ARM Skia NEON patches - 33 - Convolution filter

src/opts/SkBitmapProcState_arm_neon.cpp

 
 /*
  * Copyright 2012 Google Inc.
  *
  * Use of this source code is governed by a BSD-style license that can be
  * found in the LICENSE file.
  */
 #include "SkBitmapProcState.h"
 #include "SkBitmapProcState_filter.h"
 #include "SkColorPriv.h"
@@ skipping 72 matching lines @@
     SI8_D16_nofilter_DXDY_neon,
     SI8_D16_nofilter_DX_neon,
     SI8_D16_filter_DXDY_neon,
     SI8_D16_filter_DX_neon,
 
     // Don't support 4444 -> 565
     NULL, NULL, NULL, NULL,
     // Don't support A8 -> 565
     NULL, NULL, NULL, NULL
 };
+
+///////////////////////////////////////////////////////////////////////////////
+
+#include <arm_neon.h>
+#include "SkConvolver.h"
+
+// Convolves horizontally along a single row. The row data is given in
+// |srcData| and continues for the numValues() of the filter.
+void convolveHorizontally_neon(const unsigned char* srcData,
+                               const SkConvolutionFilter1D& filter,
+                               unsigned char* outRow,
+                               bool hasAlpha) {
+    // Loop over each pixel on this row in the output image.
+    int numValues = filter.numValues();
+    for (int outX = 0; outX < numValues; outX++) {
+        uint8x8_t coeff_mask0 = vcreate_u8(0x0100010001000100);
+        uint8x8_t coeff_mask1 = vcreate_u8(0x0302030203020302);
+        uint8x8_t coeff_mask2 = vcreate_u8(0x0504050405040504);
+        uint8x8_t coeff_mask3 = vcreate_u8(0x0706070607060706);
+        // Get the filter that determines the current output pixel.
+        int filterOffset, filterLength;
+        const SkConvolutionFilter1D::ConvolutionFixed* filterValues =
+            filter.FilterForValue(outX, &filterOffset, &filterLength);
+
+        // Compute the first pixel in this row that the filter affects. It will
+        // touch |filterLength| pixels (4 bytes each) after this.
+        const unsigned char* rowToFilter = &srcData[filterOffset * 4];
+
+        // Apply the filter to the row to get the destination pixel in |accum|.
+        int32x4_t accum = vdupq_n_s32(0);
+        for (int filterX = 0; filterX < filterLength >> 2; filterX++) {
+            // Load 4 coefficients
+            int16x4_t coeffs, coeff0, coeff1, coeff2, coeff3;
+            coeffs = vld1_s16(filterValues);
+            coeff0 = vreinterpret_s16_u8(vtbl1_u8(vreinterpret_u8_s16(coeffs), coeff_mask0));
+            coeff1 = vreinterpret_s16_u8(vtbl1_u8(vreinterpret_u8_s16(coeffs), coeff_mask1));
+            coeff2 = vreinterpret_s16_u8(vtbl1_u8(vreinterpret_u8_s16(coeffs), coeff_mask2));
+            coeff3 = vreinterpret_s16_u8(vtbl1_u8(vreinterpret_u8_s16(coeffs), coeff_mask3));
+
+            // Load pixels and calc
+            uint8x16_t pixels = vld1q_u8(rowToFilter);
+            int16x8_t p01_16 = vreinterpretq_s16_u16(vmovl_u8(vget_low_u8(pixels)));
+            int16x8_t p23_16 = vreinterpretq_s16_u16(vmovl_u8(vget_high_u8(pixels)));
+
+            int16x4_t p0_src = vget_low_s16(p01_16);
+            int16x4_t p1_src = vget_high_s16(p01_16);
+            int16x4_t p2_src = vget_low_s16(p23_16);
+            int16x4_t p3_src = vget_high_s16(p23_16);
+
+            int32x4_t p0 = vmull_s16(p0_src, coeff0);
+            int32x4_t p1 = vmull_s16(p1_src, coeff1);
+            int32x4_t p2 = vmull_s16(p2_src, coeff2);
