:
+; LA64-LAMCAS-LAM-BH: # %bb.0:
+; LA64-LAMCAS-LAM-BH-NEXT: addi.w $a2, $zero, -1
+; LA64-LAMCAS-LAM-BH-NEXT: amswap_db.b $a1, $a2, $a0
+; LA64-LAMCAS-LAM-BH-NEXT: move $a0, $a1
+; LA64-LAMCAS-LAM-BH-NEXT: ret
%1 = atomicrmw xchg ptr %a, i8 -1 release
ret i8 %1
}
@@ -1195,6 +1544,12 @@ define i16 @atomicrmw_xchg_i16_release(ptr %a, i16 %b) nounwind {
; LA64-LAMCAS-NEXT: # %bb.2: # %atomicrmw.end
; LA64-LAMCAS-NEXT: move $a0, $a2
; LA64-LAMCAS-NEXT: ret
+;
+; LA64-LAMCAS-LAM-BH-LABEL: atomicrmw_xchg_i16_release:
+; LA64-LAMCAS-LAM-BH: # %bb.0:
+; LA64-LAMCAS-LAM-BH-NEXT: amswap_db.h $a2, $a1, $a0
+; LA64-LAMCAS-LAM-BH-NEXT: move $a0, $a2
+; LA64-LAMCAS-LAM-BH-NEXT: ret
%1 = atomicrmw xchg ptr %a, i16 %b release
ret i16 %1
}
@@ -1224,6 +1579,12 @@ define i16 @atomicrmw_xchg_0_i16_release(ptr %a) nounwind {
; LA64-LAMCAS-NEXT: bne $a0, $a2, .LBB32_1
; LA64-LAMCAS-NEXT: # %bb.2: # %atomicrmw.end
; LA64-LAMCAS-NEXT: ret
+;
+; LA64-LAMCAS-LAM-BH-LABEL: atomicrmw_xchg_0_i16_release:
+; LA64-LAMCAS-LAM-BH: # %bb.0:
+; LA64-LAMCAS-LAM-BH-NEXT: amswap_db.h $a1, $zero, $a0
+; LA64-LAMCAS-LAM-BH-NEXT: move $a0, $a1
+; LA64-LAMCAS-LAM-BH-NEXT: ret
%1 = atomicrmw xchg ptr %a, i16 0 release
ret i16 %1
}
@@ -1253,6 +1614,13 @@ define i16 @atomicrmw_xchg_minus_1_i16_release(ptr %a) nounwind {
; LA64-LAMCAS-NEXT: bne $a0, $a3, .LBB33_1
; LA64-LAMCAS-NEXT: # %bb.2: # %atomicrmw.end
; LA64-LAMCAS-NEXT: ret
+;
+; LA64-LAMCAS-LAM-BH-LABEL: atomicrmw_xchg_minus_1_i16_release:
+; LA64-LAMCAS-LAM-BH: # %bb.0:
+; LA64-LAMCAS-LAM-BH-NEXT: addi.w $a2, $zero, -1
+; LA64-LAMCAS-LAM-BH-NEXT: amswap_db.h $a1, $a2, $a0
+; LA64-LAMCAS-LAM-BH-NEXT: move $a0, $a1
+; LA64-LAMCAS-LAM-BH-NEXT: ret
%1 = atomicrmw xchg ptr %a, i16 -1 release
ret i16 %1
@@ -1292,6 +1660,12 @@ define i8 @atomicrmw_add_i8_release(ptr %a, i8 %b) nounwind {
; LA64-LAMCAS-NEXT: bne $a0, $a4, .LBB34_1
; LA64-LAMCAS-NEXT: # %bb.2: # %atomicrmw.end
; LA64-LAMCAS-NEXT: ret
+;
+; LA64-LAMCAS-LAM-BH-LABEL: atomicrmw_add_i8_release:
+; LA64-LAMCAS-LAM-BH: # %bb.0:
+; LA64-LAMCAS-LAM-BH-NEXT: amadd_db.b $a2, $a1, $a0
+; LA64-LAMCAS-LAM-BH-NEXT: move $a0, $a2
+; LA64-LAMCAS-LAM-BH-NEXT: ret
%1 = atomicrmw add ptr %a, i8 %b release
ret i8 %1
}
@@ -1331,6 +1705,12 @@ define i16 @atomicrmw_add_i16_release(ptr %a, i16 %b) nounwind {
; LA64-LAMCAS-NEXT: bne $a0, $a4, .LBB35_1
; LA64-LAMCAS-NEXT: # %bb.2: # %atomicrmw.end
; LA64-LAMCAS-NEXT: ret
+;
+; LA64-LAMCAS-LAM-BH-LABEL: atomicrmw_add_i16_release:
+; LA64-LAMCAS-LAM-BH: # %bb.0:
+; LA64-LAMCAS-LAM-BH-NEXT: amadd_db.h $a2, $a1, $a0
+; LA64-LAMCAS-LAM-BH-NEXT: move $a0, $a2
+; LA64-LAMCAS-LAM-BH-NEXT: ret
%1 = atomicrmw add ptr %a, i16 %b release
ret i16 %1
@@ -1370,6 +1750,13 @@ define i8 @atomicrmw_sub_i8_release(ptr %a, i8 %b) nounwind {
; LA64-LAMCAS-NEXT: bne $a0, $a4, .LBB36_1
; LA64-LAMCAS-NEXT: # %bb.2: # %atomicrmw.end
; LA64-LAMCAS-NEXT: ret
+;
+; LA64-LAMCAS-LAM-BH-LABEL: atomicrmw_sub_i8_release:
+; LA64-LAMCAS-LAM-BH: # %bb.0:
+; LA64-LAMCAS-LAM-BH-NEXT: sub.w $a2, $zero, $a1
+; LA64-LAMCAS-LAM-BH-NEXT: amadd_db.b $a1, $a2, $a0
+; LA64-LAMCAS-LAM-BH-NEXT: move $a0, $a1
+; LA64-LAMCAS-LAM-BH-NEXT: ret
%1 = atomicrmw sub ptr %a, i8 %b release
ret i8 %1
}
@@ -1409,6 +1796,13 @@ define i16 @atomicrmw_sub_i16_release(ptr %a, i16 %b) nounwind {
; LA64-LAMCAS-NEXT: bne $a0, $a4, .LBB37_1
; LA64-LAMCAS-NEXT: # %bb.2: # %atomicrmw.end
; LA64-LAMCAS-NEXT: ret
+;
+; LA64-LAMCAS-LAM-BH-LABEL: atomicrmw_sub_i16_release:
+; LA64-LAMCAS-LAM-BH: # %bb.0:
+; LA64-LAMCAS-LAM-BH-NEXT: sub.w $a2, $zero, $a1
+; LA64-LAMCAS-LAM-BH-NEXT: amadd_db.h $a1, $a2, $a0
+; LA64-LAMCAS-LAM-BH-NEXT: move $a0, $a1
+; LA64-LAMCAS-LAM-BH-NEXT: ret
%1 = atomicrmw sub ptr %a, i16 %b release
ret i16 %1
@@ -1457,6 +1851,25 @@ define i8 @atomicrmw_umax_i8_release(ptr %a, i8 %b) nounwind {
; LA64-LAMCAS-NEXT: bne $a0, $a5, .LBB38_1
; LA64-LAMCAS-NEXT: # %bb.2: # %atomicrmw.end
; LA64-LAMCAS-NEXT: ret
+;
+; LA64-LAMCAS-LAM-BH-LABEL: atomicrmw_umax_i8_release:
+; LA64-LAMCAS-LAM-BH: # %bb.0:
+; LA64-LAMCAS-LAM-BH-NEXT: move $a2, $a0
+; LA64-LAMCAS-LAM-BH-NEXT: ld.bu $a0, $a0, 0
+; LA64-LAMCAS-LAM-BH-NEXT: andi $a3, $a1, 255
+; LA64-LAMCAS-LAM-BH-NEXT: .p2align 4, , 16
+; LA64-LAMCAS-LAM-BH-NEXT: .LBB38_1: # %atomicrmw.start
+; LA64-LAMCAS-LAM-BH-NEXT: # =>This Inner Loop Header: Depth=1
+; LA64-LAMCAS-LAM-BH-NEXT: andi $a4, $a0, 255
+; LA64-LAMCAS-LAM-BH-NEXT: sltu $a4, $a3, $a4
+; LA64-LAMCAS-LAM-BH-NEXT: masknez $a5, $a1, $a4
+; LA64-LAMCAS-LAM-BH-NEXT: maskeqz $a4, $a0, $a4
+; LA64-LAMCAS-LAM-BH-NEXT: or $a4, $a4, $a5
+; LA64-LAMCAS-LAM-BH-NEXT: ext.w.b $a5, $a0
+; LA64-LAMCAS-LAM-BH-NEXT: amcas_db.b $a0, $a4, $a2
+; LA64-LAMCAS-LAM-BH-NEXT: bne $a0, $a5, .LBB38_1
+; LA64-LAMCAS-LAM-BH-NEXT: # %bb.2: # %atomicrmw.end
+; LA64-LAMCAS-LAM-BH-NEXT: ret
%1 = atomicrmw umax ptr %a, i8 %b release
ret i8 %1
}
@@ -1505,6 +1918,25 @@ define i16 @atomicrmw_umax_i16_release(ptr %a, i16 %b) nounwind {
; LA64-LAMCAS-NEXT: bne $a0, $a5, .LBB39_1
; LA64-LAMCAS-NEXT: # %bb.2: # %atomicrmw.end
; LA64-LAMCAS-NEXT: ret
+;
+; LA64-LAMCAS-LAM-BH-LABEL: atomicrmw_umax_i16_release:
+; LA64-LAMCAS-LAM-BH: # %bb.0:
+; LA64-LAMCAS-LAM-BH-NEXT: move $a2, $a0
+; LA64-LAMCAS-LAM-BH-NEXT: ld.hu $a0, $a0, 0
+; LA64-LAMCAS-LAM-BH-NEXT: bstrpick.d $a3, $a1, 15, 0
+; LA64-LAMCAS-LAM-BH-NEXT: .p2align 4, , 16
+; LA64-LAMCAS-LAM-BH-NEXT: .LBB39_1: # %atomicrmw.start
+; LA64-LAMCAS-LAM-BH-NEXT: # =>This Inner Loop Header: Depth=1
+; LA64-LAMCAS-LAM-BH-NEXT: bstrpick.d $a4, $a0, 15, 0
+; LA64-LAMCAS-LAM-BH-NEXT: sltu $a4, $a3, $a4
+; LA64-LAMCAS-LAM-BH-NEXT: masknez $a5, $a1, $a4
+; LA64-LAMCAS-LAM-BH-NEXT: maskeqz $a4, $a0, $a4
+; LA64-LAMCAS-LAM-BH-NEXT: or $a4, $a4, $a5
+; LA64-LAMCAS-LAM-BH-NEXT: ext.w.h $a5, $a0
+; LA64-LAMCAS-LAM-BH-NEXT: amcas_db.h $a0, $a4, $a2
+; LA64-LAMCAS-LAM-BH-NEXT: bne $a0, $a5, .LBB39_1
+; LA64-LAMCAS-LAM-BH-NEXT: # %bb.2: # %atomicrmw.end
+; LA64-LAMCAS-LAM-BH-NEXT: ret
%1 = atomicrmw umax ptr %a, i16 %b release
ret i16 %1
}
@@ -1553,6 +1985,26 @@ define i8 @atomicrmw_umin_i8_release(ptr %a, i8 %b) nounwind {
; LA64-LAMCAS-NEXT: bne $a0, $a5, .LBB40_1
; LA64-LAMCAS-NEXT: # %bb.2: # %atomicrmw.end
; LA64-LAMCAS-NEXT: ret
+;
+; LA64-LAMCAS-LAM-BH-LABEL: atomicrmw_umin_i8_release:
+; LA64-LAMCAS-LAM-BH: # %bb.0:
+; LA64-LAMCAS-LAM-BH-NEXT: move $a2, $a0
+; LA64-LAMCAS-LAM-BH-NEXT: ld.bu $a0, $a0, 0
+; LA64-LAMCAS-LAM-BH-NEXT: andi $a3, $a1, 255
+; LA64-LAMCAS-LAM-BH-NEXT: .p2align 4, , 16
+; LA64-LAMCAS-LAM-BH-NEXT: .LBB40_1: # %atomicrmw.start
+; LA64-LAMCAS-LAM-BH-NEXT: # =>This Inner Loop Header: Depth=1
+; LA64-LAMCAS-LAM-BH-NEXT: andi $a4, $a0, 255
+; LA64-LAMCAS-LAM-BH-NEXT: sltu $a4, $a3, $a4
+; LA64-LAMCAS-LAM-BH-NEXT: xori $a4, $a4, 1
+; LA64-LAMCAS-LAM-BH-NEXT: masknez $a5, $a1, $a4
+; LA64-LAMCAS-LAM-BH-NEXT: maskeqz $a4, $a0, $a4
+; LA64-LAMCAS-LAM-BH-NEXT: or $a4, $a4, $a5
+; LA64-LAMCAS-LAM-BH-NEXT: ext.w.b $a5, $a0
+; LA64-LAMCAS-LAM-BH-NEXT: amcas_db.b $a0, $a4, $a2
+; LA64-LAMCAS-LAM-BH-NEXT: bne $a0, $a5, .LBB40_1
+; LA64-LAMCAS-LAM-BH-NEXT: # %bb.2: # %atomicrmw.end
+; LA64-LAMCAS-LAM-BH-NEXT: ret
%1 = atomicrmw umin ptr %a, i8 %b release
ret i8 %1
}
@@ -1602,6 +2054,26 @@ define i16 @atomicrmw_umin_i16_release(ptr %a, i16 %b) nounwind {
; LA64-LAMCAS-NEXT: bne $a0, $a5, .LBB41_1
; LA64-LAMCAS-NEXT: # %bb.2: # %atomicrmw.end
; LA64-LAMCAS-NEXT: ret
+;
+; LA64-LAMCAS-LAM-BH-LABEL: atomicrmw_umin_i16_release:
+; LA64-LAMCAS-LAM-BH: # %bb.0:
+; LA64-LAMCAS-LAM-BH-NEXT: move $a2, $a0
+; LA64-LAMCAS-LAM-BH-NEXT: ld.hu $a0, $a0, 0
+; LA64-LAMCAS-LAM-BH-NEXT: bstrpick.d $a3, $a1, 15, 0
+; LA64-LAMCAS-LAM-BH-NEXT: .p2align 4, , 16
+; LA64-LAMCAS-LAM-BH-NEXT: .LBB41_1: # %atomicrmw.start
+; LA64-LAMCAS-LAM-BH-NEXT: # =>This Inner Loop Header: Depth=1
+; LA64-LAMCAS-LAM-BH-NEXT: bstrpick.d $a4, $a0, 15, 0
+; LA64-LAMCAS-LAM-BH-NEXT: sltu $a4, $a3, $a4
+; LA64-LAMCAS-LAM-BH-NEXT: xori $a4, $a4, 1
+; LA64-LAMCAS-LAM-BH-NEXT: masknez $a5, $a1, $a4
+; LA64-LAMCAS-LAM-BH-NEXT: maskeqz $a4, $a0, $a4
+; LA64-LAMCAS-LAM-BH-NEXT: or $a4, $a4, $a5
+; LA64-LAMCAS-LAM-BH-NEXT: ext.w.h $a5, $a0
+; LA64-LAMCAS-LAM-BH-NEXT: amcas_db.h $a0, $a4, $a2
+; LA64-LAMCAS-LAM-BH-NEXT: bne $a0, $a5, .LBB41_1
+; LA64-LAMCAS-LAM-BH-NEXT: # %bb.2: # %atomicrmw.end
+; LA64-LAMCAS-LAM-BH-NEXT: ret
%1 = atomicrmw umin ptr %a, i16 %b release
ret i16 %1
}
@@ -1652,6 +2124,24 @@ define i8 @atomicrmw_max_i8_release(ptr %a, i8 %b) nounwind {
; LA64-LAMCAS-NEXT: bne $a0, $a4, .LBB42_1
; LA64-LAMCAS-NEXT: # %bb.2: # %atomicrmw.end
; LA64-LAMCAS-NEXT: ret
+;
+; LA64-LAMCAS-LAM-BH-LABEL: atomicrmw_max_i8_release:
+; LA64-LAMCAS-LAM-BH: # %bb.0:
+; LA64-LAMCAS-LAM-BH-NEXT: move $a2, $a0
+; LA64-LAMCAS-LAM-BH-NEXT: ld.bu $a0, $a0, 0
+; LA64-LAMCAS-LAM-BH-NEXT: ext.w.b $a3, $a1
+; LA64-LAMCAS-LAM-BH-NEXT: .p2align 4, , 16
+; LA64-LAMCAS-LAM-BH-NEXT: .LBB42_1: # %atomicrmw.start
+; LA64-LAMCAS-LAM-BH-NEXT: # =>This Inner Loop Header: Depth=1
+; LA64-LAMCAS-LAM-BH-NEXT: ext.w.b $a4, $a0
+; LA64-LAMCAS-LAM-BH-NEXT: slt $a5, $a3, $a4
+; LA64-LAMCAS-LAM-BH-NEXT: masknez $a6, $a1, $a5
+; LA64-LAMCAS-LAM-BH-NEXT: maskeqz $a5, $a0, $a5
+; LA64-LAMCAS-LAM-BH-NEXT: or $a5, $a5, $a6
+; LA64-LAMCAS-LAM-BH-NEXT: amcas_db.b $a0, $a5, $a2
+; LA64-LAMCAS-LAM-BH-NEXT: bne $a0, $a4, .LBB42_1
+; LA64-LAMCAS-LAM-BH-NEXT: # %bb.2: # %atomicrmw.end
+; LA64-LAMCAS-LAM-BH-NEXT: ret
%1 = atomicrmw max ptr %a, i8 %b release
ret i8 %1
}
@@ -1704,6 +2194,24 @@ define i16 @atomicrmw_max_i16_release(ptr %a, i16 %b) nounwind {
; LA64-LAMCAS-NEXT: bne $a0, $a4, .LBB43_1
; LA64-LAMCAS-NEXT: # %bb.2: # %atomicrmw.end
; LA64-LAMCAS-NEXT: ret
+;
+; LA64-LAMCAS-LAM-BH-LABEL: atomicrmw_max_i16_release:
+; LA64-LAMCAS-LAM-BH: # %bb.0:
+; LA64-LAMCAS-LAM-BH-NEXT: move $a2, $a0
+; LA64-LAMCAS-LAM-BH-NEXT: ld.hu $a0, $a0, 0
+; LA64-LAMCAS-LAM-BH-NEXT: ext.w.h $a3, $a1
+; LA64-LAMCAS-LAM-BH-NEXT: .p2align 4, , 16
+; LA64-LAMCAS-LAM-BH-NEXT: .LBB43_1: # %atomicrmw.start
+; LA64-LAMCAS-LAM-BH-NEXT: # =>This Inner Loop Header: Depth=1
+; LA64-LAMCAS-LAM-BH-NEXT: ext.w.h $a4, $a0
+; LA64-LAMCAS-LAM-BH-NEXT: slt $a5, $a3, $a4
+; LA64-LAMCAS-LAM-BH-NEXT: masknez $a6, $a1, $a5
+; LA64-LAMCAS-LAM-BH-NEXT: maskeqz $a5, $a0, $a5
+; LA64-LAMCAS-LAM-BH-NEXT: or $a5, $a5, $a6
+; LA64-LAMCAS-LAM-BH-NEXT: amcas_db.h $a0, $a5, $a2
+; LA64-LAMCAS-LAM-BH-NEXT: bne $a0, $a4, .LBB43_1
+; LA64-LAMCAS-LAM-BH-NEXT: # %bb.2: # %atomicrmw.end
+; LA64-LAMCAS-LAM-BH-NEXT: ret
%1 = atomicrmw max ptr %a, i16 %b release
ret i16 %1
}
@@ -1755,6 +2263,25 @@ define i8 @atomicrmw_min_i8_release(ptr %a, i8 %b) nounwind {
; LA64-LAMCAS-NEXT: bne $a0, $a4, .LBB44_1
; LA64-LAMCAS-NEXT: # %bb.2: # %atomicrmw.end
; LA64-LAMCAS-NEXT: ret
+;
+; LA64-LAMCAS-LAM-BH-LABEL: atomicrmw_min_i8_release:
+; LA64-LAMCAS-LAM-BH: # %bb.0:
+; LA64-LAMCAS-LAM-BH-NEXT: move $a2, $a0
+; LA64-LAMCAS-LAM-BH-NEXT: ld.bu $a0, $a0, 0
+; LA64-LAMCAS-LAM-BH-NEXT: ext.w.b $a3, $a1
+; LA64-LAMCAS-LAM-BH-NEXT: .p2align 4, , 16
+; LA64-LAMCAS-LAM-BH-NEXT: .LBB44_1: # %atomicrmw.start
+; LA64-LAMCAS-LAM-BH-NEXT: # =>This Inner Loop Header: Depth=1
+; LA64-LAMCAS-LAM-BH-NEXT: ext.w.b $a4, $a0
+; LA64-LAMCAS-LAM-BH-NEXT: slt $a5, $a3, $a4
+; LA64-LAMCAS-LAM-BH-NEXT: xori $a5, $a5, 1
+; LA64-LAMCAS-LAM-BH-NEXT: masknez $a6, $a1, $a5
+; LA64-LAMCAS-LAM-BH-NEXT: maskeqz $a5, $a0, $a5
+; LA64-LAMCAS-LAM-BH-NEXT: or $a5, $a5, $a6
+; LA64-LAMCAS-LAM-BH-NEXT: amcas_db.b $a0, $a5, $a2
+; LA64-LAMCAS-LAM-BH-NEXT: bne $a0, $a4, .LBB44_1
+; LA64-LAMCAS-LAM-BH-NEXT: # %bb.2: # %atomicrmw.end
+; LA64-LAMCAS-LAM-BH-NEXT: ret
%1 = atomicrmw min ptr %a, i8 %b release
ret i8 %1
}
@@ -1808,6 +2335,25 @@ define i16 @atomicrmw_min_i16_release(ptr %a, i16 %b) nounwind {
; LA64-LAMCAS-NEXT: bne $a0, $a4, .LBB45_1
; LA64-LAMCAS-NEXT: # %bb.2: # %atomicrmw.end
; LA64-LAMCAS-NEXT: ret
+;
+; LA64-LAMCAS-LAM-BH-LABEL: atomicrmw_min_i16_release:
+; LA64-LAMCAS-LAM-BH: # %bb.0:
+; LA64-LAMCAS-LAM-BH-NEXT: move $a2, $a0
+; LA64-LAMCAS-LAM-BH-NEXT: ld.hu $a0, $a0, 0
+; LA64-LAMCAS-LAM-BH-NEXT: ext.w.h $a3, $a1
+; LA64-LAMCAS-LAM-BH-NEXT: .p2align 4, , 16
+; LA64-LAMCAS-LAM-BH-NEXT: .LBB45_1: # %atomicrmw.start
+; LA64-LAMCAS-LAM-BH-NEXT: # =>This Inner Loop Header: Depth=1
+; LA64-LAMCAS-LAM-BH-NEXT: ext.w.h $a4, $a0
+; LA64-LAMCAS-LAM-BH-NEXT: slt $a5, $a3, $a4
+; LA64-LAMCAS-LAM-BH-NEXT: xori $a5, $a5, 1
+; LA64-LAMCAS-LAM-BH-NEXT: masknez $a6, $a1, $a5
+; LA64-LAMCAS-LAM-BH-NEXT: maskeqz $a5, $a0, $a5
+; LA64-LAMCAS-LAM-BH-NEXT: or $a5, $a5, $a6
+; LA64-LAMCAS-LAM-BH-NEXT: amcas_db.h $a0, $a5, $a2
+; LA64-LAMCAS-LAM-BH-NEXT: bne $a0, $a4, .LBB45_1
+; LA64-LAMCAS-LAM-BH-NEXT: # %bb.2: # %atomicrmw.end
+; LA64-LAMCAS-LAM-BH-NEXT: ret
%1 = atomicrmw min ptr %a, i16 %b release
ret i16 %1
}
@@ -1848,6 +2394,21 @@ define i8 @atomicrmw_nand_i8_release(ptr %a, i8 %b) nounwind {
; LA64-LAMCAS-NEXT: bne $a0, $a4, .LBB46_1
; LA64-LAMCAS-NEXT: # %bb.2: # %atomicrmw.end
; LA64-LAMCAS-NEXT: ret
+;
+; LA64-LAMCAS-LAM-BH-LABEL: atomicrmw_nand_i8_release:
+; LA64-LAMCAS-LAM-BH: # %bb.0:
+; LA64-LAMCAS-LAM-BH-NEXT: move $a2, $a0
+; LA64-LAMCAS-LAM-BH-NEXT: ld.bu $a0, $a0, 0
+; LA64-LAMCAS-LAM-BH-NEXT: .p2align 4, , 16
+; LA64-LAMCAS-LAM-BH-NEXT: .LBB46_1: # %atomicrmw.start
+; LA64-LAMCAS-LAM-BH-NEXT: # =>This Inner Loop Header: Depth=1
+; LA64-LAMCAS-LAM-BH-NEXT: and $a3, $a0, $a1
+; LA64-LAMCAS-LAM-BH-NEXT: nor $a3, $a3, $zero
+; LA64-LAMCAS-LAM-BH-NEXT: ext.w.b $a4, $a0
+; LA64-LAMCAS-LAM-BH-NEXT: amcas_db.b $a0, $a3, $a2
+; LA64-LAMCAS-LAM-BH-NEXT: bne $a0, $a4, .LBB46_1
+; LA64-LAMCAS-LAM-BH-NEXT: # %bb.2: # %atomicrmw.end
+; LA64-LAMCAS-LAM-BH-NEXT: ret
%1 = atomicrmw nand ptr %a, i8 %b release
ret i8 %1
}
@@ -1889,6 +2450,21 @@ define i16 @atomicrmw_nand_i16_release(ptr %a, i16 %b) nounwind {
; LA64-LAMCAS-NEXT: bne $a0, $a4, .LBB47_1
; LA64-LAMCAS-NEXT: # %bb.2: # %atomicrmw.end
; LA64-LAMCAS-NEXT: ret
+;
+; LA64-LAMCAS-LAM-BH-LABEL: atomicrmw_nand_i16_release:
+; LA64-LAMCAS-LAM-BH: # %bb.0:
+; LA64-LAMCAS-LAM-BH-NEXT: move $a2, $a0
+; LA64-LAMCAS-LAM-BH-NEXT: ld.hu $a0, $a0, 0
+; LA64-LAMCAS-LAM-BH-NEXT: .p2align 4, , 16
+; LA64-LAMCAS-LAM-BH-NEXT: .LBB47_1: # %atomicrmw.start
+; LA64-LAMCAS-LAM-BH-NEXT: # =>This Inner Loop Header: Depth=1
+; LA64-LAMCAS-LAM-BH-NEXT: and $a3, $a0, $a1
+; LA64-LAMCAS-LAM-BH-NEXT: nor $a3, $a3, $zero
+; LA64-LAMCAS-LAM-BH-NEXT: ext.w.h $a4, $a0
+; LA64-LAMCAS-LAM-BH-NEXT: amcas_db.h $a0, $a3, $a2
+; LA64-LAMCAS-LAM-BH-NEXT: bne $a0, $a4, .LBB47_1
+; LA64-LAMCAS-LAM-BH-NEXT: # %bb.2: # %atomicrmw.end
+; LA64-LAMCAS-LAM-BH-NEXT: ret
%1 = atomicrmw nand ptr %a, i16 %b release
ret i16 %1
@@ -1921,6 +2497,21 @@ define i32 @atomicrmw_nand_i32_release(ptr %a, i32 %b) nounwind {
; LA64-LAMCAS-NEXT: bne $a0, $a3, .LBB48_1
; LA64-LAMCAS-NEXT: # %bb.2: # %atomicrmw.end
; LA64-LAMCAS-NEXT: ret
+;
+; LA64-LAMCAS-LAM-BH-LABEL: atomicrmw_nand_i32_release:
+; LA64-LAMCAS-LAM-BH: # %bb.0:
+; LA64-LAMCAS-LAM-BH-NEXT: move $a2, $a0
+; LA64-LAMCAS-LAM-BH-NEXT: ld.w $a0, $a0, 0
+; LA64-LAMCAS-LAM-BH-NEXT: .p2align 4, , 16
+; LA64-LAMCAS-LAM-BH-NEXT: .LBB48_1: # %atomicrmw.start
+; LA64-LAMCAS-LAM-BH-NEXT: # =>This Inner Loop Header: Depth=1
+; LA64-LAMCAS-LAM-BH-NEXT: move $a3, $a0
+; LA64-LAMCAS-LAM-BH-NEXT: and $a4, $a0, $a1
+; LA64-LAMCAS-LAM-BH-NEXT: nor $a4, $a4, $zero
+; LA64-LAMCAS-LAM-BH-NEXT: amcas_db.w $a0, $a4, $a2
+; LA64-LAMCAS-LAM-BH-NEXT: bne $a0, $a3, .LBB48_1
+; LA64-LAMCAS-LAM-BH-NEXT: # %bb.2: # %atomicrmw.end
+; LA64-LAMCAS-LAM-BH-NEXT: ret
%1 = atomicrmw nand ptr %a, i32 %b release
ret i32 %1
}
@@ -1952,6 +2543,21 @@ define i64 @atomicrmw_nand_i64_release(ptr %a, i64 %b) nounwind {
; LA64-LAMCAS-NEXT: bne $a0, $a3, .LBB49_1
; LA64-LAMCAS-NEXT: # %bb.2: # %atomicrmw.end
; LA64-LAMCAS-NEXT: ret
+;
+; LA64-LAMCAS-LAM-BH-LABEL: atomicrmw_nand_i64_release:
+; LA64-LAMCAS-LAM-BH: # %bb.0:
+; LA64-LAMCAS-LAM-BH-NEXT: move $a2, $a0
+; LA64-LAMCAS-LAM-BH-NEXT: ld.d $a0, $a0, 0
+; LA64-LAMCAS-LAM-BH-NEXT: .p2align 4, , 16
+; LA64-LAMCAS-LAM-BH-NEXT: .LBB49_1: # %atomicrmw.start
+; LA64-LAMCAS-LAM-BH-NEXT: # =>This Inner Loop Header: Depth=1
+; LA64-LAMCAS-LAM-BH-NEXT: move $a3, $a0
+; LA64-LAMCAS-LAM-BH-NEXT: and $a4, $a0, $a1
+; LA64-LAMCAS-LAM-BH-NEXT: nor $a4, $a4, $zero
+; LA64-LAMCAS-LAM-BH-NEXT: amcas_db.d $a0, $a4, $a2
+; LA64-LAMCAS-LAM-BH-NEXT: bne $a0, $a3, .LBB49_1
+; LA64-LAMCAS-LAM-BH-NEXT: # %bb.2: # %atomicrmw.end
+; LA64-LAMCAS-LAM-BH-NEXT: ret
%1 = atomicrmw nand ptr %a, i64 %b release
ret i64 %1
}
@@ -1972,17 +2578,29 @@ define i8 @atomicrmw_and_i8_release(ptr %a, i8 %b) nounwind {
;
; LA64-LAMCAS-LABEL: atomicrmw_and_i8_release:
; LA64-LAMCAS: # %bb.0:
-; LA64-LAMCAS-NEXT: move $a2, $a0
-; LA64-LAMCAS-NEXT: ld.bu $a0, $a0, 0
-; LA64-LAMCAS-NEXT: .p2align 4, , 16
-; LA64-LAMCAS-NEXT: .LBB50_1: # %atomicrmw.start
-; LA64-LAMCAS-NEXT: # =>This Inner Loop Header: Depth=1
-; LA64-LAMCAS-NEXT: and $a3, $a0, $a1
-; LA64-LAMCAS-NEXT: ext.w.b $a4, $a0
-; LA64-LAMCAS-NEXT: amcas_db.b $a0, $a3, $a2
-; LA64-LAMCAS-NEXT: bne $a0, $a4, .LBB50_1
-; LA64-LAMCAS-NEXT: # %bb.2: # %atomicrmw.end
-; LA64-LAMCAS-NEXT: ret
+; LA64-LAMCAS-NEXT: slli.d $a2, $a0, 3
+; LA64-LAMCAS-NEXT: bstrins.d $a0, $zero, 1, 0
+; LA64-LAMCAS-NEXT: ori $a3, $zero, 255
+; LA64-LAMCAS-NEXT: sll.w $a3, $a3, $a2
+; LA64-LAMCAS-NEXT: andi $a1, $a1, 255
+; LA64-LAMCAS-NEXT: sll.w $a1, $a1, $a2
+; LA64-LAMCAS-NEXT: orn $a1, $a1, $a3
+; LA64-LAMCAS-NEXT: amand_db.w $a3, $a1, $a0
+; LA64-LAMCAS-NEXT: srl.w $a0, $a3, $a2
+; LA64-LAMCAS-NEXT: ret
+;
+; LA64-LAMCAS-LAM-BH-LABEL: atomicrmw_and_i8_release:
+; LA64-LAMCAS-LAM-BH: # %bb.0:
+; LA64-LAMCAS-LAM-BH-NEXT: slli.d $a2, $a0, 3
+; LA64-LAMCAS-LAM-BH-NEXT: bstrins.d $a0, $zero, 1, 0
+; LA64-LAMCAS-LAM-BH-NEXT: ori $a3, $zero, 255
+; LA64-LAMCAS-LAM-BH-NEXT: sll.w $a3, $a3, $a2
+; LA64-LAMCAS-LAM-BH-NEXT: andi $a1, $a1, 255
+; LA64-LAMCAS-LAM-BH-NEXT: sll.w $a1, $a1, $a2
+; LA64-LAMCAS-LAM-BH-NEXT: orn $a1, $a1, $a3
+; LA64-LAMCAS-LAM-BH-NEXT: amand_db.w $a3, $a1, $a0
+; LA64-LAMCAS-LAM-BH-NEXT: srl.w $a0, $a3, $a2
+; LA64-LAMCAS-LAM-BH-NEXT: ret
%1 = atomicrmw and ptr %a, i8 %b release
ret i8 %1
}
@@ -2004,17 +2622,31 @@ define i16 @atomicrmw_and_i16_release(ptr %a, i16 %b) nounwind {
;
; LA64-LAMCAS-LABEL: atomicrmw_and_i16_release:
; LA64-LAMCAS: # %bb.0:
-; LA64-LAMCAS-NEXT: move $a2, $a0
-; LA64-LAMCAS-NEXT: ld.hu $a0, $a0, 0
-; LA64-LAMCAS-NEXT: .p2align 4, , 16
-; LA64-LAMCAS-NEXT: .LBB51_1: # %atomicrmw.start
-; LA64-LAMCAS-NEXT: # =>This Inner Loop Header: Depth=1
-; LA64-LAMCAS-NEXT: and $a3, $a0, $a1
-; LA64-LAMCAS-NEXT: ext.w.h $a4, $a0
-; LA64-LAMCAS-NEXT: amcas_db.h $a0, $a3, $a2
-; LA64-LAMCAS-NEXT: bne $a0, $a4, .LBB51_1
-; LA64-LAMCAS-NEXT: # %bb.2: # %atomicrmw.end
-; LA64-LAMCAS-NEXT: ret
+; LA64-LAMCAS-NEXT: slli.d $a2, $a0, 3
+; LA64-LAMCAS-NEXT: bstrins.d $a0, $zero, 1, 0
+; LA64-LAMCAS-NEXT: lu12i.w $a3, 15
+; LA64-LAMCAS-NEXT: ori $a3, $a3, 4095
+; LA64-LAMCAS-NEXT: sll.w $a3, $a3, $a2
+; LA64-LAMCAS-NEXT: bstrpick.d $a1, $a1, 15, 0
+; LA64-LAMCAS-NEXT: sll.w $a1, $a1, $a2
+; LA64-LAMCAS-NEXT: orn $a1, $a1, $a3
+; LA64-LAMCAS-NEXT: amand_db.w $a3, $a1, $a0
+; LA64-LAMCAS-NEXT: srl.w $a0, $a3, $a2
+; LA64-LAMCAS-NEXT: ret
+;
+; LA64-LAMCAS-LAM-BH-LABEL: atomicrmw_and_i16_release:
+; LA64-LAMCAS-LAM-BH: # %bb.0:
+; LA64-LAMCAS-LAM-BH-NEXT: slli.d $a2, $a0, 3
+; LA64-LAMCAS-LAM-BH-NEXT: bstrins.d $a0, $zero, 1, 0
+; LA64-LAMCAS-LAM-BH-NEXT: lu12i.w $a3, 15
+; LA64-LAMCAS-LAM-BH-NEXT: ori $a3, $a3, 4095
+; LA64-LAMCAS-LAM-BH-NEXT: sll.w $a3, $a3, $a2
+; LA64-LAMCAS-LAM-BH-NEXT: bstrpick.d $a1, $a1, 15, 0
+; LA64-LAMCAS-LAM-BH-NEXT: sll.w $a1, $a1, $a2
+; LA64-LAMCAS-LAM-BH-NEXT: orn $a1, $a1, $a3
+; LA64-LAMCAS-LAM-BH-NEXT: amand_db.w $a3, $a1, $a0
+; LA64-LAMCAS-LAM-BH-NEXT: srl.w $a0, $a3, $a2
+; LA64-LAMCAS-LAM-BH-NEXT: ret
%1 = atomicrmw and ptr %a, i16 %b release
ret i16 %1
@@ -2033,17 +2665,23 @@ define i8 @atomicrmw_or_i8_release(ptr %a, i8 %b) nounwind {
