5 files changed, 96 insertions, 74 deletions

diff --git a/src/gallium/drivers/nv50/nv50_program.c b/src/gallium/drivers/nv50/nv50_program.c
index 0156ff95ff..97f938e698 100644
--- a/src/gallium/drivers/nv50/nv50_program.c
+++ b/src/gallium/drivers/nv50/nv50_program.c
@@ -4208,9 +4208,12 @@ static void
 nv50_program_validate_code(struct nv50_context *nv50, struct nv50_program *p)
 {
 	struct nouveau_channel *chan = nv50->screen->base.channel;
+	struct nouveau_grobj *tesla = nv50->screen->tesla;
 	struct nv50_program_exec *e;
 	uint32_t *up, i;
 	boolean upload = FALSE;
+	unsigned offset;
+	int width;
 
 	if (!p->bo) {
 		nouveau_bo_new(chan->device, NOUVEAU_BO_VRAM, 0x100,
@@ -4267,10 +4270,22 @@ nv50_program_validate_code(struct nv50_context *nv50, struct nv50_program *p)
 			NOUVEAU_ERR("0x%08x\n", e->inst[1]);
 	}
 #endif
-	nv50_upload_sifc(nv50, p->bo, 0, NOUVEAU_BO_VRAM,
-			 NV50_2D_DST_FORMAT_R8_UNORM, 65536, 1, 262144,
-			 up, NV50_2D_SIFC_FORMAT_R8_UNORM, 0,
-			 0, 0, p->exec_size * 4, 1, 1);
+
+	/* SIFC_HEIGHT/SIFC_WIDTH of 65536 do not work, and are not reported
+	 * as data error either. hw bug ? */
+#define SIFC_MAX_WIDTH (65536 - 256)
+	offset = 0;
+	width = p->exec_size * 4;
+	while (width > 0) {
+		nv50_upload_sifc(nv50, p->bo, offset, NOUVEAU_BO_VRAM,
+				 NV50_2D_DST_FORMAT_R8_UNORM, 65536, 1, 262144,
+				 &up[offset / 4], NV50_2D_SIFC_FORMAT_R8_UNORM,
+				 0, 0, 0, MIN2(SIFC_MAX_WIDTH, width), 1, 1);
+		width -= SIFC_MAX_WIDTH;
+		offset += SIFC_MAX_WIDTH;
+	}
+	BEGIN_RING(chan, tesla, NV50TCL_CODE_CB_FLUSH, 1);
+	OUT_RING  (chan, 0);
 
 	FREE(up);
 }
diff --git a/src/gallium/drivers/nv50/nv50_push.c b/src/gallium/drivers/nv50/nv50_push.c
index 244242b843..c3ac804146 100644
--- a/src/gallium/drivers/nv50/nv50_push.c
+++ b/src/gallium/drivers/nv50/nv50_push.c
@@ -108,7 +108,7 @@ emit_vertex(struct push_context *ctx, unsigned n)
    int i;
 
    if (ctx->edgeflag_attr < 16) {
-      float *edgeflag = ctx->attr[ctx->edgeflag_attr].map +
+      float *edgeflag = (uint8_t *)ctx->attr[ctx->edgeflag_attr].map +
                         ctx->attr[ctx->edgeflag_attr].stride * n;
 
       if (*edgeflag != ctx->edgeflag) {
@@ -120,7 +120,8 @@ emit_vertex(struct push_context *ctx, unsigned n)
 
    BEGIN_RING_NI(chan, tesla, NV50TCL_VERTEX_DATA, ctx->vtx_size);
    for (i = 0; i < ctx->attr_nr; i++)
-      ctx->attr[i].push(chan, ctx->attr[i].map + ctx->attr[i].stride * n);
+      ctx->attr[i].push(chan,
+			(uint8_t *)ctx->attr[i].map + ctx->attr[i].stride * n);
 }
 
 static void
@@ -243,14 +244,14 @@ nv50_push_elements_instanced(struct pipe_context *pipe,
          assert(bo->map);
          return;
       }
-      ctx.attr[n].map = bo->map + vb->buffer_offset + ve->src_offset;
+      ctx.attr[n].map = (uint8_t *)bo->map + vb->buffer_offset + ve->src_offset;
       nouveau_bo_unmap(bo);
 
       ctx.attr[n].stride = vb->stride;
       ctx.attr[n].divisor = ve->instance_divisor;
       if (ctx.attr[n].divisor) {
          ctx.attr[n].step = i_start % ve->instance_divisor;
-         ctx.attr[n].map += i_start * vb->stride;
+         ctx.attr[n].map = (uint8_t *)ctx.attr[n].map + i_start * vb->stride;
       }
 
       size = util_format_get_component_bits(ve->src_format,
@@ -331,7 +332,7 @@ nv50_push_elements_instanced(struct pipe_context *pipe,
               ctx.attr[i].divisor != ++ctx.attr[i].step)
             continue;
          ctx.attr[i].step = 0;
-         ctx.attr[i].map += ctx.attr[i].stride;
+         ctx.attr[i].map = (uint8_t *)ctx.attr[i].map + ctx.attr[i].stride;
       }
 
       u_split_prim_init(&s, mode, start, count);
diff --git a/src/gallium/drivers/nv50/nv50_screen.c b/src/gallium/drivers/nv50/nv50_screen.c
index a8cb1e25d8..2c0caada36 100644
--- a/src/gallium/drivers/nv50/nv50_screen.c
+++ b/src/gallium/drivers/nv50/nv50_screen.c
@@ -20,6 +20,7 @@
  * SOFTWARE.
  */
 
+#include "util/u_format_s3tc.h"
 #include "pipe/p_screen.h"
 
 #include "nv50_context.h"
@@ -64,6 +65,17 @@ nv50_screen_is_format_supported(struct pipe_screen *pscreen,
 			break;
 		}
 	} else {
+		if (tex_usage & PIPE_BIND_SAMPLER_VIEW) {
+			switch (format) {
+			case PIPE_FORMAT_DXT1_RGB:
+			case PIPE_FORMAT_DXT1_RGBA:
+			case PIPE_FORMAT_DXT3_RGBA:
+			case PIPE_FORMAT_DXT5_RGBA:
+				return util_format_s3tc_enabled;
+			default:
+				break;
+			}
+		}
 		switch (format) {
 		case PIPE_FORMAT_B8G8R8A8_UNORM:
 		case PIPE_FORMAT_B8G8R8X8_UNORM:
@@ -76,10 +88,6 @@ nv50_screen_is_format_supported(struct pipe_screen *pscreen,
 		case PIPE_FORMAT_A8_UNORM:
 		case PIPE_FORMAT_I8_UNORM:
 		case PIPE_FORMAT_L8A8_UNORM:
-		case PIPE_FORMAT_DXT1_RGB:
-		case PIPE_FORMAT_DXT1_RGBA:
-		case PIPE_FORMAT_DXT3_RGBA:
-		case PIPE_FORMAT_DXT5_RGBA:
 		case PIPE_FORMAT_S8_USCALED_Z24_UNORM:
 		case PIPE_FORMAT_Z24_UNORM_S8_USCALED:
 		case PIPE_FORMAT_Z32_FLOAT:
@@ -98,7 +106,7 @@ nv50_screen_is_format_supported(struct pipe_screen *pscreen,
 }
 
