; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py UTC_ARGS: --version 4 ; RUN: llc < %s -mtriple=nvptx64 -mcpu=sm_20 | FileCheck %s ; RUN: %if ptxas %{ llc < %s -mtriple=nvptx64 -mcpu=sm_20 | %ptxas-verify %} target triple = "nvptx64-nvidia-cuda" define i16 @bswap16(i16 %a) { ; CHECK-LABEL: bswap16( ; CHECK: { ; CHECK-NEXT: .reg .b32 %r<3>; ; CHECK-EMPTY: ; CHECK-NEXT: // %bb.0: ; CHECK-NEXT: ld.param.b16 %r1, [bswap16_param_0]; ; CHECK-NEXT: prmt.b32 %r2, %r1, 0, 0x7701U; ; CHECK-NEXT: st.param.b32 [func_retval0], %r2; ; CHECK-NEXT: ret; %b = tail call i16 @llvm.bswap.i16(i16 %a) ret i16 %b } define i32 @bswap32(i32 %a) { ; CHECK-LABEL: bswap32( ; CHECK: { ; CHECK-NEXT: .reg .b32 %r<3>; ; CHECK-EMPTY: ; CHECK-NEXT: // %bb.0: ; CHECK-NEXT: ld.param.b32 %r1, [bswap32_param_0]; ; CHECK-NEXT: prmt.b32 %r2, %r1, 0, 0x123U; ; CHECK-NEXT: st.param.b32 [func_retval0], %r2; ; CHECK-NEXT: ret; %b = tail call i32 @llvm.bswap.i32(i32 %a) ret i32 %b } define <2 x i16> @bswapv2i16(<2 x i16> %a) #0 { ; CHECK-LABEL: bswapv2i16( ; CHECK: { ; CHECK-NEXT: .reg .b32 %r<3>; ; CHECK-EMPTY: ; CHECK-NEXT: // %bb.0: ; CHECK-NEXT: ld.param.b32 %r1, [bswapv2i16_param_0]; ; CHECK-NEXT: prmt.b32 %r2, %r1, 0, 0x2301U; ; CHECK-NEXT: st.param.b32 [func_retval0], %r2; ; CHECK-NEXT: ret; %b = tail call <2 x i16> @llvm.bswap.v2i16(<2 x i16> %a) ret <2 x i16> %b } define i64 @bswap64(i64 %a) { ; CHECK-LABEL: bswap64( ; CHECK: { ; CHECK-NEXT: .reg .b32 %r<5>; ; CHECK-NEXT: .reg .b64 %rd<3>; ; CHECK-EMPTY: ; CHECK-NEXT: // %bb.0: ; CHECK-NEXT: ld.param.b64 %rd1, [bswap64_param_0]; ; CHECK-NEXT: mov.b64 {%r1, %r2}, %rd1; ; CHECK-NEXT: prmt.b32 %r3, %r1, 0, 0x123U; ; CHECK-NEXT: prmt.b32 %r4, %r2, 0, 0x123U; ; CHECK-NEXT: mov.b64 %rd2, {%r4, %r3}; ; CHECK-NEXT: st.param.b64 [func_retval0], %rd2; ; CHECK-NEXT: ret; %b = tail call i64 @llvm.bswap.i64(i64 %a) ret i64 %b } define <2 x i32> @bswapv2i32(<2 x i32> %a) { ; CHECK-LABEL: bswapv2i32( ; CHECK: { ; CHECK-NEXT: .reg .b32 %r<5>; ; CHECK-EMPTY: ; CHECK-NEXT: // %bb.0: ; CHECK-NEXT: ld.param.v2.b32 {%r1, %r2}, [bswapv2i32_param_0]; ; CHECK-NEXT: prmt.b32 %r3, %r2, 0, 0x123U; ; CHECK-NEXT: prmt.b32 %r4, %r1, 0, 0x123U; ; CHECK-NEXT: st.param.v2.b32 [func_retval0], {%r4, %r3}; ; CHECK-NEXT: ret; %b = tail call <2 x i32> @llvm.bswap.v2i32(<2 x i32> %a) ret <2 x i32> %b } declare i16 @llvm.bswap.i16(i16) declare i32 @llvm.bswap.i32(i32) declare <2 x i16> @llvm.bswap.v2i16(<2 x i16>) declare i64 @llvm.bswap.i64(i64) declare <2 x i32> @llvm.bswap.v2i32(<2 x i32>)