================
@@ -0,0 +1,909 @@
+; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py
+; RUN: llc < %s -verify-machineinstrs -mtriple=x86_64-unknown-unknown
--show-mc-encoding -mattr=+avx10v2aux | FileCheck %s --check-prefix=X64
+; RUN: llc < %s -verify-machineinstrs -mtriple=i686-unknown-unknown
--show-mc-encoding -mattr=+avx10v2aux | FileCheck %s --check-prefix=X86
+
+define <16 x i8> @test_int_x86_avx10_mask_vcvtps2bf8128(<4 x float> %b, <16 x
i8> %src0, i8 %mask) {
+; X64-LABEL: test_int_x86_avx10_mask_vcvtps2bf8128:
+; X64: # %bb.0:
+; X64-NEXT: kmovd %edi, %k1 # encoding: [0xc5,0xfb,0x92,0xcf]
+; X64-NEXT: vcvtps2bf8 %xmm0, %xmm1 {%k1} # encoding:
[0x62,0xf5,0x7e,0x09,0x39,0xc8]
+; X64-NEXT: vmovaps %xmm1, %xmm0 # EVEX TO VEX Compression encoding:
[0xc5,0xf8,0x28,0xc1]
+; X64-NEXT: retq # encoding: [0xc3]
+;
+; X86-LABEL: test_int_x86_avx10_mask_vcvtps2bf8128:
+; X86: # %bb.0:
+; X86-NEXT: kmovb {{[0-9]+}}(%esp), %k1 # encoding:
[0xc5,0xf9,0x90,0x4c,0x24,0x04]
+; X86-NEXT: vcvtps2bf8 %xmm0, %xmm1 {%k1} # encoding:
[0x62,0xf5,0x7e,0x09,0x39,0xc8]
+; X86-NEXT: vmovaps %xmm1, %xmm0 # EVEX TO VEX Compression encoding:
[0xc5,0xf8,0x28,0xc1]
+; X86-NEXT: retl # encoding: [0xc3]
+ %ret = call <16 x i8> @llvm.x86.avx10.mask.vcvtps2bf8128(<4 x float> %b, <16
x i8> %src0, i8 %mask)
+ ret <16 x i8> %ret
+}
+
+define <16 x i8> @test_int_x86_avx10_maskz_vcvtps2bf8128(<4 x float> %b, i8
%mask) {
+; X64-LABEL: test_int_x86_avx10_maskz_vcvtps2bf8128:
+; X64: # %bb.0:
+; X64-NEXT: kmovd %edi, %k1 # encoding: [0xc5,0xfb,0x92,0xcf]
+; X64-NEXT: vcvtps2bf8 %xmm0, %xmm0 {%k1} {z} # encoding:
[0x62,0xf5,0x7e,0x89,0x39,0xc0]
+; X64-NEXT: retq # encoding: [0xc3]
+;
+; X86-LABEL: test_int_x86_avx10_maskz_vcvtps2bf8128:
+; X86: # %bb.0:
+; X86-NEXT: kmovb {{[0-9]+}}(%esp), %k1 # encoding:
[0xc5,0xf9,0x90,0x4c,0x24,0x04]
+; X86-NEXT: vcvtps2bf8 %xmm0, %xmm0 {%k1} {z} # encoding:
[0x62,0xf5,0x7e,0x89,0x39,0xc0]
+; X86-NEXT: retl # encoding: [0xc3]
+ %ret = call <16 x i8> @llvm.x86.avx10.mask.vcvtps2bf8128(<4 x float> %b, <16
x i8> zeroinitializer, i8 %mask)
----------------
e-kud wrote:
Yeah, I faced with the same issue recently that we need new pattern to get rid
of `_mask` versions for `z` and generic mask when higher bits are zeroed. Ok
for now.
https://github.com/llvm/llvm-project/pull/206888
_______________________________________________
cfe-commits mailing list
[email protected]
https://lists.llvm.org/cgi-bin/mailman/listinfo/cfe-commits
