mirror of
https://github.com/encounter/dawn-cmake.git
synced 2025-10-24 18:50:29 +00:00
1a1b5278d5
Change the DecomposeMemoryAccess to behave more like the DirectVariableAccess transform, in that it'll inline the access of buffer variable into the load / store helper functions, instead of passing the array down. This avoids large array copies observed with FXC, which can have *severe* performance costs. Fixed: tint:1819 Change-Id: I52eb3f908813f72ab9da446743e24a2637158309 Reviewed-on: https://dawn-review.googlesource.com/c/dawn/+/121460 Kokoro: Kokoro <noreply+kokoro@google.com> Auto-Submit: Ben Clayton <bclayton@google.com> Reviewed-by: James Price <jrprice@google.com> Commit-Queue: James Price <jrprice@google.com>
104 lines
2.3 KiB
HLSL
104 lines
2.3 KiB
HLSL
struct strided_arr {
|
|
float el;
|
|
};
|
|
struct strided_arr_1 {
|
|
strided_arr el[3][2];
|
|
};
|
|
|
|
RWByteAddressBuffer s : register(u0, space0);
|
|
|
|
strided_arr s_load_4(uint offset) {
|
|
const strided_arr tint_symbol = {asfloat(s.Load((offset + 0u)))};
|
|
return tint_symbol;
|
|
}
|
|
|
|
typedef strided_arr s_load_3_ret[2];
|
|
s_load_3_ret s_load_3(uint offset) {
|
|
strided_arr arr[2] = (strided_arr[2])0;
|
|
{
|
|
for(uint i = 0u; (i < 2u); i = (i + 1u)) {
|
|
arr[i] = s_load_4((offset + (i * 8u)));
|
|
}
|
|
}
|
|
return arr;
|
|
}
|
|
|
|
typedef strided_arr s_load_2_ret[3][2];
|
|
s_load_2_ret s_load_2(uint offset) {
|
|
strided_arr arr_1[3][2] = (strided_arr[3][2])0;
|
|
{
|
|
for(uint i_1 = 0u; (i_1 < 3u); i_1 = (i_1 + 1u)) {
|
|
arr_1[i_1] = s_load_3((offset + (i_1 * 16u)));
|
|
}
|
|
}
|
|
return arr_1;
|
|
}
|
|
|
|
strided_arr_1 s_load_1(uint offset) {
|
|
const strided_arr_1 tint_symbol_1 = {s_load_2((offset + 0u))};
|
|
return tint_symbol_1;
|
|
}
|
|
|
|
typedef strided_arr_1 s_load_ret[4];
|
|
s_load_ret s_load(uint offset) {
|
|
strided_arr_1 arr_2[4] = (strided_arr_1[4])0;
|
|
{
|
|
for(uint i_2 = 0u; (i_2 < 4u); i_2 = (i_2 + 1u)) {
|
|
arr_2[i_2] = s_load_1((offset + (i_2 * 128u)));
|
|
}
|
|
}
|
|
return arr_2;
|
|
}
|
|
|
|
void s_store_4(uint offset, strided_arr value) {
|
|
s.Store((offset + 0u), asuint(value.el));
|
|
}
|
|
|
|
void s_store_3(uint offset, strided_arr value[2]) {
|
|
strided_arr array_3[2] = value;
|
|
{
|
|
for(uint i_3 = 0u; (i_3 < 2u); i_3 = (i_3 + 1u)) {
|
|
s_store_4((offset + (i_3 * 8u)), array_3[i_3]);
|
|
}
|
|
}
|
|
}
|
|
|
|
void s_store_2(uint offset, strided_arr value[3][2]) {
|
|
strided_arr array_2[3][2] = value;
|
|
{
|
|
for(uint i_4 = 0u; (i_4 < 3u); i_4 = (i_4 + 1u)) {
|
|
s_store_3((offset + (i_4 * 16u)), array_2[i_4]);
|
|
}
|
|
}
|
|
}
|
|
|
|
void s_store_1(uint offset, strided_arr_1 value) {
|
|
s_store_2((offset + 0u), value.el);
|
|
}
|
|
|
|
void s_store(uint offset, strided_arr_1 value[4]) {
|
|
strided_arr_1 array_1[4] = value;
|
|
{
|
|
for(uint i_5 = 0u; (i_5 < 4u); i_5 = (i_5 + 1u)) {
|
|
s_store_1((offset + (i_5 * 128u)), array_1[i_5]);
|
|
}
|
|
}
|
|
}
|
|
|
|
void f_1() {
|
|
const strided_arr_1 x_19[4] = s_load(0u);
|
|
const strided_arr x_24[3][2] = s_load_2(384u);
|
|
const strided_arr x_28[2] = s_load_3(416u);
|
|
const float x_32 = asfloat(s.Load(424u));
|
|
const strided_arr_1 tint_symbol_2[4] = (strided_arr_1[4])0;
|
|
s_store(0u, tint_symbol_2);
|
|
s.Store(424u, asuint(5.0f));
|
|
return;
|
|
}
|
|
|
|
[numthreads(1, 1, 1)]
|
|
void f() {
|
|
f_1();
|
|
return;
|
|
}
|