nvfx: support proper shader linkage - adds glsl support

7 files changed, 211 insertions, 88 deletions

diff --git a/src/gallium/drivers/nvfx/nvfx_context.h b/src/gallium/drivers/nvfx/nvfx_context.h
index 8899bf991e..7ec6a4f412 100644
--- a/src/gallium/drivers/nvfx/nvfx_context.h
+++ b/src/gallium/drivers/nvfx/nvfx_context.h
@@ -45,6 +45,7 @@
 #define NVFX_NEW_VERTCONST	(1 << 14)
 #define NVFX_NEW_FRAGCONST	(1 << 15)
 #define NVFX_NEW_INDEX	(1 << 16)
+#define NVFX_NEW_SPRITE  (1 << 17)
 
 struct nvfx_rasterizer_state {
 	struct pipe_rasterizer_state pipe;
diff --git a/src/gallium/drivers/nvfx/nvfx_fragprog.c b/src/gallium/drivers/nvfx/nvfx_fragprog.c
index ae4fe3aa26..0a599c62a7 100644
--- a/src/gallium/drivers/nvfx/nvfx_fragprog.c
+++ b/src/gallium/drivers/nvfx/nvfx_fragprog.c
@@ -16,8 +16,6 @@
 struct nvfx_fpc {
 	struct nvfx_fragment_program *fp;
 
-	uint attrib_map[PIPE_MAX_SHADER_INPUTS];
-
 	unsigned r_temps;
 	unsigned r_temps_discard;
 	struct nvfx_sreg r_result[PIPE_MAX_SHADER_OUTPUTS];
@@ -36,6 +34,8 @@ struct nvfx_fpc {
 
 	struct nvfx_sreg imm[MAX_IMM];
 	unsigned nr_imm;
+
+	unsigned char generic_to_slot[256]; /* semantic idx for each input semantic */
 };
 
 static INLINE struct nvfx_sreg
@@ -111,6 +111,11 @@ emit_src(struct nvfx_fpc *fpc, int pos, struct nvfx_sreg src)
 		sr |= (NVFX_FP_REG_TYPE_TEMP << NVFX_FP_REG_TYPE_SHIFT);
 		sr |= (src.index << NVFX_FP_REG_SRC_SHIFT);
 		break;
+	case NVFXSR_RELOCATED:
+		sr |= (NVFX_FP_REG_TYPE_INPUT << NVFX_FP_REG_TYPE_SHIFT);
+		//printf("adding relocation at %x for %x\n", fpc->inst_offset, src.index);
+		util_dynarray_append(&fpc->fp->slot_relocations[src.index], unsigned, fpc->inst_offset);
+		break;
 	case NVFXSR_CONST:
 		if (!fpc->have_const) {
 			grow_insns(fpc, 4);
@@ -241,8 +246,28 @@ tgsi_src(struct nvfx_fpc *fpc, const struct tgsi_full_src_register *fsrc)
 
