1*eddb79d5SFeng Zou // RUN: %clang_cc1 %s -flax-vector-conversions=none -ffreestanding -triple=x86_64-unknown-unknown \ 2*eddb79d5SFeng Zou // RUN: -target-feature +amx-tf32 -target-feature +amx-transpose \ 3*eddb79d5SFeng Zou // RUN: -target-feature +amx-bf16 -target-feature +avx512f \ 4*eddb79d5SFeng Zou // RUN: -emit-llvm -o - -Werror -pedantic | FileCheck %s 5*eddb79d5SFeng Zou 6*eddb79d5SFeng Zou #include <immintrin.h> 7*eddb79d5SFeng Zou 8*eddb79d5SFeng Zou char buf[1024]; 9*eddb79d5SFeng Zou #define STRIDE 32 10*eddb79d5SFeng Zou 11*eddb79d5SFeng Zou char buf2[1024]; 12*eddb79d5SFeng Zou 13*eddb79d5SFeng Zou void test_tile_mmultf32ps(__tile1024i a, __tile1024i b, __tile1024i c) { 14*eddb79d5SFeng Zou //CHECK-LABEL: @test_tile_mmultf32ps 15*eddb79d5SFeng Zou //CHECK-DAG: call x86_amx @llvm.x86.cast.vector.to.tile.v256i32(<256 x i32> {{%.*}}) 16*eddb79d5SFeng Zou //CHECK-DAG: call x86_amx @llvm.x86.tmmultf32ps.internal 17*eddb79d5SFeng Zou //CHECK-DAG: call <256 x i32> @llvm.x86.cast.tile.to.vector.v256i32(x86_amx {{%.*}}) 18*eddb79d5SFeng Zou __tile_mmultf32ps(&c, a, b); 19*eddb79d5SFeng Zou } 20*eddb79d5SFeng Zou 21*eddb79d5SFeng Zou void test_tile_tmmultf32ps(__tile1024i a, __tile1024i b, __tile1024i c) { 22*eddb79d5SFeng Zou //CHECK-LABEL: @test_tile_tmmultf32ps 23*eddb79d5SFeng Zou //CHECK-DAG: call x86_amx @llvm.x86.cast.vector.to.tile.v256i32(<256 x i32> {{%.*}}) 24*eddb79d5SFeng Zou //CHECK-DAG: call x86_amx @llvm.x86.ttmmultf32ps.internal 25*eddb79d5SFeng Zou //CHECK-DAG: call <256 x i32> @llvm.x86.cast.tile.to.vector.v256i32(x86_amx {{%.*}}) 26*eddb79d5SFeng Zou __tile_tmmultf32ps(&c, a, b); 27*eddb79d5SFeng Zou } 28