From 13c423629ea1c8d5822d2266d0969c75ea036e9c Mon Sep 17 00:00:00 2001 From: Rhys Perry Date: Fri, 26 Apr 2019 15:39:51 +0100 Subject: [PATCH] radv: fix set_output_usage_mask() with composite and 64-bit types It previously used var->type instead of deref_instr->type and didn't handle 64-bit outputs. This fixes lots of transform feedback CTS tests involving transform feedback and geometry shaders (mostly dEQP-VK.transform_feedback.fuzz.random_geometry.*) v2: fix writemask widening when comp != 0 v3: fix 64-bit variables when comp != 0, again Signed-off-by: Rhys Perry Cc: 19.0 19.1 Reviewed-by: Samuel Pitoiset --- src/amd/vulkan/radv_shader_info.c | 21 +++++++++++++++++---- 1 file changed, 17 insertions(+), 4 deletions(-) diff --git a/src/amd/vulkan/radv_shader_info.c b/src/amd/vulkan/radv_shader_info.c index 932a1852266..e771ad79878 100644 --- a/src/amd/vulkan/radv_shader_info.c +++ b/src/amd/vulkan/radv_shader_info.c @@ -112,6 +112,15 @@ gather_intrinsic_load_deref_info(const nir_shader *nir, } } +static uint32_t +widen_writemask(uint32_t wrmask) +{ + uint32_t new_wrmask = 0; + for(unsigned i = 0; i < 4; i++) + new_wrmask |= (wrmask & (1 << i) ? 0x3 : 0x0) << (i * 2); + return new_wrmask; +} + static void set_output_usage_mask(const nir_shader *nir, const nir_intrinsic_instr *instr, uint8_t *output_usage_mask) @@ -119,7 +128,7 @@ set_output_usage_mask(const nir_shader *nir, const nir_intrinsic_instr *instr, nir_deref_instr *deref_instr = nir_instr_as_deref(instr->src[0].ssa->parent_instr); nir_variable *var = nir_deref_instr_get_variable(deref_instr); - unsigned attrib_count = glsl_count_attribute_slots(var->type, false); + unsigned attrib_count = glsl_count_attribute_slots(deref_instr->type, false); unsigned idx = var->data.location; unsigned comp = var->data.location_frac; unsigned const_offset = 0; @@ -127,15 +136,19 @@ set_output_usage_mask(const nir_shader *nir, const nir_intrinsic_instr *instr, get_deref_offset(deref_instr, &const_offset); if (var->data.compact) { + assert(!glsl_type_is_64bit(deref_instr->type)); const_offset += comp; output_usage_mask[idx + const_offset / 4] |= 1 << (const_offset % 4); return; } - for (unsigned i = 0; i < attrib_count; i++) { + uint32_t wrmask = nir_intrinsic_write_mask(instr); + if (glsl_type_is_64bit(deref_instr->type)) + wrmask = widen_writemask(wrmask); + + for (unsigned i = 0; i < attrib_count; i++) output_usage_mask[idx + i + const_offset] |= - instr->const_index[0] << comp; - } + ((wrmask >> (i * 4)) & 0xf) << comp; } static void -- 2.30.2