[llvm-branch-commits] [llvm] release/23.x: [X86][AVX10.2] Use INTR_TYPE_1OP_MASK_SAE for 512-bit VCVTTP.*QS intrinsics (#244127) (PR #245034)

via llvm-branch-commits llvm-branch-commits at lists.llvm.org
Sun Nov 22 04:51:41 PST 2026


https://github.com/llvmbot created https://github.com/llvm/llvm-project/pull/245034

Backport bb55066964931c9287b04148eb2d97fc7329800d

Requested by: @phoebewang

>From ff96fa88831b2fae57193c9b71dc27b3be0c1813 Mon Sep 17 00:00:00 2001
From: Phoebe Wang <phoebe.wang at intel.com>
Date: Sun, 22 Nov 2026 12:19:18 +0800
Subject: [PATCH] [X86][AVX10.2] Use INTR_TYPE_1OP_MASK_SAE for 512-bit
 VCVTTP.*QS intrinsics (#244127)

They are SAE instructions like VCVTTP.*Q.

Assisted-by: Claude Opus 5.5
(cherry picked from commit bb55066964931c9287b04148eb2d97fc7329800d)
---
 llvm/lib/Target/X86/X86IntrinsicsInfo.h       |  28 +-
 .../X86/avx10_2_512satcvtds-intrinsics.ll     | 289 ++++++++++++++++++
 2 files changed, 305 insertions(+), 12 deletions(-)

diff --git a/llvm/lib/Target/X86/X86IntrinsicsInfo.h b/llvm/lib/Target/X86/X86IntrinsicsInfo.h
index a3fd8e030f031f..852e2a1d3f2644 100644
--- a/llvm/lib/Target/X86/X86IntrinsicsInfo.h
+++ b/llvm/lib/Target/X86/X86IntrinsicsInfo.h
@@ -525,26 +525,28 @@ static const IntrinsicData IntrinsicsWithoutChain[] = {
                        X86ISD::CVTTP2SIS, X86ISD::MCVTTP2SIS),
     X86_INTRINSIC_DATA(avx10_mask_vcvttpd2dqs_256, INTR_TYPE_1OP_MASK,
                        X86ISD::CVTTP2SIS, 0),
-    X86_INTRINSIC_DATA(avx10_mask_vcvttpd2dqs_round_512, INTR_TYPE_1OP_MASK,
+    X86_INTRINSIC_DATA(avx10_mask_vcvttpd2dqs_round_512, INTR_TYPE_1OP_MASK_SAE,
                        X86ISD::CVTTP2SIS, X86ISD::CVTTP2SIS_SAE),
     X86_INTRINSIC_DATA(avx10_mask_vcvttpd2qqs_128, INTR_TYPE_1OP_MASK,
                        X86ISD::CVTTP2SIS, 0),
     X86_INTRINSIC_DATA(avx10_mask_vcvttpd2qqs_256, INTR_TYPE_1OP_MASK,
                        X86ISD::CVTTP2SIS, 0),
-    X86_INTRINSIC_DATA(avx10_mask_vcvttpd2qqs_round_512, INTR_TYPE_1OP_MASK,
+    X86_INTRINSIC_DATA(avx10_mask_vcvttpd2qqs_round_512, INTR_TYPE_1OP_MASK_SAE,
                        X86ISD::CVTTP2SIS, X86ISD::CVTTP2SIS_SAE),
     X86_INTRINSIC_DATA(avx10_mask_vcvttpd2udqs_128, CVTPD2DQ_MASK,
                        X86ISD::CVTTP2UIS, X86ISD::MCVTTP2UIS),
     X86_INTRINSIC_DATA(avx10_mask_vcvttpd2udqs_256, INTR_TYPE_1OP_MASK,
                        X86ISD::CVTTP2UIS, 0),
-    X86_INTRINSIC_DATA(avx10_mask_vcvttpd2udqs_round_512, INTR_TYPE_1OP_MASK,
-                       X86ISD::CVTTP2UIS, X86ISD::CVTTP2UIS_SAE),
+    X86_INTRINSIC_DATA(avx10_mask_vcvttpd2udqs_round_512,
+                       INTR_TYPE_1OP_MASK_SAE, X86ISD::CVTTP2UIS,
+                       X86ISD::CVTTP2UIS_SAE),
     X86_INTRINSIC_DATA(avx10_mask_vcvttpd2uqqs_128, INTR_TYPE_1OP_MASK,
                        X86ISD::CVTTP2UIS, 0),
