vtn/opencl: add ctz support

ctz is a CL2.0 opcode but 3.0 requires it as well so just add support
for it.

Tested against CTS integer_ops integer_ctz test.

(long line broken up)

Reviewed-by: Francisco Jerez <currojerez@riseup.net>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/7468>
This commit is contained in:
Dave Airlie 2020-11-06 13:14:07 +10:00
parent 725df7c15d
commit 9790fdf2ce
2 changed files with 13 additions and 0 deletions

View file

@ -261,6 +261,16 @@ nir_clz_u(nir_builder *b, nir_ssa_def *a)
return nir_u2u(b, val, a->bit_size);
}
static inline nir_ssa_def *
nir_ctz_u(nir_builder *b, nir_ssa_def *a)
{
nir_ssa_def *cond = nir_ieq(b, a, nir_imm_intN_t(b, 0, a->bit_size));
return nir_bcsel(b, cond,
nir_imm_intN_t(b, a->bit_size, a->bit_size),
nir_u2u(b, nir_find_lsb(b, a), a->bit_size));
}
#ifdef __cplusplus
}
#endif

View file

@ -533,6 +533,8 @@ handle_special(struct vtn_builder *b, uint32_t opcode,
return nir_normalize(nb, srcs[0]);
case OpenCLstd_Clz:
return nir_clz_u(nb, srcs[0]);
case OpenCLstd_Ctz:
return nir_ctz_u(nb, srcs[0]);
case OpenCLstd_Select:
return nir_select(nb, srcs[0], srcs[1], srcs[2]);
case OpenCLstd_S_Upsample:
@ -876,6 +878,7 @@ vtn_handle_opencl_instruction(struct vtn_builder *b, SpvOp ext_opcode,
case OpenCLstd_S_Upsample:
case OpenCLstd_U_Upsample:
case OpenCLstd_Clz:
case OpenCLstd_Ctz:
case OpenCLstd_Native_exp:
case OpenCLstd_Native_exp10:
case OpenCLstd_Native_log: