[clang] [clang][LoongArch] Don't imply -fdirect-access-external-data for non-PIC (PR #71887)
xry111 wrote: Suggest to add "Fixes #71645" in the message. https://github.com/llvm/llvm-project/pull/71887 ___ cfe-commits mailing list cfe-commits@lists.llvm.org https://lists.llvm.org/cgi-bin/mailman/listinfo/cfe-commits
[clang] [clang][LoongArch] Don't imply -fdirect-access-external-data for non-PIC (PR #71887)
xry111 wrote: cc @MaskRay https://github.com/llvm/llvm-project/pull/71887 ___ cfe-commits mailing list cfe-commits@lists.llvm.org https://lists.llvm.org/cgi-bin/mailman/listinfo/cfe-commits
[clang] [clang] Add per-global code model attribute (PR #72078)
xry111 wrote:
> Do folks feel like the attribute name is sufficiently descriptive? i.e.
> should it be `__attribute__((code_model("asdf")))`? Are we aiming for GCC
> compat here? What guides the naming choice?
Yes, for GCC compat, so we don't need to add more #ifdef's in kernel.
As the author of the GCC attribute: the name choice was "borrowed" from IA64.
I have to admit that maybe I learnt from a bad example (anyway IA64 is in a
zombie state now since it's removed from the kernel), but for now I don't see
too much benefit to rename this.
https://github.com/llvm/llvm-project/pull/72078
___
cfe-commits mailing list
cfe-commits@lists.llvm.org
https://lists.llvm.org/cgi-bin/mailman/listinfo/cfe-commits
[clang] [clang] Add per-global code model attribute (PR #72078)
xry111 wrote: > is it too late to change the gcc attribute name? It has been released in GCC 13, and GCC 14 is in stage 3 so a change must be deferred into GCC 15. And the kernel code already relies on it. So a change will render all previous kernel releases impossible to build with GCC 13/14. https://github.com/llvm/llvm-project/pull/72078 ___ cfe-commits mailing list cfe-commits@lists.llvm.org https://lists.llvm.org/cgi-bin/mailman/listinfo/cfe-commits
[clang] [Driver] Support -mcmodel= for LoongArch (PR #72514)
xry111 wrote: > And AFAIK, gcc side doesn't plan to implement the "large" code model. Why did we distinguish "large" and "extreme" in the first place? If we don't need a different "large" code model then I guess we should make it an alias of "extreme" for GCC too. https://github.com/llvm/llvm-project/pull/72514 ___ cfe-commits mailing list cfe-commits@lists.llvm.org https://lists.llvm.org/cgi-bin/mailman/listinfo/cfe-commits
[clang] [LoongArch] Fix ABI mismatch with gcc/g++ about empty structs passing (PR #70320)
xry111 wrote:
LGTM. I've no permission to make a formal ("GitHub style") approval.
https://github.com/llvm/llvm-project/pull/70320
___
cfe-commits mailing list
cfe-commits@lists.llvm.org
https://lists.llvm.org/cgi-bin/mailman/listinfo/cfe-commits
[clang] [clang][LoongArch] Enable -fasynchronous-unwind-tables by default on Linux (PR #94711)
xry111 wrote: I'm really not a fan of enabling frame pointer by default. And enabling -fasynchronous-unwind-table is at least consistent with GCC. https://github.com/llvm/llvm-project/pull/94711 ___ cfe-commits mailing list cfe-commits@lists.llvm.org https://lists.llvm.org/cgi-bin/mailman/listinfo/cfe-commits
[clang] [LoongArch][Clang] Make the parameters and return value of {x,}vorn.v builti ns `unsigned char` vectors (PR #114514)
https://github.com/xry111 updated
https://github.com/llvm/llvm-project/pull/114514
>From bfd9e855bc1c91bee6c39e388f20be48db8dfd8f Mon Sep 17 00:00:00 2001
From: Xi Ruoyao
Date: Fri, 1 Nov 2024 00:28:12 +0800
Subject: [PATCH] [LoongArch][Clang] Make the parameters and return value of
{x,}vorn.v builtins `unsigned char` vectors
All other bitwise vector builtins use `unsigned char` vector operands,
so it does not make too much sense to use `signed char` for the orn
operation alone.
The same change has been also proposed for GCC:
https://gcc.gnu.org/pipermail/gcc-patches/2024-October/667065.html.
---
clang/include/clang/Basic/BuiltinsLoongArchLASX.def | 2 +-
clang/include/clang/Basic/BuiltinsLoongArchLSX.def | 2 +-
clang/lib/Headers/lasxintrin.h | 2 +-
clang/lib/Headers/lsxintrin.h | 2 +-
4 files changed, 4 insertions(+), 4 deletions(-)
diff --git a/clang/include/clang/Basic/BuiltinsLoongArchLASX.def
b/clang/include/clang/Basic/BuiltinsLoongArchLASX.def
index f644b820a61898..c4ea46a3bc5b51 100644
--- a/clang/include/clang/Basic/BuiltinsLoongArchLASX.def
+++ b/clang/include/clang/Basic/BuiltinsLoongArchLASX.def
@@ -371,7 +371,7 @@ TARGET_BUILTIN(__builtin_lasx_xvor_v, "V32UcV32UcV32Uc",
"nc", "lasx")
TARGET_BUILTIN(__builtin_lasx_xvxor_v, "V32UcV32UcV32Uc", "nc", "lasx")
TARGET_BUILTIN(__builtin_lasx_xvnor_v, "V32UcV32UcV32Uc", "nc", "lasx")
TARGET_BUILTIN(__builtin_lasx_xvandn_v, "V32UcV32UcV32Uc", "nc", "lasx")
-TARGET_BUILTIN(__builtin_lasx_xvorn_v, "V32ScV32ScV32Sc", "nc", "lasx")
+TARGET_BUILTIN(__builtin_lasx_xvorn_v, "V32UcV32UcV32Uc", "nc", "lasx")
TARGET_BUILTIN(__builtin_lasx_xvandi_b, "V32UcV32UcIUi", "nc", "lasx")
TARGET_BUILTIN(__builtin_lasx_xvori_b, "V32UcV32UcIUi", "nc", "lasx")
diff --git a/clang/include/clang/Basic/BuiltinsLoongArchLSX.def
b/clang/include/clang/Basic/BuiltinsLoongArchLSX.def
index b3056971986d19..a823783af88c4f 100644
--- a/clang/include/clang/Basic/BuiltinsLoongArchLSX.def
+++ b/clang/include/clang/Basic/BuiltinsLoongArchLSX.def
@@ -355,7 +355,7 @@ TARGET_BUILTIN(__builtin_lsx_vor_v, "V16UcV16UcV16Uc",
"nc", "lsx")
TARGET_BUILTIN(__builtin_lsx_vxor_v, "V16UcV16UcV16Uc", "nc", "lsx")
TARGET_BUILTIN(__builtin_lsx_vnor_v, "V16UcV16UcV16Uc", "nc", "lsx")
TARGET_BUILTIN(__builtin_lsx_vandn_v, "V16UcV16UcV16Uc", "nc", "lsx")
-TARGET_BUILTIN(__builtin_lsx_vorn_v, "V16ScV16ScV16Sc", "nc", "lsx")
+TARGET_BUILTIN(__builtin_lsx_vorn_v, "V16UcV16UcV16Uc", "nc", "lsx")
TARGET_BUILTIN(__builtin_lsx_vandi_b, "V16UcV16UcIUi", "nc", "lsx")
TARGET_BUILTIN(__builtin_lsx_vori_b, "V16UcV16UcIUi", "nc", "lsx")
diff --git a/clang/lib/Headers/lasxintrin.h b/clang/lib/Headers/lasxintrin.h
index dafc2a2f3e6a70..c065ea92a2dd52 100644
--- a/clang/lib/Headers/lasxintrin.h
+++ b/clang/lib/Headers/lasxintrin.h
@@ -2585,7 +2585,7 @@ extern __inline
extern __inline
__attribute__((__gnu_inline__, __always_inline__, __artificial__)) __m256i
__lasx_xvorn_v(__m256i _1, __m256i _2) {
- return (__m256i)__builtin_lasx_xvorn_v((v32i8)_1, (v32i8)_2);
+ return (__m256i)__builtin_lasx_xvorn_v((v32u8)_1, (v32u8)_2);
}
#define __lasx_xvldi(/*i13*/ _1) ((__m256i)__builtin_lasx_xvldi((_1)))
diff --git a/clang/lib/Headers/lsxintrin.h b/clang/lib/Headers/lsxintrin.h
index f347955ce6fb51..f020b0c18f0d29 100644
--- a/clang/lib/Headers/lsxintrin.h
+++ b/clang/lib/Headers/lsxintrin.h
@@ -3425,7 +3425,7 @@ extern __inline
extern __inline
__attribute__((__gnu_inline__, __always_inline__, __artificial__)) __m128i
__lsx_vorn_v(__m128i _1, __m128i _2) {
- return (__m128i)__builtin_lsx_vorn_v((v16i8)_1, (v16i8)_2);
+ return (__m128i)__builtin_lsx_vorn_v((v16u8)_1, (v16u8)_2);
}
#define __lsx_vldi(/*i13*/ _1) ((__m128i)__builtin_lsx_vldi((_1)))
___
cfe-commits mailing list
cfe-commits@lists.llvm.org
https://lists.llvm.org/cgi-bin/mailman/listinfo/cfe-commits
[clang] [LoongArch][Clang] Make the parameters and return value of {x,}vxor.v builti ns `unsigned char` vectors (PR #114513)
https://github.com/xry111 updated
https://github.com/llvm/llvm-project/pull/114513
>From 0d62818ddfa1d6b99afca4653a9f47478f301b32 Mon Sep 17 00:00:00 2001
From: Xi Ruoyao
Date: Fri, 1 Nov 2024 00:22:09 +0800
Subject: [PATCH] [LoongArch][Clang] Make the parameters and return value of
{x,}vxor.v builtins `unsigned char` vectors
The lsxintrin.h and and lasxintrin.h headers uses `unsigned char`
vectors instead of `signed char` vectors. GCC also uses `unsigned char`
for them, so align their definition with the headers and GCC.
Fixes #110834.
---
clang/include/clang/Basic/BuiltinsLoongArchLASX.def | 2 +-
clang/include/clang/Basic/BuiltinsLoongArchLSX.def | 2 +-
clang/test/Headers/lasxintrin.c | 5 ++---
clang/test/Headers/lsxintrin.c | 5 ++---
4 files changed, 6 insertions(+), 8 deletions(-)
diff --git a/clang/include/clang/Basic/BuiltinsLoongArchLASX.def
b/clang/include/clang/Basic/BuiltinsLoongArchLASX.def
index e6c41dd0974948..f644b820a61898 100644
--- a/clang/include/clang/Basic/BuiltinsLoongArchLASX.def
+++ b/clang/include/clang/Basic/BuiltinsLoongArchLASX.def
@@ -368,7 +368,7 @@ TARGET_BUILTIN(__builtin_lasx_xvrepli_d, "V4LLiIi", "nc",
"lasx")
TARGET_BUILTIN(__builtin_lasx_xvand_v, "V32UcV32UcV32Uc", "nc", "lasx")
TARGET_BUILTIN(__builtin_lasx_xvor_v, "V32UcV32UcV32Uc", "nc", "lasx")
-TARGET_BUILTIN(__builtin_lasx_xvxor_v, "V32ScV32ScV32Sc", "nc", "lasx")
+TARGET_BUILTIN(__builtin_lasx_xvxor_v, "V32UcV32UcV32Uc", "nc", "lasx")
TARGET_BUILTIN(__builtin_lasx_xvnor_v, "V32UcV32UcV32Uc", "nc", "lasx")
TARGET_BUILTIN(__builtin_lasx_xvandn_v, "V32UcV32UcV32Uc", "nc", "lasx")
TARGET_BUILTIN(__builtin_lasx_xvorn_v, "V32ScV32ScV32Sc", "nc", "lasx")
diff --git a/clang/include/clang/Basic/BuiltinsLoongArchLSX.def
b/clang/include/clang/Basic/BuiltinsLoongArchLSX.def
index ded6519f3ef16c..b3056971986d19 100644
--- a/clang/include/clang/Basic/BuiltinsLoongArchLSX.def
+++ b/clang/include/clang/Basic/BuiltinsLoongArchLSX.def
@@ -352,7 +352,7 @@ TARGET_BUILTIN(__builtin_lsx_vrepli_d, "V2LLiIi", "nc",
"lsx")
TARGET_BUILTIN(__builtin_lsx_vand_v, "V16UcV16UcV16Uc", "nc", "lsx")
TARGET_BUILTIN(__builtin_lsx_vor_v, "V16UcV16UcV16Uc", "nc", "lsx")
-TARGET_BUILTIN(__builtin_lsx_vxor_v, "V16ScV16ScV16Sc", "nc", "lsx")
+TARGET_BUILTIN(__builtin_lsx_vxor_v, "V16UcV16UcV16Uc", "nc", "lsx")
TARGET_BUILTIN(__builtin_lsx_vnor_v, "V16UcV16UcV16Uc", "nc", "lsx")
TARGET_BUILTIN(__builtin_lsx_vandn_v, "V16UcV16UcV16Uc", "nc", "lsx")
TARGET_BUILTIN(__builtin_lsx_vorn_v, "V16ScV16ScV16Sc", "nc", "lsx")
diff --git a/clang/test/Headers/lasxintrin.c b/clang/test/Headers/lasxintrin.c
index 6126c6350455c1..08f71791bdf36f 100644
--- a/clang/test/Headers/lasxintrin.c
+++ b/clang/test/Headers/lasxintrin.c
@@ -1,6 +1,5 @@
// RUN: %clang_cc1 %s -fsyntax-only -triple loongarch64 -target-feature +lasx
-// RUN: not %clang_cc1 %s -fsyntax-only -triple loongarch64 -target-feature
+lasx -flax-vector-conversions=none
-// RUN: not %clang_cc1 %s -fsyntax-only -triple loongarch64 -target-feature
+lasx -flax-vector-conversions=none -fno-signed-char
-// FIXME: "not" should be removed once we fix GH#110834.
+// RUN: %clang_cc1 %s -fsyntax-only -triple loongarch64 -target-feature +lasx
-flax-vector-conversions=none
+// RUN: %clang_cc1 %s -fsyntax-only -triple loongarch64 -target-feature +lasx
-flax-vector-conversions=none -fno-signed-char
#include
diff --git a/clang/test/Headers/lsxintrin.c b/clang/test/Headers/lsxintrin.c
index 930d3efe62e543..83c9879eea967d 100644
--- a/clang/test/Headers/lsxintrin.c
+++ b/clang/test/Headers/lsxintrin.c
@@ -1,6 +1,5 @@
// RUN: %clang_cc1 %s -fsyntax-only -triple loongarch64 -target-feature +lsx
-// RUN: not %clang_cc1 %s -fsyntax-only -triple loongarch64 -target-feature
+lsx -flax-vector-conversions=none
-// RUN: not %clang_cc1 %s -fsyntax-only -triple loongarch64 -target-feature
+lsx -flax-vector-conversions=none -fno-signed-char
-// FIXME: "not" should be removed once we fix GH#110834.
+// RUN: %clang_cc1 %s -fsyntax-only -triple loongarch64 -target-feature +lsx
-flax-vector-conversions=none
+// RUN: %clang_cc1 %s -fsyntax-only -triple loongarch64 -target-feature +lsx
-flax-vector-conversions=none -fno-signed-char
#include
___
cfe-commits mailing list
cfe-commits@lists.llvm.org
https://lists.llvm.org/cgi-bin/mailman/listinfo/cfe-commits
[clang] [LoongArch][Clang] Make the parameters and return value of {x,}vshuf.b builtins `signed char` vectors (PR #114512)
https://github.com/xry111 updated
https://github.com/llvm/llvm-project/pull/114512
>From ef96a6e483030e649d69ffa778c0b0f229bbb739 Mon Sep 17 00:00:00 2001
From: Xi Ruoyao
Date: Fri, 1 Nov 2024 00:13:29 +0800
Subject: [PATCH 1/2] [LoongArch][Clang] Make the parameter and return value of
{x,}vmsknz.b builtins `signed char` vector
These builtins operate on int8 vectors, not int16 vectors. So the old
definition does not make any sense.
---
clang/include/clang/Basic/BuiltinsLoongArchLASX.def | 2 +-
clang/include/clang/Basic/BuiltinsLoongArchLSX.def | 2 +-
2 files changed, 2 insertions(+), 2 deletions(-)
diff --git a/clang/include/clang/Basic/BuiltinsLoongArchLASX.def
b/clang/include/clang/Basic/BuiltinsLoongArchLASX.def
index 0d7c2df5c5c503..477f704860873c 100644
--- a/clang/include/clang/Basic/BuiltinsLoongArchLASX.def
+++ b/clang/include/clang/Basic/BuiltinsLoongArchLASX.def
@@ -357,7 +357,7 @@ TARGET_BUILTIN(__builtin_lasx_xvmskltz_w, "V8iV8i", "nc",
"lasx")
TARGET_BUILTIN(__builtin_lasx_xvmskltz_d, "V4LLiV4LLi", "nc", "lasx")
TARGET_BUILTIN(__builtin_lasx_xvmskgez_b, "V32ScV32Sc", "nc", "lasx")
-TARGET_BUILTIN(__builtin_lasx_xvmsknz_b, "V16sV16s", "nc", "lasx")
+TARGET_BUILTIN(__builtin_lasx_xvmsknz_b, "V32ScV32Sc", "nc", "lasx")
TARGET_BUILTIN(__builtin_lasx_xvldi, "V4LLiIi", "nc", "lasx")
TARGET_BUILTIN(__builtin_lasx_xvrepli_b, "V32ScIi", "nc", "lasx")
diff --git a/clang/include/clang/Basic/BuiltinsLoongArchLSX.def
b/clang/include/clang/Basic/BuiltinsLoongArchLSX.def
index 25a178e1ca98ab..e847985f390367 100644
--- a/clang/include/clang/Basic/BuiltinsLoongArchLSX.def
+++ b/clang/include/clang/Basic/BuiltinsLoongArchLSX.def
@@ -341,7 +341,7 @@ TARGET_BUILTIN(__builtin_lsx_vmskltz_w, "V4iV4i", "nc",
"lsx")
TARGET_BUILTIN(__builtin_lsx_vmskltz_d, "V2LLiV2LLi", "nc", "lsx")
TARGET_BUILTIN(__builtin_lsx_vmskgez_b, "V16ScV16Sc", "nc", "lsx")
-TARGET_BUILTIN(__builtin_lsx_vmsknz_b, "V8sV8s", "nc", "lsx")
+TARGET_BUILTIN(__builtin_lsx_vmsknz_b, "V16ScV16Sc", "nc", "lsx")
TARGET_BUILTIN(__builtin_lsx_vldi, "V2LLiIi", "nc", "lsx")
TARGET_BUILTIN(__builtin_lsx_vrepli_b, "V16ScIi", "nc", "lsx")
>From 349b79172a1ede63a25173366c1607b001bf1511 Mon Sep 17 00:00:00 2001
From: Xi Ruoyao
Date: Fri, 1 Nov 2024 00:17:44 +0800
Subject: [PATCH 2/2] [LoongArch][Clang] Make the parameters and return value
of {x,}vshuf.b builtins `signed char` vectors
The lsxintrin.h and and lasxintrin.h headers uses `signed char` vectors
instead of `unsigned char` vectors. GCC also uses `signed char` for
them, so align their definition with the headers and GCC.
