using png images in tex_unit test

hw/rtl/tex_unit/VX_tex_lsu_arb.v

@@ -1,128 +0,0 @@
-`include "../cache/VX_cache_define.vh"
-
-module VX_tex_lsu_arb #(    
-    parameter NUM_REQS      = 1, 
-    parameter LANES         = 1,
-    parameter WORD_SIZE     = 1,
-    parameter TAG_IN_WIDTH  = 1,
-    parameter TAG_OUT_WIDTH = 1,
-    parameter LOG_NUM_REQS  = `CLOG2(NUM_REQS)
-) (
-    input wire clk,
-    input wire reset,
-
-    // input requests    
-    input wire [NUM_REQS-1:0][LANES-1:0]                    req_valid_in, 
-    input wire [NUM_REQS-1:0][LANES-1:0]                    req_rw_in,   
-    input wire [NUM_REQS-1:0][LANES-1:0][WORD_SIZE-1:0]     req_byteen_in, 
-    input wire [NUM_REQS-1:0][LANES-1:0][`WORD_ADDR_WIDTH-1:0] req_addr_in, 
-    input wire [NUM_REQS-1:0][LANES-1:0][`WORD_WIDTH-1:0]   req_data_in,    
-    input wire [NUM_REQS-1:0][LANES-1:0][TAG_IN_WIDTH-1:0]  req_tag_in,  
-    output wire [NUM_REQS-1:0][LANES-1:0]                   req_ready_in,
-
-    // output request
-    output wire [LANES-1:0]                                 req_valid_out,
-    output wire [LANES-1:0]                                 req_rw_out,  
-    output wire [LANES-1:0][WORD_SIZE-1:0]                  req_byteen_out,  
-    output wire [LANES-1:0][`WORD_ADDR_WIDTH-1:0]           req_addr_out, 
-    output wire [LANES-1:0][`WORD_WIDTH-1:0]                req_data_out,   
-    output wire [LANES-1:0][TAG_OUT_WIDTH-1:0]              req_tag_out,    
-    input wire  [LANES-1:0]                                 req_ready_out,
-
-    // input response
-    input wire [LANES-1:0]                                  rsp_valid_in,    
-    input wire [LANES-1:0][`WORD_WIDTH-1:0]                 rsp_data_in,
-    input wire [TAG_OUT_WIDTH-1:0]                          rsp_tag_in,
-    output wire                                             rsp_ready_in,
-
-    // output responses
-    output wire [NUM_REQS-1:0][LANES-1:0]                   rsp_valid_out,
-    output wire [NUM_REQS-1:0][LANES-1:0][`WORD_WIDTH-1:0]  rsp_data_out,
-    output wire [NUM_REQS-1:0][TAG_IN_WIDTH-1:0]            rsp_tag_out,
-    input wire  [NUM_REQS-1:0]                              rsp_ready_out    
-);  
-    localparam REQ_DATAW = LANES * (1 + TAG_IN_WIDTH + `WORD_ADDR_WIDTH + 1 + WORD_SIZE + `WORD_WIDTH);
-
-    if (NUM_REQS > 1) begin
-
-        wire [NUM_REQS-1:0][REQ_DATAW-1:0] req_merged_data_in;
-        wire [NUM_REQS-1:0] req_valid_in_any;
-
-        for (genvar i = 0; i < NUM_REQS; i++) begin
-            assign req_merged_data_in[i] = {req_valid_in[i], req_tag_in[i], req_addr_in[i], req_rw_in[i], req_byteen_in[i], req_data_in[i]};
-            assign req_valid_in_any[i] = (| req_valid_in[i]);
-        end
-
-        wire                    sel_valid;
-        wire [LOG_NUM_REQS-1:0] sel_idx;
-        wire [NUM_REQS-1:0]     sel_1hot;
-
-        wire sel_enable = (| req_ready_out);
-
-        VX_rr_arbiter #(
-            .NUM_REQS(NUM_REQS),
-            .LOCK_ENABLE(1)
-        ) sel_arb (
-            .clk          (clk),
-            .reset        (reset),
-            .requests     (req_valid_in_any),
-            .enable       (sel_enable),      
-            .grant_valid  (sel_valid),
-            .grant_index  (sel_idx),
-            .grant_onehot (sel_1hot)
-        );
-
-        wire [LANES-1:0] req_valid_out_unqual;
-        wire [LANES-1:0][TAG_IN_WIDTH-1:0] req_tag_out_unqual;
-
-        assign {req_valid_out_unqual, req_tag_out_unqual, req_addr_out, req_rw_out, req_byteen_out, req_data_out} = req_merged_data_in[sel_idx];
-
-        assign req_valid_out = req_valid_out_unqual & {LANES{sel_valid}};
-
-        for (genvar i = 0; i < LANES; i++) begin
-            assign req_tag_out[i] = {req_tag_out_unqual[i], sel_idx};
-        end
-
-        for (genvar i = 0; i < NUM_REQS; i++) begin
-            assign req_ready_in[i] = req_ready_out & {LANES{sel_1hot[i]}};
-        end
-
-        ///////////////////////////////////////////////////////////////////////
-
-        wire [LOG_NUM_REQS-1:0] rsp_sel = rsp_tag_in[LOG_NUM_REQS-1:0];
-
-        reg [NUM_REQS-1:0][LANES-1:0] rsp_valid_out_unqual;
-        always @(*) begin
-            rsp_valid_out_unqual = '0;
-            rsp_valid_out_unqual[rsp_sel] = rsp_valid_in;
-        end
-        assign rsp_valid_out = rsp_valid_out_unqual;
-
-        for (genvar i = 0; i < NUM_REQS; i++) begin
-            assign rsp_data_out[i] = rsp_data_in;
-            assign rsp_tag_out[i]  = rsp_tag_in[LOG_NUM_REQS +: TAG_IN_WIDTH];            
-        end
-
-        assign rsp_ready_in = rsp_ready_out[rsp_sel];
-
-    end else begin
-
-        `UNUSED_VAR (clk)
-        `UNUSED_VAR (reset)
-
-        assign req_valid_out  = req_valid_in;
-        assign req_tag_out    = req_tag_in;
-        assign req_addr_out   = req_addr_in;
-        assign req_rw_out     = req_rw_in;
-        assign req_byteen_out = req_byteen_in;
-        assign req_data_out   = req_data_in;
-        assign req_ready_in   = req_ready_out;
-
-        assign rsp_valid_out  = rsp_valid_in;
-        assign rsp_tag_out    = rsp_tag_in;
-        assign rsp_data_out   = rsp_data_in;
-        assign rsp_ready_in   = rsp_ready_out;
-
-    end
-
-endmodule

