nvfx: support nv30 simulation on nv40

7 files changed, 50 insertions, 32 deletions

diff --git a/src/gallium/drivers/nvfx/nvfx_context.c b/src/gallium/drivers/nvfx/nvfx_context.c
index 9baf78b02a1..62d623b9ffe 100644
--- a/src/gallium/drivers/nvfx/nvfx_context.c
+++ b/src/gallium/drivers/nvfx/nvfx_context.c
@@ -77,6 +77,7 @@ nvfx_create(struct pipe_screen *pscreen, void *priv)
 	nvfx->pipe.flush = nvfx_flush;
 
 	nvfx->is_nv4x = screen->is_nv4x;
+	nvfx->use_nv4x = screen->use_nv4x;
 	/* TODO: it seems that nv30 might have fixed function clipping usable with vertex programs
 	 * However, my code for that doesn't work, so use vp clipping for all cards, which works.
 	 */
diff --git a/src/gallium/drivers/nvfx/nvfx_context.h b/src/gallium/drivers/nvfx/nvfx_context.h
index 48a9cd2b7c6..6ef2a6945d7 100644
--- a/src/gallium/drivers/nvfx/nvfx_context.h
+++ b/src/gallium/drivers/nvfx/nvfx_context.h
@@ -141,6 +141,7 @@ struct nvfx_context {
 	struct nvfx_screen *screen;
 
 	unsigned is_nv4x; /* either 0 or ~0 */
+	unsigned use_nv4x; /* either 0 or ~0 */
 	boolean use_vp_clipping;
 
 	struct draw_context *draw;
diff --git a/src/gallium/drivers/nvfx/nvfx_fragprog.c b/src/gallium/drivers/nvfx/nvfx_fragprog.c
index 44c7968d88e..fc9d5b96cef 100644
--- a/src/gallium/drivers/nvfx/nvfx_fragprog.c
+++ b/src/gallium/drivers/nvfx/nvfx_fragprog.c
@@ -828,7 +828,7 @@ nvfx_fragprog_parse_instruction(struct nvfx_context* nvfx, struct nvfx_fpc *fpc,
 	case TGSI_OPCODE_IF:
 		// MOVRC0 R31 (TR0.xyzw), R<src>:
 		// IF (NE.xxxx) ELSE <else> END <end>
-		if(!nvfx->is_nv4x)
+		if(!nvfx->use_nv4x)
 			goto nv3x_cflow;
 		nv40_fp_if(fpc, src[0]);
 		break;
@@ -836,7 +836,7 @@ nvfx_fragprog_parse_instruction(struct nvfx_context* nvfx, struct nvfx_fpc *fpc,
 	case TGSI_OPCODE_ELSE:
 	{
 		uint32_t *hw;
-		if(!nvfx->is_nv4x)
+		if(!nvfx->use_nv4x)
 			goto nv3x_cflow;
 		assert(util_dynarray_contains(&fpc->if_stack, unsigned));
 		hw = &fpc->fp->insn[util_dynarray_top(&fpc->if_stack, unsigned)];
@@ -847,7 +847,7 @@ nvfx_fragprog_parse_instruction(struct nvfx_context* nvfx, struct nvfx_fpc *fpc,
 	case TGSI_OPCODE_ENDIF:
 	{
 		uint32_t *hw;
-		if(!nvfx->is_nv4x)
+		if(!nvfx->use_nv4x)
 			goto nv3x_cflow;
 		assert(util_dynarray_contains(&fpc->if_stack, unsigned));
 		hw = &fpc->fp->insn[util_dynarray_pop(&fpc->if_stack, unsigned)];
@@ -870,19 +870,19 @@ nvfx_fragprog_parse_instruction(struct nvfx_context* nvfx, struct nvfx_fpc *fpc,
 		break;
 
 	case TGSI_OPCODE_CAL:
-		if(!nvfx->is_nv4x)
+		if(!nvfx->use_nv4x)
 			goto nv3x_cflow;
 		nv40_fp_cal(fpc, finst->Label.Label);
 		break;
 
 	case TGSI_OPCODE_RET:
-		if(!nvfx->is_nv4x)
+		if(!nvfx->use_nv4x)
 			goto nv3x_cflow;
 		nv40_fp_ret(fpc);
 		break;
 
 	case TGSI_OPCODE_BGNLOOP:
-		if(!nvfx->is_nv4x)
+		if(!nvfx->use_nv4x)
 			goto nv3x_cflow;
 		/* TODO: we should support using two nested REPs to allow a > 255 iteration count */
 		nv40_fp_rep(fpc, 255, finst->Label.Label);
@@ -892,7 +892,7 @@ nvfx_fragprog_parse_instruction(struct nvfx_context* nvfx, struct nvfx_fpc *fpc,
 		break;
 
