https://gcc.gnu.org/bugzilla/show_bug.cgi?id=109249
Bug ID: 109249
Summary: Missed optimization for table lookups
Product: gcc
Version: 12.2.0
Status: UNCONFIRMED
Severity: normal
Priority: P3
Component: middle-end
Assignee: unassigned at gcc dot gnu.org
Reporter: andre.schackier at gmail dot com
Target Milestone: ---
Example #1 - https://godbolt.org/z/95GjPo9bs
gcc:
GetTypeDamageMultiplier(Type, Type):
sub rsp, 1184
movsx rax, edi
movsx rdx, esi
mov ecx, 162
lea rdi, [rsp-120]
mov esi, OFFSET FLAT:.LC0
lea rax, [rax+rax*8]
rep movsq
lea rax, [rdx+rax*2]
movss xmm0, DWORD PTR [rsp-120+rax*4]
add rsp, 1184
ret
clang:
GetTypeDamageMultiplier(Type, Type): # @GetTypeDamageMultiplier(Type, Type)
movsxd rax, edi
lea rax, [rax + 8*rax]
movsxd rcx, esi
lea rax, [rcx + 2*rax]
lea rcx, [rip + .L__const.GetTypeDamageMultiplier(Type,
Type).table]
movss xmm0, dword ptr [rcx + 4*rax] # xmm0 = mem[0],zero,zero,zero
ret
Example #2 - https://godbolt.org/z/sqocKjYco
gcc:
GetTypeDamageMultiplier(Type, Type):
sub rsp, 144
movsx rdi, edi
movsx rsi, esi
movaps xmm1, XMMWORD PTR .LC2[rip]
movss xmm0, DWORD PTR .LC1[rip]
movaps xmm2, XMMWORD PTR .LC6[rip]
lea rax, [rsi+rdi*8]
movaps XMMWORD PTR [rsp-72], xmm1
movaps xmm1, XMMWORD PTR .LC3[rip]
shufps xmm0, xmm0, 0
movaps XMMWORD PTR [rsp-120], xmm0
movaps XMMWORD PTR [rsp-104], xmm0
movaps XMMWORD PTR [rsp-88], xmm0
movaps XMMWORD PTR [rsp+8], xmm0
movaps xmm0, XMMWORD PTR .LC7[rip]
movaps XMMWORD PTR [rsp-56], xmm1
movaps xmm1, XMMWORD PTR .LC4[rip]
movaps XMMWORD PTR [rsp+24], xmm0
movaps xmm0, XMMWORD PTR .LC8[rip]
movaps XMMWORD PTR [rsp-40], xmm1
movaps xmm1, XMMWORD PTR .LC5[rip]
movaps XMMWORD PTR [rsp+40], xmm0
movaps xmm0, XMMWORD PTR .LC9[rip]
movaps XMMWORD PTR [rsp-24], xmm1
movaps XMMWORD PTR [rsp+56], xmm1
movaps xmm1, XMMWORD PTR .LC10[rip]
movaps XMMWORD PTR [rsp+72], xmm0
movaps XMMWORD PTR [rsp+104], xmm0
movaps xmm0, XMMWORD PTR .LC11[rip]
movaps XMMWORD PTR [rsp-8], xmm2
movaps XMMWORD PTR [rsp+88], xmm1
movaps XMMWORD PTR [rsp+120], xmm0
movss xmm0, DWORD PTR [rsp-120+rax*4]
add rsp, 144
ret
clang:
GetTypeDamageMultiplier(Type, Type): # @GetTypeDamageMultiplier(Type, Type)
movsxd rax, edi
movsxd rcx, esi
lea rax, [rcx + 8*rax]
lea rcx, [rip + .L__const.GetTypeDamageMultiplier(Type,
Type).table]
movss xmm0, dword ptr [rcx + 4*rax] # xmm0 = mem[0],zero,zero,zero
ret