 	switch (fsrc->Register.File) {
 	case TGSI_FILE_INPUT:
-		src = nvfx_sr(NVFXSR_INPUT,
-			      fpc->attrib_map[fsrc->Register.Index]);
+		if(fpc->fp->info.input_semantic_name[fsrc->Register.Index] == TGSI_SEMANTIC_POSITION) {
+			assert(fpc->fp->info.input_semantic_index[fsrc->Register.Index] == 0);
+			src = nvfx_sr(NVFXSR_INPUT, NVFX_FP_OP_INPUT_SRC_POSITION);
+		} else if(fpc->fp->info.input_semantic_name[fsrc->Register.Index] == TGSI_SEMANTIC_COLOR) {
+			if(fpc->fp->info.input_semantic_index[fsrc->Register.Index] == 0)
+				src = nvfx_sr(NVFXSR_INPUT, NVFX_FP_OP_INPUT_SRC_COL0);
+			else if(fpc->fp->info.input_semantic_index[fsrc->Register.Index] == 1)
+				src = nvfx_sr(NVFXSR_INPUT, NVFX_FP_OP_INPUT_SRC_COL1);
+			else
+				assert(0);
+		} else if(fpc->fp->info.input_semantic_name[fsrc->Register.Index] == TGSI_SEMANTIC_FOG) {
+			assert(fpc->fp->info.input_semantic_index[fsrc->Register.Index] == 0);
+			src = nvfx_sr(NVFXSR_INPUT, NVFX_FP_OP_INPUT_SRC_FOGC);
+		} else if(fpc->fp->info.input_semantic_name[fsrc->Register.Index] == TGSI_SEMANTIC_FACE) {
+			/* TODO: check this has the correct values */
+			/* XXX: what do we do for nv30 here (assuming it lacks facing)?!  */
+			assert(fpc->fp->info.input_semantic_index[fsrc->Register.Index] == 0);
+			src = nvfx_sr(NVFXSR_INPUT, NV40_FP_OP_INPUT_SRC_FACING);
+		} else {
+			assert(fpc->fp->info.input_semantic_name[fsrc->Register.Index] == TGSI_SEMANTIC_GENERIC);
+			src = nvfx_sr(NVFXSR_RELOCATED, fpc->generic_to_slot[fpc->fp->info.input_semantic_index[fsrc->Register.Index]]);
+		}
 		break;
 	case TGSI_FILE_CONSTANT:
 		src = constant(fpc, fsrc->Register.Index, NULL);
@@ -611,48 +636,6 @@ nvfx_fragprog_parse_instruction(struct nvfx_context* nvfx, struct nvfx_fpc *fpc,
 }
 
 static boolean
-nvfx_fragprog_parse_decl_attrib(struct nvfx_context* nvfx, struct nvfx_fpc *fpc,
-				const struct tgsi_full_declaration *fdec)
-{
-	int hw;
-
-	switch (fdec->Semantic.Name) {
-	case TGSI_SEMANTIC_POSITION:
-		hw = NVFX_FP_OP_INPUT_SRC_POSITION;
-		break;
-	case TGSI_SEMANTIC_COLOR:
-		if (fdec->Semantic.Index == 0) {
-			hw = NVFX_FP_OP_INPUT_SRC_COL0;
-		} else
-		if (fdec->Semantic.Index == 1) {
-			hw = NVFX_FP_OP_INPUT_SRC_COL1;
-		} else {
-			NOUVEAU_ERR("bad colour semantic index\n");
-			return FALSE;
-		}
-		break;
-	case TGSI_SEMANTIC_FOG:
-		hw = NVFX_FP_OP_INPUT_SRC_FOGC;
-		break;
-	case TGSI_SEMANTIC_GENERIC:
-		if (fdec->Semantic.Index <= 7) {
-			hw = NVFX_FP_OP_INPUT_SRC_TC(fdec->Semantic.
-						     Index);
-		} else {
-			NOUVEAU_ERR("bad generic semantic index\n");
-			return FALSE;
-		}
-		break;
-	default:
-		NOUVEAU_ERR("bad input semantic\n");
-		return FALSE;
-	}
-
-	fpc->attrib_map[fdec->Range.First] = hw;
-	return TRUE;
-}
-
-static boolean
 nvfx_fragprog_parse_decl_output(struct nvfx_context* nvfx, struct nvfx_fpc *fpc,
 				const struct tgsi_full_declaration *fdec)
 {
@@ -691,6 +674,15 @@ nvfx_fragprog_prepare(struct nvfx_context* nvfx, struct nvfx_fpc *fpc)
 {
 	struct tgsi_parse_context p;
 	int high_temp = -1, i;
+	struct util_semantic_set set;
+
+	fpc->fp->num_slots = util_semantic_set_from_program_file(&set, fpc->fp->pipe.tokens, TGSI_FILE_INPUT);
+	if(fpc->fp->num_slots > 8)
+		return FALSE;
+	util_semantic_layout_from_set(fpc->fp->slot_to_generic, &set, 0, 8);
+	util_semantic_table_from_layout(fpc->generic_to_slot, fpc->fp->slot_to_generic, 0, 8);
+
+	memset(fpc->fp->slot_to_fp_input, 0xff, sizeof(fpc->fp->slot_to_fp_input));
 