;
; LA64-LAMCAS-LABEL: atomicrmw_or_i8_release:
; LA64-LAMCAS: # %bb.0:
-; LA64-LAMCAS-NEXT: move $a2, $a0
-; LA64-LAMCAS-NEXT: ld.bu $a0, $a0, 0
-; LA64-LAMCAS-NEXT: .p2align 4, , 16
-; LA64-LAMCAS-NEXT: .LBB52_1: # %atomicrmw.start
-; LA64-LAMCAS-NEXT: # =>This Inner Loop Header: Depth=1
-; LA64-LAMCAS-NEXT: or $a3, $a0, $a1
-; LA64-LAMCAS-NEXT: ext.w.b $a4, $a0
-; LA64-LAMCAS-NEXT: amcas_db.b $a0, $a3, $a2
-; LA64-LAMCAS-NEXT: bne $a0, $a4, .LBB52_1
-; LA64-LAMCAS-NEXT: # %bb.2: # %atomicrmw.end
-; LA64-LAMCAS-NEXT: ret
+; LA64-LAMCAS-NEXT: slli.d $a2, $a0, 3
+; LA64-LAMCAS-NEXT: bstrins.d $a0, $zero, 1, 0
+; LA64-LAMCAS-NEXT: andi $a1, $a1, 255
+; LA64-LAMCAS-NEXT: sll.w $a1, $a1, $a2
+; LA64-LAMCAS-NEXT: amor_db.w $a3, $a1, $a0
+; LA64-LAMCAS-NEXT: srl.w $a0, $a3, $a2
+; LA64-LAMCAS-NEXT: ret
+;
+; LA64-LAMCAS-LAM-BH-LABEL: atomicrmw_or_i8_release:
+; LA64-LAMCAS-LAM-BH: # %bb.0:
+; LA64-LAMCAS-LAM-BH-NEXT: slli.d $a2, $a0, 3
+; LA64-LAMCAS-LAM-BH-NEXT: bstrins.d $a0, $zero, 1, 0
+; LA64-LAMCAS-LAM-BH-NEXT: andi $a1, $a1, 255
+; LA64-LAMCAS-LAM-BH-NEXT: sll.w $a1, $a1, $a2
+; LA64-LAMCAS-LAM-BH-NEXT: amor_db.w $a3, $a1, $a0
+; LA64-LAMCAS-LAM-BH-NEXT: srl.w $a0, $a3, $a2
+; LA64-LAMCAS-LAM-BH-NEXT: ret
%1 = atomicrmw or ptr %a, i8 %b release
ret i8 %1
}
@@ -2061,17 +2699,23 @@ define i16 @atomicrmw_or_i16_release(ptr %a, i16 %b) nounwind {
;
; LA64-LAMCAS-LABEL: atomicrmw_or_i16_release:
; LA64-LAMCAS: # %bb.0:
-; LA64-LAMCAS-NEXT: move $a2, $a0
-; LA64-LAMCAS-NEXT: ld.hu $a0, $a0, 0
-; LA64-LAMCAS-NEXT: .p2align 4, , 16
-; LA64-LAMCAS-NEXT: .LBB53_1: # %atomicrmw.start
-; LA64-LAMCAS-NEXT: # =>This Inner Loop Header: Depth=1
-; LA64-LAMCAS-NEXT: or $a3, $a0, $a1
-; LA64-LAMCAS-NEXT: ext.w.h $a4, $a0
-; LA64-LAMCAS-NEXT: amcas_db.h $a0, $a3, $a2
-; LA64-LAMCAS-NEXT: bne $a0, $a4, .LBB53_1
-; LA64-LAMCAS-NEXT: # %bb.2: # %atomicrmw.end
-; LA64-LAMCAS-NEXT: ret
+; LA64-LAMCAS-NEXT: slli.d $a2, $a0, 3
+; LA64-LAMCAS-NEXT: bstrins.d $a0, $zero, 1, 0
+; LA64-LAMCAS-NEXT: bstrpick.d $a1, $a1, 15, 0
+; LA64-LAMCAS-NEXT: sll.w $a1, $a1, $a2
+; LA64-LAMCAS-NEXT: amor_db.w $a3, $a1, $a0
+; LA64-LAMCAS-NEXT: srl.w $a0, $a3, $a2
+; LA64-LAMCAS-NEXT: ret
+;
+; LA64-LAMCAS-LAM-BH-LABEL: atomicrmw_or_i16_release:
+; LA64-LAMCAS-LAM-BH: # %bb.0:
+; LA64-LAMCAS-LAM-BH-NEXT: slli.d $a2, $a0, 3
+; LA64-LAMCAS-LAM-BH-NEXT: bstrins.d $a0, $zero, 1, 0
+; LA64-LAMCAS-LAM-BH-NEXT: bstrpick.d $a1, $a1, 15, 0
+; LA64-LAMCAS-LAM-BH-NEXT: sll.w $a1, $a1, $a2
+; LA64-LAMCAS-LAM-BH-NEXT: amor_db.w $a3, $a1, $a0
+; LA64-LAMCAS-LAM-BH-NEXT: srl.w $a0, $a3, $a2
+; LA64-LAMCAS-LAM-BH-NEXT: ret
%1 = atomicrmw or ptr %a, i16 %b release
ret i16 %1
@@ -2090,17 +2734,23 @@ define i8 @atomicrmw_xor_i8_release(ptr %a, i8 %b) nounwind {
;
; LA64-LAMCAS-LABEL: atomicrmw_xor_i8_release:
; LA64-LAMCAS: # %bb.0:
-; LA64-LAMCAS-NEXT: move $a2, $a0
-; LA64-LAMCAS-NEXT: ld.bu $a0, $a0, 0
-; LA64-LAMCAS-NEXT: .p2align 4, , 16
-; LA64-LAMCAS-NEXT: .LBB54_1: # %atomicrmw.start
-; LA64-LAMCAS-NEXT: # =>This Inner Loop Header: Depth=1
-; LA64-LAMCAS-NEXT: xor $a3, $a0, $a1
-; LA64-LAMCAS-NEXT: ext.w.b $a4, $a0
-; LA64-LAMCAS-NEXT: amcas_db.b $a0, $a3, $a2
-; LA64-LAMCAS-NEXT: bne $a0, $a4, .LBB54_1
-; LA64-LAMCAS-NEXT: # %bb.2: # %atomicrmw.end
-; LA64-LAMCAS-NEXT: ret
+; LA64-LAMCAS-NEXT: slli.d $a2, $a0, 3
+; LA64-LAMCAS-NEXT: bstrins.d $a0, $zero, 1, 0
+; LA64-LAMCAS-NEXT: andi $a1, $a1, 255
+; LA64-LAMCAS-NEXT: sll.w $a1, $a1, $a2
+; LA64-LAMCAS-NEXT: amxor_db.w $a3, $a1, $a0
+; LA64-LAMCAS-NEXT: srl.w $a0, $a3, $a2
+; LA64-LAMCAS-NEXT: ret
+;
+; LA64-LAMCAS-LAM-BH-LABEL: atomicrmw_xor_i8_release:
+; LA64-LAMCAS-LAM-BH: # %bb.0:
+; LA64-LAMCAS-LAM-BH-NEXT: slli.d $a2, $a0, 3
+; LA64-LAMCAS-LAM-BH-NEXT: bstrins.d $a0, $zero, 1, 0
+; LA64-LAMCAS-LAM-BH-NEXT: andi $a1, $a1, 255
+; LA64-LAMCAS-LAM-BH-NEXT: sll.w $a1, $a1, $a2
+; LA64-LAMCAS-LAM-BH-NEXT: amxor_db.w $a3, $a1, $a0
+; LA64-LAMCAS-LAM-BH-NEXT: srl.w $a0, $a3, $a2
+; LA64-LAMCAS-LAM-BH-NEXT: ret
%1 = atomicrmw xor ptr %a, i8 %b release
ret i8 %1
}
@@ -2118,17 +2768,23 @@ define i16 @atomicrmw_xor_i16_release(ptr %a, i16 %b) nounwind {
;
; LA64-LAMCAS-LABEL: atomicrmw_xor_i16_release:
; LA64-LAMCAS: # %bb.0:
-; LA64-LAMCAS-NEXT: move $a2, $a0
-; LA64-LAMCAS-NEXT: ld.hu $a0, $a0, 0
-; LA64-LAMCAS-NEXT: .p2align 4, , 16
-; LA64-LAMCAS-NEXT: .LBB55_1: # %atomicrmw.start
-; LA64-LAMCAS-NEXT: # =>This Inner Loop Header: Depth=1
-; LA64-LAMCAS-NEXT: xor $a3, $a0, $a1
-; LA64-LAMCAS-NEXT: ext.w.h $a4, $a0
-; LA64-LAMCAS-NEXT: amcas_db.h $a0, $a3, $a2
-; LA64-LAMCAS-NEXT: bne $a0, $a4, .LBB55_1
-; LA64-LAMCAS-NEXT: # %bb.2: # %atomicrmw.end
-; LA64-LAMCAS-NEXT: ret
+; LA64-LAMCAS-NEXT: slli.d $a2, $a0, 3
+; LA64-LAMCAS-NEXT: bstrins.d $a0, $zero, 1, 0
+; LA64-LAMCAS-NEXT: bstrpick.d $a1, $a1, 15, 0
+; LA64-LAMCAS-NEXT: sll.w $a1, $a1, $a2
+; LA64-LAMCAS-NEXT: amxor_db.w $a3, $a1, $a0
+; LA64-LAMCAS-NEXT: srl.w $a0, $a3, $a2
+; LA64-LAMCAS-NEXT: ret
+;
+; LA64-LAMCAS-LAM-BH-LABEL: atomicrmw_xor_i16_release:
+; LA64-LAMCAS-LAM-BH: # %bb.0:
+; LA64-LAMCAS-LAM-BH-NEXT: slli.d $a2, $a0, 3
+; LA64-LAMCAS-LAM-BH-NEXT: bstrins.d $a0, $zero, 1, 0
+; LA64-LAMCAS-LAM-BH-NEXT: bstrpick.d $a1, $a1, 15, 0
+; LA64-LAMCAS-LAM-BH-NEXT: sll.w $a1, $a1, $a2
+; LA64-LAMCAS-LAM-BH-NEXT: amxor_db.w $a3, $a1, $a0
+; LA64-LAMCAS-LAM-BH-NEXT: srl.w $a0, $a3, $a2
+; LA64-LAMCAS-LAM-BH-NEXT: ret
%1 = atomicrmw xor ptr %a, i16 %b release
ret i16 %1
@@ -2167,6 +2823,12 @@ define i8 @atomicrmw_xchg_i8_acq_rel(ptr %a, i8 %b) nounwind {
; LA64-LAMCAS-NEXT: # %bb.2: # %atomicrmw.end
; LA64-LAMCAS-NEXT: move $a0, $a2
; LA64-LAMCAS-NEXT: ret
+;
+; LA64-LAMCAS-LAM-BH-LABEL: atomicrmw_xchg_i8_acq_rel:
+; LA64-LAMCAS-LAM-BH: # %bb.0:
+; LA64-LAMCAS-LAM-BH-NEXT: amswap_db.b $a2, $a1, $a0
+; LA64-LAMCAS-LAM-BH-NEXT: move $a0, $a2
+; LA64-LAMCAS-LAM-BH-NEXT: ret
%1 = atomicrmw xchg ptr %a, i8 %b acq_rel
ret i8 %1
}
@@ -2195,6 +2857,12 @@ define i8 @atomicrmw_xchg_0_i8_acq_rel(ptr %a) nounwind {
; LA64-LAMCAS-NEXT: bne $a0, $a2, .LBB57_1
; LA64-LAMCAS-NEXT: # %bb.2: # %atomicrmw.end
; LA64-LAMCAS-NEXT: ret
+;
+; LA64-LAMCAS-LAM-BH-LABEL: atomicrmw_xchg_0_i8_acq_rel:
+; LA64-LAMCAS-LAM-BH: # %bb.0:
+; LA64-LAMCAS-LAM-BH-NEXT: amswap_db.b $a1, $zero, $a0
+; LA64-LAMCAS-LAM-BH-NEXT: move $a0, $a1
+; LA64-LAMCAS-LAM-BH-NEXT: ret
%1 = atomicrmw xchg ptr %a, i8 0 acq_rel
ret i8 %1
}
@@ -2223,6 +2891,13 @@ define i8 @atomicrmw_xchg_minus_1_i8_acq_rel(ptr %a) nounwind {
; LA64-LAMCAS-NEXT: bne $a0, $a3, .LBB58_1
; LA64-LAMCAS-NEXT: # %bb.2: # %atomicrmw.end
; LA64-LAMCAS-NEXT: ret
+;
+; LA64-LAMCAS-LAM-BH-LABEL: atomicrmw_xchg_minus_1_i8_acq_rel:
+; LA64-LAMCAS-LAM-BH: # %bb.0:
+; LA64-LAMCAS-LAM-BH-NEXT: addi.w $a2, $zero, -1
+; LA64-LAMCAS-LAM-BH-NEXT: amswap_db.b $a1, $a2, $a0
+; LA64-LAMCAS-LAM-BH-NEXT: move $a0, $a1
+; LA64-LAMCAS-LAM-BH-NEXT: ret
%1 = atomicrmw xchg ptr %a, i8 -1 acq_rel
ret i8 %1
}
@@ -2261,6 +2936,12 @@ define i16 @atomicrmw_xchg_i16_acq_rel(ptr %a, i16 %b) nounwind {
; LA64-LAMCAS-NEXT: # %bb.2: # %atomicrmw.end
; LA64-LAMCAS-NEXT: move $a0, $a2
; LA64-LAMCAS-NEXT: ret
+;
+; LA64-LAMCAS-LAM-BH-LABEL: atomicrmw_xchg_i16_acq_rel:
+; LA64-LAMCAS-LAM-BH: # %bb.0:
+; LA64-LAMCAS-LAM-BH-NEXT: amswap_db.h $a2, $a1, $a0
+; LA64-LAMCAS-LAM-BH-NEXT: move $a0, $a2
+; LA64-LAMCAS-LAM-BH-NEXT: ret
%1 = atomicrmw xchg ptr %a, i16 %b acq_rel
ret i16 %1
}
@@ -2290,6 +2971,12 @@ define i16 @atomicrmw_xchg_0_i16_acq_rel(ptr %a) nounwind {
; LA64-LAMCAS-NEXT: bne $a0, $a2, .LBB60_1
; LA64-LAMCAS-NEXT: # %bb.2: # %atomicrmw.end
; LA64-LAMCAS-NEXT: ret
+;
+; LA64-LAMCAS-LAM-BH-LABEL: atomicrmw_xchg_0_i16_acq_rel:
+; LA64-LAMCAS-LAM-BH: # %bb.0:
+; LA64-LAMCAS-LAM-BH-NEXT: amswap_db.h $a1, $zero, $a0
+; LA64-LAMCAS-LAM-BH-NEXT: move $a0, $a1
+; LA64-LAMCAS-LAM-BH-NEXT: ret
%1 = atomicrmw xchg ptr %a, i16 0 acq_rel
ret i16 %1
}
@@ -2319,6 +3006,13 @@ define i16 @atomicrmw_xchg_minus_1_i16_acq_rel(ptr %a) nounwind {
; LA64-LAMCAS-NEXT: bne $a0, $a3, .LBB61_1
; LA64-LAMCAS-NEXT: # %bb.2: # %atomicrmw.end
; LA64-LAMCAS-NEXT: ret
+;
+; LA64-LAMCAS-LAM-BH-LABEL: atomicrmw_xchg_minus_1_i16_acq_rel:
+; LA64-LAMCAS-LAM-BH: # %bb.0:
+; LA64-LAMCAS-LAM-BH-NEXT: addi.w $a2, $zero, -1
+; LA64-LAMCAS-LAM-BH-NEXT: amswap_db.h $a1, $a2, $a0
+; LA64-LAMCAS-LAM-BH-NEXT: move $a0, $a1
+; LA64-LAMCAS-LAM-BH-NEXT: ret
%1 = atomicrmw xchg ptr %a, i16 -1 acq_rel
ret i16 %1
@@ -2358,6 +3052,12 @@ define i8 @atomicrmw_add_i8_acq_rel(ptr %a, i8 %b) nounwind {
; LA64-LAMCAS-NEXT: bne $a0, $a4, .LBB62_1
; LA64-LAMCAS-NEXT: # %bb.2: # %atomicrmw.end
; LA64-LAMCAS-NEXT: ret
+;
+; LA64-LAMCAS-LAM-BH-LABEL: atomicrmw_add_i8_acq_rel:
+; LA64-LAMCAS-LAM-BH: # %bb.0:
+; LA64-LAMCAS-LAM-BH-NEXT: amadd_db.b $a2, $a1, $a0
+; LA64-LAMCAS-LAM-BH-NEXT: move $a0, $a2
+; LA64-LAMCAS-LAM-BH-NEXT: ret
%1 = atomicrmw add ptr %a, i8 %b acq_rel
ret i8 %1
}
@@ -2397,6 +3097,12 @@ define i16 @atomicrmw_add_i16_acq_rel(ptr %a, i16 %b) nounwind {
; LA64-LAMCAS-NEXT: bne $a0, $a4, .LBB63_1
; LA64-LAMCAS-NEXT: # %bb.2: # %atomicrmw.end
; LA64-LAMCAS-NEXT: ret
+;
+; LA64-LAMCAS-LAM-BH-LABEL: atomicrmw_add_i16_acq_rel:
+; LA64-LAMCAS-LAM-BH: # %bb.0:
+; LA64-LAMCAS-LAM-BH-NEXT: amadd_db.h $a2, $a1, $a0
+; LA64-LAMCAS-LAM-BH-NEXT: move $a0, $a2
+; LA64-LAMCAS-LAM-BH-NEXT: ret
%1 = atomicrmw add ptr %a, i16 %b acq_rel
ret i16 %1
@@ -2436,6 +3142,13 @@ define i8 @atomicrmw_sub_i8_acq_rel(ptr %a, i8 %b) nounwind {
; LA64-LAMCAS-NEXT: bne $a0, $a4, .LBB64_1
; LA64-LAMCAS-NEXT: # %bb.2: # %atomicrmw.end
; LA64-LAMCAS-NEXT: ret
+;
+; LA64-LAMCAS-LAM-BH-LABEL: atomicrmw_sub_i8_acq_rel:
+; LA64-LAMCAS-LAM-BH: # %bb.0:
+; LA64-LAMCAS-LAM-BH-NEXT: sub.w $a2, $zero, $a1
+; LA64-LAMCAS-LAM-BH-NEXT: amadd_db.b $a1, $a2, $a0
+; LA64-LAMCAS-LAM-BH-NEXT: move $a0, $a1
+; LA64-LAMCAS-LAM-BH-NEXT: ret
%1 = atomicrmw sub ptr %a, i8 %b acq_rel
ret i8 %1
}
@@ -2475,6 +3188,13 @@ define i16 @atomicrmw_sub_i16_acq_rel(ptr %a, i16 %b) nounwind {
; LA64-LAMCAS-NEXT: bne $a0, $a4, .LBB65_1
; LA64-LAMCAS-NEXT: # %bb.2: # %atomicrmw.end
; LA64-LAMCAS-NEXT: ret
+;
+; LA64-LAMCAS-LAM-BH-LABEL: atomicrmw_sub_i16_acq_rel:
+; LA64-LAMCAS-LAM-BH: # %bb.0:
+; LA64-LAMCAS-LAM-BH-NEXT: sub.w $a2, $zero, $a1
+; LA64-LAMCAS-LAM-BH-NEXT: amadd_db.h $a1, $a2, $a0
+; LA64-LAMCAS-LAM-BH-NEXT: move $a0, $a1
+; LA64-LAMCAS-LAM-BH-NEXT: ret
%1 = atomicrmw sub ptr %a, i16 %b acq_rel
ret i16 %1
@@ -2523,6 +3243,25 @@ define i8 @atomicrmw_umax_i8_acq_rel(ptr %a, i8 %b) nounwind {
; LA64-LAMCAS-NEXT: bne $a0, $a5, .LBB66_1
; LA64-LAMCAS-NEXT: # %bb.2: # %atomicrmw.end
; LA64-LAMCAS-NEXT: ret
+;
+; LA64-LAMCAS-LAM-BH-LABEL: atomicrmw_umax_i8_acq_rel:
+; LA64-LAMCAS-LAM-BH: # %bb.0:
+; LA64-LAMCAS-LAM-BH-NEXT: move $a2, $a0
+; LA64-LAMCAS-LAM-BH-NEXT: ld.bu $a0, $a0, 0
+; LA64-LAMCAS-LAM-BH-NEXT: andi $a3, $a1, 255
+; LA64-LAMCAS-LAM-BH-NEXT: .p2align 4, , 16
+; LA64-LAMCAS-LAM-BH-NEXT: .LBB66_1: # %atomicrmw.start
+; LA64-LAMCAS-LAM-BH-NEXT: # =>This Inner Loop Header: Depth=1
+; LA64-LAMCAS-LAM-BH-NEXT: andi $a4, $a0, 255
+; LA64-LAMCAS-LAM-BH-NEXT: sltu $a4, $a3, $a4
+; LA64-LAMCAS-LAM-BH-NEXT: masknez $a5, $a1, $a4
+; LA64-LAMCAS-LAM-BH-NEXT: maskeqz $a4, $a0, $a4
+; LA64-LAMCAS-LAM-BH-NEXT: or $a4, $a4, $a5
+; LA64-LAMCAS-LAM-BH-NEXT: ext.w.b $a5, $a0
+; LA64-LAMCAS-LAM-BH-NEXT: amcas_db.b $a0, $a4, $a2
+; LA64-LAMCAS-LAM-BH-NEXT: bne $a0, $a5, .LBB66_1
+; LA64-LAMCAS-LAM-BH-NEXT: # %bb.2: # %atomicrmw.end
+; LA64-LAMCAS-LAM-BH-NEXT: ret
%1 = atomicrmw umax ptr %a, i8 %b acq_rel
ret i8 %1
}
@@ -2571,6 +3310,25 @@ define i16 @atomicrmw_umax_i16_acq_rel(ptr %a, i16 %b) nounwind {
; LA64-LAMCAS-NEXT: bne $a0, $a5, .LBB67_1
; LA64-LAMCAS-NEXT: # %bb.2: # %atomicrmw.end
; LA64-LAMCAS-NEXT: ret
+;
+; LA64-LAMCAS-LAM-BH-LABEL: atomicrmw_umax_i16_acq_rel:
+; LA64-LAMCAS-LAM-BH: # %bb.0:
+; LA64-LAMCAS-LAM-BH-NEXT: move $a2, $a0
+; LA64-LAMCAS-LAM-BH-NEXT: ld.hu $a0, $a0, 0
+; LA64-LAMCAS-LAM-BH-NEXT: bstrpick.d $a3, $a1, 15, 0
+; LA64-LAMCAS-LAM-BH-NEXT: .p2align 4, , 16
+; LA64-LAMCAS-LAM-BH-NEXT: .LBB67_1: # %atomicrmw.start
+; LA64-LAMCAS-LAM-BH-NEXT: # =>This Inner Loop Header: Depth=1
+; LA64-LAMCAS-LAM-BH-NEXT: bstrpick.d $a4, $a0, 15, 0
+; LA64-LAMCAS-LAM-BH-NEXT: sltu $a4, $a3, $a4
+; LA64-LAMCAS-LAM-BH-NEXT: masknez $a5, $a1, $a4
+; LA64-LAMCAS-LAM-BH-NEXT: maskeqz $a4, $a0, $a4
+; LA64-LAMCAS-LAM-BH-NEXT: or $a4, $a4, $a5
+; LA64-LAMCAS-LAM-BH-NEXT: ext.w.h $a5, $a0
+; LA64-LAMCAS-LAM-BH-NEXT: amcas_db.h $a0, $a4, $a2
+; LA64-LAMCAS-LAM-BH-NEXT: bne $a0, $a5, .LBB67_1
+; LA64-LAMCAS-LAM-BH-NEXT: # %bb.2: # %atomicrmw.end
+; LA64-LAMCAS-LAM-BH-NEXT: ret
%1 = atomicrmw umax ptr %a, i16 %b acq_rel
ret i16 %1
}
@@ -2619,6 +3377,26 @@ define i8 @atomicrmw_umin_i8_acq_rel(ptr %a, i8 %b) nounwind {
; LA64-LAMCAS-NEXT: bne $a0, $a5, .LBB68_1
; LA64-LAMCAS-NEXT: # %bb.2: # %atomicrmw.end
; LA64-LAMCAS-NEXT: ret
+;
+; LA64-LAMCAS-LAM-BH-LABEL: atomicrmw_umin_i8_acq_rel:
+; LA64-LAMCAS-LAM-BH: # %bb.0:
+; LA64-LAMCAS-LAM-BH-NEXT: move $a2, $a0
+; LA64-LAMCAS-LAM-BH-NEXT: ld.bu $a0, $a0, 0
+; LA64-LAMCAS-LAM-BH-NEXT: andi $a3, $a1, 255
+; LA64-LAMCAS-LAM-BH-NEXT: .p2align 4, , 16
+; LA64-LAMCAS-LAM-BH-NEXT: .LBB68_1: # %atomicrmw.start
+; LA64-LAMCAS-LAM-BH-NEXT: # =>This Inner Loop Header: Depth=1
+; LA64-LAMCAS-LAM-BH-NEXT: andi $a4, $a0, 255
+; LA64-LAMCAS-LAM-BH-NEXT: sltu $a4, $a3, $a4
+; LA64-LAMCAS-LAM-BH-NEXT: xori $a4, $a4, 1
+; LA64-LAMCAS-LAM-BH-NEXT: masknez $a5, $a1, $a4
+; LA64-LAMCAS-LAM-BH-NEXT: maskeqz $a4, $a0, $a4
+; LA64-LAMCAS-LAM-BH-NEXT: or $a4, $a4, $a5
+; LA64-LAMCAS-LAM-BH-NEXT: ext.w.b $a5, $a0
+; LA64-LAMCAS-LAM-BH-NEXT: amcas_db.b $a0, $a4, $a2
+; LA64-LAMCAS-LAM-BH-NEXT: bne $a0, $a5, .LBB68_1
+; LA64-LAMCAS-LAM-BH-NEXT: # %bb.2: # %atomicrmw.end
+; LA64-LAMCAS-LAM-BH-NEXT: ret
%1 = atomicrmw umin ptr %a, i8 %b acq_rel
ret i8 %1
}
@@ -2668,6 +3446,26 @@ define i16 @atomicrmw_umin_i16_acq_rel(ptr %a, i16 %b) nounwind {
; LA64-LAMCAS-NEXT: bne $a0, $a5, .LBB69_1
; LA64-LAMCAS-NEXT: # %bb.2: # %atomicrmw.end
; LA64-LAMCAS-NEXT: ret
+;
+; LA64-LAMCAS-LAM-BH-LABEL: atomicrmw_umin_i16_acq_rel:
+; LA64-LAMCAS-LAM-BH: # %bb.0:
+; LA64-LAMCAS-LAM-BH-NEXT: move $a2, $a0
+; LA64-LAMCAS-LAM-BH-NEXT: ld.hu $a0, $a0, 0
+; LA64-LAMCAS-LAM-BH-NEXT: bstrpick.d $a3, $a1, 15, 0
+; LA64-LAMCAS-LAM-BH-NEXT: .p2align 4, , 16
+; LA64-LAMCAS-LAM-BH-NEXT: .LBB69_1: # %atomicrmw.start
+; LA64-LAMCAS-LAM-BH-NEXT: # =>This Inner Loop Header: Depth=1
+; LA64-LAMCAS-LAM-BH-NEXT: bstrpick.d $a4, $a0, 15, 0
+; LA64-LAMCAS-LAM-BH-NEXT: sltu $a4, $a3, $a4
+; LA64-LAMCAS-LAM-BH-NEXT: xori $a4, $a4, 1
+; LA64-LAMCAS-LAM-BH-NEXT: masknez $a5, $a1, $a4
+; LA64-LAMCAS-LAM-BH-NEXT: maskeqz $a4, $a0, $a4
+; LA64-LAMCAS-LAM-BH-NEXT: or $a4, $a4, $a5
+; LA64-LAMCAS-LAM-BH-NEXT: ext.w.h $a5, $a0
+; LA64-LAMCAS-LAM-BH-NEXT: amcas_db.h $a0, $a4, $a2
+; LA64-LAMCAS-LAM-BH-NEXT: bne $a0, $a5, .LBB69_1
+; LA64-LAMCAS-LAM-BH-NEXT: # %bb.2: # %atomicrmw.end
+; LA64-LAMCAS-LAM-BH-NEXT: ret
%1 = atomicrmw umin ptr %a, i16 %b acq_rel
ret i16 %1
}
@@ -2718,6 +3516,24 @@ define i8 @atomicrmw_max_i8_acq_rel(ptr %a, i8 %b) nounwind {
; LA64-LAMCAS-NEXT: bne $a0, $a4, .LBB70_1
; LA64-LAMCAS-NEXT: # %bb.2: # %atomicrmw.end
; LA64-LAMCAS-NEXT: ret
+;
+; LA64-LAMCAS-LAM-BH-LABEL: atomicrmw_max_i8_acq_rel:
+; LA64-LAMCAS-LAM-BH: # %bb.0:
+; LA64-LAMCAS-LAM-BH-NEXT: move $a2, $a0
+; LA64-LAMCAS-LAM-BH-NEXT: ld.bu $a0, $a0, 0
+; LA64-LAMCAS-LAM-BH-NEXT: ext.w.b $a3, $a1
+; LA64-LAMCAS-LAM-BH-NEXT: .p2align 4, , 16
+; LA64-LAMCAS-LAM-BH-NEXT: .LBB70_1: # %atomicrmw.start
+; LA64-LAMCAS-LAM-BH-NEXT: # =>This Inner Loop Header: Depth=1
+; LA64-LAMCAS-LAM-BH-NEXT: ext.w.b $a4, $a0
+; LA64-LAMCAS-LAM-BH-NEXT: slt $a5, $a3, $a4
+; LA64-LAMCAS-LAM-BH-NEXT: masknez $a6, $a1, $a5
+; LA64-LAMCAS-LAM-BH-NEXT: maskeqz $a5, $a0, $a5
+; LA64-LAMCAS-LAM-BH-NEXT: or $a5, $a5, $a6
+; LA64-LAMCAS-LAM-BH-NEXT: amcas_db.b $a0, $a5, $a2
+; LA64-LAMCAS-LAM-BH-NEXT: bne $a0, $a4, .LBB70_1
+; LA64-LAMCAS-LAM-BH-NEXT: # %bb.2: # %atomicrmw.end
+; LA64-LAMCAS-LAM-BH-NEXT: ret
%1 = atomicrmw max ptr %a, i8 %b acq_rel
ret i8 %1
}
@@ -2770,6 +3586,24 @@ define i16 @atomicrmw_max_i16_acq_rel(ptr %a, i16 %b) nounwind {
; LA64-LAMCAS-NEXT: bne $a0, $a4, .LBB71_1
; LA64-LAMCAS-NEXT: # %bb.2: # %atomicrmw.end
; LA64-LAMCAS-NEXT: ret
+;
+; LA64-LAMCAS-LAM-BH-LABEL: atomicrmw_max_i16_acq_rel:
+; LA64-LAMCAS-LAM-BH: # %bb.0:
+; LA64-LAMCAS-LAM-BH-NEXT: move $a2, $a0
+; LA64-LAMCAS-LAM-BH-NEXT: ld.hu $a0, $a0, 0
+; LA64-LAMCAS-LAM-BH-NEXT: ext.w.h $a3, $a1
+; LA64-LAMCAS-LAM-BH-NEXT: .p2align 4, , 16
+; LA64-LAMCAS-LAM-BH-NEXT: .LBB71_1: # %atomicrmw.start
+; LA64-LAMCAS-LAM-BH-NEXT: # =>This Inner Loop Header: Depth=1
+; LA64-LAMCAS-LAM-BH-NEXT: ext.w.h $a4, $a0
+; LA64-LAMCAS-LAM-BH-NEXT: slt $a5, $a3, $a4
+; LA64-LAMCAS-LAM-BH-NEXT: masknez $a6, $a1, $a5
+; LA64-LAMCAS-LAM-BH-NEXT: maskeqz $a5, $a0, $a5
+; LA64-LAMCAS-LAM-BH-NEXT: or $a5, $a5, $a6
+; LA64-LAMCAS-LAM-BH-NEXT: amcas_db.h $a0, $a5, $a2
+; LA64-LAMCAS-LAM-BH-NEXT: bne $a0, $a4, .LBB71_1
+; LA64-LAMCAS-LAM-BH-NEXT: # %bb.2: # %atomicrmw.end
+; LA64-LAMCAS-LAM-BH-NEXT: ret
%1 = atomicrmw max ptr %a, i16 %b acq_rel
ret i16 %1
}
@@ -2821,6 +3655,25 @@ define i8 @atomicrmw_min_i8_acq_rel(ptr %a, i8 %b) nounwind {
; LA64-LAMCAS-NEXT: bne $a0, $a4, .LBB72_1
; LA64-LAMCAS-NEXT: # %bb.2: # %atomicrmw.end
; LA64-LAMCAS-NEXT: ret
+;
+; LA64-LAMCAS-LAM-BH-LABEL: atomicrmw_min_i8_acq_rel:
+; LA64-LAMCAS-LAM-BH: # %bb.0:
+; LA64-LAMCAS-LAM-BH-NEXT: move $a2, $a0
+; LA64-LAMCAS-LAM-BH-NEXT: ld.bu $a0, $a0, 0
+; LA64-LAMCAS-LAM-BH-NEXT: ext.w.b $a3, $a1
+; LA64-LAMCAS-LAM-BH-NEXT: .p2align 4, , 16
+; LA64-LAMCAS-LAM-BH-NEXT: .LBB72_1: # %atomicrmw.start
+; LA64-LAMCAS-LAM-BH-NEXT: # =>This Inner Loop Header: Depth=1
+; LA64-LAMCAS-LAM-BH-NEXT: ext.w.b $a4, $a0
+; LA64-LAMCAS-LAM-BH-NEXT: slt $a5, $a3, $a4
+; LA64-LAMCAS-LAM-BH-NEXT: xori $a5, $a5, 1
+; LA64-LAMCAS-LAM-BH-NEXT: masknez $a6, $a1, $a5
+; LA64-LAMCAS-LAM-BH-NEXT: maskeqz $a5, $a0, $a5
+; LA64-LAMCAS-LAM-BH-NEXT: or $a5, $a5, $a6
+; LA64-LAMCAS-LAM-BH-NEXT: amcas_db.b $a0, $a5, $a2
+; LA64-LAMCAS-LAM-BH-NEXT: bne $a0, $a4, .LBB72_1
+; LA64-LAMCAS-LAM-BH-NEXT: # %bb.2: # %atomicrmw.end
+; LA64-LAMCAS-LAM-BH-NEXT: ret
%1 = atomicrmw min ptr %a, i8 %b acq_rel
ret i8 %1
}
@@ -2874,6 +3727,25 @@ define i16 @atomicrmw_min_i16_acq_rel(ptr %a, i16 %b) nounwind {
; LA64-LAMCAS-NEXT: bne $a0, $a4, .LBB73_1
; LA64-LAMCAS-NEXT: # %bb.2: # %atomicrmw.end
; LA64-LAMCAS-NEXT: ret
+;
+; LA64-LAMCAS-LAM-BH-LABEL: atomicrmw_min_i16_acq_rel:
+; LA64-LAMCAS-LAM-BH: # %bb.0:
+; LA64-LAMCAS-LAM-BH-NEXT: move $a2, $a0
+; LA64-LAMCAS-LAM-BH-NEXT: ld.hu $a0, $a0, 0
+; LA64-LAMCAS-LAM-BH-NEXT: ext.w.h $a3, $a1
+; LA64-LAMCAS-LAM-BH-NEXT: .p2align 4, , 16
+; LA64-LAMCAS-LAM-BH-NEXT: .LBB73_1: # %atomicrmw.start
+; LA64-LAMCAS-LAM-BH-NEXT: # =>This Inner Loop Header: Depth=1
+; LA64-LAMCAS-LAM-BH-NEXT: ext.w.h $a4, $a0
+; LA64-LAMCAS-LAM-BH-NEXT: slt $a5, $a3, $a4
+; LA64-LAMCAS-LAM-BH-NEXT: xori $a5, $a5, 1
+; LA64-LAMCAS-LAM-BH-NEXT: masknez $a6, $a1, $a5
+; LA64-LAMCAS-LAM-BH-NEXT: maskeqz $a5, $a0, $a5
+; LA64-LAMCAS-LAM-BH-NEXT: or $a5, $a5, $a6
+; LA64-LAMCAS-LAM-BH-NEXT: amcas_db.h $a0, $a5, $a2
+; LA64-LAMCAS-LAM-BH-NEXT: bne $a0, $a4, .LBB73_1
+; LA64-LAMCAS-LAM-BH-NEXT: # %bb.2: # %atomicrmw.end
+; LA64-LAMCAS-LAM-BH-NEXT: ret
%1 = atomicrmw min ptr %a, i16 %b acq_rel
ret i16 %1
}
@@ -2914,6 +3786,21 @@ define i8 @atomicrmw_nand_i8_acq_rel(ptr %a, i8 %b) nounwind {
; LA64-LAMCAS-NEXT: bne $a0, $a4, .LBB74_1
; LA64-LAMCAS-NEXT: # %bb.2: # %atomicrmw.end
; LA64-LAMCAS-NEXT: ret
+;
+; LA64-LAMCAS-LAM-BH-LABEL: atomicrmw_nand_i8_acq_rel:
+; LA64-LAMCAS-LAM-BH: # %bb.0:
+; LA64-LAMCAS-LAM-BH-NEXT: move $a2, $a0
+; LA64-LAMCAS-LAM-BH-NEXT: ld.bu $a0, $a0, 0
+; LA64-LAMCAS-LAM-BH-NEXT: .p2align 4, , 16
+; LA64-LAMCAS-LAM-BH-NEXT: .LBB74_1: # %atomicrmw.start
+; LA64-LAMCAS-LAM-BH-NEXT: # =>This Inner Loop Header: Depth=1
