static void
sse4_1_test (void)
{
union
{
__m128i x[NUM / 16];
char i[NUM];
} dst, src1, src2;
int i, sign = 1;
char min;
for (i = 0; i < NUM; i++)
{
src1.i[i] = i * i * sign;
src2.i[i] = (i + 20) * sign;
sign = -sign;
}
for (i = 0; i < NUM; i += 16)
dst.x[i / 16] = _mm_min_epi8 (src1.x[i / 16], src2.x[i / 16]);
for (i = 0; i < NUM; i++)
{
min = src1.i[i] >= src2.i[i] ? src2.i[i] : src1.i[i];
if (min != dst.i[i])
abort ();
}
}
void nibble_sort_tom(unsigned long *buf) {
for (int i = 0; i < TEST_SIZE; ++i) {
__m128i x = _mm_and_si128(_mm_set_epi64x(buf[i] >> 4, buf[i]), g_mask);
x = S(x, 0);
x = S(x, 1);
x = S(x, 0);
x = S(x, 2);
x = S(x, 3);
x = S(x, 0);
x = S(x, 4);
x = S(x, 5);
x = S(x, 3);
/* Final step is different; the output is in the right layout
* for reassembling for the final write. */
const __m128i a0 = _mm_shuffle_epi8(x, g_shuffles[0][0]);
const __m128i b0 = _mm_shuffle_epi8(x, g_shuffles[0][1]);
const __m128i a1 = _mm_min_epi8(a0, b0);
const __m128i b1 = _mm_max_epi8(a0, b0);
const __m128i out = _mm_or_si128(a1, _mm_slli_epi64(b1, 4));
_mm_storel_epi64((__m128i *)&buf[i], out);
}
}
static __m128i S(__m128i x, int i) {
const __m128i a0 = _mm_shuffle_epi8(x, g_shuffles[i][0]);
const __m128i b0 = _mm_shuffle_epi8(x, g_shuffles[i][1]);
const __m128i a1 = _mm_min_epi8(a0, b0);
const __m128i b1 = _mm_max_epi8(a0, b0);
const __m128i a2 = _mm_shuffle_epi8(a1, g_shuffles[i][2]);
const __m128i b2 = _mm_shuffle_epi8(b1, g_shuffles[i][3]);
return _mm_or_si128(a2, b2);
}
__m128i test_mm_min_epi8(__m128i x, __m128i y) {
// CHECK-LABEL: test_mm_min_epi8
// CHECK: call <16 x i8> @llvm.x86.sse41.pminsb
// CHECK-ASM: pminsb %xmm{{.*}}, %xmm{{.*}}
return _mm_min_epi8(x, y);
}
__m128i test_mm_min_epi8(__m128i x, __m128i y) {
// CHECK-LABEL: test_mm_min_epi8
// CHECK: [[CMP:%.*]] = icmp slt <16 x i8> [[X:%.*]], [[Y:%.*]]
// CHECK-NEXT: select <16 x i1> [[CMP]], <16 x i8> [[X]], <16 x i8> [[Y]]
return _mm_min_epi8(x, y);
}