---
clang/include/clang/Basic/BuiltinsLoongArchLASX.def | 2 +-
clang/include/clang/Basic/BuiltinsLoongArchLSX.def | 2 +-
2 files changed, 2 insertions(+), 2 deletions(-)
diff --git a/clang/include/clang/Basic/BuiltinsLoongArchLASX.def
b/clang/include/clang/Basic/BuiltinsLoongArchLASX.def
index 477f704860873c..e6c41dd0974948 100644
--- a/clang/include/clang/Basic/BuiltinsLoongArchLASX.def
+++ b/clang/include/clang/Basic/BuiltinsLoongArchLASX.def
@@ -935,7 +935,7 @@ TARGET_BUILTIN(__builtin_lasx_xvilvh_h, "V16sV16sV16s",
"nc", "lasx")
TARGET_BUILTIN(__builtin_lasx_xvilvh_w, "V8iV8iV8i", "nc", "lasx")
TARGET_BUILTIN(__builtin_lasx_xvilvh_d, "V4LLiV4LLiV4LLi", "nc", "lasx")
-TARGET_BUILTIN(__builtin_lasx_xvshuf_b, "V32UcV32UcV32UcV32Uc", "nc", "lasx")
+TARGET_BUILTIN(__builtin_lasx_xvshuf_b, "V32ScV32ScV32ScV32Sc", "nc", "lasx")
TARGET_BUILTIN(__builtin_lasx_xvshuf_h, "V16sV16sV16sV16s", "nc", "lasx")
TARGET_BUILTIN(__builtin_lasx_xvshuf_w, "V8iV8iV8iV8i", "nc", "lasx")
diff --git a/clang/include/clang/Basic/BuiltinsLoongArchLSX.def
b/clang/include/clang/Basic/BuiltinsLoongArchLSX.def
index e847985f390367..ded6519f3ef16c 100644
--- a/clang/include/clang/Basic/BuiltinsLoongArchLSX.def
+++ b/clang/include/clang/Basic/BuiltinsLoongArchLSX.def
@@ -910,7 +910,7 @@ TARGET_BUILTIN(__builtin_lsx_vilvh_h, "V8sV8sV8s", "nc",
"lsx")
TARGET_BUILTIN(__builtin_lsx_vilvh_w, "V4iV4iV4i", "nc", "lsx")
TARGET_BUILTIN(__builtin_lsx_vilvh_d, "V2LLiV2LLiV2LLi", "nc", "lsx")
-TARGET_BUILTIN(__builtin_lsx_vshuf_b, "V16UcV16UcV16UcV16Uc", "nc", "lsx")
+TARGET_BUILTIN(__builtin_lsx_vshuf_b, "V16ScV16ScV16ScV16Sc", "nc", "lsx")
TARGET_BUILTIN(__builtin_lsx_vshuf_h, "V8sV8sV8sV8s", "nc", "lsx")
TARGET_BUILTIN(__builtin_lsx_vshuf_w, "V4iV4iV4iV4i", "nc", "lsx")
___
cfe-commits mailing list
cfe-commits@lists.llvm.org
https://lists.llvm.org/cgi-bin/mailman/listinfo/cfe-commits
[clang] [LoongArch][Clang] Make the parameter and return value of {x,}vmsknz.b builtins `signed char` vector (PR #114511)
xry111 wrote: > Is it the time to remove the FIXME in those tests? There are several random inconsistencies and I separated the fixes into multiple PRs as suggested by Xuerui. The FIXME is removed in #114513. https://github.com/llvm/llvm-project/pull/114511 ___ cfe-commits mailing list cfe-commits@lists.llvm.org https://lists.llvm.org/cgi-bin/mailman/listinfo/cfe-commits
[clang] [LoongArch][Clang] Make the parameters and return value of {x,}vorn.v builti ns `unsigned char` vectors (PR #114514)
https://github.com/xry111 updated https://github.com/llvm/llvm-project/pull/114514 >From 3d65a59473580f1c0e0b57bdd7fb13c9f40d3077 Mon Sep 17 00:00:00 2001 From: Xi Ruoyao Date: Fri, 1 Nov 2024 00:07:17 +0800 Subject: [PATCH 1/5] [LoongArch][clang] Use `signed char` vectors instead of `char` vectors for LSX and LASX builtins `-flax-vector-conversions=none` does not allow an implicit conversion from `signed char` vector to `char` vector, and we cannot remove `signed` from `v16i8` or `v32i8` because doing so will break our expectation with `-fno-signed-char`. So to make lsxintrin.h and lasxintrin.h work fine with `-flax-vector-conversions=none`, we must use `signed char` instead of `char`. The change is just done via sed 's/V16c/V16Sc/g' -i BuiltinsLoongArchLSX.def sed 's/V32c/V32Sc/g' -i BuiltinsLoongArchLASX.def --- .../clang/Basic/BuiltinsLoongArchLASX.def | 142 +- .../clang/Basic/BuiltinsLoongArchLSX.def | 128 2 files changed, 135 insertions(+), 135 deletions(-) diff --git a/clang/include/clang/Basic/BuiltinsLoongArchLASX.def b/clang/include/clang/Basic/BuiltinsLoongArchLASX.def index 4cf51cc000f6fc..0d7c2df5c5c503 100644 --- a/clang/include/clang/Basic/BuiltinsLoongArchLASX.def +++ b/clang/include/clang/Basic/BuiltinsLoongArchLASX.def @@ -12,29 +12,29 @@ // //===--===// -TARGET_BUILTIN(__builtin_lasx_xvadd_b, "V32cV32cV32c", "nc", "lasx") +TARGET_BUILTIN(__builtin_lasx_xvadd_b, "V32ScV32ScV32Sc", "nc", "lasx") TARGET_BUILTIN(__builtin_lasx_xvadd_h, "V16sV16sV16s", "nc", "lasx") TARGET_BUILTIN(__builtin_lasx_xvadd_w, "V8iV8iV8i", "nc", "lasx") TARGET_BUILTIN(__builtin_lasx_xvadd_d, "V4LLiV4LLiV4LLi", "nc", "lasx") TARGET_BUILTIN(__builtin_lasx_xvadd_q, "V4LLiV4LLiV4LLi", "nc", "lasx") -TARGET_BUILTIN(__builtin_lasx_xvsub_b, "V32cV32cV32c", "nc", "lasx") +TARGET_BUILTIN(__builtin_lasx_xvsub_b, "V32ScV32ScV32Sc", "nc", "lasx") TARGET_BUILTIN(__builtin_lasx_xvsub_h, "V16sV16sV16s", "nc", "lasx") TARGET_BUILTIN(__builtin_lasx_xvsub_w, "V8iV8iV8i", "nc", "lasx") TARGET_BUILTIN(__builtin_lasx_xvsub_d, "V4LLiV4LLiV4LLi", "nc", "lasx") TARGET_BUILTIN(__builtin_lasx_xvsub_q, "V4LLiV4LLiV4LLi", "nc", "lasx") -TARGET_BUILTIN(__builtin_lasx_xvaddi_bu, "V32cV32cIUi", "nc", "lasx") +TARGET_BUILTIN(__builtin_lasx_xvaddi_bu, "V32ScV32ScIUi", "nc", "lasx") TARGET_BUILTIN(__builtin_lasx_xvaddi_hu, "V16sV16sIUi", "nc", "lasx") TARGET_BUILTIN(__builtin_lasx_xvaddi_wu, "V8iV8iIUi", "nc", "lasx") TARGET_BUILTIN(__builtin_lasx_xvaddi_du, "V4LLiV4LLiIUi", "nc", "lasx") -TARGET_BUILTIN(__builtin_lasx_xvsubi_bu, "V32cV32cIUi", "nc", "lasx") +TARGET_BUILTIN(__builtin_lasx_xvsubi_bu, "V32ScV32ScIUi", "nc", "lasx") TARGET_BUILTIN(__builtin_lasx_xvsubi_hu, "V16sV16sIUi", "nc", "lasx") TARGET_BUILTIN(__builtin_lasx_xvsubi_wu, "V8iV8iIUi", "nc", "lasx") TARGET_BUILTIN(__builtin_lasx_xvsubi_du, "V4LLiV4LLiIUi", "nc", "lasx") -TARGET_BUILTIN(__builtin_lasx_xvneg_b, "V32cV32c", "nc", "lasx") +TARGET_BUILTIN(__builtin_lasx_xvneg_b, "V32ScV32Sc", "nc", "lasx") TARGET_BUILTIN(__builtin_lasx_xvneg_h, "V16sV16s", "nc", "lasx") TARGET_BUILTIN(__builtin_lasx_xvneg_w, "V8iV8i", "nc", "lasx") TARGET_BUILTIN(__builtin_lasx_xvneg_d, "V4LLiV4LLi", "nc", "lasx") @@ -79,22 +79,22 @@ TARGET_BUILTIN(__builtin_lasx_xvhsubw_wu_hu, "V8UiV16UsV16Us", "nc", "lasx") TARGET_BUILTIN(__builtin_lasx_xvhsubw_du_wu, "V4ULLiV8UiV8Ui", "nc", "lasx") TARGET_BUILTIN(__builtin_lasx_xvhsubw_qu_du, "V4ULLiV4ULLiV4ULLi", "nc", "lasx") -TARGET_BUILTIN(__builtin_lasx_xvaddwev_h_b, "V16sV32cV32c", "nc", "lasx") +TARGET_BUILTIN(__builtin_lasx_xvaddwev_h_b, "V16sV32ScV32Sc", "nc", "lasx") TARGET_BUILTIN(__builtin_lasx_xvaddwev_w_h, "V8SiV16sV16s", "nc", "lasx") TARGET_BUILTIN(__builtin_lasx_xvaddwev_d_w, "V4LLiV8SiV8Si", "nc", "lasx") TARGET_BUILTIN(__builtin_lasx_xvaddwev_q_d, "V4LLiV4LLiV4LLi", "nc", "lasx") -TARGET_BUILTIN(__builtin_lasx_xvaddwod_h_b, "V16sV32cV32c", "nc", "lasx") +TARGET_BUILTIN(__builtin_lasx_xvaddwod_h_b, "V16sV32ScV32Sc", "nc", "lasx") TARGET_BUILTIN(__builtin_lasx_xvaddwod_w_h, "V8SiV16sV16s", "nc", "lasx") TARGET_BUILTIN(__builtin_lasx_xvaddwod_d_w, "V4LLiV8SiV8Si", "nc", "lasx") TARGET_BUILTIN(__builtin_lasx_xvaddwod_q_d, "V4LLiV4LLiV4LLi", "nc", "lasx") -TARGET_BUILTIN(__builtin_lasx_xvsubwev_h_b, "V16sV32cV32c", "nc", "lasx") +TARGET_BUILTIN(__builtin_lasx_xvsubwev_h_b, "V16sV32ScV32Sc", "nc", "lasx") TARGET_BUILTIN(__builtin_lasx_xvsubwev_w_h, "V8SiV16sV16s", "nc", "lasx") TARGET_BUILTIN(__builtin_lasx_xvsubwev_d_w, "V4LLiV8SiV8Si", "nc", "lasx") TARGET_BUILTIN(__builtin_lasx_xvsubwev_q_d, "V4LLiV4LLiV4LLi", "nc", "lasx") -TARGET_BUILTIN(__builtin_lasx_xvsubwod_h_b, "V16sV32cV32c", "nc", "lasx") +TARGET_BUILTIN(__builtin_lasx_xvsubwod_h_b, "V16sV32ScV32Sc", "nc", "lasx") TARGET_BUILTIN(__builtin_lasx_xvsubwod_w_h, "V8SiV16sV16s", "nc", "lasx") TARGET_BUILTIN(__builtin_
[clang] [LoongArch][Clang] Make the parameters and return value of {x,}vshuf.b builtins `signed char` vectors (PR #114512)
https://github.com/xry111 updated https://github.com/llvm/llvm-project/pull/114512 >From 3d65a59473580f1c0e0b57bdd7fb13c9f40d3077 Mon Sep 17 00:00:00 2001 From: Xi Ruoyao Date: Fri, 1 Nov 2024 00:07:17 +0800 Subject: [PATCH 1/3] [LoongArch][clang] Use `signed char` vectors instead of `char` vectors for LSX and LASX builtins `-flax-vector-conversions=none` does not allow an implicit conversion from `signed char` vector to `char` vector, and we cannot remove `signed` from `v16i8` or `v32i8` because doing so will break our expectation with `-fno-signed-char`. So to make lsxintrin.h and lasxintrin.h work fine with `-flax-vector-conversions=none`, we must use `signed char` instead of `char`. The change is just done via sed 's/V16c/V16Sc/g' -i BuiltinsLoongArchLSX.def sed 's/V32c/V32Sc/g' -i BuiltinsLoongArchLASX.def --- .../clang/Basic/BuiltinsLoongArchLASX.def | 142 +- .../clang/Basic/BuiltinsLoongArchLSX.def | 128 2 files changed, 135 insertions(+), 135 deletions(-) diff --git a/clang/include/clang/Basic/BuiltinsLoongArchLASX.def b/clang/include/clang/Basic/BuiltinsLoongArchLASX.def index 4cf51cc000f6fc..0d7c2df5c5c503 100644 --- a/clang/include/clang/Basic/BuiltinsLoongArchLASX.def +++ b/clang/include/clang/Basic/BuiltinsLoongArchLASX.def @@ -12,29 +12,29 @@ // //===--===// -TARGET_BUILTIN(__builtin_lasx_xvadd_b, "V32cV32cV32c", "nc", "lasx") +TARGET_BUILTIN(__builtin_lasx_xvadd_b, "V32ScV32ScV32Sc", "nc", "lasx") TARGET_BUILTIN(__builtin_lasx_xvadd_h, "V16sV16sV16s", "nc", "lasx") TARGET_BUILTIN(__builtin_lasx_xvadd_w, "V8iV8iV8i", "nc", "lasx") TARGET_BUILTIN(__builtin_lasx_xvadd_d, "V4LLiV4LLiV4LLi", "nc", "lasx") TARGET_BUILTIN(__builtin_lasx_xvadd_q, "V4LLiV4LLiV4LLi", "nc", "lasx") -TARGET_BUILTIN(__builtin_lasx_xvsub_b, "V32cV32cV32c", "nc", "lasx") +TARGET_BUILTIN(__builtin_lasx_xvsub_b, "V32ScV32ScV32Sc", "nc", "lasx") TARGET_BUILTIN(__builtin_lasx_xvsub_h, "V16sV16sV16s", "nc", "lasx") TARGET_BUILTIN(__builtin_lasx_xvsub_w, "V8iV8iV8i", "nc", "lasx") TARGET_BUILTIN(__builtin_lasx_xvsub_d, "V4LLiV4LLiV4LLi", "nc", "lasx") TARGET_BUILTIN(__builtin_lasx_xvsub_q, "V4LLiV4LLiV4LLi", "nc", "lasx") -TARGET_BUILTIN(__builtin_lasx_xvaddi_bu, "V32cV32cIUi", "nc", "lasx") +TARGET_BUILTIN(__builtin_lasx_xvaddi_bu, "V32ScV32ScIUi", "nc", "lasx") TARGET_BUILTIN(__builtin_lasx_xvaddi_hu, "V16sV16sIUi", "nc", "lasx") TARGET_BUILTIN(__builtin_lasx_xvaddi_wu, "V8iV8iIUi", "nc", "lasx") TARGET_BUILTIN(__builtin_lasx_xvaddi_du, "V4LLiV4LLiIUi", "nc", "lasx") -TARGET_BUILTIN(__builtin_lasx_xvsubi_bu, "V32cV32cIUi", "nc", "lasx") +TARGET_BUILTIN(__builtin_lasx_xvsubi_bu, "V32ScV32ScIUi", "nc", "lasx") TARGET_BUILTIN(__builtin_lasx_xvsubi_hu, "V16sV16sIUi", "nc", "lasx") TARGET_BUILTIN(__builtin_lasx_xvsubi_wu, "V8iV8iIUi", "nc", "lasx") TARGET_BUILTIN(__builtin_lasx_xvsubi_du, "V4LLiV4LLiIUi", "nc", "lasx") -TARGET_BUILTIN(__builtin_lasx_xvneg_b, "V32cV32c", "nc", "lasx") +TARGET_BUILTIN(__builtin_lasx_xvneg_b, "V32ScV32Sc", "nc", "lasx") TARGET_BUILTIN(__builtin_lasx_xvneg_h, "V16sV16s", "nc", "lasx") TARGET_BUILTIN(__builtin_lasx_xvneg_w, "V8iV8i", "nc", "lasx") TARGET_BUILTIN(__builtin_lasx_xvneg_d, "V4LLiV4LLi", "nc", "lasx") @@ -79,22 +79,22 @@ TARGET_BUILTIN(__builtin_lasx_xvhsubw_wu_hu, "V8UiV16UsV16Us", "nc", "lasx") TARGET_BUILTIN(__builtin_lasx_xvhsubw_du_wu, "V4ULLiV8UiV8Ui", "nc", "lasx") TARGET_BUILTIN(__builtin_lasx_xvhsubw_qu_du, "V4ULLiV4ULLiV4ULLi", "nc", "lasx") -TARGET_BUILTIN(__builtin_lasx_xvaddwev_h_b, "V16sV32cV32c", "nc", "lasx") +TARGET_BUILTIN(__builtin_lasx_xvaddwev_h_b, "V16sV32ScV32Sc", "nc", "lasx") TARGET_BUILTIN(__builtin_lasx_xvaddwev_w_h, "V8SiV16sV16s", "nc", "lasx") TARGET_BUILTIN(__builtin_lasx_xvaddwev_d_w, "V4LLiV8SiV8Si", "nc", "lasx") TARGET_BUILTIN(__builtin_lasx_xvaddwev_q_d, "V4LLiV4LLiV4LLi", "nc", "lasx") -TARGET_BUILTIN(__builtin_lasx_xvaddwod_h_b, "V16sV32cV32c", "nc", "lasx") +TARGET_BUILTIN(__builtin_lasx_xvaddwod_h_b, "V16sV32ScV32Sc", "nc", "lasx") TARGET_BUILTIN(__builtin_lasx_xvaddwod_w_h, "V8SiV16sV16s", "nc", "lasx") TARGET_BUILTIN(__builtin_lasx_xvaddwod_d_w, "V4LLiV8SiV8Si", "nc", "lasx") TARGET_BUILTIN(__builtin_lasx_xvaddwod_q_d, "V4LLiV4LLiV4LLi", "nc", "lasx") -TARGET_BUILTIN(__builtin_lasx_xvsubwev_h_b, "V16sV32cV32c", "nc", "lasx") +TARGET_BUILTIN(__builtin_lasx_xvsubwev_h_b, "V16sV32ScV32Sc", "nc", "lasx") TARGET_BUILTIN(__builtin_lasx_xvsubwev_w_h, "V8SiV16sV16s", "nc", "lasx") TARGET_BUILTIN(__builtin_lasx_xvsubwev_d_w, "V4LLiV8SiV8Si", "nc", "lasx") TARGET_BUILTIN(__builtin_lasx_xvsubwev_q_d, "V4LLiV4LLiV4LLi", "nc", "lasx") -TARGET_BUILTIN(__builtin_lasx_xvsubwod_h_b, "V16sV32cV32c", "nc", "lasx") +TARGET_BUILTIN(__builtin_lasx_xvsubwod_h_b, "V16sV32ScV32Sc", "nc", "lasx") TARGET_BUILTIN(__builtin_lasx_xvsubwod_w_h, "V8SiV16sV16s", "nc", "lasx") TARGET_BUILTIN(__builtin_
[clang] [LoongArch][Clang] Make the parameter and return value of {x,}vmsknz.b builtins `signed char` vector (PR #114511)
https://github.com/xry111 updated https://github.com/llvm/llvm-project/pull/114511 >From 3d65a59473580f1c0e0b57bdd7fb13c9f40d3077 Mon Sep 17 00:00:00 2001 From: Xi Ruoyao Date: Fri, 1 Nov 2024 00:07:17 +0800 Subject: [PATCH 1/2] [LoongArch][clang] Use `signed char` vectors instead of `char` vectors for LSX and LASX builtins `-flax-vector-conversions=none` does not allow an implicit conversion from `signed char` vector to `char` vector, and we cannot remove `signed` from `v16i8` or `v32i8` because doing so will break our expectation with `-fno-signed-char`. So to make lsxintrin.h and lasxintrin.h work fine with `-flax-vector-conversions=none`, we must use `signed char` instead of `char`. The change is just done via sed 's/V16c/V16Sc/g' -i BuiltinsLoongArchLSX.def sed 's/V32c/V32Sc/g' -i BuiltinsLoongArchLASX.def --- .../clang/Basic/BuiltinsLoongArchLASX.def | 142 +- .../clang/Basic/BuiltinsLoongArchLSX.def | 128 2 files changed, 135 insertions(+), 135 deletions(-) diff --git a/clang/include/clang/Basic/BuiltinsLoongArchLASX.def b/clang/include/clang/Basic/BuiltinsLoongArchLASX.def index 4cf51cc000f6fc..0d7c2df5c5c503 100644 --- a/clang/include/clang/Basic/BuiltinsLoongArchLASX.def +++ b/clang/include/clang/Basic/BuiltinsLoongArchLASX.def @@ -12,29 +12,29 @@ // //===--===// -TARGET_BUILTIN(__builtin_lasx_xvadd_b, "V32cV32cV32c", "nc", "lasx") +TARGET_BUILTIN(__builtin_lasx_xvadd_b, "V32ScV32ScV32Sc", "nc", "lasx") TARGET_BUILTIN(__builtin_lasx_xvadd_h, "V16sV16sV16s", "nc", "lasx") TARGET_BUILTIN(__builtin_lasx_xvadd_w, "V8iV8iV8i", "nc", "lasx") TARGET_BUILTIN(__builtin_lasx_xvadd_d, "V4LLiV4LLiV4LLi", "nc", "lasx") TARGET_BUILTIN(__builtin_lasx_xvadd_q, "V4LLiV4LLiV4LLi", "nc", "lasx") -TARGET_BUILTIN(__builtin_lasx_xvsub_b, "V32cV32cV32c", "nc", "lasx") +TARGET_BUILTIN(__builtin_lasx_xvsub_b, "V32ScV32ScV32Sc", "nc", "lasx") TARGET_BUILTIN(__builtin_lasx_xvsub_h, "V16sV16sV16s", "nc", "lasx") TARGET_BUILTIN(__builtin_lasx_xvsub_w, "V8iV8iV8i", "nc", "lasx") TARGET_BUILTIN(__builtin_lasx_xvsub_d, "V4LLiV4LLiV4LLi", "nc", "lasx") TARGET_BUILTIN(__builtin_lasx_xvsub_q, "V4LLiV4LLiV4LLi", "nc", "lasx") -TARGET_BUILTIN(__builtin_lasx_xvaddi_bu, "V32cV32cIUi", "nc", "lasx") +TARGET_BUILTIN(__builtin_lasx_xvaddi_bu, "V32ScV32ScIUi", "nc", "lasx") TARGET_BUILTIN(__builtin_lasx_xvaddi_hu, "V16sV16sIUi", "nc", "lasx") TARGET_BUILTIN(__builtin_lasx_xvaddi_wu, "V8iV8iIUi", "nc", "lasx") TARGET_BUILTIN(__builtin_lasx_xvaddi_du, "V4LLiV4LLiIUi", "nc", "lasx") -TARGET_BUILTIN(__builtin_lasx_xvsubi_bu, "V32cV32cIUi", "nc", "lasx") +TARGET_BUILTIN(__builtin_lasx_xvsubi_bu, "V32ScV32ScIUi", "nc", "lasx") TARGET_BUILTIN(__builtin_lasx_xvsubi_hu, "V16sV16sIUi", "nc", "lasx") TARGET_BUILTIN(__builtin_lasx_xvsubi_wu, "V8iV8iIUi", "nc", "lasx") TARGET_BUILTIN(__builtin_lasx_xvsubi_du, "V4LLiV4LLiIUi", "nc", "lasx") -TARGET_BUILTIN(__builtin_lasx_xvneg_b, "V32cV32c", "nc", "lasx") +TARGET_BUILTIN(__builtin_lasx_xvneg_b, "V32ScV32Sc", "nc", "lasx") TARGET_BUILTIN(__builtin_lasx_xvneg_h, "V16sV16s", "nc", "lasx") TARGET_BUILTIN(__builtin_lasx_xvneg_w, "V8iV8i", "nc", "lasx") TARGET_BUILTIN(__builtin_lasx_xvneg_d, "V4LLiV4LLi", "nc", "lasx") @@ -79,22 +79,22 @@ TARGET_BUILTIN(__builtin_lasx_xvhsubw_wu_hu, "V8UiV16UsV16Us", "nc", "lasx") TARGET_BUILTIN(__builtin_lasx_xvhsubw_du_wu, "V4ULLiV8UiV8Ui", "nc", "lasx") TARGET_BUILTIN(__builtin_lasx_xvhsubw_qu_du, "V4ULLiV4ULLiV4ULLi", "nc", "lasx") -TARGET_BUILTIN(__builtin_lasx_xvaddwev_h_b, "V16sV32cV32c", "nc", "lasx") +TARGET_BUILTIN(__builtin_lasx_xvaddwev_h_b, "V16sV32ScV32Sc", "nc", "lasx") TARGET_BUILTIN(__builtin_lasx_xvaddwev_w_h, "V8SiV16sV16s", "nc", "lasx") TARGET_BUILTIN(__builtin_lasx_xvaddwev_d_w, "V4LLiV8SiV8Si", "nc", "lasx") TARGET_BUILTIN(__builtin_lasx_xvaddwev_q_d, "V4LLiV4LLiV4LLi", "nc", "lasx") -TARGET_BUILTIN(__builtin_lasx_xvaddwod_h_b, "V16sV32cV32c", "nc", "lasx") +TARGET_BUILTIN(__builtin_lasx_xvaddwod_h_b, "V16sV32ScV32Sc", "nc", "lasx") TARGET_BUILTIN(__builtin_lasx_xvaddwod_w_h, "V8SiV16sV16s", "nc", "lasx") TARGET_BUILTIN(__builtin_lasx_xvaddwod_d_w, "V4LLiV8SiV8Si", "nc", "lasx") TARGET_BUILTIN(__builtin_lasx_xvaddwod_q_d, "V4LLiV4LLiV4LLi", "nc", "lasx") -TARGET_BUILTIN(__builtin_lasx_xvsubwev_h_b, "V16sV32cV32c", "nc", "lasx") +TARGET_BUILTIN(__builtin_lasx_xvsubwev_h_b, "V16sV32ScV32Sc", "nc", "lasx") TARGET_BUILTIN(__builtin_lasx_xvsubwev_w_h, "V8SiV16sV16s", "nc", "lasx") TARGET_BUILTIN(__builtin_lasx_xvsubwev_d_w, "V4LLiV8SiV8Si", "nc", "lasx") TARGET_BUILTIN(__builtin_lasx_xvsubwev_q_d, "V4LLiV4LLiV4LLi", "nc", "lasx") -TARGET_BUILTIN(__builtin_lasx_xvsubwod_h_b, "V16sV32cV32c", "nc", "lasx") +TARGET_BUILTIN(__builtin_lasx_xvsubwod_h_b, "V16sV32ScV32Sc", "nc", "lasx") TARGET_BUILTIN(__builtin_lasx_xvsubwod_w_h, "V8SiV16sV16s", "nc", "lasx") TARGET_BUILTIN(__builtin_