+; LA64-LAMCAS-LAM-BH-NEXT: and $a3, $a0, $a1
+; LA64-LAMCAS-LAM-BH-NEXT: nor $a3, $a3, $zero
+; LA64-LAMCAS-LAM-BH-NEXT: ext.w.b $a4, $a0
+; LA64-LAMCAS-LAM-BH-NEXT: amcas_db.b $a0, $a3, $a2
+; LA64-LAMCAS-LAM-BH-NEXT: bne $a0, $a4, .LBB74_1
+; LA64-LAMCAS-LAM-BH-NEXT: # %bb.2: # %atomicrmw.end
+; LA64-LAMCAS-LAM-BH-NEXT: ret
%1 = atomicrmw nand ptr %a, i8 %b acq_rel
ret i8 %1
}
@@ -2955,6 +3842,21 @@ define i16 @atomicrmw_nand_i16_acq_rel(ptr %a, i16 %b) nounwind {
; LA64-LAMCAS-NEXT: bne $a0, $a4, .LBB75_1
; LA64-LAMCAS-NEXT: # %bb.2: # %atomicrmw.end
; LA64-LAMCAS-NEXT: ret
+;
+; LA64-LAMCAS-LAM-BH-LABEL: atomicrmw_nand_i16_acq_rel:
+; LA64-LAMCAS-LAM-BH: # %bb.0:
+; LA64-LAMCAS-LAM-BH-NEXT: move $a2, $a0
+; LA64-LAMCAS-LAM-BH-NEXT: ld.hu $a0, $a0, 0
+; LA64-LAMCAS-LAM-BH-NEXT: .p2align 4, , 16
+; LA64-LAMCAS-LAM-BH-NEXT: .LBB75_1: # %atomicrmw.start
+; LA64-LAMCAS-LAM-BH-NEXT: # =>This Inner Loop Header: Depth=1
+; LA64-LAMCAS-LAM-BH-NEXT: and $a3, $a0, $a1
+; LA64-LAMCAS-LAM-BH-NEXT: nor $a3, $a3, $zero
+; LA64-LAMCAS-LAM-BH-NEXT: ext.w.h $a4, $a0
+; LA64-LAMCAS-LAM-BH-NEXT: amcas_db.h $a0, $a3, $a2
+; LA64-LAMCAS-LAM-BH-NEXT: bne $a0, $a4, .LBB75_1
+; LA64-LAMCAS-LAM-BH-NEXT: # %bb.2: # %atomicrmw.end
+; LA64-LAMCAS-LAM-BH-NEXT: ret
%1 = atomicrmw nand ptr %a, i16 %b acq_rel
ret i16 %1
@@ -2987,6 +3889,21 @@ define i32 @atomicrmw_nand_i32_acq_rel(ptr %a, i32 %b) nounwind {
; LA64-LAMCAS-NEXT: bne $a0, $a3, .LBB76_1
; LA64-LAMCAS-NEXT: # %bb.2: # %atomicrmw.end
; LA64-LAMCAS-NEXT: ret
+;
+; LA64-LAMCAS-LAM-BH-LABEL: atomicrmw_nand_i32_acq_rel:
+; LA64-LAMCAS-LAM-BH: # %bb.0:
+; LA64-LAMCAS-LAM-BH-NEXT: move $a2, $a0
+; LA64-LAMCAS-LAM-BH-NEXT: ld.w $a0, $a0, 0
+; LA64-LAMCAS-LAM-BH-NEXT: .p2align 4, , 16
+; LA64-LAMCAS-LAM-BH-NEXT: .LBB76_1: # %atomicrmw.start
+; LA64-LAMCAS-LAM-BH-NEXT: # =>This Inner Loop Header: Depth=1
+; LA64-LAMCAS-LAM-BH-NEXT: move $a3, $a0
+; LA64-LAMCAS-LAM-BH-NEXT: and $a4, $a0, $a1
+; LA64-LAMCAS-LAM-BH-NEXT: nor $a4, $a4, $zero
+; LA64-LAMCAS-LAM-BH-NEXT: amcas_db.w $a0, $a4, $a2
+; LA64-LAMCAS-LAM-BH-NEXT: bne $a0, $a3, .LBB76_1
+; LA64-LAMCAS-LAM-BH-NEXT: # %bb.2: # %atomicrmw.end
+; LA64-LAMCAS-LAM-BH-NEXT: ret
%1 = atomicrmw nand ptr %a, i32 %b acq_rel
ret i32 %1
}
@@ -3018,6 +3935,21 @@ define i64 @atomicrmw_nand_i64_acq_rel(ptr %a, i64 %b) nounwind {
; LA64-LAMCAS-NEXT: bne $a0, $a3, .LBB77_1
; LA64-LAMCAS-NEXT: # %bb.2: # %atomicrmw.end
; LA64-LAMCAS-NEXT: ret
+;
+; LA64-LAMCAS-LAM-BH-LABEL: atomicrmw_nand_i64_acq_rel:
+; LA64-LAMCAS-LAM-BH: # %bb.0:
+; LA64-LAMCAS-LAM-BH-NEXT: move $a2, $a0
+; LA64-LAMCAS-LAM-BH-NEXT: ld.d $a0, $a0, 0
+; LA64-LAMCAS-LAM-BH-NEXT: .p2align 4, , 16
+; LA64-LAMCAS-LAM-BH-NEXT: .LBB77_1: # %atomicrmw.start
+; LA64-LAMCAS-LAM-BH-NEXT: # =>This Inner Loop Header: Depth=1
+; LA64-LAMCAS-LAM-BH-NEXT: move $a3, $a0
+; LA64-LAMCAS-LAM-BH-NEXT: and $a4, $a0, $a1
+; LA64-LAMCAS-LAM-BH-NEXT: nor $a4, $a4, $zero
+; LA64-LAMCAS-LAM-BH-NEXT: amcas_db.d $a0, $a4, $a2
+; LA64-LAMCAS-LAM-BH-NEXT: bne $a0, $a3, .LBB77_1
+; LA64-LAMCAS-LAM-BH-NEXT: # %bb.2: # %atomicrmw.end
+; LA64-LAMCAS-LAM-BH-NEXT: ret
%1 = atomicrmw nand ptr %a, i64 %b acq_rel
ret i64 %1
}
@@ -3040,17 +3972,29 @@ define i8 @atomicrmw_and_i8_acq_rel(ptr %a, i8 %b) nounwind {
;
; LA64-LAMCAS-LABEL: atomicrmw_and_i8_acq_rel:
; LA64-LAMCAS: # %bb.0:
-; LA64-LAMCAS-NEXT: move $a2, $a0
-; LA64-LAMCAS-NEXT: ld.bu $a0, $a0, 0
-; LA64-LAMCAS-NEXT: .p2align 4, , 16
-; LA64-LAMCAS-NEXT: .LBB78_1: # %atomicrmw.start
-; LA64-LAMCAS-NEXT: # =>This Inner Loop Header: Depth=1
-; LA64-LAMCAS-NEXT: and $a3, $a0, $a1
-; LA64-LAMCAS-NEXT: ext.w.b $a4, $a0
-; LA64-LAMCAS-NEXT: amcas_db.b $a0, $a3, $a2
-; LA64-LAMCAS-NEXT: bne $a0, $a4, .LBB78_1
-; LA64-LAMCAS-NEXT: # %bb.2: # %atomicrmw.end
-; LA64-LAMCAS-NEXT: ret
+; LA64-LAMCAS-NEXT: slli.d $a2, $a0, 3
+; LA64-LAMCAS-NEXT: bstrins.d $a0, $zero, 1, 0
+; LA64-LAMCAS-NEXT: ori $a3, $zero, 255
+; LA64-LAMCAS-NEXT: sll.w $a3, $a3, $a2
+; LA64-LAMCAS-NEXT: andi $a1, $a1, 255
+; LA64-LAMCAS-NEXT: sll.w $a1, $a1, $a2
+; LA64-LAMCAS-NEXT: orn $a1, $a1, $a3
+; LA64-LAMCAS-NEXT: amand_db.w $a3, $a1, $a0
+; LA64-LAMCAS-NEXT: srl.w $a0, $a3, $a2
+; LA64-LAMCAS-NEXT: ret
+;
+; LA64-LAMCAS-LAM-BH-LABEL: atomicrmw_and_i8_acq_rel:
+; LA64-LAMCAS-LAM-BH: # %bb.0:
+; LA64-LAMCAS-LAM-BH-NEXT: slli.d $a2, $a0, 3
+; LA64-LAMCAS-LAM-BH-NEXT: bstrins.d $a0, $zero, 1, 0
+; LA64-LAMCAS-LAM-BH-NEXT: ori $a3, $zero, 255
+; LA64-LAMCAS-LAM-BH-NEXT: sll.w $a3, $a3, $a2
+; LA64-LAMCAS-LAM-BH-NEXT: andi $a1, $a1, 255
+; LA64-LAMCAS-LAM-BH-NEXT: sll.w $a1, $a1, $a2
+; LA64-LAMCAS-LAM-BH-NEXT: orn $a1, $a1, $a3
+; LA64-LAMCAS-LAM-BH-NEXT: amand_db.w $a3, $a1, $a0
+; LA64-LAMCAS-LAM-BH-NEXT: srl.w $a0, $a3, $a2
+; LA64-LAMCAS-LAM-BH-NEXT: ret
%1 = atomicrmw and ptr %a, i8 %b acq_rel
ret i8 %1
}
@@ -3072,17 +4016,31 @@ define i16 @atomicrmw_and_i16_acq_rel(ptr %a, i16 %b) nounwind {
;
; LA64-LAMCAS-LABEL: atomicrmw_and_i16_acq_rel:
; LA64-LAMCAS: # %bb.0:
-; LA64-LAMCAS-NEXT: move $a2, $a0
-; LA64-LAMCAS-NEXT: ld.hu $a0, $a0, 0
-; LA64-LAMCAS-NEXT: .p2align 4, , 16
-; LA64-LAMCAS-NEXT: .LBB79_1: # %atomicrmw.start
-; LA64-LAMCAS-NEXT: # =>This Inner Loop Header: Depth=1
-; LA64-LAMCAS-NEXT: and $a3, $a0, $a1
-; LA64-LAMCAS-NEXT: ext.w.h $a4, $a0
-; LA64-LAMCAS-NEXT: amcas_db.h $a0, $a3, $a2
-; LA64-LAMCAS-NEXT: bne $a0, $a4, .LBB79_1
-; LA64-LAMCAS-NEXT: # %bb.2: # %atomicrmw.end
-; LA64-LAMCAS-NEXT: ret
+; LA64-LAMCAS-NEXT: slli.d $a2, $a0, 3
+; LA64-LAMCAS-NEXT: bstrins.d $a0, $zero, 1, 0
+; LA64-LAMCAS-NEXT: lu12i.w $a3, 15
+; LA64-LAMCAS-NEXT: ori $a3, $a3, 4095
+; LA64-LAMCAS-NEXT: sll.w $a3, $a3, $a2
+; LA64-LAMCAS-NEXT: bstrpick.d $a1, $a1, 15, 0
+; LA64-LAMCAS-NEXT: sll.w $a1, $a1, $a2
+; LA64-LAMCAS-NEXT: orn $a1, $a1, $a3
+; LA64-LAMCAS-NEXT: amand_db.w $a3, $a1, $a0
+; LA64-LAMCAS-NEXT: srl.w $a0, $a3, $a2
+; LA64-LAMCAS-NEXT: ret
+;
+; LA64-LAMCAS-LAM-BH-LABEL: atomicrmw_and_i16_acq_rel:
+; LA64-LAMCAS-LAM-BH: # %bb.0:
+; LA64-LAMCAS-LAM-BH-NEXT: slli.d $a2, $a0, 3
+; LA64-LAMCAS-LAM-BH-NEXT: bstrins.d $a0, $zero, 1, 0
+; LA64-LAMCAS-LAM-BH-NEXT: lu12i.w $a3, 15
+; LA64-LAMCAS-LAM-BH-NEXT: ori $a3, $a3, 4095
+; LA64-LAMCAS-LAM-BH-NEXT: sll.w $a3, $a3, $a2
+; LA64-LAMCAS-LAM-BH-NEXT: bstrpick.d $a1, $a1, 15, 0
+; LA64-LAMCAS-LAM-BH-NEXT: sll.w $a1, $a1, $a2
+; LA64-LAMCAS-LAM-BH-NEXT: orn $a1, $a1, $a3
+; LA64-LAMCAS-LAM-BH-NEXT: amand_db.w $a3, $a1, $a0
+; LA64-LAMCAS-LAM-BH-NEXT: srl.w $a0, $a3, $a2
+; LA64-LAMCAS-LAM-BH-NEXT: ret
%1 = atomicrmw and ptr %a, i16 %b acq_rel
ret i16 %1
@@ -3101,17 +4059,23 @@ define i8 @atomicrmw_or_i8_acq_rel(ptr %a, i8 %b) nounwind {
;
; LA64-LAMCAS-LABEL: atomicrmw_or_i8_acq_rel:
; LA64-LAMCAS: # %bb.0:
-; LA64-LAMCAS-NEXT: move $a2, $a0
-; LA64-LAMCAS-NEXT: ld.bu $a0, $a0, 0
-; LA64-LAMCAS-NEXT: .p2align 4, , 16
-; LA64-LAMCAS-NEXT: .LBB80_1: # %atomicrmw.start
-; LA64-LAMCAS-NEXT: # =>This Inner Loop Header: Depth=1
-; LA64-LAMCAS-NEXT: or $a3, $a0, $a1
-; LA64-LAMCAS-NEXT: ext.w.b $a4, $a0
-; LA64-LAMCAS-NEXT: amcas_db.b $a0, $a3, $a2
-; LA64-LAMCAS-NEXT: bne $a0, $a4, .LBB80_1
-; LA64-LAMCAS-NEXT: # %bb.2: # %atomicrmw.end
-; LA64-LAMCAS-NEXT: ret
+; LA64-LAMCAS-NEXT: slli.d $a2, $a0, 3
+; LA64-LAMCAS-NEXT: bstrins.d $a0, $zero, 1, 0
+; LA64-LAMCAS-NEXT: andi $a1, $a1, 255
+; LA64-LAMCAS-NEXT: sll.w $a1, $a1, $a2
+; LA64-LAMCAS-NEXT: amor_db.w $a3, $a1, $a0
+; LA64-LAMCAS-NEXT: srl.w $a0, $a3, $a2
+; LA64-LAMCAS-NEXT: ret
+;
+; LA64-LAMCAS-LAM-BH-LABEL: atomicrmw_or_i8_acq_rel:
+; LA64-LAMCAS-LAM-BH: # %bb.0:
+; LA64-LAMCAS-LAM-BH-NEXT: slli.d $a2, $a0, 3
+; LA64-LAMCAS-LAM-BH-NEXT: bstrins.d $a0, $zero, 1, 0
+; LA64-LAMCAS-LAM-BH-NEXT: andi $a1, $a1, 255
+; LA64-LAMCAS-LAM-BH-NEXT: sll.w $a1, $a1, $a2
+; LA64-LAMCAS-LAM-BH-NEXT: amor_db.w $a3, $a1, $a0
+; LA64-LAMCAS-LAM-BH-NEXT: srl.w $a0, $a3, $a2
+; LA64-LAMCAS-LAM-BH-NEXT: ret
%1 = atomicrmw or ptr %a, i8 %b acq_rel
ret i8 %1
}
@@ -3129,17 +4093,23 @@ define i16 @atomicrmw_or_i16_acq_rel(ptr %a, i16 %b) nounwind {
;
; LA64-LAMCAS-LABEL: atomicrmw_or_i16_acq_rel:
; LA64-LAMCAS: # %bb.0:
-; LA64-LAMCAS-NEXT: move $a2, $a0
-; LA64-LAMCAS-NEXT: ld.hu $a0, $a0, 0
-; LA64-LAMCAS-NEXT: .p2align 4, , 16
-; LA64-LAMCAS-NEXT: .LBB81_1: # %atomicrmw.start
-; LA64-LAMCAS-NEXT: # =>This Inner Loop Header: Depth=1
-; LA64-LAMCAS-NEXT: or $a3, $a0, $a1
-; LA64-LAMCAS-NEXT: ext.w.h $a4, $a0
-; LA64-LAMCAS-NEXT: amcas_db.h $a0, $a3, $a2
-; LA64-LAMCAS-NEXT: bne $a0, $a4, .LBB81_1
-; LA64-LAMCAS-NEXT: # %bb.2: # %atomicrmw.end
-; LA64-LAMCAS-NEXT: ret
+; LA64-LAMCAS-NEXT: slli.d $a2, $a0, 3
+; LA64-LAMCAS-NEXT: bstrins.d $a0, $zero, 1, 0
+; LA64-LAMCAS-NEXT: bstrpick.d $a1, $a1, 15, 0
+; LA64-LAMCAS-NEXT: sll.w $a1, $a1, $a2
+; LA64-LAMCAS-NEXT: amor_db.w $a3, $a1, $a0
+; LA64-LAMCAS-NEXT: srl.w $a0, $a3, $a2
+; LA64-LAMCAS-NEXT: ret
+;
+; LA64-LAMCAS-LAM-BH-LABEL: atomicrmw_or_i16_acq_rel:
+; LA64-LAMCAS-LAM-BH: # %bb.0:
+; LA64-LAMCAS-LAM-BH-NEXT: slli.d $a2, $a0, 3
+; LA64-LAMCAS-LAM-BH-NEXT: bstrins.d $a0, $zero, 1, 0
+; LA64-LAMCAS-LAM-BH-NEXT: bstrpick.d $a1, $a1, 15, 0
+; LA64-LAMCAS-LAM-BH-NEXT: sll.w $a1, $a1, $a2
+; LA64-LAMCAS-LAM-BH-NEXT: amor_db.w $a3, $a1, $a0
+; LA64-LAMCAS-LAM-BH-NEXT: srl.w $a0, $a3, $a2
+; LA64-LAMCAS-LAM-BH-NEXT: ret
%1 = atomicrmw or ptr %a, i16 %b acq_rel
ret i16 %1
@@ -3158,17 +4128,23 @@ define i8 @atomicrmw_xor_i8_acq_rel(ptr %a, i8 %b) nounwind {
;
; LA64-LAMCAS-LABEL: atomicrmw_xor_i8_acq_rel:
; LA64-LAMCAS: # %bb.0:
-; LA64-LAMCAS-NEXT: move $a2, $a0
-; LA64-LAMCAS-NEXT: ld.bu $a0, $a0, 0
-; LA64-LAMCAS-NEXT: .p2align 4, , 16
-; LA64-LAMCAS-NEXT: .LBB82_1: # %atomicrmw.start
-; LA64-LAMCAS-NEXT: # =>This Inner Loop Header: Depth=1
-; LA64-LAMCAS-NEXT: xor $a3, $a0, $a1
-; LA64-LAMCAS-NEXT: ext.w.b $a4, $a0
-; LA64-LAMCAS-NEXT: amcas_db.b $a0, $a3, $a2
-; LA64-LAMCAS-NEXT: bne $a0, $a4, .LBB82_1
-; LA64-LAMCAS-NEXT: # %bb.2: # %atomicrmw.end
-; LA64-LAMCAS-NEXT: ret
+; LA64-LAMCAS-NEXT: slli.d $a2, $a0, 3
+; LA64-LAMCAS-NEXT: bstrins.d $a0, $zero, 1, 0
+; LA64-LAMCAS-NEXT: andi $a1, $a1, 255
+; LA64-LAMCAS-NEXT: sll.w $a1, $a1, $a2
+; LA64-LAMCAS-NEXT: amxor_db.w $a3, $a1, $a0
+; LA64-LAMCAS-NEXT: srl.w $a0, $a3, $a2
+; LA64-LAMCAS-NEXT: ret
+;
+; LA64-LAMCAS-LAM-BH-LABEL: atomicrmw_xor_i8_acq_rel:
+; LA64-LAMCAS-LAM-BH: # %bb.0:
+; LA64-LAMCAS-LAM-BH-NEXT: slli.d $a2, $a0, 3
+; LA64-LAMCAS-LAM-BH-NEXT: bstrins.d $a0, $zero, 1, 0
+; LA64-LAMCAS-LAM-BH-NEXT: andi $a1, $a1, 255
+; LA64-LAMCAS-LAM-BH-NEXT: sll.w $a1, $a1, $a2
+; LA64-LAMCAS-LAM-BH-NEXT: amxor_db.w $a3, $a1, $a0
+; LA64-LAMCAS-LAM-BH-NEXT: srl.w $a0, $a3, $a2
+; LA64-LAMCAS-LAM-BH-NEXT: ret
%1 = atomicrmw xor ptr %a, i8 %b acq_rel
ret i8 %1
}
@@ -3186,17 +4162,23 @@ define i16 @atomicrmw_xor_i16_acq_rel(ptr %a, i16 %b) nounwind {
;
; LA64-LAMCAS-LABEL: atomicrmw_xor_i16_acq_rel:
; LA64-LAMCAS: # %bb.0:
-; LA64-LAMCAS-NEXT: move $a2, $a0
-; LA64-LAMCAS-NEXT: ld.hu $a0, $a0, 0
-; LA64-LAMCAS-NEXT: .p2align 4, , 16
-; LA64-LAMCAS-NEXT: .LBB83_1: # %atomicrmw.start
-; LA64-LAMCAS-NEXT: # =>This Inner Loop Header: Depth=1
-; LA64-LAMCAS-NEXT: xor $a3, $a0, $a1
-; LA64-LAMCAS-NEXT: ext.w.h $a4, $a0
-; LA64-LAMCAS-NEXT: amcas_db.h $a0, $a3, $a2
-; LA64-LAMCAS-NEXT: bne $a0, $a4, .LBB83_1
-; LA64-LAMCAS-NEXT: # %bb.2: # %atomicrmw.end
-; LA64-LAMCAS-NEXT: ret
+; LA64-LAMCAS-NEXT: slli.d $a2, $a0, 3
+; LA64-LAMCAS-NEXT: bstrins.d $a0, $zero, 1, 0
+; LA64-LAMCAS-NEXT: bstrpick.d $a1, $a1, 15, 0
+; LA64-LAMCAS-NEXT: sll.w $a1, $a1, $a2
+; LA64-LAMCAS-NEXT: amxor_db.w $a3, $a1, $a0
+; LA64-LAMCAS-NEXT: srl.w $a0, $a3, $a2
+; LA64-LAMCAS-NEXT: ret
+;
+; LA64-LAMCAS-LAM-BH-LABEL: atomicrmw_xor_i16_acq_rel:
+; LA64-LAMCAS-LAM-BH: # %bb.0:
+; LA64-LAMCAS-LAM-BH-NEXT: slli.d $a2, $a0, 3
+; LA64-LAMCAS-LAM-BH-NEXT: bstrins.d $a0, $zero, 1, 0
+; LA64-LAMCAS-LAM-BH-NEXT: bstrpick.d $a1, $a1, 15, 0
+; LA64-LAMCAS-LAM-BH-NEXT: sll.w $a1, $a1, $a2
+; LA64-LAMCAS-LAM-BH-NEXT: amxor_db.w $a3, $a1, $a0
+; LA64-LAMCAS-LAM-BH-NEXT: srl.w $a0, $a3, $a2
+; LA64-LAMCAS-LAM-BH-NEXT: ret
%1 = atomicrmw xor ptr %a, i16 %b acq_rel
ret i16 %1
@@ -3235,6 +4217,12 @@ define i8 @atomicrmw_xchg_i8_seq_cst(ptr %a, i8 %b) nounwind {
; LA64-LAMCAS-NEXT: # %bb.2: # %atomicrmw.end
; LA64-LAMCAS-NEXT: move $a0, $a2
; LA64-LAMCAS-NEXT: ret
+;
+; LA64-LAMCAS-LAM-BH-LABEL: atomicrmw_xchg_i8_seq_cst:
+; LA64-LAMCAS-LAM-BH: # %bb.0:
+; LA64-LAMCAS-LAM-BH-NEXT: amswap_db.b $a2, $a1, $a0
+; LA64-LAMCAS-LAM-BH-NEXT: move $a0, $a2
+; LA64-LAMCAS-LAM-BH-NEXT: ret
%1 = atomicrmw xchg ptr %a, i8 %b seq_cst
ret i8 %1
}
@@ -3263,6 +4251,12 @@ define i8 @atomicrmw_xchg_0_i8_seq_cst(ptr %a) nounwind {
; LA64-LAMCAS-NEXT: bne $a0, $a2, .LBB85_1
; LA64-LAMCAS-NEXT: # %bb.2: # %atomicrmw.end
; LA64-LAMCAS-NEXT: ret
+;
+; LA64-LAMCAS-LAM-BH-LABEL: atomicrmw_xchg_0_i8_seq_cst:
+; LA64-LAMCAS-LAM-BH: # %bb.0:
+; LA64-LAMCAS-LAM-BH-NEXT: amswap_db.b $a1, $zero, $a0
+; LA64-LAMCAS-LAM-BH-NEXT: move $a0, $a1
+; LA64-LAMCAS-LAM-BH-NEXT: ret
%1 = atomicrmw xchg ptr %a, i8 0 seq_cst
ret i8 %1
}
@@ -3291,6 +4285,13 @@ define i8 @atomicrmw_xchg_minus_1_i8_seq_cst(ptr %a) nounwind {
; LA64-LAMCAS-NEXT: bne $a0, $a3, .LBB86_1
; LA64-LAMCAS-NEXT: # %bb.2: # %atomicrmw.end
; LA64-LAMCAS-NEXT: ret
+;
+; LA64-LAMCAS-LAM-BH-LABEL: atomicrmw_xchg_minus_1_i8_seq_cst:
+; LA64-LAMCAS-LAM-BH: # %bb.0:
+; LA64-LAMCAS-LAM-BH-NEXT: addi.w $a2, $zero, -1
+; LA64-LAMCAS-LAM-BH-NEXT: amswap_db.b $a1, $a2, $a0
+; LA64-LAMCAS-LAM-BH-NEXT: move $a0, $a1
+; LA64-LAMCAS-LAM-BH-NEXT: ret
%1 = atomicrmw xchg ptr %a, i8 -1 seq_cst
ret i8 %1
}
@@ -3329,6 +4330,12 @@ define i16 @atomicrmw_xchg_i16_seq_cst(ptr %a, i16 %b) nounwind {
; LA64-LAMCAS-NEXT: # %bb.2: # %atomicrmw.end
; LA64-LAMCAS-NEXT: move $a0, $a2
; LA64-LAMCAS-NEXT: ret
+;
+; LA64-LAMCAS-LAM-BH-LABEL: atomicrmw_xchg_i16_seq_cst:
+; LA64-LAMCAS-LAM-BH: # %bb.0:
+; LA64-LAMCAS-LAM-BH-NEXT: amswap_db.h $a2, $a1, $a0
+; LA64-LAMCAS-LAM-BH-NEXT: move $a0, $a2
+; LA64-LAMCAS-LAM-BH-NEXT: ret
%1 = atomicrmw xchg ptr %a, i16 %b seq_cst
ret i16 %1
}
@@ -3358,6 +4365,12 @@ define i16 @atomicrmw_xchg_0_i16_seq_cst(ptr %a) nounwind {
; LA64-LAMCAS-NEXT: bne $a0, $a2, .LBB88_1
; LA64-LAMCAS-NEXT: # %bb.2: # %atomicrmw.end
; LA64-LAMCAS-NEXT: ret
+;
+; LA64-LAMCAS-LAM-BH-LABEL: atomicrmw_xchg_0_i16_seq_cst:
+; LA64-LAMCAS-LAM-BH: # %bb.0:
+; LA64-LAMCAS-LAM-BH-NEXT: amswap_db.h $a1, $zero, $a0
+; LA64-LAMCAS-LAM-BH-NEXT: move $a0, $a1
+; LA64-LAMCAS-LAM-BH-NEXT: ret
%1 = atomicrmw xchg ptr %a, i16 0 seq_cst
ret i16 %1
}
@@ -3387,6 +4400,13 @@ define i16 @atomicrmw_xchg_minus_1_i16_seq_cst(ptr %a) nounwind {
; LA64-LAMCAS-NEXT: bne $a0, $a3, .LBB89_1
; LA64-LAMCAS-NEXT: # %bb.2: # %atomicrmw.end
; LA64-LAMCAS-NEXT: ret
+;
+; LA64-LAMCAS-LAM-BH-LABEL: atomicrmw_xchg_minus_1_i16_seq_cst:
+; LA64-LAMCAS-LAM-BH: # %bb.0:
+; LA64-LAMCAS-LAM-BH-NEXT: addi.w $a2, $zero, -1
+; LA64-LAMCAS-LAM-BH-NEXT: amswap_db.h $a1, $a2, $a0
+; LA64-LAMCAS-LAM-BH-NEXT: move $a0, $a1
+; LA64-LAMCAS-LAM-BH-NEXT: ret
%1 = atomicrmw xchg ptr %a, i16 -1 seq_cst
ret i16 %1
@@ -3426,6 +4446,12 @@ define i8 @atomicrmw_add_i8_seq_cst(ptr %a, i8 %b) nounwind {
; LA64-LAMCAS-NEXT: bne $a0, $a4, .LBB90_1
; LA64-LAMCAS-NEXT: # %bb.2: # %atomicrmw.end
; LA64-LAMCAS-NEXT: ret
+;
+; LA64-LAMCAS-LAM-BH-LABEL: atomicrmw_add_i8_seq_cst:
+; LA64-LAMCAS-LAM-BH: # %bb.0:
+; LA64-LAMCAS-LAM-BH-NEXT: amadd_db.b $a2, $a1, $a0
+; LA64-LAMCAS-LAM-BH-NEXT: move $a0, $a2
+; LA64-LAMCAS-LAM-BH-NEXT: ret
%1 = atomicrmw add ptr %a, i8 %b seq_cst
ret i8 %1
}
@@ -3465,6 +4491,12 @@ define i16 @atomicrmw_add_i16_seq_cst(ptr %a, i16 %b) nounwind {
; LA64-LAMCAS-NEXT: bne $a0, $a4, .LBB91_1
; LA64-LAMCAS-NEXT: # %bb.2: # %atomicrmw.end
; LA64-LAMCAS-NEXT: ret
+;
+; LA64-LAMCAS-LAM-BH-LABEL: atomicrmw_add_i16_seq_cst:
+; LA64-LAMCAS-LAM-BH: # %bb.0:
+; LA64-LAMCAS-LAM-BH-NEXT: amadd_db.h $a2, $a1, $a0
+; LA64-LAMCAS-LAM-BH-NEXT: move $a0, $a2
+; LA64-LAMCAS-LAM-BH-NEXT: ret
%1 = atomicrmw add ptr %a, i16 %b seq_cst
ret i16 %1
@@ -3504,6 +4536,13 @@ define i8 @atomicrmw_sub_i8_seq_cst(ptr %a, i8 %b) nounwind {
; LA64-LAMCAS-NEXT: bne $a0, $a4, .LBB92_1
; LA64-LAMCAS-NEXT: # %bb.2: # %atomicrmw.end
; LA64-LAMCAS-NEXT: ret
+;
+; LA64-LAMCAS-LAM-BH-LABEL: atomicrmw_sub_i8_seq_cst:
+; LA64-LAMCAS-LAM-BH: # %bb.0:
+; LA64-LAMCAS-LAM-BH-NEXT: sub.w $a2, $zero, $a1
+; LA64-LAMCAS-LAM-BH-NEXT: amadd_db.b $a1, $a2, $a0
+; LA64-LAMCAS-LAM-BH-NEXT: move $a0, $a1
+; LA64-LAMCAS-LAM-BH-NEXT: ret
%1 = atomicrmw sub ptr %a, i8 %b seq_cst
ret i8 %1
}
@@ -3543,6 +4582,13 @@ define i16 @atomicrmw_sub_i16_seq_cst(ptr %a, i16 %b) nounwind {
; LA64-LAMCAS-NEXT: bne $a0, $a4, .LBB93_1
; LA64-LAMCAS-NEXT: # %bb.2: # %atomicrmw.end
; LA64-LAMCAS-NEXT: ret
+;
+; LA64-LAMCAS-LAM-BH-LABEL: atomicrmw_sub_i16_seq_cst:
+; LA64-LAMCAS-LAM-BH: # %bb.0:
+; LA64-LAMCAS-LAM-BH-NEXT: sub.w $a2, $zero, $a1
+; LA64-LAMCAS-LAM-BH-NEXT: amadd_db.h $a1, $a2, $a0
+; LA64-LAMCAS-LAM-BH-NEXT: move $a0, $a1
+; LA64-LAMCAS-LAM-BH-NEXT: ret
%1 = atomicrmw sub ptr %a, i16 %b seq_cst
ret i16 %1
@@ -3591,6 +4637,25 @@ define i8 @atomicrmw_umax_i8_seq_cst(ptr %a, i8 %b) nounwind {
; LA64-LAMCAS-NEXT: bne $a0, $a5, .LBB94_1
; LA64-LAMCAS-NEXT: # %bb.2: # %atomicrmw.end
; LA64-LAMCAS-NEXT: ret
+;
+; LA64-LAMCAS-LAM-BH-LABEL: atomicrmw_umax_i8_seq_cst:
+; LA64-LAMCAS-LAM-BH: # %bb.0:
+; LA64-LAMCAS-LAM-BH-NEXT: move $a2, $a0
+; LA64-LAMCAS-LAM-BH-NEXT: ld.bu $a0, $a0, 0
+; LA64-LAMCAS-LAM-BH-NEXT: andi $a3, $a1, 255
+; LA64-LAMCAS-LAM-BH-NEXT: .p2align 4, , 16
+; LA64-LAMCAS-LAM-BH-NEXT: .LBB94_1: # %atomicrmw.start
+; LA64-LAMCAS-LAM-BH-NEXT: # =>This Inner Loop Header: Depth=1
+; LA64-LAMCAS-LAM-BH-NEXT: andi $a4, $a0, 255
+; LA64-LAMCAS-LAM-BH-NEXT: sltu $a4, $a3, $a4
+; LA64-LAMCAS-LAM-BH-NEXT: masknez $a5, $a1, $a4
+; LA64-LAMCAS-LAM-BH-NEXT: maskeqz $a4, $a0, $a4
+; LA64-LAMCAS-LAM-BH-NEXT: or $a4, $a4, $a5
+; LA64-LAMCAS-LAM-BH-NEXT: ext.w.b $a5, $a0
+; LA64-LAMCAS-LAM-BH-NEXT: amcas_db.b $a0, $a4, $a2
+; LA64-LAMCAS-LAM-BH-NEXT: bne $a0, $a5, .LBB94_1
+; LA64-LAMCAS-LAM-BH-NEXT: # %bb.2: # %atomicrmw.end
+; LA64-LAMCAS-LAM-BH-NEXT: ret
%1 = atomicrmw umax ptr %a, i8 %b seq_cst
ret i8 %1
}
@@ -3639,6 +4704,25 @@ define i16 @atomicrmw_umax_i16_seq_cst(ptr %a, i16 %b) nounwind {
; LA64-LAMCAS-NEXT: bne $a0, $a5, .LBB95_1
; LA64-LAMCAS-NEXT: # %bb.2: # %atomicrmw.end
; LA64-LAMCAS-NEXT: ret
+;
+; LA64-LAMCAS-LAM-BH-LABEL: atomicrmw_umax_i16_seq_cst:
+; LA64-LAMCAS-LAM-BH: # %bb.0:
+; LA64-LAMCAS-LAM-BH-NEXT: move $a2, $a0
+; LA64-LAMCAS-LAM-BH-NEXT: ld.hu $a0, $a0, 0
+; LA64-LAMCAS-LAM-BH-NEXT: bstrpick.d $a3, $a1, 15, 0
+; LA64-LAMCAS-LAM-BH-NEXT: .p2align 4, , 16
+; LA64-LAMCAS-LAM-BH-NEXT: .LBB95_1: # %atomicrmw.start
+; LA64-LAMCAS-LAM-BH-NEXT: # =>This Inner Loop Header: Depth=1
+; LA64-LAMCAS-LAM-BH-NEXT: bstrpick.d $a4, $a0, 15, 0
+; LA64-LAMCAS-LAM-BH-NEXT: sltu $a4, $a3, $a4
+; LA64-LAMCAS-LAM-BH-NEXT: masknez $a5, $a1, $a4
+; LA64-LAMCAS-LAM-BH-NEXT: maskeqz $a4, $a0, $a4
+; LA64-LAMCAS-LAM-BH-NEXT: or $a4, $a4, $a5
+; LA64-LAMCAS-LAM-BH-NEXT: ext.w.h $a5, $a0
+; LA64-LAMCAS-LAM-BH-NEXT: amcas_db.h $a0, $a4, $a2
+; LA64-LAMCAS-LAM-BH-NEXT: bne $a0, $a5, .LBB95_1
+; LA64-LAMCAS-LAM-BH-NEXT: # %bb.2: # %atomicrmw.end
+; LA64-LAMCAS-LAM-BH-NEXT: ret
%1 = atomicrmw umax ptr %a, i16 %b seq_cst
ret i16 %1
}
@@ -3687,6 +4771,26 @@ define i8 @atomicrmw_umin_i8_seq_cst(ptr %a, i8 %b) nounwind {
; LA64-LAMCAS-NEXT: bne $a0, $a5, .LBB96_1
; LA64-LAMCAS-NEXT: # %bb.2: # %atomicrmw.end
; LA64-LAMCAS-NEXT: ret
+;
+; LA64-LAMCAS-LAM-BH-LABEL: atomicrmw_umin_i8_seq_cst:
+; LA64-LAMCAS-LAM-BH: # %bb.0:
+; LA64-LAMCAS-LAM-BH-NEXT: move $a2, $a0
+; LA64-LAMCAS-LAM-BH-NEXT: ld.bu $a0, $a0, 0
+; LA64-LAMCAS-LAM-BH-NEXT: andi $a3, $a1, 255
+; LA64-LAMCAS-LAM-BH-NEXT: .p2align 4, , 16
+; LA64-LAMCAS-LAM-BH-NEXT: .LBB96_1: # %atomicrmw.start
+; LA64-LAMCAS-LAM-BH-NEXT: # =>This Inner Loop Header: Depth=1
+; LA64-LAMCAS-LAM-BH-NEXT: andi $a4, $a0, 255
+; LA64-LAMCAS-LAM-BH-NEXT: sltu $a4, $a3, $a4
+; LA64-LAMCAS-LAM-BH-NEXT: xori $a4, $a4, 1
+; LA64-LAMCAS-LAM-BH-NEXT: masknez $a5, $a1, $a4
+; LA64-LAMCAS-LAM-BH-NEXT: maskeqz $a4, $a0, $a4
+; LA64-LAMCAS-LAM-BH-NEXT: or $a4, $a4, $a5
+; LA64-LAMCAS-LAM-BH-NEXT: ext.w.b $a5, $a0
+; LA64-LAMCAS-LAM-BH-NEXT: amcas_db.b $a0, $a4, $a2
+; LA64-LAMCAS-LAM-BH-NEXT: bne $a0, $a5, .LBB96_1
+; LA64-LAMCAS-LAM-BH-NEXT: # %bb.2: # %atomicrmw.end
+; LA64-LAMCAS-LAM-BH-NEXT: ret
%1 = atomicrmw umin ptr %a, i8 %b seq_cst
ret i8 %1
}
@@ -3736,6 +4840,26 @@ define i16 @atomicrmw_umin_i16_seq_cst(ptr %a, i16 %b) nounwind {
; LA64-LAMCAS-NEXT: bne $a0, $a5, .LBB97_1
; LA64-LAMCAS-NEXT: # %bb.2: # %atomicrmw.end
; LA64-LAMCAS-NEXT: ret
+;
+; LA64-LAMCAS-LAM-BH-LABEL: atomicrmw_umin_i16_seq_cst:
+; LA64-LAMCAS-LAM-BH: # %bb.0:
+; LA64-LAMCAS-LAM-BH-NEXT: move $a2, $a0
+; LA64-LAMCAS-LAM-BH-NEXT: ld.hu $a0, $a0, 0
+; LA64-LAMCAS-LAM-BH-NEXT: bstrpick.d $a3, $a1, 15, 0
+; LA64-LAMCAS-LAM-BH-NEXT: .p2align 4, , 16
+; LA64-LAMCAS-LAM-BH-NEXT: .LBB97_1: # %atomicrmw.start
+; LA64-LAMCAS-LAM-BH-NEXT: # =>This Inner Loop Header: Depth=1
+; LA64-LAMCAS-LAM-BH-NEXT: bstrpick.d $a4, $a0, 15, 0
+; LA64-LAMCAS-LAM-BH-NEXT: sltu $a4, $a3, $a4
+; LA64-LAMCAS-LAM-BH-NEXT: xori $a4, $a4, 1
+; LA64-LAMCAS-LAM-BH-NEXT: masknez $a5, $a1, $a4
+; LA64-LAMCAS-LAM-BH-NEXT: maskeqz $a4, $a0, $a4
+; LA64-LAMCAS-LAM-BH-NEXT: or $a4, $a4, $a5
+; LA64-LAMCAS-LAM-BH-NEXT: ext.w.h $a5, $a0
+; LA64-LAMCAS-LAM-BH-NEXT: amcas_db.h $a0, $a4, $a2
+; LA64-LAMCAS-LAM-BH-NEXT: bne $a0, $a5, .LBB97_1
+; LA64-LAMCAS-LAM-BH-NEXT: # %bb.2: # %atomicrmw.end
+; LA64-LAMCAS-LAM-BH-NEXT: ret
%1 = atomicrmw umin ptr %a, i16 %b seq_cst
ret i16 %1
}
@@ -3786,6 +4910,24 @@ define i8 @atomicrmw_max_i8_seq_cst(ptr %a, i8 %b) nounwind {
; LA64-LAMCAS-NEXT: bne $a0, $a4, .LBB98_1
; LA64-LAMCAS-NEXT: # %bb.2: # %atomicrmw.end
; LA64-LAMCAS-NEXT: ret
+;
+; LA64-LAMCAS-LAM-BH-LABEL: atomicrmw_max_i8_seq_cst:
+; LA64-LAMCAS-LAM-BH: # %bb.0:
+; LA64-LAMCAS-LAM-BH-NEXT: move $a2, $a0
+; LA64-LAMCAS-LAM-BH-NEXT: ld.bu $a0, $a0, 0
+; LA64-LAMCAS-LAM-BH-NEXT: ext.w.b $a3, $a1
+; LA64-LAMCAS-LAM-BH-NEXT: .p2align 4, , 16
+; LA64-LAMCAS-LAM-BH-NEXT: .LBB98_1: # %atomicrmw.start
+; LA64-LAMCAS-LAM-BH-NEXT: # =>This Inner Loop Header: Depth=1
+; LA64-LAMCAS-LAM-BH-NEXT: ext.w.b $a4, $a0
+; LA64-LAMCAS-LAM-BH-NEXT: slt $a5, $a3, $a4
+; LA64-LAMCAS-LAM-BH-NEXT: masknez $a6, $a1, $a5
+; LA64-LAMCAS-LAM-BH-NEXT: maskeqz $a5, $a0, $a5
+; LA64-LAMCAS-LAM-BH-NEXT: or $a5, $a5, $a6
+; LA64-LAMCAS-LAM-BH-NEXT: amcas_db.b $a0, $a5, $a2
+; LA64-LAMCAS-LAM-BH-NEXT: bne $a0, $a4, .LBB98_1
+; LA64-LAMCAS-LAM-BH-NEXT: # %bb.2: # %atomicrmw.end
+; LA64-LAMCAS-LAM-BH-NEXT: ret
%1 = atomicrmw max ptr %a, i8 %b seq_cst
ret i8 %1
}
@@ -3838,6 +4980,24 @@ define i16 @atomicrmw_max_i16_seq_cst(ptr %a, i16 %b) nounwind {
; LA64-LAMCAS-NEXT: bne $a0, $a4, .LBB99_1
; LA64-LAMCAS-NEXT: # %bb.2: # %atomicrmw.end
; LA64-LAMCAS-NEXT: ret
+;
+; LA64-LAMCAS-LAM-BH-LABEL: atomicrmw_max_i16_seq_cst:
+; LA64-LAMCAS-LAM-BH: # %bb.0:
+; LA64-LAMCAS-LAM-BH-NEXT: move $a2, $a0
+; LA64-LAMCAS-LAM-BH-NEXT: ld.hu $a0, $a0, 0
+; LA64-LAMCAS-LAM-BH-NEXT: ext.w.h $a3, $a1
+; LA64-LAMCAS-LAM-BH-NEXT: .p2align 4, , 16
+; LA64-LAMCAS-LAM-BH-NEXT: .LBB99_1: # %atomicrmw.start
+; LA64-LAMCAS-LAM-BH-NEXT: # =>This Inner Loop Header: Depth=1
+; LA64-LAMCAS-LAM-BH-NEXT: ext.w.h $a4, $a0
+; LA64-LAMCAS-LAM-BH-NEXT: slt $a5, $a3, $a4
+; LA64-LAMCAS-LAM-BH-NEXT: masknez $a6, $a1, $a5
+; LA64-LAMCAS-LAM-BH-NEXT: maskeqz $a5, $a0, $a5
+; LA64-LAMCAS-LAM-BH-NEXT: or $a5, $a5, $a6
+; LA64-LAMCAS-LAM-BH-NEXT: amcas_db.h $a0, $a5, $a2
+; LA64-LAMCAS-LAM-BH-NEXT: bne $a0, $a4, .LBB99_1
+; LA64-LAMCAS-LAM-BH-NEXT: # %bb.2: # %atomicrmw.end
+; LA64-LAMCAS-LAM-BH-NEXT: ret
%1 = atomicrmw max ptr %a, i16 %b seq_cst
ret i16 %1
}
@@ -3889,6 +5049,25 @@ define i8 @atomicrmw_min_i8_seq_cst(ptr %a, i8 %b) nounwind {
; LA64-LAMCAS-NEXT: bne $a0, $a4, .LBB100_1
; LA64-LAMCAS-NEXT: # %bb.2: # %atomicrmw.end
; LA64-LAMCAS-NEXT: ret
+;
+; LA64-LAMCAS-LAM-BH-LABEL: atomicrmw_min_i8_seq_cst:
+; LA64-LAMCAS-LAM-BH: # %bb.0:
+; LA64-LAMCAS-LAM-BH-NEXT: move $a2, $a0
+; LA64-LAMCAS-LAM-BH-NEXT: ld.bu $a0, $a0, 0
+; LA64-LAMCAS-LAM-BH-NEXT: ext.w.b $a3, $a1
+; LA64-LAMCAS-LAM-BH-NEXT: .p2align 4, , 16
+; LA64-LAMCAS-LAM-BH-NEXT: .LBB100_1: # %atomicrmw.start
+; LA64-LAMCAS-LAM-BH-NEXT: # =>This Inner Loop Header: Depth=1
+; LA64-LAMCAS-LAM-BH-NEXT: ext.w.b $a4, $a0
+; LA64-LAMCAS-LAM-BH-NEXT: slt $a5, $a3, $a4
+; LA64-LAMCAS-LAM-BH-NEXT: xori $a5, $a5, 1
+; LA64-LAMCAS-LAM-BH-NEXT: masknez $a6, $a1, $a5
+; LA64-LAMCAS-LAM-BH-NEXT: maskeqz $a5, $a0, $a5
+; LA64-LAMCAS-LAM-BH-NEXT: or $a5, $a5, $a6
+; LA64-LAMCAS-LAM-BH-NEXT: amcas_db.b $a0, $a5, $a2
+; LA64-LAMCAS-LAM-BH-NEXT: bne $a0, $a4, .LBB100_1
+; LA64-LAMCAS-LAM-BH-NEXT: # %bb.2: # %atomicrmw.end
+; LA64-LAMCAS-LAM-BH-NEXT: ret
%1 = atomicrmw min ptr %a, i8 %b seq_cst
ret i8 %1
}
@@ -3942,6 +5121,25 @@ define i16 @atomicrmw_min_i16_seq_cst(ptr %a, i16 %b) nounwind {
; LA64-LAMCAS-NEXT: bne $a0, $a4, .LBB101_1
; LA64-LAMCAS-NEXT: # %bb.2: # %atomicrmw.end
; LA64-LAMCAS-NEXT: ret
+;
+; LA64-LAMCAS-LAM-BH-LABEL: atomicrmw_min_i16_seq_cst:
+; LA64-LAMCAS-LAM-BH: # %bb.0:
+; LA64-LAMCAS-LAM-BH-NEXT: move $a2, $a0
+; LA64-LAMCAS-LAM-BH-NEXT: ld.hu $a0, $a0, 0
+; LA64-LAMCAS-LAM-BH-NEXT: ext.w.h $a3, $a1
+; LA64-LAMCAS-LAM-BH-NEXT: .p2align 4, , 16
+; LA64-LAMCAS-LAM-BH-NEXT: .LBB101_1: # %atomicrmw.start
+; LA64-LAMCAS-LAM-BH-NEXT: # =>This Inner Loop Header: Depth=1
+; LA64-LAMCAS-LAM-BH-NEXT: ext.w.h $a4, $a0
+; LA64-LAMCAS-LAM-BH-NEXT: slt $a5, $a3, $a4
+; LA64-LAMCAS-LAM-BH-NEXT: xori $a5, $a5, 1
+; LA64-LAMCAS-LAM-BH-NEXT: masknez $a6, $a1, $a5
+; LA64-LAMCAS-LAM-BH-NEXT: maskeqz $a5, $a0, $a5
+; LA64-LAMCAS-LAM-BH-NEXT: or $a5, $a5, $a6
+; LA64-LAMCAS-LAM-BH-NEXT: amcas_db.h $a0, $a5, $a2
+; LA64-LAMCAS-LAM-BH-NEXT: bne $a0, $a4, .LBB101_1
+; LA64-LAMCAS-LAM-BH-NEXT: # %bb.2: # %atomicrmw.end
+; LA64-LAMCAS-LAM-BH-NEXT: ret
%1 = atomicrmw min ptr %a, i16 %b seq_cst
ret i16 %1
}
@@ -3982,6 +5180,21 @@ define i8 @atomicrmw_nand_i8_seq_cst(ptr %a, i8 %b) nounwind {
; LA64-LAMCAS-NEXT: bne $a0, $a4, .LBB102_1
; LA64-LAMCAS-NEXT: # %bb.2: # %atomicrmw.end
; LA64-LAMCAS-NEXT: ret
+;
+; LA64-LAMCAS-LAM-BH-LABEL: atomicrmw_nand_i8_seq_cst:
+; LA64-LAMCAS-LAM-BH: # %bb.0:
+; LA64-LAMCAS-LAM-BH-NEXT: move $a2, $a0
+; LA64-LAMCAS-LAM-BH-NEXT: ld.bu $a0, $a0, 0
+; LA64-LAMCAS-LAM-BH-NEXT: .p2align 4, , 16
+; LA64-LAMCAS-LAM-BH-NEXT: .LBB102_1: # %atomicrmw.start
+; LA64-LAMCAS-LAM-BH-NEXT: # =>This Inner Loop Header: Depth=1
+; LA64-LAMCAS-LAM-BH-NEXT: and $a3, $a0, $a1
+; LA64-LAMCAS-LAM-BH-NEXT: nor $a3, $a3, $zero
+; LA64-LAMCAS-LAM-BH-NEXT: ext.w.b $a4, $a0
+; LA64-LAMCAS-LAM-BH-NEXT: amcas_db.b $a0, $a3, $a2
+; LA64-LAMCAS-LAM-BH-NEXT: bne $a0, $a4, .LBB102_1
+; LA64-LAMCAS-LAM-BH-NEXT: # %bb.2: # %atomicrmw.end
+; LA64-LAMCAS-LAM-BH-NEXT: ret
%1 = atomicrmw nand ptr %a, i8 %b seq_cst
ret i8 %1
}
@@ -4023,6 +5236,21 @@ define i16 @atomicrmw_nand_i16_seq_cst(ptr %a, i16 %b) nounwind {
; LA64-LAMCAS-NEXT: bne $a0, $a4, .LBB103_1
; LA64-LAMCAS-NEXT: # %bb.2: # %atomicrmw.end
; LA64-LAMCAS-NEXT: ret
+;
+; LA64-LAMCAS-LAM-BH-LABEL: atomicrmw_nand_i16_seq_cst:
+; LA64-LAMCAS-LAM-BH: # %bb.0:
+; LA64-LAMCAS-LAM-BH-NEXT: move $a2, $a0
+; LA64-LAMCAS-LAM-BH-NEXT: ld.hu $a0, $a0, 0
+; LA64-LAMCAS-LAM-BH-NEXT: .p2align 4, , 16
+; LA64-LAMCAS-LAM-BH-NEXT: .LBB103_1: # %atomicrmw.start
+; LA64-LAMCAS-LAM-BH-NEXT: # =>This Inner Loop Header: Depth=1
+; LA64-LAMCAS-LAM-BH-NEXT: and $a3, $a0, $a1
+; LA64-LAMCAS-LAM-BH-NEXT: nor $a3, $a3, $zero
+; LA64-LAMCAS-LAM-BH-NEXT: ext.w.h $a4, $a0
+; LA64-LAMCAS-LAM-BH-NEXT: amcas_db.h $a0, $a3, $a2
+; LA64-LAMCAS-LAM-BH-NEXT: bne $a0, $a4, .LBB103_1
+; LA64-LAMCAS-LAM-BH-NEXT: # %bb.2: # %atomicrmw.end
+; LA64-LAMCAS-LAM-BH-NEXT: ret
%1 = atomicrmw nand ptr %a, i16 %b seq_cst
ret i16 %1
@@ -4055,6 +5283,21 @@ define i32 @atomicrmw_nand_i32_seq_cst(ptr %a, i32 %b) nounwind {
; LA64-LAMCAS-NEXT: bne $a0, $a3, .LBB104_1
; LA64-LAMCAS-NEXT: # %bb.2: # %atomicrmw.end
; LA64-LAMCAS-NEXT: ret
+;
+; LA64-LAMCAS-LAM-BH-LABEL: atomicrmw_nand_i32_seq_cst:
+; LA64-LAMCAS-LAM-BH: # %bb.0:
+; LA64-LAMCAS-LAM-BH-NEXT: move $a2, $a0
+; LA64-LAMCAS-LAM-BH-NEXT: ld.w $a0, $a0, 0
+; LA64-LAMCAS-LAM-BH-NEXT: .p2align 4, , 16
+; LA64-LAMCAS-LAM-BH-NEXT: .LBB104_1: # %atomicrmw.start
+; LA64-LAMCAS-LAM-BH-NEXT: # =>This Inner Loop Header: Depth=1
+; LA64-LAMCAS-LAM-BH-NEXT: move $a3, $a0
+; LA64-LAMCAS-LAM-BH-NEXT: and $a4, $a0, $a1
+; LA64-LAMCAS-LAM-BH-NEXT: nor $a4, $a4, $zero
+; LA64-LAMCAS-LAM-BH-NEXT: amcas_db.w $a0, $a4, $a2
+; LA64-LAMCAS-LAM-BH-NEXT: bne $a0, $a3, .LBB104_1
+; LA64-LAMCAS-LAM-BH-NEXT: # %bb.2: # %atomicrmw.end
+; LA64-LAMCAS-LAM-BH-NEXT: ret
%1 = atomicrmw nand ptr %a, i32 %b seq_cst
ret i32 %1
}
@@ -4086,6 +5329,21 @@ define i64 @atomicrmw_nand_i64_seq_cst(ptr %a, i64 %b) nounwind {
; LA64-LAMCAS-NEXT: bne $a0, $a3, .LBB105_1
; LA64-LAMCAS-NEXT: # %bb.2: # %atomicrmw.end
; LA64-LAMCAS-NEXT: ret
+;
+; LA64-LAMCAS-LAM-BH-LABEL: atomicrmw_nand_i64_seq_cst:
+; LA64-LAMCAS-LAM-BH: # %bb.0:
+; LA64-LAMCAS-LAM-BH-NEXT: move $a2, $a0
+; LA64-LAMCAS-LAM-BH-NEXT: ld.d $a0, $a0, 0
+; LA64-LAMCAS-LAM-BH-NEXT: .p2align 4, , 16
+; LA64-LAMCAS-LAM-BH-NEXT: .LBB105_1: # %atomicrmw.start
+; LA64-LAMCAS-LAM-BH-NEXT: # =>This Inner Loop Header: Depth=1
+; LA64-LAMCAS-LAM-BH-NEXT: move $a3, $a0
+; LA64-LAMCAS-LAM-BH-NEXT: and $a4, $a0, $a1
+; LA64-LAMCAS-LAM-BH-NEXT: nor $a4, $a4, $zero
+; LA64-LAMCAS-LAM-BH-NEXT: amcas_db.d $a0, $a4, $a2
+; LA64-LAMCAS-LAM-BH-NEXT: bne $a0, $a3, .LBB105_1
+; LA64-LAMCAS-LAM-BH-NEXT: # %bb.2: # %atomicrmw.end
+; LA64-LAMCAS-LAM-BH-NEXT: ret
%1 = atomicrmw nand ptr %a, i64 %b seq_cst
ret i64 %1
}
@@ -4108,17 +5366,29 @@ define i8 @atomicrmw_and_i8_seq_cst(ptr %a, i8 %b) nounwind {
;
; LA64-LAMCAS-LABEL: atomicrmw_and_i8_seq_cst:
; LA64-LAMCAS: # %bb.0:
-; LA64-LAMCAS-NEXT: move $a2, $a0
-; LA64-LAMCAS-NEXT: ld.bu $a0, $a0, 0
-; LA64-LAMCAS-NEXT: .p2align 4, , 16
-; LA64-LAMCAS-NEXT: .LBB106_1: # %atomicrmw.start
-; LA64-LAMCAS-NEXT: # =>This Inner Loop Header: Depth=1
-; LA64-LAMCAS-NEXT: and $a3, $a0, $a1
-; LA64-LAMCAS-NEXT: ext.w.b $a4, $a0
-; LA64-LAMCAS-NEXT: amcas_db.b $a0, $a3, $a2
-; LA64-LAMCAS-NEXT: bne $a0, $a4, .LBB106_1
-; LA64-LAMCAS-NEXT: # %bb.2: # %atomicrmw.end
-; LA64-LAMCAS-NEXT: ret
+; LA64-LAMCAS-NEXT: slli.d $a2, $a0, 3
+; LA64-LAMCAS-NEXT: bstrins.d $a0, $zero, 1, 0
+; LA64-LAMCAS-NEXT: ori $a3, $zero, 255
+; LA64-LAMCAS-NEXT: sll.w $a3, $a3, $a2
+; LA64-LAMCAS-NEXT: andi $a1, $a1, 255
+; LA64-LAMCAS-NEXT: sll.w $a1, $a1, $a2
+; LA64-LAMCAS-NEXT: orn $a1, $a1, $a3
+; LA64-LAMCAS-NEXT: amand_db.w $a3, $a1, $a0
+; LA64-LAMCAS-NEXT: srl.w $a0, $a3, $a2
+; LA64-LAMCAS-NEXT: ret
+;
+; LA64-LAMCAS-LAM-BH-LABEL: atomicrmw_and_i8_seq_cst:
+; LA64-LAMCAS-LAM-BH: # %bb.0:
+; LA64-LAMCAS-LAM-BH-NEXT: slli.d $a2, $a0, 3
+; LA64-LAMCAS-LAM-BH-NEXT: bstrins.d $a0, $zero, 1, 0
+; LA64-LAMCAS-LAM-BH-NEXT: ori $a3, $zero, 255
+; LA64-LAMCAS-LAM-BH-NEXT: sll.w $a3, $a3, $a2
+; LA64-LAMCAS-LAM-BH-NEXT: andi $a1, $a1, 255
+; LA64-LAMCAS-LAM-BH-NEXT: sll.w $a1, $a1, $a2
+; LA64-LAMCAS-LAM-BH-NEXT: orn $a1, $a1, $a3
+; LA64-LAMCAS-LAM-BH-NEXT: amand_db.w $a3, $a1, $a0
+; LA64-LAMCAS-LAM-BH-NEXT: srl.w $a0, $a3, $a2
+; LA64-LAMCAS-LAM-BH-NEXT: ret
%1 = atomicrmw and ptr %a, i8 %b seq_cst
ret i8 %1
}
@@ -4140,17 +5410,31 @@ define i16 @atomicrmw_and_i16_seq_cst(ptr %a, i16 %b) nounwind {
;
; LA64-LAMCAS-LABEL: atomicrmw_and_i16_seq_cst:
; LA64-LAMCAS: # %bb.0:
-; LA64-LAMCAS-NEXT: move $a2, $a0
-; LA64-LAMCAS-NEXT: ld.hu $a0, $a0, 0
-; LA64-LAMCAS-NEXT: .p2align 4, , 16
-; LA64-LAMCAS-NEXT: .LBB107_1: # %atomicrmw.start
-; LA64-LAMCAS-NEXT: # =>This Inner Loop Header: Depth=1
-; LA64-LAMCAS-NEXT: and $a3, $a0, $a1
-; LA64-LAMCAS-NEXT: ext.w.h $a4, $a0
-; LA64-LAMCAS-NEXT: amcas_db.h $a0, $a3, $a2
-; LA64-LAMCAS-NEXT: bne $a0, $a4, .LBB107_1
-; LA64-LAMCAS-NEXT: # %bb.2: # %atomicrmw.end
-; LA64-LAMCAS-NEXT: ret
+; LA64-LAMCAS-NEXT: slli.d $a2, $a0, 3
+; LA64-LAMCAS-NEXT: bstrins.d $a0, $zero, 1, 0
+; LA64-LAMCAS-NEXT: lu12i.w $a3, 15
+; LA64-LAMCAS-NEXT: ori $a3, $a3, 4095
+; LA64-LAMCAS-NEXT: sll.w $a3, $a3, $a2
+; LA64-LAMCAS-NEXT: bstrpick.d $a1, $a1, 15, 0
+; LA64-LAMCAS-NEXT: sll.w $a1, $a1, $a2
+; LA64-LAMCAS-NEXT: orn $a1, $a1, $a3
+; LA64-LAMCAS-NEXT: amand_db.w $a3, $a1, $a0
+; LA64-LAMCAS-NEXT: srl.w $a0, $a3, $a2
+; LA64-LAMCAS-NEXT: ret
+;
+; LA64-LAMCAS-LAM-BH-LABEL: atomicrmw_and_i16_seq_cst:
+; LA64-LAMCAS-LAM-BH: # %bb.0:
+; LA64-LAMCAS-LAM-BH-NEXT: slli.d $a2, $a0, 3
+; LA64-LAMCAS-LAM-BH-NEXT: bstrins.d $a0, $zero, 1, 0
+; LA64-LAMCAS-LAM-BH-NEXT: lu12i.w $a3, 15
+; LA64-LAMCAS-LAM-BH-NEXT: ori $a3, $a3, 4095
+; LA64-LAMCAS-LAM-BH-NEXT: sll.w $a3, $a3, $a2
+; LA64-LAMCAS-LAM-BH-NEXT: bstrpick.d $a1, $a1, 15, 0
+; LA64-LAMCAS-LAM-BH-NEXT: sll.w $a1, $a1, $a2
+; LA64-LAMCAS-LAM-BH-NEXT: orn $a1, $a1, $a3
+; LA64-LAMCAS-LAM-BH-NEXT: amand_db.w $a3, $a1, $a0
+; LA64-LAMCAS-LAM-BH-NEXT: srl.w $a0, $a3, $a2
+; LA64-LAMCAS-LAM-BH-NEXT: ret
%1 = atomicrmw and ptr %a, i16 %b seq_cst
ret i16 %1
@@ -4169,17 +5453,23 @@ define i8 @atomicrmw_or_i8_seq_cst(ptr %a, i8 %b) nounwind {
;
; LA64-LAMCAS-LABEL: atomicrmw_or_i8_seq_cst:
; LA64-LAMCAS: # %bb.0:
-; LA64-LAMCAS-NEXT: move $a2, $a0
-; LA64-LAMCAS-NEXT: ld.bu $a0, $a0, 0
-; LA64-LAMCAS-NEXT: .p2align 4, , 16
-; LA64-LAMCAS-NEXT: .LBB108_1: # %atomicrmw.start
-; LA64-LAMCAS-NEXT: # =>This Inner Loop Header: Depth=1
-; LA64-LAMCAS-NEXT: or $a3, $a0, $a1
-; LA64-LAMCAS-NEXT: ext.w.b $a4, $a0
-; LA64-LAMCAS-NEXT: amcas_db.b $a0, $a3, $a2
-; LA64-LAMCAS-NEXT: bne $a0, $a4, .LBB108_1
-; LA64-LAMCAS-NEXT: # %bb.2: # %atomicrmw.end
-; LA64-LAMCAS-NEXT: ret
+; LA64-LAMCAS-NEXT: slli.d $a2, $a0, 3
+; LA64-LAMCAS-NEXT: bstrins.d $a0, $zero, 1, 0
+; LA64-LAMCAS-NEXT: andi $a1, $a1, 255
+; LA64-LAMCAS-NEXT: sll.w $a1, $a1, $a2
+; LA64-LAMCAS-NEXT: amor_db.w $a3, $a1, $a0
+; LA64-LAMCAS-NEXT: srl.w $a0, $a3, $a2
+; LA64-LAMCAS-NEXT: ret
+;
+; LA64-LAMCAS-LAM-BH-LABEL: atomicrmw_or_i8_seq_cst:
+; LA64-LAMCAS-LAM-BH: # %bb.0:
+; LA64-LAMCAS-LAM-BH-NEXT: slli.d $a2, $a0, 3
+; LA64-LAMCAS-LAM-BH-NEXT: bstrins.d $a0, $zero, 1, 0
+; LA64-LAMCAS-LAM-BH-NEXT: andi $a1, $a1, 255
+; LA64-LAMCAS-LAM-BH-NEXT: sll.w $a1, $a1, $a2
+; LA64-LAMCAS-LAM-BH-NEXT: amor_db.w $a3, $a1, $a0
+; LA64-LAMCAS-LAM-BH-NEXT: srl.w $a0, $a3, $a2
+; LA64-LAMCAS-LAM-BH-NEXT: ret
%1 = atomicrmw or ptr %a, i8 %b seq_cst
ret i8 %1
}
@@ -4197,17 +5487,23 @@ define i16 @atomicrmw_or_i16_seq_cst(ptr %a, i16 %b) nounwind {
;
; LA64-LAMCAS-LABEL: atomicrmw_or_i16_seq_cst:
; LA64-LAMCAS: # %bb.0:
-; LA64-LAMCAS-NEXT: move $a2, $a0
-; LA64-LAMCAS-NEXT: ld.hu $a0, $a0, 0
-; LA64-LAMCAS-NEXT: .p2align 4, , 16
-; LA64-LAMCAS-NEXT: .LBB109_1: # %atomicrmw.start
-; LA64-LAMCAS-NEXT: # =>This Inner Loop Header: Depth=1
-; LA64-LAMCAS-NEXT: or $a3, $a0, $a1
-; LA64-LAMCAS-NEXT: ext.w.h $a4, $a0
-; LA64-LAMCAS-NEXT: amcas_db.h $a0, $a3, $a2
-; LA64-LAMCAS-NEXT: bne $a0, $a4, .LBB109_1
-; LA64-LAMCAS-NEXT: # %bb.2: # %atomicrmw.end
-; LA64-LAMCAS-NEXT: ret
+; LA64-LAMCAS-NEXT: slli.d $a2, $a0, 3
+; LA64-LAMCAS-NEXT: bstrins.d $a0, $zero, 1, 0
+; LA64-LAMCAS-NEXT: bstrpick.d $a1, $a1, 15, 0
+; LA64-LAMCAS-NEXT: sll.w $a1, $a1, $a2
+; LA64-LAMCAS-NEXT: amor_db.w $a3, $a1, $a0
+; LA64-LAMCAS-NEXT: srl.w $a0, $a3, $a2
+; LA64-LAMCAS-NEXT: ret
+;
+; LA64-LAMCAS-LAM-BH-LABEL: atomicrmw_or_i16_seq_cst:
+; LA64-LAMCAS-LAM-BH: # %bb.0:
+; LA64-LAMCAS-LAM-BH-NEXT: slli.d $a2, $a0, 3
+; LA64-LAMCAS-LAM-BH-NEXT: bstrins.d $a0, $zero, 1, 0
+; LA64-LAMCAS-LAM-BH-NEXT: bstrpick.d $a1, $a1, 15, 0
+; LA64-LAMCAS-LAM-BH-NEXT: sll.w $a1, $a1, $a2
+; LA64-LAMCAS-LAM-BH-NEXT: amor_db.w $a3, $a1, $a0
+; LA64-LAMCAS-LAM-BH-NEXT: srl.w $a0, $a3, $a2
+; LA64-LAMCAS-LAM-BH-NEXT: ret
%1 = atomicrmw or ptr %a, i16 %b seq_cst
ret i16 %1
@@ -4226,17 +5522,23 @@ define i8 @atomicrmw_xor_i8_seq_cst(ptr %a, i8 %b) nounwind {
;
; LA64-LAMCAS-LABEL: atomicrmw_xor_i8_seq_cst:
; LA64-LAMCAS: # %bb.0:
-; LA64-LAMCAS-NEXT: move $a2, $a0
-; LA64-LAMCAS-NEXT: ld.bu $a0, $a0, 0
-; LA64-LAMCAS-NEXT: .p2align 4, , 16
-; LA64-LAMCAS-NEXT: .LBB110_1: # %atomicrmw.start
-; LA64-LAMCAS-NEXT: # =>This Inner Loop Header: Depth=1
-; LA64-LAMCAS-NEXT: xor $a3, $a0, $a1
-; LA64-LAMCAS-NEXT: ext.w.b $a4, $a0
-; LA64-LAMCAS-NEXT: amcas_db.b $a0, $a3, $a2
-; LA64-LAMCAS-NEXT: bne $a0, $a4, .LBB110_1
-; LA64-LAMCAS-NEXT: # %bb.2: # %atomicrmw.end
-; LA64-LAMCAS-NEXT: ret
+; LA64-LAMCAS-NEXT: slli.d $a2, $a0, 3
+; LA64-LAMCAS-NEXT: bstrins.d $a0, $zero, 1, 0
+; LA64-LAMCAS-NEXT: andi $a1, $a1, 255
+; LA64-LAMCAS-NEXT: sll.w $a1, $a1, $a2
+; LA64-LAMCAS-NEXT: amxor_db.w $a3, $a1, $a0
+; LA64-LAMCAS-NEXT: srl.w $a0, $a3, $a2
+; LA64-LAMCAS-NEXT: ret
+;
+; LA64-LAMCAS-LAM-BH-LABEL: atomicrmw_xor_i8_seq_cst:
+; LA64-LAMCAS-LAM-BH: # %bb.0:
+; LA64-LAMCAS-LAM-BH-NEXT: slli.d $a2, $a0, 3
+; LA64-LAMCAS-LAM-BH-NEXT: bstrins.d $a0, $zero, 1, 0
+; LA64-LAMCAS-LAM-BH-NEXT: andi $a1, $a1, 255
+; LA64-LAMCAS-LAM-BH-NEXT: sll.w $a1, $a1, $a2
+; LA64-LAMCAS-LAM-BH-NEXT: amxor_db.w $a3, $a1, $a0
+; LA64-LAMCAS-LAM-BH-NEXT: srl.w $a0, $a3, $a2
+; LA64-LAMCAS-LAM-BH-NEXT: ret
%1 = atomicrmw xor ptr %a, i8 %b seq_cst
ret i8 %1
}
@@ -4254,17 +5556,23 @@ define i16 @atomicrmw_xor_i16_seq_cst(ptr %a, i16 %b) nounwind {
;
; LA64-LAMCAS-LABEL: atomicrmw_xor_i16_seq_cst:
; LA64-LAMCAS: # %bb.0:
-; LA64-LAMCAS-NEXT: move $a2, $a0
-; LA64-LAMCAS-NEXT: ld.hu $a0, $a0, 0
-; LA64-LAMCAS-NEXT: .p2align 4, , 16
-; LA64-LAMCAS-NEXT: .LBB111_1: # %atomicrmw.start
-; LA64-LAMCAS-NEXT: # =>This Inner Loop Header: Depth=1
-; LA64-LAMCAS-NEXT: xor $a3, $a0, $a1
-; LA64-LAMCAS-NEXT: ext.w.h $a4, $a0
-; LA64-LAMCAS-NEXT: amcas_db.h $a0, $a3, $a2
-; LA64-LAMCAS-NEXT: bne $a0, $a4, .LBB111_1
-; LA64-LAMCAS-NEXT: # %bb.2: # %atomicrmw.end
-; LA64-LAMCAS-NEXT: ret
+; LA64-LAMCAS-NEXT: slli.d $a2, $a0, 3
+; LA64-LAMCAS-NEXT: bstrins.d $a0, $zero, 1, 0
+; LA64-LAMCAS-NEXT: bstrpick.d $a1, $a1, 15, 0
+; LA64-LAMCAS-NEXT: sll.w $a1, $a1, $a2
+; LA64-LAMCAS-NEXT: amxor_db.w $a3, $a1, $a0
+; LA64-LAMCAS-NEXT: srl.w $a0, $a3, $a2
+; LA64-LAMCAS-NEXT: ret
+;
+; LA64-LAMCAS-LAM-BH-LABEL: atomicrmw_xor_i16_seq_cst:
+; LA64-LAMCAS-LAM-BH: # %bb.0:
+; LA64-LAMCAS-LAM-BH-NEXT: slli.d $a2, $a0, 3
+; LA64-LAMCAS-LAM-BH-NEXT: bstrins.d $a0, $zero, 1, 0
+; LA64-LAMCAS-LAM-BH-NEXT: bstrpick.d $a1, $a1, 15, 0
+; LA64-LAMCAS-LAM-BH-NEXT: sll.w $a1, $a1, $a2
+; LA64-LAMCAS-LAM-BH-NEXT: amxor_db.w $a3, $a1, $a0
+; LA64-LAMCAS-LAM-BH-NEXT: srl.w $a0, $a3, $a2
+; LA64-LAMCAS-LAM-BH-NEXT: ret
%1 = atomicrmw xor ptr %a, i16 %b seq_cst
ret i16 %1
@@ -4303,6 +5611,12 @@ define i8 @atomicrmw_xchg_i8_monotonic(ptr %a, i8 %b) nounwind {
; LA64-LAMCAS-NEXT: # %bb.2: # %atomicrmw.end
; LA64-LAMCAS-NEXT: move $a0, $a2
; LA64-LAMCAS-NEXT: ret
+;
+; LA64-LAMCAS-LAM-BH-LABEL: atomicrmw_xchg_i8_monotonic:
+; LA64-LAMCAS-LAM-BH: # %bb.0:
+; LA64-LAMCAS-LAM-BH-NEXT: amswap.b $a2, $a1, $a0
+; LA64-LAMCAS-LAM-BH-NEXT: move $a0, $a2
+; LA64-LAMCAS-LAM-BH-NEXT: ret
%1 = atomicrmw xchg ptr %a, i8 %b monotonic
ret i8 %1
}
@@ -4331,6 +5645,12 @@ define i8 @atomicrmw_xchg_0_i8_monotonic(ptr %a) nounwind {
; LA64-LAMCAS-NEXT: bne $a0, $a2, .LBB113_1
; LA64-LAMCAS-NEXT: # %bb.2: # %atomicrmw.end
; LA64-LAMCAS-NEXT: ret
+;
+; LA64-LAMCAS-LAM-BH-LABEL: atomicrmw_xchg_0_i8_monotonic:
+; LA64-LAMCAS-LAM-BH: # %bb.0:
+; LA64-LAMCAS-LAM-BH-NEXT: amswap.b $a1, $zero, $a0
+; LA64-LAMCAS-LAM-BH-NEXT: move $a0, $a1
+; LA64-LAMCAS-LAM-BH-NEXT: ret
%1 = atomicrmw xchg ptr %a, i8 0 monotonic
ret i8 %1
}
@@ -4359,6 +5679,13 @@ define i8 @atomicrmw_xchg_minus_1_i8_monotonic(ptr %a) nounwind {
; LA64-LAMCAS-NEXT: bne $a0, $a3, .LBB114_1
; LA64-LAMCAS-NEXT: # %bb.2: # %atomicrmw.end
; LA64-LAMCAS-NEXT: ret
+;
+; LA64-LAMCAS-LAM-BH-LABEL: atomicrmw_xchg_minus_1_i8_monotonic:
+; LA64-LAMCAS-LAM-BH: # %bb.0:
+; LA64-LAMCAS-LAM-BH-NEXT: addi.w $a2, $zero, -1
+; LA64-LAMCAS-LAM-BH-NEXT: amswap.b $a1, $a2, $a0
+; LA64-LAMCAS-LAM-BH-NEXT: move $a0, $a1
+; LA64-LAMCAS-LAM-BH-NEXT: ret
%1 = atomicrmw xchg ptr %a, i8 -1 monotonic
ret i8 %1
}
@@ -4397,6 +5724,12 @@ define i16 @atomicrmw_xchg_i16_monotonic(ptr %a, i16 %b) nounwind {
; LA64-LAMCAS-NEXT: # %bb.2: # %atomicrmw.end
; LA64-LAMCAS-NEXT: move $a0, $a2
; LA64-LAMCAS-NEXT: ret
+;
+; LA64-LAMCAS-LAM-BH-LABEL: atomicrmw_xchg_i16_monotonic:
+; LA64-LAMCAS-LAM-BH: # %bb.0:
+; LA64-LAMCAS-LAM-BH-NEXT: amswap.h $a2, $a1, $a0
+; LA64-LAMCAS-LAM-BH-NEXT: move $a0, $a2
+; LA64-LAMCAS-LAM-BH-NEXT: ret
%1 = atomicrmw xchg ptr %a, i16 %b monotonic
ret i16 %1
}
@@ -4426,6 +5759,12 @@ define i16 @atomicrmw_xchg_0_i16_monotonic(ptr %a) nounwind {
; LA64-LAMCAS-NEXT: bne $a0, $a2, .LBB116_1
; LA64-LAMCAS-NEXT: # %bb.2: # %atomicrmw.end
; LA64-LAMCAS-NEXT: ret
+;
+; LA64-LAMCAS-LAM-BH-LABEL: atomicrmw_xchg_0_i16_monotonic:
+; LA64-LAMCAS-LAM-BH: # %bb.0:
+; LA64-LAMCAS-LAM-BH-NEXT: amswap.h $a1, $zero, $a0
+; LA64-LAMCAS-LAM-BH-NEXT: move $a0, $a1
+; LA64-LAMCAS-LAM-BH-NEXT: ret
%1 = atomicrmw xchg ptr %a, i16 0 monotonic
ret i16 %1
}
@@ -4455,6 +5794,13 @@ define i16 @atomicrmw_xchg_minus_1_i16_monotonic(ptr %a) nounwind {
; LA64-LAMCAS-NEXT: bne $a0, $a3, .LBB117_1
; LA64-LAMCAS-NEXT: # %bb.2: # %atomicrmw.end
; LA64-LAMCAS-NEXT: ret
+;
+; LA64-LAMCAS-LAM-BH-LABEL: atomicrmw_xchg_minus_1_i16_monotonic:
+; LA64-LAMCAS-LAM-BH: # %bb.0:
+; LA64-LAMCAS-LAM-BH-NEXT: addi.w $a2, $zero, -1
+; LA64-LAMCAS-LAM-BH-NEXT: amswap.h $a1, $a2, $a0
+; LA64-LAMCAS-LAM-BH-NEXT: move $a0, $a1
+; LA64-LAMCAS-LAM-BH-NEXT: ret
%1 = atomicrmw xchg ptr %a, i16 -1 monotonic
ret i16 %1
@@ -4494,6 +5840,12 @@ define i8 @atomicrmw_add_i8_monotonic(ptr %a, i8 %b) nounwind {
; LA64-LAMCAS-NEXT: bne $a0, $a4, .LBB118_1
; LA64-LAMCAS-NEXT: # %bb.2: # %atomicrmw.end
; LA64-LAMCAS-NEXT: ret
+;
+; LA64-LAMCAS-LAM-BH-LABEL: atomicrmw_add_i8_monotonic:
+; LA64-LAMCAS-LAM-BH: # %bb.0:
+; LA64-LAMCAS-LAM-BH-NEXT: amadd.b $a2, $a1, $a0
+; LA64-LAMCAS-LAM-BH-NEXT: move $a0, $a2
+; LA64-LAMCAS-LAM-BH-NEXT: ret
%1 = atomicrmw add ptr %a, i8 %b monotonic
ret i8 %1
}
@@ -4533,6 +5885,12 @@ define i16 @atomicrmw_add_i16_monotonic(ptr %a, i16 %b) nounwind {
; LA64-LAMCAS-NEXT: bne $a0, $a4, .LBB119_1
; LA64-LAMCAS-NEXT: # %bb.2: # %atomicrmw.end
; LA64-LAMCAS-NEXT: ret
+;
+; LA64-LAMCAS-LAM-BH-LABEL: atomicrmw_add_i16_monotonic:
+; LA64-LAMCAS-LAM-BH: # %bb.0:
+; LA64-LAMCAS-LAM-BH-NEXT: amadd.h $a2, $a1, $a0
+; LA64-LAMCAS-LAM-BH-NEXT: move $a0, $a2
+; LA64-LAMCAS-LAM-BH-NEXT: ret
%1 = atomicrmw add ptr %a, i16 %b monotonic
ret i16 %1
@@ -4572,6 +5930,13 @@ define i8 @atomicrmw_sub_i8_monotonic(ptr %a, i8 %b) nounwind {
; LA64-LAMCAS-NEXT: bne $a0, $a4, .LBB120_1
; LA64-LAMCAS-NEXT: # %bb.2: # %atomicrmw.end
; LA64-LAMCAS-NEXT: ret
+;
+; LA64-LAMCAS-LAM-BH-LABEL: atomicrmw_sub_i8_monotonic:
+; LA64-LAMCAS-LAM-BH: # %bb.0:
+; LA64-LAMCAS-LAM-BH-NEXT: sub.w $a2, $zero, $a1
+; LA64-LAMCAS-LAM-BH-NEXT: amadd.b $a1, $a2, $a0
+; LA64-LAMCAS-LAM-BH-NEXT: move $a0, $a1
+; LA64-LAMCAS-LAM-BH-NEXT: ret
%1 = atomicrmw sub ptr %a, i8 %b monotonic
ret i8 %1
}
@@ -4611,6 +5976,13 @@ define i16 @atomicrmw_sub_i16_monotonic(ptr %a, i16 %b) nounwind {
; LA64-LAMCAS-NEXT: bne $a0, $a4, .LBB121_1
; LA64-LAMCAS-NEXT: # %bb.2: # %atomicrmw.end
; LA64-LAMCAS-NEXT: ret
+;
+; LA64-LAMCAS-LAM-BH-LABEL: atomicrmw_sub_i16_monotonic:
+; LA64-LAMCAS-LAM-BH: # %bb.0:
+; LA64-LAMCAS-LAM-BH-NEXT: sub.w $a2, $zero, $a1
+; LA64-LAMCAS-LAM-BH-NEXT: amadd.h $a1, $a2, $a0
+; LA64-LAMCAS-LAM-BH-NEXT: move $a0, $a1
+; LA64-LAMCAS-LAM-BH-NEXT: ret
%1 = atomicrmw sub ptr %a, i16 %b monotonic
ret i16 %1
@@ -4659,6 +6031,25 @@ define i8 @atomicrmw_umax_i8_monotonic(ptr %a, i8 %b) nounwind {
; LA64-LAMCAS-NEXT: bne $a0, $a5, .LBB122_1
; LA64-LAMCAS-NEXT: # %bb.2: # %atomicrmw.end
; LA64-LAMCAS-NEXT: ret
+;
+; LA64-LAMCAS-LAM-BH-LABEL: atomicrmw_umax_i8_monotonic:
+; LA64-LAMCAS-LAM-BH: # %bb.0:
+; LA64-LAMCAS-LAM-BH-NEXT: move $a2, $a0
+; LA64-LAMCAS-LAM-BH-NEXT: ld.bu $a0, $a0, 0
+; LA64-LAMCAS-LAM-BH-NEXT: andi $a3, $a1, 255
+; LA64-LAMCAS-LAM-BH-NEXT: .p2align 4, , 16
+; LA64-LAMCAS-LAM-BH-NEXT: .LBB122_1: # %atomicrmw.start
+; LA64-LAMCAS-LAM-BH-NEXT: # =>This Inner Loop Header: Depth=1
+; LA64-LAMCAS-LAM-BH-NEXT: andi $a4, $a0, 255
+; LA64-LAMCAS-LAM-BH-NEXT: sltu $a4, $a3, $a4
+; LA64-LAMCAS-LAM-BH-NEXT: masknez $a5, $a1, $a4
+; LA64-LAMCAS-LAM-BH-NEXT: maskeqz $a4, $a0, $a4
+; LA64-LAMCAS-LAM-BH-NEXT: or $a4, $a4, $a5
+; LA64-LAMCAS-LAM-BH-NEXT: ext.w.b $a5, $a0
+; LA64-LAMCAS-LAM-BH-NEXT: amcas.b $a0, $a4, $a2
+; LA64-LAMCAS-LAM-BH-NEXT: bne $a0, $a5, .LBB122_1
+; LA64-LAMCAS-LAM-BH-NEXT: # %bb.2: # %atomicrmw.end
+; LA64-LAMCAS-LAM-BH-NEXT: ret
%1 = atomicrmw umax ptr %a, i8 %b monotonic
ret i8 %1
}
@@ -4707,6 +6098,25 @@ define i16 @atomicrmw_umax_i16_monotonic(ptr %a, i16 %b) nounwind {
; LA64-LAMCAS-NEXT: bne $a0, $a5, .LBB123_1
; LA64-LAMCAS-NEXT: # %bb.2: # %atomicrmw.end
; LA64-LAMCAS-NEXT: ret
+;
+; LA64-LAMCAS-LAM-BH-LABEL: atomicrmw_umax_i16_monotonic:
+; LA64-LAMCAS-LAM-BH: # %bb.0:
+; LA64-LAMCAS-LAM-BH-NEXT: move $a2, $a0
+; LA64-LAMCAS-LAM-BH-NEXT: ld.hu $a0, $a0, 0
+; LA64-LAMCAS-LAM-BH-NEXT: bstrpick.d $a3, $a1, 15, 0
+; LA64-LAMCAS-LAM-BH-NEXT: .p2align 4, , 16
+; LA64-LAMCAS-LAM-BH-NEXT: .LBB123_1: # %atomicrmw.start
+; LA64-LAMCAS-LAM-BH-NEXT: # =>This Inner Loop Header: Depth=1
+; LA64-LAMCAS-LAM-BH-NEXT: bstrpick.d $a4, $a0, 15, 0
+; LA64-LAMCAS-LAM-BH-NEXT: sltu $a4, $a3, $a4
+; LA64-LAMCAS-LAM-BH-NEXT: masknez $a5, $a1, $a4
+; LA64-LAMCAS-LAM-BH-NEXT: maskeqz $a4, $a0, $a4
+; LA64-LAMCAS-LAM-BH-NEXT: or $a4, $a4, $a5
+; LA64-LAMCAS-LAM-BH-NEXT: ext.w.h $a5, $a0
+; LA64-LAMCAS-LAM-BH-NEXT: amcas.h $a0, $a4, $a2
+; LA64-LAMCAS-LAM-BH-NEXT: bne $a0, $a5, .LBB123_1
+; LA64-LAMCAS-LAM-BH-NEXT: # %bb.2: # %atomicrmw.end
+; LA64-LAMCAS-LAM-BH-NEXT: ret
%1 = atomicrmw umax ptr %a, i16 %b monotonic
ret i16 %1
}
@@ -4755,6 +6165,26 @@ define i8 @atomicrmw_umin_i8_monotonic(ptr %a, i8 %b) nounwind {
; LA64-LAMCAS-NEXT: bne $a0, $a5, .LBB124_1
; LA64-LAMCAS-NEXT: # %bb.2: # %atomicrmw.end
; LA64-LAMCAS-NEXT: ret
+;
+; LA64-LAMCAS-LAM-BH-LABEL: atomicrmw_umin_i8_monotonic:
+; LA64-LAMCAS-LAM-BH: # %bb.0:
+; LA64-LAMCAS-LAM-BH-NEXT: move $a2, $a0
+; LA64-LAMCAS-LAM-BH-NEXT: ld.bu $a0, $a0, 0
+; LA64-LAMCAS-LAM-BH-NEXT: andi $a3, $a1, 255
+; LA64-LAMCAS-LAM-BH-NEXT: .p2align 4, , 16
+; LA64-LAMCAS-LAM-BH-NEXT: .LBB124_1: # %atomicrmw.start
+; LA64-LAMCAS-LAM-BH-NEXT: # =>This Inner Loop Header: Depth=1
+; LA64-LAMCAS-LAM-BH-NEXT: andi $a4, $a0, 255
+; LA64-LAMCAS-LAM-BH-NEXT: sltu $a4, $a3, $a4
+; LA64-LAMCAS-LAM-BH-NEXT: xori $a4, $a4, 1
+; LA64-LAMCAS-LAM-BH-NEXT: masknez $a5, $a1, $a4
+; LA64-LAMCAS-LAM-BH-NEXT: maskeqz $a4, $a0, $a4
+; LA64-LAMCAS-LAM-BH-NEXT: or $a4, $a4, $a5
+; LA64-LAMCAS-LAM-BH-NEXT: ext.w.b $a5, $a0
+; LA64-LAMCAS-LAM-BH-NEXT: amcas.b $a0, $a4, $a2
+; LA64-LAMCAS-LAM-BH-NEXT: bne $a0, $a5, .LBB124_1
+; LA64-LAMCAS-LAM-BH-NEXT: # %bb.2: # %atomicrmw.end
+; LA64-LAMCAS-LAM-BH-NEXT: ret
%1 = atomicrmw umin ptr %a, i8 %b monotonic
ret i8 %1
}
@@ -4804,6 +6234,26 @@ define i16 @atomicrmw_umin_i16_monotonic(ptr %a, i16 %b) nounwind {
; LA64-LAMCAS-NEXT: bne $a0, $a5, .LBB125_1
; LA64-LAMCAS-NEXT: # %bb.2: # %atomicrmw.end
; LA64-LAMCAS-NEXT: ret
+;
+; LA64-LAMCAS-LAM-BH-LABEL: atomicrmw_umin_i16_monotonic:
+; LA64-LAMCAS-LAM-BH: # %bb.0:
+; LA64-LAMCAS-LAM-BH-NEXT: move $a2, $a0
+; LA64-LAMCAS-LAM-BH-NEXT: ld.hu $a0, $a0, 0
+; LA64-LAMCAS-LAM-BH-NEXT: bstrpick.d $a3, $a1, 15, 0
+; LA64-LAMCAS-LAM-BH-NEXT: .p2align 4, , 16
+; LA64-LAMCAS-LAM-BH-NEXT: .LBB125_1: # %atomicrmw.start
+; LA64-LAMCAS-LAM-BH-NEXT: # =>This Inner Loop Header: Depth=1
+; LA64-LAMCAS-LAM-BH-NEXT: bstrpick.d $a4, $a0, 15, 0
+; LA64-LAMCAS-LAM-BH-NEXT: sltu $a4, $a3, $a4
+; LA64-LAMCAS-LAM-BH-NEXT: xori $a4, $a4, 1
+; LA64-LAMCAS-LAM-BH-NEXT: masknez $a5, $a1, $a4
+; LA64-LAMCAS-LAM-BH-NEXT: maskeqz $a4, $a0, $a4
+; LA64-LAMCAS-LAM-BH-NEXT: or $a4, $a4, $a5
+; LA64-LAMCAS-LAM-BH-NEXT: ext.w.h $a5, $a0
+; LA64-LAMCAS-LAM-BH-NEXT: amcas.h $a0, $a4, $a2
+; LA64-LAMCAS-LAM-BH-NEXT: bne $a0, $a5, .LBB125_1
+; LA64-LAMCAS-LAM-BH-NEXT: # %bb.2: # %atomicrmw.end
+; LA64-LAMCAS-LAM-BH-NEXT: ret
%1 = atomicrmw umin ptr %a, i16 %b monotonic
ret i16 %1
}
@@ -4854,6 +6304,24 @@ define i8 @atomicrmw_max_i8_monotonic(ptr %a, i8 %b) nounwind {
; LA64-LAMCAS-NEXT: bne $a0, $a4, .LBB126_1
; LA64-LAMCAS-NEXT: # %bb.2: # %atomicrmw.end
; LA64-LAMCAS-NEXT: ret
+;
+; LA64-LAMCAS-LAM-BH-LABEL: atomicrmw_max_i8_monotonic:
+; LA64-LAMCAS-LAM-BH: # %bb.0:
+; LA64-LAMCAS-LAM-BH-NEXT: move $a2, $a0
+; LA64-LAMCAS-LAM-BH-NEXT: ld.bu $a0, $a0, 0
+; LA64-LAMCAS-LAM-BH-NEXT: ext.w.b $a3, $a1
+; LA64-LAMCAS-LAM-BH-NEXT: .p2align 4, , 16
+; LA64-LAMCAS-LAM-BH-NEXT: .LBB126_1: # %atomicrmw.start
+; LA64-LAMCAS-LAM-BH-NEXT: # =>This Inner Loop Header: Depth=1
+; LA64-LAMCAS-LAM-BH-NEXT: ext.w.b $a4, $a0
+; LA64-LAMCAS-LAM-BH-NEXT: slt $a5, $a3, $a4
+; LA64-LAMCAS-LAM-BH-NEXT: masknez $a6, $a1, $a5
+; LA64-LAMCAS-LAM-BH-NEXT: maskeqz $a5, $a0, $a5
+; LA64-LAMCAS-LAM-BH-NEXT: or $a5, $a5, $a6
+; LA64-LAMCAS-LAM-BH-NEXT: amcas.b $a0, $a5, $a2
+; LA64-LAMCAS-LAM-BH-NEXT: bne $a0, $a4, .LBB126_1
+; LA64-LAMCAS-LAM-BH-NEXT: # %bb.2: # %atomicrmw.end
+; LA64-LAMCAS-LAM-BH-NEXT: ret
%1 = atomicrmw max ptr %a, i8 %b monotonic
ret i8 %1
}
@@ -4906,6 +6374,24 @@ define i16 @atomicrmw_max_i16_monotonic(ptr %a, i16 %b) nounwind {
; LA64-LAMCAS-NEXT: bne $a0, $a4, .LBB127_1
; LA64-LAMCAS-NEXT: # %bb.2: # %atomicrmw.end
; LA64-LAMCAS-NEXT: ret
+;
+; LA64-LAMCAS-LAM-BH-LABEL: atomicrmw_max_i16_monotonic:
+; LA64-LAMCAS-LAM-BH: # %bb.0:
+; LA64-LAMCAS-LAM-BH-NEXT: move $a2, $a0
+; LA64-LAMCAS-LAM-BH-NEXT: ld.hu $a0, $a0, 0
+; LA64-LAMCAS-LAM-BH-NEXT: ext.w.h $a3, $a1
+; LA64-LAMCAS-LAM-BH-NEXT: .p2align 4, , 16
+; LA64-LAMCAS-LAM-BH-NEXT: .LBB127_1: # %atomicrmw.start
+; LA64-LAMCAS-LAM-BH-NEXT: # =>This Inner Loop Header: Depth=1
+; LA64-LAMCAS-LAM-BH-NEXT: ext.w.h $a4, $a0
+; LA64-LAMCAS-LAM-BH-NEXT: slt $a5, $a3, $a4
+; LA64-LAMCAS-LAM-BH-NEXT: masknez $a6, $a1, $a5
+; LA64-LAMCAS-LAM-BH-NEXT: maskeqz $a5, $a0, $a5
+; LA64-LAMCAS-LAM-BH-NEXT: or $a5, $a5, $a6
+; LA64-LAMCAS-LAM-BH-NEXT: amcas.h $a0, $a5, $a2
+; LA64-LAMCAS-LAM-BH-NEXT: bne $a0, $a4, .LBB127_1
+; LA64-LAMCAS-LAM-BH-NEXT: # %bb.2: # %atomicrmw.end
+; LA64-LAMCAS-LAM-BH-NEXT: ret
%1 = atomicrmw max ptr %a, i16 %b monotonic
ret i16 %1
}
@@ -4957,6 +6443,25 @@ define i8 @atomicrmw_min_i8_monotonic(ptr %a, i8 %b) nounwind {
; LA64-LAMCAS-NEXT: bne $a0, $a4, .LBB128_1
; LA64-LAMCAS-NEXT: # %bb.2: # %atomicrmw.end
; LA64-LAMCAS-NEXT: ret
+;
+; LA64-LAMCAS-LAM-BH-LABEL: atomicrmw_min_i8_monotonic:
+; LA64-LAMCAS-LAM-BH: # %bb.0:
+; LA64-LAMCAS-LAM-BH-NEXT: move $a2, $a0
+; LA64-LAMCAS-LAM-BH-NEXT: ld.bu $a0, $a0, 0
+; LA64-LAMCAS-LAM-BH-NEXT: ext.w.b $a3, $a1
+; LA64-LAMCAS-LAM-BH-NEXT: .p2align 4, , 16
+; LA64-LAMCAS-LAM-BH-NEXT: .LBB128_1: # %atomicrmw.start
+; LA64-LAMCAS-LAM-BH-NEXT: # =>This Inner Loop Header: Depth=1
+; LA64-LAMCAS-LAM-BH-NEXT: ext.w.b $a4, $a0
+; LA64-LAMCAS-LAM-BH-NEXT: slt $a5, $a3, $a4
+; LA64-LAMCAS-LAM-BH-NEXT: xori $a5, $a5, 1
+; LA64-LAMCAS-LAM-BH-NEXT: masknez $a6, $a1, $a5
+; LA64-LAMCAS-LAM-BH-NEXT: maskeqz $a5, $a0, $a5
+; LA64-LAMCAS-LAM-BH-NEXT: or $a5, $a5, $a6
+; LA64-LAMCAS-LAM-BH-NEXT: amcas.b $a0, $a5, $a2
+; LA64-LAMCAS-LAM-BH-NEXT: bne $a0, $a4, .LBB128_1
+; LA64-LAMCAS-LAM-BH-NEXT: # %bb.2: # %atomicrmw.end
+; LA64-LAMCAS-LAM-BH-NEXT: ret
%1 = atomicrmw min ptr %a, i8 %b monotonic
ret i8 %1
}
@@ -5010,6 +6515,25 @@ define i16 @atomicrmw_min_i16_monotonic(ptr %a, i16 %b) nounwind {
; LA64-LAMCAS-NEXT: bne $a0, $a4, .LBB129_1
; LA64-LAMCAS-NEXT: # %bb.2: # %atomicrmw.end
; LA64-LAMCAS-NEXT: ret
+;
+; LA64-LAMCAS-LAM-BH-LABEL: atomicrmw_min_i16_monotonic:
+; LA64-LAMCAS-LAM-BH: # %bb.0:
+; LA64-LAMCAS-LAM-BH-NEXT: move $a2, $a0
+; LA64-LAMCAS-LAM-BH-NEXT: ld.hu $a0, $a0, 0
+; LA64-LAMCAS-LAM-BH-NEXT: ext.w.h $a3, $a1
+; LA64-LAMCAS-LAM-BH-NEXT: .p2align 4, , 16
+; LA64-LAMCAS-LAM-BH-NEXT: .LBB129_1: # %atomicrmw.start
+; LA64-LAMCAS-LAM-BH-NEXT: # =>This Inner Loop Header: Depth=1
+; LA64-LAMCAS-LAM-BH-NEXT: ext.w.h $a4, $a0
+; LA64-LAMCAS-LAM-BH-NEXT: slt $a5, $a3, $a4
+; LA64-LAMCAS-LAM-BH-NEXT: xori $a5, $a5, 1
+; LA64-LAMCAS-LAM-BH-NEXT: masknez $a6, $a1, $a5
+; LA64-LAMCAS-LAM-BH-NEXT: maskeqz $a5, $a0, $a5
+; LA64-LAMCAS-LAM-BH-NEXT: or $a5, $a5, $a6
+; LA64-LAMCAS-LAM-BH-NEXT: amcas.h $a0, $a5, $a2
+; LA64-LAMCAS-LAM-BH-NEXT: bne $a0, $a4, .LBB129_1
+; LA64-LAMCAS-LAM-BH-NEXT: # %bb.2: # %atomicrmw.end
+; LA64-LAMCAS-LAM-BH-NEXT: ret
%1 = atomicrmw min ptr %a, i16 %b monotonic
ret i16 %1
}
@@ -5052,6 +6576,21 @@ define i8 @atomicrmw_nand_i8_monotonic(ptr %a, i8 %b) nounwind {
; LA64-LAMCAS-NEXT: bne $a0, $a4, .LBB130_1
; LA64-LAMCAS-NEXT: # %bb.2: # %atomicrmw.end
; LA64-LAMCAS-NEXT: ret
+;
+; LA64-LAMCAS-LAM-BH-LABEL: atomicrmw_nand_i8_monotonic:
+; LA64-LAMCAS-LAM-BH: # %bb.0:
+; LA64-LAMCAS-LAM-BH-NEXT: move $a2, $a0
+; LA64-LAMCAS-LAM-BH-NEXT: ld.bu $a0, $a0, 0
+; LA64-LAMCAS-LAM-BH-NEXT: .p2align 4, , 16
+; LA64-LAMCAS-LAM-BH-NEXT: .LBB130_1: # %atomicrmw.start
+; LA64-LAMCAS-LAM-BH-NEXT: # =>This Inner Loop Header: Depth=1
+; LA64-LAMCAS-LAM-BH-NEXT: and $a3, $a0, $a1
+; LA64-LAMCAS-LAM-BH-NEXT: nor $a3, $a3, $zero
+; LA64-LAMCAS-LAM-BH-NEXT: ext.w.b $a4, $a0
+; LA64-LAMCAS-LAM-BH-NEXT: amcas.b $a0, $a3, $a2
+; LA64-LAMCAS-LAM-BH-NEXT: bne $a0, $a4, .LBB130_1
+; LA64-LAMCAS-LAM-BH-NEXT: # %bb.2: # %atomicrmw.end
+; LA64-LAMCAS-LAM-BH-NEXT: ret
%1 = atomicrmw nand ptr %a, i8 %b monotonic
ret i8 %1
}
@@ -5093,6 +6632,21 @@ define i16 @atomicrmw_nand_i16_monotonic(ptr %a, i16 %b) nounwind {
; LA64-LAMCAS-NEXT: bne $a0, $a4, .LBB131_1
; LA64-LAMCAS-NEXT: # %bb.2: # %atomicrmw.end
; LA64-LAMCAS-NEXT: ret
+;
+; LA64-LAMCAS-LAM-BH-LABEL: atomicrmw_nand_i16_monotonic:
+; LA64-LAMCAS-LAM-BH: # %bb.0:
+; LA64-LAMCAS-LAM-BH-NEXT: move $a2, $a0
+; LA64-LAMCAS-LAM-BH-NEXT: ld.hu $a0, $a0, 0
+; LA64-LAMCAS-LAM-BH-NEXT: .p2align 4, , 16
+; LA64-LAMCAS-LAM-BH-NEXT: .LBB131_1: # %atomicrmw.start
+; LA64-LAMCAS-LAM-BH-NEXT: # =>This Inner Loop Header: Depth=1
+; LA64-LAMCAS-LAM-BH-NEXT: and $a3, $a0, $a1
+; LA64-LAMCAS-LAM-BH-NEXT: nor $a3, $a3, $zero
+; LA64-LAMCAS-LAM-BH-NEXT: ext.w.h $a4, $a0
+; LA64-LAMCAS-LAM-BH-NEXT: amcas.h $a0, $a3, $a2
+; LA64-LAMCAS-LAM-BH-NEXT: bne $a0, $a4, .LBB131_1
+; LA64-LAMCAS-LAM-BH-NEXT: # %bb.2: # %atomicrmw.end
+; LA64-LAMCAS-LAM-BH-NEXT: ret
%1 = atomicrmw nand ptr %a, i16 %b monotonic
ret i16 %1
@@ -5125,6 +6679,21 @@ define i32 @atomicrmw_nand_i32_monotonic(ptr %a, i32 %b) nounwind {
; LA64-LAMCAS-NEXT: bne $a0, $a3, .LBB132_1
; LA64-LAMCAS-NEXT: # %bb.2: # %atomicrmw.end
; LA64-LAMCAS-NEXT: ret
+;
+; LA64-LAMCAS-LAM-BH-LABEL: atomicrmw_nand_i32_monotonic:
+; LA64-LAMCAS-LAM-BH: # %bb.0:
+; LA64-LAMCAS-LAM-BH-NEXT: move $a2, $a0
+; LA64-LAMCAS-LAM-BH-NEXT: ld.w $a0, $a0, 0
+; LA64-LAMCAS-LAM-BH-NEXT: .p2align 4, , 16
+; LA64-LAMCAS-LAM-BH-NEXT: .LBB132_1: # %atomicrmw.start
+; LA64-LAMCAS-LAM-BH-NEXT: # =>This Inner Loop Header: Depth=1
+; LA64-LAMCAS-LAM-BH-NEXT: move $a3, $a0
+; LA64-LAMCAS-LAM-BH-NEXT: and $a4, $a0, $a1
+; LA64-LAMCAS-LAM-BH-NEXT: nor $a4, $a4, $zero
+; LA64-LAMCAS-LAM-BH-NEXT: amcas.w $a0, $a4, $a2
+; LA64-LAMCAS-LAM-BH-NEXT: bne $a0, $a3, .LBB132_1
+; LA64-LAMCAS-LAM-BH-NEXT: # %bb.2: # %atomicrmw.end
+; LA64-LAMCAS-LAM-BH-NEXT: ret
%1 = atomicrmw nand ptr %a, i32 %b monotonic
ret i32 %1
}
@@ -5156,6 +6725,21 @@ define i64 @atomicrmw_nand_i64_monotonic(ptr %a, i64 %b) nounwind {
; LA64-LAMCAS-NEXT: bne $a0, $a3, .LBB133_1
; LA64-LAMCAS-NEXT: # %bb.2: # %atomicrmw.end
; LA64-LAMCAS-NEXT: ret
+;
+; LA64-LAMCAS-LAM-BH-LABEL: atomicrmw_nand_i64_monotonic:
+; LA64-LAMCAS-LAM-BH: # %bb.0:
+; LA64-LAMCAS-LAM-BH-NEXT: move $a2, $a0
+; LA64-LAMCAS-LAM-BH-NEXT: ld.d $a0, $a0, 0
+; LA64-LAMCAS-LAM-BH-NEXT: .p2align 4, , 16
+; LA64-LAMCAS-LAM-BH-NEXT: .LBB133_1: # %atomicrmw.start
+; LA64-LAMCAS-LAM-BH-NEXT: # =>This Inner Loop Header: Depth=1
+; LA64-LAMCAS-LAM-BH-NEXT: move $a3, $a0
+; LA64-LAMCAS-LAM-BH-NEXT: and $a4, $a0, $a1
+; LA64-LAMCAS-LAM-BH-NEXT: nor $a4, $a4, $zero
+; LA64-LAMCAS-LAM-BH-NEXT: amcas.d $a0, $a4, $a2
+; LA64-LAMCAS-LAM-BH-NEXT: bne $a0, $a3, .LBB133_1
+; LA64-LAMCAS-LAM-BH-NEXT: # %bb.2: # %atomicrmw.end
+; LA64-LAMCAS-LAM-BH-NEXT: ret
%1 = atomicrmw nand ptr %a, i64 %b monotonic
ret i64 %1
}
@@ -5178,17 +6762,29 @@ define i8 @atomicrmw_and_i8_monotonic(ptr %a, i8 %b) nounwind {
;
; LA64-LAMCAS-LABEL: atomicrmw_and_i8_monotonic:
; LA64-LAMCAS: # %bb.0:
-; LA64-LAMCAS-NEXT: move $a2, $a0
-; LA64-LAMCAS-NEXT: ld.bu $a0, $a0, 0
-; LA64-LAMCAS-NEXT: .p2align 4, , 16
-; LA64-LAMCAS-NEXT: .LBB134_1: # %atomicrmw.start
-; LA64-LAMCAS-NEXT: # =>This Inner Loop Header: Depth=1
-; LA64-LAMCAS-NEXT: and $a3, $a0, $a1
-; LA64-LAMCAS-NEXT: ext.w.b $a4, $a0
-; LA64-LAMCAS-NEXT: amcas.b $a0, $a3, $a2
-; LA64-LAMCAS-NEXT: bne $a0, $a4, .LBB134_1
-; LA64-LAMCAS-NEXT: # %bb.2: # %atomicrmw.end
-; LA64-LAMCAS-NEXT: ret
+; LA64-LAMCAS-NEXT: slli.d $a2, $a0, 3
+; LA64-LAMCAS-NEXT: bstrins.d $a0, $zero, 1, 0
+; LA64-LAMCAS-NEXT: ori $a3, $zero, 255
+; LA64-LAMCAS-NEXT: sll.w $a3, $a3, $a2
+; LA64-LAMCAS-NEXT: andi $a1, $a1, 255
+; LA64-LAMCAS-NEXT: sll.w $a1, $a1, $a2
+; LA64-LAMCAS-NEXT: orn $a1, $a1, $a3
+; LA64-LAMCAS-NEXT: amand.w $a3, $a1, $a0
+; LA64-LAMCAS-NEXT: srl.w $a0, $a3, $a2
+; LA64-LAMCAS-NEXT: ret
+;
+; LA64-LAMCAS-LAM-BH-LABEL: atomicrmw_and_i8_monotonic:
+; LA64-LAMCAS-LAM-BH: # %bb.0:
+; LA64-LAMCAS-LAM-BH-NEXT: slli.d $a2, $a0, 3
+; LA64-LAMCAS-LAM-BH-NEXT: bstrins.d $a0, $zero, 1, 0
+; LA64-LAMCAS-LAM-BH-NEXT: ori $a3, $zero, 255
+; LA64-LAMCAS-LAM-BH-NEXT: sll.w $a3, $a3, $a2
+; LA64-LAMCAS-LAM-BH-NEXT: andi $a1, $a1, 255
+; LA64-LAMCAS-LAM-BH-NEXT: sll.w $a1, $a1, $a2
+; LA64-LAMCAS-LAM-BH-NEXT: orn $a1, $a1, $a3
+; LA64-LAMCAS-LAM-BH-NEXT: amand.w $a3, $a1, $a0
+; LA64-LAMCAS-LAM-BH-NEXT: srl.w $a0, $a3, $a2
+; LA64-LAMCAS-LAM-BH-NEXT: ret
%1 = atomicrmw and ptr %a, i8 %b monotonic
ret i8 %1
}
@@ -5210,17 +6806,31 @@ define i16 @atomicrmw_and_i16_monotonic(ptr %a, i16 %b) nounwind {
;
; LA64-LAMCAS-LABEL: atomicrmw_and_i16_monotonic:
; LA64-LAMCAS: # %bb.0:
-; LA64-LAMCAS-NEXT: move $a2, $a0
-; LA64-LAMCAS-NEXT: ld.hu $a0, $a0, 0
-; LA64-LAMCAS-NEXT: .p2align 4, , 16
-; LA64-LAMCAS-NEXT: .LBB135_1: # %atomicrmw.start
-; LA64-LAMCAS-NEXT: # =>This Inner Loop Header: Depth=1
-; LA64-LAMCAS-NEXT: and $a3, $a0, $a1
-; LA64-LAMCAS-NEXT: ext.w.h $a4, $a0
-; LA64-LAMCAS-NEXT: amcas.h $a0, $a3, $a2
-; LA64-LAMCAS-NEXT: bne $a0, $a4, .LBB135_1
-; LA64-LAMCAS-NEXT: # %bb.2: # %atomicrmw.end
-; LA64-LAMCAS-NEXT: ret
+; LA64-LAMCAS-NEXT: slli.d $a2, $a0, 3
+; LA64-LAMCAS-NEXT: bstrins.d $a0, $zero, 1, 0
+; LA64-LAMCAS-NEXT: lu12i.w $a3, 15
+; LA64-LAMCAS-NEXT: ori $a3, $a3, 4095
+; LA64-LAMCAS-NEXT: sll.w $a3, $a3, $a2
+; LA64-LAMCAS-NEXT: bstrpick.d $a1, $a1, 15, 0
+; LA64-LAMCAS-NEXT: sll.w $a1, $a1, $a2
+; LA64-LAMCAS-NEXT: orn $a1, $a1, $a3
+; LA64-LAMCAS-NEXT: amand.w $a3, $a1, $a0
+; LA64-LAMCAS-NEXT: srl.w $a0, $a3, $a2
+; LA64-LAMCAS-NEXT: ret
+;
+; LA64-LAMCAS-LAM-BH-LABEL: atomicrmw_and_i16_monotonic:
+; LA64-LAMCAS-LAM-BH: # %bb.0:
+; LA64-LAMCAS-LAM-BH-NEXT: slli.d $a2, $a0, 3
+; LA64-LAMCAS-LAM-BH-NEXT: bstrins.d $a0, $zero, 1, 0
+; LA64-LAMCAS-LAM-BH-NEXT: lu12i.w $a3, 15
+; LA64-LAMCAS-LAM-BH-NEXT: ori $a3, $a3, 4095
+; LA64-LAMCAS-LAM-BH-NEXT: sll.w $a3, $a3, $a2
+; LA64-LAMCAS-LAM-BH-NEXT: bstrpick.d $a1, $a1, 15, 0
+; LA64-LAMCAS-LAM-BH-NEXT: sll.w $a1, $a1, $a2
+; LA64-LAMCAS-LAM-BH-NEXT: orn $a1, $a1, $a3
+; LA64-LAMCAS-LAM-BH-NEXT: amand.w $a3, $a1, $a0
+; LA64-LAMCAS-LAM-BH-NEXT: srl.w $a0, $a3, $a2
