[llvm-branch-commits] [clang] release/20.x: [AVX10.2] Fix wrong intrinsic names after rename (#126390) (PR #126687)
via llvm-branch-commits
llvm-branch-commits at lists.llvm.org
Mon Feb 10 22:48:56 PST 2025
https://github.com/llvmbot created https://github.com/llvm/llvm-project/pull/126687
Backport 161cfc6f39bef8994eb944687033ebd3570196e8
Requested by: @phoebewang
>From 8322903232de1a8586471367187fec23f60bb98b Mon Sep 17 00:00:00 2001
From: =?UTF-8?q?Miko=C5=82aj=20Pir=C3=B3g?= <mikolaj.maciej.pirog at intel.com>
Date: Mon, 10 Feb 2025 05:48:02 +0100
Subject: [PATCH] [AVX10.2] Fix wrong intrinsic names after rename (#126390)
In my previous PR (#123656) to update the names of AVX10.2 intrinsics
and mnemonics, I have erroneously deleted `_ph` from few intrinsics.
This PR corrects this.
(cherry picked from commit 161cfc6f39bef8994eb944687033ebd3570196e8)
---
clang/lib/Headers/avx10_2_512convertintrin.h | 6 ++--
clang/lib/Headers/avx10_2convertintrin.h | 17 +++++----
.../CodeGen/X86/avx10_2_512convert-builtins.c | 18 +++++-----
.../CodeGen/X86/avx10_2convert-builtins.c | 36 +++++++++----------
4 files changed, 38 insertions(+), 39 deletions(-)
diff --git a/clang/lib/Headers/avx10_2_512convertintrin.h b/clang/lib/Headers/avx10_2_512convertintrin.h
index 0b5fca5cda5228f..516ccc68672d636 100644
--- a/clang/lib/Headers/avx10_2_512convertintrin.h
+++ b/clang/lib/Headers/avx10_2_512convertintrin.h
@@ -213,19 +213,19 @@ _mm512_maskz_cvts2ph_hf8(__mmask64 __U, __m512h __A, __m512h __B) {
(__v64qi)(__m512i)_mm512_setzero_si512());
}
-static __inline__ __m512h __DEFAULT_FN_ATTRS512 _mm512_cvthf8(__m256i __A) {
+static __inline__ __m512h __DEFAULT_FN_ATTRS512 _mm512_cvthf8_ph(__m256i __A) {
return (__m512h)__builtin_ia32_vcvthf8_2ph512_mask(
(__v32qi)__A, (__v32hf)(__m512h)_mm512_undefined_ph(), (__mmask32)-1);
}
static __inline__ __m512h __DEFAULT_FN_ATTRS512
-_mm512_mask_cvthf8(__m512h __W, __mmask32 __U, __m256i __A) {
+_mm512_mask_cvthf8_ph(__m512h __W, __mmask32 __U, __m256i __A) {
return (__m512h)__builtin_ia32_vcvthf8_2ph512_mask(
(__v32qi)__A, (__v32hf)(__m512h)__W, (__mmask32)__U);
}
static __inline__ __m512h __DEFAULT_FN_ATTRS512
-_mm512_maskz_cvthf8(__mmask32 __U, __m256i __A) {
+_mm512_maskz_cvthf8_ph(__mmask32 __U, __m256i __A) {
return (__m512h)__builtin_ia32_vcvthf8_2ph512_mask(
(__v32qi)__A, (__v32hf)(__m512h)_mm512_setzero_ph(), (__mmask32)__U);
}
diff --git a/clang/lib/Headers/avx10_2convertintrin.h b/clang/lib/Headers/avx10_2convertintrin.h
index c67a5b890f1957d..c419323910f187a 100644
--- a/clang/lib/Headers/avx10_2convertintrin.h
+++ b/clang/lib/Headers/avx10_2convertintrin.h
@@ -381,37 +381,36 @@ _mm256_maskz_cvts2ph_hf8(__mmask32 __U, __m256h __A, __m256h __B) {
(__v32qi)(__m256i)_mm256_setzero_si256());
}
-static __inline__ __m128h __DEFAULT_FN_ATTRS128 _mm_cvthf8(__m128i __A) {
+static __inline__ __m128h __DEFAULT_FN_ATTRS128 _mm_cvthf8_ph(__m128i __A) {
