[llvm-bugs] [Bug 47023] New: Improve store merging

[via llvm-bugs]

https://bugs.llvm.org/show_bug.cgi?id=47023

            Bug ID: 47023
           Summary: Improve store merging
           Product: libraries
           Version: trunk
          Hardware: PC
                OS: Linux
            Status: NEW
          Severity: enhancement
          Priority: P
         Component: Common Code Generator Code
          Assignee: unassignedbugs at nondot.org
          Reporter: david.bolvansky at gmail.com
                CC: llvm-bugs at lists.llvm.org

struct data_t {
  char data[16];
};

void process(data_t);
void process3(data_t, data_t, data_t);

void bad1() { process({1, 2, 3, 4, 5}); }

void bad2() { process({1, 2, 3, 4, 5, 1, 2, 3, 4, 5, 1, 2, 3, 4, 5}); }

void bad3() {
  process3({1, 2, 3, 4, 5, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0},
           {11, 12, 13, 14, 15, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10},
           {21, 22, 23, 24, 25, 20, 20, 20, 10, 20, 20, 20, 20, 20, 20});
}

void bad4() {
  process3({1, 2, 3, 4, 5, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0},
           {11, 12, 13, 14, 15, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10},
           {21, 22, 23, 24, 25, 20, 20, 20, 10, 20, 20, 20, 20, 20, 20, 11});
}


Clang:
bad1():                               # @bad1()
        sub     rsp, 24
        mov     dword ptr [rsp + 8], 67305985
        mov     byte ptr [rsp + 12], 5
        mov     qword ptr [rsp + 13], 0
        mov     dword ptr [rsp + 20], 0
        mov     rdi, qword ptr [rsp + 8]
        xor     esi, esi
        call    process(data_t)
        add     rsp, 24
        ret
.LCPI1_0:
        .byte   1                               # 0x1
        .byte   2                               # 0x2
        .byte   3                               # 0x3
        .byte   4                               # 0x4
        .byte   5                               # 0x5
        .byte   1                               # 0x1
        .byte   2                               # 0x2
        .byte   3                               # 0x3
        .byte   4                               # 0x4
        .byte   5                               # 0x5
        .byte   1                               # 0x1
        .byte   2                               # 0x2
        .byte   3                               # 0x3
        .byte   4                               # 0x4
        .byte   5                               # 0x5
        .byte   0                               # 0x0
bad2():                               # @bad2()
        sub     rsp, 24
        movaps  xmm0, xmmword ptr [rip + .LCPI1_0] # xmm0 =
[1,2,3,4,5,1,2,3,4,5,1,2,3,4,5,0]
        movaps  xmmword ptr [rsp], xmm0
        mov     rdi, qword ptr [rsp]
        mov     rsi, qword ptr [rsp + 8]
        call    process(data_t)
        add     rsp, 24
        ret
bad3():                               # @bad3()
        sub     rsp, 24
        movabs  r8, 1446803478303675925
        mov     qword ptr [rsp + 8], r8
        mov     byte ptr [rsp + 16], 10
        mov     dword ptr [rsp + 17], 336860180
        mov     word ptr [rsp + 21], 5140
        mov     byte ptr [rsp + 23], 0
        mov     r9, qword ptr [rsp + 16]
        movabs  rdi, 21542142465
        movabs  rdx, 723401749922909195
        movabs  rcx, 723401728380766730
        mov     esi, 0
        call    process3(data_t, data_t, data_t)
        add     rsp, 24
        ret
bad4():                               # @bad4()
        movabs  rdi, 21542142465
        movabs  rdx, 723401749922909195
        movabs  rcx, 723401728380766730
        movabs  r8, 1446803478303675925
        movabs  r9, 798285110420182026
        mov     esi, 0
        jmp     process3(data_t, data_t, data_t)   


GCC:
bad1():
        movabs  rdi, 21542142465
        xor     esi, esi
        jmp     process(data_t)
bad2():
        mov     rdi, QWORD PTR .LC0[rip]
        movabs  rsi, 1411785848587524
        jmp     process(data_t)
bad3():
        mov     r8, QWORD PTR .LC1[rip]
        xor     esi, esi
        movabs  r9, 5651576002974730
        movabs  rdx, 723401749922909195
        movabs  rcx, 723401728380766730
        movabs  rdi, 21542142465
        jmp     process3(data_t, data_t, data_t)
bad4():
        movabs  r8, 1446803478303675925
        xor     esi, esi
        movabs  r9, 798285110420182026
        movabs  rdx, 723401749922909195
        movabs  rcx, 723401728380766730
        movabs  rdi, 21542142465
        jmp     process3(data_t, data_t, data_t)
.LC0:
        .byte   1
        .byte   2
        .byte   3
        .byte   4
        .byte   5
        .byte   1
        .byte   2
        .byte   3
.LC1:
        .byte   21
        .byte   22
        .byte   23
        .byte   24
        .byte   25
        .byte   20
        .byte   20
        .byte   20


GCC's codegen is better here..

Godbolt: https://godbolt.org/z/xGY79q

-- 
You are receiving this mail because:
You are on the CC list for the bug.
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.llvm.org/pipermail/llvm-bugs/attachments/20200806/1b54257b/attachment.html>