 	case TGSI_OPCODE_BRK:
-		if(!nvfx->is_nv4x)
+		if(!nvfx->use_nv4x)
 			goto nv3x_cflow;
 		nv40_fp_brk(fpc);
 		break;
@@ -947,7 +947,7 @@ nvfx_fragprog_parse_decl_output(struct nvfx_context* nvfx, struct nvfx_fpc *fpc,
 		case 2: hw = 3; break;
 		case 3: hw = 4; break;
 		}
-		if(hw > ((nvfx->is_nv4x) ? 4 : 2)) {
+		if(hw > ((nvfx->use_nv4x) ? 4 : 2)) {
 			NOUVEAU_ERR("bad rcol index\n");
 			return FALSE;
 		}
@@ -968,7 +968,7 @@ 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;
-	unsigned num_texcoords = nvfx->is_nv4x ? 10 : 8;
+	unsigned num_texcoords = nvfx->use_nv4x ? 10 : 8;
 
 	fpc->fp->num_slots = util_semantic_set_from_program_file(&set, fpc->pfp->pipe.tokens, TGSI_FILE_INPUT);
 	if(fpc->fp->num_slots > num_texcoords)
@@ -1062,7 +1062,7 @@ nvfx_fragprog_translate(struct nvfx_context *nvfx,
 	if (!fpc)
 		goto out_err;
 