     X86_INTRINSIC_DATA(avx10_mask_vcvttpd2uqqs_256, INTR_TYPE_1OP_MASK,
                        X86ISD::CVTTP2UIS, 0),
-    X86_INTRINSIC_DATA(avx10_mask_vcvttpd2uqqs_round_512, INTR_TYPE_1OP_MASK,
-                       X86ISD::CVTTP2UIS, X86ISD::CVTTP2UIS_SAE),
+    X86_INTRINSIC_DATA(avx10_mask_vcvttpd2uqqs_round_512,
+                       INTR_TYPE_1OP_MASK_SAE, X86ISD::CVTTP2UIS,
+                       X86ISD::CVTTP2UIS_SAE),
     X86_INTRINSIC_DATA(avx10_mask_vcvttph2ibs128, INTR_TYPE_1OP_MASK,
                        X86ISD::CVTTP2IBS, 0),
     X86_INTRINSIC_DATA(avx10_mask_vcvttph2ibs256, INTR_TYPE_1OP_MASK,
@@ -561,7 +563,7 @@ static const IntrinsicData IntrinsicsWithoutChain[] = {
                        X86ISD::CVTTP2SIS, 0),
     X86_INTRINSIC_DATA(avx10_mask_vcvttps2dqs_256, INTR_TYPE_1OP_MASK,
                        X86ISD::CVTTP2SIS, 0),
-    X86_INTRINSIC_DATA(avx10_mask_vcvttps2dqs_round_512, INTR_TYPE_1OP_MASK,
+    X86_INTRINSIC_DATA(avx10_mask_vcvttps2dqs_round_512, INTR_TYPE_1OP_MASK_SAE,
                        X86ISD::CVTTP2SIS, X86ISD::CVTTP2SIS_SAE),
     X86_INTRINSIC_DATA(avx10_mask_vcvttps2ibs128, INTR_TYPE_1OP_MASK,
                        X86ISD::CVTTP2IBS, 0),
@@ -579,20 +581,22 @@ static const IntrinsicData IntrinsicsWithoutChain[] = {
                        X86ISD::CVTTP2SIS, 0),
     X86_INTRINSIC_DATA(avx10_mask_vcvttps2qqs_256, INTR_TYPE_1OP_MASK,
                        X86ISD::CVTTP2SIS, 0),
-    X86_INTRINSIC_DATA(avx10_mask_vcvttps2qqs_round_512, INTR_TYPE_1OP_MASK,
+    X86_INTRINSIC_DATA(avx10_mask_vcvttps2qqs_round_512, INTR_TYPE_1OP_MASK_SAE,
                        X86ISD::CVTTP2SIS, X86ISD::CVTTP2SIS_SAE),
     X86_INTRINSIC_DATA(avx10_mask_vcvttps2udqs_128, INTR_TYPE_1OP_MASK,
                        X86ISD::CVTTP2UIS, 0),
     X86_INTRINSIC_DATA(avx10_mask_vcvttps2udqs_256, INTR_TYPE_1OP_MASK,
                        X86ISD::CVTTP2UIS, 0),
-    X86_INTRINSIC_DATA(avx10_mask_vcvttps2udqs_round_512, INTR_TYPE_1OP_MASK,
-                       X86ISD::CVTTP2UIS, X86ISD::CVTTP2UIS_SAE),
+    X86_INTRINSIC_DATA(avx10_mask_vcvttps2udqs_round_512,
+                       INTR_TYPE_1OP_MASK_SAE, X86ISD::CVTTP2UIS,
+                       X86ISD::CVTTP2UIS_SAE),
     X86_INTRINSIC_DATA(avx10_mask_vcvttps2uqqs_128, INTR_TYPE_1OP_MASK,
                        X86ISD::CVTTP2UIS, 0),
     X86_INTRINSIC_DATA(avx10_mask_vcvttps2uqqs_256, INTR_TYPE_1OP_MASK,
                        X86ISD::CVTTP2UIS, 0),
-    X86_INTRINSIC_DATA(avx10_mask_vcvttps2uqqs_round_512, INTR_TYPE_1OP_MASK,
-                       X86ISD::CVTTP2UIS, X86ISD::CVTTP2UIS_SAE),
+    X86_INTRINSIC_DATA(avx10_mask_vcvttps2uqqs_round_512,
+                       INTR_TYPE_1OP_MASK_SAE, X86ISD::CVTTP2UIS,
+                       X86ISD::CVTTP2UIS_SAE),
