| /* APPLE LOCAL file 4736174 */ |
| /* { dg-do compile { target i?86-*-* } } */ |
| /* { dg-options "-O3 -msse2" } */ |
| typedef double __v2df __attribute__ ((__vector_size__ (16))); |
| typedef __v2df __m128d; |
| __m128d _mm_cmpgt_sd (__m128d __A, __m128d __B) |
| { |
| return (__m128d) __builtin_ia32_movsd ((__m128d) __A, |
| (__m128d) __builtin_ia32_cmpltsd ((__m128d) __B, |
| (__m128d) __A)); |
| } |
| |
| __m128d _mm_and_pd (__m128d __A, __m128d __B) |
| { |
| return (__m128d)__builtin_ia32_andpd ((__m128d)__A, (__m128d)__B); |
| } |
| |
| __m128d _mm_add_sd (__m128d __A, __m128d __B) |
| { |
| return (__m128d)__builtin_ia32_addsd ((__m128d)__A, (__m128d)__B); |
| } |
| |
| __m128d _mm_sub_sd (__m128d __A, __m128d __B) |
| { |
| return (__m128d)__builtin_ia32_subsd ((__m128d)__A, (__m128d)__B); |
| } |
| |
| __m128d _mm_mul_sd (__m128d __A, __m128d __B) |
| { |
| return (__m128d)__builtin_ia32_mulsd ((__m128d)__A, (__m128d)__B); |
| } |
| |
| __m128d _mm_storer_pd (double *__P, __m128d __A) |
| { |
| _mm_store_pd (__P, __builtin_ia32_shufpd (__A, __A, (((0) << 1) | (1)))); |
| } |
| |
| __m128d _mm_add_pd (__m128d __A, __m128d __B) |
| { |
| return (__m128d)__builtin_ia32_addpd ((__m128d)__A, (__m128d)__B); |
| } |
| |
| __m128d _mm_sub_pd (__m128d __A, __m128d __B) |
| { |
| return (__m128d)__builtin_ia32_subpd ((__m128d)__A, (__m128d)__B); |
| } |
| |
| void foo () |
| { |
| __m128d t1r, t1i, t2r, t2i, t3r, t3i, t5r, ttp4x, carry4x, bj4x; |
| __m128d MM_bigA, MM_bc[4], MM_c; |
| double * px, pd1, pd2, pd3; |
| int i, j, k, l, ll, UPDATE, pad2; |
| for (i = 0, j = 0; j < pad2; j += UPDATE, i++) |
| { |
| for (k = 0; k < UPDATE; k += 4, px += 8) |
| { |
| { |
| __m128d maskj; |
| int imaskj; |
| maskj = _mm_cmpgt_sd (bj4x, MM_c); |
| maskj = _mm_and_pd (maskj, ttp4x); |
| __asm__ volatile ("addsd %2, %0 \n" " subsd %2, %0" : "=&x" (carry4x) : "0" (t5r), "X" (MM_bigA)); |
| ttp4x = _mm_add_sd (ttp4x, maskj); |
| t5r = _mm_sub_sd (t5r, carry4x); |
| t5r = _mm_mul_sd (t5r, ttp4x); |
| bj4x = _mm_add_sd (bj4x, MM_bc[imaskj]); |
| } |
| { |
| __m128d _ar, _ai, _br, _bi; |
| t1r = _mm_sub_pd (t1r, t3i); |
| t1i = _mm_add_pd (t1i, t3r); |
| _mm_store_pd (pd1 + ll, t1r); |
| _mm_store_pd ((pd1 + ll + 2), t1i); |
| t3r = _mm_add_pd (t2r, _ai); |
| t3i = _mm_sub_pd (t2i, _ar); |
| t2r = _mm_sub_pd (t2r, _ai); |
| t2i = _mm_add_pd (t2i, _ar); |
| _mm_store_pd (pd3 + ll, t3r); |
| _mm_store_pd ((pd3 + ll + 2), t3i); |
| _mm_store_pd (pd2 + ll, t2r); |
| _mm_store_pd ((pd2 + ll + 2), t2i); |
| } |
| } |
| } |
| } |