 	tgsi_parse_init(&p, fpc->fp->pipe.tokens);
 	while (!tgsi_parse_end_of_tokens(&p)) {
@@ -703,10 +695,6 @@ nvfx_fragprog_prepare(struct nvfx_context* nvfx, struct nvfx_fpc *fpc)
 			const struct tgsi_full_declaration *fdec;
 			fdec = &p.FullToken.FullDeclaration;
 			switch (fdec->Declaration.File) {
-			case TGSI_FILE_INPUT:
-				if (!nvfx_fragprog_parse_decl_attrib(nvfx, fpc, fdec))
-					goto out_err;
-				break;
 			case TGSI_FILE_OUTPUT:
 				if (!nvfx_fragprog_parse_decl_output(nvfx, fpc, fdec))
 					goto out_err;
@@ -805,7 +793,7 @@ nvfx_fragprog_translate(struct nvfx_context *nvfx,
 
 	/* Terminate final instruction */
 	if(fp->insn)
-                fp->insn[fpc->inst_offset] |= 0x00000001;
+		fp->insn[fpc->inst_offset] |= 0x00000001;
 
 	/* Append NOP + END instruction, may or may not be necessary. */
 	fpc->inst_offset = fp->insn_len;
@@ -881,9 +869,70 @@ nvfx_fragprog_validate(struct nvfx_context *nvfx)
 	if (nvfx->dirty & (NVFX_NEW_FRAGCONST | NVFX_NEW_FRAGPROG))
 		update = TRUE;
 
-	if(update) {
-		int offset;
+	struct nvfx_vertex_program* vp = nvfx->render_mode == HW ? nvfx->vertprog : nvfx->swtnl.vertprog;
+	if (fp->last_vp_id != vp->id) {
+		char* vp_sem_table = vp->generic_to_fp_input;
+		unsigned char* fp_semantics = fp->slot_to_generic;
+		unsigned diff = 0;
+		fp->last_vp_id = nvfx->vertprog->id;
+		unsigned char* cur_slots = fp->slot_to_fp_input;
+		for(unsigned i = 0; i < fp->num_slots; ++i) {
+			unsigned char slot_mask = vp_sem_table[fp_semantics[i]];
+			diff |= (slot_mask >> 4) & (slot_mask ^ cur_slots[i]);
+		}
+
+		if(diff)
+		{
+			for(unsigned i = 0; i < fp->num_slots; ++i) {
+				/* if 0xff, then this will write to the dummy value at fp->last_layout_mask[0] */
+				fp->slot_to_fp_input[i] = vp_sem_table[fp_semantics[i]] & 0xf;
+				//printf("fp: GENERIC[%i] from fpreg %i\n", fp_semantics[i], fp->slot_to_fp_input[i]);
+			}
+
+			fp->progs_left_with_obsolete_slot_assignments = fp->progs;
+			update = TRUE;
+		}
+	}
+
+	// last_sprite_coord_enable
+	unsigned sprite_coord_enable = nvfx->rasterizer->pipe.point_quad_rasterization * nvfx->rasterizer->pipe.sprite_coord_enable;
+	if(fp->last_sprite_coord_enable != sprite_coord_enable)
+	{
+		unsigned texcoord_mask = vp->texcoord_ouput_mask;
+		fp->last_sprite_coord_enable = sprite_coord_enable;
+		fp->point_sprite_control = 0;
+		for(unsigned i = 0; i < fp->num_slots; ++i) {
+			if((1 << fp->slot_to_generic[i]) & sprite_coord_enable)
+			{
+				unsigned fpin = fp->slot_to_fp_input[i];
+				//printf("sprite: slot %i generic %i had texcoord %i\n", i, fp->slot_to_generic[i], fpin - NVFX_FP_OP_INPUT_SRC_TC0);
+				if(fpin >= 0x0f)
+				{
+					unsigned tc = __builtin_ctz(~texcoord_mask);
+					texcoord_mask |= (1 << tc);
+					fp->slot_to_fp_input[i] = fpin = NVFX_FP_OP_INPUT_SRC_TC(tc);
+
+					fp->progs_left_with_obsolete_slot_assignments = fp->progs;
+					update = TRUE;
+				}
+				//printf("sprite: slot %i texcoord %i\n", i, fpin - NVFX_FP_OP_INPUT_SRC_TC0);
+				fp->point_sprite_control |= (1 << (fpin - NVFX_FP_OP_INPUT_SRC_TC0 + 8));
+			}
+			else
+			{
+				unsigned fpin = fp->slot_to_fp_input[i];
+				if(!(vp->texcoord_ouput_mask & (1 << (fpin - NVFX_FP_OP_INPUT_SRC_TC0))))
+				{
+					fp->slot_to_fp_input[i] = 0x0f;
 