-	fpc->max_temps = nvfx->is_nv4x ? 48 : 32;
+	fpc->max_temps = nvfx->use_nv4x ? 48 : 32;
 	fpc->pfp = pfp;
 	fpc->fp = fp;
 	fpc->num_regs = 2;
diff --git a/src/gallium/drivers/nvfx/nvfx_screen.c b/src/gallium/drivers/nvfx/nvfx_screen.c
index b72e6aa1a85..0290370d989 100644
--- a/src/gallium/drivers/nvfx/nvfx_screen.c
+++ b/src/gallium/drivers/nvfx/nvfx_screen.c
@@ -27,7 +27,7 @@ nvfx_screen_get_param(struct pipe_screen *pscreen, enum pipe_cap param)
 	case PIPE_CAP_MAX_TEXTURE_IMAGE_UNITS:
 		return 16;
 	case PIPE_CAP_NPOT_TEXTURES:
-		return !!screen->is_nv4x;
+		return screen->advertise_npot;
 	case PIPE_CAP_TWO_SIDED_STENCIL:
 		return 1;
 	case PIPE_CAP_GLSL:
@@ -37,7 +37,7 @@ nvfx_screen_get_param(struct pipe_screen *pscreen, enum pipe_cap param)
 	case PIPE_CAP_POINT_SPRITE:
 		return 1;
 	case PIPE_CAP_MAX_RENDER_TARGETS:
-		return screen->is_nv4x ? 4 : 2;
+		return screen->use_nv4x ? 4 : 2;
 	case PIPE_CAP_OCCLUSION_QUERY:
 		return 1;
         case PIPE_CAP_TIMER_QUERY:
@@ -53,7 +53,7 @@ nvfx_screen_get_param(struct pipe_screen *pscreen, enum pipe_cap param)
 	case PIPE_CAP_MAX_TEXTURE_CUBE_LEVELS:
 		return 13;
 	case PIPE_CAP_TEXTURE_MIRROR_CLAMP:
-		return !!screen->is_nv4x;
+		return !!screen->use_nv4x;
 	case PIPE_CAP_TEXTURE_MIRROR_REPEAT:
 		return 1;
 	case PIPE_CAP_MAX_VERTEX_TEXTURE_UNITS:
@@ -61,7 +61,7 @@ nvfx_screen_get_param(struct pipe_screen *pscreen, enum pipe_cap param)
 	case PIPE_CAP_TGSI_CONT_SUPPORTED:
 		return 0;
 	case PIPE_CAP_BLEND_EQUATION_SEPARATE:
-		return !!screen->is_nv4x;
+		return screen->advertise_blend_equation_separate;
 	case PIPE_CAP_MAX_COMBINED_SAMPLERS:
 		return 16;
 	case PIPE_CAP_INDEP_BLEND_ENABLE:
@@ -85,35 +85,35 @@ nvfx_screen_get_param(struct pipe_screen *pscreen, enum pipe_cap param)
 	case PIPE_CAP_MAX_FS_CONTROL_FLOW_DEPTH:
 		/* FIXME: is it the dynamic (nv30:0/nv40:24) or the static
 		   value (nv30:0/nv40:4) ? */
-		return screen->is_nv4x ? 4 : 0;
+		return screen->use_nv4x ? 4 : 0;
 	case PIPE_CAP_MAX_FS_INPUTS:
-		return screen->is_nv4x ? 12 : 10;
+		return screen->use_nv4x ? 12 : 10;
 	case PIPE_CAP_MAX_FS_CONSTS:
-		return screen->is_nv4x ? 224 : 32;
+		return screen->use_nv4x ? 224 : 32;
 	case PIPE_CAP_MAX_FS_TEMPS:
 		return 32;
 	case PIPE_CAP_MAX_FS_ADDRS:
-		return screen->is_nv4x ? 1 : 0;
+		return screen->use_nv4x ? 1 : 0;
 	case PIPE_CAP_MAX_FS_PREDS:
 		return 0; /* we could expose these, but nothing uses them */
 	case PIPE_CAP_MAX_VS_INSTRUCTIONS:
 	case PIPE_CAP_MAX_VS_ALU_INSTRUCTIONS:
-		return screen->is_nv4x ? 512 : 256;
+		return screen->use_nv4x ? 512 : 256;
 	case PIPE_CAP_MAX_VS_TEX_INSTRUCTIONS:
 	case PIPE_CAP_MAX_VS_TEX_INDIRECTIONS:
-		return screen->is_nv4x ? 512 : 0;
+		return screen->use_nv4x ? 512 : 0;
 	case PIPE_CAP_MAX_VS_CONTROL_FLOW_DEPTH:
 		/* FIXME: is it the dynamic (nv30:24/nv40:24) or the static
 		   value (nv30:1/nv40:4) ? */
-		return screen->is_nv4x ? 4 : 1;
+		return screen->use_nv4x ? 4 : 1;
 	case PIPE_CAP_MAX_VS_INPUTS:
 		return 16;
 	case PIPE_CAP_MAX_VS_CONSTS:
 		/* - 6 is for clip planes; Gallium should be fixed to put
 		 * them in the vertex shader itself, so we don't need to reserve these */
-		return (screen->is_nv4x ? 468 : 256) - 6;
+		return (screen->use_nv4x ? 468 : 256) - 6;
 	case PIPE_CAP_MAX_VS_TEMPS:
-		return screen->is_nv4x ? 32 : 13;
+		return screen->use_nv4x ? 32 : 13;
 	case PIPE_CAP_MAX_VS_ADDRS:
 		return 2;
 	case PIPE_CAP_MAX_VS_PREDS:
@@ -141,7 +141,7 @@ nvfx_screen_get_paramf(struct pipe_screen *pscreen, enum pipe_cap param)
 	case PIPE_CAP_MAX_POINT_WIDTH_AA:
 		return 64.0;
 	case PIPE_CAP_MAX_TEXTURE_ANISOTROPY:
-		return screen->is_nv4x ? 16.0 : 8.0;
+		return screen->use_nv4x ? 16.0 : 8.0;
 	case PIPE_CAP_MAX_TEXTURE_LOD_BIAS:
 		return 15.0;
 	default:
@@ -200,7 +200,7 @@ nvfx_screen_is_format_supported(struct pipe_screen *pscreen,
 			return FALSE;
 		if(format == PIPE_FORMAT_R32G32B32A32_FLOAT && !screen->advertise_fp32)
 			return FALSE;
-		if(screen->is_nv4x)
+		if(screen->use_nv4x)
 		{
 			if(tf->fmt[4] < 0)
 				return FALSE;
@@ -432,6 +432,19 @@ nvfx_screen_create(struct pipe_winsys *ws, struct nouveau_device *dev)
 		return NULL;
 	}
 
+	screen->advertise_npot = !!screen->is_nv4x;
+	screen->advertise_blend_equation_separate = !!screen->is_nv4x;
+	screen->use_nv4x = screen->is_nv4x;
+
+	if(screen->is_nv4x) {
+		if(debug_get_bool_option("NVFX_SIMULATE_NV30", FALSE))
+			screen->use_nv4x = 0;
+		if(!debug_get_bool_option("NVFX_NPOT", TRUE))
+			screen->advertise_npot = 0;
+		if(!debug_get_bool_option("NVFX_BLEND_EQ_SEP", TRUE))
+			screen->advertise_blend_equation_separate = 0;
+	}
+
 	screen->force_swtnl = debug_get_bool_option("NVFX_SWTNL", FALSE);
 	screen->trace_draw = debug_get_bool_option("NVFX_TRACE_DRAW", FALSE);
 