 static int
-nv50_screen_get_param(struct pipe_screen *pscreen, int param)
+nv50_screen_get_param(struct pipe_screen *pscreen, enum pipe_cap param)
 {
 	switch (param) {
 	case PIPE_CAP_MAX_TEXTURE_IMAGE_UNITS:
@@ -121,6 +129,8 @@ nv50_screen_get_param(struct pipe_screen *pscreen, int param)
 		return 8;
 	case PIPE_CAP_OCCLUSION_QUERY:
 		return 1;
+        case PIPE_CAP_TIMER_QUERY:
+		return 0;
 	case PIPE_CAP_TEXTURE_SHADOW_MAP:
 		return 1;
 	case PIPE_CAP_MAX_TEXTURE_2D_LEVELS:
@@ -146,6 +156,34 @@ nv50_screen_get_param(struct pipe_screen *pscreen, int param)
 	case PIPE_CAP_TGSI_FS_COORD_ORIGIN_LOWER_LEFT:
 	case PIPE_CAP_TGSI_FS_COORD_PIXEL_CENTER_INTEGER:
 		return 0;
+	case PIPE_CAP_MAX_VS_INSTRUCTIONS:
+	case PIPE_CAP_MAX_FS_INSTRUCTIONS:
+	case PIPE_CAP_MAX_VS_ALU_INSTRUCTIONS:
+	case PIPE_CAP_MAX_FS_ALU_INSTRUCTIONS:
+	case PIPE_CAP_MAX_VS_TEX_INSTRUCTIONS:
+	case PIPE_CAP_MAX_FS_TEX_INSTRUCTIONS:
+	case PIPE_CAP_MAX_VS_TEX_INDIRECTIONS:
+	case PIPE_CAP_MAX_FS_TEX_INDIRECTIONS: /* arbitrary limit */
+		return 16384;
+	case PIPE_CAP_MAX_VS_CONTROL_FLOW_DEPTH:
+	case PIPE_CAP_MAX_FS_CONTROL_FLOW_DEPTH: /* need stack bo */
+		return 4;
+	case PIPE_CAP_MAX_VS_INPUTS:
+		return 16;
+	case PIPE_CAP_MAX_FS_INPUTS: /* 128 / 4 with GP */
+		return 64 / 4;
+	case PIPE_CAP_MAX_VS_CONSTS:
+	case PIPE_CAP_MAX_FS_CONSTS:
+		return 65536 / 16;
+	case PIPE_CAP_MAX_VS_ADDRS:
+	case PIPE_CAP_MAX_FS_ADDRS: /* no spilling atm */
+		return 1;
+	case PIPE_CAP_MAX_VS_PREDS:
+	case PIPE_CAP_MAX_FS_PREDS: /* not yet handled */
+		return 0;
+	case PIPE_CAP_MAX_VS_TEMPS:
+	case PIPE_CAP_MAX_FS_TEMPS: /* no spilling atm */
+		return 128 / 4;
 	default:
 		NOUVEAU_ERR("Unknown PIPE_CAP %d\n", param);
 		return 0;
@@ -153,7 +191,7 @@ nv50_screen_get_param(struct pipe_screen *pscreen, int param)
 }
 
 static float
-nv50_screen_get_paramf(struct pipe_screen *pscreen, int param)
+nv50_screen_get_paramf(struct pipe_screen *pscreen, enum pipe_cap param)
 {
 	switch (param) {
 	case PIPE_CAP_MAX_LINE_WIDTH:
diff --git a/src/gallium/drivers/nv50/nv50_state.c b/src/gallium/drivers/nv50/nv50_state.c
index cf4105bcb7..f8bff764f2 100644
--- a/src/gallium/drivers/nv50/nv50_state.c
+++ b/src/gallium/drivers/nv50/nv50_state.c
@@ -345,7 +345,7 @@ nv50_rasterizer_state_create(struct pipe_context *pipe,
 		CALLOC_STRUCT(nv50_rasterizer_stateobj);
 
 	/*XXX: ignored
-	 * 	- light_twosize
+	 * 	- light_twoside
 	 * 	- point_smooth
 	 * 	- multisample
 	 * 	- point_sprite / sprite_coord_mode
@@ -385,72 +385,44 @@ nv50_rasterizer_state_create(struct pipe_context *pipe,
 	so_data  (so, cso->point_quad_rasterization ? 1 : 0);
 
 	so_method(so, tesla, NV50TCL_POLYGON_MODE_FRONT, 3);
-	if (cso->front_winding == PIPE_WINDING_CCW) {
-		so_data(so, nvgl_polygon_mode(cso->fill_ccw));
-		so_data(so, nvgl_polygon_mode(cso->fill_cw));
-	} else {
-		so_data(so, nvgl_polygon_mode(cso->fill_cw));
-		so_data(so, nvgl_polygon_mode(cso->fill_ccw));
-	}
+        so_data(so, nvgl_polygon_mode(cso->fill_front));
+        so_data(so, nvgl_polygon_mode(cso->fill_back));
 	so_data(so, cso->poly_smooth ? 1 : 0);
 