+            int32x4_t p3 = vmull_s16(p3_src, coeff3);
+
+            accum += p0;
+            accum += p1;
+            accum += p2;
+            accum += p3;
+
+            // Advance the pointers
+            rowToFilter += 16;
+            filterValues += 4;
+        }
+        int r = filterLength & 3;
+        if (r) {
+            const uint16_t mask[4][4] = {
+                {0, 0, 0, 0},
+                {0xFFFF, 0, 0, 0},
+                {0xFFFF, 0xFFFF, 0, 0},
+                {0xFFFF, 0xFFFF, 0xFFFF, 0}
+            };
+            uint16x4_t coeffs;
+            int16x4_t coeff0, coeff1, coeff2;
+            coeffs = vld1_u16(reinterpret_cast<const uint16_t*>(filterValues));
+            coeffs &= vld1_u16(&mask[r][0]);
+            coeff0 = vreinterpret_s16_u8(vtbl1_u8(vreinterpret_u8_u16(coeffs), coeff_mask0));
+            coeff1 = vreinterpret_s16_u8(vtbl1_u8(vreinterpret_u8_u16(coeffs), coeff_mask1));
+            coeff2 = vreinterpret_s16_u8(vtbl1_u8(vreinterpret_u8_u16(coeffs), coeff_mask2));
+
+            // Load pixels and calc
+            uint8x16_t pixels = vld1q_u8(rowToFilter);
+            int16x8_t p01_16 = vreinterpretq_s16_u16(vmovl_u8(vget_low_u8(pixels)));
+            int16x8_t p23_16 = vreinterpretq_s16_u16(vmovl_u8(vget_high_u8(pixels)));
+            int32x4_t p0 = vmull_s16(vget_low_s16(p01_16), coeff0);
+            int32x4_t p1 = vmull_s16(vget_high_s16(p01_16), coeff1);
+            int32x4_t p2 = vmull_s16(vget_low_s16(p23_16), coeff2);
+
+            accum += p0;
+            accum += p1;
+            accum += p2;
+        }
+
+        // Bring this value back in range. All of the filter scaling factors
+        // are in fixed point with kShiftBits bits of fractional part.
+        accum = vshrq_n_s32(accum, SkConvolutionFilter1D::kShiftBits);
+
+        // Pack and store the new pixel.
+        int16x4_t accum16 = vqmovn_s32(accum);
+        uint8x8_t accum8 = vqmovun_s16(vcombine_s16(accum16, accum16));
+        vst1_lane_u32(reinterpret_cast<uint32_t*>(outRow), vreinterpret_u32_u8(accum8), 0);
+        outRow += 4;
+    }
+}
+
+// Does vertical convolution to produce one output row. The filter values and
+// length are given in the first two parameters. These are applied to each
+// of the rows pointed to in the |sourceDataRows| array, with each row
+// being |pixelWidth| wide.
+//
+// The output must have room for |pixelWidth * 4| bytes.
+template<bool hasAlpha>
+void convolveVertically_neon(const SkConvolutionFilter1D::ConvolutionFixed* filterValues,
+                             int filterLength,
+                             unsigned char* const* sourceDataRows,
+                             int pixelWidth,
+                             unsigned char* outRow) {
+    int width = pixelWidth & ~3;
+
+    int32x4_t accum0, accum1, accum2, accum3;
+    int16x4_t coeff16;
+
+    // Output four pixels per iteration (16 bytes).
+    for (int outX = 0; outX < width; outX += 4) {
+
+        // Accumulated result for each pixel. 32 bits per RGBA channel.
+        accum0 = accum1 = accum2 = accum3 = vdupq_n_s32(0);
+
+        // Convolve with one filter coefficient per iteration.
+        for (int filterY = 0; filterY < filterLength; filterY++) {
+
+            // Duplicate the filter coefficient 4 times.
+            // [16] cj cj cj cj
+            coeff16 = vdup_n_s16(filterValues[filterY]);
+
+            // Load four pixels (16 bytes) together.
