evas: add SSE3 op_blend_pixel blend functions

_op_blend_p_dp_sse3
_op_blend_pas_dp_sse3
_op_blend_rel_p_dp_sse3
_op_blend_rel_pan_dp_sse3

Patch by: Jim Kukunas <james.t.kukunas@linux.intel.com>

SVN revision: 63701

diff --git a/legacy/evas/src/lib/engines/common/evas_op_blend/op_blend_pixel_sse3.c b/legacy/evas/src/lib/engines/common/evas_op_blend/op_blend_pixel_sse3.c
new file mode 100644 (file)
index 0000000..e6fae52
--- /dev/null
+++ b/legacy/evas/src/lib/engines/common/evas_op_blend/op_blend_pixel_sse3.c
@@ -0,0 +1,314 @@
+/* blend pixel --> dst */
+
+#ifdef BUILD_SSE3
+
+static void
+_op_blend_p_dp_sse3(DATA32 *s, DATA8 *m __UNUSED__, DATA32 c __UNUSED__, DATA32 *d, int l) {
+
+   LOOP_ALIGNED_U1_A48_SSE3(d, l,
+      { /* UOP */
+
+         int alpha = 256 - (*s >> 24);
+         *d = *s + MUL_256(alpha, *d);
+         s++; d++; l--;
+      },
+      { /* A4OP */
+
+         __m128i s0 = _mm_lddqu_si128((__m128i *)s);
+         __m128i d0 = _mm_load_si128((__m128i *)d);
+
+         __m128i a0 = sub4_alpha_sse3(s0);
+         __m128i mul0 = mul_256_sse3(a0, d0);
+         d0 = _mm_add_epi32(mul0, s0);
+
+         _mm_store_si128((__m128i *)d, d0);
+
+         s += 4; d += 4; l -= 4;
+      },
+      { /* A8OP */
+
+         __m128i s0 = _mm_lddqu_si128((__m128i *)s);
+         __m128i d0 = _mm_load_si128((__m128i *)d);
+
+         __m128i s1 = _mm_lddqu_si128((__m128i *)(s+4));
+         __m128i d1 = _mm_load_si128((__m128i *)(d+4));
+
+         __m128i a0 = sub4_alpha_sse3(s0);
+         __m128i a1 = sub4_alpha_sse3(s1);
+
+         __m128i mul0 = mul_256_sse3(a0, d0);
+         __m128i mul1 = mul_256_sse3(a1, d1);
+
+         d0 = _mm_add_epi32(mul0, s0);
+         d1 = _mm_add_epi32(mul1, s1);
+
+         _mm_store_si128((__m128i *)d, d0);
+         _mm_store_si128((__m128i *)(d+4), d1);
+
+         s += 8; d += 8; l -= 8;
+      })
+}
+
+static void
+_op_blend_pas_dp_sse3(DATA32 *s, DATA8 *m __UNUSED__, DATA32 c __UNUSED__, DATA32 *d, int l) {
+
+   int alpha;
+
+   const __m128i zero = _mm_setzero_si128();
+
+   LOOP_ALIGNED_U1_A48_SSE3(d, l,
+      { /* UOP */
+         switch (*s & 0xff000000)
+           {
+           case 0:
+              break;
+           case 0xff000000:
+              *d = *s;
+              break;
+           default:
+              alpha = 256 - (*s >> 24);
+              *d = *s + MUL_256(alpha, *d);
+              break;
+           }
+         s++;  d++; l--;
+      },
+      { /* A4OP */
+
+         __m128i s0 = _mm_lddqu_si128((__m128i *)s);
+         __m128i d0 = _mm_load_si128((__m128i *)d);
+
+         __m128i a0 = sub4_alpha_sse3(s0);
