fdo-mirrors/mesa
1
0
Fork 0
mirror of https://gitlab.freedesktop.org/mesa/mesa.git synced 2026-01-07 06:30:11 +01:00
Code Issues Projects Releases Packages Wiki Activity Actions 2

aco/gfx11: use v_cvt_i32_i16/v_cvt_u32_u16

Browse source 
fossil-db (gfx1100):
Totals from 52753 (39.07% of 135032) affected shaders:
CodeSize: 153603860 -> 153163384 (-0.29%)

Signed-off-by: Rhys Perry <pendingchaos02@gmail.com>
Reviewed-by: Daniel Schürmann <daniel@schuermann.dev>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/19933>
This commit is contained in:
Rhys Perry 2022-11-16 17:08:09 +00:00 • committed by Marge Bot
parent 8f0177b334
commit ce5838599d
2 changed files with 8 additions and 1 deletions
Show stats Download patch file Download diff file Expand all files Collapse all files

5
src/amd/compiler/aco_lower_to_hw_instr.cpp
View file

@ -1478,6 +1478,8 @@ do_pack_2x16(lower_context* ctx, Builder& bld, Definition def, Operand lo, Opera
/* move lo and zero high bits */
if (lo.physReg().byte() == 2)
bld.vop2(aco_opcode::v_lshrrev_b32, def_lo, Operand::c32(16u), lo);
else if (ctx->program->gfx_level >= GFX11)
bld.vop1(aco_opcode::v_cvt_u32_u16, def, lo);
else
bld.vop2(aco_opcode::v_and_b32, def_lo, Operand::c32(0xFFFFu), lo);
bld.vop2(aco_opcode::v_or_b32, def, Operand::c32(hi.constantValue() << 16u),
@ -2237,6 +2239,9 @@ lower_to_hw_instr(Program* program)
if (offset == (32 - bits) && op.regClass() != s1) {
bld.vop2(signext ? aco_opcode::v_ashrrev_i32 : aco_opcode::v_lshrrev_b32, dst,
Operand::c32(offset), op);
} else if (offset == 0 && bits == 16 && ctx.program->gfx_level >= GFX11) {
bld.vop1(signext ? aco_opcode::v_cvt_i32_i16 : aco_opcode::v_cvt_u32_u16, dst,
op);
} else {
bld.vop3(signext ? aco_opcode::v_bfe_i32 : aco_opcode::v_bfe_u32, dst, op,
Operand::c32(offset), Operand::c32(bits));

4
src/amd/compiler/tests/test_to_hw_instr.cpp
View file

@ -626,7 +626,9 @@ BEGIN_TEST(to_hw_instr.extract)
EXT(2, 8)
//! v1: %_:v[0] = @v_shr 24, %_:v[1]
EXT(3, 8)
//! v1: %_:v[0] = @v_bfe %_:v[1], 0, 16
//~gfx(7|8|9)_.*! v1: %_:v[0] = @v_bfe %_:v[1], 0, 16
//~gfx11_unsigned! v1: %_:v[0] = v_cvt_u32_u16 %_:v[1]
//~gfx11_signed! v1: %_:v[0] = v_cvt_i32_i16 %_:v[1]
EXT(0, 16)
//! v1: %_:v[0] = @v_shr 16, %_:v[1]
EXT(1, 16)