 	so_method(so, tesla, NV50TCL_CULL_FACE_ENABLE, 3);
-	so_data  (so, cso->cull_mode != PIPE_WINDING_NONE);
-	if (cso->front_winding == PIPE_WINDING_CCW) {
+	so_data  (so, cso->cull_face != PIPE_FACE_NONE);
+	if (cso->front_ccw) {
 		so_data(so, NV50TCL_FRONT_FACE_CCW);
-		switch (cso->cull_mode) {
-		case PIPE_WINDING_CCW:
-			so_data(so, NV50TCL_CULL_FACE_FRONT);
-			break;
-		case PIPE_WINDING_CW:
-			so_data(so, NV50TCL_CULL_FACE_BACK);
-			break;
-		case PIPE_WINDING_BOTH:
-			so_data(so, NV50TCL_CULL_FACE_FRONT_AND_BACK);
-			break;
-		default:
-			so_data(so, NV50TCL_CULL_FACE_BACK);
-			break;
-		}
-	} else {
+        }
+        else {
 		so_data(so, NV50TCL_FRONT_FACE_CW);
-		switch (cso->cull_mode) {
-		case PIPE_WINDING_CCW:
-			so_data(so, NV50TCL_CULL_FACE_BACK);
-			break;
-		case PIPE_WINDING_CW:
-			so_data(so, NV50TCL_CULL_FACE_FRONT);
-			break;
-		case PIPE_WINDING_BOTH:
-			so_data(so, NV50TCL_CULL_FACE_FRONT_AND_BACK);
-			break;
-		default:
-			so_data(so, NV50TCL_CULL_FACE_BACK);
-			break;
-		}
+        }
+	switch (cso->cull_face) {
+	case PIPE_FACE_FRONT:
+		so_data(so, NV50TCL_CULL_FACE_FRONT);
+		break;
+	case PIPE_FACE_BACK:
+		so_data(so, NV50TCL_CULL_FACE_BACK);
+		break;
+	case PIPE_FACE_FRONT_AND_BACK:
+		so_data(so, NV50TCL_CULL_FACE_FRONT_AND_BACK);
+		break;
+	default:
+		so_data(so, NV50TCL_CULL_FACE_BACK);
+		break;
 	}
 
 	so_method(so, tesla, NV50TCL_POLYGON_STIPPLE_ENABLE, 1);
 	so_data  (so, cso->poly_stipple_enable ? 1 : 0);
 
 	so_method(so, tesla, NV50TCL_POLYGON_OFFSET_POINT_ENABLE, 3);
-	if ((cso->offset_cw && cso->fill_cw == PIPE_POLYGON_MODE_POINT) ||
-	    (cso->offset_ccw && cso->fill_ccw == PIPE_POLYGON_MODE_POINT))
-		so_data(so, 1);
-	else
-		so_data(so, 0);
-	if ((cso->offset_cw && cso->fill_cw == PIPE_POLYGON_MODE_LINE) ||
-	    (cso->offset_ccw && cso->fill_ccw == PIPE_POLYGON_MODE_LINE))
-		so_data(so, 1);
-	else
-		so_data(so, 0);
-	if ((cso->offset_cw && cso->fill_cw == PIPE_POLYGON_MODE_FILL) ||
-	    (cso->offset_ccw && cso->fill_ccw == PIPE_POLYGON_MODE_FILL))
-		so_data(so, 1);
-	else
-		so_data(so, 0);
+        so_data(so, cso->offset_point);
+        so_data(so, cso->offset_line);
+        so_data(so, cso->offset_tri);
 
-	if (cso->offset_cw || cso->offset_ccw) {
+	if (cso->offset_point ||
+            cso->offset_line ||
+            cso->offset_tri) {
 		so_method(so, tesla, NV50TCL_POLYGON_OFFSET_FACTOR, 1);
 		so_data  (so, fui(cso->offset_scale));
 		so_method(so, tesla, NV50TCL_POLYGON_OFFSET_UNITS, 1);
diff --git a/src/gallium/drivers/nv50/nv50_transfer.c b/src/gallium/drivers/nv50/nv50_transfer.c
index c5581a6f9d..f973cf24b9 100644
--- a/src/gallium/drivers/nv50/nv50_transfer.c
+++ b/src/gallium/drivers/nv50/nv50_transfer.c
@@ -274,7 +274,6 @@ nv50_upload_sifc(struct nv50_context *nv50,
 {
 	struct nouveau_channel *chan = nv50->screen->base.channel;
 	struct nouveau_grobj *eng2d = nv50->screen->eng2d;
-	struct nouveau_grobj *tesla = nv50->screen->tesla;
 	unsigned line_dwords = (w * cpp + 3) / 4;
 
 	reloc |= NOUVEAU_BO_WR;
@@ -346,7 +345,4 @@ nv50_upload_sifc(struct nv50_context *nv50,
 
 		src = (uint8_t *) src + src_pitch;
 	}
-
-	BEGIN_RING(chan, tesla, NV50TCL_CODE_CB_FLUSH, 1);
-	OUT_RING  (chan, 0);
 }