Module: Mesa Branch: staging/19.2 Commit: 2d30c0fc5905c88f0208c2757e6e07976a8660c9 URL: http://cgit.freedesktop.org/mesa/mesa/commit/?id=2d30c0fc5905c88f0208c2757e6e07976a8660c9
Author: Ian Romanick <[email protected]> Date: Sat Aug 31 11:40:32 2019 -0700 nir/algebraic: Do not apply late DPH optimization in vertex processing stages Some shaders do not use 'invariant' in vertex and (possibly) geometry shader stages on some outputs that are intended to be invariant. For various reasons, this optimization may not be fully applied in all shaders used for different rendering passes of the same geometry. This can result in Z-fighting artifacts (at best). For now, disable this optimization in these stages. In tessellation stages applications seem to use 'precise' when necessary, so allow the optimization in those stages. Reviewed-by: Kenneth Graunke <[email protected]> Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=111490 Fixes: 09705747d72 ("nir/algebraic: Reassociate fadd into fmul in DPH-like pattern") All Gen8+ platforms had similar results. (Ice Lake shown) total instructions in shared programs: 16194726 -> 16344745 (0.93%) instructions in affected programs: 2855172 -> 3005191 (5.25%) helped: 6 HURT: 20279 helped stats (abs) min: 1 max: 3 x̄: 1.33 x̃: 1 helped stats (rel) min: 0.44% max: 1.00% x̄: 0.54% x̃: 0.44% HURT stats (abs) min: 1 max: 32 x̄: 7.40 x̃: 7 HURT stats (rel) min: 0.14% max: 42.86% x̄: 8.58% x̃: 6.56% 95% mean confidence interval for instructions value: 7.34 7.45 95% mean confidence interval for instructions %-change: 8.48% 8.67% Instructions are HURT. total cycles in shared programs: 364471296 -> 365014683 (0.15%) cycles in affected programs: 32421530 -> 32964917 (1.68%) helped: 2925 HURT: 16144 helped stats (abs) min: 1 max: 403 x̄: 18.39 x̃: 5 helped stats (rel) min: <.01% max: 22.61% x̄: 1.97% x̃: 1.15% HURT stats (abs) min: 1 max: 18471 x̄: 36.99 x̃: 15 HURT stats (rel) min: 0.02% max: 52.58% x̄: 5.60% x̃: 3.87% 95% mean confidence interval for cycles value: 21.58 35.41 95% mean confidence interval for cycles %-change: 4.36% 4.52% Cycles are HURT. (cherry picked from commit 92f70df8c38a36d913334c596ce26af64b6c569b) --- src/compiler/nir/nir_opt_algebraic.py | 15 ++++++++++++--- 1 file changed, 12 insertions(+), 3 deletions(-) diff --git a/src/compiler/nir/nir_opt_algebraic.py b/src/compiler/nir/nir_opt_algebraic.py index cf7f48b0c5c..6892f1c630b 100644 --- a/src/compiler/nir/nir_opt_algebraic.py +++ b/src/compiler/nir/nir_opt_algebraic.py @@ -1524,10 +1524,19 @@ late_optimizations = [ # fadd(ffma(v1.z, v2.z, ffma(v1.y, v2.y, fmul(v1.x, v2.x))), v1.w) # # Reassociate the last addition into the first multiplication. + # + # Some shaders do not use 'invariant' in vertex and (possibly) geometry + # shader stages on some outputs that are intended to be invariant. For + # various reasons, this optimization may not be fully applied in all + # shaders used for different rendering passes of the same geometry. This + # can result in Z-fighting artifacts (at best). For now, disable this + # optimization in these stages. See bugzilla #111490. In tessellation + # stages applications seem to use 'precise' when necessary, so allow the + # optimization in those stages. (('~fadd', ('ffma(is_used_once)', a, b, ('ffma', c, d, ('fmul', 'e(is_not_const_and_not_fsign)', 'f(is_not_const_and_not_fsign)'))), 'g(is_not_const)'), - ('ffma', a, b, ('ffma', c, d, ('ffma', e, 'f', 'g'))), '!options->intel_vec4'), - (('~fadd', ('ffma(is_used_once)', a, b, ('fmul', 'e(is_not_const_and_not_fsign)', 'f(is_not_const_and_not_fsign)') ), 'g(is_not_const)'), - ('ffma', a, b, ('ffma', e, 'f', 'g') ), '!options->intel_vec4'), + ('ffma', a, b, ('ffma', c, d, ('ffma', e, 'f', 'g'))), '(info->stage != MESA_SHADER_VERTEX && info->stage != MESA_SHADER_GEOMETRY) && !options->intel_vec4'), + (('~fadd', ('ffma(is_used_once)', a, b, ('fmul', 'c(is_not_const_and_not_fsign)', 'd(is_not_const_and_not_fsign)') ), 'e(is_not_const)'), + ('ffma', a, b, ('ffma', c, d, e)), '(info->stage != MESA_SHADER_VERTEX && info->stage != MESA_SHADER_GEOMETRY) && !options->intel_vec4'), ] print(nir_algebraic.AlgebraicPass("nir_opt_algebraic", optimizations).render()) _______________________________________________ mesa-commit mailing list [email protected] https://lists.freedesktop.org/mailman/listinfo/mesa-commit
