From 03a72d96d8dacc32e817089b94bec08ac70b898b Mon Sep 17 00:00:00 2001 From: Jason Ekstrand Date: Fri, 29 Mar 2019 22:51:20 -0500 Subject: [PATCH] nir/algebraic: Drop some @bool specifiers Now that we have one-bit booleans, we don't need to rely on looking at parent instructions in order to figure out if a value is a Boolean most of the time. We can drop these specifiers and now the optimizations will apply more generally. Shader-DB results on Kaby Lake: total instructions in shared programs: 15321168 -> 15321227 (<.01%) instructions in affected programs: 8836 -> 8895 (0.67%) helped: 1 HURT: 31 total cycles in shared programs: 357481781 -> 357481321 (<.01%) cycles in affected programs: 146524 -> 146064 (-0.31%) helped: 22 HURT: 10 total spills in shared programs: 23675 -> 23673 (<.01%) spills in affected programs: 11 -> 9 (-18.18%) helped: 1 HURT: 0 total fills in shared programs: 32040 -> 32036 (-0.01%) fills in affected programs: 27 -> 23 (-14.81%) helped: 1 HURT: 0 No change in VkPipeline-DB Looking at the instructions hurt, a bunch of them seem to be a case where doing exactly the right thing in NIR ends up doing the wrong-ish thing in the back-end because flags are dumb. In particular, there's a case where we have a MUL followed by a CMP followed by a SEL and when we turn that SEL into an OR, it uses the GRF result of the CMP rather than the flag result so the CMP can't be merged with the MUL. Those shaders appear to schedule better according to the cycle estimates so I guess it's a win? Also it helps spilling in one Car Chase compute shader. Reviewed-by: Timothy Arceri --- src/compiler/nir/nir_opt_algebraic.py | 14 +++++++------- 1 file changed, 7 insertions(+), 7 deletions(-) diff --git a/src/compiler/nir/nir_opt_algebraic.py b/src/compiler/nir/nir_opt_algebraic.py index bdf7078..597d479 100644 --- a/src/compiler/nir/nir_opt_algebraic.py +++ b/src/compiler/nir/nir_opt_algebraic.py @@ -312,7 +312,7 @@ optimizations = [ (('bcsel', a, ('bcsel(is_used_once)', b, c, d), ('bcsel', b, c, 'e')), ('bcsel', b, c, ('bcsel', a, d, 'e'))), (('bcsel', a, ('bcsel', b, c, d), ('bcsel(is_used_once)', b, 'e', d)), ('bcsel', b, ('bcsel', a, c, 'e'), d)), (('bcsel', a, ('bcsel(is_used_once)', b, c, d), ('bcsel', b, 'e', d)), ('bcsel', b, ('bcsel', a, c, 'e'), d)), - (('bcsel', a, True, 'b@bool'), ('ior', a, b)), + (('bcsel', a, True, b), ('ior', a, b)), (('fmin', a, a), a), (('fmax', a, a), a), (('imin', a, a), a), @@ -390,7 +390,7 @@ optimizations = [ (('ior', ('uge', 1, a), ('ieq', a, 2)), ('uge', 2, a)), (('ior', ('uge', 2, a), ('ieq', a, 3)), ('uge', 3, a)), - (('ior', 'a@bool', ('ieq', a, False)), True), + (('ior', a, ('ieq', a, False)), True), (('ior', a, ('inot', a)), -1), (('iand', ('ieq', 'a@32', 0), ('ieq', 'b@32', 0)), ('ieq', ('ior', 'a@32', 'b@32'), 0)), @@ -535,10 +535,10 @@ optimizations = [ # Boolean simplifications (('i2b32(is_used_by_if)', a), ('ine32', a, 0)), (('i2b1(is_used_by_if)', a), ('ine', a, 0)), - (('ieq', 'a@bool', True), a), - (('ine(is_not_used_by_if)', 'a@bool', True), ('inot', a)), - (('ine', 'a@bool', False), a), - (('ieq(is_not_used_by_if)', 'a@bool', False), ('inot', 'a')), + (('ieq', a, True), a), + (('ine(is_not_used_by_if)', a, True), ('inot', a)), + (('ine', a, False), a), + (('ieq(is_not_used_by_if)', a, False), ('inot', 'a')), (('bcsel', a, True, False), a), (('bcsel', a, False, True), ('inot', a)), (('bcsel@32', a, 1.0, 0.0), ('b2f', a)), @@ -1018,7 +1018,7 @@ late_optimizations = [ (('fmin', ('fadd(is_used_once)', '#c', a), ('fadd(is_used_once)', '#c', b)), ('fadd', c, ('fmin', a, b))), (('fmax', ('fadd(is_used_once)', '#c', a), ('fadd(is_used_once)', '#c', b)), ('fadd', c, ('fmax', a, b))), - (('bcsel', 'a@bool', 0, ('b2f32', ('inot', 'b@bool'))), ('b2f32', ('inot', ('ior', a, b)))), + (('bcsel', a, 0, ('b2f32', ('inot', 'b@bool'))), ('b2f32', ('inot', ('ior', a, b)))), ] print(nir_algebraic.AlgebraicPass("nir_opt_algebraic", optimizations).render()) -- 2.7.4