+            // [8] a3 b3 g3 r3 a2 b2 g2 r2 a1 b1 g1 r1 a0 b0 g0 r0
+            uint8x16_t src8 = vld1q_u8(&sourceDataRows[filterY][outX << 2]);
+
+            int16x8_t src16_01 = vreinterpretq_s16_u16(vmovl_u8(vget_low_u8(src8)));
+            int16x8_t src16_23 = vreinterpretq_s16_u16(vmovl_u8(vget_high_u8(src8)));
+            int16x4_t src16_0 = vget_low_s16(src16_01);
+            int16x4_t src16_1 = vget_high_s16(src16_01);
+            int16x4_t src16_2 = vget_low_s16(src16_23);
+            int16x4_t src16_3 = vget_high_s16(src16_23);
+
+            accum0 += vmull_s16(src16_0, coeff16);
+            accum1 += vmull_s16(src16_1, coeff16);
+            accum2 += vmull_s16(src16_2, coeff16);
+            accum3 += vmull_s16(src16_3, coeff16);
+        }
+
+        // Shift right for fixed point implementation.
+        accum0 = vshrq_n_s32(accum0, SkConvolutionFilter1D::kShiftBits);
+        accum1 = vshrq_n_s32(accum1, SkConvolutionFilter1D::kShiftBits);
+        accum2 = vshrq_n_s32(accum2, SkConvolutionFilter1D::kShiftBits);
+        accum3 = vshrq_n_s32(accum3, SkConvolutionFilter1D::kShiftBits);
+
+        // Packing 32 bits |accum| to 16 bits per channel (signed saturation).
+        // [16] a1 b1 g1 r1 a0 b0 g0 r0
+        int16x8_t accum16_0 = vcombine_s16(vqmovn_s32(accum0), vqmovn_s32(accum1));
+        // [16] a3 b3 g3 r3 a2 b2 g2 r2
+        int16x8_t accum16_1 = vcombine_s16(vqmovn_s32(accum2), vqmovn_s32(accum3));
+
+        // Packing 16 bits |accum| to 8 bits per channel (unsigned saturation).
+        // [8] a3 b3 g3 r3 a2 b2 g2 r2 a1 b1 g1 r1 a0 b0 g0 r0
+        uint8x16_t accum8 = vcombine_u8(vqmovun_s16(accum16_0), vqmovun_s16(accum16_1));
+
+        if (hasAlpha) {
+            // Compute the max(ri, gi, bi) for each pixel.
+            // [8] xx a3 b3 g3 xx a2 b2 g2 xx a1 b1 g1 xx a0 b0 g0
+            uint8x16_t a = vreinterpretq_u8_u32(vshrq_n_u32(vreinterpretq_u32_u8(accum8), 8));
+            // [8] xx xx xx max3 xx xx xx max2 xx xx xx max1 xx xx xx max0
+            uint8x16_t b = vmaxq_u8(a, accum8); // Max of r and g
+            // [8] xx xx a3 b3 xx xx a2 b2 xx xx a1 b1 xx xx a0 b0
+            a = vreinterpretq_u8_u32(vshrq_n_u32(vreinterpretq_u32_u8(accum8), 16));
+            // [8] xx xx xx max3 xx xx xx max2 xx xx xx max1 xx xx xx max0
+            b = vmaxq_u8(a, b); // Max of r and g and b.
+            // [8] max3 00 00 00 max2 00 00 00 max1 00 00 00 max0 00 00 00
+            b = vreinterpretq_u8_u32(vshlq_n_u32(vreinterpretq_u32_u8(b), 24));
+
+            // Make sure the value of alpha channel is always larger than maximum
+            // value of color channels.
+            accum8 = vmaxq_u8(b, accum8);
+        } else {
+            // Set value of alpha channels to 0xFF.
+            accum8 = vreinterpretq_u8_u32(vreinterpretq_u32_u8(accum8) | vdupq_n_u32(0xFF000000));
+        }
+
+        // Store the convolution result (16 bytes) and advance the pixel pointers.
+        vst1q_u8(outRow, accum8);
+        outRow += 16;
+    }
+
+    // Process the leftovers when the width of the output is not divisible
+    // by 4, that is at most 3 pixels.
