jmp .L1_test
.p2align 4
-.loop1:
+.Loop1:
movlps 0(%rdi), %xmm0
shufps $0x50, %xmm0, %xmm0 # b01010000
addps %xmm0, %xmm4
.L1_test:
dec %rax
- jge .loop1
+ jge .Loop1
# set up for primary loop which is unrolled 4 times
movaps %xmm5, %xmm7
shr $2, %rdx # n_2_complex_blocks / 4
- je .cleanup # if zero, take short path
+ je .Lcleanup # if zero, take short path
# finish setup and loop priming
# hence enter loop at top
.p2align 4
-.loop2:
+.Loop2:
addps %xmm2, %xmm6
movlps 0x10(%rdi), %xmm2
add $0x40, %rsi
add $0x20, %rdi
dec %rdx
- jne .loop2
+ jne .Loop2
# OK, now we've done with all the multiplies, but
# we still need to handle the unaccumulated
# At this point, xmm4 contains 2x2 partial sums. We need
# to compute a "horizontal complex add" across xmm4.
-.cleanup: # xmm4 = r1 i2 r3 i4
+.Lcleanup: # xmm4 = r1 i2 r3 i4
movhlps %xmm4, %xmm0 # xmm0 = ?? ?? r1 r2
addps %xmm4, %xmm0 # xmm0 = ?? ?? r1+r3 i2+i4
movlps %xmm0, (%rcx) # store low 2x32 bits (complex) to memory
FUNC_TAIL(fcomplex_dotprod_sse)
.ident "Hand coded x86_64 SSE assembly"
+
+#if defined(__linux__) && defined(__ELF__)
+.section .note.GNU-stack,"",%progbits
+#endif