[clang] [LoongArch][clang] Use `signed char` vectors instead of `char` vectors for LSX and LASX builtins (PR #114510)
https://github.com/xry111 updated https://github.com/llvm/llvm-project/pull/114510 >From 3d65a59473580f1c0e0b57bdd7fb13c9f40d3077 Mon Sep 17 00:00:00 2001 From: Xi Ruoyao Date: Fri, 1 Nov 2024 00:07:17 +0800 Subject: [PATCH] [LoongArch][clang] Use `signed char` vectors instead of `char` vectors for LSX and LASX builtins `-flax-vector-conversions=none` does not allow an implicit conversion from `signed char` vector to `char` vector, and we cannot remove `signed` from `v16i8` or `v32i8` because doing so will break our expectation with `-fno-signed-char`. So to make lsxintrin.h and lasxintrin.h work fine with `-flax-vector-conversions=none`, we must use `signed char` instead of `char`. The change is just done via sed 's/V16c/V16Sc/g' -i BuiltinsLoongArchLSX.def sed 's/V32c/V32Sc/g' -i BuiltinsLoongArchLASX.def --- .../clang/Basic/BuiltinsLoongArchLASX.def | 142 +- .../clang/Basic/BuiltinsLoongArchLSX.def | 128 2 files changed, 135 insertions(+), 135 deletions(-) diff --git a/clang/include/clang/Basic/BuiltinsLoongArchLASX.def b/clang/include/clang/Basic/BuiltinsLoongArchLASX.def index 4cf51cc000f6fc..0d7c2df5c5c503 100644 --- a/clang/include/clang/Basic/BuiltinsLoongArchLASX.def +++ b/clang/include/clang/Basic/BuiltinsLoongArchLASX.def @@ -12,29 +12,29 @@ // //===--===// -TARGET_BUILTIN(__builtin_lasx_xvadd_b, "V32cV32cV32c", "nc", "lasx") +TARGET_BUILTIN(__builtin_lasx_xvadd_b, "V32ScV32ScV32Sc", "nc", "lasx") TARGET_BUILTIN(__builtin_lasx_xvadd_h, "V16sV16sV16s", "nc", "lasx") TARGET_BUILTIN(__builtin_lasx_xvadd_w, "V8iV8iV8i", "nc", "lasx") TARGET_BUILTIN(__builtin_lasx_xvadd_d, "V4LLiV4LLiV4LLi", "nc", "lasx") TARGET_BUILTIN(__builtin_lasx_xvadd_q, "V4LLiV4LLiV4LLi", "nc", "lasx") -TARGET_BUILTIN(__builtin_lasx_xvsub_b, "V32cV32cV32c", "nc", "lasx") +TARGET_BUILTIN(__builtin_lasx_xvsub_b, "V32ScV32ScV32Sc", "nc", "lasx") TARGET_BUILTIN(__builtin_lasx_xvsub_h, "V16sV16sV16s", "nc", "lasx") TARGET_BUILTIN(__builtin_lasx_xvsub_w, "V8iV8iV8i", "nc", "lasx") TARGET_BUILTIN(__builtin_lasx_xvsub_d, "V4LLiV4LLiV4LLi", "nc", "lasx") TARGET_BUILTIN(__builtin_lasx_xvsub_q, "V4LLiV4LLiV4LLi", "nc", "lasx") -TARGET_BUILTIN(__builtin_lasx_xvaddi_bu, "V32cV32cIUi", "nc", "lasx") +TARGET_BUILTIN(__builtin_lasx_xvaddi_bu, "V32ScV32ScIUi", "nc", "lasx") TARGET_BUILTIN(__builtin_lasx_xvaddi_hu, "V16sV16sIUi", "nc", "lasx") TARGET_BUILTIN(__builtin_lasx_xvaddi_wu, "V8iV8iIUi", "nc", "lasx") TARGET_BUILTIN(__builtin_lasx_xvaddi_du, "V4LLiV4LLiIUi", "nc", "lasx") -TARGET_BUILTIN(__builtin_lasx_xvsubi_bu, "V32cV32cIUi", "nc", "lasx") +TARGET_BUILTIN(__builtin_lasx_xvsubi_bu, "V32ScV32ScIUi", "nc", "lasx") TARGET_BUILTIN(__builtin_lasx_xvsubi_hu, "V16sV16sIUi", "nc", "lasx") TARGET_BUILTIN(__builtin_lasx_xvsubi_wu, "V8iV8iIUi", "nc", "lasx") TARGET_BUILTIN(__builtin_lasx_xvsubi_du, "V4LLiV4LLiIUi", "nc", "lasx") -TARGET_BUILTIN(__builtin_lasx_xvneg_b, "V32cV32c", "nc", "lasx") +TARGET_BUILTIN(__builtin_lasx_xvneg_b, "V32ScV32Sc", "nc", "lasx") TARGET_BUILTIN(__builtin_lasx_xvneg_h, "V16sV16s", "nc", "lasx") TARGET_BUILTIN(__builtin_lasx_xvneg_w, "V8iV8i", "nc", "lasx") TARGET_BUILTIN(__builtin_lasx_xvneg_d, "V4LLiV4LLi", "nc", "lasx") @@ -79,22 +79,22 @@ TARGET_BUILTIN(__builtin_lasx_xvhsubw_wu_hu, "V8UiV16UsV16Us", "nc", "lasx") TARGET_BUILTIN(__builtin_lasx_xvhsubw_du_wu, "V4ULLiV8UiV8Ui", "nc", "lasx") TARGET_BUILTIN(__builtin_lasx_xvhsubw_qu_du, "V4ULLiV4ULLiV4ULLi", "nc", "lasx") -TARGET_BUILTIN(__builtin_lasx_xvaddwev_h_b, "V16sV32cV32c", "nc", "lasx") +TARGET_BUILTIN(__builtin_lasx_xvaddwev_h_b, "V16sV32ScV32Sc", "nc", "lasx") TARGET_BUILTIN(__builtin_lasx_xvaddwev_w_h, "V8SiV16sV16s", "nc", "lasx") TARGET_BUILTIN(__builtin_lasx_xvaddwev_d_w, "V4LLiV8SiV8Si", "nc", "lasx") TARGET_BUILTIN(__builtin_lasx_xvaddwev_q_d, "V4LLiV4LLiV4LLi", "nc", "lasx") -TARGET_BUILTIN(__builtin_lasx_xvaddwod_h_b, "V16sV32cV32c", "nc", "lasx") +TARGET_BUILTIN(__builtin_lasx_xvaddwod_h_b, "V16sV32ScV32Sc", "nc", "lasx") TARGET_BUILTIN(__builtin_lasx_xvaddwod_w_h, "V8SiV16sV16s", "nc", "lasx") TARGET_BUILTIN(__builtin_lasx_xvaddwod_d_w, "V4LLiV8SiV8Si", "nc", "lasx") TARGET_BUILTIN(__builtin_lasx_xvaddwod_q_d, "V4LLiV4LLiV4LLi", "nc", "lasx") -TARGET_BUILTIN(__builtin_lasx_xvsubwev_h_b, "V16sV32cV32c", "nc", "lasx") +TARGET_BUILTIN(__builtin_lasx_xvsubwev_h_b, "V16sV32ScV32Sc", "nc", "lasx") TARGET_BUILTIN(__builtin_lasx_xvsubwev_w_h, "V8SiV16sV16s", "nc", "lasx") TARGET_BUILTIN(__builtin_lasx_xvsubwev_d_w, "V4LLiV8SiV8Si", "nc", "lasx") TARGET_BUILTIN(__builtin_lasx_xvsubwev_q_d, "V4LLiV4LLiV4LLi", "nc", "lasx") -TARGET_BUILTIN(__builtin_lasx_xvsubwod_h_b, "V16sV32cV32c", "nc", "lasx") +TARGET_BUILTIN(__builtin_lasx_xvsubwod_h_b, "V16sV32ScV32Sc", "nc", "lasx") TARGET_BUILTIN(__builtin_lasx_xvsubwod_w_h, "V8SiV16sV16s", "nc", "lasx") TARGET_BUILTIN(__builtin_lasx
[clang] [LoongArch][Clang] Make the parameter and return value of {x,}vmsknz.b builtins `signed char` vector (PR #114511)
https://github.com/xry111 updated
https://github.com/llvm/llvm-project/pull/114511
>From ef96a6e483030e649d69ffa778c0b0f229bbb739 Mon Sep 17 00:00:00 2001
From: Xi Ruoyao
Date: Fri, 1 Nov 2024 00:13:29 +0800
Subject: [PATCH] [LoongArch][Clang] Make the parameter and return value of
{x,}vmsknz.b builtins `signed char` vector
These builtins operate on int8 vectors, not int16 vectors. So the old
definition does not make any sense.
---
clang/include/clang/Basic/BuiltinsLoongArchLASX.def | 2 +-
clang/include/clang/Basic/BuiltinsLoongArchLSX.def | 2 +-
2 files changed, 2 insertions(+), 2 deletions(-)
diff --git a/clang/include/clang/Basic/BuiltinsLoongArchLASX.def
b/clang/include/clang/Basic/BuiltinsLoongArchLASX.def
index 0d7c2df5c5c503..477f704860873c 100644
--- a/clang/include/clang/Basic/BuiltinsLoongArchLASX.def
+++ b/clang/include/clang/Basic/BuiltinsLoongArchLASX.def
@@ -357,7 +357,7 @@ TARGET_BUILTIN(__builtin_lasx_xvmskltz_w, "V8iV8i", "nc",
"lasx")
TARGET_BUILTIN(__builtin_lasx_xvmskltz_d, "V4LLiV4LLi", "nc", "lasx")
TARGET_BUILTIN(__builtin_lasx_xvmskgez_b, "V32ScV32Sc", "nc", "lasx")
-TARGET_BUILTIN(__builtin_lasx_xvmsknz_b, "V16sV16s", "nc", "lasx")
+TARGET_BUILTIN(__builtin_lasx_xvmsknz_b, "V32ScV32Sc", "nc", "lasx")
TARGET_BUILTIN(__builtin_lasx_xvldi, "V4LLiIi", "nc", "lasx")
TARGET_BUILTIN(__builtin_lasx_xvrepli_b, "V32ScIi", "nc", "lasx")
diff --git a/clang/include/clang/Basic/BuiltinsLoongArchLSX.def
b/clang/include/clang/Basic/BuiltinsLoongArchLSX.def
index 25a178e1ca98ab..e847985f390367 100644
--- a/clang/include/clang/Basic/BuiltinsLoongArchLSX.def
+++ b/clang/include/clang/Basic/BuiltinsLoongArchLSX.def
@@ -341,7 +341,7 @@ TARGET_BUILTIN(__builtin_lsx_vmskltz_w, "V4iV4i", "nc",
"lsx")
TARGET_BUILTIN(__builtin_lsx_vmskltz_d, "V2LLiV2LLi", "nc", "lsx")
TARGET_BUILTIN(__builtin_lsx_vmskgez_b, "V16ScV16Sc", "nc", "lsx")
-TARGET_BUILTIN(__builtin_lsx_vmsknz_b, "V8sV8s", "nc", "lsx")
+TARGET_BUILTIN(__builtin_lsx_vmsknz_b, "V16ScV16Sc", "nc", "lsx")
TARGET_BUILTIN(__builtin_lsx_vldi, "V2LLiIi", "nc", "lsx")
TARGET_BUILTIN(__builtin_lsx_vrepli_b, "V16ScIi", "nc", "lsx")
___
cfe-commits mailing list
cfe-commits@lists.llvm.org
https://lists.llvm.org/cgi-bin/mailman/listinfo/cfe-commits
[clang] [LoongArch][Clang] Make the parameters and return value of {x,}vxor.v builti ns `unsigned char` vectors (PR #114513)
https://github.com/xry111 updated https://github.com/llvm/llvm-project/pull/114513 >From 3d65a59473580f1c0e0b57bdd7fb13c9f40d3077 Mon Sep 17 00:00:00 2001 From: Xi Ruoyao Date: Fri, 1 Nov 2024 00:07:17 +0800 Subject: [PATCH 1/4] [LoongArch][clang] Use `signed char` vectors instead of `char` vectors for LSX and LASX builtins `-flax-vector-conversions=none` does not allow an implicit conversion from `signed char` vector to `char` vector, and we cannot remove `signed` from `v16i8` or `v32i8` because doing so will break our expectation with `-fno-signed-char`. So to make lsxintrin.h and lasxintrin.h work fine with `-flax-vector-conversions=none`, we must use `signed char` instead of `char`. The change is just done via sed 's/V16c/V16Sc/g' -i BuiltinsLoongArchLSX.def sed 's/V32c/V32Sc/g' -i BuiltinsLoongArchLASX.def --- .../clang/Basic/BuiltinsLoongArchLASX.def | 142 +- .../clang/Basic/BuiltinsLoongArchLSX.def | 128 2 files changed, 135 insertions(+), 135 deletions(-) diff --git a/clang/include/clang/Basic/BuiltinsLoongArchLASX.def b/clang/include/clang/Basic/BuiltinsLoongArchLASX.def index 4cf51cc000f6fc..0d7c2df5c5c503 100644 --- a/clang/include/clang/Basic/BuiltinsLoongArchLASX.def +++ b/clang/include/clang/Basic/BuiltinsLoongArchLASX.def @@ -12,29 +12,29 @@ // //===--===// -TARGET_BUILTIN(__builtin_lasx_xvadd_b, "V32cV32cV32c", "nc", "lasx") +TARGET_BUILTIN(__builtin_lasx_xvadd_b, "V32ScV32ScV32Sc", "nc", "lasx") TARGET_BUILTIN(__builtin_lasx_xvadd_h, "V16sV16sV16s", "nc", "lasx") TARGET_BUILTIN(__builtin_lasx_xvadd_w, "V8iV8iV8i", "nc", "lasx") TARGET_BUILTIN(__builtin_lasx_xvadd_d, "V4LLiV4LLiV4LLi", "nc", "lasx") TARGET_BUILTIN(__builtin_lasx_xvadd_q, "V4LLiV4LLiV4LLi", "nc", "lasx") -TARGET_BUILTIN(__builtin_lasx_xvsub_b, "V32cV32cV32c", "nc", "lasx") +TARGET_BUILTIN(__builtin_lasx_xvsub_b, "V32ScV32ScV32Sc", "nc", "lasx") TARGET_BUILTIN(__builtin_lasx_xvsub_h, "V16sV16sV16s", "nc", "lasx") TARGET_BUILTIN(__builtin_lasx_xvsub_w, "V8iV8iV8i", "nc", "lasx") TARGET_BUILTIN(__builtin_lasx_xvsub_d, "V4LLiV4LLiV4LLi", "nc", "lasx") TARGET_BUILTIN(__builtin_lasx_xvsub_q, "V4LLiV4LLiV4LLi", "nc", "lasx") -TARGET_BUILTIN(__builtin_lasx_xvaddi_bu, "V32cV32cIUi", "nc", "lasx") +TARGET_BUILTIN(__builtin_lasx_xvaddi_bu, "V32ScV32ScIUi", "nc", "lasx") TARGET_BUILTIN(__builtin_lasx_xvaddi_hu, "V16sV16sIUi", "nc", "lasx") TARGET_BUILTIN(__builtin_lasx_xvaddi_wu, "V8iV8iIUi", "nc", "lasx") TARGET_BUILTIN(__builtin_lasx_xvaddi_du, "V4LLiV4LLiIUi", "nc", "lasx") -TARGET_BUILTIN(__builtin_lasx_xvsubi_bu, "V32cV32cIUi", "nc", "lasx") +TARGET_BUILTIN(__builtin_lasx_xvsubi_bu, "V32ScV32ScIUi", "nc", "lasx") TARGET_BUILTIN(__builtin_lasx_xvsubi_hu, "V16sV16sIUi", "nc", "lasx") TARGET_BUILTIN(__builtin_lasx_xvsubi_wu, "V8iV8iIUi", "nc", "lasx") TARGET_BUILTIN(__builtin_lasx_xvsubi_du, "V4LLiV4LLiIUi", "nc", "lasx") -TARGET_BUILTIN(__builtin_lasx_xvneg_b, "V32cV32c", "nc", "lasx") +TARGET_BUILTIN(__builtin_lasx_xvneg_b, "V32ScV32Sc", "nc", "lasx") TARGET_BUILTIN(__builtin_lasx_xvneg_h, "V16sV16s", "nc", "lasx") TARGET_BUILTIN(__builtin_lasx_xvneg_w, "V8iV8i", "nc", "lasx") TARGET_BUILTIN(__builtin_lasx_xvneg_d, "V4LLiV4LLi", "nc", "lasx") @@ -79,22 +79,22 @@ TARGET_BUILTIN(__builtin_lasx_xvhsubw_wu_hu, "V8UiV16UsV16Us", "nc", "lasx") TARGET_BUILTIN(__builtin_lasx_xvhsubw_du_wu, "V4ULLiV8UiV8Ui", "nc", "lasx") TARGET_BUILTIN(__builtin_lasx_xvhsubw_qu_du, "V4ULLiV4ULLiV4ULLi", "nc", "lasx") -TARGET_BUILTIN(__builtin_lasx_xvaddwev_h_b, "V16sV32cV32c", "nc", "lasx") +TARGET_BUILTIN(__builtin_lasx_xvaddwev_h_b, "V16sV32ScV32Sc", "nc", "lasx") TARGET_BUILTIN(__builtin_lasx_xvaddwev_w_h, "V8SiV16sV16s", "nc", "lasx") TARGET_BUILTIN(__builtin_lasx_xvaddwev_d_w, "V4LLiV8SiV8Si", "nc", "lasx") TARGET_BUILTIN(__builtin_lasx_xvaddwev_q_d, "V4LLiV4LLiV4LLi", "nc", "lasx") -TARGET_BUILTIN(__builtin_lasx_xvaddwod_h_b, "V16sV32cV32c", "nc", "lasx") +TARGET_BUILTIN(__builtin_lasx_xvaddwod_h_b, "V16sV32ScV32Sc", "nc", "lasx") TARGET_BUILTIN(__builtin_lasx_xvaddwod_w_h, "V8SiV16sV16s", "nc", "lasx") TARGET_BUILTIN(__builtin_lasx_xvaddwod_d_w, "V4LLiV8SiV8Si", "nc", "lasx") TARGET_BUILTIN(__builtin_lasx_xvaddwod_q_d, "V4LLiV4LLiV4LLi", "nc", "lasx") -TARGET_BUILTIN(__builtin_lasx_xvsubwev_h_b, "V16sV32cV32c", "nc", "lasx") +TARGET_BUILTIN(__builtin_lasx_xvsubwev_h_b, "V16sV32ScV32Sc", "nc", "lasx") TARGET_BUILTIN(__builtin_lasx_xvsubwev_w_h, "V8SiV16sV16s", "nc", "lasx") TARGET_BUILTIN(__builtin_lasx_xvsubwev_d_w, "V4LLiV8SiV8Si", "nc", "lasx") TARGET_BUILTIN(__builtin_lasx_xvsubwev_q_d, "V4LLiV4LLiV4LLi", "nc", "lasx") -TARGET_BUILTIN(__builtin_lasx_xvsubwod_h_b, "V16sV32cV32c", "nc", "lasx") +TARGET_BUILTIN(__builtin_lasx_xvsubwod_h_b, "V16sV32ScV32Sc", "nc", "lasx") TARGET_BUILTIN(__builtin_lasx_xvsubwod_w_h, "V8SiV16sV16s", "nc", "lasx") TARGET_BUILTIN(__builtin_
[clang] [LoongArch][Clang] Make the parameters and return value of {x,}vorn.v builti ns `unsigned char` vectors (PR #114514)
https://github.com/xry111 updated
https://github.com/llvm/llvm-project/pull/114514
>From ef96a6e483030e649d69ffa778c0b0f229bbb739 Mon Sep 17 00:00:00 2001
From: Xi Ruoyao
Date: Fri, 1 Nov 2024 00:13:29 +0800
Subject: [PATCH 1/4] [LoongArch][Clang] Make the parameter and return value of
{x,}vmsknz.b builtins `signed char` vector
These builtins operate on int8 vectors, not int16 vectors. So the old
definition does not make any sense.