+    int r = pixelWidth & 3;
+    if (r) {
+
+        accum0 = accum1 = accum2 = vdupq_n_s32(0);
+
+        for (int filterY = 0; filterY < filterLength; ++filterY) {
+            coeff16 = vdup_n_s16(filterValues[filterY]);
+
+            // [8] a3 b3 g3 r3 a2 b2 g2 r2 a1 b1 g1 r1 a0 b0 g0 r0
+            uint8x16_t src8 = vld1q_u8(&sourceDataRows[filterY][width << 2]);
+
+            int16x8_t src16_01 = vreinterpretq_s16_u16(vmovl_u8(vget_low_u8(src8)));
+            int16x8_t src16_23 = vreinterpretq_s16_u16(vmovl_u8(vget_high_u8(src8)));
+            int16x4_t src16_0 = vget_low_s16(src16_01);
+            int16x4_t src16_1 = vget_high_s16(src16_01);
+            int16x4_t src16_2 = vget_low_s16(src16_23);
+
+            accum0 += vmull_s16(src16_0, coeff16);
+            accum1 += vmull_s16(src16_1, coeff16);
+            accum2 += vmull_s16(src16_2, coeff16);
+        }
+
+        accum0 = vshrq_n_s32(accum0, SkConvolutionFilter1D::kShiftBits);
+        accum1 = vshrq_n_s32(accum1, SkConvolutionFilter1D::kShiftBits);
+        accum2 = vshrq_n_s32(accum2, SkConvolutionFilter1D::kShiftBits);
+
+        int16x8_t accum16_0 = vcombine_s16(vqmovn_s32(accum0), vqmovn_s32(accum1));
+        int16x8_t accum16_1 = vcombine_s16(vqmovn_s32(accum2), vqmovn_s32(accum2));
+
+        uint8x16_t accum8 = vcombine_u8(vqmovun_s16(accum16_0), vqmovun_s16(accum16_1));
+
+        if (hasAlpha) {
+            // Compute the max(ri, gi, bi) for each pixel.
+            // [8] xx a3 b3 g3 xx a2 b2 g2 xx a1 b1 g1 xx a0 b0 g0
+            uint8x16_t a = vreinterpretq_u8_u32(vshrq_n_u32(vreinterpretq_u32_u8(accum8), 8));
+            // [8] xx xx xx max3 xx xx xx max2 xx xx xx max1 xx xx xx max0
+            uint8x16_t b = vmaxq_u8(a, accum8); // Max of r and g
+            // [8] xx xx a3 b3 xx xx a2 b2 xx xx a1 b1 xx xx a0 b0
+            a = vreinterpretq_u8_u32(vshrq_n_u32(vreinterpretq_u32_u8(accum8), 16));
+            // [8] xx xx xx max3 xx xx xx max2 xx xx xx max1 xx xx xx max0
+            b = vmaxq_u8(a, b); // Max of r and g and b.
+            // [8] max3 00 00 00 max2 00 00 00 max1 00 00 00 max0 00 00 00
+            b = vreinterpretq_u8_u32(vshlq_n_u32(vreinterpretq_u32_u8(b), 24));
+
+            // Make sure the value of alpha channel is always larger than maximum
+            // value of color channels.
+            accum8 = vmaxq_u8(b, accum8);
+        } else {
+            // Set value of alpha channels to 0xFF.
+            accum8 = vreinterpretq_u8_u32(vreinterpretq_u32_u8(accum8) | vdupq_n_u32(0xFF000000));
+        }
+
+        switch(r) {
+        case 1:
+            vst1q_lane_u32(reinterpret_cast<uint32_t*>(outRow), vreinterpretq_u32_u8(accum8), 0);
+            break;
+        case 2:
+            vst1_u32(reinterpret_cast<uint32_t*>(outRow),
+                     vreinterpret_u32_u8(vget_low_u8(accum8)));
+            break;
+        case 3:
+            vst1_u32(reinterpret_cast<uint32_t*>(outRow),
+                     vreinterpret_u32_u8(vget_low_u8(accum8)));
+            vst1q_lane_u32(reinterpret_cast<uint32_t*>(outRow+8), vreinterpretq_u32_u8(accum8), 2);
+            break;
+        }
+    }
+}
+
+void convolveVertically_neon(const SkConvolutionFilter1D::ConvolutionFixed* filterValues,
+                             int filterLength,
+                             unsigned char* const* sourceDataRows,
+                             int pixelWidth,
+                             unsigned char* outRow,
+                             bool sourceHasAlpha) {
+    if (sourceHasAlpha) {
+        convolveVertically_neon<true>(filterValues, filterLength,
+                                      sourceDataRows, pixelWidth,
+                                      outRow);
+    } else {
+        convolveVertically_neon<false>(filterValues, filterLength,
+                                       sourceDataRows, pixelWidth,
+                                       outRow);
+    }
+}
+
+// Convolves horizontally along four rows. The row data is given in
+// |src_data| and continues for the num_values() of the filter.