tests/regression/tex/Makefile

@@ -19,13 +19,15 @@ VX_SRCS = kernel.c
 #CXXFLAGS += -std=c++11 -O2 -Wall -Wextra -Wfatal-errors
 CXXFLAGS += -std=c++11 -O0 -g -Wall -Wextra -Wfatal-errors
 
+CXXFLAGS += -DLUPNG_USE_ZLIB
+
 CXXFLAGS += -I$(VORTEX_DRV_PATH)/include
 
-LDFLAGS += -L$(VORTEX_DRV_PATH)/stub -lvortex
+LDFLAGS += -L$(VORTEX_DRV_PATH)/stub -lvortex  -lz
 
 PROJECT = tex
 
-SRCS = main.cpp utils.cpp
+SRCS = main.cpp utils.cpp tga.cpp lupng.c
 
 all: $(PROJECT) kernel.bin kernel.dump
  

tests/regression/tex/main.cpp

@@ -22,8 +22,8 @@
 ///////////////////////////////////////////////////////////////////////////////
 
 const char* kernel_file = "kernel.bin";
-const char* input_file  = "palette64.tga";
-const char* output_file = "output.tga";
+const char* input_file  = "palette64.png";
+const char* output_file = "output.png";
 int wrap    = 0;
 int filter  = 0;
 float scale = 1.0f;
@@ -101,8 +101,7 @@ void cleanup() {
 int run_test(const kernel_arg_t& kernel_arg, 
              uint32_t buf_size, 
              uint32_t width, 
-             uint32_t height, 
-             uint32_t bpp) {
+             uint32_t height) {
   auto time_start = std::chrono::high_resolution_clock::now();
 