---
clang/include/clang/Basic/BuiltinsLoongArchLASX.def | 2 +-
clang/include/clang/Basic/BuiltinsLoongArchLSX.def | 2 +-
2 files changed, 2 insertions(+), 2 deletions(-)
diff --git a/clang/include/clang/Basic/BuiltinsLoongArchLASX.def
b/clang/include/clang/Basic/BuiltinsLoongArchLASX.def
index 0d7c2df5c5c503..477f704860873c 100644
--- a/clang/include/clang/Basic/BuiltinsLoongArchLASX.def
+++ b/clang/include/clang/Basic/BuiltinsLoongArchLASX.def
@@ -357,7 +357,7 @@ TARGET_BUILTIN(__builtin_lasx_xvmskltz_w, "V8iV8i", "nc",
"lasx")
TARGET_BUILTIN(__builtin_lasx_xvmskltz_d, "V4LLiV4LLi", "nc", "lasx")
TARGET_BUILTIN(__builtin_lasx_xvmskgez_b, "V32ScV32Sc", "nc", "lasx")
-TARGET_BUILTIN(__builtin_lasx_xvmsknz_b, "V16sV16s", "nc", "lasx")
+TARGET_BUILTIN(__builtin_lasx_xvmsknz_b, "V32ScV32Sc", "nc", "lasx")
TARGET_BUILTIN(__builtin_lasx_xvldi, "V4LLiIi", "nc", "lasx")
TARGET_BUILTIN(__builtin_lasx_xvrepli_b, "V32ScIi", "nc", "lasx")
diff --git a/clang/include/clang/Basic/BuiltinsLoongArchLSX.def
b/clang/include/clang/Basic/BuiltinsLoongArchLSX.def
index 25a178e1ca98ab..e847985f390367 100644
--- a/clang/include/clang/Basic/BuiltinsLoongArchLSX.def
+++ b/clang/include/clang/Basic/BuiltinsLoongArchLSX.def
@@ -341,7 +341,7 @@ TARGET_BUILTIN(__builtin_lsx_vmskltz_w, "V4iV4i", "nc",
"lsx")
TARGET_BUILTIN(__builtin_lsx_vmskltz_d, "V2LLiV2LLi", "nc", "lsx")
TARGET_BUILTIN(__builtin_lsx_vmskgez_b, "V16ScV16Sc", "nc", "lsx")
-TARGET_BUILTIN(__builtin_lsx_vmsknz_b, "V8sV8s", "nc", "lsx")
+TARGET_BUILTIN(__builtin_lsx_vmsknz_b, "V16ScV16Sc", "nc", "lsx")
TARGET_BUILTIN(__builtin_lsx_vldi, "V2LLiIi", "nc", "lsx")
TARGET_BUILTIN(__builtin_lsx_vrepli_b, "V16ScIi", "nc", "lsx")
>From 349b79172a1ede63a25173366c1607b001bf1511 Mon Sep 17 00:00:00 2001
From: Xi Ruoyao
Date: Fri, 1 Nov 2024 00:17:44 +0800
Subject: [PATCH 2/4] [LoongArch][Clang] Make the parameters and return value
of {x,}vshuf.b builtins `signed char` vectors
The lsxintrin.h and and lasxintrin.h headers uses `signed char` vectors
instead of `unsigned char` vectors. GCC also uses `signed char` for
them, so align their definition with the headers and GCC.
---
clang/include/clang/Basic/BuiltinsLoongArchLASX.def | 2 +-
clang/include/clang/Basic/BuiltinsLoongArchLSX.def | 2 +-
2 files changed, 2 insertions(+), 2 deletions(-)
diff --git a/clang/include/clang/Basic/BuiltinsLoongArchLASX.def
b/clang/include/clang/Basic/BuiltinsLoongArchLASX.def
index 477f704860873c..e6c41dd0974948 100644
--- a/clang/include/clang/Basic/BuiltinsLoongArchLASX.def
+++ b/clang/include/clang/Basic/BuiltinsLoongArchLASX.def
@@ -935,7 +935,7 @@ TARGET_BUILTIN(__builtin_lasx_xvilvh_h, "V16sV16sV16s",
"nc", "lasx")
TARGET_BUILTIN(__builtin_lasx_xvilvh_w, "V8iV8iV8i", "nc", "lasx")
TARGET_BUILTIN(__builtin_lasx_xvilvh_d, "V4LLiV4LLiV4LLi", "nc", "lasx")
-TARGET_BUILTIN(__builtin_lasx_xvshuf_b, "V32UcV32UcV32UcV32Uc", "nc", "lasx")
+TARGET_BUILTIN(__builtin_lasx_xvshuf_b, "V32ScV32ScV32ScV32Sc", "nc", "lasx")
TARGET_BUILTIN(__builtin_lasx_xvshuf_h, "V16sV16sV16sV16s", "nc", "lasx")
TARGET_BUILTIN(__builtin_lasx_xvshuf_w, "V8iV8iV8iV8i", "nc", "lasx")
diff --git a/clang/include/clang/Basic/BuiltinsLoongArchLSX.def
b/clang/include/clang/Basic/BuiltinsLoongArchLSX.def
index e847985f390367..ded6519f3ef16c 100644
--- a/clang/include/clang/Basic/BuiltinsLoongArchLSX.def
+++ b/clang/include/clang/Basic/BuiltinsLoongArchLSX.def
@@ -910,7 +910,7 @@ TARGET_BUILTIN(__builtin_lsx_vilvh_h, "V8sV8sV8s", "nc",
"lsx")
TARGET_BUILTIN(__builtin_lsx_vilvh_w, "V4iV4iV4i", "nc", "lsx")
TARGET_BUILTIN(__builtin_lsx_vilvh_d, "V2LLiV2LLiV2LLi", "nc", "lsx")
-TARGET_BUILTIN(__builtin_lsx_vshuf_b, "V16UcV16UcV16UcV16Uc", "nc", "lsx")
+TARGET_BUILTIN(__builtin_lsx_vshuf_b, "V16ScV16ScV16ScV16Sc", "nc", "lsx")
TARGET_BUILTIN(__builtin_lsx_vshuf_h, "V8sV8sV8sV8s", "nc", "lsx")
TARGET_BUILTIN(__builtin_lsx_vshuf_w, "V4iV4iV4iV4i", "nc", "lsx")
>From bf82a2490d77457f2afd92da7a09020f59d465e4 Mon Sep 17 00:00:00 2001
From: Xi Ruoyao
Date: Fri, 1 Nov 2024 00:22:09 +0800
Subject: [PATCH 3/4] [LoongArch][Clang] Make the parameters and return value
of {x,}vxor.v builtins `unsigned char` vectors
The lsxintrin.h and and lasxintrin.h headers uses `unsigned char`
vectors instead of `signed char` vectors. GCC also uses `unsigned char`
for them, so align their definition with the headers and GCC.
Fixes #110834.
---
clang/include/clang/Basic/BuiltinsLoongArchLASX.def | 2 +-
clang/include/clang/Basic/BuiltinsLoongArchLSX.def | 2 +-
clang/test/Head
[clang] [LoongArch][Clang] Make the parameters and return value of {x,}vxor.v builti ns `unsigned char` vectors (PR #114513)
https://github.com/xry111 updated
https://github.com/llvm/llvm-project/pull/114513
>From ef96a6e483030e649d69ffa778c0b0f229bbb739 Mon Sep 17 00:00:00 2001
From: Xi Ruoyao
Date: Fri, 1 Nov 2024 00:13:29 +0800
Subject: [PATCH 1/3] [LoongArch][Clang] Make the parameter and return value of
{x,}vmsknz.b builtins `signed char` vector
These builtins operate on int8 vectors, not int16 vectors. So the old
definition does not make any sense.
---
clang/include/clang/Basic/BuiltinsLoongArchLASX.def | 2 +-
clang/include/clang/Basic/BuiltinsLoongArchLSX.def | 2 +-
2 files changed, 2 insertions(+), 2 deletions(-)
diff --git a/clang/include/clang/Basic/BuiltinsLoongArchLASX.def
b/clang/include/clang/Basic/BuiltinsLoongArchLASX.def
index 0d7c2df5c5c503..477f704860873c 100644
--- a/clang/include/clang/Basic/BuiltinsLoongArchLASX.def
+++ b/clang/include/clang/Basic/BuiltinsLoongArchLASX.def
@@ -357,7 +357,7 @@ TARGET_BUILTIN(__builtin_lasx_xvmskltz_w, "V8iV8i", "nc",
"lasx")
TARGET_BUILTIN(__builtin_lasx_xvmskltz_d, "V4LLiV4LLi", "nc", "lasx")
TARGET_BUILTIN(__builtin_lasx_xvmskgez_b, "V32ScV32Sc", "nc", "lasx")
-TARGET_BUILTIN(__builtin_lasx_xvmsknz_b, "V16sV16s", "nc", "lasx")
+TARGET_BUILTIN(__builtin_lasx_xvmsknz_b, "V32ScV32Sc", "nc", "lasx")
TARGET_BUILTIN(__builtin_lasx_xvldi, "V4LLiIi", "nc", "lasx")
TARGET_BUILTIN(__builtin_lasx_xvrepli_b, "V32ScIi", "nc", "lasx")
diff --git a/clang/include/clang/Basic/BuiltinsLoongArchLSX.def
b/clang/include/clang/Basic/BuiltinsLoongArchLSX.def
index 25a178e1ca98ab..e847985f390367 100644
--- a/clang/include/clang/Basic/BuiltinsLoongArchLSX.def
+++ b/clang/include/clang/Basic/BuiltinsLoongArchLSX.def
@@ -341,7 +341,7 @@ TARGET_BUILTIN(__builtin_lsx_vmskltz_w, "V4iV4i", "nc",
"lsx")
TARGET_BUILTIN(__builtin_lsx_vmskltz_d, "V2LLiV2LLi", "nc", "lsx")
TARGET_BUILTIN(__builtin_lsx_vmskgez_b, "V16ScV16Sc", "nc", "lsx")
-TARGET_BUILTIN(__builtin_lsx_vmsknz_b, "V8sV8s", "nc", "lsx")
+TARGET_BUILTIN(__builtin_lsx_vmsknz_b, "V16ScV16Sc", "nc", "lsx")
TARGET_BUILTIN(__builtin_lsx_vldi, "V2LLiIi", "nc", "lsx")
TARGET_BUILTIN(__builtin_lsx_vrepli_b, "V16ScIi", "nc", "lsx")
>From 349b79172a1ede63a25173366c1607b001bf1511 Mon Sep 17 00:00:00 2001
From: Xi Ruoyao
Date: Fri, 1 Nov 2024 00:17:44 +0800
Subject: [PATCH 2/3] [LoongArch][Clang] Make the parameters and return value
of {x,}vshuf.b builtins `signed char` vectors
The lsxintrin.h and and lasxintrin.h headers uses `signed char` vectors
instead of `unsigned char` vectors. GCC also uses `signed char` for
them, so align their definition with the headers and GCC.
---
clang/include/clang/Basic/BuiltinsLoongArchLASX.def | 2 +-
clang/include/clang/Basic/BuiltinsLoongArchLSX.def | 2 +-
2 files changed, 2 insertions(+), 2 deletions(-)
diff --git a/clang/include/clang/Basic/BuiltinsLoongArchLASX.def
b/clang/include/clang/Basic/BuiltinsLoongArchLASX.def
index 477f704860873c..e6c41dd0974948 100644
--- a/clang/include/clang/Basic/BuiltinsLoongArchLASX.def
+++ b/clang/include/clang/Basic/BuiltinsLoongArchLASX.def
@@ -935,7 +935,7 @@ TARGET_BUILTIN(__builtin_lasx_xvilvh_h, "V16sV16sV16s",
"nc", "lasx")
TARGET_BUILTIN(__builtin_lasx_xvilvh_w, "V8iV8iV8i", "nc", "lasx")
TARGET_BUILTIN(__builtin_lasx_xvilvh_d, "V4LLiV4LLiV4LLi", "nc", "lasx")
-TARGET_BUILTIN(__builtin_lasx_xvshuf_b, "V32UcV32UcV32UcV32Uc", "nc", "lasx")
+TARGET_BUILTIN(__builtin_lasx_xvshuf_b, "V32ScV32ScV32ScV32Sc", "nc", "lasx")
TARGET_BUILTIN(__builtin_lasx_xvshuf_h, "V16sV16sV16sV16s", "nc", "lasx")
TARGET_BUILTIN(__builtin_lasx_xvshuf_w, "V8iV8iV8iV8i", "nc", "lasx")
diff --git a/clang/include/clang/Basic/BuiltinsLoongArchLSX.def
b/clang/include/clang/Basic/BuiltinsLoongArchLSX.def
index e847985f390367..ded6519f3ef16c 100644
--- a/clang/include/clang/Basic/BuiltinsLoongArchLSX.def
+++ b/clang/include/clang/Basic/BuiltinsLoongArchLSX.def
@@ -910,7 +910,7 @@ TARGET_BUILTIN(__builtin_lsx_vilvh_h, "V8sV8sV8s", "nc",
"lsx")
TARGET_BUILTIN(__builtin_lsx_vilvh_w, "V4iV4iV4i", "nc", "lsx")
TARGET_BUILTIN(__builtin_lsx_vilvh_d, "V2LLiV2LLiV2LLi", "nc", "lsx")
-TARGET_BUILTIN(__builtin_lsx_vshuf_b, "V16UcV16UcV16UcV16Uc", "nc", "lsx")
+TARGET_BUILTIN(__builtin_lsx_vshuf_b, "V16ScV16ScV16ScV16Sc", "nc", "lsx")
TARGET_BUILTIN(__builtin_lsx_vshuf_h, "V8sV8sV8sV8s", "nc", "lsx")
TARGET_BUILTIN(__builtin_lsx_vshuf_w, "V4iV4iV4iV4i", "nc", "lsx")
>From bf82a2490d77457f2afd92da7a09020f59d465e4 Mon Sep 17 00:00:00 2001
From: Xi Ruoyao
Date: Fri, 1 Nov 2024 00:22:09 +0800
Subject: [PATCH 3/3] [LoongArch][Clang] Make the parameters and return value
of {x,}vxor.v builtins `unsigned char` vectors
The lsxintrin.h and and lasxintrin.h headers uses `unsigned char`
vectors instead of `signed char` vectors. GCC also uses `unsigned char`
for them, so align their definition with the headers and GCC.
Fixes #110834.
---
clang/include/clang/Basic/BuiltinsLoongArchLASX.def | 2 +-
clang/include/clang/Basic/BuiltinsLoongArchLSX.def | 2 +-
clang/test/Head
[clang] [LoongArch][Clang] Make the parameters and return value of {x,}vorn.v builti ns `unsigned char` vectors (PR #114514)
https://github.com/xry111 updated
https://github.com/llvm/llvm-project/pull/114514
>From c1b7b79e7027ad7fed410c1783113228173843f1 Mon Sep 17 00:00:00 2001
From: Xi Ruoyao
Date: Fri, 1 Nov 2024 00:17:44 +0800
Subject: [PATCH 1/3] [LoongArch][Clang] Make the parameters and return value
of {x,}vshuf.b builtins `signed char` vectors
The lsxintrin.h and and lasxintrin.h headers uses `signed char` vectors
instead of `unsigned char` vectors. GCC also uses `signed char` for
them, so align their definition with the headers and GCC.
---
clang/include/clang/Basic/BuiltinsLoongArchLASX.def | 2 +-
clang/include/clang/Basic/BuiltinsLoongArchLSX.def | 2 +-
2 files changed, 2 insertions(+), 2 deletions(-)
diff --git a/clang/include/clang/Basic/BuiltinsLoongArchLASX.def
b/clang/include/clang/Basic/BuiltinsLoongArchLASX.def
index 477f704860873c..e6c41dd0974948 100644
--- a/clang/include/clang/Basic/BuiltinsLoongArchLASX.def
+++ b/clang/include/clang/Basic/BuiltinsLoongArchLASX.def
@@ -935,7 +935,7 @@ TARGET_BUILTIN(__builtin_lasx_xvilvh_h, "V16sV16sV16s",
"nc", "lasx")
TARGET_BUILTIN(__builtin_lasx_xvilvh_w, "V8iV8iV8i", "nc", "lasx")
TARGET_BUILTIN(__builtin_lasx_xvilvh_d, "V4LLiV4LLiV4LLi", "nc", "lasx")
-TARGET_BUILTIN(__builtin_lasx_xvshuf_b, "V32UcV32UcV32UcV32Uc", "nc", "lasx")
+TARGET_BUILTIN(__builtin_lasx_xvshuf_b, "V32ScV32ScV32ScV32Sc", "nc", "lasx")
TARGET_BUILTIN(__builtin_lasx_xvshuf_h, "V16sV16sV16sV16s", "nc", "lasx")
TARGET_BUILTIN(__builtin_lasx_xvshuf_w, "V8iV8iV8iV8i", "nc", "lasx")
diff --git a/clang/include/clang/Basic/BuiltinsLoongArchLSX.def
b/clang/include/clang/Basic/BuiltinsLoongArchLSX.def
index e847985f390367..ded6519f3ef16c 100644
--- a/clang/include/clang/Basic/BuiltinsLoongArchLSX.def
+++ b/clang/include/clang/Basic/BuiltinsLoongArchLSX.def
@@ -910,7 +910,7 @@ TARGET_BUILTIN(__builtin_lsx_vilvh_h, "V8sV8sV8s", "nc",
"lsx")
TARGET_BUILTIN(__builtin_lsx_vilvh_w, "V4iV4iV4i", "nc", "lsx")
TARGET_BUILTIN(__builtin_lsx_vilvh_d, "V2LLiV2LLiV2LLi", "nc", "lsx")
-TARGET_BUILTIN(__builtin_lsx_vshuf_b, "V16UcV16UcV16UcV16Uc", "nc", "lsx")
+TARGET_BUILTIN(__builtin_lsx_vshuf_b, "V16ScV16ScV16ScV16Sc", "nc", "lsx")
TARGET_BUILTIN(__builtin_lsx_vshuf_h, "V8sV8sV8sV8s", "nc", "lsx")
TARGET_BUILTIN(__builtin_lsx_vshuf_w, "V4iV4iV4iV4i", "nc", "lsx")
>From b162c48b1cf4c0312d87f094f232e807cea75bc9 Mon Sep 17 00:00:00 2001
From: Xi Ruoyao
Date: Fri, 1 Nov 2024 00:22:09 +0800
Subject: [PATCH 2/3] [LoongArch][Clang] Make the parameters and return value
of {x,}vxor.v builtins `unsigned char` vectors
The lsxintrin.h and and lasxintrin.h headers uses `unsigned char`
vectors instead of `signed char` vectors. GCC also uses `unsigned char`
for them, so align their definition with the headers and GCC.
Fixes #110834.
---
clang/include/clang/Basic/BuiltinsLoongArchLASX.def | 2 +-
clang/include/clang/Basic/BuiltinsLoongArchLSX.def | 2 +-
clang/test/Headers/lasxintrin.c | 5 ++---
clang/test/Headers/lsxintrin.c | 5 ++---
4 files changed, 6 insertions(+), 8 deletions(-)
diff --git a/clang/include/clang/Basic/BuiltinsLoongArchLASX.def
b/clang/include/clang/Basic/BuiltinsLoongArchLASX.def
index e6c41dd0974948..f644b820a61898 100644
--- a/clang/include/clang/Basic/BuiltinsLoongArchLASX.def
+++ b/clang/include/clang/Basic/BuiltinsLoongArchLASX.def
@@ -368,7 +368,7 @@ TARGET_BUILTIN(__builtin_lasx_xvrepli_d, "V4LLiIi", "nc",
"lasx")
TARGET_BUILTIN(__builtin_lasx_xvand_v, "V32UcV32UcV32Uc", "nc", "lasx")
TARGET_BUILTIN(__builtin_lasx_xvor_v, "V32UcV32UcV32Uc", "nc", "lasx")
-TARGET_BUILTIN(__builtin_lasx_xvxor_v, "V32ScV32ScV32Sc", "nc", "lasx")
+TARGET_BUILTIN(__builtin_lasx_xvxor_v, "V32UcV32UcV32Uc", "nc", "lasx")
TARGET_BUILTIN(__builtin_lasx_xvnor_v, "V32UcV32UcV32Uc", "nc", "lasx")
TARGET_BUILTIN(__builtin_lasx_xvandn_v, "V32UcV32UcV32Uc", "nc", "lasx")
TARGET_BUILTIN(__builtin_lasx_xvorn_v, "V32ScV32ScV32Sc", "nc", "lasx")
diff --git a/clang/include/clang/Basic/BuiltinsLoongArchLSX.def
b/clang/include/clang/Basic/BuiltinsLoongArchLSX.def
index ded6519f3ef16c..b3056971986d19 100644
--- a/clang/include/clang/Basic/BuiltinsLoongArchLSX.def
+++ b/clang/include/clang/Basic/BuiltinsLoongArchLSX.def
@@ -352,7 +352,7 @@ TARGET_BUILTIN(__builtin_lsx_vrepli_d, "V2LLiIi", "nc",
"lsx")
TARGET_BUILTIN(__builtin_lsx_vand_v, "V16UcV16UcV16Uc", "nc", "lsx")
TARGET_BUILTIN(__builtin_lsx_vor_v, "V16UcV16UcV16Uc", "nc", "lsx")
-TARGET_BUILTIN(__builtin_lsx_vxor_v, "V16ScV16ScV16Sc", "nc", "lsx")
+TARGET_BUILTIN(__builtin_lsx_vxor_v, "V16UcV16UcV16Uc", "nc", "lsx")
TARGET_BUILTIN(__builtin_lsx_vnor_v, "V16UcV16UcV16Uc", "nc", "lsx")
TARGET_BUILTIN(__builtin_lsx_vandn_v, "V16UcV16UcV16Uc", "nc", "lsx")
TARGET_BUILTIN(__builtin_lsx_vorn_v, "V16ScV16ScV16Sc", "nc", "lsx")
diff --git a/clang/test/Headers/lasxintrin.c b/clang/test/Headers/lasxintrin.c
index 6126c6350455c1..08f71791bdf36f 100644
--- a/clang/test/Headers/lasxintrin.c
+++ b/clang/te
[clang] [LoongArch][Clang] Make the parameters and return value of {x,}vxor.v builti ns `unsigned char` vectors (PR #114513)
https://github.com/xry111 updated
https://github.com/llvm/llvm-project/pull/114513
>From c1b7b79e7027ad7fed410c1783113228173843f1 Mon Sep 17 00:00:00 2001
From: Xi Ruoyao
Date: Fri, 1 Nov 2024 00:17:44 +0800
Subject: [PATCH 1/2] [LoongArch][Clang] Make the parameters and return value
of {x,}vshuf.b builtins `signed char` vectors
The lsxintrin.h and and lasxintrin.h headers uses `signed char` vectors
instead of `unsigned char` vectors. GCC also uses `signed char` for
them, so align their definition with the headers and GCC.