return (__m128h)__builtin_ia32_vcvthf8_2ph128_mask(
(__v16qi)__A, (__v8hf)(__m128h)_mm_undefined_ph(), (__mmask8)-1);
}
-static __inline__ __m128h __DEFAULT_FN_ATTRS128 _mm_mask_cvthf8(__m128h __W,
- __mmask8 __U,
- __m128i __A) {
+static __inline__ __m128h __DEFAULT_FN_ATTRS128
+_mm_mask_cvthf8_ph(__m128h __W, __mmask8 __U, __m128i __A) {
return (__m128h)__builtin_ia32_vcvthf8_2ph128_mask(
(__v16qi)__A, (__v8hf)(__m128h)__W, (__mmask8)__U);
}
-static __inline__ __m128h __DEFAULT_FN_ATTRS128 _mm_maskz_cvthf8(__mmask8 __U,
- __m128i __A) {
+static __inline__ __m128h __DEFAULT_FN_ATTRS128
+_mm_maskz_cvthf8_ph(__mmask8 __U, __m128i __A) {
return (__m128h)__builtin_ia32_vcvthf8_2ph128_mask(
(__v16qi)__A, (__v8hf)(__m128h)_mm_setzero_ph(), (__mmask8)__U);
}
-static __inline__ __m256h __DEFAULT_FN_ATTRS256 _mm256_cvthf8(__m128i __A) {
+static __inline__ __m256h __DEFAULT_FN_ATTRS256 _mm256_cvthf8_ph(__m128i __A) {
return (__m256h)__builtin_ia32_vcvthf8_2ph256_mask(
(__v16qi)__A, (__v16hf)(__m256h)_mm256_undefined_ph(), (__mmask16)-1);
}
static __inline__ __m256h __DEFAULT_FN_ATTRS256
-_mm256_mask_cvthf8(__m256h __W, __mmask16 __U, __m128i __A) {
+_mm256_mask_cvthf8_ph(__m256h __W, __mmask16 __U, __m128i __A) {
return (__m256h)__builtin_ia32_vcvthf8_2ph256_mask(
(__v16qi)__A, (__v16hf)(__m256h)__W, (__mmask16)__U);
}
static __inline__ __m256h __DEFAULT_FN_ATTRS256
-_mm256_maskz_cvthf8(__mmask16 __U, __m128i __A) {
+_mm256_maskz_cvthf8_ph(__mmask16 __U, __m128i __A) {
return (__m256h)__builtin_ia32_vcvthf8_2ph256_mask(
(__v16qi)__A, (__v16hf)(__m256h)_mm256_setzero_ph(), (__mmask16)__U);
}
diff --git a/clang/test/CodeGen/X86/avx10_2_512convert-builtins.c b/clang/test/CodeGen/X86/avx10_2_512convert-builtins.c
index 22503c640a727f2..dcf7bbc005a7c6c 100644
--- a/clang/test/CodeGen/X86/avx10_2_512convert-builtins.c
+++ b/clang/test/CodeGen/X86/avx10_2_512convert-builtins.c
@@ -201,22 +201,22 @@ __m512i test_mm512_maskz_cvts2ph_hf8(__mmask64 __U, __m512h __A, __m512h __B) {
return _mm512_maskz_cvts2ph_hf8(__U, __A, __B);
}
-__m512h test_mm512_cvthf8(__m256i __A) {
- // CHECK-LABEL: @test_mm512_cvthf8(
+__m512h test_mm512_cvthf8_ph(__m256i __A) {
+ // CHECK-LABEL: @test_mm512_cvthf8_ph(
// CHECK: call <32 x half> @llvm.x86.avx10.mask.vcvthf82ph512(
- return _mm512_cvthf8(__A);
+ return _mm512_cvthf8_ph(__A);
}
-__m512h test_mm512_mask_cvthf8(__m512h __A, __mmask32 __B, __m256i __C) {
- // CHECK-LABEL: @test_mm512_mask_cvthf8(
+__m512h test_mm512_mask_cvthf8_ph(__m512h __A, __mmask32 __B, __m256i __C) {
+ // CHECK-LABEL: @test_mm512_mask_cvthf8_ph(
// CHECK: call <32 x half> @llvm.x86.avx10.mask.vcvthf82ph512(
- return _mm512_mask_cvthf8(__A, __B, __C);
+ return _mm512_mask_cvthf8_ph(__A, __B, __C);
}
-__m512h test_mm512_maskz_cvthf8(__mmask32 __A, __m256i __B) {
- // CHECK-LABEL: @test_mm512_maskz_cvthf8(