+// The algorithm is almost same as |ConvolveHorizontally_SSE2|. Please
+// refer to that function for detailed comments.
+void convolve4RowsHorizontally_neon(const unsigned char* srcData[4],
+                                    const SkConvolutionFilter1D& filter,
+                                    unsigned char* outRow[4]) {
+
+    uint8x8_t coeff_mask0 = vcreate_u8(0x0100010001000100);
+    uint8x8_t coeff_mask1 = vcreate_u8(0x0302030203020302);
+    uint8x8_t coeff_mask2 = vcreate_u8(0x0504050405040504);
+    uint8x8_t coeff_mask3 = vcreate_u8(0x0706070607060706);
+    int num_values = filter.numValues();
+
+    int filterOffset, filterLength;
+    // |mask| will be used to decimate all extra filter coefficients that are
+    // loaded by SIMD when |filter_length| is not divisible by 4.
+    // mask[0] is not used in following algorithm.
+    const uint16_t mask[4][4] = {
+        {0, 0, 0, 0},
+        {0xFFFF, 0, 0, 0},
+        {0xFFFF, 0xFFFF, 0, 0},
+        {0xFFFF, 0xFFFF, 0xFFFF, 0}
+    };
+
+    // Output one pixel each iteration, calculating all channels (RGBA) together.
+    for (int outX = 0; outX < num_values; outX++) {
+
+        const SkConvolutionFilter1D::ConvolutionFixed* filterValues =
+        filter.FilterForValue(outX, &filterOffset, &filterLength);
+
+        // four pixels in a column per iteration.
+        int32x4_t accum0 = vdupq_n_s32(0);
+        int32x4_t accum1 = vdupq_n_s32(0);
+        int32x4_t accum2 = vdupq_n_s32(0);
+        int32x4_t accum3 = vdupq_n_s32(0);
+
+        int start = (filterOffset<<2);
+
+        // We will load and accumulate with four coefficients per iteration.
+        for (int filter_x = 0; filter_x < (filterLength >> 2); filter_x++) {
+            int16x4_t coeffs, coeff0, coeff1, coeff2, coeff3;
+
+            coeffs = vld1_s16(filterValues);
+            coeff0 = vreinterpret_s16_u8(vtbl1_u8(vreinterpret_u8_s16(coeffs), coeff_mask0));
+            coeff1 = vreinterpret_s16_u8(vtbl1_u8(vreinterpret_u8_s16(coeffs), coeff_mask1));
+            coeff2 = vreinterpret_s16_u8(vtbl1_u8(vreinterpret_u8_s16(coeffs), coeff_mask2));
+            coeff3 = vreinterpret_s16_u8(vtbl1_u8(vreinterpret_u8_s16(coeffs), coeff_mask3));
+
+            uint8x16_t pixels;
+            int16x8_t p01_16, p23_16;
+            int32x4_t p0, p1, p2, p3;
+
+
+#define ITERATION(src, accum)                                       \
+    pixels = vld1q_u8(src);                                         \
+    p01_16 = vreinterpretq_s16_u16(vmovl_u8(vget_low_u8(pixels)));  \
+    p23_16 = vreinterpretq_s16_u16(vmovl_u8(vget_high_u8(pixels))); \
+    p0 = vmull_s16(vget_low_s16(p01_16), coeff0);                   \
+    p1 = vmull_s16(vget_high_s16(p01_16), coeff1);                  \
+    p2 = vmull_s16(vget_low_s16(p23_16), coeff2);                   \
+    p3 = vmull_s16(vget_high_s16(p23_16), coeff3);                  \
+    accum += p0;                                                    \
+    accum += p1;                                                    \
+    accum += p2;                                                    \
+    accum += p3
+