+					fp->progs_left_with_obsolete_slot_assignments = fp->progs;
+					update = TRUE;
+				}
+			}
+		}
+	}
+
+	if(update) {
 		++fp->bo_prog_idx;
 		if(fp->bo_prog_idx >= fp->progs_per_bo)
 		{
@@ -893,10 +942,9 @@ nvfx_fragprog_validate(struct nvfx_context *nvfx)
 			}
 			else
 			{
-				struct nvfx_fragment_program_bo* fpbo = os_malloc_aligned(sizeof(struct nvfx_fragment_program) + fp->prog_size * fp->progs_per_bo, 16);
-				char *map, *buf;
-				int i;
-
+				struct nvfx_fragment_program_bo* fpbo = os_malloc_aligned(sizeof(struct nvfx_fragment_program) + (fp->prog_size + 8) * fp->progs_per_bo, 16);
+				fpbo->slots = (unsigned char*)&fpbo->insn[(fp->prog_size) * fp->progs_per_bo];
+				memset(fpbo->slots, 0, 8 * fp->progs_per_bo);
 				if(fp->fpbo)
 				{
 					fpbo->next = fp->fpbo->next;
@@ -906,12 +954,14 @@ nvfx_fragprog_validate(struct nvfx_context *nvfx)
 					fpbo->next = fpbo;
 				fp->fpbo = fpbo;
 				fpbo->bo = 0;
+				fp->progs += fp->progs_per_bo;
+				fp->progs_left_with_obsolete_slot_assignments += fp->progs_per_bo;
 				nouveau_bo_new(nvfx->screen->base.device, NOUVEAU_BO_VRAM | NOUVEAU_BO_MAP, 64, fp->prog_size * fp->progs_per_bo, &fpbo->bo);
 				nouveau_bo_map(fpbo->bo, NOUVEAU_BO_NOSYNC);
 
-				map = fpbo->bo->map;
-				buf = fpbo->insn;
-				for(i = 0; i < fp->progs_per_bo; ++i)
+				uint8_t* map = fpbo->bo->map;
+				uint8_t* buf = (uint8_t*)fpbo->insn;
+				for(unsigned i = 0; i < fp->progs_per_bo; ++i)
 				{
 					memcpy(buf, fp->insn, fp->insn_len * 4);
 					nvfx_fp_memcpy(map, fp->insn, fp->insn_len * 4);
@@ -922,7 +972,8 @@ nvfx_fragprog_validate(struct nvfx_context *nvfx)
 			fp->bo_prog_idx = 0;
 		}
 