@@ -443,7 +456,7 @@ nvfx_screen_create(struct pipe_winsys *ws, struct nouveau_device *dev)
 	 * it should, due to several restrictions.
 	 * The only exception is fp16 on nv40.
 	 */
-	screen->advertise_fp16 = debug_get_bool_option("NVFX_FP16", !!screen->is_nv4x);
+	screen->advertise_fp16 = debug_get_bool_option("NVFX_FP16", !!screen->use_nv4x);
 	screen->advertise_fp32 = debug_get_bool_option("NVFX_FP32", 0);
 
 	screen->vertex_buffer_reloc_flags = nvfx_screen_get_vertex_buffer_flags(screen);
@@ -498,8 +511,8 @@ nvfx_screen_create(struct pipe_winsys *ws, struct nouveau_device *dev)
 	LIST_INITHEAD(&screen->query_list);
 
 	/* Vtxprog resources */
-	if (nouveau_resource_init(&screen->vp_exec_heap, 0, screen->is_nv4x ? 512 : 256) ||
-	    nouveau_resource_init(&screen->vp_data_heap, 0, screen->is_nv4x ? 468 : 256)) {
+	if (nouveau_resource_init(&screen->vp_exec_heap, 0, screen->use_nv4x ? 512 : 256) ||
+	    nouveau_resource_init(&screen->vp_data_heap, 0, screen->use_nv4x ? 468 : 256)) {
 		nvfx_screen_destroy(pscreen);
 		return NULL;
 	}
diff --git a/src/gallium/drivers/nvfx/nvfx_screen.h b/src/gallium/drivers/nvfx/nvfx_screen.h
index 566fcb12085..b1f07187c78 100644
--- a/src/gallium/drivers/nvfx/nvfx_screen.h
+++ b/src/gallium/drivers/nvfx/nvfx_screen.h
@@ -15,12 +15,15 @@ struct nvfx_screen {
 	struct nvfx_context *cur_ctx;
 
 	unsigned is_nv4x; /* either 0 or ~0 */
+	unsigned use_nv4x; /* either 0 or ~0 */
 	boolean force_swtnl;
 	boolean trace_draw;
 	unsigned vertex_buffer_reloc_flags;
 	unsigned index_buffer_reloc_flags;
 	unsigned advertise_fp16;
 	unsigned advertise_fp32;
+	unsigned advertise_npot;
+	unsigned advertise_blend_equation_separate;
 
 	/* HW graphics objects */
 	struct nouveau_grobj *eng3d;
diff --git a/src/gallium/drivers/nvfx/nvfx_surface.h b/src/gallium/drivers/nvfx/nvfx_surface.h
new file mode 100644
index 00000000000..e69de29bb2d
--- /dev/null
+++ b/src/gallium/drivers/nvfx/nvfx_surface.h
diff --git a/src/gallium/drivers/nvfx/nvfx_vertprog.c b/src/gallium/drivers/nvfx/nvfx_vertprog.c
index 570bb452b5e..6e5b6c09363 100644
--- a/src/gallium/drivers/nvfx/nvfx_vertprog.c
+++ b/src/gallium/drivers/nvfx/nvfx_vertprog.c
@@ -297,8 +297,8 @@ nvfx_vp_emit(struct nvfx_vpc *vpc, struct nvfx_insn insn)
 
 	if(insn.sat)
 	{
-		assert(nvfx->is_nv4x);
-		if(nvfx->is_nv4x)
+		assert(nvfx->use_nv4x);
+		if(nvfx->use_nv4x)
 			hw[0] |= NV40_VP_INST_SATURATE;
 	}
 
@@ -519,7 +519,7 @@ nvfx_vertprog_parse_instruction(struct nvfx_context* nvfx, struct nvfx_vpc *vpc,
 	if(finst->Instruction.Saturate == TGSI_SAT_ZERO_ONE)
 	{
 		assert(finst->Instruction.Opcode != TGSI_OPCODE_ARL);
-		if(nvfx->is_nv4x)
+		if(nvfx->use_nv4x)
 			sat = TRUE;
 		else if(dst.type != NVFXSR_TEMP)
 			dst = temp(vpc);
@@ -757,7 +757,7 @@ nvfx_vertprog_parse_instruction(struct nvfx_context* nvfx, struct nvfx_vpc *vpc,
 		return FALSE;
 	}
 
-	if(finst->Instruction.Saturate == TGSI_SAT_ZERO_ONE && !nvfx->is_nv4x)
+	if(finst->Instruction.Saturate == TGSI_SAT_ZERO_ONE && !nvfx->use_nv4x)
 	{
 		if(!vpc->r_0_1.type)
 			vpc->r_0_1 = constant(vpc, -1, 0, 1, 0, 0);