radv: fix sample_mask_in loading. (v3.1)

This is ported from radeonsi and fixes:
dEQP-VK.pipeline.multisample_shader_builtin.sample_mask.bit_*

v2: don't call this path for radeonsi, it does it in the epilog.
use the radeonsi code path.
v3: handle NULL pCreateInfo->pMultisampleState properly (Samuel)
v3.1: set ps_iter_samples default to 1 (Bas)

Reviewed-by: Bas Nieuwenhuizen <[email protected]>
Fixes: bdcbe7c76 (radv: add sample mask input support)
Signed-off-by: Dave Airlie <[email protected]>
(cherry picked from commit 766589d89a211e67f313e8cb38f2d05b09975f96)

4 files changed, 56 insertions, 6 deletions

diff --git a/src/amd/common/ac_nir_to_llvm.c b/src/amd/common/ac_nir_to_llvm.c
index 13792e71611..bb4344d5366 100644
--- a/src/amd/common/ac_nir_to_llvm.c
+++ b/src/amd/common/ac_nir_to_llvm.c
@@ -4049,6 +4049,30 @@ static LLVMValueRef load_sample_pos(struct ac_nir_context *ctx)
 	return ac_build_gather_values(&ctx->ac, values, 2);
 }
 
+static LLVMValueRef load_sample_mask_in(struct ac_nir_context *ctx)
+{
+	uint8_t log2_ps_iter_samples = ctx->nctx->shader_info->info.ps.force_persample ? ctx->nctx->options->key.fs.log2_num_samples : ctx->nctx->options->key.fs.log2_ps_iter_samples;
+
+	/* The bit pattern matches that used by fixed function fragment
+	 * processing. */
+	static const uint16_t ps_iter_masks[] = {
+		0xffff, /* not used */
+		0x5555,
+		0x1111,
+		0x0101,
+		0x0001,
+	};
+	assert(log2_ps_iter_samples < ARRAY_SIZE(ps_iter_masks));
+
+	uint32_t ps_iter_mask = ps_iter_masks[log2_ps_iter_samples];
+
+	LLVMValueRef result, sample_id;
+	sample_id = unpack_param(&ctx->ac, ctx->abi->ancillary, 8, 4);
+	sample_id = LLVMBuildShl(ctx->ac.builder, LLVMConstInt(ctx->ac.i32, ps_iter_mask, false), sample_id, "");
+	result = LLVMBuildAnd(ctx->ac.builder, sample_id, ctx->abi->sample_coverage, "");
+	return result;
+}
+
 static LLVMValueRef visit_interp(struct nir_to_llvm_context *ctx,
 				 const nir_intrinsic_instr *instr)
 {
@@ -4353,7 +4377,10 @@ static void visit_intrinsic(struct ac_nir_context *ctx,
 		result = load_sample_pos(ctx);
 		break;
 	case nir_intrinsic_load_sample_mask_in:
-		result = ctx->abi->sample_coverage;
+		if (ctx->nctx)
+			result = load_sample_mask_in(ctx);
+		else
+			result = ctx->abi->sample_coverage;
 		break;
 	case nir_intrinsic_load_frag_coord: {
 		LLVMValueRef values[4] = {
diff --git a/src/amd/common/ac_nir_to_llvm.h b/src/amd/common/ac_nir_to_llvm.h
index 62ea38be373..1656289e06e 100644
--- a/src/amd/common/ac_nir_to_llvm.h
+++ b/src/amd/common/ac_nir_to_llvm.h
@@ -60,6 +60,8 @@ struct ac_tcs_variant_key {
 
 struct ac_fs_variant_key {
 	uint32_t col_format;
+	uint8_t log2_ps_iter_samples;
+	uint8_t log2_num_samples;
 	uint32_t is_int8;
 	uint32_t is_int10;
 	uint32_t multisample : 1;
diff --git a/src/amd/vulkan/radv_pipeline.c b/src/amd/vulkan/radv_pipeline.c
index a49fe059934..21333b808ab 100644
--- a/src/amd/vulkan/radv_pipeline.c
+++ b/src/amd/vulkan/radv_pipeline.c
@@ -798,6 +798,18 @@ radv_pipeline_init_raster_state(struct radv_pipeline *pipeline,
 