-		offset = fp->bo_prog_idx * fp->prog_size;
+		int offset = fp->bo_prog_idx * fp->prog_size;
+		uint32_t* fpmap = (uint32_t*)((char*)fp->fpbo->bo->map + offset);
 
 		if(nvfx->constbuf[PIPE_SHADER_FRAGMENT]) {
 			struct pipe_resource* constbuf = nvfx->constbuf[PIPE_SHADER_FRAGMENT];
@@ -941,6 +992,25 @@ nvfx_fragprog_validate(struct nvfx_context *nvfx)
 				}
 			}
 		}
+
+		if(fp->progs_left_with_obsolete_slot_assignments) {
+			unsigned char* fpbo_slots = &fp->fpbo->slots[fp->bo_prog_idx * 8];
+			for(unsigned i = 0; i < fp->num_slots; ++i) {
+				unsigned value = fp->slot_to_fp_input[i];;
+				if(value != fpbo_slots[i]) {
+					unsigned* p = (unsigned*)fp->slot_relocations[i].data;
+					unsigned* pend = (unsigned*)((char*)fp->slot_relocations[i].data + fp->slot_relocations[i].size);
+					for(; p != pend; ++p) {
+						unsigned off = *p;
+						unsigned dw = fp->insn[off];
+						dw = (dw & ~NVFX_FP_OP_INPUT_SRC_MASK) | (value << NVFX_FP_OP_INPUT_SRC_SHIFT);
+						nvfx_fp_memcpy(&fpmap[*p], &dw, sizeof(dw));
+					}
+					fpbo_slots[i] = value;
+				}
+			}
+			--fp->progs_left_with_obsolete_slot_assignments;
+		}
 	}
 
 	if(update || (nvfx->dirty & NVFX_NEW_FRAGPROG)) {
@@ -960,6 +1030,13 @@ nvfx_fragprog_validate(struct nvfx_context *nvfx)
 			OUT_RING(chan, fp->samplers);
 		}
 	}
+
+	if(nvfx->dirty & (NVFX_NEW_FRAGPROG | NVFX_NEW_SPRITE))
+	{
+		WAIT_RING(chan, 2);
+		OUT_RING(chan, RING_3D(NV34TCL_POINT_SPRITE, 1));
+		OUT_RING(chan, fp->point_sprite_control | nvfx->rasterizer->pipe.point_quad_rasterization);
+	}
 }
 
 void
@@ -982,6 +1059,7 @@ void
 nvfx_fragprog_destroy(struct nvfx_context *nvfx,
 		      struct nvfx_fragment_program *fp)
 {
+	unsigned i;
 	struct nvfx_fragment_program_bo* fpbo = fp->fpbo;
 	if(fpbo)
 	{
@@ -996,7 +1074,9 @@ nvfx_fragprog_destroy(struct nvfx_context *nvfx,
 		while(fpbo != fp->fpbo);
 	}
 
+	for(i = 0; i < 8; ++i)
+		util_dynarray_fini(&fp->slot_relocations[i]);
+
 	if (fp->insn_len)
 		FREE(fp->insn);
 }
-
diff --git a/src/gallium/drivers/nvfx/nvfx_shader.h b/src/gallium/drivers/nvfx/nvfx_shader.h
index 50830b3916..88cf91b95f 100644
--- a/src/gallium/drivers/nvfx/nvfx_shader.h
+++ b/src/gallium/drivers/nvfx/nvfx_shader.h
@@ -323,6 +323,7 @@
 #define NVFXSR_INPUT	2
 #define NVFXSR_TEMP	3
 #define NVFXSR_CONST	4
+#define NVFXSR_RELOCATED	5
 
