[LLVMdev] [NVPTX] CUDA inline PTX asm definitions scoping "{" "}" is broken

Dmitry N. Mikushin maemarcus at gmail.com
Tue Jul 10 15:15:05 PDT 2012


Hi,

Looks like "{" and "}" are lost when trying to use the combination of Clang
and NVPTX, which may result into clash of definitions of the function-scope
and asm-scope. Here is an example:

> cat test.cu
__attribute__((device)) __attribute__((nv_linkonce_odr)) __inline__ int
__any(int a) {
    int result;
    asm __volatile__ ("{ \n\t"
        ".reg .pred \t%%p1; \n\t"
        ".reg .pred \t%%p2; \n\t"
        "setp.ne.u32 \t%%p1, %1, 0; \n\t"
        "vote.any.pred \t%%p2, %%p1; \n\t"
        "selp.s32 \t%0, 1, 0, %%p2; \n\t"
        "}" : "=r"(result) : "r"(a));
    return result;
}

> clang -cc1 -emit-llvm -fcuda-is-device -triple ptx64-unknown-unknown
test.cu -o test.ll
> cat test.ll
; ModuleID = 'test.cu'
target datalayout = "e-p:64:64-i64:64:64-f64:64:64-n1:8:16:32:64"
target triple = "ptx64-unknown-unknown"

define ptx_device i32 @_Z5__anyi(i32 %a) nounwind inlinehint {
entry:
  %a.addr = alloca i32, align 4
  %result = alloca i32, align 4
  store i32 %a, i32* %a.addr, align 4
  %0 = load i32* %a.addr, align 4
  %1 = call i32 asm sideeffect "$( \0A\09.reg .pred \09%p1; \0A\09.reg
.pred \09%p2; \0A\09setp.ne.u32 \09%p1, $1, 0; \0A\09vote.any.pred \09%p2,
%p1; \0A\09selp.s32 \09$0, 1, 0, %p2; \0A\09$)", "=r,r"(i32 %0) nounwind,
!srcloc !0
  store i32 %1, i32* %result, align 4
  %2 = load i32* %result, align 4
  ret i32 %2
}

!0 = metadata !{i32 127, i32 132, i32 166, i32 200, i32 242, i32 285, i32
327}

> llc -march=nvptx64 test.ll -o test.ptx
> cat test.ptx
//
// Generated by LLVM NVPTX Back-End
//

.version 3.0
.target sm_10, texmode_independent
.address_size 64


    // .globl    _Z5__anyi
.visible .global .align 4 .b8     __local_depot0[8];

.func  (.reg .b32 func_retval0) _Z5__anyi(
    .reg .b32 _Z5__anyi_param_0
)                                       // @_Z5__anyi
{
    .reg .b64     %SP;
    .reg .b64     %SPL;
    .reg .pred %p<396>;
    .reg .s16 %rc<396>;
    .reg .s16 %rs<396>;
    .reg .s32 %r<396>;
    .reg .s64 %rl<396>;
    .reg .f32 %f<396>;
    .reg .f64 %fl<396>;

// BB#0:                                // %entry
    mov.u64     %SP, __local_depot0;
    mov.b32    %r0, _Z5__anyi_param_0;
    st.global.u32     [%SP+0], %r0;
    // inline asm

    .reg .pred     %p1;
    .reg .pred     %p2;
    setp.ne.u32     %p1, %r0, 0;
    vote.any.pred     %p2, %p1;
    selp.s32     %r0, 1, 0, %p2;

    // inline asm
    st.global.u32     [%SP+4], %r0;
    mov.b32    func_retval0, %r0;
    ret;
}

> ptxas test.ptx -o test.cubin
ptxas test.ptx, line 33; error   : Duplicate definition of variable '%p1'
ptxas test.ptx, line 34; error   : Duplicate definition of variable '%p2'
ptxas test.ptx, line 36; error   : Instruction 'vote' requires .target
sm_12 or higher
ptxas fatal   : Ptx assembly aborted due to errors

- D.
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.llvm.org/pipermail/llvm-dev/attachments/20120711/56136610/attachment.html>


More information about the llvm-dev mailing list