   // start device
@@ -130,7 +129,7 @@ int run_test(const kernel_arg_t& kernel_arg,
   // save output image
   std::cout << "save output image" << std::endl;  
   //dump_image(dst_pixels, width, height, bpp);
-  RT_CHECK(SaveTGA(output_file, dst_pixels, width, height, bpp));
+  RT_CHECK(SaveImage(output_file, FORMAT_A8R8G8B8, dst_pixels, width, height));
 
   return 0;
 }
@@ -140,13 +139,11 @@ int main(int argc, char *argv[]) {
   std::vector<uint8_t> src_pixels;
   uint32_t src_width;
   uint32_t src_height;
-  uint32_t src_bpp;
- 
+  
   // parse command arguments
   parse_args(argc, argv);
 
-  std::vector<uint8_t> tmp_pixels;
-  RT_CHECK(LoadTGA(input_file, tmp_pixels, &src_width, &src_height));
+  RT_CHECK(LoadImage(input_file, eformat, src_pixels, &src_width, &src_height));
 
   // check power of two support
   if (!ISPOW2(src_width) || !ISPOW2(src_height)) {
@@ -154,8 +151,7 @@ int main(int argc, char *argv[]) {
     return -1;
   }
 
-  RT_CHECK(ConvertImage(src_pixels, tmp_pixels, src_width, src_height, FORMAT_A8R8G8B8, eformat));
-  src_bpp = Format::GetInfo(eformat).BytePerPixel;
+  uint32_t src_bpp = Format::GetInfo(eformat).BytePerPixel;
   
   //dump_image(src_pixels, src_width, src_height, src_bpp);
 
@@ -248,7 +244,7 @@ int main(int argc, char *argv[]) {
 
   // run tests
   std::cout << "run tests" << std::endl;
-  RT_CHECK(run_test(kernel_arg, dst_bufsize, dst_width, dst_height, dst_bpp));
+  RT_CHECK(run_test(kernel_arg, dst_bufsize, dst_width, dst_height));
 
   // cleanup
   std::cout << "cleanup" << std::endl;  

tests/regression/tex/utils.cpp

@@ -1,146 +1,123 @@
 #include "utils.h"
-#include <fstream>
 #include <assert.h>
+#include <cstring>
+#include "blitter.h"
 #include "format.h"
+#include "tga.h"
+#include "lupng.h"
 
-struct __attribute__((__packed__)) tga_header_t {
-  int8_t idlength;
-  int8_t colormaptype;
-  int8_t imagetype;
-  int16_t colormaporigin;
-  int16_t colormaplength;
-  int8_t colormapdepth;
-  int16_t xoffset;
-  int16_t yoffset;
-  int16_t width;
-  int16_t height;
-  int8_t bitsperpixel;
-  int8_t imagedescriptor;
-};
+std::string getFileExt(const std::string& str) {
+   auto i = str.rfind('.');
+   if (i != std::string::npos) {
+      return str.substr(i+1);
+   }
+   return("");
+}
 
-int LoadTGA(const char *filename, 
-            std::vector<uint8_t> &pixels, 
-            uint32_t *width, 
-            uint32_t *height) {
-  std::ifstream ifs(filename, std::ios::in | std::ios::binary);
-  if (!ifs.is_open()) {
-    std::cerr << "couldn't open file: " << filename << "!" << std::endl;
-    return -1;
-  }
+bool iequals(const std::string& a, const std::string& b) {
+    auto sz = a.size();
+    if (b.size() != sz)
+        return false;
+    for (size_t i = 0; i < sz; ++i) {
+        if (tolower(a[i]) != tolower(b[i]))
+            return false;
+    }
+    return true;
+}
 