---
clang/include/clang/Basic/BuiltinsLoongArchLASX.def | 2 +-
clang/include/clang/Basic/BuiltinsLoongArchLSX.def | 2 +-
2 files changed, 2 insertions(+), 2 deletions(-)
diff --git a/clang/include/clang/Basic/BuiltinsLoongArchLASX.def
b/clang/include/clang/Basic/BuiltinsLoongArchLASX.def
index 477f704860873c..e6c41dd0974948 100644
--- a/clang/include/clang/Basic/BuiltinsLoongArchLASX.def
+++ b/clang/include/clang/Basic/BuiltinsLoongArchLASX.def
@@ -935,7 +935,7 @@ TARGET_BUILTIN(__builtin_lasx_xvilvh_h, "V16sV16sV16s",
"nc", "lasx")
TARGET_BUILTIN(__builtin_lasx_xvilvh_w, "V8iV8iV8i", "nc", "lasx")
TARGET_BUILTIN(__builtin_lasx_xvilvh_d, "V4LLiV4LLiV4LLi", "nc", "lasx")
-TARGET_BUILTIN(__builtin_lasx_xvshuf_b, "V32UcV32UcV32UcV32Uc", "nc", "lasx")
+TARGET_BUILTIN(__builtin_lasx_xvshuf_b, "V32ScV32ScV32ScV32Sc", "nc", "lasx")
TARGET_BUILTIN(__builtin_lasx_xvshuf_h, "V16sV16sV16sV16s", "nc", "lasx")
TARGET_BUILTIN(__builtin_lasx_xvshuf_w, "V8iV8iV8iV8i", "nc", "lasx")
diff --git a/clang/include/clang/Basic/BuiltinsLoongArchLSX.def
b/clang/include/clang/Basic/BuiltinsLoongArchLSX.def
index e847985f390367..ded6519f3ef16c 100644
--- a/clang/include/clang/Basic/BuiltinsLoongArchLSX.def
+++ b/clang/include/clang/Basic/BuiltinsLoongArchLSX.def
@@ -910,7 +910,7 @@ TARGET_BUILTIN(__builtin_lsx_vilvh_h, "V8sV8sV8s", "nc",
"lsx")
TARGET_BUILTIN(__builtin_lsx_vilvh_w, "V4iV4iV4i", "nc", "lsx")
TARGET_BUILTIN(__builtin_lsx_vilvh_d, "V2LLiV2LLiV2LLi", "nc", "lsx")
-TARGET_BUILTIN(__builtin_lsx_vshuf_b, "V16UcV16UcV16UcV16Uc", "nc", "lsx")
+TARGET_BUILTIN(__builtin_lsx_vshuf_b, "V16ScV16ScV16ScV16Sc", "nc", "lsx")
TARGET_BUILTIN(__builtin_lsx_vshuf_h, "V8sV8sV8sV8s", "nc", "lsx")
TARGET_BUILTIN(__builtin_lsx_vshuf_w, "V4iV4iV4iV4i", "nc", "lsx")
>From b162c48b1cf4c0312d87f094f232e807cea75bc9 Mon Sep 17 00:00:00 2001
From: Xi Ruoyao
Date: Fri, 1 Nov 2024 00:22:09 +0800
Subject: [PATCH 2/2] [LoongArch][Clang] Make the parameters and return value
of {x,}vxor.v builtins `unsigned char` vectors
The lsxintrin.h and and lasxintrin.h headers uses `unsigned char`
vectors instead of `signed char` vectors. GCC also uses `unsigned char`
for them, so align their definition with the headers and GCC.
Fixes #110834.
---
clang/include/clang/Basic/BuiltinsLoongArchLASX.def | 2 +-
clang/include/clang/Basic/BuiltinsLoongArchLSX.def | 2 +-
clang/test/Headers/lasxintrin.c | 5 ++---
clang/test/Headers/lsxintrin.c | 5 ++---
4 files changed, 6 insertions(+), 8 deletions(-)
diff --git a/clang/include/clang/Basic/BuiltinsLoongArchLASX.def
b/clang/include/clang/Basic/BuiltinsLoongArchLASX.def
index e6c41dd0974948..f644b820a61898 100644
--- a/clang/include/clang/Basic/BuiltinsLoongArchLASX.def
+++ b/clang/include/clang/Basic/BuiltinsLoongArchLASX.def
@@ -368,7 +368,7 @@ TARGET_BUILTIN(__builtin_lasx_xvrepli_d, "V4LLiIi", "nc",
"lasx")
TARGET_BUILTIN(__builtin_lasx_xvand_v, "V32UcV32UcV32Uc", "nc", "lasx")
TARGET_BUILTIN(__builtin_lasx_xvor_v, "V32UcV32UcV32Uc", "nc", "lasx")
-TARGET_BUILTIN(__builtin_lasx_xvxor_v, "V32ScV32ScV32Sc", "nc", "lasx")
+TARGET_BUILTIN(__builtin_lasx_xvxor_v, "V32UcV32UcV32Uc", "nc", "lasx")
TARGET_BUILTIN(__builtin_lasx_xvnor_v, "V32UcV32UcV32Uc", "nc", "lasx")
TARGET_BUILTIN(__builtin_lasx_xvandn_v, "V32UcV32UcV32Uc", "nc", "lasx")
TARGET_BUILTIN(__builtin_lasx_xvorn_v, "V32ScV32ScV32Sc", "nc", "lasx")
diff --git a/clang/include/clang/Basic/BuiltinsLoongArchLSX.def
b/clang/include/clang/Basic/BuiltinsLoongArchLSX.def
index ded6519f3ef16c..b3056971986d19 100644
--- a/clang/include/clang/Basic/BuiltinsLoongArchLSX.def
+++ b/clang/include/clang/Basic/BuiltinsLoongArchLSX.def
@@ -352,7 +352,7 @@ TARGET_BUILTIN(__builtin_lsx_vrepli_d, "V2LLiIi", "nc",
"lsx")
TARGET_BUILTIN(__builtin_lsx_vand_v, "V16UcV16UcV16Uc", "nc", "lsx")
TARGET_BUILTIN(__builtin_lsx_vor_v, "V16UcV16UcV16Uc", "nc", "lsx")
-TARGET_BUILTIN(__builtin_lsx_vxor_v, "V16ScV16ScV16Sc", "nc", "lsx")
+TARGET_BUILTIN(__builtin_lsx_vxor_v, "V16UcV16UcV16Uc", "nc", "lsx")
TARGET_BUILTIN(__builtin_lsx_vnor_v, "V16UcV16UcV16Uc", "nc", "lsx")
TARGET_BUILTIN(__builtin_lsx_vandn_v, "V16UcV16UcV16Uc", "nc", "lsx")
TARGET_BUILTIN(__builtin_lsx_vorn_v, "V16ScV16ScV16Sc", "nc", "lsx")
diff --git a/clang/test/Headers/lasxintrin.c b/clang/test/Headers/lasxintrin.c
index 6126c6350455c1..08f71791bdf36f 100644
--- a/clang/test/Headers/lasxintrin.c
+++ b/clang/te
[clang] [LoongArch][Clang] Make the parameters and return value of {x,}vshuf.b builtins `signed char` vectors (PR #114512)
https://github.com/xry111 updated
https://github.com/llvm/llvm-project/pull/114512
>From c1b7b79e7027ad7fed410c1783113228173843f1 Mon Sep 17 00:00:00 2001
From: Xi Ruoyao
Date: Fri, 1 Nov 2024 00:17:44 +0800
Subject: [PATCH] [LoongArch][Clang] Make the parameters and return value of
{x,}vshuf.b builtins `signed char` vectors
The lsxintrin.h and and lasxintrin.h headers uses `signed char` vectors
instead of `unsigned char` vectors. GCC also uses `signed char` for
them, so align their definition with the headers and GCC.
---
clang/include/clang/Basic/BuiltinsLoongArchLASX.def | 2 +-
clang/include/clang/Basic/BuiltinsLoongArchLSX.def | 2 +-
2 files changed, 2 insertions(+), 2 deletions(-)
diff --git a/clang/include/clang/Basic/BuiltinsLoongArchLASX.def
b/clang/include/clang/Basic/BuiltinsLoongArchLASX.def
index 477f704860873c..e6c41dd0974948 100644
--- a/clang/include/clang/Basic/BuiltinsLoongArchLASX.def
+++ b/clang/include/clang/Basic/BuiltinsLoongArchLASX.def
@@ -935,7 +935,7 @@ TARGET_BUILTIN(__builtin_lasx_xvilvh_h, "V16sV16sV16s",
"nc", "lasx")
TARGET_BUILTIN(__builtin_lasx_xvilvh_w, "V8iV8iV8i", "nc", "lasx")
TARGET_BUILTIN(__builtin_lasx_xvilvh_d, "V4LLiV4LLiV4LLi", "nc", "lasx")
-TARGET_BUILTIN(__builtin_lasx_xvshuf_b, "V32UcV32UcV32UcV32Uc", "nc", "lasx")
+TARGET_BUILTIN(__builtin_lasx_xvshuf_b, "V32ScV32ScV32ScV32Sc", "nc", "lasx")
TARGET_BUILTIN(__builtin_lasx_xvshuf_h, "V16sV16sV16sV16s", "nc", "lasx")
TARGET_BUILTIN(__builtin_lasx_xvshuf_w, "V8iV8iV8iV8i", "nc", "lasx")
diff --git a/clang/include/clang/Basic/BuiltinsLoongArchLSX.def
b/clang/include/clang/Basic/BuiltinsLoongArchLSX.def
index e847985f390367..ded6519f3ef16c 100644
--- a/clang/include/clang/Basic/BuiltinsLoongArchLSX.def
+++ b/clang/include/clang/Basic/BuiltinsLoongArchLSX.def
@@ -910,7 +910,7 @@ TARGET_BUILTIN(__builtin_lsx_vilvh_h, "V8sV8sV8s", "nc",
"lsx")
TARGET_BUILTIN(__builtin_lsx_vilvh_w, "V4iV4iV4i", "nc", "lsx")
TARGET_BUILTIN(__builtin_lsx_vilvh_d, "V2LLiV2LLiV2LLi", "nc", "lsx")
-TARGET_BUILTIN(__builtin_lsx_vshuf_b, "V16UcV16UcV16UcV16Uc", "nc", "lsx")
+TARGET_BUILTIN(__builtin_lsx_vshuf_b, "V16ScV16ScV16ScV16Sc", "nc", "lsx")
TARGET_BUILTIN(__builtin_lsx_vshuf_h, "V8sV8sV8sV8s", "nc", "lsx")
TARGET_BUILTIN(__builtin_lsx_vshuf_w, "V4iV4iV4iV4i", "nc", "lsx")
___
cfe-commits mailing list
cfe-commits@lists.llvm.org
https://lists.llvm.org/cgi-bin/mailman/listinfo/cfe-commits
[clang] [LoongArch][clang] Use `signed char` vectors instead of `char` vectors for L SX and LASX builtins (PR #114510)
https://github.com/xry111 created https://github.com/llvm/llvm-project/pull/114510 `-flax-vector-conversions=none` does not allow an implicit conversion from `signed char` vector to `char` vector, and we cannot remove `signed` from `v16i8` or `v32i8` because doing so will break our expectation with `-fno-signed-char`. So to make lsxintrin.h and lasxintrin.h work fine with `-flax-vector-conversions=none`, we must use `signed char` instead of `char`. The change is just done via sed 's/V16c/V16Sc/g' -i BuiltinsLoongArchLSX.def sed 's/V32c/V32Sc/g' -i BuiltinsLoongArchLASX.def Depends on #114509. Part of #110834 fix. >From 1bbbc1c38715f42e6014ee319d76af9a4015e4d5 Mon Sep 17 00:00:00 2001 From: Xi Ruoyao Date: Fri, 1 Nov 2024 00:00:14 +0800 Subject: [PATCH 1/2] [LoongArch][Clang] Add tests for #110834 --- clang/test/Headers/lasxintrin.c | 6 ++ clang/test/Headers/lsxintrin.c | 6 ++ 2 files changed, 12 insertions(+) create mode 100644 clang/test/Headers/lasxintrin.c create mode 100644 clang/test/Headers/lsxintrin.c diff --git a/clang/test/Headers/lasxintrin.c b/clang/test/Headers/lasxintrin.c new file mode 100644 index 00..530b72a7e3eb73 --- /dev/null +++ b/clang/test/Headers/lasxintrin.c @@ -0,0 +1,6 @@ +// REQUIRES: loongarch-registered-target +// RUN: %clang_cc1 %s -fsyntax-only -triple loongarch64-linux -target-feature +lasx +// RUN: %clang_cc1 %s -fsyntax-only -triple loongarch64-linux -target-feature +lasx -flax-vector-conversions=none +// RUN: %clang_cc1 %s -fsyntax-only -triple loongarch64-linux -target-feature +lasx -flax-vector-conversions=none -fno-signed-char + +#include diff --git a/clang/test/Headers/lsxintrin.c b/clang/test/Headers/lsxintrin.c new file mode 100644 index 00..d9ac76b5543c12 --- /dev/null +++ b/clang/test/Headers/lsxintrin.c @@ -0,0 +1,6 @@ +// REQUIRES: loongarch-registered-target +// RUN: %clang_cc1 %s -fsyntax-only -triple loongarch64-linux -target-feature +lsx +// RUN: %clang_cc1 %s -fsyntax-only -triple loongarch64-linux -target-feature +lsx -flax-vector-conversions=none +// RUN: %clang_cc1 %s -fsyntax-only -triple loongarch64-linux -target-feature +lsx -flax-vector-conversions=none -fno-signed-char + +#include >From e97149fd58e55e8bdf2b538b53bf343b082cf17b Mon Sep 17 00:00:00 2001 From: Xi Ruoyao Date: Fri, 1 Nov 2024 00:07:17 +0800 Subject: [PATCH 2/2] [LoongArch][clang] Use `signed char` vectors instead of `char` vectors for LSX and LASX builtins `-flax-vector-conversions=none` does not allow an implicit conversion from `signed char` vector to `char` vector, and we cannot remove `signed` from `v16i8` or `v32i8` because doing so will break our expectation with `-fno-signed-char`. So to make lsxintrin.h and lasxintrin.h work fine with `-flax-vector-conversions=none`, we must use `signed char` instead of `char`. The change is just done via sed 's/V16c/V16Sc/g' -i BuiltinsLoongArchLSX.def sed 's/V32c/V32Sc/g' -i BuiltinsLoongArchLASX.def --- .../clang/Basic/BuiltinsLoongArchLASX.def | 142 +- .../clang/Basic/BuiltinsLoongArchLSX.def | 128 2 files changed, 135 insertions(+), 135 deletions(-) diff --git a/clang/include/clang/Basic/BuiltinsLoongArchLASX.def b/clang/include/clang/Basic/BuiltinsLoongArchLASX.def index 4cf51cc000f6fc..0d7c2df5c5c503 100644 --- a/clang/include/clang/Basic/BuiltinsLoongArchLASX.def +++ b/clang/include/clang/Basic/BuiltinsLoongArchLASX.def @@ -12,29 +12,29 @@ // //===--===// -TARGET_BUILTIN(__builtin_lasx_xvadd_b, "V32cV32cV32c", "nc", "lasx") +TARGET_BUILTIN(__builtin_lasx_xvadd_b, "V32ScV32ScV32Sc", "nc", "lasx") TARGET_BUILTIN(__builtin_lasx_xvadd_h, "V16sV16sV16s", "nc", "lasx") TARGET_BUILTIN(__builtin_lasx_xvadd_w, "V8iV8iV8i", "nc", "lasx") TARGET_BUILTIN(__builtin_lasx_xvadd_d, "V4LLiV4LLiV4LLi", "nc", "lasx") TARGET_BUILTIN(__builtin_lasx_xvadd_q, "V4LLiV4LLiV4LLi", "nc", "lasx") -TARGET_BUILTIN(__builtin_lasx_xvsub_b, "V32cV32cV32c", "nc", "lasx") +TARGET_BUILTIN(__builtin_lasx_xvsub_b, "V32ScV32ScV32Sc", "nc", "lasx") TARGET_BUILTIN(__builtin_lasx_xvsub_h, "V16sV16sV16s", "nc", "lasx") TARGET_BUILTIN(__builtin_lasx_xvsub_w, "V8iV8iV8i", "nc", "lasx") TARGET_BUILTIN(__builtin_lasx_xvsub_d, "V4LLiV4LLiV4LLi", "nc", "lasx") TARGET_BUILTIN(__builtin_lasx_xvsub_q, "V4LLiV4LLiV4LLi", "nc", "lasx") -TARGET_BUILTIN(__builtin_lasx_xvaddi_bu, "V32cV32cIUi", "nc", "lasx") +TARGET_BUILTIN(__builtin_lasx_xvaddi_bu, "V32ScV32ScIUi", "nc", "lasx") TARGET_BUILTIN(__builtin_lasx_xvaddi_hu, "V16sV16sIUi", "nc", "lasx") TARGET_BUILTIN(__builtin_lasx_xvaddi_wu, "V8iV8iIUi", "nc", "lasx") TARGET_BUILTIN(__builtin_lasx_xvaddi_du, "V4LLiV4LLiIUi", "nc", "lasx") -TARGET_BUILTIN(__builtin_lasx_xvsubi_bu, "V32cV32cIUi", "nc", "lasx") +TARGET_BUILTIN(__builtin_lasx_xvsubi_bu, "V32ScV32ScIUi", "nc", "lasx") TARGET_BUILTIN(__builtin_lasx_xvsubi_hu, "V1
[clang] [LoongArch][Clang] Make the parameters and return value of {x,}vxor.v builti ns `unsigned char` vectors (PR #114513)
https://github.com/xry111 created https://github.com/llvm/llvm-project/pull/114513 The lsxintrin.h and and lasxintrin.h headers uses `unsigned char` vectors instead of `signed char` vectors. GCC also uses `unsigned char` for them, so align their definition with the headers and GCC. Fixes #110834. Depends on #114512. >From 1bbbc1c38715f42e6014ee319d76af9a4015e4d5 Mon Sep 17 00:00:00 2001 From: Xi Ruoyao Date: Fri, 1 Nov 2024 00:00:14 +0800 Subject: [PATCH 1/5] [LoongArch][Clang] Add tests for #110834 --- clang/test/Headers/lasxintrin.c | 6 ++ clang/test/Headers/lsxintrin.c | 6 ++ 2 files changed, 12 insertions(+) create mode 100644 clang/test/Headers/lasxintrin.c create mode 100644 clang/test/Headers/lsxintrin.c diff --git a/clang/test/Headers/lasxintrin.c b/clang/test/Headers/lasxintrin.c new file mode 100644 index 00..530b72a7e3eb73 --- /dev/null +++ b/clang/test/Headers/lasxintrin.c @@ -0,0 +1,6 @@ +// REQUIRES: loongarch-registered-target +// RUN: %clang_cc1 %s -fsyntax-only -triple loongarch64-linux -target-feature +lasx +// RUN: %clang_cc1 %s -fsyntax-only -triple loongarch64-linux -target-feature +lasx -flax-vector-conversions=none +// RUN: %clang_cc1 %s -fsyntax-only -triple loongarch64-linux -target-feature +lasx -flax-vector-conversions=none -fno-signed-char + +#include diff --git a/clang/test/Headers/lsxintrin.c b/clang/test/Headers/lsxintrin.c new file mode 100644 index 00..d9ac76b5543c12 --- /dev/null +++ b/clang/test/Headers/lsxintrin.c @@ -0,0 +1,6 @@ +// REQUIRES: loongarch-registered-target +// RUN: %clang_cc1 %s -fsyntax-only -triple loongarch64-linux -target-feature +lsx +// RUN: %clang_cc1 %s -fsyntax-only -triple loongarch64-linux -target-feature +lsx -flax-vector-conversions=none +// RUN: %clang_cc1 %s -fsyntax-only -triple loongarch64-linux -target-feature +lsx -flax-vector-conversions=none -fno-signed-char + +#include >From e97149fd58e55e8bdf2b538b53bf343b082cf17b Mon Sep 17 00:00:00 2001 From: Xi Ruoyao Date: Fri, 1 Nov 2024 00:07:17 +0800 Subject: [PATCH 2/5] [LoongArch][clang] Use `signed char` vectors instead of `char` vectors for LSX and LASX builtins `-flax-vector-conversions=none` does not allow an implicit conversion from `signed char` vector to `char` vector, and we cannot remove `signed` from `v16i8` or `v32i8` because doing so will break our expectation with `-fno-signed-char`. So to make lsxintrin.h and lasxintrin.h work fine with `-flax-vector-conversions=none`, we must use `signed char` instead of `char`. The change is just done via sed 's/V16c/V16Sc/g' -i BuiltinsLoongArchLSX.def sed 's/V32c/V32Sc/g' -i BuiltinsLoongArchLASX.def --- .../clang/Basic/BuiltinsLoongArchLASX.def | 142 +- .../clang/Basic/BuiltinsLoongArchLSX.def | 128 2 files changed, 135 insertions(+), 135 deletions(-) diff --git a/clang/include/clang/Basic/BuiltinsLoongArchLASX.def b/clang/include/clang/Basic/BuiltinsLoongArchLASX.def index 4cf51cc000f6fc..0d7c2df5c5c503 100644 --- a/clang/include/clang/Basic/BuiltinsLoongArchLASX.def +++ b/clang/include/clang/Basic/BuiltinsLoongArchLASX.def @@ -12,29 +12,29 @@ // //===--===// -TARGET_BUILTIN(__builtin_lasx_xvadd_b, "V32cV32cV32c", "nc", "lasx") +TARGET_BUILTIN(__builtin_lasx_xvadd_b, "V32ScV32ScV32Sc", "nc", "lasx") TARGET_BUILTIN(__builtin_lasx_xvadd_h, "V16sV16sV16s", "nc", "lasx") TARGET_BUILTIN(__builtin_lasx_xvadd_w, "V8iV8iV8i", "nc", "lasx") TARGET_BUILTIN(__builtin_lasx_xvadd_d, "V4LLiV4LLiV4LLi", "nc", "lasx") TARGET_BUILTIN(__builtin_lasx_xvadd_q, "V4LLiV4LLiV4LLi", "nc", "lasx") -TARGET_BUILTIN(__builtin_lasx_xvsub_b, "V32cV32cV32c", "nc", "lasx") +TARGET_BUILTIN(__builtin_lasx_xvsub_b, "V32ScV32ScV32Sc", "nc", "lasx") TARGET_BUILTIN(__builtin_lasx_xvsub_h, "V16sV16sV16s", "nc", "lasx") TARGET_BUILTIN(__builtin_lasx_xvsub_w, "V8iV8iV8i", "nc", "lasx") TARGET_BUILTIN(__builtin_lasx_xvsub_d, "V4LLiV4LLiV4LLi", "nc", "lasx") TARGET_BUILTIN(__builtin_lasx_xvsub_q, "V4LLiV4LLiV4LLi", "nc", "lasx") -TARGET_BUILTIN(__builtin_lasx_xvaddi_bu, "V32cV32cIUi", "nc", "lasx") +TARGET_BUILTIN(__builtin_lasx_xvaddi_bu, "V32ScV32ScIUi", "nc", "lasx") TARGET_BUILTIN(__builtin_lasx_xvaddi_hu, "V16sV16sIUi", "nc", "lasx") TARGET_BUILTIN(__builtin_lasx_xvaddi_wu, "V8iV8iIUi", "nc", "lasx") TARGET_BUILTIN(__builtin_lasx_xvaddi_du, "V4LLiV4LLiIUi", "nc", "lasx") -TARGET_BUILTIN(__builtin_lasx_xvsubi_bu, "V32cV32cIUi", "nc", "lasx") +TARGET_BUILTIN(__builtin_lasx_xvsubi_bu, "V32ScV32ScIUi", "nc", "lasx") TARGET_BUILTIN(__builtin_lasx_xvsubi_hu, "V16sV16sIUi", "nc", "lasx") TARGET_BUILTIN(__builtin_lasx_xvsubi_wu, "V8iV8iIUi", "nc", "lasx") TARGET_BUILTIN(__builtin_lasx_xvsubi_du, "V4LLiV4LLiIUi", "nc", "lasx") -TARGET_BUILTIN(__builtin_lasx_xvneg_b, "V32cV32c", "nc", "lasx") +TARGET_BUILTIN(__builtin_lasx_xvneg_b, "V32ScV32Sc", "nc", "lasx") TARGET_BUILT
