fdo-mirrors/mesa
1
0
Fork 0
mirror of https://gitlab.freedesktop.org/mesa/mesa.git synced 2025-12-20 07:20:10 +01:00
Code Issues Projects Releases Packages Wiki Activity Actions 2

nir/opt_algebraic: optimize 16bit vec2 comparison followed by b2i16 using usub_sat

Browse source 
Helps vectorized emulated fp16 -> fp8 conversions

No Foz-DB changes.

Reviewed-by: Rhys Perry <pendingchaos02@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/35876>
This commit is contained in:
Georg Lehmann 2025-07-02 08:13:03 +02:00 • committed by Marge Bot
parent b21e62b71a
commit 045ddb992a
2 changed files with 31 additions and 0 deletions
Show stats Download patch file Download diff file Expand all files Collapse all files

11
src/compiler/nir/nir_opt_algebraic.py
View file

@ -2704,6 +2704,17 @@ optimizations.extend([
(('imul_high@16', a, b), ('i2i16', ('ishr', ('imul24_relaxed', ('i2i32', a), ('i2i32', b)), 16)), 'options->lower_mul_high16'), (('imul_high@16', a, b), ('i2i16', ('ishr', ('imul24_relaxed', ('i2i32', a), ('i2i32', b)), 16)), 'options->lower_mul_high16'),
(('umul_high@16', a, b), ('u2u16', ('ushr', ('umul24_relaxed', ('u2u32', a), ('u2u32', b)), 16)), 'options->lower_mul_high16'), (('umul_high@16', a, b), ('u2u16', ('ushr', ('umul24_relaxed', ('u2u32', a), ('u2u32', b)), 16)), 'options->lower_mul_high16'),
# Optimize vec2 unsigned comparison predicates to usub_sat with clamp.
(('b2i16', ('vec2', ('ult', 'a@16', b), ('ult', 'c@16', d))),
('umin', 1, ('usub_sat', ('vec2', b, d), ('vec2', a, c))),
'options->vectorize_vec2_16bit && !options->lower_usub_sat'),
(('b2i16', ('vec2', ('uge', 'a@16', '#b(is_not_zero)'), ('uge', 'c@16', '#d(is_not_zero)'))),
('umin', 1, ('usub_sat', ('vec2', a, c), ('iadd', ('vec2', b, d), -1))),
'options->vectorize_vec2_16bit && !options->lower_usub_sat'),
(('b2i16', ('vec2', ('uge', '#a(is_not_uint_max)', 'b@16'), ('uge', '#c(is_not_uint_max)', 'd@16'))),
('umin', 1, ('usub_sat', ('iadd', ('vec2', a, c), 1), ('vec2', b, d))),
'options->vectorize_vec2_16bit && !options->lower_usub_sat'),
]) ])
for bit_size in [8, 16, 32, 64]: for bit_size in [8, 16, 32, 64]:

20
src/compiler/nir/nir_search_helpers.h
View file

@ -862,6 +862,26 @@ is_5lsb_not_zero(UNUSED struct hash_table *ht, const nir_alu_instr *instr,
return true; return true;
} }
/**
* Returns whether at least one bit is 0.
*/
static inline bool
is_not_uint_max(UNUSED struct hash_table *ht, const nir_alu_instr *instr,
unsigned src, unsigned num_components,
const uint8_t *swizzle)
{
if (nir_src_as_const_value(instr->src[src].src) == NULL)
return false;
for (unsigned i = 0; i < num_components; i++) {
const int64_t c = nir_src_comp_as_int(instr->src[src].src, swizzle[i]);
if (c == -1)
return false;
}
return true;
}
static inline bool static inline bool
no_signed_wrap(const nir_alu_instr *instr) no_signed_wrap(const nir_alu_instr *instr)
{ {