fdo-mirrors/mesa
1
0
Fork 0
mirror of https://gitlab.freedesktop.org/mesa/mesa.git synced 2026-05-05 16:08:04 +02:00
Code Issues Projects Releases Packages Wiki Activity Actions 2

aco: fix u2f16 with 32bit input

Browse source 
The vulkan spec says all conversions are correctly rounded, so if the input
is larger than the largest fp16 value, we need to return MAX_FLOAT/inf
instead of cutting off the msbs.

Cc: mesa-stable
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/24826>
(cherry picked from commit 6d949e18fd)
This commit is contained in:
Georg Lehmann 2023-08-22 15:15:40 +02:00 • committed by Dylan Baker
parent f179d999fc
commit 5d0248db61
2 changed files with 8 additions and 9 deletions
Show stats Download patch file Download diff file Expand all files Collapse all files

2
.pick_status.json
View file

@ -4014,7 +4014,7 @@
"description": "aco: fix u2f16 with 32bit input",
"nominated": true,
"nomination_type": 0,
"resolution": 0,
"resolution": 1,
"main_sha": null,
"because_sha": null,
"notes": null

15
src/amd/compiler/aco_instruction_selection.cpp
View file

@ -2938,10 +2938,7 @@ visit_alu_instr(isel_context* ctx, nir_alu_instr* instr)
if (ctx->program->gfx_level >= GFX8 && input_size <= 16) {
bld.vop1(aco_opcode::v_cvt_f16_i16, Definition(dst), src);
} else {
/* Convert to f32 and then down to f16. This is needed to handle
* inputs slightly outside the range [INT16_MIN, INT16_MAX],
* which are representable via f16 but wouldn't be converted
* correctly by v_cvt_f16_i16.
/* Large 32bit inputs need to return +-inf/FLOAT_MAX.
*
* This is also the fallback-path taken on GFX7 and earlier, which
* do not support direct f16⟷i16 conversions.
@ -2989,12 +2986,14 @@ visit_alu_instr(isel_context* ctx, nir_alu_instr* instr)
}
}
if (ctx->program->gfx_level >= GFX8) {
/* float16 has a range of [0, 65519]. Converting from larger
* inputs is UB, so we just need to consider the lower 16 bits */
if (ctx->program->gfx_level >= GFX8 && input_size <= 16) {
bld.vop1(aco_opcode::v_cvt_f16_u16, Definition(dst), src);
} else {
/* GFX7 and earlier do not support direct f16⟷u16 conversions */
/* Large 32bit inputs need to return inf/FLOAT_MAX.
*
* This is also the fallback-path taken on GFX7 and earlier, which
* do not support direct f16⟷u16 conversions.
*/
src = bld.vop1(aco_opcode::v_cvt_f32_u32, bld.def(v1), src);
bld.vop1(aco_opcode::v_cvt_f16_f32, Definition(dst), src);
}