+            ITERATION(srcData[0] + start, accum0);
+            ITERATION(srcData[1] + start, accum1);
+            ITERATION(srcData[2] + start, accum2);
+            ITERATION(srcData[3] + start, accum3);
+
+            start += 16;
+            filterValues += 4;
+        }
+
+        int r = filterLength & 3;
+        if (r) {
+            int16x4_t coeffs, coeff0, coeff1, coeff2, coeff3;
+            coeffs = vld1_s16(filterValues);
+            coeffs &= vreinterpret_s16_u16(vld1_u16(&mask[r][0]));
+            coeff0 = vreinterpret_s16_u8(vtbl1_u8(vreinterpret_u8_s16(coeffs), coeff_mask0));
+            coeff1 = vreinterpret_s16_u8(vtbl1_u8(vreinterpret_u8_s16(coeffs), coeff_mask1));
+            coeff2 = vreinterpret_s16_u8(vtbl1_u8(vreinterpret_u8_s16(coeffs), coeff_mask2));
+            coeff3 = vreinterpret_s16_u8(vtbl1_u8(vreinterpret_u8_s16(coeffs), coeff_mask3));
+
+            uint8x16_t pixels;
+            int16x8_t p01_16, p23_16;
+            int32x4_t p0, p1, p2, p3;
+
+            ITERATION(srcData[0] + start, accum0);
+            ITERATION(srcData[1] + start, accum1);
+            ITERATION(srcData[2] + start, accum2);
+            ITERATION(srcData[3] + start, accum3);
+        }
+
+        int16x4_t accum16;
+        uint8x8_t res0, res1, res2, res3;
+
+#define PACK_RESULT(accum, res)                                         \
+        accum = vshrq_n_s32(accum, SkConvolutionFilter1D::kShiftBits);  \
+        accum16 = vqmovn_s32(accum);                                    \
+        res = vqmovun_s16(vcombine_s16(accum16, accum16));
+
+        PACK_RESULT(accum0, res0);
+        PACK_RESULT(accum1, res1);
+        PACK_RESULT(accum2, res2);
+        PACK_RESULT(accum3, res3);
+
+        vst1_lane_u32(reinterpret_cast<uint32_t*>(outRow[0]), vreinterpret_u32_u8(res0), 0);
+        vst1_lane_u32(reinterpret_cast<uint32_t*>(outRow[1]), vreinterpret_u32_u8(res1), 0);
+        vst1_lane_u32(reinterpret_cast<uint32_t*>(outRow[2]), vreinterpret_u32_u8(res2), 0);
+        vst1_lane_u32(reinterpret_cast<uint32_t*>(outRow[3]), vreinterpret_u32_u8(res3), 0);
+        outRow[0] += 4;
+        outRow[1] += 4;
+        outRow[2] += 4;
+        outRow[3] += 4;
+    }
+}
+
+void applySIMDPadding_neon(SkConvolutionFilter1D *filter) {
+    // Padding |paddingCount| of more dummy coefficients after the coefficients
+    // of last filter to prevent SIMD instructions which load 8 or 16 bytes
+    // together to access invalid memory areas. We are not trying to align the
+    // coefficients right now due to the opaqueness of <vector> implementation.
+    // This has to be done after all |AddFilter| calls.
+    for (int i = 0; i < 8; ++i) {
+        filter->addFilterValue(static_cast<SkConvolutionFilter1D::ConvolutionFixed>(0));
+    }
+}
+
+void platformConvolutionProcs_arm_neon(SkConvolutionProcs* procs) {
+    procs->fExtraHorizontalReads = 3;
+    procs->fConvolveVertically = &convolveVertically_neon;
+    procs->fConvolve4RowsHorizontally = &convolve4RowsHorizontally_neon;
+    procs->fConvolveHorizontally = &convolveHorizontally_neon;
+    procs->fApplySIMDPadding = &applySIMDPadding_neon;
+}
+