[clang] [LoongArch][clang] Use `signed char` vectors instead of `char` vectors for LSX and LASX builtins (PR #114510)
https://github.com/xry111 edited https://github.com/llvm/llvm-project/pull/114510 ___ cfe-commits mailing list cfe-commits@lists.llvm.org https://lists.llvm.org/cgi-bin/mailman/listinfo/cfe-commits
[clang] [LoongArch][Clang] Add tests for #110834 (PR #114509)
https://github.com/xry111 updated https://github.com/llvm/llvm-project/pull/114509 >From dd6a7a7d8446281b948c75b10dd40bacbb23daa8 Mon Sep 17 00:00:00 2001 From: Xi Ruoyao Date: Fri, 1 Nov 2024 00:00:14 +0800 Subject: [PATCH] [LoongArch][Clang] Add tests for #110834 --- clang/test/Headers/lasxintrin.c | 6 ++ clang/test/Headers/lsxintrin.c | 6 ++ 2 files changed, 12 insertions(+) create mode 100644 clang/test/Headers/lasxintrin.c create mode 100644 clang/test/Headers/lsxintrin.c diff --git a/clang/test/Headers/lasxintrin.c b/clang/test/Headers/lasxintrin.c new file mode 100644 index 00..6126c6350455c1 --- /dev/null +++ b/clang/test/Headers/lasxintrin.c @@ -0,0 +1,6 @@ +// RUN: %clang_cc1 %s -fsyntax-only -triple loongarch64 -target-feature +lasx +// RUN: not %clang_cc1 %s -fsyntax-only -triple loongarch64 -target-feature +lasx -flax-vector-conversions=none +// RUN: not %clang_cc1 %s -fsyntax-only -triple loongarch64 -target-feature +lasx -flax-vector-conversions=none -fno-signed-char +// FIXME: "not" should be removed once we fix GH#110834. + +#include diff --git a/clang/test/Headers/lsxintrin.c b/clang/test/Headers/lsxintrin.c new file mode 100644 index 00..930d3efe62e543 --- /dev/null +++ b/clang/test/Headers/lsxintrin.c @@ -0,0 +1,6 @@ +// RUN: %clang_cc1 %s -fsyntax-only -triple loongarch64 -target-feature +lsx +// RUN: not %clang_cc1 %s -fsyntax-only -triple loongarch64 -target-feature +lsx -flax-vector-conversions=none +// RUN: not %clang_cc1 %s -fsyntax-only -triple loongarch64 -target-feature +lsx -flax-vector-conversions=none -fno-signed-char +// FIXME: "not" should be removed once we fix GH#110834. + +#include ___ cfe-commits mailing list cfe-commits@lists.llvm.org https://lists.llvm.org/cgi-bin/mailman/listinfo/cfe-commits
[clang] [LoongArch][Clang] Add tests for #110834 (PR #114509)
@@ -0,0 +1,6 @@ +// REQUIRES: loongarch-registered-target xry111 wrote: Dropped. https://github.com/llvm/llvm-project/pull/114509 ___ cfe-commits mailing list cfe-commits@lists.llvm.org https://lists.llvm.org/cgi-bin/mailman/listinfo/cfe-commits
[clang] [LoongArch][Clang] Add tests for #110834 (PR #114509)
@@ -0,0 +1,6 @@ +// REQUIRES: loongarch-registered-target +// RUN: %clang_cc1 %s -fsyntax-only -triple loongarch64-linux -target-feature +lasx xry111 wrote: Done. https://github.com/llvm/llvm-project/pull/114509 ___ cfe-commits mailing list cfe-commits@lists.llvm.org https://lists.llvm.org/cgi-bin/mailman/listinfo/cfe-commits
[clang] [LoongArch][Clang] Add tests for #110834 (PR #114509)
@@ -0,0 +1,6 @@ +// REQUIRES: loongarch-registered-target +// RUN: %clang_cc1 %s -fsyntax-only -triple loongarch64-linux -target-feature +lasx xry111 wrote: -linux dropped. I just mistakenly thought a triple had to have the os field. https://github.com/llvm/llvm-project/pull/114509 ___ cfe-commits mailing list cfe-commits@lists.llvm.org https://lists.llvm.org/cgi-bin/mailman/listinfo/cfe-commits
[clang] [LoongArch][clang] Use `signed char` vectors instead of `char` vectors for LSX and LASX builtins (PR #114510)
https://github.com/xry111 updated https://github.com/llvm/llvm-project/pull/114510 >From dd6a7a7d8446281b948c75b10dd40bacbb23daa8 Mon Sep 17 00:00:00 2001 From: Xi Ruoyao Date: Fri, 1 Nov 2024 00:00:14 +0800 Subject: [PATCH 1/2] [LoongArch][Clang] Add tests for #110834 --- clang/test/Headers/lasxintrin.c | 6 ++ clang/test/Headers/lsxintrin.c | 6 ++ 2 files changed, 12 insertions(+) create mode 100644 clang/test/Headers/lasxintrin.c create mode 100644 clang/test/Headers/lsxintrin.c diff --git a/clang/test/Headers/lasxintrin.c b/clang/test/Headers/lasxintrin.c new file mode 100644 index 00..6126c6350455c1 --- /dev/null +++ b/clang/test/Headers/lasxintrin.c @@ -0,0 +1,6 @@ +// RUN: %clang_cc1 %s -fsyntax-only -triple loongarch64 -target-feature +lasx +// RUN: not %clang_cc1 %s -fsyntax-only -triple loongarch64 -target-feature +lasx -flax-vector-conversions=none +// RUN: not %clang_cc1 %s -fsyntax-only -triple loongarch64 -target-feature +lasx -flax-vector-conversions=none -fno-signed-char +// FIXME: "not" should be removed once we fix GH#110834. + +#include diff --git a/clang/test/Headers/lsxintrin.c b/clang/test/Headers/lsxintrin.c new file mode 100644 index 00..930d3efe62e543 --- /dev/null +++ b/clang/test/Headers/lsxintrin.c @@ -0,0 +1,6 @@ +// RUN: %clang_cc1 %s -fsyntax-only -triple loongarch64 -target-feature +lsx +// RUN: not %clang_cc1 %s -fsyntax-only -triple loongarch64 -target-feature +lsx -flax-vector-conversions=none +// RUN: not %clang_cc1 %s -fsyntax-only -triple loongarch64 -target-feature +lsx -flax-vector-conversions=none -fno-signed-char +// FIXME: "not" should be removed once we fix GH#110834. + +#include >From b25a14535dc7b8b15d14a74da66299400e58409b Mon Sep 17 00:00:00 2001 From: Xi Ruoyao Date: Fri, 1 Nov 2024 00:07:17 +0800 Subject: [PATCH 2/2] [LoongArch][clang] Use `signed char` vectors instead of `char` vectors for LSX and LASX builtins `-flax-vector-conversions=none` does not allow an implicit conversion from `signed char` vector to `char` vector, and we cannot remove `signed` from `v16i8` or `v32i8` because doing so will break our expectation with `-fno-signed-char`. So to make lsxintrin.h and lasxintrin.h work fine with `-flax-vector-conversions=none`, we must use `signed char` instead of `char`. The change is just done via sed 's/V16c/V16Sc/g' -i BuiltinsLoongArchLSX.def sed 's/V32c/V32Sc/g' -i BuiltinsLoongArchLASX.def --- .../clang/Basic/BuiltinsLoongArchLASX.def | 142 +- .../clang/Basic/BuiltinsLoongArchLSX.def | 128 2 files changed, 135 insertions(+), 135 deletions(-) diff --git a/clang/include/clang/Basic/BuiltinsLoongArchLASX.def b/clang/include/clang/Basic/BuiltinsLoongArchLASX.def index 4cf51cc000f6fc..0d7c2df5c5c503 100644 --- a/clang/include/clang/Basic/BuiltinsLoongArchLASX.def +++ b/clang/include/clang/Basic/BuiltinsLoongArchLASX.def @@ -12,29 +12,29 @@ // //===--===// -TARGET_BUILTIN(__builtin_lasx_xvadd_b, "V32cV32cV32c", "nc", "lasx") +TARGET_BUILTIN(__builtin_lasx_xvadd_b, "V32ScV32ScV32Sc", "nc", "lasx") TARGET_BUILTIN(__builtin_lasx_xvadd_h, "V16sV16sV16s", "nc", "lasx") TARGET_BUILTIN(__builtin_lasx_xvadd_w, "V8iV8iV8i", "nc", "lasx") TARGET_BUILTIN(__builtin_lasx_xvadd_d, "V4LLiV4LLiV4LLi", "nc", "lasx") TARGET_BUILTIN(__builtin_lasx_xvadd_q, "V4LLiV4LLiV4LLi", "nc", "lasx") -TARGET_BUILTIN(__builtin_lasx_xvsub_b, "V32cV32cV32c", "nc", "lasx") +TARGET_BUILTIN(__builtin_lasx_xvsub_b, "V32ScV32ScV32Sc", "nc", "lasx") TARGET_BUILTIN(__builtin_lasx_xvsub_h, "V16sV16sV16s", "nc", "lasx") TARGET_BUILTIN(__builtin_lasx_xvsub_w, "V8iV8iV8i", "nc", "lasx") TARGET_BUILTIN(__builtin_lasx_xvsub_d, "V4LLiV4LLiV4LLi", "nc", "lasx") TARGET_BUILTIN(__builtin_lasx_xvsub_q, "V4LLiV4LLiV4LLi", "nc", "lasx") -TARGET_BUILTIN(__builtin_lasx_xvaddi_bu, "V32cV32cIUi", "nc", "lasx") +TARGET_BUILTIN(__builtin_lasx_xvaddi_bu, "V32ScV32ScIUi", "nc", "lasx") TARGET_BUILTIN(__builtin_lasx_xvaddi_hu, "V16sV16sIUi", "nc", "lasx") TARGET_BUILTIN(__builtin_lasx_xvaddi_wu, "V8iV8iIUi", "nc", "lasx") TARGET_BUILTIN(__builtin_lasx_xvaddi_du, "V4LLiV4LLiIUi", "nc", "lasx") -TARGET_BUILTIN(__builtin_lasx_xvsubi_bu, "V32cV32cIUi", "nc", "lasx") +TARGET_BUILTIN(__builtin_lasx_xvsubi_bu, "V32ScV32ScIUi", "nc", "lasx") TARGET_BUILTIN(__builtin_lasx_xvsubi_hu, "V16sV16sIUi", "nc", "lasx") TARGET_BUILTIN(__builtin_lasx_xvsubi_wu, "V8iV8iIUi", "nc", "lasx") TARGET_BUILTIN(__builtin_lasx_xvsubi_du, "V4LLiV4LLiIUi", "nc", "lasx") -TARGET_BUILTIN(__builtin_lasx_xvneg_b, "V32cV32c", "nc", "lasx") +TARGET_BUILTIN(__builtin_lasx_xvneg_b, "V32ScV32Sc", "nc", "lasx") TARGET_BUILTIN(__builtin_lasx_xvneg_h, "V16sV16s", "nc", "lasx") TARGET_BUILTIN(__builtin_lasx_xvneg_w, "V8iV8i", "nc", "lasx") TARGET_BUILTIN(__builtin_lasx_xvneg_d, "V4LLiV4LLi", "nc", "lasx") @@ -79,22 +79,22 @@ TARGET_BUILTIN(__builtin_las
[clang] [LoongArch][Clang] Make the parameters and return value of {x,}vxor.v builti ns `unsigned char` vectors (PR #114513)
https://github.com/xry111 updated https://github.com/llvm/llvm-project/pull/114513 >From dd6a7a7d8446281b948c75b10dd40bacbb23daa8 Mon Sep 17 00:00:00 2001 From: Xi Ruoyao Date: Fri, 1 Nov 2024 00:00:14 +0800 Subject: [PATCH 1/5] [LoongArch][Clang] Add tests for #110834 --- clang/test/Headers/lasxintrin.c | 6 ++ clang/test/Headers/lsxintrin.c | 6 ++ 2 files changed, 12 insertions(+) create mode 100644 clang/test/Headers/lasxintrin.c create mode 100644 clang/test/Headers/lsxintrin.c diff --git a/clang/test/Headers/lasxintrin.c b/clang/test/Headers/lasxintrin.c new file mode 100644 index 00..6126c6350455c1 --- /dev/null +++ b/clang/test/Headers/lasxintrin.c @@ -0,0 +1,6 @@ +// RUN: %clang_cc1 %s -fsyntax-only -triple loongarch64 -target-feature +lasx +// RUN: not %clang_cc1 %s -fsyntax-only -triple loongarch64 -target-feature +lasx -flax-vector-conversions=none +// RUN: not %clang_cc1 %s -fsyntax-only -triple loongarch64 -target-feature +lasx -flax-vector-conversions=none -fno-signed-char +// FIXME: "not" should be removed once we fix GH#110834. + +#include diff --git a/clang/test/Headers/lsxintrin.c b/clang/test/Headers/lsxintrin.c new file mode 100644 index 00..930d3efe62e543 --- /dev/null +++ b/clang/test/Headers/lsxintrin.c @@ -0,0 +1,6 @@ +// RUN: %clang_cc1 %s -fsyntax-only -triple loongarch64 -target-feature +lsx +// RUN: not %clang_cc1 %s -fsyntax-only -triple loongarch64 -target-feature +lsx -flax-vector-conversions=none +// RUN: not %clang_cc1 %s -fsyntax-only -triple loongarch64 -target-feature +lsx -flax-vector-conversions=none -fno-signed-char +// FIXME: "not" should be removed once we fix GH#110834. + +#include >From b25a14535dc7b8b15d14a74da66299400e58409b Mon Sep 17 00:00:00 2001 From: Xi Ruoyao Date: Fri, 1 Nov 2024 00:07:17 +0800 Subject: [PATCH 2/5] [LoongArch][clang] Use `signed char` vectors instead of `char` vectors for LSX and LASX builtins `-flax-vector-conversions=none` does not allow an implicit conversion from `signed char` vector to `char` vector, and we cannot remove `signed` from `v16i8` or `v32i8` because doing so will break our expectation with `-fno-signed-char`. So to make lsxintrin.h and lasxintrin.h work fine with `-flax-vector-conversions=none`, we must use `signed char` instead of `char`. The change is just done via sed 's/V16c/V16Sc/g' -i BuiltinsLoongArchLSX.def sed 's/V32c/V32Sc/g' -i BuiltinsLoongArchLASX.def --- .../clang/Basic/BuiltinsLoongArchLASX.def | 142 +- .../clang/Basic/BuiltinsLoongArchLSX.def | 128 2 files changed, 135 insertions(+), 135 deletions(-) diff --git a/clang/include/clang/Basic/BuiltinsLoongArchLASX.def b/clang/include/clang/Basic/BuiltinsLoongArchLASX.def index 4cf51cc000f6fc..0d7c2df5c5c503 100644 --- a/clang/include/clang/Basic/BuiltinsLoongArchLASX.def +++ b/clang/include/clang/Basic/BuiltinsLoongArchLASX.def @@ -12,29 +12,29 @@ // //===--===// -TARGET_BUILTIN(__builtin_lasx_xvadd_b, "V32cV32cV32c", "nc", "lasx") +TARGET_BUILTIN(__builtin_lasx_xvadd_b, "V32ScV32ScV32Sc", "nc", "lasx") TARGET_BUILTIN(__builtin_lasx_xvadd_h, "V16sV16sV16s", "nc", "lasx") TARGET_BUILTIN(__builtin_lasx_xvadd_w, "V8iV8iV8i", "nc", "lasx") TARGET_BUILTIN(__builtin_lasx_xvadd_d, "V4LLiV4LLiV4LLi", "nc", "lasx") TARGET_BUILTIN(__builtin_lasx_xvadd_q, "V4LLiV4LLiV4LLi", "nc", "lasx") -TARGET_BUILTIN(__builtin_lasx_xvsub_b, "V32cV32cV32c", "nc", "lasx") +TARGET_BUILTIN(__builtin_lasx_xvsub_b, "V32ScV32ScV32Sc", "nc", "lasx") TARGET_BUILTIN(__builtin_lasx_xvsub_h, "V16sV16sV16s", "nc", "lasx") TARGET_BUILTIN(__builtin_lasx_xvsub_w, "V8iV8iV8i", "nc", "lasx") TARGET_BUILTIN(__builtin_lasx_xvsub_d, "V4LLiV4LLiV4LLi", "nc", "lasx") TARGET_BUILTIN(__builtin_lasx_xvsub_q, "V4LLiV4LLiV4LLi", "nc", "lasx") -TARGET_BUILTIN(__builtin_lasx_xvaddi_bu, "V32cV32cIUi", "nc", "lasx") +TARGET_BUILTIN(__builtin_lasx_xvaddi_bu, "V32ScV32ScIUi", "nc", "lasx") TARGET_BUILTIN(__builtin_lasx_xvaddi_hu, "V16sV16sIUi", "nc", "lasx") TARGET_BUILTIN(__builtin_lasx_xvaddi_wu, "V8iV8iIUi", "nc", "lasx") TARGET_BUILTIN(__builtin_lasx_xvaddi_du, "V4LLiV4LLiIUi", "nc", "lasx") -TARGET_BUILTIN(__builtin_lasx_xvsubi_bu, "V32cV32cIUi", "nc", "lasx") +TARGET_BUILTIN(__builtin_lasx_xvsubi_bu, "V32ScV32ScIUi", "nc", "lasx") TARGET_BUILTIN(__builtin_lasx_xvsubi_hu, "V16sV16sIUi", "nc", "lasx") TARGET_BUILTIN(__builtin_lasx_xvsubi_wu, "V8iV8iIUi", "nc", "lasx") TARGET_BUILTIN(__builtin_lasx_xvsubi_du, "V4LLiV4LLiIUi", "nc", "lasx") -TARGET_BUILTIN(__builtin_lasx_xvneg_b, "V32cV32c", "nc", "lasx") +TARGET_BUILTIN(__builtin_lasx_xvneg_b, "V32ScV32Sc", "nc", "lasx") TARGET_BUILTIN(__builtin_lasx_xvneg_h, "V16sV16s", "nc", "lasx") TARGET_BUILTIN(__builtin_lasx_xvneg_w, "V8iV8i", "nc", "lasx") TARGET_BUILTIN(__builtin_lasx_xvneg_d, "V4LLiV4LLi", "nc", "lasx") @@ -79,22 +79,22 @@ TARGET_BUILTIN(__builtin_las
[clang] [LoongArch][Clang] Make the parameters and return value of {x,}vorn.v builti ns `unsigned char` vectors (PR #114514)