     X86_INTRINSIC_DATA(avx10_mask_vminmaxpd_round, INTR_TYPE_3OP_MASK_SAE,
                        X86ISD::VMINMAX, X86ISD::VMINMAX_SAE),
     X86_INTRINSIC_DATA(avx10_mask_vminmaxpd128, INTR_TYPE_3OP_MASK_SAE,
diff --git a/llvm/test/CodeGen/X86/avx10_2_512satcvtds-intrinsics.ll b/llvm/test/CodeGen/X86/avx10_2_512satcvtds-intrinsics.ll
index 591349aabef454..ca96ed93eb82af 100644
--- a/llvm/test/CodeGen/X86/avx10_2_512satcvtds-intrinsics.ll
+++ b/llvm/test/CodeGen/X86/avx10_2_512satcvtds-intrinsics.ll
@@ -524,3 +524,292 @@ define <8 x i64> @test_int_x86_mask_vcvtt_ps2uqqs_512_default(<8 x float> %x0) {
   ret <8 x i64> %res
 }
 declare <8 x i64> @llvm.x86.avx10.mask.vcvttps2uqqs.round.512(<8 x float>, <8 x i64>, i8 , i32)
+
+define <8 x i32> @test_int_x86_mask_vcvtt_pd2dqs_512_sae_8(<8 x double> %x0, <8 x i32> %src, i8 %mask) {
+; X64-LABEL: test_int_x86_mask_vcvtt_pd2dqs_512_sae_8:
+; X64:       # %bb.0:
+; X64-NEXT:    kmovd %edi, %k1 # encoding: [0xc5,0xfb,0x92,0xcf]
+; X64-NEXT:    vcvttpd2dqs {sae}, %zmm0, %ymm1 {%k1} # encoding: [0x62,0xf5,0xfc,0x19,0x6d,0xc8]
+; X64-NEXT:    vmovaps %ymm1, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfc,0x28,0xc1]
+; X64-NEXT:    retq # encoding: [0xc3]
+;
+; X86-LABEL: test_int_x86_mask_vcvtt_pd2dqs_512_sae_8:
+; X86:       # %bb.0:
+; X86-NEXT:    kmovb {{[0-9]+}}(%esp), %k1 # encoding: [0xc5,0xf9,0x90,0x4c,0x24,0x04]
+; X86-NEXT:    vcvttpd2dqs {sae}, %zmm0, %ymm1 {%k1} # encoding: [0x62,0xf5,0xfc,0x19,0x6d,0xc8]
+; X86-NEXT:    vmovaps %ymm1, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfc,0x28,0xc1]
+; X86-NEXT:    retl # encoding: [0xc3]
+  %res = call <8 x i32> @llvm.x86.avx10.mask.vcvttpd2dqs.round.512(<8 x double> %x0, <8 x i32> %src, i8 %mask, i32 8)
+  ret <8 x i32> %res
+}
+
+define <8 x i32> @test_int_x86_mask_vcvtt_pd2dqs_512_sae_12(<8 x double> %x0, <8 x i32> %src, i8 %mask) {
+; X64-LABEL: test_int_x86_mask_vcvtt_pd2dqs_512_sae_12:
+; X64:       # %bb.0:
+; X64-NEXT:    kmovd %edi, %k1 # encoding: [0xc5,0xfb,0x92,0xcf]
+; X64-NEXT:    vcvttpd2dqs {sae}, %zmm0, %ymm1 {%k1} # encoding: [0x62,0xf5,0xfc,0x19,0x6d,0xc8]
+; X64-NEXT:    vmovaps %ymm1, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfc,0x28,0xc1]
+; X64-NEXT:    retq # encoding: [0xc3]
+;
+; X86-LABEL: test_int_x86_mask_vcvtt_pd2dqs_512_sae_12:
+; X86:       # %bb.0:
+; X86-NEXT:    kmovb {{[0-9]+}}(%esp), %k1 # encoding: [0xc5,0xf9,0x90,0x4c,0x24,0x04]
+; X86-NEXT:    vcvttpd2dqs {sae}, %zmm0, %ymm1 {%k1} # encoding: [0x62,0xf5,0xfc,0x19,0x6d,0xc8]
+; X86-NEXT:    vmovaps %ymm1, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfc,0x28,0xc1]
+; X86-NEXT:    retl # encoding: [0xc3]
+  %res = call <8 x i32> @llvm.x86.avx10.mask.vcvttpd2dqs.round.512(<8 x double> %x0, <8 x i32> %src, i8 %mask, i32 12)
+  ret <8 x i32> %res
+}
+
+define <8 x i32> @test_int_x86_mask_vcvtt_pd2udqs_512_sae_8(<8 x double> %x0, <8 x i32> %src, i8 %mask) {
+; X64-LABEL: test_int_x86_mask_vcvtt_pd2udqs_512_sae_8:
+; X64:       # %bb.0:
+; X64-NEXT:    kmovd %edi, %k1 # encoding: [0xc5,0xfb,0x92,0xcf]
+; X64-NEXT:    vcvttpd2udqs {sae}, %zmm0, %ymm1 {%k1} # encoding: [0x62,0xf5,0xfc,0x19,0x6c,0xc8]
+; X64-NEXT:    vmovaps %ymm1, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfc,0x28,0xc1]
+; X64-NEXT:    retq # encoding: [0xc3]
+;
+; X86-LABEL: test_int_x86_mask_vcvtt_pd2udqs_512_sae_8:
+; X86:       # %bb.0:
+; X86-NEXT:    kmovb {{[0-9]+}}(%esp), %k1 # encoding: [0xc5,0xf9,0x90,0x4c,0x24,0x04]
+; X86-NEXT:    vcvttpd2udqs {sae}, %zmm0, %ymm1 {%k1} # encoding: [0x62,0xf5,0xfc,0x19,0x6c,0xc8]
+; X86-NEXT:    vmovaps %ymm1, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfc,0x28,0xc1]
+; X86-NEXT:    retl # encoding: [0xc3]
+  %res = call <8 x i32> @llvm.x86.avx10.mask.vcvttpd2udqs.round.512(<8 x double> %x0, <8 x i32> %src, i8 %mask, i32 8)
+  ret <8 x i32> %res
+}
+
+define <8 x i32> @test_int_x86_mask_vcvtt_pd2udqs_512_sae_12(<8 x double> %x0, <8 x i32> %src, i8 %mask) {
+; X64-LABEL: test_int_x86_mask_vcvtt_pd2udqs_512_sae_12:
+; X64:       # %bb.0:
+; X64-NEXT:    kmovd %edi, %k1 # encoding: [0xc5,0xfb,0x92,0xcf]
+; X64-NEXT:    vcvttpd2udqs {sae}, %zmm0, %ymm1 {%k1} # encoding: [0x62,0xf5,0xfc,0x19,0x6c,0xc8]
+; X64-NEXT:    vmovaps %ymm1, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfc,0x28,0xc1]
+; X64-NEXT:    retq # encoding: [0xc3]
+;
+; X86-LABEL: test_int_x86_mask_vcvtt_pd2udqs_512_sae_12:
+; X86:       # %bb.0:
+; X86-NEXT:    kmovb {{[0-9]+}}(%esp), %k1 # encoding: [0xc5,0xf9,0x90,0x4c,0x24,0x04]
+; X86-NEXT:    vcvttpd2udqs {sae}, %zmm0, %ymm1 {%k1} # encoding: [0x62,0xf5,0xfc,0x19,0x6c,0xc8]
+; X86-NEXT:    vmovaps %ymm1, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfc,0x28,0xc1]
+; X86-NEXT:    retl # encoding: [0xc3]
+  %res = call <8 x i32> @llvm.x86.avx10.mask.vcvttpd2udqs.round.512(<8 x double> %x0, <8 x i32> %src, i8 %mask, i32 12)
+  ret <8 x i32> %res
+}
+
+define <8 x i64> @test_int_x86_mask_vcvtt_pd2qqs_512_sae_8(<8 x double> %x0, <8 x i64> %src, i8 %mask) {
+; X64-LABEL: test_int_x86_mask_vcvtt_pd2qqs_512_sae_8:
+; X64:       # %bb.0:
+; X64-NEXT:    kmovd %edi, %k1 # encoding: [0xc5,0xfb,0x92,0xcf]
+; X64-NEXT:    vcvttpd2qqs {sae}, %zmm0, %zmm1 {%k1} # encoding: [0x62,0xf5,0xfd,0x19,0x6d,0xc8]
+; X64-NEXT:    vmovaps %zmm1, %zmm0 # encoding: [0x62,0xf1,0x7c,0x48,0x28,0xc1]
+; X64-NEXT:    retq # encoding: [0xc3]
+;
+; X86-LABEL: test_int_x86_mask_vcvtt_pd2qqs_512_sae_8:
+; X86:       # %bb.0:
+; X86-NEXT:    kmovb {{[0-9]+}}(%esp), %k1 # encoding: [0xc5,0xf9,0x90,0x4c,0x24,0x04]
+; X86-NEXT:    vcvttpd2qqs {sae}, %zmm0, %zmm1 {%k1} # encoding: [0x62,0xf5,0xfd,0x19,0x6d,0xc8]
+; X86-NEXT:    vmovaps %zmm1, %zmm0 # encoding: [0x62,0xf1,0x7c,0x48,0x28,0xc1]