-  tga_header_t header;
-  ifs.read(reinterpret_cast<char *>(&header), sizeof(tga_header_t));
-  if (ifs.fail()) {
-    std::cerr << "invalid TGA file header!" << std::endl;
-    return -1;
-  }
+int LoadImage(const char *filename,
+              ePixelFormat format, 
+              std::vector<uint8_t> &pixels, 
+              uint32_t *width,
+              uint32_t *height) {
+  uint32_t img_width;
+  uint32_t img_height;
+  uint32_t img_bpp;
 
-  if (header.imagetype != 2) {
-    std::cerr << "unsupported TGA encoding format!" << std::endl;
-    return -1;
-  }
-
-  ifs.seekg(header.idlength, std::ios::cur); // skip string
-  if (ifs.fail()) {
-    std::cerr << "invalid TGA file!" << std::endl;
-    return -1;
-  }
-
-  switch (header.bitsperpixel) {
-  case 16:
-  case 24:
-  case 32: {
-    auto stride = header.bitsperpixel / 8;
-    std::vector<uint8_t> staging(stride * header.width * header.height);
-
-    // Read pixels data
-    ifs.read((char*)staging.data(), staging.size());
-    if (ifs.fail()) {
-      std::cerr << "invalid TGA file!" << std::endl;
+  auto ext = getFileExt(filename);
+  if (iequals(ext, "tga")) {
+    int ret = LoadTGA(filename, pixels, &img_width, &img_height, &img_bpp);
+    if (ret)
+      return ret;
+  } else 
+  if (iequals(ext, "png")) {
+    auto image = luPngReadFile(filename, NULL);
+    if (image == NULL)
+      return -1;
+    if (image->depth != 8 
+      || (image->channels != 3 
+       && image->channels != 4)) {
+      luImageRelease(image, NULL);
+      std::cerr << "invalid png file format!" << std::endl;  
       return -1;
     }
-
-    // format conversion to RGBA
-    pixels.resize(4 * header.width * header.height);
-    const uint8_t* src_bytes = staging.data();
-    uint32_t* dst_bytes = (uint32_t*)pixels.data();
-    for (const uint8_t* const src_end = src_bytes + staging.size(); 
-         src_bytes != src_end; 
-         src_bytes += stride) {
-      ColorARGB color;        
-      switch (stride) {
-      case 2: 
-        color = Format::ConvertFrom<FORMAT_A1R5G5B5, true>(src_bytes);         
-        break;
-      case 3: 
-        color = Format::ConvertFrom<FORMAT_R8G8B8, true>(src_bytes); 
-        break;
-      case 4: 
-        color = Format::ConvertFrom<FORMAT_A8R8G8B8, true>(src_bytes); 
-        break;
-      default:
-        std::abort();            
-      }
-      *dst_bytes++ = color;
-    }
-    break;
-  }
-  default:
-    std::cerr << "unsupported TGA bitsperpixel!" << std::endl;
+    pixels.resize(image->channels * image->width * image->height);
+    memcpy(pixels.data(), image->data, pixels.size());
+    img_width  = image->width;
+    img_height = image->height;
+    img_bpp    = image->channels;
+    luImageRelease(image, NULL);
+  } else {
+    std::cerr << "invalid file extension: " << ext << "!" << std::endl;
     return -1;
-  } 
+  }
 
-  *width = header.width;
-  *height = header.height;
+  ePixelFormat img_format;  
+  switch (img_bpp) {
+  case 1: 
+    img_format = FORMAT_A8;
+    break;
+  case 2: 
+    img_format = FORMAT_A1R5G5B5;
+    break;
+  case 3: 
+    img_format = FORMAT_R8G8B8; 
+    break;
+  case 4: 
+    img_format = FORMAT_A8R8G8B8; 
+    break;
+  default:
+    std::abort();            
+  }
 
+  if (img_format != format) {
+    // format conversion to RGBA
+    std::vector<uint8_t> staging;    
+    int ret = ConvertImage(staging, pixels, img_width, img_height, img_format, format);
+    if (ret)
+      return ret;
+    pixels.swap(staging);
+  }
+
+  *width  = img_width;
+  *height = img_height;
+  
   return 0;
 }
 