 #define NVFX_COND_FL  0
 #define NVFX_COND_LT  1
diff --git a/src/gallium/drivers/nvfx/nvfx_state.c b/src/gallium/drivers/nvfx/nvfx_state.c
index 25d29720a8..c3addf1114 100644
--- a/src/gallium/drivers/nvfx/nvfx_state.c
+++ b/src/gallium/drivers/nvfx/nvfx_state.c
@@ -91,6 +91,7 @@ nvfx_rasterizer_state_create(struct pipe_context *pipe,
 	/*XXX: ignored:
 	 * 	point_smooth -nohw
 	 * 	multisample
+	 *     sprite_coord_origin
 	 */
 
 	sb_method(sb, NV34TCL_SHADE_MODEL, 1);
@@ -150,20 +151,6 @@ nvfx_rasterizer_state_create(struct pipe_context *pipe,
 		sb_data(sb, fui(cso->offset_units * 2));
 	}
 
-	sb_method(sb, NV34TCL_POINT_SPRITE, 1);
-	if (cso->point_quad_rasterization) {
-		unsigned psctl = (1 << 0), i;
-
-		for (i = 0; i < 8; i++) {
-			if ((cso->sprite_coord_enable >> i) & 1)
-				psctl |= (1 << (8 + i));
-		}
-
-		sb_data(sb, psctl);
-	} else {
-		sb_data(sb, 0);
-	}
-
 	rsso->pipe = *cso;
 	rsso->sb_len = sb_len(sb, rsso->sb);
 	return (void *)rsso;
@@ -189,6 +176,12 @@ nvfx_rasterizer_state_bind(struct pipe_context *pipe, void *hwcso)
 			nvfx->dirty |= NVFX_NEW_STIPPLE;
 			nvfx->draw_dirty |= NVFX_NEW_STIPPLE;
 		}
+
+		if(((struct nvfx_rasterizer_state*)hwcso)->pipe.point_quad_rasterization != nvfx->rasterizer->pipe.point_quad_rasterization
+				|| ((struct nvfx_rasterizer_state*)hwcso)->pipe.sprite_coord_enable != nvfx->rasterizer->pipe.sprite_coord_enable)
+		{
+			nvfx->dirty |= NVFX_NEW_SPRITE;
+		}
 	}
 
 	nvfx->rasterizer = hwcso;
@@ -280,9 +273,13 @@ nvfx_vp_state_create(struct pipe_context *pipe,
 	struct nvfx_context *nvfx = nvfx_context(pipe);
 	struct nvfx_vertex_program *vp;
 
+	// TODO: use a 64-bit atomic here!
+	static unsigned long long id = 0;
+
 	vp = CALLOC(1, sizeof(struct nvfx_vertex_program));
 	vp->pipe.tokens = tgsi_dup_tokens(cso->tokens);
 	vp->draw = draw_create_vertex_shader(nvfx->draw, &vp->pipe);
+	vp->id = ++id;
 
 	return (void *)vp;
 }
diff --git a/src/gallium/drivers/nvfx/nvfx_state.h b/src/gallium/drivers/nvfx/nvfx_state.h
index 9ceb2577ec..e1fa3c7e04 100644
--- a/src/gallium/drivers/nvfx/nvfx_state.h
+++ b/src/gallium/drivers/nvfx/nvfx_state.h
@@ -4,6 +4,8 @@
 #include "pipe/p_state.h"
 #include "tgsi/tgsi_scan.h"
 #include "nouveau/nouveau_statebuf.h"
+#include "util/u_dynarray.h"
+#include "util/u_linkage.h"
 
 struct nvfx_vertex_program_exec {
 	uint32_t data[4];
@@ -18,6 +20,7 @@ struct nvfx_vertex_program_data {
 
 struct nvfx_vertex_program {
 	struct pipe_shader_state pipe;
+	unsigned long long id;
 
 	struct draw_vertex_shader *draw;
 
@@ -30,6 +33,9 @@ struct nvfx_vertex_program {
 	struct nvfx_vertex_program_data *consts;
 	unsigned nr_consts;
 