+; X86-NEXT:    retl # encoding: [0xc3]
+  %res = call <8 x i64> @llvm.x86.avx10.mask.vcvttpd2qqs.round.512(<8 x double> %x0, <8 x i64> %src, i8 %mask, i32 8)
+  ret <8 x i64> %res
+}
+
+define <8 x i64> @test_int_x86_mask_vcvtt_pd2qqs_512_sae_12(<8 x double> %x0, <8 x i64> %src, i8 %mask) {
+; X64-LABEL: test_int_x86_mask_vcvtt_pd2qqs_512_sae_12:
+; X64:       # %bb.0:
+; X64-NEXT:    kmovd %edi, %k1 # encoding: [0xc5,0xfb,0x92,0xcf]
+; X64-NEXT:    vcvttpd2qqs {sae}, %zmm0, %zmm1 {%k1} # encoding: [0x62,0xf5,0xfd,0x19,0x6d,0xc8]
+; X64-NEXT:    vmovaps %zmm1, %zmm0 # encoding: [0x62,0xf1,0x7c,0x48,0x28,0xc1]
+; X64-NEXT:    retq # encoding: [0xc3]
+;
+; X86-LABEL: test_int_x86_mask_vcvtt_pd2qqs_512_sae_12:
+; X86:       # %bb.0:
+; X86-NEXT:    kmovb {{[0-9]+}}(%esp), %k1 # encoding: [0xc5,0xf9,0x90,0x4c,0x24,0x04]
+; X86-NEXT:    vcvttpd2qqs {sae}, %zmm0, %zmm1 {%k1} # encoding: [0x62,0xf5,0xfd,0x19,0x6d,0xc8]
+; X86-NEXT:    vmovaps %zmm1, %zmm0 # encoding: [0x62,0xf1,0x7c,0x48,0x28,0xc1]
+; X86-NEXT:    retl # encoding: [0xc3]
+  %res = call <8 x i64> @llvm.x86.avx10.mask.vcvttpd2qqs.round.512(<8 x double> %x0, <8 x i64> %src, i8 %mask, i32 12)
+  ret <8 x i64> %res
+}
+
+define <8 x i64> @test_int_x86_mask_vcvtt_pd2uqqs_512_sae_8(<8 x double> %x0, <8 x i64> %src, i8 %mask) {
+; X64-LABEL: test_int_x86_mask_vcvtt_pd2uqqs_512_sae_8:
+; X64:       # %bb.0:
+; X64-NEXT:    kmovd %edi, %k1 # encoding: [0xc5,0xfb,0x92,0xcf]
+; X64-NEXT:    vcvttpd2uqqs {sae}, %zmm0, %zmm1 {%k1} # encoding: [0x62,0xf5,0xfd,0x19,0x6c,0xc8]
+; X64-NEXT:    vmovaps %zmm1, %zmm0 # encoding: [0x62,0xf1,0x7c,0x48,0x28,0xc1]
+; X64-NEXT:    retq # encoding: [0xc3]
+;
+; X86-LABEL: test_int_x86_mask_vcvtt_pd2uqqs_512_sae_8:
+; X86:       # %bb.0:
+; X86-NEXT:    kmovb {{[0-9]+}}(%esp), %k1 # encoding: [0xc5,0xf9,0x90,0x4c,0x24,0x04]
+; X86-NEXT:    vcvttpd2uqqs {sae}, %zmm0, %zmm1 {%k1} # encoding: [0x62,0xf5,0xfd,0x19,0x6c,0xc8]
+; X86-NEXT:    vmovaps %zmm1, %zmm0 # encoding: [0x62,0xf1,0x7c,0x48,0x28,0xc1]
+; X86-NEXT:    retl # encoding: [0xc3]
+  %res = call <8 x i64> @llvm.x86.avx10.mask.vcvttpd2uqqs.round.512(<8 x double> %x0, <8 x i64> %src, i8 %mask, i32 8)
+  ret <8 x i64> %res
+}
+
+define <8 x i64> @test_int_x86_mask_vcvtt_pd2uqqs_512_sae_12(<8 x double> %x0, <8 x i64> %src, i8 %mask) {
+; X64-LABEL: test_int_x86_mask_vcvtt_pd2uqqs_512_sae_12:
+; X64:       # %bb.0:
+; X64-NEXT:    kmovd %edi, %k1 # encoding: [0xc5,0xfb,0x92,0xcf]
+; X64-NEXT:    vcvttpd2uqqs {sae}, %zmm0, %zmm1 {%k1} # encoding: [0x62,0xf5,0xfd,0x19,0x6c,0xc8]
+; X64-NEXT:    vmovaps %zmm1, %zmm0 # encoding: [0x62,0xf1,0x7c,0x48,0x28,0xc1]
+; X64-NEXT:    retq # encoding: [0xc3]
+;
+; X86-LABEL: test_int_x86_mask_vcvtt_pd2uqqs_512_sae_12:
+; X86:       # %bb.0:
+; X86-NEXT:    kmovb {{[0-9]+}}(%esp), %k1 # encoding: [0xc5,0xf9,0x90,0x4c,0x24,0x04]