+         __m128i mul0 = mul_256_sse3(a0, d0);
+
+         mul0 = _mm_add_epi32(s0, mul0);
+
+         __m128i zmask0 = _mm_cmpeq_epi32(_mm_srli_epi32(s0, 24), zero);
+         __m128i imask0  = ~zmask0;
+
+         mul0 = _mm_and_si128(imask0, mul0);
+         d0 = _mm_and_si128(zmask0, d0);
+
+         d0 = _mm_add_epi32(mul0, d0);
+
+         _mm_store_si128((__m128i *)d, d0);
+
+        s += 4; d += 4; l -= 4;
+      },
+      { /* A8OP */
+
+         __m128i s0 = _mm_lddqu_si128((__m128i *)s);
+         __m128i d0 = _mm_load_si128((__m128i *)d);
+
+         __m128i s1 = _mm_lddqu_si128((__m128i *)(s+4));
+         __m128i d1 = _mm_load_si128((__m128i *)(d+4));
+
+         __m128i a0 = sub4_alpha_sse3(s0);
+         __m128i a1 = sub4_alpha_sse3(s1);
+
+         __m128i mul0 = mul_256_sse3(a0, d0);
+         __m128i mul1 = mul_256_sse3(a1, d1);
+
+         mul0 = _mm_add_epi32(s0, mul0);
+         mul1 = _mm_add_epi32(s1, mul1);
+
+         __m128i zmask0 = _mm_cmpeq_epi32(_mm_srli_epi32(s0, 24), zero);
+         __m128i zmask1 = _mm_cmpeq_epi32(_mm_srli_epi32(s1, 24),  zero);
+
+         __m128i imask0 = ~zmask0;
+         __m128i imask1 = ~zmask1;
+
+         mul0 = _mm_and_si128(imask0, mul0);
+         d0 = _mm_and_si128(zmask0, d0);
+
+         mul1 = _mm_and_si128(imask1, mul1);
+         d1 = _mm_and_si128(zmask1, d1);
+
+         d0 = _mm_add_epi32(mul0, d0);
+         d1 = _mm_add_epi32(mul1, d1);
+
+         _mm_store_si128((__m128i *)d, d0);
+         _mm_store_si128((__m128i *)(d+4), d1);
+
+         s += 8; d += 8; l -= 8;
+      })
+}
+
+#define _op_blend_pan_dp_sse3 NULL
+
+#define _op_blend_p_dpan_sse3 _op_blend_p_dp_sse3
+#define _op_blend_pas_dpan_sse3 _op_blend_pas_dp_sse3
+#define _op_blend_pan_dpan_sse3 _op_blend_pan_dp_sse3
+
+static void
+init_blend_pixel_span_funcs_sse3(void)
+{
+   op_blend_span_funcs[SP][SM_N][SC_N][DP][CPU_SSE3] = _op_blend_p_dp_sse3;
+   op_blend_span_funcs[SP_AS][SM_N][SC_N][DP][CPU_SSE3] = _op_blend_pas_dp_sse3;
+   op_blend_span_funcs[SP_AN][SM_N][SC_N][DP][CPU_SSE3] = _op_blend_pan_dp_sse3;
+
+   op_blend_span_funcs[SP][SM_N][SC_N][DP_AN][CPU_SSE3] = _op_blend_p_dpan_sse3;
+   op_blend_span_funcs[SP_AS][SM_N][SC_N][DP_AN][CPU_SSE3] = _op_blend_pas_dpan_sse3;
+   op_blend_span_funcs[SP_AN][SM_N][SC_N][DP_AN][CPU_SSE3] = _op_blend_pan_dpan_sse3;
+}
+
+#define _op_blend_pt_p_dp_sse3 NULL
+
+#define _op_blend_pt_pas_dp_sse3 _op_blend_pt_p_dp_sse3
+#define _op_blend_pt_pan_dp_sse3 NULL
+
+#define _op_blend_pt_p_dpan_sse3 _op_blend_pt_p_dp_sse3
+#define _op_blend_pt_pan_dpan_sse3 _op_blend_pt_pan_dp_sse3
+#define _op_blend_pt_pas_dpan_sse3 _op_blend_pt_pas_dp_sse3
+
+static void
+init_blend_pixel_pt_funcs_sse3(void)
+{
+   op_blend_pt_funcs[SP][SM_N][SC_N][DP][CPU_SSE3] = _op_blend_pt_p_dp_sse3;
+   op_blend_pt_funcs[SP_AS][SM_N][SC_N][DP][CPU_SSE3] = _op_blend_pt_pas_dp_sse3;
+   op_blend_pt_funcs[SP_AN][SM_N][SC_N][DP][CPU_SSE3] = _op_blend_pt_pan_dp_sse3;
+
+   op_blend_pt_funcs[SP][SM_N][SC_N][DP_AN][CPU_SSE3] = _op_blend_pt_p_dpan_sse3;
+   op_blend_pt_funcs[SP_AS][SM_N][SC_N][DP_AN][CPU_SSE3] = _op_blend_pt_pas_dpan_sse3;
+   op_blend_pt_funcs[SP_AN][SM_N][SC_N][DP_AN][CPU_SSE3] = _op_blend_pt_pan_dpan_sse3;
+}
+
+/*-----*/
+
+/* blend_rel pixel -> dst */
+
+static void
+_op_blend_rel_p_dp_sse3(DATA32 *s, DATA8 *m __UNUSED__, DATA32 c, DATA32 *d, int l) {
+
+   const __m128i ones = _mm_set_epi32(1, 1, 1, 1);
+
+   LOOP_ALIGNED_U1_A48_SSE3(d, l,
+      { /* UOP */
+
+         int alpha = 256 - (*s >> 24);
+         c = 1 + (*d >> 24);
+         *d = MUL_256(c, *s) + MUL_256(alpha, *d);
+         d++; s++; l--;
+      },
+      { /*A4OP */
+
+         __m128i s0 = _mm_lddqu_si128((__m128i *)s);
+         __m128i d0 = _mm_load_si128((__m128i *)d);
+
+         __m128i c0 = _mm_add_epi32(_mm_srli_epi32(d0, 24), ones);
+         __m128i a0 = sub4_alpha_sse3(s0);
+
+         d0 = _mm_add_epi32(mul_256_sse3(c0, s0), mul_256_sse3(a0, d0));
+
+         _mm_store_si128((__m128i *)d, d0);
+
+         d += 4; s += 4; l -= 4;
+      },
+      { /* A8OP */
+
+         __m128i s0 = _mm_lddqu_si128((__m128i *)s);
+         __m128i d0 = _mm_load_si128 ((__m128i *)d);
+
+         __m128i s1 = _mm_lddqu_si128((__m128i *)(s+4));
+         __m128i d1 = _mm_load_si128 ((__m128i *)(d+4));
+
+         __m128i c0 = _mm_add_epi32(_mm_srli_epi32(d0, 24), ones);
+         __m128i c1 = _mm_add_epi32(_mm_srli_epi32(d1, 24), ones);
+
+         __m128i a0 = sub4_alpha_sse3(s0);
+         __m128i a1 = sub4_alpha_sse3(s1);
+
+         d0 = _mm_add_epi32(mul_256_sse3(c0, s0), mul_256_sse3(a0, d0));
+         d1 = _mm_add_epi32(mul_256_sse3(c1, s1), mul_256_sse3(a1, d1));
+
+         _mm_store_si128((__m128i *)d, d0);
+         _mm_store_si128((__m128i *)(d+4), d1);
+
+         d += 8; s += 8; l -= 8;
+      })
+}
+
+static void
+_op_blend_rel_pan_dp_sse3(DATA32 *s, DATA8 *m __UNUSED__, DATA32 c, DATA32 *d, int l) {
+
+   const __m128i ones = _mm_set_epi32(1, 1, 1, 1);
+
+   LOOP_ALIGNED_U1_A48_SSE3(d, l,
+      { /* UOP */
+
+         c = 1 + (*d >> 24);
+         *d++ = MUL_256(c, *s);
+         s++; l--;
+      },
+      { /* A4OP */
+
+         __m128i s0 = _mm_lddqu_si128((__m128i *)s);
+         __m128i d0 = _mm_load_si128((__m128i *)d);
+
