From: Marek Olšák <maraeo@gmail.com>
Date: Wed, 1 Sep 2010 04:14:58 +0000 (+0200)
Subject: r300/compiler: make optimizations not use 0.5 swizzles in vertex shaders
X-Git-Url: https://git.libre-soc.org/?a=commitdiff_plain;h=f46fd041369a1c2c188cbd65918c900900724d37;p=mesa.git

r300/compiler: make optimizations not use 0.5 swizzles in vertex shaders
---

diff --git a/src/gallium/drivers/r300/r300_fs.c b/src/gallium/drivers/r300/r300_fs.c
index 7a5ed1b03bc..e8ebe90b33b 100644
--- a/src/gallium/drivers/r300/r300_fs.c
+++ b/src/gallium/drivers/r300/r300_fs.c
@@ -385,6 +385,7 @@ static void r300_translate_fragment_shader(
     compiler.code = &shader->code;
     compiler.state = shader->compare_state;
     compiler.Base.is_r500 = r300->screen->caps.is_r500;
+    compiler.Base.has_half_swizzles = TRUE;
     compiler.Base.max_temp_regs = compiler.Base.is_r500 ? 128 : 32;
     compiler.Base.max_constants = compiler.Base.is_r500 ? 256 : 32;
     compiler.Base.max_alu_insts = compiler.Base.is_r500 ? 512 : 64;
diff --git a/src/gallium/drivers/r300/r300_vs.c b/src/gallium/drivers/r300/r300_vs.c
index 4df6464224c..2db7b52a87e 100644
--- a/src/gallium/drivers/r300/r300_vs.c
+++ b/src/gallium/drivers/r300/r300_vs.c
@@ -205,6 +205,7 @@ void r300_translate_vertex_shader(struct r300_context *r300,
     compiler.code = &vs->code;
     compiler.UserData = vs;
     compiler.Base.is_r500 = r300->screen->caps.is_r500;
+    compiler.Base.has_half_swizzles = FALSE;
     compiler.Base.max_temp_regs = 32;
     compiler.Base.max_constants = 256;
     compiler.Base.max_alu_insts = r300->screen->caps.is_r500 ? 1024 : 256;
diff --git a/src/mesa/drivers/dri/r300/compiler/radeon_compiler.h b/src/mesa/drivers/dri/r300/compiler/radeon_compiler.h
index fcb0aee9f1b..f6a8e834342 100644
--- a/src/mesa/drivers/dri/r300/compiler/radeon_compiler.h
+++ b/src/mesa/drivers/dri/r300/compiler/radeon_compiler.h
@@ -41,6 +41,7 @@ struct radeon_compiler {
 
 	/* Hardware specification. */
 	unsigned is_r500:1;
+	unsigned has_half_swizzles:1;
 	unsigned max_temp_regs;
 	unsigned max_constants;
 	int max_alu_insts;
diff --git a/src/mesa/drivers/dri/r300/compiler/radeon_optimize.c b/src/mesa/drivers/dri/r300/compiler/radeon_optimize.c
index 8db521e51b9..a9be30706bb 100644
--- a/src/mesa/drivers/dri/r300/compiler/radeon_optimize.c
+++ b/src/mesa/drivers/dri/r300/compiler/radeon_optimize.c
@@ -410,7 +410,7 @@ static void constant_folding(struct radeon_compiler * c, struct rc_instruction *
 				newswz = RC_SWIZZLE_ZERO;
 			} else if (baseimm == 1.0) {
 				newswz = RC_SWIZZLE_ONE;
-			} else if (baseimm == 0.5) {
+			} else if (baseimm == 0.5 && c->has_half_swizzles) {
 				newswz = RC_SWIZZLE_HALF;
 			} else {
 				have_real_reference = 1;
diff --git a/src/mesa/drivers/dri/r300/r300_blit.c b/src/mesa/drivers/dri/r300/r300_blit.c
index c22890c29e9..e43f0eeb889 100644
--- a/src/mesa/drivers/dri/r300/r300_blit.c
+++ b/src/mesa/drivers/dri/r300/r300_blit.c
@@ -89,6 +89,7 @@ static void create_vertex_program(struct r300_context *r300)
     compiler.SetHwInputOutput = vp_ins_outs;
     compiler.code = &r300->blit.vp_code;
     compiler.Base.is_r500 = r300->radeon.radeonScreen->chip_family >= CHIP_FAMILY_RV515;
+    compiler.Base.has_half_swizzles = 0;
     compiler.Base.max_temp_regs = 32;
     compiler.Base.max_constants = 256;
     compiler.Base.max_alu_insts = compiler.Base.is_r500 ? 1024 : 256;
@@ -123,6 +124,7 @@ static void create_fragment_program(struct r300_context *r300)
     compiler.OutputDepth = FRAG_RESULT_DEPTH;
     compiler.enable_shadow_ambient = GL_TRUE;
     compiler.Base.is_r500 = (r300->radeon.radeonScreen->chip_family >= CHIP_FAMILY_RV515);
+    compiler.Base.has_half_swizzles = 1;
     compiler.Base.max_temp_regs = (compiler.Base.is_r500) ? 128 : 32;
     compiler.Base.max_constants = compiler.Base.is_r500 ? 256 : 32;
     compiler.Base.max_alu_insts = compiler.Base.is_r500 ? 512 : 64;
diff --git a/src/mesa/drivers/dri/r300/r300_fragprog_common.c b/src/mesa/drivers/dri/r300/r300_fragprog_common.c
index cd0f6b3fe64..47cf3e49008 100644
--- a/src/mesa/drivers/dri/r300/r300_fragprog_common.c
+++ b/src/mesa/drivers/dri/r300/r300_fragprog_common.c
@@ -220,6 +220,7 @@ static void translate_fragment_program(GLcontext *ctx, struct r300_fragment_prog
 	compiler.state = fp->state;
 	compiler.enable_shadow_ambient = GL_TRUE;
 	compiler.Base.is_r500 = (r300->radeon.radeonScreen->chip_family >= CHIP_FAMILY_RV515) ? GL_TRUE : GL_FALSE;
+	compiler.Base.has_half_swizzles = 1;
 	compiler.Base.max_temp_regs = (compiler.Base.is_r500) ? 128 : 32;
 	compiler.Base.max_constants = compiler.Base.is_r500 ? 256 : 32;
 	compiler.Base.max_alu_insts = compiler.Base.is_r500 ? 512 : 64;
diff --git a/src/mesa/drivers/dri/r300/r300_vertprog.c b/src/mesa/drivers/dri/r300/r300_vertprog.c
index e7dcdef9b00..1d24266fbdd 100644
--- a/src/mesa/drivers/dri/r300/r300_vertprog.c
+++ b/src/mesa/drivers/dri/r300/r300_vertprog.c
@@ -245,6 +245,7 @@ static struct r300_vertex_program *build_program(GLcontext *ctx,
 	compiler.RequiredOutputs = compute_required_outputs(vp->Base, vp->key.FpReads);
 	compiler.SetHwInputOutput = &t_inputs_outputs;
 	compiler.Base.is_r500 = R300_CONTEXT(ctx)->radeon.radeonScreen->chip_family >= CHIP_FAMILY_RV515;
+	compiler.Base.has_half_swizzles = 0;
 	compiler.Base.max_temp_regs = 32;
 	compiler.Base.max_constants = 256;
 	compiler.Base.max_alu_insts = compiler.Base.is_r500 ? 1024 : 256;