KonstantinSeurer
/
mesa
1
0
Fork 0
Code Issues Pull Requests Packages Projects Releases Wiki Activity

gallivm: add 64-bit atomic global support 

Reviewed-by: Roland Scheidegger <sroland@vmware.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/9591>
This commit is contained in:
Dave Airlie 2021-03-17 10:12:52 +10:00 committed by Marge Bot
parent 92eafe42af
commit 8c1d5fcb7c
4 changed files with 28 additions and 5 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

6
src/gallium/auxiliary/gallivm/lp_bld_const.h
View File

@ -138,6 +138,12 @@ lp_build_const_float(struct gallivm_state *gallivm, float x)
return LLVMConstReal(LLVMFloatTypeInContext(gallivm->context), x);
}
static inline LLVMValueRef
lp_build_const_double(struct gallivm_state *gallivm, float x)
{
return LLVMConstReal(LLVMDoubleTypeInContext(gallivm->context), x);
}
/** Return constant-valued pointer to int */
static inline LLVMValueRef

4
src/gallium/auxiliary/gallivm/lp_bld_nir.c
View File

@ -1572,10 +1572,12 @@ static void visit_global_atomic(struct lp_build_nir_context *bld_base,
LLVMValueRef val = get_src(bld_base, instr->src[1]);
LLVMValueRef val2 = NULL;
int addr_bitsize = nir_src_bit_size(instr->src[0]);
int val_bitsize = nir_src_bit_size(instr->src[1]);
if (instr->intrinsic == nir_intrinsic_global_atomic_comp_swap)
val2 = get_src(bld_base, instr->src[2]);
bld_base->atomic_global(bld_base, instr->intrinsic, addr_bitsize, addr, val, val2, &result[0]);
bld_base->atomic_global(bld_base, instr->intrinsic, addr_bitsize,
val_bitsize, addr, val, val2, &result[0]);
}
static void visit_interp(struct lp_build_nir_context *bld_base,

1
src/gallium/auxiliary/gallivm/lp_bld_nir.h
View File

@ -86,6 +86,7 @@ struct lp_build_nir_context
void (*atomic_global)(struct lp_build_nir_context *bld_base,
nir_intrinsic_op op,
unsigned addr_bit_size,
unsigned val_bit_size,
LLVMValueRef addr,
LLVMValueRef val, LLVMValueRef val2,
LLVMValueRef *result);

22
src/gallium/auxiliary/gallivm/lp_bld_nir_soa.c
View File

@ -870,6 +870,7 @@ static void emit_store_global(struct lp_build_nir_context *bld_base,
static void emit_atomic_global(struct lp_build_nir_context *bld_base,
nir_intrinsic_op nir_op,
unsigned addr_bit_size,
unsigned val_bit_size,
LLVMValueRef addr,
LLVMValueRef val, LLVMValueRef val2,
LLVMValueRef *result)
@ -877,7 +878,7 @@ static void emit_atomic_global(struct lp_build_nir_context *bld_base,
struct gallivm_state *gallivm = bld_base->base.gallivm;
LLVMBuilderRef builder = gallivm->builder;
struct lp_build_context *uint_bld = &bld_base->uint_bld;
struct lp_build_context *atom_bld = get_int_bld(bld_base, true, val_bit_size);
LLVMValueRef atom_res = lp_build_alloca(gallivm,
LLVMTypeOf(val), "");
LLVMValueRef exec_mask = mask_vec(bld_base);
@ -897,10 +898,11 @@ static void emit_atomic_global(struct lp_build_nir_context *bld_base,
cond = LLVMBuildExtractElement(gallivm->builder, cond, loop_state.counter, "");
lp_build_if(&ifthen, gallivm, cond);
addr_ptr = LLVMBuildBitCast(gallivm->builder, addr_ptr, LLVMPointerType(LLVMTypeOf(value_ptr), 0), "");
if (nir_op == nir_intrinsic_global_atomic_comp_swap) {
LLVMValueRef cas_src_ptr = LLVMBuildExtractElement(gallivm->builder, val2,
loop_state.counter, "");
cas_src_ptr = LLVMBuildBitCast(gallivm->builder, cas_src_ptr, uint_bld->elem_type, "");
cas_src_ptr = LLVMBuildBitCast(gallivm->builder, cas_src_ptr, atom_bld->elem_type, "");
scalar = LLVMBuildAtomicCmpXchg(builder, addr_ptr, value_ptr,
cas_src_ptr,
LLVMAtomicOrderingSequentiallyConsistent,
@ -914,7 +916,7 @@ static void emit_atomic_global(struct lp_build_nir_context *bld_base,
op = LLVMAtomicRMWBinOpAdd;
break;
case nir_intrinsic_global_atomic_exchange:
addr_ptr = LLVMBuildBitCast(gallivm->builder, addr_ptr, LLVMPointerType(LLVMTypeOf(value_ptr), 0), "");
op = LLVMAtomicRMWBinOpXchg;
break;
case nir_intrinsic_global_atomic_and:
@ -953,7 +955,19 @@ static void emit_atomic_global(struct lp_build_nir_context *bld_base,
lp_build_else(&ifthen);
temp_res = LLVMBuildLoad(builder, atom_res, "");
bool is_float = LLVMTypeOf(val) == bld_base->base.vec_type;
LLVMValueRef zero_val = is_float ? lp_build_const_float(gallivm, 0) : lp_build_const_int32(gallivm, 0);
LLVMValueRef zero_val;
if (is_float) {
if (val_bit_size == 64)
zero_val = lp_build_const_double(gallivm, 0);
else
zero_val = lp_build_const_float(gallivm, 0);
} else {
if (val_bit_size == 64)
zero_val = lp_build_const_int64(gallivm, 0);
else
zero_val = lp_build_const_int32(gallivm, 0);
}
temp_res = LLVMBuildInsertElement(builder, temp_res, zero_val, loop_state.counter, "");
LLVMBuildStore(builder, temp_res, atom_res);
lp_build_endif(&ifthen);