+; X86-NEXT:    vcvttpd2uqqs {sae}, %zmm0, %zmm1 {%k1} # encoding: [0x62,0xf5,0xfd,0x19,0x6c,0xc8]
+; X86-NEXT:    vmovaps %zmm1, %zmm0 # encoding: [0x62,0xf1,0x7c,0x48,0x28,0xc1]
+; X86-NEXT:    retl # encoding: [0xc3]
+  %res = call <8 x i64> @llvm.x86.avx10.mask.vcvttpd2uqqs.round.512(<8 x double> %x0, <8 x i64> %src, i8 %mask, i32 12)
+  ret <8 x i64> %res
+}
+
+define <16 x i32> @test_int_x86_mask_vcvtt_ps2dqs_512_sae_8(<16 x float> %x0, <16 x i32> %src, i16 %mask) {
+; X64-LABEL: test_int_x86_mask_vcvtt_ps2dqs_512_sae_8:
+; X64:       # %bb.0:
+; X64-NEXT:    kmovd %edi, %k1 # encoding: [0xc5,0xfb,0x92,0xcf]
+; X64-NEXT:    vcvttps2dqs {sae}, %zmm0, %zmm1 {%k1} # encoding: [0x62,0xf5,0x7c,0x19,0x6d,0xc8]
+; X64-NEXT:    vmovaps %zmm1, %zmm0 # encoding: [0x62,0xf1,0x7c,0x48,0x28,0xc1]
+; X64-NEXT:    retq # encoding: [0xc3]
+;
+; X86-LABEL: test_int_x86_mask_vcvtt_ps2dqs_512_sae_8:
+; X86:       # %bb.0:
+; X86-NEXT:    kmovw {{[0-9]+}}(%esp), %k1 # encoding: [0xc5,0xf8,0x90,0x4c,0x24,0x04]
+; X86-NEXT:    vcvttps2dqs {sae}, %zmm0, %zmm1 {%k1} # encoding: [0x62,0xf5,0x7c,0x19,0x6d,0xc8]
+; X86-NEXT:    vmovaps %zmm1, %zmm0 # encoding: [0x62,0xf1,0x7c,0x48,0x28,0xc1]
+; X86-NEXT:    retl # encoding: [0xc3]
+  %res = call <16 x i32> @llvm.x86.avx10.mask.vcvttps2dqs.round.512(<16 x float> %x0, <16 x i32> %src, i16 %mask, i32 8)
+  ret <16 x i32> %res
+}
+
+define <16 x i32> @test_int_x86_mask_vcvtt_ps2dqs_512_sae_12(<16 x float> %x0, <16 x i32> %src, i16 %mask) {
+; X64-LABEL: test_int_x86_mask_vcvtt_ps2dqs_512_sae_12:
+; X64:       # %bb.0:
+; X64-NEXT:    kmovd %edi, %k1 # encoding: [0xc5,0xfb,0x92,0xcf]
+; X64-NEXT:    vcvttps2dqs {sae}, %zmm0, %zmm1 {%k1} # encoding: [0x62,0xf5,0x7c,0x19,0x6d,0xc8]
+; X64-NEXT:    vmovaps %zmm1, %zmm0 # encoding: [0x62,0xf1,0x7c,0x48,0x28,0xc1]
+; X64-NEXT:    retq # encoding: [0xc3]
+;
+; X86-LABEL: test_int_x86_mask_vcvtt_ps2dqs_512_sae_12:
+; X86:       # %bb.0:
+; X86-NEXT:    kmovw {{[0-9]+}}(%esp), %k1 # encoding: [0xc5,0xf8,0x90,0x4c,0x24,0x04]
+; X86-NEXT:    vcvttps2dqs {sae}, %zmm0, %zmm1 {%k1} # encoding: [0x62,0xf5,0x7c,0x19,0x6d,0xc8]
+; X86-NEXT:    vmovaps %zmm1, %zmm0 # encoding: [0x62,0xf1,0x7c,0x48,0x28,0xc1]
+; X86-NEXT:    retl # encoding: [0xc3]
+  %res = call <16 x i32> @llvm.x86.avx10.mask.vcvttps2dqs.round.512(<16 x float> %x0, <16 x i32> %src, i16 %mask, i32 12)
+  ret <16 x i32> %res
+}
+
+define <16 x i32> @test_int_x86_mask_vcvtt_ps2udqs_512_sae_8(<16 x float> %x0, <16 x i32> %src, i16 %mask) {
+; X64-LABEL: test_int_x86_mask_vcvtt_ps2udqs_512_sae_8:
+; X64:       # %bb.0:
+; X64-NEXT:    kmovd %edi, %k1 # encoding: [0xc5,0xfb,0x92,0xcf]
+; X64-NEXT:    vcvttps2udqs {sae}, %zmm0, %zmm1 {%k1} # encoding: [0x62,0xf5,0x7c,0x19,0x6c,0xc8]
+; X64-NEXT:    vmovaps %zmm1, %zmm0 # encoding: [0x62,0xf1,0x7c,0x48,0x28,0xc1]
+; X64-NEXT:    retq # encoding: [0xc3]
+;
+; X86-LABEL: test_int_x86_mask_vcvtt_ps2udqs_512_sae_8:
+; X86:       # %bb.0:
+; X86-NEXT:    kmovw {{[0-9]+}}(%esp), %k1 # encoding: [0xc5,0xf8,0x90,0x4c,0x24,0x04]
+; X86-NEXT:    vcvttps2udqs {sae}, %zmm0, %zmm1 {%k1} # encoding: [0x62,0xf5,0x7c,0x19,0x6c,0xc8]
+; X86-NEXT:    vmovaps %zmm1, %zmm0 # encoding: [0x62,0xf1,0x7c,0x48,0x28,0xc1]
+; X86-NEXT:    retl # encoding: [0xc3]
+  %res = call <16 x i32> @llvm.x86.avx10.mask.vcvttps2udqs.round.512(<16 x float> %x0, <16 x i32> %src, i16 %mask, i32 8)
+  ret <16 x i32> %res
+}
+
+define <16 x i32> @test_int_x86_mask_vcvtt_ps2udqs_512_sae_12(<16 x float> %x0, <16 x i32> %src, i16 %mask) {
+; X64-LABEL: test_int_x86_mask_vcvtt_ps2udqs_512_sae_12:
+; X64:       # %bb.0:
+; X64-NEXT:    kmovd %edi, %k1 # encoding: [0xc5,0xfb,0x92,0xcf]
+; X64-NEXT:    vcvttps2udqs {sae}, %zmm0, %zmm1 {%k1} # encoding: [0x62,0xf5,0x7c,0x19,0x6c,0xc8]
+; X64-NEXT:    vmovaps %zmm1, %zmm0 # encoding: [0x62,0xf1,0x7c,0x48,0x28,0xc1]
+; X64-NEXT:    retq # encoding: [0xc3]
+;
+; X86-LABEL: test_int_x86_mask_vcvtt_ps2udqs_512_sae_12:
+; X86:       # %bb.0:
+; X86-NEXT:    kmovw {{[0-9]+}}(%esp), %k1 # encoding: [0xc5,0xf8,0x90,0x4c,0x24,0x04]
+; X86-NEXT:    vcvttps2udqs {sae}, %zmm0, %zmm1 {%k1} # encoding: [0x62,0xf5,0x7c,0x19,0x6c,0xc8]
+; X86-NEXT:    vmovaps %zmm1, %zmm0 # encoding: [0x62,0xf1,0x7c,0x48,0x28,0xc1]
+; X86-NEXT:    retl # encoding: [0xc3]
+  %res = call <16 x i32> @llvm.x86.avx10.mask.vcvttps2udqs.round.512(<16 x float> %x0, <16 x i32> %src, i16 %mask, i32 12)
+  ret <16 x i32> %res
+}
+
+define <8 x i64> @test_int_x86_mask_vcvtt_ps2qqs_512_sae_8(<8 x float> %x0, <8 x i64> %src, i8 %mask) {
+; X64-LABEL: test_int_x86_mask_vcvtt_ps2qqs_512_sae_8:
+; X64:       # %bb.0:
+; X64-NEXT:    kmovd %edi, %k1 # encoding: [0xc5,0xfb,0x92,0xcf]
+; X64-NEXT:    vcvttps2qqs {sae}, %ymm0, %zmm1 {%k1} # encoding: [0x62,0xf5,0x7d,0x19,0x6d,0xc8]
+; X64-NEXT:    vmovaps %zmm1, %zmm0 # encoding: [0x62,0xf1,0x7c,0x48,0x28,0xc1]
+; X64-NEXT:    retq # encoding: [0xc3]
+;
+; X86-LABEL: test_int_x86_mask_vcvtt_ps2qqs_512_sae_8:
+; X86:       # %bb.0:
+; X86-NEXT:    kmovb {{[0-9]+}}(%esp), %k1 # encoding: [0xc5,0xf9,0x90,0x4c,0x24,0x04]
+; X86-NEXT:    vcvttps2qqs {sae}, %ymm0, %zmm1 {%k1} # encoding: [0x62,0xf5,0x7d,0x19,0x6d,0xc8]
+; X86-NEXT:    vmovaps %zmm1, %zmm0 # encoding: [0x62,0xf1,0x7c,0x48,0x28,0xc1]
+; X86-NEXT:    retl # encoding: [0xc3]
+  %res = call <8 x i64> @llvm.x86.avx10.mask.vcvttps2qqs.round.512(<8 x float> %x0, <8 x i64> %src, i8 %mask, i32 8)
+  ret <8 x i64> %res
+}
+
+define <8 x i64> @test_int_x86_mask_vcvtt_ps2qqs_512_sae_12(<8 x float> %x0, <8 x i64> %src, i8 %mask) {
+; X64-LABEL: test_int_x86_mask_vcvtt_ps2qqs_512_sae_12:
+; X64:       # %bb.0:
+; X64-NEXT:    kmovd %edi, %k1 # encoding: [0xc5,0xfb,0x92,0xcf]
