https://gcc.gnu.org/bugzilla/show_bug.cgi?id=89021
Bug ID: 89021 Summary: Implement mmintrin.h in SSE Product: gcc Version: 9.0 Status: UNCONFIRMED Severity: normal Priority: P3 Component: target Assignee: unassigned at gcc dot gnu.org Reporter: hjl.tools at gmail dot com CC: crazylht at gmail dot com, ubizjak at gmail dot com, xuepeng.guo at intel dot com Target Milestone: --- Target: x86-64 On x86-64, __m64 is returned and passed in XMM registers. We generate [hjl@gnu-cfl-1 pr88828]$ cat i.c #include <mmintrin.h> __m64 foo (__m64 x, __m64 y) { return _m_packssdw (x, y); } [hjl@gnu-cfl-1 pr88828]$ gcc -S -O2 i.c -march=haswell [hjl@gnu-cfl-1 pr88828]$ cat i.s .file "i.c" .text .p2align 4,,15 .globl foo .type foo, @function foo: .LFB129: .cfi_startproc movdq2q %xmm0, %mm0 movdq2q %xmm1, %mm1 packssdw %mm1, %mm0 movq2dq %mm0, %xmm0 ret .cfi_endproc .LFE129: .size foo, .-foo .ident "GCC: (GNU) 8.2.1 20190109 (Red Hat 8.2.1-7)" .section .note.GNU-stack,"",@progbits [hjl@gnu-cfl-1 pr88828]$ We can generate packssdw %xmm1, %xmm0 ret instead.