+__m512h test_mm512_maskz_cvthf8_ph(__mmask32 __A, __m256i __B) {
+ // CHECK-LABEL: @test_mm512_maskz_cvthf8_ph(
// CHECK: call <32 x half> @llvm.x86.avx10.mask.vcvthf82ph512(
- return _mm512_maskz_cvthf8(__A, __B);
+ return _mm512_maskz_cvthf8_ph(__A, __B);
}
__m256i test_mm512_cvtph_bf8(__m512h __A) {
diff --git a/clang/test/CodeGen/X86/avx10_2convert-builtins.c b/clang/test/CodeGen/X86/avx10_2convert-builtins.c
index efd9a31c40875ab..d73a223922ce03d 100644
--- a/clang/test/CodeGen/X86/avx10_2convert-builtins.c
+++ b/clang/test/CodeGen/X86/avx10_2convert-builtins.c
@@ -379,40 +379,40 @@ __m256i test_mm256_maskz_cvts2ph_hf8(__mmask16 __U, __m256h __A, __m256h __B) {
return _mm256_maskz_cvts2ph_hf8(__U, __A, __B);
}
-__m128h test_mm_cvthf8(__m128i __A) {
- // CHECK-LABEL: @test_mm_cvthf8(
+__m128h test_mm_cvthf8_ph(__m128i __A) {
+ // CHECK-LABEL: @test_mm_cvthf8_ph(
// CHECK: call <8 x half> @llvm.x86.avx10.mask.vcvthf82ph128(
- return _mm_cvthf8(__A);
+ return _mm_cvthf8_ph(__A);
}
-__m128h test_mm_mask_cvthf8(__m128h __A, __mmask8 __B, __m128i __C) {
- // CHECK-LABEL: @test_mm_mask_cvthf8(
+__m128h test_mm_mask_cvthf8_ph(__m128h __A, __mmask8 __B, __m128i __C) {
+ // CHECK-LABEL: @test_mm_mask_cvthf8_ph(
// CHECK: call <8 x half> @llvm.x86.avx10.mask.vcvthf82ph128(
- return _mm_mask_cvthf8(__A, __B, __C);
+ return _mm_mask_cvthf8_ph(__A, __B, __C);
}
-__m128h test_mm_maskz_cvthf8(__mmask8 __A, __m128i __B) {
- // CHECK-LABEL: @test_mm_maskz_cvthf8(
+__m128h test_mm_maskz_cvthf8_ph(__mmask8 __A, __m128i __B) {
+ // CHECK-LABEL: @test_mm_maskz_cvthf8_ph(
// CHECK: call <8 x half> @llvm.x86.avx10.mask.vcvthf82ph128(
- return _mm_maskz_cvthf8(__A, __B);
+ return _mm_maskz_cvthf8_ph(__A, __B);
}
-__m256h test_mm256_cvthf8(__m128i __A) {
- // CHECK-LABEL: @test_mm256_cvthf8(
+__m256h test_mm256_cvthf8_ph(__m128i __A) {
+ // CHECK-LABEL: @test_mm256_cvthf8_ph(
// CHECK: call <16 x half> @llvm.x86.avx10.mask.vcvthf82ph256(
- return _mm256_cvthf8(__A);
+ return _mm256_cvthf8_ph(__A);
}
-__m256h test_mm256_mask_cvthf8(__m256h __A, __mmask16 __B, __m128i __C) {
- // CHECK-LABEL: @test_mm256_mask_cvthf8(
+__m256h test_mm256_mask_cvthf8_ph(__m256h __A, __mmask16 __B, __m128i __C) {
+ // CHECK-LABEL: @test_mm256_mask_cvthf8_ph(
// CHECK: call <16 x half> @llvm.x86.avx10.mask.vcvthf82ph256(
- return _mm256_mask_cvthf8(__A, __B, __C);
+ return _mm256_mask_cvthf8_ph(__A, __B, __C);
}
-__m256h test_mm256_maskz_cvthf8(__mmask16 __A, __m128i __B) {
- // CHECK-LABEL: @test_mm256_maskz_cvthf8(
+__m256h test_mm256_maskz_cvthf8_ph(__mmask16 __A, __m128i __B) {
+ // CHECK-LABEL: @test_mm256_maskz_cvthf8_ph(
// CHECK: call <16 x half> @llvm.x86.avx10.mask.vcvthf82ph256(
- return _mm256_maskz_cvthf8(__A, __B);
+ return _mm256_maskz_cvthf8_ph(__A, __B);
}
__m128i test_mm_cvtph_bf8(__m128h __A) {
More information about the llvm-branch-commits
mailing list