https://github.com/xry111 updated https://github.com/llvm/llvm-project/pull/114514 >From dd6a7a7d8446281b948c75b10dd40bacbb23daa8 Mon Sep 17 00:00:00 2001 From: Xi Ruoyao Date: Fri, 1 Nov 2024 00:00:14 +0800 Subject: [PATCH 1/6] [LoongArch][Clang] Add tests for #110834 --- clang/test/Headers/lasxintrin.c | 6 ++ clang/test/Headers/lsxintrin.c | 6 ++ 2 files changed, 12 insertions(+) create mode 100644 clang/test/Headers/lasxintrin.c create mode 100644 clang/test/Headers/lsxintrin.c diff --git a/clang/test/Headers/lasxintrin.c b/clang/test/Headers/lasxintrin.c new file mode 100644 index 00..6126c6350455c1 --- /dev/null +++ b/clang/test/Headers/lasxintrin.c @@ -0,0 +1,6 @@ +// RUN: %clang_cc1 %s -fsyntax-only -triple loongarch64 -target-feature +lasx +// RUN: not %clang_cc1 %s -fsyntax-only -triple loongarch64 -target-feature +lasx -flax-vector-conversions=none +// RUN: not %clang_cc1 %s -fsyntax-only -triple loongarch64 -target-feature +lasx -flax-vector-conversions=none -fno-signed-char +// FIXME: "not" should be removed once we fix GH#110834. + +#include diff --git a/clang/test/Headers/lsxintrin.c b/clang/test/Headers/lsxintrin.c new file mode 100644 index 00..930d3efe62e543 --- /dev/null +++ b/clang/test/Headers/lsxintrin.c @@ -0,0 +1,6 @@ +// RUN: %clang_cc1 %s -fsyntax-only -triple loongarch64 -target-feature +lsx +// RUN: not %clang_cc1 %s -fsyntax-only -triple loongarch64 -target-feature +lsx -flax-vector-conversions=none +// RUN: not %clang_cc1 %s -fsyntax-only -triple loongarch64 -target-feature +lsx -flax-vector-conversions=none -fno-signed-char +// FIXME: "not" should be removed once we fix GH#110834. + +#include >From b25a14535dc7b8b15d14a74da66299400e58409b Mon Sep 17 00:00:00 2001 From: Xi Ruoyao Date: Fri, 1 Nov 2024 00:07:17 +0800 Subject: [PATCH 2/6] [LoongArch][clang] Use `signed char` vectors instead of `char` vectors for LSX and LASX builtins `-flax-vector-conversions=none` does not allow an implicit conversion from `signed char` vector to `char` vector, and we cannot remove `signed` from `v16i8` or `v32i8` because doing so will break our expectation with `-fno-signed-char`. So to make lsxintrin.h and lasxintrin.h work fine with `-flax-vector-conversions=none`, we must use `signed char` instead of `char`. The change is just done via sed 's/V16c/V16Sc/g' -i BuiltinsLoongArchLSX.def sed 's/V32c/V32Sc/g' -i BuiltinsLoongArchLASX.def --- .../clang/Basic/BuiltinsLoongArchLASX.def | 142 +- .../clang/Basic/BuiltinsLoongArchLSX.def | 128 2 files changed, 135 insertions(+), 135 deletions(-) diff --git a/clang/include/clang/Basic/BuiltinsLoongArchLASX.def b/clang/include/clang/Basic/BuiltinsLoongArchLASX.def index 4cf51cc000f6fc..0d7c2df5c5c503 100644 --- a/clang/include/clang/Basic/BuiltinsLoongArchLASX.def +++ b/clang/include/clang/Basic/BuiltinsLoongArchLASX.def @@ -12,29 +12,29 @@ // //===--===// -TARGET_BUILTIN(__builtin_lasx_xvadd_b, "V32cV32cV32c", "nc", "lasx") +TARGET_BUILTIN(__builtin_lasx_xvadd_b, "V32ScV32ScV32Sc", "nc", "lasx") TARGET_BUILTIN(__builtin_lasx_xvadd_h, "V16sV16sV16s", "nc", "lasx") TARGET_BUILTIN(__builtin_lasx_xvadd_w, "V8iV8iV8i", "nc", "lasx") TARGET_BUILTIN(__builtin_lasx_xvadd_d, "V4LLiV4LLiV4LLi", "nc", "lasx") TARGET_BUILTIN(__builtin_lasx_xvadd_q, "V4LLiV4LLiV4LLi", "nc", "lasx") -TARGET_BUILTIN(__builtin_lasx_xvsub_b, "V32cV32cV32c", "nc", "lasx") +TARGET_BUILTIN(__builtin_lasx_xvsub_b, "V32ScV32ScV32Sc", "nc", "lasx") TARGET_BUILTIN(__builtin_lasx_xvsub_h, "V16sV16sV16s", "nc", "lasx") TARGET_BUILTIN(__builtin_lasx_xvsub_w, "V8iV8iV8i", "nc", "lasx") TARGET_BUILTIN(__builtin_lasx_xvsub_d, "V4LLiV4LLiV4LLi", "nc", "lasx") TARGET_BUILTIN(__builtin_lasx_xvsub_q, "V4LLiV4LLiV4LLi", "nc", "lasx") -TARGET_BUILTIN(__builtin_lasx_xvaddi_bu, "V32cV32cIUi", "nc", "lasx") +TARGET_BUILTIN(__builtin_lasx_xvaddi_bu, "V32ScV32ScIUi", "nc", "lasx") TARGET_BUILTIN(__builtin_lasx_xvaddi_hu, "V16sV16sIUi", "nc", "lasx") TARGET_BUILTIN(__builtin_lasx_xvaddi_wu, "V8iV8iIUi", "nc", "lasx") TARGET_BUILTIN(__builtin_lasx_xvaddi_du, "V4LLiV4LLiIUi", "nc", "lasx") -TARGET_BUILTIN(__builtin_lasx_xvsubi_bu, "V32cV32cIUi", "nc", "lasx") +TARGET_BUILTIN(__builtin_lasx_xvsubi_bu, "V32ScV32ScIUi", "nc", "lasx") TARGET_BUILTIN(__builtin_lasx_xvsubi_hu, "V16sV16sIUi", "nc", "lasx") TARGET_BUILTIN(__builtin_lasx_xvsubi_wu, "V8iV8iIUi", "nc", "lasx") TARGET_BUILTIN(__builtin_lasx_xvsubi_du, "V4LLiV4LLiIUi", "nc", "lasx") -TARGET_BUILTIN(__builtin_lasx_xvneg_b, "V32cV32c", "nc", "lasx") +TARGET_BUILTIN(__builtin_lasx_xvneg_b, "V32ScV32Sc", "nc", "lasx") TARGET_BUILTIN(__builtin_lasx_xvneg_h, "V16sV16s", "nc", "lasx") TARGET_BUILTIN(__builtin_lasx_xvneg_w, "V8iV8i", "nc", "lasx") TARGET_BUILTIN(__builtin_lasx_xvneg_d, "V4LLiV4LLi", "nc", "lasx") @@ -79,22 +79,22 @@ TARGET_BUILTIN(__builtin_las
[clang] [LoongArch][Clang] Make the parameter and return value of {x,}vmsknz.b builtins `signed char` vector (PR #114511)
https://github.com/xry111 updated https://github.com/llvm/llvm-project/pull/114511 >From dd6a7a7d8446281b948c75b10dd40bacbb23daa8 Mon Sep 17 00:00:00 2001 From: Xi Ruoyao Date: Fri, 1 Nov 2024 00:00:14 +0800 Subject: [PATCH 1/3] [LoongArch][Clang] Add tests for #110834 --- clang/test/Headers/lasxintrin.c | 6 ++ clang/test/Headers/lsxintrin.c | 6 ++ 2 files changed, 12 insertions(+) create mode 100644 clang/test/Headers/lasxintrin.c create mode 100644 clang/test/Headers/lsxintrin.c diff --git a/clang/test/Headers/lasxintrin.c b/clang/test/Headers/lasxintrin.c new file mode 100644 index 00..6126c6350455c1 --- /dev/null +++ b/clang/test/Headers/lasxintrin.c @@ -0,0 +1,6 @@ +// RUN: %clang_cc1 %s -fsyntax-only -triple loongarch64 -target-feature +lasx +// RUN: not %clang_cc1 %s -fsyntax-only -triple loongarch64 -target-feature +lasx -flax-vector-conversions=none +// RUN: not %clang_cc1 %s -fsyntax-only -triple loongarch64 -target-feature +lasx -flax-vector-conversions=none -fno-signed-char +// FIXME: "not" should be removed once we fix GH#110834. + +#include diff --git a/clang/test/Headers/lsxintrin.c b/clang/test/Headers/lsxintrin.c new file mode 100644 index 00..930d3efe62e543 --- /dev/null +++ b/clang/test/Headers/lsxintrin.c @@ -0,0 +1,6 @@ +// RUN: %clang_cc1 %s -fsyntax-only -triple loongarch64 -target-feature +lsx +// RUN: not %clang_cc1 %s -fsyntax-only -triple loongarch64 -target-feature +lsx -flax-vector-conversions=none +// RUN: not %clang_cc1 %s -fsyntax-only -triple loongarch64 -target-feature +lsx -flax-vector-conversions=none -fno-signed-char +// FIXME: "not" should be removed once we fix GH#110834. + +#include >From b25a14535dc7b8b15d14a74da66299400e58409b Mon Sep 17 00:00:00 2001 From: Xi Ruoyao Date: Fri, 1 Nov 2024 00:07:17 +0800 Subject: [PATCH 2/3] [LoongArch][clang] Use `signed char` vectors instead of `char` vectors for LSX and LASX builtins `-flax-vector-conversions=none` does not allow an implicit conversion from `signed char` vector to `char` vector, and we cannot remove `signed` from `v16i8` or `v32i8` because doing so will break our expectation with `-fno-signed-char`. So to make lsxintrin.h and lasxintrin.h work fine with `-flax-vector-conversions=none`, we must use `signed char` instead of `char`. The change is just done via sed 's/V16c/V16Sc/g' -i BuiltinsLoongArchLSX.def sed 's/V32c/V32Sc/g' -i BuiltinsLoongArchLASX.def --- .../clang/Basic/BuiltinsLoongArchLASX.def | 142 +- .../clang/Basic/BuiltinsLoongArchLSX.def | 128 2 files changed, 135 insertions(+), 135 deletions(-) diff --git a/clang/include/clang/Basic/BuiltinsLoongArchLASX.def b/clang/include/clang/Basic/BuiltinsLoongArchLASX.def index 4cf51cc000f6fc..0d7c2df5c5c503 100644 --- a/clang/include/clang/Basic/BuiltinsLoongArchLASX.def +++ b/clang/include/clang/Basic/BuiltinsLoongArchLASX.def @@ -12,29 +12,29 @@ // //===--===// -TARGET_BUILTIN(__builtin_lasx_xvadd_b, "V32cV32cV32c", "nc", "lasx") +TARGET_BUILTIN(__builtin_lasx_xvadd_b, "V32ScV32ScV32Sc", "nc", "lasx") TARGET_BUILTIN(__builtin_lasx_xvadd_h, "V16sV16sV16s", "nc", "lasx") TARGET_BUILTIN(__builtin_lasx_xvadd_w, "V8iV8iV8i", "nc", "lasx") TARGET_BUILTIN(__builtin_lasx_xvadd_d, "V4LLiV4LLiV4LLi", "nc", "lasx") TARGET_BUILTIN(__builtin_lasx_xvadd_q, "V4LLiV4LLiV4LLi", "nc", "lasx") -TARGET_BUILTIN(__builtin_lasx_xvsub_b, "V32cV32cV32c", "nc", "lasx") +TARGET_BUILTIN(__builtin_lasx_xvsub_b, "V32ScV32ScV32Sc", "nc", "lasx") TARGET_BUILTIN(__builtin_lasx_xvsub_h, "V16sV16sV16s", "nc", "lasx") TARGET_BUILTIN(__builtin_lasx_xvsub_w, "V8iV8iV8i", "nc", "lasx") TARGET_BUILTIN(__builtin_lasx_xvsub_d, "V4LLiV4LLiV4LLi", "nc", "lasx") TARGET_BUILTIN(__builtin_lasx_xvsub_q, "V4LLiV4LLiV4LLi", "nc", "lasx") -TARGET_BUILTIN(__builtin_lasx_xvaddi_bu, "V32cV32cIUi", "nc", "lasx") +TARGET_BUILTIN(__builtin_lasx_xvaddi_bu, "V32ScV32ScIUi", "nc", "lasx") TARGET_BUILTIN(__builtin_lasx_xvaddi_hu, "V16sV16sIUi", "nc", "lasx") TARGET_BUILTIN(__builtin_lasx_xvaddi_wu, "V8iV8iIUi", "nc", "lasx") TARGET_BUILTIN(__builtin_lasx_xvaddi_du, "V4LLiV4LLiIUi", "nc", "lasx") -TARGET_BUILTIN(__builtin_lasx_xvsubi_bu, "V32cV32cIUi", "nc", "lasx") +TARGET_BUILTIN(__builtin_lasx_xvsubi_bu, "V32ScV32ScIUi", "nc", "lasx") TARGET_BUILTIN(__builtin_lasx_xvsubi_hu, "V16sV16sIUi", "nc", "lasx") TARGET_BUILTIN(__builtin_lasx_xvsubi_wu, "V8iV8iIUi", "nc", "lasx") TARGET_BUILTIN(__builtin_lasx_xvsubi_du, "V4LLiV4LLiIUi", "nc", "lasx") -TARGET_BUILTIN(__builtin_lasx_xvneg_b, "V32cV32c", "nc", "lasx") +TARGET_BUILTIN(__builtin_lasx_xvneg_b, "V32ScV32Sc", "nc", "lasx") TARGET_BUILTIN(__builtin_lasx_xvneg_h, "V16sV16s", "nc", "lasx") TARGET_BUILTIN(__builtin_lasx_xvneg_w, "V8iV8i", "nc", "lasx") TARGET_BUILTIN(__builtin_lasx_xvneg_d, "V4LLiV4LLi", "nc", "lasx") @@ -79,22 +79,22 @@ TARGET_BUILTIN(__builtin_las
[clang] [LoongArch][Clang] Make the parameters and return value of {x,}vshuf.b builtins `signed char` vectors (PR #114512)
https://github.com/xry111 updated https://github.com/llvm/llvm-project/pull/114512 >From dd6a7a7d8446281b948c75b10dd40bacbb23daa8 Mon Sep 17 00:00:00 2001 From: Xi Ruoyao Date: Fri, 1 Nov 2024 00:00:14 +0800 Subject: [PATCH 1/4] [LoongArch][Clang] Add tests for #110834 --- clang/test/Headers/lasxintrin.c | 6 ++ clang/test/Headers/lsxintrin.c | 6 ++ 2 files changed, 12 insertions(+) create mode 100644 clang/test/Headers/lasxintrin.c create mode 100644 clang/test/Headers/lsxintrin.c diff --git a/clang/test/Headers/lasxintrin.c b/clang/test/Headers/lasxintrin.c new file mode 100644 index 00..6126c6350455c1 --- /dev/null +++ b/clang/test/Headers/lasxintrin.c @@ -0,0 +1,6 @@ +// RUN: %clang_cc1 %s -fsyntax-only -triple loongarch64 -target-feature +lasx +// RUN: not %clang_cc1 %s -fsyntax-only -triple loongarch64 -target-feature +lasx -flax-vector-conversions=none +// RUN: not %clang_cc1 %s -fsyntax-only -triple loongarch64 -target-feature +lasx -flax-vector-conversions=none -fno-signed-char +// FIXME: "not" should be removed once we fix GH#110834. + +#include diff --git a/clang/test/Headers/lsxintrin.c b/clang/test/Headers/lsxintrin.c new file mode 100644 index 00..930d3efe62e543 --- /dev/null +++ b/clang/test/Headers/lsxintrin.c @@ -0,0 +1,6 @@ +// RUN: %clang_cc1 %s -fsyntax-only -triple loongarch64 -target-feature +lsx +// RUN: not %clang_cc1 %s -fsyntax-only -triple loongarch64 -target-feature +lsx -flax-vector-conversions=none +// RUN: not %clang_cc1 %s -fsyntax-only -triple loongarch64 -target-feature +lsx -flax-vector-conversions=none -fno-signed-char +// FIXME: "not" should be removed once we fix GH#110834. + +#include >From b25a14535dc7b8b15d14a74da66299400e58409b Mon Sep 17 00:00:00 2001 From: Xi Ruoyao Date: Fri, 1 Nov 2024 00:07:17 +0800 Subject: [PATCH 2/4] [LoongArch][clang] Use `signed char` vectors instead of `char` vectors for LSX and LASX builtins `-flax-vector-conversions=none` does not allow an implicit conversion from `signed char` vector to `char` vector, and we cannot remove `signed` from `v16i8` or `v32i8` because doing so will break our expectation with `-fno-signed-char`. So to make lsxintrin.h and lasxintrin.h work fine with `-flax-vector-conversions=none`, we must use `signed char` instead of `char`. The change is just done via sed 's/V16c/V16Sc/g' -i BuiltinsLoongArchLSX.def sed 's/V32c/V32Sc/g' -i BuiltinsLoongArchLASX.def --- .../clang/Basic/BuiltinsLoongArchLASX.def | 142 +- .../clang/Basic/BuiltinsLoongArchLSX.def | 128 2 files changed, 135 insertions(+), 135 deletions(-) diff --git a/clang/include/clang/Basic/BuiltinsLoongArchLASX.def b/clang/include/clang/Basic/BuiltinsLoongArchLASX.def index 4cf51cc000f6fc..0d7c2df5c5c503 100644 --- a/clang/include/clang/Basic/BuiltinsLoongArchLASX.def +++ b/clang/include/clang/Basic/BuiltinsLoongArchLASX.def @@ -12,29 +12,29 @@ // //===--===// -TARGET_BUILTIN(__builtin_lasx_xvadd_b, "V32cV32cV32c", "nc", "lasx") +TARGET_BUILTIN(__builtin_lasx_xvadd_b, "V32ScV32ScV32Sc", "nc", "lasx") TARGET_BUILTIN(__builtin_lasx_xvadd_h, "V16sV16sV16s", "nc", "lasx") TARGET_BUILTIN(__builtin_lasx_xvadd_w, "V8iV8iV8i", "nc", "lasx") TARGET_BUILTIN(__builtin_lasx_xvadd_d, "V4LLiV4LLiV4LLi", "nc", "lasx") TARGET_BUILTIN(__builtin_lasx_xvadd_q, "V4LLiV4LLiV4LLi", "nc", "lasx") -TARGET_BUILTIN(__builtin_lasx_xvsub_b, "V32cV32cV32c", "nc", "lasx") +TARGET_BUILTIN(__builtin_lasx_xvsub_b, "V32ScV32ScV32Sc", "nc", "lasx") TARGET_BUILTIN(__builtin_lasx_xvsub_h, "V16sV16sV16s", "nc", "lasx") TARGET_BUILTIN(__builtin_lasx_xvsub_w, "V8iV8iV8i", "nc", "lasx") TARGET_BUILTIN(__builtin_lasx_xvsub_d, "V4LLiV4LLiV4LLi", "nc", "lasx") TARGET_BUILTIN(__builtin_lasx_xvsub_q, "V4LLiV4LLiV4LLi", "nc", "lasx") -TARGET_BUILTIN(__builtin_lasx_xvaddi_bu, "V32cV32cIUi", "nc", "lasx") +TARGET_BUILTIN(__builtin_lasx_xvaddi_bu, "V32ScV32ScIUi", "nc", "lasx") TARGET_BUILTIN(__builtin_lasx_xvaddi_hu, "V16sV16sIUi", "nc", "lasx") TARGET_BUILTIN(__builtin_lasx_xvaddi_wu, "V8iV8iIUi", "nc", "lasx") TARGET_BUILTIN(__builtin_lasx_xvaddi_du, "V4LLiV4LLiIUi", "nc", "lasx") -TARGET_BUILTIN(__builtin_lasx_xvsubi_bu, "V32cV32cIUi", "nc", "lasx") +TARGET_BUILTIN(__builtin_lasx_xvsubi_bu, "V32ScV32ScIUi", "nc", "lasx") TARGET_BUILTIN(__builtin_lasx_xvsubi_hu, "V16sV16sIUi", "nc", "lasx") TARGET_BUILTIN(__builtin_lasx_xvsubi_wu, "V8iV8iIUi", "nc", "lasx") TARGET_BUILTIN(__builtin_lasx_xvsubi_du, "V4LLiV4LLiIUi", "nc", "lasx") -TARGET_BUILTIN(__builtin_lasx_xvneg_b, "V32cV32c", "nc", "lasx") +TARGET_BUILTIN(__builtin_lasx_xvneg_b, "V32ScV32Sc", "nc", "lasx") TARGET_BUILTIN(__builtin_lasx_xvneg_h, "V16sV16s", "nc", "lasx") TARGET_BUILTIN(__builtin_lasx_xvneg_w, "V8iV8i", "nc", "lasx") TARGET_BUILTIN(__builtin_lasx_xvneg_d, "V4LLiV4LLi", "nc", "lasx") @@ -79,22 +79,22 @@ TARGET_BUILTIN(__builtin_las
[clang] [LoongArch][Clang] Add tests for #110834 (PR #114509)
https://github.com/xry111 created https://github.com/llvm/llvm-project/pull/114509 None >From 1bbbc1c38715f42e6014ee319d76af9a4015e4d5 Mon Sep 17 00:00:00 2001 From: Xi Ruoyao Date: Fri, 1 Nov 2024 00:00:14 +0800 Subject: [PATCH] [LoongArch][Clang] Add tests for #110834 --- clang/test/Headers/lasxintrin.c | 6 ++ clang/test/Headers/lsxintrin.c | 6 ++ 2 files changed, 12 insertions(+) create mode 100644 clang/test/Headers/lasxintrin.c create mode 100644 clang/test/Headers/lsxintrin.c diff --git a/clang/test/Headers/lasxintrin.c b/clang/test/Headers/lasxintrin.c new file mode 100644 index 00..530b72a7e3eb73 --- /dev/null +++ b/clang/test/Headers/lasxintrin.c @@ -0,0 +1,6 @@ +// REQUIRES: loongarch-registered-target +// RUN: %clang_cc1 %s -fsyntax-only -triple loongarch64-linux -target-feature +lasx +// RUN: %clang_cc1 %s -fsyntax-only -triple loongarch64-linux -target-feature +lasx -flax-vector-conversions=none +// RUN: %clang_cc1 %s -fsyntax-only -triple loongarch64-linux -target-feature +lasx -flax-vector-conversions=none -fno-signed-char + +#include diff --git a/clang/test/Headers/lsxintrin.c b/clang/test/Headers/lsxintrin.c new file mode 100644 index 00..d9ac76b5543c12 --- /dev/null +++ b/clang/test/Headers/lsxintrin.c @@ -0,0 +1,6 @@ +// REQUIRES: loongarch-registered-target +// RUN: %clang_cc1 %s -fsyntax-only -triple loongarch64-linux -target-feature +lsx +// RUN: %clang_cc1 %s -fsyntax-only -triple loongarch64-linux -target-feature +lsx -flax-vector-conversions=none +// RUN: %clang_cc1 %s -fsyntax-only -triple loongarch64-linux -target-feature +lsx -flax-vector-conversions=none -fno-signed-char + +#include ___ cfe-commits mailing list cfe-commits@lists.llvm.org https://lists.llvm.org/cgi-bin/mailman/listinfo/cfe-commits