+; X64-NEXT:    vcvttps2qqs {sae}, %ymm0, %zmm1 {%k1} # encoding: [0x62,0xf5,0x7d,0x19,0x6d,0xc8]
+; X64-NEXT:    vmovaps %zmm1, %zmm0 # encoding: [0x62,0xf1,0x7c,0x48,0x28,0xc1]
+; X64-NEXT:    retq # encoding: [0xc3]
+;
+; X86-LABEL: test_int_x86_mask_vcvtt_ps2qqs_512_sae_12:
+; X86:       # %bb.0:
+; X86-NEXT:    kmovb {{[0-9]+}}(%esp), %k1 # encoding: [0xc5,0xf9,0x90,0x4c,0x24,0x04]
+; X86-NEXT:    vcvttps2qqs {sae}, %ymm0, %zmm1 {%k1} # encoding: [0x62,0xf5,0x7d,0x19,0x6d,0xc8]
+; X86-NEXT:    vmovaps %zmm1, %zmm0 # encoding: [0x62,0xf1,0x7c,0x48,0x28,0xc1]
+; X86-NEXT:    retl # encoding: [0xc3]
+  %res = call <8 x i64> @llvm.x86.avx10.mask.vcvttps2qqs.round.512(<8 x float> %x0, <8 x i64> %src, i8 %mask, i32 12)
+  ret <8 x i64> %res
+}
+
+define <8 x i64> @test_int_x86_mask_vcvtt_ps2uqqs_512_sae_8(<8 x float> %x0, <8 x i64> %src, i8 %mask) {
+; X64-LABEL: test_int_x86_mask_vcvtt_ps2uqqs_512_sae_8:
+; X64:       # %bb.0:
+; X64-NEXT:    kmovd %edi, %k1 # encoding: [0xc5,0xfb,0x92,0xcf]
+; X64-NEXT:    vcvttps2uqqs {sae}, %ymm0, %zmm1 {%k1} # encoding: [0x62,0xf5,0x7d,0x19,0x6c,0xc8]
+; X64-NEXT:    vmovaps %zmm1, %zmm0 # encoding: [0x62,0xf1,0x7c,0x48,0x28,0xc1]
+; X64-NEXT:    retq # encoding: [0xc3]
+;
+; X86-LABEL: test_int_x86_mask_vcvtt_ps2uqqs_512_sae_8:
+; X86:       # %bb.0:
+; X86-NEXT:    kmovb {{[0-9]+}}(%esp), %k1 # encoding: [0xc5,0xf9,0x90,0x4c,0x24,0x04]
+; X86-NEXT:    vcvttps2uqqs {sae}, %ymm0, %zmm1 {%k1} # encoding: [0x62,0xf5,0x7d,0x19,0x6c,0xc8]
+; X86-NEXT:    vmovaps %zmm1, %zmm0 # encoding: [0x62,0xf1,0x7c,0x48,0x28,0xc1]
+; X86-NEXT:    retl # encoding: [0xc3]
+  %res = call <8 x i64> @llvm.x86.avx10.mask.vcvttps2uqqs.round.512(<8 x float> %x0, <8 x i64> %src, i8 %mask, i32 8)
+  ret <8 x i64> %res
+}
+
+define <8 x i64> @test_int_x86_mask_vcvtt_ps2uqqs_512_sae_12(<8 x float> %x0, <8 x i64> %src, i8 %mask) {
+; X64-LABEL: test_int_x86_mask_vcvtt_ps2uqqs_512_sae_12:
+; X64:       # %bb.0:
+; X64-NEXT:    kmovd %edi, %k1 # encoding: [0xc5,0xfb,0x92,0xcf]
+; X64-NEXT:    vcvttps2uqqs {sae}, %ymm0, %zmm1 {%k1} # encoding: [0x62,0xf5,0x7d,0x19,0x6c,0xc8]
+; X64-NEXT:    vmovaps %zmm1, %zmm0 # encoding: [0x62,0xf1,0x7c,0x48,0x28,0xc1]
+; X64-NEXT:    retq # encoding: [0xc3]
+;
+; X86-LABEL: test_int_x86_mask_vcvtt_ps2uqqs_512_sae_12:
+; X86:       # %bb.0:
+; X86-NEXT:    kmovb {{[0-9]+}}(%esp), %k1 # encoding: [0xc5,0xf9,0x90,0x4c,0x24,0x04]
+; X86-NEXT:    vcvttps2uqqs {sae}, %ymm0, %zmm1 {%k1} # encoding: [0x62,0xf5,0x7d,0x19,0x6c,0xc8]
+; X86-NEXT:    vmovaps %zmm1, %zmm0 # encoding: [0x62,0xf1,0x7c,0x48,0x28,0xc1]
+; X86-NEXT:    retl # encoding: [0xc3]
+  %res = call <8 x i64> @llvm.x86.avx10.mask.vcvttps2uqqs.round.512(<8 x float> %x0, <8 x i64> %src, i8 %mask, i32 12)
+  ret <8 x i64> %res
+}
+



More information about the llvm-branch-commits mailing list