-int SaveTGA(const char *filename, 
-            const std::vector<uint8_t> &pixels, 
-            uint32_t width, 
-            uint32_t height, 
-            uint32_t bpp) {              
-  std::ofstream ofs(filename, std::ios::out | std::ios::binary);
-  if (!ofs.is_open()) {
-    std::cerr << "couldn't create file: " << filename << "!" << std::endl;
+int SaveImage(const char *filename,
+              ePixelFormat format,
+              const std::vector<uint8_t> &pixels, 
+              uint32_t width,
+              uint32_t height) {
+  uint32_t bpp = Format::GetInfo(format).BytePerPixel;
+  auto ext = getFileExt(filename);
+  if (iequals(ext, "tga")) {
+    return SaveTGA(filename, pixels, width, height, bpp);
+  } else 
+  if (iequals(ext, "png")) {
+    LuImage image;
+    image.width    = width;
+    image.height   = height;
+    image.depth    = 8;
+    image.channels = bpp;
+    image.data     = (uint8_t*)pixels.data();
+    return luPngWriteFile(filename, &image);
+  } else {
+    std::cerr << "invalid file extension: " << ext << "!" << std::endl;
     return -1;
   }
 
-  if (bpp < 2 || bpp > 4) {        
-    std::cerr << "unsupported pixel stride: " << bpp << "!" << std::endl;
-    return -1;
-  }
-
-  tga_header_t header;
-  header.idlength = 0;
-  header.colormaptype = 0; // no palette
-  header.imagetype = 2; // color mapped data
-  header.colormaporigin = 0;
-  header.colormaplength = 0;
-  header.colormapdepth = 0;
-  header.xoffset = 0;
-  header.yoffset = 0;
-  header.width = width;
-  header.height = height;
-  header.bitsperpixel = bpp * 8;
-  header.imagedescriptor = 0;
-
-  // write header
-  ofs.write(reinterpret_cast<char *>(&header), sizeof(tga_header_t));
-
-  // write pixel data
-  uint32_t pitch = bpp * width;
-  const uint8_t* pixel_bytes = pixels.data() + (height - 1) * pitch;
-  for (uint32_t y = 0; y < height; ++y) {
-    const uint8_t* pixel_row = pixel_bytes;
-    for (uint32_t x = 0; x < width; ++x) {
-      ofs.write((const char*)pixel_row, bpp);      
-      pixel_row += bpp;
-    }
-    pixel_bytes -= pitch;
-  }
-
   return 0;
 }
 

tests/regression/tex/utils.h

@@ -1,7 +1,7 @@
 #include <cstdint>
 #include <vector>
 #include <iostream>
-#include "blitter.h"
+#include "surfacedesc.h"
 
 #define ISPOW2(x)   (((x) != 0) && (0 == ((x) & ((x) - 1))))
 
@@ -9,16 +9,17 @@ inline uint32_t ilog2 (uint32_t value) {
   	return (uint32_t)(sizeof(uint32_t) * 8UL) - (uint32_t)__builtin_clzl((value << 1) - 1UL) - 1;
 }
 
-int LoadTGA(const char *filename, 
-            std::vector<uint8_t> &pixels, 
-            uint32_t *width,
-            uint32_t *height);
+int LoadImage(const char *filename,
+              ePixelFormat format,
+              std::vector<uint8_t> &pixels,
+              uint32_t *width,
+              uint32_t *height);
 
-int SaveTGA(const char *filename, 
-            const std::vector<uint8_t> &pixels, 
-            uint32_t width,
-            uint32_t height, 
-            uint32_t bpp);
+int SaveImage(const char *filename,
+              ePixelFormat format,
+              const std::vector<uint8_t> &pixels,
+              uint32_t width,
+              uint32_t height);
 
 int CopyBuffers(SurfaceDesc &dstDesc, 
                 int32_t dstOffsetX,