 }
 
+static uint8_t radv_pipeline_get_ps_iter_samples(const VkPipelineMultisampleStateCreateInfo *vkms)
+{
+	uint32_t num_samples = vkms->rasterizationSamples;
+	uint32_t ps_iter_samples = 1;
+
+	if (vkms->sampleShadingEnable) {
+		ps_iter_samples = ceil(vkms->minSampleShading * num_samples);
+		ps_iter_samples = util_next_power_of_two(ps_iter_samples);
+	}
+	return ps_iter_samples;
+}
+
 static void
 radv_pipeline_init_multisample_state(struct radv_pipeline *pipeline,
 				     const VkGraphicsPipelineCreateInfo *pCreateInfo)
@@ -813,9 +825,9 @@ radv_pipeline_init_multisample_state(struct radv_pipeline *pipeline,
 	else
 		ms->num_samples = 1;
 
-	if (vkms && vkms->sampleShadingEnable) {
-		ps_iter_samples = ceil(vkms->minSampleShading * ms->num_samples);
-	} else if (pipeline->shaders[MESA_SHADER_FRAGMENT]->info.info.ps.force_persample) {
+	if (vkms)
+		ps_iter_samples = radv_pipeline_get_ps_iter_samples(vkms);
+	if (vkms && !vkms->sampleShadingEnable && pipeline->shaders[MESA_SHADER_FRAGMENT]->info.info.ps.force_persample) {
 		ps_iter_samples = ms->num_samples;
 	}
 
@@ -838,7 +850,7 @@ radv_pipeline_init_multisample_state(struct radv_pipeline *pipeline,
 
 	if (ms->num_samples > 1) {
 		unsigned log_samples = util_logbase2(ms->num_samples);
-		unsigned log_ps_iter_samples = util_logbase2(util_next_power_of_two(ps_iter_samples));
+		unsigned log_ps_iter_samples = util_logbase2(ps_iter_samples);
 		ms->pa_sc_mode_cntl_0 |= S_028A48_MSAA_ENABLE(1);
 		ms->pa_sc_line_cntl |= S_028BDC_EXPAND_LINE_WIDTH(1); /* CM_R_028BDC_PA_SC_LINE_CNTL */
 		ms->db_eqaa |= S_028804_MAX_ANCHOR_SAMPLES(log_samples) |
@@ -1745,8 +1757,13 @@ radv_generate_graphics_pipeline_key(struct radv_pipeline *pipeline,
 
 
 	if (pCreateInfo->pMultisampleState &&
-	    pCreateInfo->pMultisampleState->rasterizationSamples > 1)
+	    pCreateInfo->pMultisampleState->rasterizationSamples > 1) {
+		uint32_t num_samples = pCreateInfo->pMultisampleState->rasterizationSamples;
+		uint32_t ps_iter_samples = radv_pipeline_get_ps_iter_samples(pCreateInfo->pMultisampleState);
 		key.multisample = true;
+		key.log2_num_samples = util_logbase2(num_samples);
+		key.log2_ps_iter_samples = util_logbase2(ps_iter_samples);
+	}
 
 	key.col_format = pipeline->graphics.blend.spi_shader_col_format;
 	if (pipeline->device->physical_device->rad_info.chip_class < VI)
@@ -1784,6 +1801,8 @@ radv_fill_shader_keys(struct ac_shader_variant_key *keys,
 	keys[MESA_SHADER_FRAGMENT].fs.col_format = key->col_format;
 	keys[MESA_SHADER_FRAGMENT].fs.is_int8 = key->is_int8;
 	keys[MESA_SHADER_FRAGMENT].fs.is_int10 = key->is_int10;
+	keys[MESA_SHADER_FRAGMENT].fs.log2_ps_iter_samples = key->log2_ps_iter_samples;
+	keys[MESA_SHADER_FRAGMENT].fs.log2_num_samples = key->log2_num_samples;
 }
 
 static void
diff --git a/src/amd/vulkan/radv_private.h b/src/amd/vulkan/radv_private.h
index 808db0007dc..711ae719266 100644
--- a/src/amd/vulkan/radv_private.h
+++ b/src/amd/vulkan/radv_private.h
@@ -331,6 +331,8 @@ struct radv_pipeline_key {
 	uint32_t col_format;
 	uint32_t is_int8;
 	uint32_t is_int10;
+	uint8_t log2_ps_iter_samples;
+	uint8_t log2_num_samples;
 	uint32_t multisample : 1;
 	uint32_t has_multiview_view_index : 1;
 };