+; LA64-LAMCAS-LAM-BH-NEXT: ret
%1 = atomicrmw and ptr %a, i16 %b monotonic
ret i16 %1
@@ -5239,17 +6849,23 @@ define i8 @atomicrmw_or_i8_monotonic(ptr %a, i8 %b) nounwind {
;
; LA64-LAMCAS-LABEL: atomicrmw_or_i8_monotonic:
; LA64-LAMCAS: # %bb.0:
-; LA64-LAMCAS-NEXT: move $a2, $a0
-; LA64-LAMCAS-NEXT: ld.bu $a0, $a0, 0
-; LA64-LAMCAS-NEXT: .p2align 4, , 16
-; LA64-LAMCAS-NEXT: .LBB136_1: # %atomicrmw.start
-; LA64-LAMCAS-NEXT: # =>This Inner Loop Header: Depth=1
-; LA64-LAMCAS-NEXT: or $a3, $a0, $a1
-; LA64-LAMCAS-NEXT: ext.w.b $a4, $a0
-; LA64-LAMCAS-NEXT: amcas.b $a0, $a3, $a2
-; LA64-LAMCAS-NEXT: bne $a0, $a4, .LBB136_1
-; LA64-LAMCAS-NEXT: # %bb.2: # %atomicrmw.end
-; LA64-LAMCAS-NEXT: ret
+; LA64-LAMCAS-NEXT: slli.d $a2, $a0, 3
+; LA64-LAMCAS-NEXT: bstrins.d $a0, $zero, 1, 0
+; LA64-LAMCAS-NEXT: andi $a1, $a1, 255
+; LA64-LAMCAS-NEXT: sll.w $a1, $a1, $a2
+; LA64-LAMCAS-NEXT: amor.w $a3, $a1, $a0
+; LA64-LAMCAS-NEXT: srl.w $a0, $a3, $a2
+; LA64-LAMCAS-NEXT: ret
+;
+; LA64-LAMCAS-LAM-BH-LABEL: atomicrmw_or_i8_monotonic:
+; LA64-LAMCAS-LAM-BH: # %bb.0:
+; LA64-LAMCAS-LAM-BH-NEXT: slli.d $a2, $a0, 3
+; LA64-LAMCAS-LAM-BH-NEXT: bstrins.d $a0, $zero, 1, 0
+; LA64-LAMCAS-LAM-BH-NEXT: andi $a1, $a1, 255
+; LA64-LAMCAS-LAM-BH-NEXT: sll.w $a1, $a1, $a2
+; LA64-LAMCAS-LAM-BH-NEXT: amor.w $a3, $a1, $a0
+; LA64-LAMCAS-LAM-BH-NEXT: srl.w $a0, $a3, $a2
+; LA64-LAMCAS-LAM-BH-NEXT: ret
%1 = atomicrmw or ptr %a, i8 %b monotonic
ret i8 %1
}
@@ -5267,17 +6883,23 @@ define i16 @atomicrmw_or_i16_monotonic(ptr %a, i16 %b) nounwind {
;
; LA64-LAMCAS-LABEL: atomicrmw_or_i16_monotonic:
; LA64-LAMCAS: # %bb.0:
-; LA64-LAMCAS-NEXT: move $a2, $a0
-; LA64-LAMCAS-NEXT: ld.hu $a0, $a0, 0
-; LA64-LAMCAS-NEXT: .p2align 4, , 16
-; LA64-LAMCAS-NEXT: .LBB137_1: # %atomicrmw.start
-; LA64-LAMCAS-NEXT: # =>This Inner Loop Header: Depth=1
-; LA64-LAMCAS-NEXT: or $a3, $a0, $a1
-; LA64-LAMCAS-NEXT: ext.w.h $a4, $a0
-; LA64-LAMCAS-NEXT: amcas.h $a0, $a3, $a2
-; LA64-LAMCAS-NEXT: bne $a0, $a4, .LBB137_1
-; LA64-LAMCAS-NEXT: # %bb.2: # %atomicrmw.end
-; LA64-LAMCAS-NEXT: ret
+; LA64-LAMCAS-NEXT: slli.d $a2, $a0, 3
+; LA64-LAMCAS-NEXT: bstrins.d $a0, $zero, 1, 0
+; LA64-LAMCAS-NEXT: bstrpick.d $a1, $a1, 15, 0
+; LA64-LAMCAS-NEXT: sll.w $a1, $a1, $a2
+; LA64-LAMCAS-NEXT: amor.w $a3, $a1, $a0
+; LA64-LAMCAS-NEXT: srl.w $a0, $a3, $a2
+; LA64-LAMCAS-NEXT: ret
+;
+; LA64-LAMCAS-LAM-BH-LABEL: atomicrmw_or_i16_monotonic:
+; LA64-LAMCAS-LAM-BH: # %bb.0:
+; LA64-LAMCAS-LAM-BH-NEXT: slli.d $a2, $a0, 3
+; LA64-LAMCAS-LAM-BH-NEXT: bstrins.d $a0, $zero, 1, 0
+; LA64-LAMCAS-LAM-BH-NEXT: bstrpick.d $a1, $a1, 15, 0
+; LA64-LAMCAS-LAM-BH-NEXT: sll.w $a1, $a1, $a2
+; LA64-LAMCAS-LAM-BH-NEXT: amor.w $a3, $a1, $a0
+; LA64-LAMCAS-LAM-BH-NEXT: srl.w $a0, $a3, $a2
+; LA64-LAMCAS-LAM-BH-NEXT: ret
%1 = atomicrmw or ptr %a, i16 %b monotonic
ret i16 %1
@@ -5296,17 +6918,23 @@ define i8 @atomicrmw_xor_i8_monotonic(ptr %a, i8 %b) nounwind {
;
; LA64-LAMCAS-LABEL: atomicrmw_xor_i8_monotonic:
; LA64-LAMCAS: # %bb.0:
-; LA64-LAMCAS-NEXT: move $a2, $a0
-; LA64-LAMCAS-NEXT: ld.bu $a0, $a0, 0
-; LA64-LAMCAS-NEXT: .p2align 4, , 16
-; LA64-LAMCAS-NEXT: .LBB138_1: # %atomicrmw.start
-; LA64-LAMCAS-NEXT: # =>This Inner Loop Header: Depth=1
-; LA64-LAMCAS-NEXT: xor $a3, $a0, $a1
-; LA64-LAMCAS-NEXT: ext.w.b $a4, $a0
-; LA64-LAMCAS-NEXT: amcas.b $a0, $a3, $a2
-; LA64-LAMCAS-NEXT: bne $a0, $a4, .LBB138_1
-; LA64-LAMCAS-NEXT: # %bb.2: # %atomicrmw.end
-; LA64-LAMCAS-NEXT: ret
+; LA64-LAMCAS-NEXT: slli.d $a2, $a0, 3
+; LA64-LAMCAS-NEXT: bstrins.d $a0, $zero, 1, 0
+; LA64-LAMCAS-NEXT: andi $a1, $a1, 255
+; LA64-LAMCAS-NEXT: sll.w $a1, $a1, $a2
+; LA64-LAMCAS-NEXT: amxor.w $a3, $a1, $a0
+; LA64-LAMCAS-NEXT: srl.w $a0, $a3, $a2
+; LA64-LAMCAS-NEXT: ret
+;
+; LA64-LAMCAS-LAM-BH-LABEL: atomicrmw_xor_i8_monotonic:
+; LA64-LAMCAS-LAM-BH: # %bb.0:
+; LA64-LAMCAS-LAM-BH-NEXT: slli.d $a2, $a0, 3
+; LA64-LAMCAS-LAM-BH-NEXT: bstrins.d $a0, $zero, 1, 0
+; LA64-LAMCAS-LAM-BH-NEXT: andi $a1, $a1, 255
+; LA64-LAMCAS-LAM-BH-NEXT: sll.w $a1, $a1, $a2
+; LA64-LAMCAS-LAM-BH-NEXT: amxor.w $a3, $a1, $a0
+; LA64-LAMCAS-LAM-BH-NEXT: srl.w $a0, $a3, $a2
+; LA64-LAMCAS-LAM-BH-NEXT: ret
%1 = atomicrmw xor ptr %a, i8 %b monotonic
ret i8 %1
}
@@ -5324,17 +6952,23 @@ define i16 @atomicrmw_xor_i16_monotonic(ptr %a, i16 %b) nounwind {
;
; LA64-LAMCAS-LABEL: atomicrmw_xor_i16_monotonic:
; LA64-LAMCAS: # %bb.0:
-; LA64-LAMCAS-NEXT: move $a2, $a0
-; LA64-LAMCAS-NEXT: ld.hu $a0, $a0, 0
-; LA64-LAMCAS-NEXT: .p2align 4, , 16
-; LA64-LAMCAS-NEXT: .LBB139_1: # %atomicrmw.start
-; LA64-LAMCAS-NEXT: # =>This Inner Loop Header: Depth=1
-; LA64-LAMCAS-NEXT: xor $a3, $a0, $a1
-; LA64-LAMCAS-NEXT: ext.w.h $a4, $a0
-; LA64-LAMCAS-NEXT: amcas.h $a0, $a3, $a2
-; LA64-LAMCAS-NEXT: bne $a0, $a4, .LBB139_1
-; LA64-LAMCAS-NEXT: # %bb.2: # %atomicrmw.end
-; LA64-LAMCAS-NEXT: ret
+; LA64-LAMCAS-NEXT: slli.d $a2, $a0, 3
+; LA64-LAMCAS-NEXT: bstrins.d $a0, $zero, 1, 0
+; LA64-LAMCAS-NEXT: bstrpick.d $a1, $a1, 15, 0
+; LA64-LAMCAS-NEXT: sll.w $a1, $a1, $a2
+; LA64-LAMCAS-NEXT: amxor.w $a3, $a1, $a0
+; LA64-LAMCAS-NEXT: srl.w $a0, $a3, $a2
+; LA64-LAMCAS-NEXT: ret
+;
+; LA64-LAMCAS-LAM-BH-LABEL: atomicrmw_xor_i16_monotonic:
+; LA64-LAMCAS-LAM-BH: # %bb.0:
+; LA64-LAMCAS-LAM-BH-NEXT: slli.d $a2, $a0, 3
+; LA64-LAMCAS-LAM-BH-NEXT: bstrins.d $a0, $zero, 1, 0
+; LA64-LAMCAS-LAM-BH-NEXT: bstrpick.d $a1, $a1, 15, 0
+; LA64-LAMCAS-LAM-BH-NEXT: sll.w $a1, $a1, $a2
+; LA64-LAMCAS-LAM-BH-NEXT: amxor.w $a3, $a1, $a0
+; LA64-LAMCAS-LAM-BH-NEXT: srl.w $a0, $a3, $a2
+; LA64-LAMCAS-LAM-BH-NEXT: ret
%1 = atomicrmw xor ptr %a, i16 %b monotonic
ret i16 %1
>From 680bf845742bb79e1cf33826ac8eed5ea3ed51b4 Mon Sep 17 00:00:00 2001
From: tangaac <tangyan01 at loongson.cn>
Date: Tue, 26 Nov 2024 16:33:55 +0800
Subject: [PATCH 3/3] update & add extra tests, support lamcas in
sys:getHostCPUFeatures and add some additional descriptions
---
clang/test/Driver/loongarch-march.c | 2 +-
clang/test/Preprocessor/init-loongarch.c | 2 +-
.../LoongArch/LoongArchISelLowering.cpp | 4 +
llvm/lib/TargetParser/Host.cpp | 2 +-
.../ir-instruction/atomic-cmpxchg.ll | 98 ++-
.../LoongArch/atomicrmw-expand.ll | 812 ++++++++++++++++++
6 files changed, 916 insertions(+), 4 deletions(-)
create mode 100644 llvm/test/Transforms/AtomicExpand/LoongArch/atomicrmw-expand.ll
diff --git a/clang/test/Driver/loongarch-march.c b/clang/test/Driver/loongarch-march.c
index b3a6557f6f231b..f3d1dd342ed7e1 100644
--- a/clang/test/Driver/loongarch-march.c
+++ b/clang/test/Driver/loongarch-march.c
@@ -53,7 +53,7 @@
// IR-LA464: attributes #[[#]] ={{.*}}"target-cpu"="la464" {{.*}}"target-features"="+64bit,+d,+f,+lasx,+lsx,+ual"
// IR-LA64V1P0: attributes #[[#]] ={{.*}}"target-cpu"="loongarch64" {{.*}}"target-features"="+64bit,+d,+lsx,+ual"
// IR-LA64V1P1: attributes #[[#]] ={{.*}}"target-cpu"="loongarch64" {{.*}}"target-features"="+64bit,+d,+frecipe,+lam-bh,+lamcas,+ld-seq-sa,+lsx,+ual"
-// IR-LA664: attributes #[[#]] ={{.*}}"target-cpu"="la664" {{.*}}"target-features"="+64bit,+d,+f,+frecipe,+lam-bh,+lamcas,+ld-seq-sa,+lasx,+lsx,+ual"
+// IR-LA664: attributes #[[#]] ={{.*}}"target-cpu"="la664" {{.*}}"target-features"="+64bit,+d,+f,+frecipe,+lam-bh,+lamcas,+lasx,+ld-seq-sa,+lsx,+ual"
int foo(void) {
return 3;
diff --git a/clang/test/Preprocessor/init-loongarch.c b/clang/test/Preprocessor/init-loongarch.c
index aa8b1bbc1e7aea..069a108cfb8695 100644
--- a/clang/test/Preprocessor/init-loongarch.c
+++ b/clang/test/Preprocessor/init-loongarch.c
@@ -825,7 +825,7 @@
// RUN: %clang --target=loongarch64 -x c -E -dM %s -o - -march=la64v1.1 | \
// RUN: FileCheck --match-full-lines --check-prefixes=ARCH-TUNE,FRECIPE,LAM-BH,LAMCAS,LD-SEQ-SA -DARCH=la64v1.1 -DTUNE=loongarch64 %s
// RUN: %clang --target=loongarch64 -x c -E -dM %s -o - -march=la64v1.1 -Xclang -target-feature -Xclang -frecipe | \
-// RUN: FileCheck --match-full-lines --check-prefixes=ARCH-TUNE,LAM-BH,LAMCAS,lD-SEQ-SA -DARCH=la64v1.0 -DTUNE=loongarch64 %s
+// RUN: FileCheck --match-full-lines --check-prefixes=ARCH-TUNE,LAM-BH,LAMCAS,LD-SEQ-SA -DARCH=la64v1.0 -DTUNE=loongarch64 %s
// RUN: %clang --target=loongarch64 -x c -E -dM %s -o - -march=la64v1.1 -Xclang -target-feature -Xclang -lsx | \
// RUN: FileCheck --match-full-lines --check-prefixes=ARCH-TUNE,FRECIPE,LAM-BH,LAMCAS,LD-SEQ-SA -DARCH=loongarch64 -DTUNE=loongarch64 %s
// RUN: %clang --target=loongarch64 -x c -E -dM %s -o - -march=loongarch64 -Xclang -target-feature -Xclang +frecipe | \
diff --git a/llvm/lib/Target/LoongArch/LoongArchISelLowering.cpp b/llvm/lib/Target/LoongArch/LoongArchISelLowering.cpp
index da4b8353ca2b01..7297409871edc9 100644
--- a/llvm/lib/Target/LoongArch/LoongArchISelLowering.cpp
+++ b/llvm/lib/Target/LoongArch/LoongArchISelLowering.cpp
@@ -5748,6 +5748,10 @@ bool LoongArchTargetLowering::getTgtMemIntrinsic(IntrinsicInfo &Info,
}
}
+// When -mlamcas is enabled, MinCmpXchgSizeInBits will be set to 8,
+// atomicrmw and/or/xor operations with operands less than 32 bits cannot be
+// expanded to am{and/or/xor}[_db].w through AtomicExpandPass. To prevent
+// regression, we need to implement it manually.
void LoongArchTargetLowering::emitExpandAtomicRMW(AtomicRMWInst *AI) const {
AtomicRMWInst::BinOp Op = AI->getOperation();
diff --git a/llvm/lib/TargetParser/Host.cpp b/llvm/lib/TargetParser/Host.cpp
index 51d6b7cb9b1fd6..b9ce0fdf8ec3d5 100644
--- a/llvm/lib/TargetParser/Host.cpp
+++ b/llvm/lib/TargetParser/Host.cpp
@@ -2026,12 +2026,12 @@ const StringMap<bool> sys::getHostCPUFeatures() {
Features["frecipe"] = cpucfg2 & (1U << 25); // CPUCFG.2.FRECIPE
Features["lam-bh"] = cpucfg2 & (1U << 27); // CPUCFG.2.LAM_BH
+ Features["lamcas"] = cpucfg2 & (1U << 28); // CPUCFG.2.LAMCAS
Features["ld-seq-sa"] = cpucfg3 & (1U << 23); // CPUCFG.3.LD_SEQ_SA
// TODO: Need to complete.
// Features["div32"] = cpucfg2 & (1U << 26); // CPUCFG.2.DIV32
- // Features["lamcas"] = cpucfg2 & (1U << 28); // CPUCFG.2.LAMCAS
// Features["llacq-screl"] = cpucfg2 & (1U << 29); // CPUCFG.2.LLACQ_SCREL
// Features["scq"] = cpucfg2 & (1U << 30); // CPUCFG.2.SCQ
return Features;
diff --git a/llvm/test/CodeGen/LoongArch/ir-instruction/atomic-cmpxchg.ll b/llvm/test/CodeGen/LoongArch/ir-instruction/atomic-cmpxchg.ll
index bbc9052549117a..159ffa8c7238a9 100644
--- a/llvm/test/CodeGen/LoongArch/ir-instruction/atomic-cmpxchg.ll
+++ b/llvm/test/CodeGen/LoongArch/ir-instruction/atomic-cmpxchg.ll
@@ -1,5 +1,5 @@
; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py
-; RUN: llc --mtriple=loongarch64 -mattr=+d,-lamcas,-ld-seq-sa < %s | FileCheck %s --check-prefixes=LA64,NO-LD-SEQ-SA
+; RUN: llc --mtriple=loongarch64 -mattr=+d,-ld-seq-sa < %s | FileCheck %s --check-prefixes=LA64,NO-LD-SEQ-SA
; RUN: llc --mtriple=loongarch64 -mattr=+d,+ld-seq-sa < %s | FileCheck %s --check-prefixes=LA64,LD-SEQ-SA
; RUN: llc --mtriple=loongarch64 -mattr=+d,+lamcas < %s | FileCheck %s --check-prefix=LA64-LAMCAS
@@ -170,6 +170,11 @@ define void @cmpxchg_i8_acquire_monotonic(ptr %ptr, i8 %cmp, i8 %val) nounwind {
; LD-SEQ-SA-NEXT: .LBB4_3:
; LD-SEQ-SA-NEXT: .LBB4_4:
; LD-SEQ-SA-NEXT: ret
+;
+; LA64-LAMCAS-LABEL: cmpxchg_i8_acquire_monotonic:
+; LA64-LAMCAS: # %bb.0:
+; LA64-LAMCAS-NEXT: amcas_db.b $a1, $a2, $a0
+; LA64-LAMCAS-NEXT: ret
%res = cmpxchg ptr %ptr, i8 %cmp, i8 %val acquire monotonic
ret void
}
@@ -225,6 +230,11 @@ define void @cmpxchg_i16_acquire_monotonic(ptr %ptr, i16 %cmp, i16 %val) nounwin
; LD-SEQ-SA-NEXT: .LBB5_3:
; LD-SEQ-SA-NEXT: .LBB5_4:
; LD-SEQ-SA-NEXT: ret
+;
+; LA64-LAMCAS-LABEL: cmpxchg_i16_acquire_monotonic:
+; LA64-LAMCAS: # %bb.0:
+; LA64-LAMCAS-NEXT: amcas_db.h $a1, $a2, $a0
+; LA64-LAMCAS-NEXT: ret
%res = cmpxchg ptr %ptr, i16 %cmp, i16 %val acquire monotonic
ret void
}
@@ -260,6 +270,11 @@ define void @cmpxchg_i32_acquire_monotonic(ptr %ptr, i32 %cmp, i32 %val) nounwin
; LD-SEQ-SA-NEXT: .LBB6_3:
; LD-SEQ-SA-NEXT: .LBB6_4:
; LD-SEQ-SA-NEXT: ret
+;
+; LA64-LAMCAS-LABEL: cmpxchg_i32_acquire_monotonic:
+; LA64-LAMCAS: # %bb.0:
+; LA64-LAMCAS-NEXT: amcas_db.w $a1, $a2, $a0
+; LA64-LAMCAS-NEXT: ret
%res = cmpxchg ptr %ptr, i32 %cmp, i32 %val acquire monotonic
ret void
}
@@ -293,6 +308,11 @@ define void @cmpxchg_i64_acquire_monotonic(ptr %ptr, i64 %cmp, i64 %val) nounwin
; LD-SEQ-SA-NEXT: .LBB7_3:
; LD-SEQ-SA-NEXT: .LBB7_4:
; LD-SEQ-SA-NEXT: ret
+;
+; LA64-LAMCAS-LABEL: cmpxchg_i64_acquire_monotonic:
+; LA64-LAMCAS: # %bb.0:
+; LA64-LAMCAS-NEXT: amcas_db.d $a1, $a2, $a0
+; LA64-LAMCAS-NEXT: ret
%res = cmpxchg ptr %ptr, i64 %cmp, i64 %val acquire monotonic
ret void
}
@@ -620,6 +640,11 @@ define void @cmpxchg_i8_monotonic_monotonic(ptr %ptr, i8 %cmp, i8 %val) nounwind
; LD-SEQ-SA-NEXT: .LBB16_3:
; LD-SEQ-SA-NEXT: .LBB16_4:
; LD-SEQ-SA-NEXT: ret
+;
+; LA64-LAMCAS-LABEL: cmpxchg_i8_monotonic_monotonic:
+; LA64-LAMCAS: # %bb.0:
+; LA64-LAMCAS-NEXT: amcas.b $a1, $a2, $a0
+; LA64-LAMCAS-NEXT: ret
%res = cmpxchg ptr %ptr, i8 %cmp, i8 %val monotonic monotonic
ret void
}
@@ -675,6 +700,11 @@ define void @cmpxchg_i16_monotonic_monotonic(ptr %ptr, i16 %cmp, i16 %val) nounw
; LD-SEQ-SA-NEXT: .LBB17_3:
; LD-SEQ-SA-NEXT: .LBB17_4:
; LD-SEQ-SA-NEXT: ret
+;
+; LA64-LAMCAS-LABEL: cmpxchg_i16_monotonic_monotonic:
+; LA64-LAMCAS: # %bb.0:
+; LA64-LAMCAS-NEXT: amcas.h $a1, $a2, $a0
+; LA64-LAMCAS-NEXT: ret
%res = cmpxchg ptr %ptr, i16 %cmp, i16 %val monotonic monotonic
ret void
}
@@ -710,6 +740,11 @@ define void @cmpxchg_i32_monotonic_monotonic(ptr %ptr, i32 %cmp, i32 %val) nounw
; LD-SEQ-SA-NEXT: .LBB18_3:
; LD-SEQ-SA-NEXT: .LBB18_4:
; LD-SEQ-SA-NEXT: ret
+;
+; LA64-LAMCAS-LABEL: cmpxchg_i32_monotonic_monotonic:
+; LA64-LAMCAS: # %bb.0:
+; LA64-LAMCAS-NEXT: amcas.w $a1, $a2, $a0
+; LA64-LAMCAS-NEXT: ret
%res = cmpxchg ptr %ptr, i32 %cmp, i32 %val monotonic monotonic
ret void
}
@@ -743,6 +778,11 @@ define void @cmpxchg_i64_monotonic_monotonic(ptr %ptr, i64 %cmp, i64 %val) nounw
; LD-SEQ-SA-NEXT: .LBB19_3:
; LD-SEQ-SA-NEXT: .LBB19_4:
; LD-SEQ-SA-NEXT: ret
+;
+; LA64-LAMCAS-LABEL: cmpxchg_i64_monotonic_monotonic:
+; LA64-LAMCAS: # %bb.0:
+; LA64-LAMCAS-NEXT: amcas.d $a1, $a2, $a0
+; LA64-LAMCAS-NEXT: ret
%res = cmpxchg ptr %ptr, i64 %cmp, i64 %val monotonic monotonic
ret void
}
@@ -798,6 +838,12 @@ define i8 @cmpxchg_i8_monotonic_monotonic_reti8(ptr %ptr, i8 %cmp, i8 %val) noun
; LD-SEQ-SA-NEXT: .LBB20_4:
; LD-SEQ-SA-NEXT: srl.w $a0, $a5, $a3
; LD-SEQ-SA-NEXT: ret
+;
+; LA64-LAMCAS-LABEL: cmpxchg_i8_monotonic_monotonic_reti8:
+; LA64-LAMCAS: # %bb.0:
+; LA64-LAMCAS-NEXT: amcas.b $a1, $a2, $a0
+; LA64-LAMCAS-NEXT: move $a0, $a1
+; LA64-LAMCAS-NEXT: ret
%tmp = cmpxchg ptr %ptr, i8 %cmp, i8 %val monotonic monotonic
%res = extractvalue { i8, i1 } %tmp, 0
ret i8 %res
@@ -856,6 +902,12 @@ define i16 @cmpxchg_i16_monotonic_monotonic_reti16(ptr %ptr, i16 %cmp, i16 %val)
; LD-SEQ-SA-NEXT: .LBB21_4:
; LD-SEQ-SA-NEXT: srl.w $a0, $a5, $a3
; LD-SEQ-SA-NEXT: ret
+;
+; LA64-LAMCAS-LABEL: cmpxchg_i16_monotonic_monotonic_reti16:
+; LA64-LAMCAS: # %bb.0:
+; LA64-LAMCAS-NEXT: amcas.h $a1, $a2, $a0
+; LA64-LAMCAS-NEXT: move $a0, $a1
+; LA64-LAMCAS-NEXT: ret
%tmp = cmpxchg ptr %ptr, i16 %cmp, i16 %val monotonic monotonic
%res = extractvalue { i16, i1 } %tmp, 0
ret i16 %res
@@ -894,6 +946,12 @@ define i32 @cmpxchg_i32_monotonic_monotonic_reti32(ptr %ptr, i32 %cmp, i32 %val)
; LD-SEQ-SA-NEXT: .LBB22_4:
; LD-SEQ-SA-NEXT: move $a0, $a1
; LD-SEQ-SA-NEXT: ret
+;
+; LA64-LAMCAS-LABEL: cmpxchg_i32_monotonic_monotonic_reti32:
+; LA64-LAMCAS: # %bb.0:
+; LA64-LAMCAS-NEXT: amcas.w $a1, $a2, $a0
+; LA64-LAMCAS-NEXT: move $a0, $a1
+; LA64-LAMCAS-NEXT: ret
%tmp = cmpxchg ptr %ptr, i32 %cmp, i32 %val monotonic monotonic
%res = extractvalue { i32, i1 } %tmp, 0
ret i32 %res
@@ -930,6 +988,12 @@ define i64 @cmpxchg_i64_monotonic_monotonic_reti64(ptr %ptr, i64 %cmp, i64 %val)
; LD-SEQ-SA-NEXT: .LBB23_4:
; LD-SEQ-SA-NEXT: move $a0, $a3
; LD-SEQ-SA-NEXT: ret
+;
+; LA64-LAMCAS-LABEL: cmpxchg_i64_monotonic_monotonic_reti64:
+; LA64-LAMCAS: # %bb.0:
+; LA64-LAMCAS-NEXT: amcas.d $a1, $a2, $a0
+; LA64-LAMCAS-NEXT: move $a0, $a1
+; LA64-LAMCAS-NEXT: ret
%tmp = cmpxchg ptr %ptr, i64 %cmp, i64 %val monotonic monotonic
%res = extractvalue { i64, i1 } %tmp, 0
ret i64 %res
@@ -990,6 +1054,14 @@ define i1 @cmpxchg_i8_monotonic_monotonic_reti1(ptr %ptr, i8 %cmp, i8 %val) noun
; LD-SEQ-SA-NEXT: xor $a0, $a1, $a0
; LD-SEQ-SA-NEXT: sltui $a0, $a0, 1
; LD-SEQ-SA-NEXT: ret
+;
+; LA64-LAMCAS-LABEL: cmpxchg_i8_monotonic_monotonic_reti1:
+; LA64-LAMCAS: # %bb.0:
+; LA64-LAMCAS-NEXT: ext.w.b $a3, $a1
+; LA64-LAMCAS-NEXT: amcas.b $a1, $a2, $a0
+; LA64-LAMCAS-NEXT: xor $a0, $a1, $a3
+; LA64-LAMCAS-NEXT: sltui $a0, $a0, 1
+; LA64-LAMCAS-NEXT: ret
%tmp = cmpxchg ptr %ptr, i8 %cmp, i8 %val monotonic monotonic
%res = extractvalue { i8, i1 } %tmp, 1
ret i1 %res
@@ -1052,6 +1124,14 @@ define i1 @cmpxchg_i16_monotonic_monotonic_reti1(ptr %ptr, i16 %cmp, i16 %val) n
; LD-SEQ-SA-NEXT: xor $a0, $a1, $a0
; LD-SEQ-SA-NEXT: sltui $a0, $a0, 1
; LD-SEQ-SA-NEXT: ret
+;
+; LA64-LAMCAS-LABEL: cmpxchg_i16_monotonic_monotonic_reti1:
+; LA64-LAMCAS: # %bb.0:
+; LA64-LAMCAS-NEXT: ext.w.h $a3, $a1
+; LA64-LAMCAS-NEXT: amcas.h $a1, $a2, $a0
+; LA64-LAMCAS-NEXT: xor $a0, $a1, $a3
+; LA64-LAMCAS-NEXT: sltui $a0, $a0, 1
+; LA64-LAMCAS-NEXT: ret
%tmp = cmpxchg ptr %ptr, i16 %cmp, i16 %val monotonic monotonic
%res = extractvalue { i16, i1 } %tmp, 1
ret i1 %res
@@ -1092,6 +1172,14 @@ define i1 @cmpxchg_i32_monotonic_monotonic_reti1(ptr %ptr, i32 %cmp, i32 %val) n
; LD-SEQ-SA-NEXT: xor $a0, $a3, $a1
; LD-SEQ-SA-NEXT: sltui $a0, $a0, 1
; LD-SEQ-SA-NEXT: ret
+;
+; LA64-LAMCAS-LABEL: cmpxchg_i32_monotonic_monotonic_reti1:
+; LA64-LAMCAS: # %bb.0:
+; LA64-LAMCAS-NEXT: addi.w $a3, $a1, 0
+; LA64-LAMCAS-NEXT: amcas.w $a1, $a2, $a0
+; LA64-LAMCAS-NEXT: xor $a0, $a1, $a3
+; LA64-LAMCAS-NEXT: sltui $a0, $a0, 1
+; LA64-LAMCAS-NEXT: ret
%tmp = cmpxchg ptr %ptr, i32 %cmp, i32 %val monotonic monotonic
%res = extractvalue { i32, i1 } %tmp, 1
ret i1 %res
@@ -1130,6 +1218,14 @@ define i1 @cmpxchg_i64_monotonic_monotonic_reti1(ptr %ptr, i64 %cmp, i64 %val) n
; LD-SEQ-SA-NEXT: xor $a0, $a3, $a1
; LD-SEQ-SA-NEXT: sltui $a0, $a0, 1
; LD-SEQ-SA-NEXT: ret
+;
+; LA64-LAMCAS-LABEL: cmpxchg_i64_monotonic_monotonic_reti1:
+; LA64-LAMCAS: # %bb.0:
+; LA64-LAMCAS-NEXT: move $a3, $a1
+; LA64-LAMCAS-NEXT: amcas.d $a3, $a2, $a0
+; LA64-LAMCAS-NEXT: xor $a0, $a3, $a1
+; LA64-LAMCAS-NEXT: sltui $a0, $a0, 1
+; LA64-LAMCAS-NEXT: ret
%tmp = cmpxchg ptr %ptr, i64 %cmp, i64 %val monotonic monotonic
%res = extractvalue { i64, i1 } %tmp, 1
ret i1 %res
diff --git a/llvm/test/Transforms/AtomicExpand/LoongArch/atomicrmw-expand.ll b/llvm/test/Transforms/AtomicExpand/LoongArch/atomicrmw-expand.ll
new file mode 100644
index 00000000000000..7f06a926a70345
--- /dev/null
+++ b/llvm/test/Transforms/AtomicExpand/LoongArch/atomicrmw-expand.ll
@@ -0,0 +1,812 @@
+; NOTE: Assertions have been autogenerated by utils/update_test_checks.py UTC_ARGS: --version 5
+; RUN: opt -S --mtriple=loongarch64 --mattr=+d,+lamcas %s | FileCheck %s --check-prefix=NO-EXPAND
+; RUN: opt -S --mtriple=loongarch64 --passes=atomic-expand --mattr=+d,+lamcas %s | FileCheck %s --check-prefix=EXPAND
+
+; When -mlamcas is enabled, all atomicrmw and/or/xor ptr %a,a i8/i16 %b ordering
+; will be expanded to am{and/or/xo}[_db].w by LoongArchTargetLowering::emitExpandAtomicRMW
+
+define i8 @atomicrmw_and_i8_acquire(ptr %a, i8 %b) nounwind {
+; NO-EXPAND-LABEL: define i8 @atomicrmw_and_i8_acquire(
+; NO-EXPAND-SAME: ptr [[A:%.*]], i8 [[B:%.*]]) #[[ATTR0:[0-9]+]] {
+; NO-EXPAND-NEXT: [[TMP1:%.*]] = atomicrmw and ptr [[A]], i8 [[B]] acquire, align 1
+; NO-EXPAND-NEXT: ret i8 [[TMP1]]
+;
+; EXPAND-LABEL: define i8 @atomicrmw_and_i8_acquire(
+; EXPAND-SAME: ptr [[A:%.*]], i8 [[B:%.*]]) #[[ATTR0:[0-9]+]] {
+; EXPAND-NEXT: [[ALIGNEDADDR:%.*]] = call ptr @llvm.ptrmask.p0.i64(ptr [[A]], i64 -4)
+; EXPAND-NEXT: [[TMP1:%.*]] = ptrtoint ptr [[A]] to i64
+; EXPAND-NEXT: [[PTRLSB:%.*]] = and i64 [[TMP1]], 3
+; EXPAND-NEXT: [[TMP2:%.*]] = shl i64 [[PTRLSB]], 3
+; EXPAND-NEXT: [[SHIFTAMT:%.*]] = trunc i64 [[TMP2]] to i32
+; EXPAND-NEXT: [[MASK:%.*]] = shl i32 255, [[SHIFTAMT]]
+; EXPAND-NEXT: [[INV_MASK:%.*]] = xor i32 [[MASK]], -1
+; EXPAND-NEXT: [[TMP3:%.*]] = zext i8 [[B]] to i32
+; EXPAND-NEXT: [[VALOPERAND_SHIFTED:%.*]] = shl i32 [[TMP3]], [[SHIFTAMT]]
+; EXPAND-NEXT: [[ANDOPERAND:%.*]] = or i32 [[VALOPERAND_SHIFTED]], [[INV_MASK]]
+; EXPAND-NEXT: [[TMP4:%.*]] = atomicrmw and ptr [[ALIGNEDADDR]], i32 [[ANDOPERAND]] acquire, align 4
+; EXPAND-NEXT: [[SHIFTED:%.*]] = lshr i32 [[TMP4]], [[SHIFTAMT]]
+; EXPAND-NEXT: [[EXTRACTED:%.*]] = trunc i32 [[SHIFTED]] to i8
+; EXPAND-NEXT: ret i8 [[EXTRACTED]]
+;
+ %1 = atomicrmw and ptr %a, i8 %b acquire
+ ret i8 %1
+}
+
+define i16 @atomicrmw_and_i16_acquire(ptr %a, i16 %b) nounwind {
+; NO-EXPAND-LABEL: define i16 @atomicrmw_and_i16_acquire(
+; NO-EXPAND-SAME: ptr [[A:%.*]], i16 [[B:%.*]]) #[[ATTR0]] {
+; NO-EXPAND-NEXT: [[TMP1:%.*]] = atomicrmw and ptr [[A]], i16 [[B]] acquire, align 2
+; NO-EXPAND-NEXT: ret i16 [[TMP1]]
+;
+; EXPAND-LABEL: define i16 @atomicrmw_and_i16_acquire(
+; EXPAND-SAME: ptr [[A:%.*]], i16 [[B:%.*]]) #[[ATTR0]] {
+; EXPAND-NEXT: [[ALIGNEDADDR:%.*]] = call ptr @llvm.ptrmask.p0.i64(ptr [[A]], i64 -4)
+; EXPAND-NEXT: [[TMP1:%.*]] = ptrtoint ptr [[A]] to i64
+; EXPAND-NEXT: [[PTRLSB:%.*]] = and i64 [[TMP1]], 3
+; EXPAND-NEXT: [[TMP2:%.*]] = shl i64 [[PTRLSB]], 3
+; EXPAND-NEXT: [[SHIFTAMT:%.*]] = trunc i64 [[TMP2]] to i32
+; EXPAND-NEXT: [[MASK:%.*]] = shl i32 65535, [[SHIFTAMT]]
+; EXPAND-NEXT: [[INV_MASK:%.*]] = xor i32 [[MASK]], -1
+; EXPAND-NEXT: [[TMP3:%.*]] = zext i16 [[B]] to i32
+; EXPAND-NEXT: [[VALOPERAND_SHIFTED:%.*]] = shl i32 [[TMP3]], [[SHIFTAMT]]
+; EXPAND-NEXT: [[ANDOPERAND:%.*]] = or i32 [[VALOPERAND_SHIFTED]], [[INV_MASK]]
+; EXPAND-NEXT: [[TMP4:%.*]] = atomicrmw and ptr [[ALIGNEDADDR]], i32 [[ANDOPERAND]] acquire, align 4
+; EXPAND-NEXT: [[SHIFTED:%.*]] = lshr i32 [[TMP4]], [[SHIFTAMT]]
+; EXPAND-NEXT: [[EXTRACTED:%.*]] = trunc i32 [[SHIFTED]] to i16
+; EXPAND-NEXT: ret i16 [[EXTRACTED]]
+;
+ %1 = atomicrmw and ptr %a, i16 %b acquire
+ ret i16 %1
+
+}
+
+define i8 @atomicrmw_or_i8_acquire(ptr %a, i8 %b) nounwind {
+; NO-EXPAND-LABEL: define i8 @atomicrmw_or_i8_acquire(
+; NO-EXPAND-SAME: ptr [[A:%.*]], i8 [[B:%.*]]) #[[ATTR0]] {
+; NO-EXPAND-NEXT: [[TMP1:%.*]] = atomicrmw or ptr [[A]], i8 [[B]] acquire, align 1
+; NO-EXPAND-NEXT: ret i8 [[TMP1]]
+;
+; EXPAND-LABEL: define i8 @atomicrmw_or_i8_acquire(
+; EXPAND-SAME: ptr [[A:%.*]], i8 [[B:%.*]]) #[[ATTR0]] {
+; EXPAND-NEXT: [[ALIGNEDADDR:%.*]] = call ptr @llvm.ptrmask.p0.i64(ptr [[A]], i64 -4)
+; EXPAND-NEXT: [[TMP1:%.*]] = ptrtoint ptr [[A]] to i64
+; EXPAND-NEXT: [[PTRLSB:%.*]] = and i64 [[TMP1]], 3
+; EXPAND-NEXT: [[TMP2:%.*]] = shl i64 [[PTRLSB]], 3
+; EXPAND-NEXT: [[SHIFTAMT:%.*]] = trunc i64 [[TMP2]] to i32
+; EXPAND-NEXT: [[MASK:%.*]] = shl i32 255, [[SHIFTAMT]]
+; EXPAND-NEXT: [[INV_MASK:%.*]] = xor i32 [[MASK]], -1
+; EXPAND-NEXT: [[TMP3:%.*]] = zext i8 [[B]] to i32
+; EXPAND-NEXT: [[VALOPERAND_SHIFTED:%.*]] = shl i32 [[TMP3]], [[SHIFTAMT]]
+; EXPAND-NEXT: [[TMP4:%.*]] = atomicrmw or ptr [[ALIGNEDADDR]], i32 [[VALOPERAND_SHIFTED]] acquire, align 4
+; EXPAND-NEXT: [[SHIFTED:%.*]] = lshr i32 [[TMP4]], [[SHIFTAMT]]
+; EXPAND-NEXT: [[EXTRACTED:%.*]] = trunc i32 [[SHIFTED]] to i8
+; EXPAND-NEXT: ret i8 [[EXTRACTED]]
+;
+ %1 = atomicrmw or ptr %a, i8 %b acquire
+ ret i8 %1
+}
+
+define i16 @atomicrmw_or_i16_acquire(ptr %a, i16 %b) nounwind {
+; NO-EXPAND-LABEL: define i16 @atomicrmw_or_i16_acquire(
+; NO-EXPAND-SAME: ptr [[A:%.*]], i16 [[B:%.*]]) #[[ATTR0]] {
+; NO-EXPAND-NEXT: [[TMP1:%.*]] = atomicrmw or ptr [[A]], i16 [[B]] acquire, align 2
+; NO-EXPAND-NEXT: ret i16 [[TMP1]]
+;
+; EXPAND-LABEL: define i16 @atomicrmw_or_i16_acquire(
+; EXPAND-SAME: ptr [[A:%.*]], i16 [[B:%.*]]) #[[ATTR0]] {
+; EXPAND-NEXT: [[ALIGNEDADDR:%.*]] = call ptr @llvm.ptrmask.p0.i64(ptr [[A]], i64 -4)
+; EXPAND-NEXT: [[TMP1:%.*]] = ptrtoint ptr [[A]] to i64
+; EXPAND-NEXT: [[PTRLSB:%.*]] = and i64 [[TMP1]], 3
+; EXPAND-NEXT: [[TMP2:%.*]] = shl i64 [[PTRLSB]], 3
+; EXPAND-NEXT: [[SHIFTAMT:%.*]] = trunc i64 [[TMP2]] to i32
+; EXPAND-NEXT: [[MASK:%.*]] = shl i32 65535, [[SHIFTAMT]]
+; EXPAND-NEXT: [[INV_MASK:%.*]] = xor i32 [[MASK]], -1
+; EXPAND-NEXT: [[TMP3:%.*]] = zext i16 [[B]] to i32
+; EXPAND-NEXT: [[VALOPERAND_SHIFTED:%.*]] = shl i32 [[TMP3]], [[SHIFTAMT]]
+; EXPAND-NEXT: [[TMP4:%.*]] = atomicrmw or ptr [[ALIGNEDADDR]], i32 [[VALOPERAND_SHIFTED]] acquire, align 4
+; EXPAND-NEXT: [[SHIFTED:%.*]] = lshr i32 [[TMP4]], [[SHIFTAMT]]
+; EXPAND-NEXT: [[EXTRACTED:%.*]] = trunc i32 [[SHIFTED]] to i16
+; EXPAND-NEXT: ret i16 [[EXTRACTED]]
+;
+ %1 = atomicrmw or ptr %a, i16 %b acquire
+ ret i16 %1
+
+}
+
+define i8 @atomicrmw_xor_i8_acquire(ptr %a, i8 %b) nounwind {
+; NO-EXPAND-LABEL: define i8 @atomicrmw_xor_i8_acquire(
+; NO-EXPAND-SAME: ptr [[A:%.*]], i8 [[B:%.*]]) #[[ATTR0]] {
+; NO-EXPAND-NEXT: [[TMP1:%.*]] = atomicrmw xor ptr [[A]], i8 [[B]] acquire, align 1
+; NO-EXPAND-NEXT: ret i8 [[TMP1]]
+;
+; EXPAND-LABEL: define i8 @atomicrmw_xor_i8_acquire(
+; EXPAND-SAME: ptr [[A:%.*]], i8 [[B:%.*]]) #[[ATTR0]] {
+; EXPAND-NEXT: [[ALIGNEDADDR:%.*]] = call ptr @llvm.ptrmask.p0.i64(ptr [[A]], i64 -4)
+; EXPAND-NEXT: [[TMP1:%.*]] = ptrtoint ptr [[A]] to i64
+; EXPAND-NEXT: [[PTRLSB:%.*]] = and i64 [[TMP1]], 3
+; EXPAND-NEXT: [[TMP2:%.*]] = shl i64 [[PTRLSB]], 3
+; EXPAND-NEXT: [[SHIFTAMT:%.*]] = trunc i64 [[TMP2]] to i32
+; EXPAND-NEXT: [[MASK:%.*]] = shl i32 255, [[SHIFTAMT]]
+; EXPAND-NEXT: [[INV_MASK:%.*]] = xor i32 [[MASK]], -1
+; EXPAND-NEXT: [[TMP3:%.*]] = zext i8 [[B]] to i32
+; EXPAND-NEXT: [[VALOPERAND_SHIFTED:%.*]] = shl i32 [[TMP3]], [[SHIFTAMT]]
+; EXPAND-NEXT: [[TMP4:%.*]] = atomicrmw xor ptr [[ALIGNEDADDR]], i32 [[VALOPERAND_SHIFTED]] acquire, align 4
+; EXPAND-NEXT: [[SHIFTED:%.*]] = lshr i32 [[TMP4]], [[SHIFTAMT]]
+; EXPAND-NEXT: [[EXTRACTED:%.*]] = trunc i32 [[SHIFTED]] to i8
+; EXPAND-NEXT: ret i8 [[EXTRACTED]]
+;
+ %1 = atomicrmw xor ptr %a, i8 %b acquire
+ ret i8 %1
+}
+
+define i16 @atomicrmw_xor_i16_acquire(ptr %a, i16 %b) nounwind {
+; NO-EXPAND-LABEL: define i16 @atomicrmw_xor_i16_acquire(
+; NO-EXPAND-SAME: ptr [[A:%.*]], i16 [[B:%.*]]) #[[ATTR0]] {
+; NO-EXPAND-NEXT: [[TMP1:%.*]] = atomicrmw xor ptr [[A]], i16 [[B]] acquire, align 2
+; NO-EXPAND-NEXT: ret i16 [[TMP1]]
+;
+; EXPAND-LABEL: define i16 @atomicrmw_xor_i16_acquire(
+; EXPAND-SAME: ptr [[A:%.*]], i16 [[B:%.*]]) #[[ATTR0]] {
+; EXPAND-NEXT: [[ALIGNEDADDR:%.*]] = call ptr @llvm.ptrmask.p0.i64(ptr [[A]], i64 -4)
+; EXPAND-NEXT: [[TMP1:%.*]] = ptrtoint ptr [[A]] to i64
+; EXPAND-NEXT: [[PTRLSB:%.*]] = and i64 [[TMP1]], 3
+; EXPAND-NEXT: [[TMP2:%.*]] = shl i64 [[PTRLSB]], 3
+; EXPAND-NEXT: [[SHIFTAMT:%.*]] = trunc i64 [[TMP2]] to i32
+; EXPAND-NEXT: [[MASK:%.*]] = shl i32 65535, [[SHIFTAMT]]
+; EXPAND-NEXT: [[INV_MASK:%.*]] = xor i32 [[MASK]], -1
+; EXPAND-NEXT: [[TMP3:%.*]] = zext i16 [[B]] to i32
+; EXPAND-NEXT: [[VALOPERAND_SHIFTED:%.*]] = shl i32 [[TMP3]], [[SHIFTAMT]]
+; EXPAND-NEXT: [[TMP4:%.*]] = atomicrmw xor ptr [[ALIGNEDADDR]], i32 [[VALOPERAND_SHIFTED]] acquire, align 4
+; EXPAND-NEXT: [[SHIFTED:%.*]] = lshr i32 [[TMP4]], [[SHIFTAMT]]
+; EXPAND-NEXT: [[EXTRACTED:%.*]] = trunc i32 [[SHIFTED]] to i16
+; EXPAND-NEXT: ret i16 [[EXTRACTED]]
+;
+ %1 = atomicrmw xor ptr %a, i16 %b acquire
+ ret i16 %1
+
+}
+
+define i8 @atomicrmw_and_i8_release(ptr %a, i8 %b) nounwind {
+; NO-EXPAND-LABEL: define i8 @atomicrmw_and_i8_release(
+; NO-EXPAND-SAME: ptr [[A:%.*]], i8 [[B:%.*]]) #[[ATTR0]] {
+; NO-EXPAND-NEXT: [[TMP1:%.*]] = atomicrmw and ptr [[A]], i8 [[B]] release, align 1
+; NO-EXPAND-NEXT: ret i8 [[TMP1]]
+;
+; EXPAND-LABEL: define i8 @atomicrmw_and_i8_release(
+; EXPAND-SAME: ptr [[A:%.*]], i8 [[B:%.*]]) #[[ATTR0]] {
+; EXPAND-NEXT: [[ALIGNEDADDR:%.*]] = call ptr @llvm.ptrmask.p0.i64(ptr [[A]], i64 -4)
+; EXPAND-NEXT: [[TMP1:%.*]] = ptrtoint ptr [[A]] to i64
+; EXPAND-NEXT: [[PTRLSB:%.*]] = and i64 [[TMP1]], 3
+; EXPAND-NEXT: [[TMP2:%.*]] = shl i64 [[PTRLSB]], 3
+; EXPAND-NEXT: [[SHIFTAMT:%.*]] = trunc i64 [[TMP2]] to i32
+; EXPAND-NEXT: [[MASK:%.*]] = shl i32 255, [[SHIFTAMT]]
+; EXPAND-NEXT: [[INV_MASK:%.*]] = xor i32 [[MASK]], -1
+; EXPAND-NEXT: [[TMP3:%.*]] = zext i8 [[B]] to i32
+; EXPAND-NEXT: [[VALOPERAND_SHIFTED:%.*]] = shl i32 [[TMP3]], [[SHIFTAMT]]
+; EXPAND-NEXT: [[ANDOPERAND:%.*]] = or i32 [[VALOPERAND_SHIFTED]], [[INV_MASK]]
+; EXPAND-NEXT: [[TMP4:%.*]] = atomicrmw and ptr [[ALIGNEDADDR]], i32 [[ANDOPERAND]] release, align 4
+; EXPAND-NEXT: [[SHIFTED:%.*]] = lshr i32 [[TMP4]], [[SHIFTAMT]]
+; EXPAND-NEXT: [[EXTRACTED:%.*]] = trunc i32 [[SHIFTED]] to i8
+; EXPAND-NEXT: ret i8 [[EXTRACTED]]
+;
+ %1 = atomicrmw and ptr %a, i8 %b release
+ ret i8 %1
+}
+
+define i16 @atomicrmw_and_i16_release(ptr %a, i16 %b) nounwind {
+; NO-EXPAND-LABEL: define i16 @atomicrmw_and_i16_release(
+; NO-EXPAND-SAME: ptr [[A:%.*]], i16 [[B:%.*]]) #[[ATTR0]] {
+; NO-EXPAND-NEXT: [[TMP1:%.*]] = atomicrmw and ptr [[A]], i16 [[B]] release, align 2
+; NO-EXPAND-NEXT: ret i16 [[TMP1]]
+;
+; EXPAND-LABEL: define i16 @atomicrmw_and_i16_release(
+; EXPAND-SAME: ptr [[A:%.*]], i16 [[B:%.*]]) #[[ATTR0]] {
+; EXPAND-NEXT: [[ALIGNEDADDR:%.*]] = call ptr @llvm.ptrmask.p0.i64(ptr [[A]], i64 -4)
+; EXPAND-NEXT: [[TMP1:%.*]] = ptrtoint ptr [[A]] to i64
+; EXPAND-NEXT: [[PTRLSB:%.*]] = and i64 [[TMP1]], 3
+; EXPAND-NEXT: [[TMP2:%.*]] = shl i64 [[PTRLSB]], 3
+; EXPAND-NEXT: [[SHIFTAMT:%.*]] = trunc i64 [[TMP2]] to i32
+; EXPAND-NEXT: [[MASK:%.*]] = shl i32 65535, [[SHIFTAMT]]
+; EXPAND-NEXT: [[INV_MASK:%.*]] = xor i32 [[MASK]], -1
+; EXPAND-NEXT: [[TMP3:%.*]] = zext i16 [[B]] to i32
+; EXPAND-NEXT: [[VALOPERAND_SHIFTED:%.*]] = shl i32 [[TMP3]], [[SHIFTAMT]]
+; EXPAND-NEXT: [[ANDOPERAND:%.*]] = or i32 [[VALOPERAND_SHIFTED]], [[INV_MASK]]
+; EXPAND-NEXT: [[TMP4:%.*]] = atomicrmw and ptr [[ALIGNEDADDR]], i32 [[ANDOPERAND]] release, align 4
+; EXPAND-NEXT: [[SHIFTED:%.*]] = lshr i32 [[TMP4]], [[SHIFTAMT]]
+; EXPAND-NEXT: [[EXTRACTED:%.*]] = trunc i32 [[SHIFTED]] to i16
+; EXPAND-NEXT: ret i16 [[EXTRACTED]]
+;
+ %1 = atomicrmw and ptr %a, i16 %b release
+ ret i16 %1
+
+}
+
+define i8 @atomicrmw_or_i8_release(ptr %a, i8 %b) nounwind {
+; NO-EXPAND-LABEL: define i8 @atomicrmw_or_i8_release(
+; NO-EXPAND-SAME: ptr [[A:%.*]], i8 [[B:%.*]]) #[[ATTR0]] {
+; NO-EXPAND-NEXT: [[TMP1:%.*]] = atomicrmw or ptr [[A]], i8 [[B]] release, align 1
+; NO-EXPAND-NEXT: ret i8 [[TMP1]]
+;
+; EXPAND-LABEL: define i8 @atomicrmw_or_i8_release(
+; EXPAND-SAME: ptr [[A:%.*]], i8 [[B:%.*]]) #[[ATTR0]] {
+; EXPAND-NEXT: [[ALIGNEDADDR:%.*]] = call ptr @llvm.ptrmask.p0.i64(ptr [[A]], i64 -4)
+; EXPAND-NEXT: [[TMP1:%.*]] = ptrtoint ptr [[A]] to i64
+; EXPAND-NEXT: [[PTRLSB:%.*]] = and i64 [[TMP1]], 3
+; EXPAND-NEXT: [[TMP2:%.*]] = shl i64 [[PTRLSB]], 3
+; EXPAND-NEXT: [[SHIFTAMT:%.*]] = trunc i64 [[TMP2]] to i32
+; EXPAND-NEXT: [[MASK:%.*]] = shl i32 255, [[SHIFTAMT]]
+; EXPAND-NEXT: [[INV_MASK:%.*]] = xor i32 [[MASK]], -1
+; EXPAND-NEXT: [[TMP3:%.*]] = zext i8 [[B]] to i32
+; EXPAND-NEXT: [[VALOPERAND_SHIFTED:%.*]] = shl i32 [[TMP3]], [[SHIFTAMT]]
+; EXPAND-NEXT: [[TMP4:%.*]] = atomicrmw or ptr [[ALIGNEDADDR]], i32 [[VALOPERAND_SHIFTED]] release, align 4
+; EXPAND-NEXT: [[SHIFTED:%.*]] = lshr i32 [[TMP4]], [[SHIFTAMT]]
+; EXPAND-NEXT: [[EXTRACTED:%.*]] = trunc i32 [[SHIFTED]] to i8
+; EXPAND-NEXT: ret i8 [[EXTRACTED]]
+;
+ %1 = atomicrmw or ptr %a, i8 %b release
+ ret i8 %1
+}
+
+define i16 @atomicrmw_or_i16_release(ptr %a, i16 %b) nounwind {
+; NO-EXPAND-LABEL: define i16 @atomicrmw_or_i16_release(
+; NO-EXPAND-SAME: ptr [[A:%.*]], i16 [[B:%.*]]) #[[ATTR0]] {
+; NO-EXPAND-NEXT: [[TMP1:%.*]] = atomicrmw or ptr [[A]], i16 [[B]] release, align 2
+; NO-EXPAND-NEXT: ret i16 [[TMP1]]
+;
+; EXPAND-LABEL: define i16 @atomicrmw_or_i16_release(
+; EXPAND-SAME: ptr [[A:%.*]], i16 [[B:%.*]]) #[[ATTR0]] {
+; EXPAND-NEXT: [[ALIGNEDADDR:%.*]] = call ptr @llvm.ptrmask.p0.i64(ptr [[A]], i64 -4)
+; EXPAND-NEXT: [[TMP1:%.*]] = ptrtoint ptr [[A]] to i64
+; EXPAND-NEXT: [[PTRLSB:%.*]] = and i64 [[TMP1]], 3
+; EXPAND-NEXT: [[TMP2:%.*]] = shl i64 [[PTRLSB]], 3
+; EXPAND-NEXT: [[SHIFTAMT:%.*]] = trunc i64 [[TMP2]] to i32
+; EXPAND-NEXT: [[MASK:%.*]] = shl i32 65535, [[SHIFTAMT]]
+; EXPAND-NEXT: [[INV_MASK:%.*]] = xor i32 [[MASK]], -1
+; EXPAND-NEXT: [[TMP3:%.*]] = zext i16 [[B]] to i32
+; EXPAND-NEXT: [[VALOPERAND_SHIFTED:%.*]] = shl i32 [[TMP3]], [[SHIFTAMT]]
+; EXPAND-NEXT: [[TMP4:%.*]] = atomicrmw or ptr [[ALIGNEDADDR]], i32 [[VALOPERAND_SHIFTED]] release, align 4
+; EXPAND-NEXT: [[SHIFTED:%.*]] = lshr i32 [[TMP4]], [[SHIFTAMT]]
+; EXPAND-NEXT: [[EXTRACTED:%.*]] = trunc i32 [[SHIFTED]] to i16
+; EXPAND-NEXT: ret i16 [[EXTRACTED]]
+;
+ %1 = atomicrmw or ptr %a, i16 %b release
+ ret i16 %1
+
+}
+
+define i8 @atomicrmw_xor_i8_release(ptr %a, i8 %b) nounwind {
+; NO-EXPAND-LABEL: define i8 @atomicrmw_xor_i8_release(
+; NO-EXPAND-SAME: ptr [[A:%.*]], i8 [[B:%.*]]) #[[ATTR0]] {
+; NO-EXPAND-NEXT: [[TMP1:%.*]] = atomicrmw xor ptr [[A]], i8 [[B]] release, align 1
+; NO-EXPAND-NEXT: ret i8 [[TMP1]]
+;
+; EXPAND-LABEL: define i8 @atomicrmw_xor_i8_release(
+; EXPAND-SAME: ptr [[A:%.*]], i8 [[B:%.*]]) #[[ATTR0]] {
+; EXPAND-NEXT: [[ALIGNEDADDR:%.*]] = call ptr @llvm.ptrmask.p0.i64(ptr [[A]], i64 -4)
+; EXPAND-NEXT: [[TMP1:%.*]] = ptrtoint ptr [[A]] to i64
+; EXPAND-NEXT: [[PTRLSB:%.*]] = and i64 [[TMP1]], 3
+; EXPAND-NEXT: [[TMP2:%.*]] = shl i64 [[PTRLSB]], 3
+; EXPAND-NEXT: [[SHIFTAMT:%.*]] = trunc i64 [[TMP2]] to i32
+; EXPAND-NEXT: [[MASK:%.*]] = shl i32 255, [[SHIFTAMT]]
+; EXPAND-NEXT: [[INV_MASK:%.*]] = xor i32 [[MASK]], -1
+; EXPAND-NEXT: [[TMP3:%.*]] = zext i8 [[B]] to i32
+; EXPAND-NEXT: [[VALOPERAND_SHIFTED:%.*]] = shl i32 [[TMP3]], [[SHIFTAMT]]
+; EXPAND-NEXT: [[TMP4:%.*]] = atomicrmw xor ptr [[ALIGNEDADDR]], i32 [[VALOPERAND_SHIFTED]] release, align 4
+; EXPAND-NEXT: [[SHIFTED:%.*]] = lshr i32 [[TMP4]], [[SHIFTAMT]]
+; EXPAND-NEXT: [[EXTRACTED:%.*]] = trunc i32 [[SHIFTED]] to i8
+; EXPAND-NEXT: ret i8 [[EXTRACTED]]
+;
+ %1 = atomicrmw xor ptr %a, i8 %b release
+ ret i8 %1
+}
+
+define i16 @atomicrmw_xor_i16_release(ptr %a, i16 %b) nounwind {
+; NO-EXPAND-LABEL: define i16 @atomicrmw_xor_i16_release(
+; NO-EXPAND-SAME: ptr [[A:%.*]], i16 [[B:%.*]]) #[[ATTR0]] {
+; NO-EXPAND-NEXT: [[TMP1:%.*]] = atomicrmw xor ptr [[A]], i16 [[B]] release, align 2
+; NO-EXPAND-NEXT: ret i16 [[TMP1]]
+;
+; EXPAND-LABEL: define i16 @atomicrmw_xor_i16_release(
+; EXPAND-SAME: ptr [[A:%.*]], i16 [[B:%.*]]) #[[ATTR0]] {
+; EXPAND-NEXT: [[ALIGNEDADDR:%.*]] = call ptr @llvm.ptrmask.p0.i64(ptr [[A]], i64 -4)
+; EXPAND-NEXT: [[TMP1:%.*]] = ptrtoint ptr [[A]] to i64
+; EXPAND-NEXT: [[PTRLSB:%.*]] = and i64 [[TMP1]], 3
+; EXPAND-NEXT: [[TMP2:%.*]] = shl i64 [[PTRLSB]], 3
+; EXPAND-NEXT: [[SHIFTAMT:%.*]] = trunc i64 [[TMP2]] to i32
+; EXPAND-NEXT: [[MASK:%.*]] = shl i32 65535, [[SHIFTAMT]]
+; EXPAND-NEXT: [[INV_MASK:%.*]] = xor i32 [[MASK]], -1
+; EXPAND-NEXT: [[TMP3:%.*]] = zext i16 [[B]] to i32
+; EXPAND-NEXT: [[VALOPERAND_SHIFTED:%.*]] = shl i32 [[TMP3]], [[SHIFTAMT]]
+; EXPAND-NEXT: [[TMP4:%.*]] = atomicrmw xor ptr [[ALIGNEDADDR]], i32 [[VALOPERAND_SHIFTED]] release, align 4
+; EXPAND-NEXT: [[SHIFTED:%.*]] = lshr i32 [[TMP4]], [[SHIFTAMT]]
+; EXPAND-NEXT: [[EXTRACTED:%.*]] = trunc i32 [[SHIFTED]] to i16
+; EXPAND-NEXT: ret i16 [[EXTRACTED]]
+;
+ %1 = atomicrmw xor ptr %a, i16 %b release
+ ret i16 %1
+
+}
+
+define i8 @atomicrmw_and_i8_acq_rel(ptr %a, i8 %b) nounwind {
+; NO-EXPAND-LABEL: define i8 @atomicrmw_and_i8_acq_rel(
+; NO-EXPAND-SAME: ptr [[A:%.*]], i8 [[B:%.*]]) #[[ATTR0]] {
+; NO-EXPAND-NEXT: [[TMP1:%.*]] = atomicrmw and ptr [[A]], i8 [[B]] acq_rel, align 1
+; NO-EXPAND-NEXT: ret i8 [[TMP1]]
+;
+; EXPAND-LABEL: define i8 @atomicrmw_and_i8_acq_rel(
+; EXPAND-SAME: ptr [[A:%.*]], i8 [[B:%.*]]) #[[ATTR0]] {
+; EXPAND-NEXT: [[ALIGNEDADDR:%.*]] = call ptr @llvm.ptrmask.p0.i64(ptr [[A]], i64 -4)
+; EXPAND-NEXT: [[TMP1:%.*]] = ptrtoint ptr [[A]] to i64
+; EXPAND-NEXT: [[PTRLSB:%.*]] = and i64 [[TMP1]], 3
+; EXPAND-NEXT: [[TMP2:%.*]] = shl i64 [[PTRLSB]], 3
+; EXPAND-NEXT: [[SHIFTAMT:%.*]] = trunc i64 [[TMP2]] to i32
+; EXPAND-NEXT: [[MASK:%.*]] = shl i32 255, [[SHIFTAMT]]
+; EXPAND-NEXT: [[INV_MASK:%.*]] = xor i32 [[MASK]], -1
+; EXPAND-NEXT: [[TMP3:%.*]] = zext i8 [[B]] to i32
+; EXPAND-NEXT: [[VALOPERAND_SHIFTED:%.*]] = shl i32 [[TMP3]], [[SHIFTAMT]]
+; EXPAND-NEXT: [[ANDOPERAND:%.*]] = or i32 [[VALOPERAND_SHIFTED]], [[INV_MASK]]
+; EXPAND-NEXT: [[TMP4:%.*]] = atomicrmw and ptr [[ALIGNEDADDR]], i32 [[ANDOPERAND]] acq_rel, align 4
+; EXPAND-NEXT: [[SHIFTED:%.*]] = lshr i32 [[TMP4]], [[SHIFTAMT]]
+; EXPAND-NEXT: [[EXTRACTED:%.*]] = trunc i32 [[SHIFTED]] to i8
+; EXPAND-NEXT: ret i8 [[EXTRACTED]]
+;
+ %1 = atomicrmw and ptr %a, i8 %b acq_rel
+ ret i8 %1
+}
+
+define i16 @atomicrmw_and_i16_acq_rel(ptr %a, i16 %b) nounwind {
+; NO-EXPAND-LABEL: define i16 @atomicrmw_and_i16_acq_rel(
+; NO-EXPAND-SAME: ptr [[A:%.*]], i16 [[B:%.*]]) #[[ATTR0]] {
+; NO-EXPAND-NEXT: [[TMP1:%.*]] = atomicrmw and ptr [[A]], i16 [[B]] acq_rel, align 2
+; NO-EXPAND-NEXT: ret i16 [[TMP1]]
+;
+; EXPAND-LABEL: define i16 @atomicrmw_and_i16_acq_rel(
+; EXPAND-SAME: ptr [[A:%.*]], i16 [[B:%.*]]) #[[ATTR0]] {
+; EXPAND-NEXT: [[ALIGNEDADDR:%.*]] = call ptr @llvm.ptrmask.p0.i64(ptr [[A]], i64 -4)
+; EXPAND-NEXT: [[TMP1:%.*]] = ptrtoint ptr [[A]] to i64
+; EXPAND-NEXT: [[PTRLSB:%.*]] = and i64 [[TMP1]], 3
+; EXPAND-NEXT: [[TMP2:%.*]] = shl i64 [[PTRLSB]], 3
+; EXPAND-NEXT: [[SHIFTAMT:%.*]] = trunc i64 [[TMP2]] to i32
+; EXPAND-NEXT: [[MASK:%.*]] = shl i32 65535, [[SHIFTAMT]]
+; EXPAND-NEXT: [[INV_MASK:%.*]] = xor i32 [[MASK]], -1
+; EXPAND-NEXT: [[TMP3:%.*]] = zext i16 [[B]] to i32
+; EXPAND-NEXT: [[VALOPERAND_SHIFTED:%.*]] = shl i32 [[TMP3]], [[SHIFTAMT]]
+; EXPAND-NEXT: [[ANDOPERAND:%.*]] = or i32 [[VALOPERAND_SHIFTED]], [[INV_MASK]]
+; EXPAND-NEXT: [[TMP4:%.*]] = atomicrmw and ptr [[ALIGNEDADDR]], i32 [[ANDOPERAND]] acq_rel, align 4
+; EXPAND-NEXT: [[SHIFTED:%.*]] = lshr i32 [[TMP4]], [[SHIFTAMT]]
+; EXPAND-NEXT: [[EXTRACTED:%.*]] = trunc i32 [[SHIFTED]] to i16
+; EXPAND-NEXT: ret i16 [[EXTRACTED]]
+;
+ %1 = atomicrmw and ptr %a, i16 %b acq_rel
+ ret i16 %1
+
+}
+
+define i8 @atomicrmw_or_i8_acq_rel(ptr %a, i8 %b) nounwind {
+; NO-EXPAND-LABEL: define i8 @atomicrmw_or_i8_acq_rel(
+; NO-EXPAND-SAME: ptr [[A:%.*]], i8 [[B:%.*]]) #[[ATTR0]] {
+; NO-EXPAND-NEXT: [[TMP1:%.*]] = atomicrmw or ptr [[A]], i8 [[B]] acq_rel, align 1
+; NO-EXPAND-NEXT: ret i8 [[TMP1]]
+;
+; EXPAND-LABEL: define i8 @atomicrmw_or_i8_acq_rel(
+; EXPAND-SAME: ptr [[A:%.*]], i8 [[B:%.*]]) #[[ATTR0]] {
+; EXPAND-NEXT: [[ALIGNEDADDR:%.*]] = call ptr @llvm.ptrmask.p0.i64(ptr [[A]], i64 -4)
+; EXPAND-NEXT: [[TMP1:%.*]] = ptrtoint ptr [[A]] to i64
+; EXPAND-NEXT: [[PTRLSB:%.*]] = and i64 [[TMP1]], 3
+; EXPAND-NEXT: [[TMP2:%.*]] = shl i64 [[PTRLSB]], 3
+; EXPAND-NEXT: [[SHIFTAMT:%.*]] = trunc i64 [[TMP2]] to i32
+; EXPAND-NEXT: [[MASK:%.*]] = shl i32 255, [[SHIFTAMT]]
+; EXPAND-NEXT: [[INV_MASK:%.*]] = xor i32 [[MASK]], -1
+; EXPAND-NEXT: [[TMP3:%.*]] = zext i8 [[B]] to i32
+; EXPAND-NEXT: [[VALOPERAND_SHIFTED:%.*]] = shl i32 [[TMP3]], [[SHIFTAMT]]
+; EXPAND-NEXT: [[TMP4:%.*]] = atomicrmw or ptr [[ALIGNEDADDR]], i32 [[VALOPERAND_SHIFTED]] acq_rel, align 4
+; EXPAND-NEXT: [[SHIFTED:%.*]] = lshr i32 [[TMP4]], [[SHIFTAMT]]
+; EXPAND-NEXT: [[EXTRACTED:%.*]] = trunc i32 [[SHIFTED]] to i8
+; EXPAND-NEXT: ret i8 [[EXTRACTED]]
+;
+ %1 = atomicrmw or ptr %a, i8 %b acq_rel
+ ret i8 %1
+}
+
+define i16 @atomicrmw_or_i16_acq_rel(ptr %a, i16 %b) nounwind {
+; NO-EXPAND-LABEL: define i16 @atomicrmw_or_i16_acq_rel(
+; NO-EXPAND-SAME: ptr [[A:%.*]], i16 [[B:%.*]]) #[[ATTR0]] {
+; NO-EXPAND-NEXT: [[TMP1:%.*]] = atomicrmw or ptr [[A]], i16 [[B]] acq_rel, align 2
+; NO-EXPAND-NEXT: ret i16 [[TMP1]]