+	char generic_to_fp_input[256];
+	unsigned texcoord_ouput_mask;
+
 	struct nouveau_resource *exec;
 	unsigned exec_start;
 	struct nouveau_resource *data;
@@ -49,6 +55,7 @@ struct nvfx_fragment_program_data {
 struct nvfx_fragment_program_bo {
 	struct nvfx_fragment_program_bo* next;
 	struct nouveau_bo* bo;
+	unsigned char* slots;
 	char insn[] __attribute__((aligned(16)));
 };
 
@@ -58,6 +65,7 @@ struct nvfx_fragment_program {
 
 	boolean translated;
 	unsigned samplers;
+	unsigned point_sprite_control;
 
 	uint32_t *insn;
 	int       insn_len;
@@ -65,11 +73,27 @@ struct nvfx_fragment_program {
 	struct nvfx_fragment_program_data *consts;
 	unsigned nr_consts;
 
+	unsigned num_slots; /* how many input semantics? */
+	unsigned char slot_to_generic[8]; /* semantics */
+	unsigned char slot_to_fp_input[8]; /* current assignment of slots for each used semantic */
+	struct util_dynarray slot_relocations[8];
+
+	/* This is reset to progs on any relocation update, and decreases every time we
+	 * move to a new prog due to a constant update
+	 * When this is the same as progs, applying relocations is no longer necessary.
+	 */
+	unsigned progs_left_with_obsolete_slot_assignments;
+
+	unsigned long long last_vp_id;
+	unsigned last_sprite_coord_enable;
+
 	uint32_t fp_control;
 
 	unsigned bo_prog_idx;
 	unsigned prog_size;
 	unsigned progs_per_bo;
+	unsigned progs;
+
 	struct nvfx_fragment_program_bo* fpbo;
 };
 
diff --git a/src/gallium/drivers/nvfx/nvfx_state_emit.c b/src/gallium/drivers/nvfx/nvfx_state_emit.c
index b9d1897791..2e0e366ca3 100644
--- a/src/gallium/drivers/nvfx/nvfx_state_emit.c
+++ b/src/gallium/drivers/nvfx/nvfx_state_emit.c
@@ -15,6 +15,7 @@ nvfx_state_validate_common(struct nvfx_context *nvfx)
 	if(nvfx != nvfx->screen->cur_ctx)
 	{
 		nvfx->dirty = ~0;
+		nvfx->hw_vtxelt_nr = 16;
 		nvfx->screen->cur_ctx = nvfx;
 	}
 
@@ -86,7 +87,7 @@ nvfx_state_validate_common(struct nvfx_context *nvfx)
 	if(dirty & NVFX_NEW_STIPPLE)
 		nvfx_state_stipple_validate(nvfx);
 
-	if(dirty & (NVFX_NEW_FRAGPROG | NVFX_NEW_FRAGCONST))
+	if(dirty & (NVFX_NEW_FRAGPROG | NVFX_NEW_FRAGCONST | NVFX_NEW_VERTPROG | NVFX_NEW_SPRITE))
 	{
 		nvfx_fragprog_validate(nvfx);
 		if(dirty & NVFX_NEW_FRAGPROG)
diff --git a/src/gallium/drivers/nvfx/nvfx_vertprog.c b/src/gallium/drivers/nvfx/nvfx_vertprog.c
index 939d2b83ae..3d2f2b9fba 100644
--- a/src/gallium/drivers/nvfx/nvfx_vertprog.c
+++ b/src/gallium/drivers/nvfx/nvfx_vertprog.c
@@ -1,7 +1,7 @@
 #include "pipe/p_context.h"
 #include "pipe/p_defines.h"
 #include "pipe/p_state.h"
-#include "util/u_inlines.h"
+#include "util/u_linkage.h"
 
 #include "pipe/p_shader_tokens.h"
 #include "tgsi/tgsi_parse.h"
@@ -61,7 +61,7 @@ temp(struct nvfx_vpc *vpc)
 	return nvfx_sr(NVFXSR_TEMP, idx);
 }
 
