Mesa (master): radeonsi: Always pre-load separate VGPRs for centroid vs. center interpolation

Michel Dänzer daenzer at kemper.freedesktop.org
Tue Aug 20 17:08:46 UTC 2013


Module: Mesa
Branch: master
Commit: be301f707e8c6ba3bb0574e64a0a4f1653af8fb7
URL:    http://cgit.freedesktop.org/mesa/mesa/commit/?id=be301f707e8c6ba3bb0574e64a0a4f1653af8fb7

Author: Michel Dänzer <michel.daenzer at amd.com>
Date:   Mon Aug 19 15:45:32 2013 +0200

radeonsi: Always pre-load separate VGPRs for centroid vs. center interpolation

The LLVM R600 backend currently always uses separate VGPRs for these.

Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=68162
(Centroid interpolation is identical to center interpolation without
multisampling, so the shader hardware was only pre-loading one set of
interpolation coefficients, and the pixel shader code was using
uninitialized values as the centroid interpolation coefficients)

Cc: mesa-stable at lists.freedesktop.org
Tested-by: Laurent Carlier <lordheavym at gmail.com>

---

 src/gallium/drivers/radeonsi/si_state_draw.c |    3 ++-
 1 files changed, 2 insertions(+), 1 deletions(-)

diff --git a/src/gallium/drivers/radeonsi/si_state_draw.c b/src/gallium/drivers/radeonsi/si_state_draw.c
index 7ecaf77..15cb87f 100644
--- a/src/gallium/drivers/radeonsi/si_state_draw.c
+++ b/src/gallium/drivers/radeonsi/si_state_draw.c
@@ -183,7 +183,8 @@ static void si_pipe_shader_ps(struct pipe_context *ctx, struct si_pipe_shader *s
 		exports_ps = 2;
 	}
 
-	spi_ps_in_control = S_0286D8_NUM_INTERP(shader->shader.ninterp);
+	spi_ps_in_control = S_0286D8_NUM_INTERP(shader->shader.ninterp) |
+		S_0286D8_BC_OPTIMIZE_DISABLE(1);
 
 	si_pm4_set_reg(pm4, R_0286E0_SPI_BARYC_CNTL, spi_baryc_cntl);
 	spi_ps_input_ena = shader->spi_ps_input_ena;




More information about the mesa-commit mailing list