freedreno/ir3: fix lockups with lame FRAG shaders

Shaders like:

  FRAG
  PROPERTY FS_COLOR0_WRITES_ALL_CBUFS 1
  DCL IN[0], GENERIC[0], PERSPECTIVE
  DCL OUT[0], COLOR
  DCL SAMP[0]
  DCL TEMP[0], LOCAL
  IMM[0] FLT32 {    0.0000,     1.0000,     0.0000,     0.0000}
    0: TEX TEMP[0], IN[0].xyyy, SAMP[0], 2D
    1: MOV OUT[0], IMM[0].xyxx
    2: END

cause unhappyness.  They have an IN[], but once this is compiled the
useless TEX instruction goes away.  Leaving a varying that is never
fetched, which makes the hw unhappy.

In the process fix a signed vs unsigned compare.  If the vertex shader
has max_reg=-1, MAX2() vs an unsigned would not give the desired result.

Signed-off-by: Rob Clark <[email protected]>

4 files changed, 17 insertions, 6 deletions

diff --git a/src/gallium/drivers/freedreno/ir3/ir3.h b/src/gallium/drivers/freedreno/ir3/ir3.h
index c73e9cc8d7a..b92a57a43f8 100644
--- a/src/gallium/drivers/freedreno/ir3/ir3.h
+++ b/src/gallium/drivers/freedreno/ir3/ir3.h
@@ -413,7 +413,7 @@ int ir3_block_sched(struct ir3_block *block);
 /* register assignment: */
 int ir3_block_ra(struct ir3_block *block, enum shader_t type,
 		bool half_precision, bool frag_coord, bool frag_face,
-		bool *has_samp);
+		bool *has_samp, int *max_bary);
 
 #ifndef ARRAY_SIZE
 #  define ARRAY_SIZE(arr) (sizeof(arr) / sizeof((arr)[0]))
diff --git a/src/gallium/drivers/freedreno/ir3/ir3_compiler.c b/src/gallium/drivers/freedreno/ir3/ir3_compiler.c
index 66f78a80f6d..80676830dd7 100644
--- a/src/gallium/drivers/freedreno/ir3/ir3_compiler.c
+++ b/src/gallium/drivers/freedreno/ir3/ir3_compiler.c
@@ -3068,7 +3068,7 @@ ir3_compile_shader(struct ir3_shader_variant *so,
 	struct ir3_block *block;
 	struct ir3_instruction **inputs;
 	unsigned i, j, actual_in;
-	int ret = 0;
+	int ret = 0, max_bary;
 
 	assert(!so->ir);
 
@@ -3183,7 +3183,7 @@ ir3_compile_shader(struct ir3_shader_variant *so,
 	}
 
 	ret = ir3_block_ra(block, so->type, key.half_precision,
-			so->frag_coord, so->frag_face, &so->has_samp);
+			so->frag_coord, so->frag_face, &so->has_samp, &max_bary);
 	if (ret) {
 		DBG("RA failed!");
 		goto out;
@@ -3230,6 +3230,8 @@ ir3_compile_shader(struct ir3_shader_variant *so,
 	 */
 	if (so->type == SHADER_VERTEX)
 		so->total_in = actual_in;
+	else
+		so->total_in = align(max_bary + 1, 4);
 
 out:
 	if (ret) {
diff --git a/src/gallium/drivers/freedreno/ir3/ir3_ra.c b/src/gallium/drivers/freedreno/ir3/ir3_ra.c
index 3ac626ca3b6..aa8ad513e04 100644
--- a/src/gallium/drivers/freedreno/ir3/ir3_ra.c
+++ b/src/gallium/drivers/freedreno/ir3/ir3_ra.c
@@ -58,6 +58,7 @@ struct ir3_ra_ctx {
 	bool frag_face;
 	bool has_samp;
 	int cnt;
+	int max_bary;
 	bool error;
 };
 
@@ -614,6 +615,12 @@ static void legalize(struct ir3_ra_ctx *ctx, struct ir3_block *block)
 		if (is_meta(n))
 			continue;
 
+		if (is_input(n)) {
+			struct ir3_register *inloc = n->regs[1];
+			assert(inloc->flags & IR3_REG_IMMED);
+			ctx->max_bary = MAX2(ctx->max_bary, inloc->iim_val);
+		}
+
 		for (i = 1; i < n->regs_count; i++) {
 			reg = n->regs[i];
 
@@ -775,7 +782,7 @@ static int block_ra(struct ir3_ra_ctx *ctx, struct ir3_block *block)
 
 int ir3_block_ra(struct ir3_block *block, enum shader_t type,
 		bool half_precision, bool frag_coord, bool frag_face,
-		bool *has_samp)
+		bool *has_samp, int *max_bary)
 {
 	struct ir3_ra_ctx ctx = {
 			.block = block,
@@ -783,12 +790,14 @@ int ir3_block_ra(struct ir3_block *block, enum shader_t type,
 			.half_precision = half_precision,
 			.frag_coord = frag_coord,
 			.frag_face = frag_face,
+			.max_bary = -1,
 	};
 	int ret;
 
 	ir3_clear_mark(block->shader);
 	ret = block_ra(&ctx, block);
 	*has_samp = ctx.has_samp;
+	*max_bary = ctx.max_bary;
 
 	return ret;
 }
diff --git a/src/gallium/drivers/freedreno/ir3/ir3_shader.c b/src/gallium/drivers/freedreno/ir3/ir3_shader.c
index ed7c639c930..31e302b92c2 100644
--- a/src/gallium/drivers/freedreno/ir3/ir3_shader.c
+++ b/src/gallium/drivers/freedreno/ir3/ir3_shader.c
@@ -91,12 +91,12 @@ fixup_vp_regfootprint(struct ir3_shader_variant *v)
 	unsigned i;
 	for (i = 0; i < v->inputs_count; i++) {
 		if (v->inputs[i].compmask) {
-			uint32_t regid = (v->inputs[i].regid + 3) >> 2;
+			int32_t regid = (v->inputs[i].regid + 3) >> 2;
 			v->info.max_reg = MAX2(v->info.max_reg, regid);
 		}
 	}
 	for (i = 0; i < v->outputs_count; i++) {
-		uint32_t regid = (v->outputs[i].regid + 3) >> 2;
+		int32_t regid = (v->outputs[i].regid + 3) >> 2;
 		v->info.max_reg = MAX2(v->info.max_reg, regid);
 	}
 }