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

radeonsi: don't use lp_build_if for the prim discard compute shader

Browse Source
This commit is contained in:
Marek Olšák
2019-07-23 20:49:27 -04:00
parent 5562b6b067
commit 33a8eab7a9
1 changed files with 18 additions and 23 deletions
Download Patch File Download Diff File Expand all files Collapse all files
src/gallium/drivers/radeonsi/si_compute_prim_discard.c
+18 -23
View File
@@ -267,7 +267,6 @@ static LLVMValueRef si_expand_32bit_pointer(struct si_shader_context *ctx, LLVMV
struct si_thread0_section {
struct si_shader_context *ctx;
struct lp_build_if_state if_thread0;
LLVMValueRef vgpr_result; /* a VGPR for the value on thread 0. */
LLVMValueRef saved_exec;
};
@@ -288,9 +287,9 @@ static void si_enter_thread0_section(struct si_shader_context *ctx,
*
* It could just be s_and_saveexec_b64 s, 1.
*/
lp_build_if(&section->if_thread0, &ctx->gallivm,
LLVMBuildICmp(ctx->ac.builder, LLVMIntEQ, thread_id,
ctx->i32_0, ""));
ac_build_ifcc(&ctx->ac,
LLVMBuildICmp(ctx->ac.builder, LLVMIntEQ, thread_id,
ctx->i32_0, ""), 12601);
}
/* Exit a section that only executes on thread 0 and broadcast the result
@@ -302,7 +301,7 @@ static void si_exit_thread0_section(struct si_thread0_section *section,
LLVMBuildStore(ctx->ac.builder, *result, section->vgpr_result);
lp_build_endif(&section->if_thread0);
ac_build_endif(&ctx->ac, 12601);
/* Broadcast the result from thread 0 to all threads. */
*result = ac_build_readlane(&ctx->ac,
@@ -597,10 +596,9 @@ void si_build_prim_discard_compute_shader(struct si_shader_context *ctx)
* // Just read the value that previous waves stored.
* first_is_odd = ds.ordered.add(0);
*/
struct lp_build_if_state if_overwrite_counter;
lp_build_if(&if_overwrite_counter, &ctx->gallivm,
LLVMBuildOr(builder, is_first_wave,
current_wave_resets_index, ""));
ac_build_ifcc(&ctx->ac,
LLVMBuildOr(builder, is_first_wave,
current_wave_resets_index, ""), 12602);
{
/* The GDS address is always 0 with ordered append. */
tmp = si_build_ds_ordered_op(ctx, "swap",
@@ -608,7 +606,7 @@ void si_build_prim_discard_compute_shader(struct si_shader_context *ctx)
1, true, false);
LLVMBuildStore(builder, tmp, ret);
}
lp_build_else(&if_overwrite_counter);
ac_build_else(&ctx->ac, 12603);
{
/* Just read the value from GDS. */
tmp = si_build_ds_ordered_op(ctx, "add",
@@ -616,7 +614,7 @@ void si_build_prim_discard_compute_shader(struct si_shader_context *ctx)
1, true, false);
LLVMBuildStore(builder, tmp, ret);
}
lp_build_endif(&if_overwrite_counter);
ac_build_endif(&ctx->ac, 12602);
prev_wave_state = LLVMBuildLoad(builder, ret, "");
/* Ignore the return value if this is the first wave. */
@@ -760,15 +758,14 @@ void si_build_prim_discard_compute_shader(struct si_shader_context *ctx)
* previous = ds.ordered.add(num_prims_accepted) // add the primitive count
* }
*/
struct lp_build_if_state if_first_wave;
lp_build_if(&if_first_wave, &ctx->gallivm, is_first_wave);
ac_build_ifcc(&ctx->ac, is_first_wave, 12604);
{
/* The GDS address is always 0 with ordered append. */
si_build_ds_ordered_op(ctx, "swap", ordered_wave_id,
num_prims_accepted, 0, true, true);
LLVMBuildStore(builder, ctx->i32_0, tmp_store);
}
lp_build_else(&if_first_wave);
ac_build_else(&ctx->ac, 12605);
{
LLVMBuildStore(builder,
si_build_ds_ordered_op(ctx, "add", ordered_wave_id,
@@ -776,7 +773,7 @@ void si_build_prim_discard_compute_shader(struct si_shader_context *ctx)
true, true),
tmp_store);
}
lp_build_endif(&if_first_wave);
ac_build_endif(&ctx->ac, 12604);
start = LLVMBuildLoad(builder, tmp_store, "");
}
@@ -789,10 +786,9 @@ void si_build_prim_discard_compute_shader(struct si_shader_context *ctx)
* event like this.
*/
if (VERTEX_COUNTER_GDS_MODE == 2) {
struct lp_build_if_state if_last_wave;
lp_build_if(&if_last_wave, &ctx->gallivm,
LLVMBuildICmp(builder, LLVMIntEQ, global_thread_id,
last_wave_prim_id, ""));
ac_build_ifcc(&ctx->ac,
LLVMBuildICmp(builder, LLVMIntEQ, global_thread_id,
last_wave_prim_id, ""), 12606);
LLVMValueRef count = LLVMBuildAdd(builder, start, num_prims_accepted, "");
count = LLVMBuildMul(builder, count,
LLVMConstInt(ctx->i32, vertices_per_prim, 0), "");
@@ -816,7 +812,7 @@ void si_build_prim_discard_compute_shader(struct si_shader_context *ctx)
LLVMBuildStore(builder, count,
si_expand_32bit_pointer(ctx, vertex_count_addr));
}
lp_build_endif(&if_last_wave);
ac_build_endif(&ctx->ac, 12606);
} else {
/* For unordered modes that increment a vertex count instead of
* primitive count, convert it into the primitive index.
@@ -828,8 +824,7 @@ void si_build_prim_discard_compute_shader(struct si_shader_context *ctx)
/* Now we need to store the indices of accepted primitives into
* the output index buffer.
*/
struct lp_build_if_state if_accepted;
lp_build_if(&if_accepted, &ctx->gallivm, accepted);
ac_build_ifcc(&ctx->ac, accepted, 16607);
{
/* Get the number of bits set before the index of this thread. */
LLVMValueRef prim_index = ac_build_mbcnt(&ctx->ac, accepted_threadmask);
@@ -866,7 +861,7 @@ void si_build_prim_discard_compute_shader(struct si_shader_context *ctx)
vindex, ctx->i32_0, 3,
ac_glc | (INDEX_STORES_USE_SLC ? ac_slc : 0));
}
lp_build_endif(&if_accepted);
ac_build_endif(&ctx->ac, 16607);
LLVMBuildRetVoid(builder);
}