+         __m128i c0 = _mm_add_epi32(_mm_srli_epi32(d0, 24), ones);
+         d0 = mul_256_sse3(c0, s0);
+
+         _mm_store_si128((__m128i *)d, d0);
+
+         d += 4; s += 4; l -= 4;
+      },
+      { /* A8OP */
+
+         __m128i s0 = _mm_lddqu_si128((__m128i *)s);
+         __m128i d0 = _mm_load_si128 ((__m128i *)d);
+
+         __m128i s1 = _mm_lddqu_si128((__m128i *)(s+4));
+         __m128i d1 = _mm_load_si128 ((__m128i *)(d+4));
+
+         __m128i c0 = _mm_add_epi32(_mm_srli_epi32(d0, 24), ones);
+         __m128i c1 = _mm_add_epi32(_mm_srli_epi32(d1, 24), ones);
+
+         d0 = mul_256_sse3(c0, s0);
+         d1 = mul_256_sse3(c1, s1);
+
+         _mm_store_si128((__m128i *)d, d0);
+         _mm_store_si128((__m128i *)(d+4), d1);
+
+         d += 8; s += 8; l -= 8;
+      })
+}
+
+#define _op_blend_rel_pas_dp_sse3 _op_blend_rel_p_dp_sse3
+
+#define _op_blend_rel_p_dpan_sse3 _op_blend_p_dpan_sse3
+#define _op_blend_rel_pan_dpan_sse3 _op_blend_pan_dpan_sse3
+#define _op_blend_rel_pas_dpan_sse3 _op_blend_pas_dpan_sse3
+
+static void
+init_blend_rel_pixel_span_funcs_sse3(void)
+{
+   op_blend_rel_span_funcs[SP][SM_N][SC_N][DP][CPU_SSE3] = _op_blend_rel_p_dp_sse3;
+   op_blend_rel_span_funcs[SP_AS][SM_N][SC_N][DP][CPU_SSE3] = _op_blend_rel_pas_dp_sse3;
+   op_blend_rel_span_funcs[SP_AN][SM_N][SC_N][DP][CPU_SSE3] = _op_blend_rel_pan_dp_sse3;
+
+   op_blend_rel_span_funcs[SP][SM_N][SC_N][DP_AN][CPU_SSE3] = _op_blend_rel_p_dpan_sse3;
+   op_blend_rel_span_funcs[SP_AS][SM_N][SC_N][DP_AN][CPU_SSE3] = _op_blend_rel_pas_dpan_sse3;
+   op_blend_rel_span_funcs[SP_AN][SM_N][SC_N][DP_AN][CPU_SSE3] = _op_blend_rel_pan_dpan_sse3;
+}
+
+#define _op_blend_rel_pt_p_dp_sse3 NULL
+#define _op_blend_rel_pt_pan_dp_sse3 NULL
+
+#define _op_blend_rel_pt_pas_dp_sse3 _op_blend_rel_pt_p_dp_sse3
+
+#define _op_blend_rel_pt_p_dpan_sse3 _op_blend_pt_p_dpan_sse3
+#define _op_blend_rel_pt_pan_dpan_sse3 _op_blend_pt_pan_dpan_sse3
+#define _op_blend_rel_pt_pas_dpan_sse3 _op_blend_pt_pas_dpan_sse3
+
+static void
+init_blend_rel_pixel_pt_funcs_sse3(void)
+{
+   op_blend_rel_pt_funcs[SP][SM_N][SC_N][DP][CPU_SSE3] = _op_blend_rel_pt_p_dp_sse3;
+   op_blend_rel_pt_funcs[SP_AS][SM_N][SC_N][DP][CPU_SSE3] = _op_blend_rel_pt_pas_dp_sse3;
+   op_blend_rel_pt_funcs[SP_AN][SM_N][SC_N][DP][CPU_SSE3] = _op_blend_rel_pt_pan_dp_sse3;
+
+   op_blend_rel_pt_funcs[SP][SM_N][SC_N][DP_AN][CPU_SSE3] = _op_blend_rel_pt_p_dpan_sse3;
+   op_blend_rel_pt_funcs[SP_AS][SM_N][SC_N][DP_AN][CPU_SSE3] = _op_blend_rel_pt_pas_dpan_sse3;
+   op_blend_rel_pt_funcs[SP_AN][SM_N][SC_N][DP_AN][CPU_SSE3] = _op_blend_rel_pt_pan_dpan_sse3;
+}
+
+#endif