[clang] [LoongArch][Clang] Make the parameters and return value of {x,}vshuf.b builtins `signed char` vectors (PR #114512)
https://github.com/xry111 created https://github.com/llvm/llvm-project/pull/114512 The lsxintrin.h and and lasxintrin.h headers uses `signed char` vectors instead of `unsigned char` vectors. GCC also uses `signed char` for them, so align their definition with the headers and GCC. Depends on #114511. Part of #110834 fix. >From 1bbbc1c38715f42e6014ee319d76af9a4015e4d5 Mon Sep 17 00:00:00 2001 From: Xi Ruoyao Date: Fri, 1 Nov 2024 00:00:14 +0800 Subject: [PATCH 1/4] [LoongArch][Clang] Add tests for #110834 --- clang/test/Headers/lasxintrin.c | 6 ++ clang/test/Headers/lsxintrin.c | 6 ++ 2 files changed, 12 insertions(+) create mode 100644 clang/test/Headers/lasxintrin.c create mode 100644 clang/test/Headers/lsxintrin.c diff --git a/clang/test/Headers/lasxintrin.c b/clang/test/Headers/lasxintrin.c new file mode 100644 index 00..530b72a7e3eb73 --- /dev/null +++ b/clang/test/Headers/lasxintrin.c @@ -0,0 +1,6 @@ +// REQUIRES: loongarch-registered-target +// RUN: %clang_cc1 %s -fsyntax-only -triple loongarch64-linux -target-feature +lasx +// RUN: %clang_cc1 %s -fsyntax-only -triple loongarch64-linux -target-feature +lasx -flax-vector-conversions=none +// RUN: %clang_cc1 %s -fsyntax-only -triple loongarch64-linux -target-feature +lasx -flax-vector-conversions=none -fno-signed-char + +#include diff --git a/clang/test/Headers/lsxintrin.c b/clang/test/Headers/lsxintrin.c new file mode 100644 index 00..d9ac76b5543c12 --- /dev/null +++ b/clang/test/Headers/lsxintrin.c @@ -0,0 +1,6 @@ +// REQUIRES: loongarch-registered-target +// RUN: %clang_cc1 %s -fsyntax-only -triple loongarch64-linux -target-feature +lsx +// RUN: %clang_cc1 %s -fsyntax-only -triple loongarch64-linux -target-feature +lsx -flax-vector-conversions=none +// RUN: %clang_cc1 %s -fsyntax-only -triple loongarch64-linux -target-feature +lsx -flax-vector-conversions=none -fno-signed-char + +#include >From e97149fd58e55e8bdf2b538b53bf343b082cf17b Mon Sep 17 00:00:00 2001 From: Xi Ruoyao Date: Fri, 1 Nov 2024 00:07:17 +0800 Subject: [PATCH 2/4] [LoongArch][clang] Use `signed char` vectors instead of `char` vectors for LSX and LASX builtins `-flax-vector-conversions=none` does not allow an implicit conversion from `signed char` vector to `char` vector, and we cannot remove `signed` from `v16i8` or `v32i8` because doing so will break our expectation with `-fno-signed-char`. So to make lsxintrin.h and lasxintrin.h work fine with `-flax-vector-conversions=none`, we must use `signed char` instead of `char`. The change is just done via sed 's/V16c/V16Sc/g' -i BuiltinsLoongArchLSX.def sed 's/V32c/V32Sc/g' -i BuiltinsLoongArchLASX.def --- .../clang/Basic/BuiltinsLoongArchLASX.def | 142 +- .../clang/Basic/BuiltinsLoongArchLSX.def | 128 2 files changed, 135 insertions(+), 135 deletions(-) diff --git a/clang/include/clang/Basic/BuiltinsLoongArchLASX.def b/clang/include/clang/Basic/BuiltinsLoongArchLASX.def index 4cf51cc000f6fc..0d7c2df5c5c503 100644 --- a/clang/include/clang/Basic/BuiltinsLoongArchLASX.def +++ b/clang/include/clang/Basic/BuiltinsLoongArchLASX.def @@ -12,29 +12,29 @@ // //===--===// -TARGET_BUILTIN(__builtin_lasx_xvadd_b, "V32cV32cV32c", "nc", "lasx") +TARGET_BUILTIN(__builtin_lasx_xvadd_b, "V32ScV32ScV32Sc", "nc", "lasx") TARGET_BUILTIN(__builtin_lasx_xvadd_h, "V16sV16sV16s", "nc", "lasx") TARGET_BUILTIN(__builtin_lasx_xvadd_w, "V8iV8iV8i", "nc", "lasx") TARGET_BUILTIN(__builtin_lasx_xvadd_d, "V4LLiV4LLiV4LLi", "nc", "lasx") TARGET_BUILTIN(__builtin_lasx_xvadd_q, "V4LLiV4LLiV4LLi", "nc", "lasx") -TARGET_BUILTIN(__builtin_lasx_xvsub_b, "V32cV32cV32c", "nc", "lasx") +TARGET_BUILTIN(__builtin_lasx_xvsub_b, "V32ScV32ScV32Sc", "nc", "lasx") TARGET_BUILTIN(__builtin_lasx_xvsub_h, "V16sV16sV16s", "nc", "lasx") TARGET_BUILTIN(__builtin_lasx_xvsub_w, "V8iV8iV8i", "nc", "lasx") TARGET_BUILTIN(__builtin_lasx_xvsub_d, "V4LLiV4LLiV4LLi", "nc", "lasx") TARGET_BUILTIN(__builtin_lasx_xvsub_q, "V4LLiV4LLiV4LLi", "nc", "lasx") -TARGET_BUILTIN(__builtin_lasx_xvaddi_bu, "V32cV32cIUi", "nc", "lasx") +TARGET_BUILTIN(__builtin_lasx_xvaddi_bu, "V32ScV32ScIUi", "nc", "lasx") TARGET_BUILTIN(__builtin_lasx_xvaddi_hu, "V16sV16sIUi", "nc", "lasx") TARGET_BUILTIN(__builtin_lasx_xvaddi_wu, "V8iV8iIUi", "nc", "lasx") TARGET_BUILTIN(__builtin_lasx_xvaddi_du, "V4LLiV4LLiIUi", "nc", "lasx") -TARGET_BUILTIN(__builtin_lasx_xvsubi_bu, "V32cV32cIUi", "nc", "lasx") +TARGET_BUILTIN(__builtin_lasx_xvsubi_bu, "V32ScV32ScIUi", "nc", "lasx") TARGET_BUILTIN(__builtin_lasx_xvsubi_hu, "V16sV16sIUi", "nc", "lasx") TARGET_BUILTIN(__builtin_lasx_xvsubi_wu, "V8iV8iIUi", "nc", "lasx") TARGET_BUILTIN(__builtin_lasx_xvsubi_du, "V4LLiV4LLiIUi", "nc", "lasx") -TARGET_BUILTIN(__builtin_lasx_xvneg_b, "V32cV32c", "nc", "lasx") +TARGET_BUILTIN(__builtin_lasx_xvneg_b, "V32ScV32Sc", "nc", "lasx") TARGET_BUILT
[clang] [LoongArch][Clang] Make the parameter and return value of {x,}vmsknz.b builtins `signed char` vector (PR #114511)
https://github.com/xry111 created https://github.com/llvm/llvm-project/pull/114511 These builtins operate on int8 vectors, not int16 vectors. So the old definition does not make any sense. Depends on #114510. Part of #110834 fix. >From 1bbbc1c38715f42e6014ee319d76af9a4015e4d5 Mon Sep 17 00:00:00 2001 From: Xi Ruoyao Date: Fri, 1 Nov 2024 00:00:14 +0800 Subject: [PATCH 1/3] [LoongArch][Clang] Add tests for #110834 --- clang/test/Headers/lasxintrin.c | 6 ++ clang/test/Headers/lsxintrin.c | 6 ++ 2 files changed, 12 insertions(+) create mode 100644 clang/test/Headers/lasxintrin.c create mode 100644 clang/test/Headers/lsxintrin.c diff --git a/clang/test/Headers/lasxintrin.c b/clang/test/Headers/lasxintrin.c new file mode 100644 index 00..530b72a7e3eb73 --- /dev/null +++ b/clang/test/Headers/lasxintrin.c @@ -0,0 +1,6 @@ +// REQUIRES: loongarch-registered-target +// RUN: %clang_cc1 %s -fsyntax-only -triple loongarch64-linux -target-feature +lasx +// RUN: %clang_cc1 %s -fsyntax-only -triple loongarch64-linux -target-feature +lasx -flax-vector-conversions=none +// RUN: %clang_cc1 %s -fsyntax-only -triple loongarch64-linux -target-feature +lasx -flax-vector-conversions=none -fno-signed-char + +#include diff --git a/clang/test/Headers/lsxintrin.c b/clang/test/Headers/lsxintrin.c new file mode 100644 index 00..d9ac76b5543c12 --- /dev/null +++ b/clang/test/Headers/lsxintrin.c @@ -0,0 +1,6 @@ +// REQUIRES: loongarch-registered-target +// RUN: %clang_cc1 %s -fsyntax-only -triple loongarch64-linux -target-feature +lsx +// RUN: %clang_cc1 %s -fsyntax-only -triple loongarch64-linux -target-feature +lsx -flax-vector-conversions=none +// RUN: %clang_cc1 %s -fsyntax-only -triple loongarch64-linux -target-feature +lsx -flax-vector-conversions=none -fno-signed-char + +#include >From e97149fd58e55e8bdf2b538b53bf343b082cf17b Mon Sep 17 00:00:00 2001 From: Xi Ruoyao Date: Fri, 1 Nov 2024 00:07:17 +0800 Subject: [PATCH 2/3] [LoongArch][clang] Use `signed char` vectors instead of `char` vectors for LSX and LASX builtins `-flax-vector-conversions=none` does not allow an implicit conversion from `signed char` vector to `char` vector, and we cannot remove `signed` from `v16i8` or `v32i8` because doing so will break our expectation with `-fno-signed-char`. So to make lsxintrin.h and lasxintrin.h work fine with `-flax-vector-conversions=none`, we must use `signed char` instead of `char`. The change is just done via sed 's/V16c/V16Sc/g' -i BuiltinsLoongArchLSX.def sed 's/V32c/V32Sc/g' -i BuiltinsLoongArchLASX.def --- .../clang/Basic/BuiltinsLoongArchLASX.def | 142 +- .../clang/Basic/BuiltinsLoongArchLSX.def | 128 2 files changed, 135 insertions(+), 135 deletions(-) diff --git a/clang/include/clang/Basic/BuiltinsLoongArchLASX.def b/clang/include/clang/Basic/BuiltinsLoongArchLASX.def index 4cf51cc000f6fc..0d7c2df5c5c503 100644 --- a/clang/include/clang/Basic/BuiltinsLoongArchLASX.def +++ b/clang/include/clang/Basic/BuiltinsLoongArchLASX.def @@ -12,29 +12,29 @@ // //===--===// -TARGET_BUILTIN(__builtin_lasx_xvadd_b, "V32cV32cV32c", "nc", "lasx") +TARGET_BUILTIN(__builtin_lasx_xvadd_b, "V32ScV32ScV32Sc", "nc", "lasx") TARGET_BUILTIN(__builtin_lasx_xvadd_h, "V16sV16sV16s", "nc", "lasx") TARGET_BUILTIN(__builtin_lasx_xvadd_w, "V8iV8iV8i", "nc", "lasx") TARGET_BUILTIN(__builtin_lasx_xvadd_d, "V4LLiV4LLiV4LLi", "nc", "lasx") TARGET_BUILTIN(__builtin_lasx_xvadd_q, "V4LLiV4LLiV4LLi", "nc", "lasx") -TARGET_BUILTIN(__builtin_lasx_xvsub_b, "V32cV32cV32c", "nc", "lasx") +TARGET_BUILTIN(__builtin_lasx_xvsub_b, "V32ScV32ScV32Sc", "nc", "lasx") TARGET_BUILTIN(__builtin_lasx_xvsub_h, "V16sV16sV16s", "nc", "lasx") TARGET_BUILTIN(__builtin_lasx_xvsub_w, "V8iV8iV8i", "nc", "lasx") TARGET_BUILTIN(__builtin_lasx_xvsub_d, "V4LLiV4LLiV4LLi", "nc", "lasx") TARGET_BUILTIN(__builtin_lasx_xvsub_q, "V4LLiV4LLiV4LLi", "nc", "lasx") -TARGET_BUILTIN(__builtin_lasx_xvaddi_bu, "V32cV32cIUi", "nc", "lasx") +TARGET_BUILTIN(__builtin_lasx_xvaddi_bu, "V32ScV32ScIUi", "nc", "lasx") TARGET_BUILTIN(__builtin_lasx_xvaddi_hu, "V16sV16sIUi", "nc", "lasx") TARGET_BUILTIN(__builtin_lasx_xvaddi_wu, "V8iV8iIUi", "nc", "lasx") TARGET_BUILTIN(__builtin_lasx_xvaddi_du, "V4LLiV4LLiIUi", "nc", "lasx") -TARGET_BUILTIN(__builtin_lasx_xvsubi_bu, "V32cV32cIUi", "nc", "lasx") +TARGET_BUILTIN(__builtin_lasx_xvsubi_bu, "V32ScV32ScIUi", "nc", "lasx") TARGET_BUILTIN(__builtin_lasx_xvsubi_hu, "V16sV16sIUi", "nc", "lasx") TARGET_BUILTIN(__builtin_lasx_xvsubi_wu, "V8iV8iIUi", "nc", "lasx") TARGET_BUILTIN(__builtin_lasx_xvsubi_du, "V4LLiV4LLiIUi", "nc", "lasx") -TARGET_BUILTIN(__builtin_lasx_xvneg_b, "V32cV32c", "nc", "lasx") +TARGET_BUILTIN(__builtin_lasx_xvneg_b, "V32ScV32Sc", "nc", "lasx") TARGET_BUILTIN(__builtin_lasx_xvneg_h, "V16sV16s", "nc", "lasx") TARGET_BUILTIN(__builtin_lasx_xvneg_w,
[clang] [LoongArch][Clang] Make the parameters and return value of {x,}vorn.v builti ns `unsigned char` vectors (PR #114514)
https://github.com/xry111 created https://github.com/llvm/llvm-project/pull/114514 All other bitwise vector builtins use `unsigned char` vector operands, so it does not make too much sense to use `signed char` for the orn operation alone. The same change has been also proposed for GCC: https://gcc.gnu.org/pipermail/gcc-patches/2024-October/667065.html. Depends on #114513. >From 1bbbc1c38715f42e6014ee319d76af9a4015e4d5 Mon Sep 17 00:00:00 2001 From: Xi Ruoyao Date: Fri, 1 Nov 2024 00:00:14 +0800 Subject: [PATCH 1/6] [LoongArch][Clang] Add tests for #110834 --- clang/test/Headers/lasxintrin.c | 6 ++ clang/test/Headers/lsxintrin.c | 6 ++ 2 files changed, 12 insertions(+) create mode 100644 clang/test/Headers/lasxintrin.c create mode 100644 clang/test/Headers/lsxintrin.c diff --git a/clang/test/Headers/lasxintrin.c b/clang/test/Headers/lasxintrin.c new file mode 100644 index 00..530b72a7e3eb73 --- /dev/null +++ b/clang/test/Headers/lasxintrin.c @@ -0,0 +1,6 @@ +// REQUIRES: loongarch-registered-target +// RUN: %clang_cc1 %s -fsyntax-only -triple loongarch64-linux -target-feature +lasx +// RUN: %clang_cc1 %s -fsyntax-only -triple loongarch64-linux -target-feature +lasx -flax-vector-conversions=none +// RUN: %clang_cc1 %s -fsyntax-only -triple loongarch64-linux -target-feature +lasx -flax-vector-conversions=none -fno-signed-char + +#include diff --git a/clang/test/Headers/lsxintrin.c b/clang/test/Headers/lsxintrin.c new file mode 100644 index 00..d9ac76b5543c12 --- /dev/null +++ b/clang/test/Headers/lsxintrin.c @@ -0,0 +1,6 @@ +// REQUIRES: loongarch-registered-target +// RUN: %clang_cc1 %s -fsyntax-only -triple loongarch64-linux -target-feature +lsx +// RUN: %clang_cc1 %s -fsyntax-only -triple loongarch64-linux -target-feature +lsx -flax-vector-conversions=none +// RUN: %clang_cc1 %s -fsyntax-only -triple loongarch64-linux -target-feature +lsx -flax-vector-conversions=none -fno-signed-char + +#include >From e97149fd58e55e8bdf2b538b53bf343b082cf17b Mon Sep 17 00:00:00 2001 From: Xi Ruoyao Date: Fri, 1 Nov 2024 00:07:17 +0800 Subject: [PATCH 2/6] [LoongArch][clang] Use `signed char` vectors instead of `char` vectors for LSX and LASX builtins `-flax-vector-conversions=none` does not allow an implicit conversion from `signed char` vector to `char` vector, and we cannot remove `signed` from `v16i8` or `v32i8` because doing so will break our expectation with `-fno-signed-char`. So to make lsxintrin.h and lasxintrin.h work fine with `-flax-vector-conversions=none`, we must use `signed char` instead of `char`. The change is just done via sed 's/V16c/V16Sc/g' -i BuiltinsLoongArchLSX.def sed 's/V32c/V32Sc/g' -i BuiltinsLoongArchLASX.def --- .../clang/Basic/BuiltinsLoongArchLASX.def | 142 +- .../clang/Basic/BuiltinsLoongArchLSX.def | 128 2 files changed, 135 insertions(+), 135 deletions(-) diff --git a/clang/include/clang/Basic/BuiltinsLoongArchLASX.def b/clang/include/clang/Basic/BuiltinsLoongArchLASX.def index 4cf51cc000f6fc..0d7c2df5c5c503 100644 --- a/clang/include/clang/Basic/BuiltinsLoongArchLASX.def +++ b/clang/include/clang/Basic/BuiltinsLoongArchLASX.def @@ -12,29 +12,29 @@ // //===--===// -TARGET_BUILTIN(__builtin_lasx_xvadd_b, "V32cV32cV32c", "nc", "lasx") +TARGET_BUILTIN(__builtin_lasx_xvadd_b, "V32ScV32ScV32Sc", "nc", "lasx") TARGET_BUILTIN(__builtin_lasx_xvadd_h, "V16sV16sV16s", "nc", "lasx") TARGET_BUILTIN(__builtin_lasx_xvadd_w, "V8iV8iV8i", "nc", "lasx") TARGET_BUILTIN(__builtin_lasx_xvadd_d, "V4LLiV4LLiV4LLi", "nc", "lasx") TARGET_BUILTIN(__builtin_lasx_xvadd_q, "V4LLiV4LLiV4LLi", "nc", "lasx") -TARGET_BUILTIN(__builtin_lasx_xvsub_b, "V32cV32cV32c", "nc", "lasx") +TARGET_BUILTIN(__builtin_lasx_xvsub_b, "V32ScV32ScV32Sc", "nc", "lasx") TARGET_BUILTIN(__builtin_lasx_xvsub_h, "V16sV16sV16s", "nc", "lasx") TARGET_BUILTIN(__builtin_lasx_xvsub_w, "V8iV8iV8i", "nc", "lasx") TARGET_BUILTIN(__builtin_lasx_xvsub_d, "V4LLiV4LLiV4LLi", "nc", "lasx") TARGET_BUILTIN(__builtin_lasx_xvsub_q, "V4LLiV4LLiV4LLi", "nc", "lasx") -TARGET_BUILTIN(__builtin_lasx_xvaddi_bu, "V32cV32cIUi", "nc", "lasx") +TARGET_BUILTIN(__builtin_lasx_xvaddi_bu, "V32ScV32ScIUi", "nc", "lasx") TARGET_BUILTIN(__builtin_lasx_xvaddi_hu, "V16sV16sIUi", "nc", "lasx") TARGET_BUILTIN(__builtin_lasx_xvaddi_wu, "V8iV8iIUi", "nc", "lasx") TARGET_BUILTIN(__builtin_lasx_xvaddi_du, "V4LLiV4LLiIUi", "nc", "lasx") -TARGET_BUILTIN(__builtin_lasx_xvsubi_bu, "V32cV32cIUi", "nc", "lasx") +TARGET_BUILTIN(__builtin_lasx_xvsubi_bu, "V32ScV32ScIUi", "nc", "lasx") TARGET_BUILTIN(__builtin_lasx_xvsubi_hu, "V16sV16sIUi", "nc", "lasx") TARGET_BUILTIN(__builtin_lasx_xvsubi_wu, "V8iV8iIUi", "nc", "lasx") TARGET_BUILTIN(__builtin_lasx_xvsubi_du, "V4LLiV4LLiIUi", "nc", "lasx") -TARGET_BUILTIN(__builtin_lasx_xvneg_b, "V32cV32c", "nc", "lasx") +TARGET_BUILTIN(__builti
[clang] [llvm] [Clang][LoongArch] Add inline asm support for the `q` constraint (PR #141037)
xry111 wrote:
I have a C test case:
```c
unsigned long
test (unsigned long x, unsigned long y)
{
register unsigned long ra asm ("ra");
asm("" : "+r"(ra));
unsigned long t = ra;
asm ("csrxchg %0, %1, 0" : "+r" (x) : "r" (t));
return x;
}
```
clang-20.1.4 errors out:
```
t.c:8:8: error: must not be $r0 or $r1
8 | asm ("csrxchg %0, %1, 0" : "+r" (x) : "r" (t));
|^
:1:16: note: instantiated into assembly here
1 | csrxchg $a0, $ra, 0
| ^
1 error generated.
```
And I suppose it should work after this PR and changing the second `"r"` to
`"q"`.
https://github.com/llvm/llvm-project/pull/141037
___
cfe-commits mailing list
cfe-commits@lists.llvm.org
https://lists.llvm.org/cgi-bin/mailman/listinfo/cfe-commits