-static INLINE void
+static inline void
 release_temps(struct nvfx_vpc *vpc)
 {
 	vpc->r_temps &= ~vpc->r_temps_discard;
@@ -339,7 +339,7 @@ nvfx_vp_arith(struct nvfx_context* nvfx, struct nvfx_vpc *vpc, int slot, int op,
 	emit_src(nvfx, vpc, hw, 2, s2);
 }
 
-static INLINE struct nvfx_sreg
+static inline struct nvfx_sreg
 tgsi_src(struct nvfx_vpc *vpc, const struct tgsi_full_src_register *fsrc) {
 	struct nvfx_sreg src = { 0 };
 
@@ -385,14 +385,14 @@ tgsi_dst(struct nvfx_vpc *vpc, const struct tgsi_full_dst_register *fdst) {
 		dst = vpc->r_address[fdst->Register.Index];
 		break;
 	default:
-		NOUVEAU_ERR("bad dst file\n");
+		NOUVEAU_ERR("bad dst file %i\n", fdst->Register.File);
 		break;
 	}
 
 	return dst;
 }
 
-static INLINE int
+static inline int
 tgsi_mask(uint tgsi)
 {
 	int mask = 0;
@@ -650,12 +650,8 @@ nvfx_vertprog_parse_decl_output(struct nvfx_context* nvfx, struct nvfx_vpc *vpc,
 		hw = NVFX_VP(INST_DEST_PSZ);
 		break;
 	case TGSI_SEMANTIC_GENERIC:
-		if (fdec->Semantic.Index <= 7) {
-			hw = NVFX_VP(INST_DEST_TC(fdec->Semantic.Index));
-		} else {
-			NOUVEAU_ERR("bad generic semantic index\n");
-			return FALSE;
-		}
+		hw = (vpc->vp->generic_to_fp_input[fdec->Semantic.Index] & 0xf)
+			+ NVFX_VP(INST_DEST_TC(0)) - NVFX_FP_OP_INPUT_SRC_TC(0);
 		break;
 	case TGSI_SEMANTIC_EDGEFLAG:
 		/* not really an error just a fallback */
@@ -675,6 +671,29 @@ nvfx_vertprog_prepare(struct nvfx_context* nvfx, struct nvfx_vpc *vpc)
 {
 	struct tgsi_parse_context p;
 	int high_temp = -1, high_addr = -1, nr_imm = 0, i;
+	struct util_semantic_set set;
+	unsigned char sem_layout[8];
+	unsigned sem_layout_size;
+	unsigned num_outputs;
+
+	num_outputs = util_semantic_set_from_program_file(&set, vpc->vp->pipe.tokens, TGSI_FILE_OUTPUT);
+
+	if(num_outputs > 8) {
+		NOUVEAU_ERR("too many vertex program outputs: %i\n", num_outputs);
+		return FALSE;
+	}
+	util_semantic_layout_from_set(sem_layout, &set, 8, 8);
+
+	/* hope 0xf is (0, 0, 0, 1) initialized; otherwise, we are _probably_ not required to do this */
+	memset(vpc->vp->generic_to_fp_input, 0x0f, sizeof(vpc->vp->generic_to_fp_input));
+	vpc->vp->texcoord_ouput_mask = 0;
+	for(int i = 0; i < 8; ++i) {
+		if(sem_layout[i] == 0xff)
+			continue;
+		vpc->vp->texcoord_ouput_mask |= (1 << i);
+		//printf("vp: GENERIC[%i] to fpreg %i\n", sem_layout[i], NVFX_FP_OP_INPUT_SRC_TC(0) + i);
+		vpc->vp->generic_to_fp_input[sem_layout[i]] = 0xf0 | (NVFX_FP_OP_INPUT_SRC_TC(0) + i);
+	}
 
 	tgsi_parse_init(&p, vpc->vp->pipe.tokens);
 	while (!tgsi_parse_end_of_tokens(&p)) {