+;
+; EXPAND-LABEL: define i16 @atomicrmw_or_i16_acq_rel(
+; EXPAND-SAME: ptr [[A:%.*]], i16 [[B:%.*]]) #[[ATTR0]] {
+; EXPAND-NEXT: [[ALIGNEDADDR:%.*]] = call ptr @llvm.ptrmask.p0.i64(ptr [[A]], i64 -4)
+; EXPAND-NEXT: [[TMP1:%.*]] = ptrtoint ptr [[A]] to i64
+; EXPAND-NEXT: [[PTRLSB:%.*]] = and i64 [[TMP1]], 3
+; EXPAND-NEXT: [[TMP2:%.*]] = shl i64 [[PTRLSB]], 3
+; EXPAND-NEXT: [[SHIFTAMT:%.*]] = trunc i64 [[TMP2]] to i32
+; EXPAND-NEXT: [[MASK:%.*]] = shl i32 65535, [[SHIFTAMT]]
+; EXPAND-NEXT: [[INV_MASK:%.*]] = xor i32 [[MASK]], -1
+; EXPAND-NEXT: [[TMP3:%.*]] = zext i16 [[B]] to i32
+; EXPAND-NEXT: [[VALOPERAND_SHIFTED:%.*]] = shl i32 [[TMP3]], [[SHIFTAMT]]
+; EXPAND-NEXT: [[TMP4:%.*]] = atomicrmw or ptr [[ALIGNEDADDR]], i32 [[VALOPERAND_SHIFTED]] acq_rel, align 4
+; EXPAND-NEXT: [[SHIFTED:%.*]] = lshr i32 [[TMP4]], [[SHIFTAMT]]
+; EXPAND-NEXT: [[EXTRACTED:%.*]] = trunc i32 [[SHIFTED]] to i16
+; EXPAND-NEXT: ret i16 [[EXTRACTED]]
+;
+ %1 = atomicrmw or ptr %a, i16 %b acq_rel
+ ret i16 %1
+
+}
+
+define i8 @atomicrmw_xor_i8_acq_rel(ptr %a, i8 %b) nounwind {
+; NO-EXPAND-LABEL: define i8 @atomicrmw_xor_i8_acq_rel(
+; NO-EXPAND-SAME: ptr [[A:%.*]], i8 [[B:%.*]]) #[[ATTR0]] {
+; NO-EXPAND-NEXT: [[TMP1:%.*]] = atomicrmw xor ptr [[A]], i8 [[B]] acq_rel, align 1
+; NO-EXPAND-NEXT: ret i8 [[TMP1]]
+;
+; EXPAND-LABEL: define i8 @atomicrmw_xor_i8_acq_rel(
+; EXPAND-SAME: ptr [[A:%.*]], i8 [[B:%.*]]) #[[ATTR0]] {
+; EXPAND-NEXT: [[ALIGNEDADDR:%.*]] = call ptr @llvm.ptrmask.p0.i64(ptr [[A]], i64 -4)
+; EXPAND-NEXT: [[TMP1:%.*]] = ptrtoint ptr [[A]] to i64
+; EXPAND-NEXT: [[PTRLSB:%.*]] = and i64 [[TMP1]], 3
+; EXPAND-NEXT: [[TMP2:%.*]] = shl i64 [[PTRLSB]], 3
+; EXPAND-NEXT: [[SHIFTAMT:%.*]] = trunc i64 [[TMP2]] to i32
+; EXPAND-NEXT: [[MASK:%.*]] = shl i32 255, [[SHIFTAMT]]
+; EXPAND-NEXT: [[INV_MASK:%.*]] = xor i32 [[MASK]], -1
+; EXPAND-NEXT: [[TMP3:%.*]] = zext i8 [[B]] to i32
+; EXPAND-NEXT: [[VALOPERAND_SHIFTED:%.*]] = shl i32 [[TMP3]], [[SHIFTAMT]]
+; EXPAND-NEXT: [[TMP4:%.*]] = atomicrmw xor ptr [[ALIGNEDADDR]], i32 [[VALOPERAND_SHIFTED]] acq_rel, align 4
+; EXPAND-NEXT: [[SHIFTED:%.*]] = lshr i32 [[TMP4]], [[SHIFTAMT]]
+; EXPAND-NEXT: [[EXTRACTED:%.*]] = trunc i32 [[SHIFTED]] to i8
+; EXPAND-NEXT: ret i8 [[EXTRACTED]]
+;
+ %1 = atomicrmw xor ptr %a, i8 %b acq_rel
+ ret i8 %1
+}
+
+define i16 @atomicrmw_xor_i16_acq_rel(ptr %a, i16 %b) nounwind {
+; NO-EXPAND-LABEL: define i16 @atomicrmw_xor_i16_acq_rel(
+; NO-EXPAND-SAME: ptr [[A:%.*]], i16 [[B:%.*]]) #[[ATTR0]] {
+; NO-EXPAND-NEXT: [[TMP1:%.*]] = atomicrmw xor ptr [[A]], i16 [[B]] acq_rel, align 2
+; NO-EXPAND-NEXT: ret i16 [[TMP1]]
+;
+; EXPAND-LABEL: define i16 @atomicrmw_xor_i16_acq_rel(
+; EXPAND-SAME: ptr [[A:%.*]], i16 [[B:%.*]]) #[[ATTR0]] {
+; EXPAND-NEXT: [[ALIGNEDADDR:%.*]] = call ptr @llvm.ptrmask.p0.i64(ptr [[A]], i64 -4)
+; EXPAND-NEXT: [[TMP1:%.*]] = ptrtoint ptr [[A]] to i64
+; EXPAND-NEXT: [[PTRLSB:%.*]] = and i64 [[TMP1]], 3
+; EXPAND-NEXT: [[TMP2:%.*]] = shl i64 [[PTRLSB]], 3
+; EXPAND-NEXT: [[SHIFTAMT:%.*]] = trunc i64 [[TMP2]] to i32
+; EXPAND-NEXT: [[MASK:%.*]] = shl i32 65535, [[SHIFTAMT]]
+; EXPAND-NEXT: [[INV_MASK:%.*]] = xor i32 [[MASK]], -1
+; EXPAND-NEXT: [[TMP3:%.*]] = zext i16 [[B]] to i32
+; EXPAND-NEXT: [[VALOPERAND_SHIFTED:%.*]] = shl i32 [[TMP3]], [[SHIFTAMT]]
+; EXPAND-NEXT: [[TMP4:%.*]] = atomicrmw xor ptr [[ALIGNEDADDR]], i32 [[VALOPERAND_SHIFTED]] acq_rel, align 4
+; EXPAND-NEXT: [[SHIFTED:%.*]] = lshr i32 [[TMP4]], [[SHIFTAMT]]
+; EXPAND-NEXT: [[EXTRACTED:%.*]] = trunc i32 [[SHIFTED]] to i16
+; EXPAND-NEXT: ret i16 [[EXTRACTED]]
+;
+ %1 = atomicrmw xor ptr %a, i16 %b acq_rel
+ ret i16 %1
+
+}
+
+define i8 @atomicrmw_and_i8_seq_cst(ptr %a, i8 %b) nounwind {
+; NO-EXPAND-LABEL: define i8 @atomicrmw_and_i8_seq_cst(
+; NO-EXPAND-SAME: ptr [[A:%.*]], i8 [[B:%.*]]) #[[ATTR0]] {
+; NO-EXPAND-NEXT: [[TMP1:%.*]] = atomicrmw and ptr [[A]], i8 [[B]] seq_cst, align 1
+; NO-EXPAND-NEXT: ret i8 [[TMP1]]
+;
+; EXPAND-LABEL: define i8 @atomicrmw_and_i8_seq_cst(
+; EXPAND-SAME: ptr [[A:%.*]], i8 [[B:%.*]]) #[[ATTR0]] {
+; EXPAND-NEXT: [[ALIGNEDADDR:%.*]] = call ptr @llvm.ptrmask.p0.i64(ptr [[A]], i64 -4)
+; EXPAND-NEXT: [[TMP1:%.*]] = ptrtoint ptr [[A]] to i64
+; EXPAND-NEXT: [[PTRLSB:%.*]] = and i64 [[TMP1]], 3
+; EXPAND-NEXT: [[TMP2:%.*]] = shl i64 [[PTRLSB]], 3
+; EXPAND-NEXT: [[SHIFTAMT:%.*]] = trunc i64 [[TMP2]] to i32
+; EXPAND-NEXT: [[MASK:%.*]] = shl i32 255, [[SHIFTAMT]]
+; EXPAND-NEXT: [[INV_MASK:%.*]] = xor i32 [[MASK]], -1
+; EXPAND-NEXT: [[TMP3:%.*]] = zext i8 [[B]] to i32
+; EXPAND-NEXT: [[VALOPERAND_SHIFTED:%.*]] = shl i32 [[TMP3]], [[SHIFTAMT]]
+; EXPAND-NEXT: [[ANDOPERAND:%.*]] = or i32 [[VALOPERAND_SHIFTED]], [[INV_MASK]]
+; EXPAND-NEXT: [[TMP4:%.*]] = atomicrmw and ptr [[ALIGNEDADDR]], i32 [[ANDOPERAND]] seq_cst, align 4
+; EXPAND-NEXT: [[SHIFTED:%.*]] = lshr i32 [[TMP4]], [[SHIFTAMT]]
+; EXPAND-NEXT: [[EXTRACTED:%.*]] = trunc i32 [[SHIFTED]] to i8
+; EXPAND-NEXT: ret i8 [[EXTRACTED]]
+;
+ %1 = atomicrmw and ptr %a, i8 %b seq_cst
+ ret i8 %1
+}
+
+define i16 @atomicrmw_and_i16_seq_cst(ptr %a, i16 %b) nounwind {
+; NO-EXPAND-LABEL: define i16 @atomicrmw_and_i16_seq_cst(
+; NO-EXPAND-SAME: ptr [[A:%.*]], i16 [[B:%.*]]) #[[ATTR0]] {
+; NO-EXPAND-NEXT: [[TMP1:%.*]] = atomicrmw and ptr [[A]], i16 [[B]] seq_cst, align 2
+; NO-EXPAND-NEXT: ret i16 [[TMP1]]
+;
+; EXPAND-LABEL: define i16 @atomicrmw_and_i16_seq_cst(
+; EXPAND-SAME: ptr [[A:%.*]], i16 [[B:%.*]]) #[[ATTR0]] {
+; EXPAND-NEXT: [[ALIGNEDADDR:%.*]] = call ptr @llvm.ptrmask.p0.i64(ptr [[A]], i64 -4)
+; EXPAND-NEXT: [[TMP1:%.*]] = ptrtoint ptr [[A]] to i64
+; EXPAND-NEXT: [[PTRLSB:%.*]] = and i64 [[TMP1]], 3
+; EXPAND-NEXT: [[TMP2:%.*]] = shl i64 [[PTRLSB]], 3
+; EXPAND-NEXT: [[SHIFTAMT:%.*]] = trunc i64 [[TMP2]] to i32
+; EXPAND-NEXT: [[MASK:%.*]] = shl i32 65535, [[SHIFTAMT]]
+; EXPAND-NEXT: [[INV_MASK:%.*]] = xor i32 [[MASK]], -1
+; EXPAND-NEXT: [[TMP3:%.*]] = zext i16 [[B]] to i32
+; EXPAND-NEXT: [[VALOPERAND_SHIFTED:%.*]] = shl i32 [[TMP3]], [[SHIFTAMT]]
+; EXPAND-NEXT: [[ANDOPERAND:%.*]] = or i32 [[VALOPERAND_SHIFTED]], [[INV_MASK]]
+; EXPAND-NEXT: [[TMP4:%.*]] = atomicrmw and ptr [[ALIGNEDADDR]], i32 [[ANDOPERAND]] seq_cst, align 4
+; EXPAND-NEXT: [[SHIFTED:%.*]] = lshr i32 [[TMP4]], [[SHIFTAMT]]
+; EXPAND-NEXT: [[EXTRACTED:%.*]] = trunc i32 [[SHIFTED]] to i16
+; EXPAND-NEXT: ret i16 [[EXTRACTED]]
+;
+ %1 = atomicrmw and ptr %a, i16 %b seq_cst
+ ret i16 %1
+
+}
+
+define i8 @atomicrmw_or_i8_seq_cst(ptr %a, i8 %b) nounwind {
+; NO-EXPAND-LABEL: define i8 @atomicrmw_or_i8_seq_cst(
+; NO-EXPAND-SAME: ptr [[A:%.*]], i8 [[B:%.*]]) #[[ATTR0]] {
+; NO-EXPAND-NEXT: [[TMP1:%.*]] = atomicrmw or ptr [[A]], i8 [[B]] seq_cst, align 1
+; NO-EXPAND-NEXT: ret i8 [[TMP1]]
+;
+; EXPAND-LABEL: define i8 @atomicrmw_or_i8_seq_cst(
+; EXPAND-SAME: ptr [[A:%.*]], i8 [[B:%.*]]) #[[ATTR0]] {
+; EXPAND-NEXT: [[ALIGNEDADDR:%.*]] = call ptr @llvm.ptrmask.p0.i64(ptr [[A]], i64 -4)
+; EXPAND-NEXT: [[TMP1:%.*]] = ptrtoint ptr [[A]] to i64
+; EXPAND-NEXT: [[PTRLSB:%.*]] = and i64 [[TMP1]], 3
+; EXPAND-NEXT: [[TMP2:%.*]] = shl i64 [[PTRLSB]], 3
+; EXPAND-NEXT: [[SHIFTAMT:%.*]] = trunc i64 [[TMP2]] to i32
+; EXPAND-NEXT: [[MASK:%.*]] = shl i32 255, [[SHIFTAMT]]
+; EXPAND-NEXT: [[INV_MASK:%.*]] = xor i32 [[MASK]], -1
+; EXPAND-NEXT: [[TMP3:%.*]] = zext i8 [[B]] to i32
+; EXPAND-NEXT: [[VALOPERAND_SHIFTED:%.*]] = shl i32 [[TMP3]], [[SHIFTAMT]]
+; EXPAND-NEXT: [[TMP4:%.*]] = atomicrmw or ptr [[ALIGNEDADDR]], i32 [[VALOPERAND_SHIFTED]] seq_cst, align 4
+; EXPAND-NEXT: [[SHIFTED:%.*]] = lshr i32 [[TMP4]], [[SHIFTAMT]]
+; EXPAND-NEXT: [[EXTRACTED:%.*]] = trunc i32 [[SHIFTED]] to i8
+; EXPAND-NEXT: ret i8 [[EXTRACTED]]
+;
+ %1 = atomicrmw or ptr %a, i8 %b seq_cst
+ ret i8 %1
+}
+
+define i16 @atomicrmw_or_i16_seq_cst(ptr %a, i16 %b) nounwind {
+; NO-EXPAND-LABEL: define i16 @atomicrmw_or_i16_seq_cst(
+; NO-EXPAND-SAME: ptr [[A:%.*]], i16 [[B:%.*]]) #[[ATTR0]] {
+; NO-EXPAND-NEXT: [[TMP1:%.*]] = atomicrmw or ptr [[A]], i16 [[B]] seq_cst, align 2
+; NO-EXPAND-NEXT: ret i16 [[TMP1]]
+;
+; EXPAND-LABEL: define i16 @atomicrmw_or_i16_seq_cst(
+; EXPAND-SAME: ptr [[A:%.*]], i16 [[B:%.*]]) #[[ATTR0]] {
+; EXPAND-NEXT: [[ALIGNEDADDR:%.*]] = call ptr @llvm.ptrmask.p0.i64(ptr [[A]], i64 -4)
+; EXPAND-NEXT: [[TMP1:%.*]] = ptrtoint ptr [[A]] to i64
+; EXPAND-NEXT: [[PTRLSB:%.*]] = and i64 [[TMP1]], 3
+; EXPAND-NEXT: [[TMP2:%.*]] = shl i64 [[PTRLSB]], 3
+; EXPAND-NEXT: [[SHIFTAMT:%.*]] = trunc i64 [[TMP2]] to i32
+; EXPAND-NEXT: [[MASK:%.*]] = shl i32 65535, [[SHIFTAMT]]
+; EXPAND-NEXT: [[INV_MASK:%.*]] = xor i32 [[MASK]], -1
+; EXPAND-NEXT: [[TMP3:%.*]] = zext i16 [[B]] to i32
+; EXPAND-NEXT: [[VALOPERAND_SHIFTED:%.*]] = shl i32 [[TMP3]], [[SHIFTAMT]]
+; EXPAND-NEXT: [[TMP4:%.*]] = atomicrmw or ptr [[ALIGNEDADDR]], i32 [[VALOPERAND_SHIFTED]] seq_cst, align 4
+; EXPAND-NEXT: [[SHIFTED:%.*]] = lshr i32 [[TMP4]], [[SHIFTAMT]]
+; EXPAND-NEXT: [[EXTRACTED:%.*]] = trunc i32 [[SHIFTED]] to i16
+; EXPAND-NEXT: ret i16 [[EXTRACTED]]
+;
+ %1 = atomicrmw or ptr %a, i16 %b seq_cst
+ ret i16 %1
+
+}
+
+define i8 @atomicrmw_xor_i8_seq_cst(ptr %a, i8 %b) nounwind {
+; NO-EXPAND-LABEL: define i8 @atomicrmw_xor_i8_seq_cst(
+; NO-EXPAND-SAME: ptr [[A:%.*]], i8 [[B:%.*]]) #[[ATTR0]] {
+; NO-EXPAND-NEXT: [[TMP1:%.*]] = atomicrmw xor ptr [[A]], i8 [[B]] seq_cst, align 1
+; NO-EXPAND-NEXT: ret i8 [[TMP1]]
+;
+; EXPAND-LABEL: define i8 @atomicrmw_xor_i8_seq_cst(
+; EXPAND-SAME: ptr [[A:%.*]], i8 [[B:%.*]]) #[[ATTR0]] {
+; EXPAND-NEXT: [[ALIGNEDADDR:%.*]] = call ptr @llvm.ptrmask.p0.i64(ptr [[A]], i64 -4)
+; EXPAND-NEXT: [[TMP1:%.*]] = ptrtoint ptr [[A]] to i64
+; EXPAND-NEXT: [[PTRLSB:%.*]] = and i64 [[TMP1]], 3
+; EXPAND-NEXT: [[TMP2:%.*]] = shl i64 [[PTRLSB]], 3
+; EXPAND-NEXT: [[SHIFTAMT:%.*]] = trunc i64 [[TMP2]] to i32
+; EXPAND-NEXT: [[MASK:%.*]] = shl i32 255, [[SHIFTAMT]]
+; EXPAND-NEXT: [[INV_MASK:%.*]] = xor i32 [[MASK]], -1
+; EXPAND-NEXT: [[TMP3:%.*]] = zext i8 [[B]] to i32
+; EXPAND-NEXT: [[VALOPERAND_SHIFTED:%.*]] = shl i32 [[TMP3]], [[SHIFTAMT]]
+; EXPAND-NEXT: [[TMP4:%.*]] = atomicrmw xor ptr [[ALIGNEDADDR]], i32 [[VALOPERAND_SHIFTED]] seq_cst, align 4
+; EXPAND-NEXT: [[SHIFTED:%.*]] = lshr i32 [[TMP4]], [[SHIFTAMT]]
+; EXPAND-NEXT: [[EXTRACTED:%.*]] = trunc i32 [[SHIFTED]] to i8
+; EXPAND-NEXT: ret i8 [[EXTRACTED]]
+;
+ %1 = atomicrmw xor ptr %a, i8 %b seq_cst
+ ret i8 %1
+}
+
+define i16 @atomicrmw_xor_i16_seq_cst(ptr %a, i16 %b) nounwind {
+; NO-EXPAND-LABEL: define i16 @atomicrmw_xor_i16_seq_cst(
+; NO-EXPAND-SAME: ptr [[A:%.*]], i16 [[B:%.*]]) #[[ATTR0]] {
+; NO-EXPAND-NEXT: [[TMP1:%.*]] = atomicrmw xor ptr [[A]], i16 [[B]] seq_cst, align 2
+; NO-EXPAND-NEXT: ret i16 [[TMP1]]
+;
+; EXPAND-LABEL: define i16 @atomicrmw_xor_i16_seq_cst(
+; EXPAND-SAME: ptr [[A:%.*]], i16 [[B:%.*]]) #[[ATTR0]] {
+; EXPAND-NEXT: [[ALIGNEDADDR:%.*]] = call ptr @llvm.ptrmask.p0.i64(ptr [[A]], i64 -4)
+; EXPAND-NEXT: [[TMP1:%.*]] = ptrtoint ptr [[A]] to i64
+; EXPAND-NEXT: [[PTRLSB:%.*]] = and i64 [[TMP1]], 3
+; EXPAND-NEXT: [[TMP2:%.*]] = shl i64 [[PTRLSB]], 3
+; EXPAND-NEXT: [[SHIFTAMT:%.*]] = trunc i64 [[TMP2]] to i32
+; EXPAND-NEXT: [[MASK:%.*]] = shl i32 65535, [[SHIFTAMT]]
+; EXPAND-NEXT: [[INV_MASK:%.*]] = xor i32 [[MASK]], -1
+; EXPAND-NEXT: [[TMP3:%.*]] = zext i16 [[B]] to i32
+; EXPAND-NEXT: [[VALOPERAND_SHIFTED:%.*]] = shl i32 [[TMP3]], [[SHIFTAMT]]
+; EXPAND-NEXT: [[TMP4:%.*]] = atomicrmw xor ptr [[ALIGNEDADDR]], i32 [[VALOPERAND_SHIFTED]] seq_cst, align 4
+; EXPAND-NEXT: [[SHIFTED:%.*]] = lshr i32 [[TMP4]], [[SHIFTAMT]]
+; EXPAND-NEXT: [[EXTRACTED:%.*]] = trunc i32 [[SHIFTED]] to i16
+; EXPAND-NEXT: ret i16 [[EXTRACTED]]
+;
+ %1 = atomicrmw xor ptr %a, i16 %b seq_cst
+ ret i16 %1
+
+}
+
+define i8 @atomicrmw_and_i8_monotonic(ptr %a, i8 %b) nounwind {
+; NO-EXPAND-LABEL: define i8 @atomicrmw_and_i8_monotonic(
+; NO-EXPAND-SAME: ptr [[A:%.*]], i8 [[B:%.*]]) #[[ATTR0]] {
+; NO-EXPAND-NEXT: [[TMP1:%.*]] = atomicrmw and ptr [[A]], i8 [[B]] monotonic, align 1
+; NO-EXPAND-NEXT: ret i8 [[TMP1]]
+;
+; EXPAND-LABEL: define i8 @atomicrmw_and_i8_monotonic(
+; EXPAND-SAME: ptr [[A:%.*]], i8 [[B:%.*]]) #[[ATTR0]] {
+; EXPAND-NEXT: [[ALIGNEDADDR:%.*]] = call ptr @llvm.ptrmask.p0.i64(ptr [[A]], i64 -4)
+; EXPAND-NEXT: [[TMP1:%.*]] = ptrtoint ptr [[A]] to i64
+; EXPAND-NEXT: [[PTRLSB:%.*]] = and i64 [[TMP1]], 3
+; EXPAND-NEXT: [[TMP2:%.*]] = shl i64 [[PTRLSB]], 3
+; EXPAND-NEXT: [[SHIFTAMT:%.*]] = trunc i64 [[TMP2]] to i32
+; EXPAND-NEXT: [[MASK:%.*]] = shl i32 255, [[SHIFTAMT]]
+; EXPAND-NEXT: [[INV_MASK:%.*]] = xor i32 [[MASK]], -1
+; EXPAND-NEXT: [[TMP3:%.*]] = zext i8 [[B]] to i32
+; EXPAND-NEXT: [[VALOPERAND_SHIFTED:%.*]] = shl i32 [[TMP3]], [[SHIFTAMT]]
+; EXPAND-NEXT: [[ANDOPERAND:%.*]] = or i32 [[VALOPERAND_SHIFTED]], [[INV_MASK]]
+; EXPAND-NEXT: [[TMP4:%.*]] = atomicrmw and ptr [[ALIGNEDADDR]], i32 [[ANDOPERAND]] monotonic, align 4
+; EXPAND-NEXT: [[SHIFTED:%.*]] = lshr i32 [[TMP4]], [[SHIFTAMT]]
+; EXPAND-NEXT: [[EXTRACTED:%.*]] = trunc i32 [[SHIFTED]] to i8
+; EXPAND-NEXT: ret i8 [[EXTRACTED]]
+;
+ %1 = atomicrmw and ptr %a, i8 %b monotonic
+ ret i8 %1
+}
+
+define i16 @atomicrmw_and_i16_monotonic(ptr %a, i16 %b) nounwind {
+; NO-EXPAND-LABEL: define i16 @atomicrmw_and_i16_monotonic(
+; NO-EXPAND-SAME: ptr [[A:%.*]], i16 [[B:%.*]]) #[[ATTR0]] {
+; NO-EXPAND-NEXT: [[TMP1:%.*]] = atomicrmw and ptr [[A]], i16 [[B]] monotonic, align 2
+; NO-EXPAND-NEXT: ret i16 [[TMP1]]
+;
+; EXPAND-LABEL: define i16 @atomicrmw_and_i16_monotonic(
+; EXPAND-SAME: ptr [[A:%.*]], i16 [[B:%.*]]) #[[ATTR0]] {
+; EXPAND-NEXT: [[ALIGNEDADDR:%.*]] = call ptr @llvm.ptrmask.p0.i64(ptr [[A]], i64 -4)
+; EXPAND-NEXT: [[TMP1:%.*]] = ptrtoint ptr [[A]] to i64
+; EXPAND-NEXT: [[PTRLSB:%.*]] = and i64 [[TMP1]], 3
+; EXPAND-NEXT: [[TMP2:%.*]] = shl i64 [[PTRLSB]], 3
+; EXPAND-NEXT: [[SHIFTAMT:%.*]] = trunc i64 [[TMP2]] to i32
+; EXPAND-NEXT: [[MASK:%.*]] = shl i32 65535, [[SHIFTAMT]]
+; EXPAND-NEXT: [[INV_MASK:%.*]] = xor i32 [[MASK]], -1
+; EXPAND-NEXT: [[TMP3:%.*]] = zext i16 [[B]] to i32
+; EXPAND-NEXT: [[VALOPERAND_SHIFTED:%.*]] = shl i32 [[TMP3]], [[SHIFTAMT]]
+; EXPAND-NEXT: [[ANDOPERAND:%.*]] = or i32 [[VALOPERAND_SHIFTED]], [[INV_MASK]]
+; EXPAND-NEXT: [[TMP4:%.*]] = atomicrmw and ptr [[ALIGNEDADDR]], i32 [[ANDOPERAND]] monotonic, align 4
+; EXPAND-NEXT: [[SHIFTED:%.*]] = lshr i32 [[TMP4]], [[SHIFTAMT]]
+; EXPAND-NEXT: [[EXTRACTED:%.*]] = trunc i32 [[SHIFTED]] to i16
+; EXPAND-NEXT: ret i16 [[EXTRACTED]]
+;
+ %1 = atomicrmw and ptr %a, i16 %b monotonic
+ ret i16 %1
+
+}
+
+define i8 @atomicrmw_or_i8_monotonic(ptr %a, i8 %b) nounwind {
+; NO-EXPAND-LABEL: define i8 @atomicrmw_or_i8_monotonic(
+; NO-EXPAND-SAME: ptr [[A:%.*]], i8 [[B:%.*]]) #[[ATTR0]] {
+; NO-EXPAND-NEXT: [[TMP1:%.*]] = atomicrmw or ptr [[A]], i8 [[B]] monotonic, align 1
+; NO-EXPAND-NEXT: ret i8 [[TMP1]]
+;
+; EXPAND-LABEL: define i8 @atomicrmw_or_i8_monotonic(
+; EXPAND-SAME: ptr [[A:%.*]], i8 [[B:%.*]]) #[[ATTR0]] {
+; EXPAND-NEXT: [[ALIGNEDADDR:%.*]] = call ptr @llvm.ptrmask.p0.i64(ptr [[A]], i64 -4)
+; EXPAND-NEXT: [[TMP1:%.*]] = ptrtoint ptr [[A]] to i64
+; EXPAND-NEXT: [[PTRLSB:%.*]] = and i64 [[TMP1]], 3
+; EXPAND-NEXT: [[TMP2:%.*]] = shl i64 [[PTRLSB]], 3
+; EXPAND-NEXT: [[SHIFTAMT:%.*]] = trunc i64 [[TMP2]] to i32
+; EXPAND-NEXT: [[MASK:%.*]] = shl i32 255, [[SHIFTAMT]]
+; EXPAND-NEXT: [[INV_MASK:%.*]] = xor i32 [[MASK]], -1
+; EXPAND-NEXT: [[TMP3:%.*]] = zext i8 [[B]] to i32
+; EXPAND-NEXT: [[VALOPERAND_SHIFTED:%.*]] = shl i32 [[TMP3]], [[SHIFTAMT]]
+; EXPAND-NEXT: [[TMP4:%.*]] = atomicrmw or ptr [[ALIGNEDADDR]], i32 [[VALOPERAND_SHIFTED]] monotonic, align 4
+; EXPAND-NEXT: [[SHIFTED:%.*]] = lshr i32 [[TMP4]], [[SHIFTAMT]]
+; EXPAND-NEXT: [[EXTRACTED:%.*]] = trunc i32 [[SHIFTED]] to i8
+; EXPAND-NEXT: ret i8 [[EXTRACTED]]
+;
+ %1 = atomicrmw or ptr %a, i8 %b monotonic
+ ret i8 %1
+}
+
+define i16 @atomicrmw_or_i16_monotonic(ptr %a, i16 %b) nounwind {
+; NO-EXPAND-LABEL: define i16 @atomicrmw_or_i16_monotonic(
+; NO-EXPAND-SAME: ptr [[A:%.*]], i16 [[B:%.*]]) #[[ATTR0]] {
+; NO-EXPAND-NEXT: [[TMP1:%.*]] = atomicrmw or ptr [[A]], i16 [[B]] monotonic, align 2
+; NO-EXPAND-NEXT: ret i16 [[TMP1]]
+;
+; EXPAND-LABEL: define i16 @atomicrmw_or_i16_monotonic(
+; EXPAND-SAME: ptr [[A:%.*]], i16 [[B:%.*]]) #[[ATTR0]] {
+; EXPAND-NEXT: [[ALIGNEDADDR:%.*]] = call ptr @llvm.ptrmask.p0.i64(ptr [[A]], i64 -4)
+; EXPAND-NEXT: [[TMP1:%.*]] = ptrtoint ptr [[A]] to i64
+; EXPAND-NEXT: [[PTRLSB:%.*]] = and i64 [[TMP1]], 3
+; EXPAND-NEXT: [[TMP2:%.*]] = shl i64 [[PTRLSB]], 3
+; EXPAND-NEXT: [[SHIFTAMT:%.*]] = trunc i64 [[TMP2]] to i32
+; EXPAND-NEXT: [[MASK:%.*]] = shl i32 65535, [[SHIFTAMT]]
+; EXPAND-NEXT: [[INV_MASK:%.*]] = xor i32 [[MASK]], -1
+; EXPAND-NEXT: [[TMP3:%.*]] = zext i16 [[B]] to i32
+; EXPAND-NEXT: [[VALOPERAND_SHIFTED:%.*]] = shl i32 [[TMP3]], [[SHIFTAMT]]
+; EXPAND-NEXT: [[TMP4:%.*]] = atomicrmw or ptr [[ALIGNEDADDR]], i32 [[VALOPERAND_SHIFTED]] monotonic, align 4
+; EXPAND-NEXT: [[SHIFTED:%.*]] = lshr i32 [[TMP4]], [[SHIFTAMT]]
+; EXPAND-NEXT: [[EXTRACTED:%.*]] = trunc i32 [[SHIFTED]] to i16
+; EXPAND-NEXT: ret i16 [[EXTRACTED]]
+;
+ %1 = atomicrmw or ptr %a, i16 %b monotonic
+ ret i16 %1
+
+}
+
+define i8 @atomicrmw_xor_i8_monotonic(ptr %a, i8 %b) nounwind {
+; NO-EXPAND-LABEL: define i8 @atomicrmw_xor_i8_monotonic(
+; NO-EXPAND-SAME: ptr [[A:%.*]], i8 [[B:%.*]]) #[[ATTR0]] {
+; NO-EXPAND-NEXT: [[TMP1:%.*]] = atomicrmw xor ptr [[A]], i8 [[B]] monotonic, align 1
+; NO-EXPAND-NEXT: ret i8 [[TMP1]]
+;
+; EXPAND-LABEL: define i8 @atomicrmw_xor_i8_monotonic(
+; EXPAND-SAME: ptr [[A:%.*]], i8 [[B:%.*]]) #[[ATTR0]] {
+; EXPAND-NEXT: [[ALIGNEDADDR:%.*]] = call ptr @llvm.ptrmask.p0.i64(ptr [[A]], i64 -4)
+; EXPAND-NEXT: [[TMP1:%.*]] = ptrtoint ptr [[A]] to i64
+; EXPAND-NEXT: [[PTRLSB:%.*]] = and i64 [[TMP1]], 3
+; EXPAND-NEXT: [[TMP2:%.*]] = shl i64 [[PTRLSB]], 3
+; EXPAND-NEXT: [[SHIFTAMT:%.*]] = trunc i64 [[TMP2]] to i32
+; EXPAND-NEXT: [[MASK:%.*]] = shl i32 255, [[SHIFTAMT]]
+; EXPAND-NEXT: [[INV_MASK:%.*]] = xor i32 [[MASK]], -1
+; EXPAND-NEXT: [[TMP3:%.*]] = zext i8 [[B]] to i32
+; EXPAND-NEXT: [[VALOPERAND_SHIFTED:%.*]] = shl i32 [[TMP3]], [[SHIFTAMT]]
+; EXPAND-NEXT: [[TMP4:%.*]] = atomicrmw xor ptr [[ALIGNEDADDR]], i32 [[VALOPERAND_SHIFTED]] monotonic, align 4
+; EXPAND-NEXT: [[SHIFTED:%.*]] = lshr i32 [[TMP4]], [[SHIFTAMT]]
+; EXPAND-NEXT: [[EXTRACTED:%.*]] = trunc i32 [[SHIFTED]] to i8
+; EXPAND-NEXT: ret i8 [[EXTRACTED]]
+;
+ %1 = atomicrmw xor ptr %a, i8 %b monotonic
+ ret i8 %1
+}
+
+define i16 @atomicrmw_xor_i16_monotonic(ptr %a, i16 %b) nounwind {
+; NO-EXPAND-LABEL: define i16 @atomicrmw_xor_i16_monotonic(
+; NO-EXPAND-SAME: ptr [[A:%.*]], i16 [[B:%.*]]) #[[ATTR0]] {
+; NO-EXPAND-NEXT: [[TMP1:%.*]] = atomicrmw xor ptr [[A]], i16 [[B]] monotonic, align 2
+; NO-EXPAND-NEXT: ret i16 [[TMP1]]
+;
+; EXPAND-LABEL: define i16 @atomicrmw_xor_i16_monotonic(
+; EXPAND-SAME: ptr [[A:%.*]], i16 [[B:%.*]]) #[[ATTR0]] {
+; EXPAND-NEXT: [[ALIGNEDADDR:%.*]] = call ptr @llvm.ptrmask.p0.i64(ptr [[A]], i64 -4)
+; EXPAND-NEXT: [[TMP1:%.*]] = ptrtoint ptr [[A]] to i64
+; EXPAND-NEXT: [[PTRLSB:%.*]] = and i64 [[TMP1]], 3
+; EXPAND-NEXT: [[TMP2:%.*]] = shl i64 [[PTRLSB]], 3
+; EXPAND-NEXT: [[SHIFTAMT:%.*]] = trunc i64 [[TMP2]] to i32
+; EXPAND-NEXT: [[MASK:%.*]] = shl i32 65535, [[SHIFTAMT]]
+; EXPAND-NEXT: [[INV_MASK:%.*]] = xor i32 [[MASK]], -1
+; EXPAND-NEXT: [[TMP3:%.*]] = zext i16 [[B]] to i32
+; EXPAND-NEXT: [[VALOPERAND_SHIFTED:%.*]] = shl i32 [[TMP3]], [[SHIFTAMT]]
+; EXPAND-NEXT: [[TMP4:%.*]] = atomicrmw xor ptr [[ALIGNEDADDR]], i32 [[VALOPERAND_SHIFTED]] monotonic, align 4
+; EXPAND-NEXT: [[SHIFTED:%.*]] = lshr i32 [[TMP4]], [[SHIFTAMT]]
+; EXPAND-NEXT: [[EXTRACTED:%.*]] = trunc i32 [[SHIFTED]] to i16
+; EXPAND-NEXT: ret i16 [[EXTRACTED]]
+;
+ %1 = atomicrmw xor ptr %a, i16 %b monotonic
+ ret i16 %1
+
+}
+
More information about the cfe-commits
mailing list