+
+#else
+
+#if defined(_ARCH_PWR8) && defined(PIPE_ARCH_LITTLE_ENDIAN)
+
+#include <altivec.h>
+#include "util/u_pwr8.h"
+
+static inline void
+build_masks_ppc(int c,
+ int cdiff,
+ int dcdx,
+ int dcdy,
+ unsigned *outmask,
+ unsigned *partmask)
+{
+ __m128i cstep0 = vec_setr_epi32(c, c+dcdx, c+dcdx*2, c+dcdx*3);
+ __m128i xdcdy = (__m128i) vec_splats(dcdy);
+
+ /* Get values across the quad
+ */
+ __m128i cstep1 = vec_add_epi32(cstep0, xdcdy);
+ __m128i cstep2 = vec_add_epi32(cstep1, xdcdy);
+ __m128i cstep3 = vec_add_epi32(cstep2, xdcdy);
+
+ {
+ __m128i cstep01, cstep23, result;
+
+ cstep01 = vec_packs_epi32(cstep0, cstep1);
+ cstep23 = vec_packs_epi32(cstep2, cstep3);
+ result = vec_packs_epi16(cstep01, cstep23);
+
+ *outmask |= vec_movemask_epi8(result);
+ }
+
+
+ {
+ __m128i cio4 = (__m128i) vec_splats(cdiff);
+ __m128i cstep01, cstep23, result;
+
+ cstep0 = vec_add_epi32(cstep0, cio4);
+ cstep1 = vec_add_epi32(cstep1, cio4);
+ cstep2 = vec_add_epi32(cstep2, cio4);
+ cstep3 = vec_add_epi32(cstep3, cio4);
+
+ cstep01 = vec_packs_epi32(cstep0, cstep1);
+ cstep23 = vec_packs_epi32(cstep2, cstep3);
+ result = vec_packs_epi16(cstep01, cstep23);
+
+ *partmask |= vec_movemask_epi8(result);
+ }
+}
+
+static inline unsigned
+build_mask_linear_ppc(int c, int dcdx, int dcdy)
+{
+ __m128i cstep0 = vec_setr_epi32(c, c+dcdx, c+dcdx*2, c+dcdx*3);
+ __m128i xdcdy = (__m128i) vec_splats(dcdy);
+
+ /* Get values across the quad
+ */
+ __m128i cstep1 = vec_add_epi32(cstep0, xdcdy);
+ __m128i cstep2 = vec_add_epi32(cstep1, xdcdy);
+ __m128i cstep3 = vec_add_epi32(cstep2, xdcdy);
+
+ /* pack pairs of results into epi16
+ */
+ __m128i cstep01 = vec_packs_epi32(cstep0, cstep1);
+ __m128i cstep23 = vec_packs_epi32(cstep2, cstep3);
+
+ /* pack into epi8, preserving sign bits
+ */
+ __m128i result = vec_packs_epi16(cstep01, cstep23);
+
+ /* extract sign bits to create mask
+ */
+ return vec_movemask_epi8(result);
+}
+
+static inline __m128i
+lp_plane_to_m128i(const struct lp_rast_plane *plane)
+{
+ return vec_setr_epi32((int32_t)plane->c, (int32_t)plane->dcdx,
+ (int32_t)plane->dcdy, (int32_t)plane->eo);
+}
+
+#define NR_PLANES 3
+
+void
+lp_rast_triangle_32_3_16(struct lp_rasterizer_task *task,
+ const union lp_rast_cmd_arg arg)
+{
+ const struct lp_rast_triangle *tri = arg.triangle.tri;
+ const struct lp_rast_plane *plane = GET_PLANES(tri);
+ int x = (arg.triangle.plane_mask & 0xff) + task->x;
+ int y = (arg.triangle.plane_mask >> 8) + task->y;
+ unsigned i, j;
+
+ struct { unsigned mask:16; unsigned i:8; unsigned j:8; } out[16];
+ unsigned nr = 0;
+
+ __m128i p0 = lp_plane_to_m128i(&plane[0]); /* c, dcdx, dcdy, eo */
+ __m128i p1 = lp_plane_to_m128i(&plane[1]); /* c, dcdx, dcdy, eo */
+ __m128i p2 = lp_plane_to_m128i(&plane[2]); /* c, dcdx, dcdy, eo */
+ __m128i zero = vec_splats((unsigned char) 0);
+
+ __m128i c;
+ __m128i dcdx;
+ __m128i dcdy;
+ __m128i rej4;
+
+ __m128i dcdx2;
+ __m128i dcdx3;
+
+ __m128i span_0; /* 0,dcdx,2dcdx,3dcdx for plane 0 */
+ __m128i span_1; /* 0,dcdx,2dcdx,3dcdx for plane 1 */
+ __m128i span_2; /* 0,dcdx,2dcdx,3dcdx for plane 2 */
+ __m128i unused;
+
+ __m128i vshuf_mask0;
+ __m128i vshuf_mask1;
+ __m128i vshuf_mask2;
+
+#ifdef PIPE_ARCH_LITTLE_ENDIAN
+ vshuf_mask0 = (__m128i) vec_splats((unsigned int) 0x03020100);
+ vshuf_mask1 = (__m128i) vec_splats((unsigned int) 0x07060504);
+ vshuf_mask2 = (__m128i) vec_splats((unsigned int) 0x0B0A0908);
+#else
+ vshuf_mask0 = (__m128i) vec_splats((unsigned int) 0x0C0D0E0F);
+ vshuf_mask1 = (__m128i) vec_splats((unsigned int) 0x08090A0B);
+ vshuf_mask2 = (__m128i) vec_splats((unsigned int) 0x04050607);