From 869be76fae4c595451dc6a544856ce4c6a89feeb Mon Sep 17 00:00:00 2001 From: Nigel Tao Date: Thu, 18 Feb 2021 14:32:35 +1100 Subject: [PATCH] Rename some std/png c/p variables --- release/c/wuffs-unsupported-snapshot.c | 982 ++++++++++++------------- std/png/decode_filter_arm_neon.wuffs | 60 +- std/png/decode_filter_fallback.wuffs | 168 +++-- std/png/decode_filter_x86_sse42.wuffs | 73 +- 4 files changed, 649 insertions(+), 634 deletions(-) diff --git a/release/c/wuffs-unsupported-snapshot.c b/release/c/wuffs-unsupported-snapshot.c index 3bf33c4b..98b42636 100644 --- a/release/c/wuffs-unsupported-snapshot.c +++ b/release/c/wuffs-unsupported-snapshot.c @@ -31593,37 +31593,37 @@ static wuffs_base__empty_struct wuffs_png__decoder__filter_1_distance_4_arm_neon( wuffs_png__decoder* self, wuffs_base__slice_u8 a_curr) { - wuffs_base__slice_u8 v_c = {0}; + wuffs_base__slice_u8 v_curr = {0}; uint8x8_t v_fa = {0}; uint8x8_t v_fx = {0}; { - wuffs_base__slice_u8 i_slice_c = a_curr; - v_c.ptr = i_slice_c.ptr; - v_c.len = 4; - uint8_t* i_end0_c = v_c.ptr + (((i_slice_c.len - (size_t)(v_c.ptr - i_slice_c.ptr)) / 8) * 8); - while (v_c.ptr < i_end0_c) { - v_fx = vreinterpret_u8_u32(vdup_n_u32(wuffs_base__peek_u32le__no_bounds_check(v_c.ptr))); + wuffs_base__slice_u8 i_slice_curr = a_curr; + v_curr.ptr = i_slice_curr.ptr; + v_curr.len = 4; + uint8_t* i_end0_curr = v_curr.ptr + (((i_slice_curr.len - (size_t)(v_curr.ptr - i_slice_curr.ptr)) / 8) * 8); + while (v_curr.ptr < i_end0_curr) { + v_fx = vreinterpret_u8_u32(vdup_n_u32(wuffs_base__peek_u32le__no_bounds_check(v_curr.ptr))); v_fx = vadd_u8(v_fx, v_fa); - wuffs_base__poke_u32le__no_bounds_check(v_c.ptr, vget_lane_u32(vreinterpret_u32_u8(v_fx), 0)); + wuffs_base__poke_u32le__no_bounds_check(v_curr.ptr, vget_lane_u32(vreinterpret_u32_u8(v_fx), 0)); v_fa = v_fx; - v_c.ptr += 4; - v_fx = vreinterpret_u8_u32(vdup_n_u32(wuffs_base__peek_u32le__no_bounds_check(v_c.ptr))); + v_curr.ptr += 4; + v_fx = vreinterpret_u8_u32(vdup_n_u32(wuffs_base__peek_u32le__no_bounds_check(v_curr.ptr))); v_fx = vadd_u8(v_fx, v_fa); - wuffs_base__poke_u32le__no_bounds_check(v_c.ptr, vget_lane_u32(vreinterpret_u32_u8(v_fx), 0)); + wuffs_base__poke_u32le__no_bounds_check(v_curr.ptr, vget_lane_u32(vreinterpret_u32_u8(v_fx), 0)); v_fa = v_fx; - v_c.ptr += 4; + v_curr.ptr += 4; } - v_c.len = 4; - uint8_t* i_end1_c = v_c.ptr + (((i_slice_c.len - (size_t)(v_c.ptr - i_slice_c.ptr)) / 4) * 4); - while (v_c.ptr < i_end1_c) { - v_fx = vreinterpret_u8_u32(vdup_n_u32(wuffs_base__peek_u32le__no_bounds_check(v_c.ptr))); + v_curr.len = 4; + uint8_t* i_end1_curr = v_curr.ptr + (((i_slice_curr.len - (size_t)(v_curr.ptr - i_slice_curr.ptr)) / 4) * 4); + while (v_curr.ptr < i_end1_curr) { + v_fx = vreinterpret_u8_u32(vdup_n_u32(wuffs_base__peek_u32le__no_bounds_check(v_curr.ptr))); v_fx = vadd_u8(v_fx, v_fa); - wuffs_base__poke_u32le__no_bounds_check(v_c.ptr, vget_lane_u32(vreinterpret_u32_u8(v_fx), 0)); + wuffs_base__poke_u32le__no_bounds_check(v_curr.ptr, vget_lane_u32(vreinterpret_u32_u8(v_fx), 0)); v_fa = v_fx; - v_c.ptr += 4; + v_curr.ptr += 4; } - v_c.len = 0; + v_curr.len = 0; } return wuffs_base__make_empty_struct(); } @@ -31637,81 +31637,81 @@ wuffs_png__decoder__filter_3_distance_4_arm_neon( wuffs_png__decoder* self, wuffs_base__slice_u8 a_curr, wuffs_base__slice_u8 a_prev) { - wuffs_base__slice_u8 v_c = {0}; - wuffs_base__slice_u8 v_p = {0}; + wuffs_base__slice_u8 v_curr = {0}; + wuffs_base__slice_u8 v_prev = {0}; uint8x8_t v_fa = {0}; uint8x8_t v_fb = {0}; uint8x8_t v_fx = {0}; if (((uint64_t)(a_prev.len)) == 0) { { - wuffs_base__slice_u8 i_slice_c = a_curr; - v_c.ptr = i_slice_c.ptr; - v_c.len = 4; - uint8_t* i_end0_c = v_c.ptr + (((i_slice_c.len - (size_t)(v_c.ptr - i_slice_c.ptr)) / 8) * 8); - while (v_c.ptr < i_end0_c) { - v_fx = vreinterpret_u8_u32(vdup_n_u32(wuffs_base__peek_u32le__no_bounds_check(v_c.ptr))); + wuffs_base__slice_u8 i_slice_curr = a_curr; + v_curr.ptr = i_slice_curr.ptr; + v_curr.len = 4; + uint8_t* i_end0_curr = v_curr.ptr + (((i_slice_curr.len - (size_t)(v_curr.ptr - i_slice_curr.ptr)) / 8) * 8); + while (v_curr.ptr < i_end0_curr) { + v_fx = vreinterpret_u8_u32(vdup_n_u32(wuffs_base__peek_u32le__no_bounds_check(v_curr.ptr))); v_fx = vadd_u8(v_fx, vhadd_u8(v_fa, v_fb)); - wuffs_base__poke_u32le__no_bounds_check(v_c.ptr, vget_lane_u32(vreinterpret_u32_u8(v_fx), 0)); + wuffs_base__poke_u32le__no_bounds_check(v_curr.ptr, vget_lane_u32(vreinterpret_u32_u8(v_fx), 0)); v_fa = v_fx; - v_c.ptr += 4; - v_fx = vreinterpret_u8_u32(vdup_n_u32(wuffs_base__peek_u32le__no_bounds_check(v_c.ptr))); + v_curr.ptr += 4; + v_fx = vreinterpret_u8_u32(vdup_n_u32(wuffs_base__peek_u32le__no_bounds_check(v_curr.ptr))); v_fx = vadd_u8(v_fx, vhadd_u8(v_fa, v_fb)); - wuffs_base__poke_u32le__no_bounds_check(v_c.ptr, vget_lane_u32(vreinterpret_u32_u8(v_fx), 0)); + wuffs_base__poke_u32le__no_bounds_check(v_curr.ptr, vget_lane_u32(vreinterpret_u32_u8(v_fx), 0)); v_fa = v_fx; - v_c.ptr += 4; + v_curr.ptr += 4; } - v_c.len = 4; - uint8_t* i_end1_c = v_c.ptr + (((i_slice_c.len - (size_t)(v_c.ptr - i_slice_c.ptr)) / 4) * 4); - while (v_c.ptr < i_end1_c) { - v_fx = vreinterpret_u8_u32(vdup_n_u32(wuffs_base__peek_u32le__no_bounds_check(v_c.ptr))); + v_curr.len = 4; + uint8_t* i_end1_curr = v_curr.ptr + (((i_slice_curr.len - (size_t)(v_curr.ptr - i_slice_curr.ptr)) / 4) * 4); + while (v_curr.ptr < i_end1_curr) { + v_fx = vreinterpret_u8_u32(vdup_n_u32(wuffs_base__peek_u32le__no_bounds_check(v_curr.ptr))); v_fx = vadd_u8(v_fx, vhadd_u8(v_fa, v_fb)); - wuffs_base__poke_u32le__no_bounds_check(v_c.ptr, vget_lane_u32(vreinterpret_u32_u8(v_fx), 0)); + wuffs_base__poke_u32le__no_bounds_check(v_curr.ptr, vget_lane_u32(vreinterpret_u32_u8(v_fx), 0)); v_fa = v_fx; - v_c.ptr += 4; + v_curr.ptr += 4; } - v_c.len = 0; + v_curr.len = 0; } } else { { - wuffs_base__slice_u8 i_slice_c = a_curr; - v_c.ptr = i_slice_c.ptr; - wuffs_base__slice_u8 i_slice_p = a_prev; - v_p.ptr = i_slice_p.ptr; - i_slice_c.len = ((size_t)(wuffs_base__u64__min(i_slice_c.len, i_slice_p.len))); - v_c.len = 4; - v_p.len = 4; - uint8_t* i_end0_c = v_c.ptr + (((i_slice_c.len - (size_t)(v_c.ptr - i_slice_c.ptr)) / 8) * 8); - while (v_c.ptr < i_end0_c) { - v_fb = vreinterpret_u8_u32(vdup_n_u32(wuffs_base__peek_u32le__no_bounds_check(v_p.ptr))); - v_fx = vreinterpret_u8_u32(vdup_n_u32(wuffs_base__peek_u32le__no_bounds_check(v_c.ptr))); + wuffs_base__slice_u8 i_slice_curr = a_curr; + v_curr.ptr = i_slice_curr.ptr; + wuffs_base__slice_u8 i_slice_prev = a_prev; + v_prev.ptr = i_slice_prev.ptr; + i_slice_curr.len = ((size_t)(wuffs_base__u64__min(i_slice_curr.len, i_slice_prev.len))); + v_curr.len = 4; + v_prev.len = 4; + uint8_t* i_end0_curr = v_curr.ptr + (((i_slice_curr.len - (size_t)(v_curr.ptr - i_slice_curr.ptr)) / 8) * 8); + while (v_curr.ptr < i_end0_curr) { + v_fb = vreinterpret_u8_u32(vdup_n_u32(wuffs_base__peek_u32le__no_bounds_check(v_prev.ptr))); + v_fx = vreinterpret_u8_u32(vdup_n_u32(wuffs_base__peek_u32le__no_bounds_check(v_curr.ptr))); v_fx = vadd_u8(v_fx, vhadd_u8(v_fa, v_fb)); - wuffs_base__poke_u32le__no_bounds_check(v_c.ptr, vget_lane_u32(vreinterpret_u32_u8(v_fx), 0)); + wuffs_base__poke_u32le__no_bounds_check(v_curr.ptr, vget_lane_u32(vreinterpret_u32_u8(v_fx), 0)); v_fa = v_fx; - v_c.ptr += 4; - v_p.ptr += 4; - v_fb = vreinterpret_u8_u32(vdup_n_u32(wuffs_base__peek_u32le__no_bounds_check(v_p.ptr))); - v_fx = vreinterpret_u8_u32(vdup_n_u32(wuffs_base__peek_u32le__no_bounds_check(v_c.ptr))); + v_curr.ptr += 4; + v_prev.ptr += 4; + v_fb = vreinterpret_u8_u32(vdup_n_u32(wuffs_base__peek_u32le__no_bounds_check(v_prev.ptr))); + v_fx = vreinterpret_u8_u32(vdup_n_u32(wuffs_base__peek_u32le__no_bounds_check(v_curr.ptr))); v_fx = vadd_u8(v_fx, vhadd_u8(v_fa, v_fb)); - wuffs_base__poke_u32le__no_bounds_check(v_c.ptr, vget_lane_u32(vreinterpret_u32_u8(v_fx), 0)); + wuffs_base__poke_u32le__no_bounds_check(v_curr.ptr, vget_lane_u32(vreinterpret_u32_u8(v_fx), 0)); v_fa = v_fx; - v_c.ptr += 4; - v_p.ptr += 4; + v_curr.ptr += 4; + v_prev.ptr += 4; } - v_c.len = 4; - v_p.len = 4; - uint8_t* i_end1_c = v_c.ptr + (((i_slice_c.len - (size_t)(v_c.ptr - i_slice_c.ptr)) / 4) * 4); - while (v_c.ptr < i_end1_c) { - v_fb = vreinterpret_u8_u32(vdup_n_u32(wuffs_base__peek_u32le__no_bounds_check(v_p.ptr))); - v_fx = vreinterpret_u8_u32(vdup_n_u32(wuffs_base__peek_u32le__no_bounds_check(v_c.ptr))); + v_curr.len = 4; + v_prev.len = 4; + uint8_t* i_end1_curr = v_curr.ptr + (((i_slice_curr.len - (size_t)(v_curr.ptr - i_slice_curr.ptr)) / 4) * 4); + while (v_curr.ptr < i_end1_curr) { + v_fb = vreinterpret_u8_u32(vdup_n_u32(wuffs_base__peek_u32le__no_bounds_check(v_prev.ptr))); + v_fx = vreinterpret_u8_u32(vdup_n_u32(wuffs_base__peek_u32le__no_bounds_check(v_curr.ptr))); v_fx = vadd_u8(v_fx, vhadd_u8(v_fa, v_fb)); - wuffs_base__poke_u32le__no_bounds_check(v_c.ptr, vget_lane_u32(vreinterpret_u32_u8(v_fx), 0)); + wuffs_base__poke_u32le__no_bounds_check(v_curr.ptr, vget_lane_u32(vreinterpret_u32_u8(v_fx), 0)); v_fa = v_fx; - v_c.ptr += 4; - v_p.ptr += 4; + v_curr.ptr += 4; + v_prev.ptr += 4; } - v_c.len = 0; - v_p.len = 0; + v_curr.len = 0; + v_prev.len = 0; } } return wuffs_base__make_empty_struct(); @@ -31726,8 +31726,8 @@ wuffs_png__decoder__filter_4_distance_3_arm_neon( wuffs_png__decoder* self, wuffs_base__slice_u8 a_curr, wuffs_base__slice_u8 a_prev) { - wuffs_base__slice_u8 v_c = {0}; - wuffs_base__slice_u8 v_p = {0}; + wuffs_base__slice_u8 v_curr = {0}; + wuffs_base__slice_u8 v_prev = {0}; uint8x8_t v_fa = {0}; uint8x8_t v_fb = {0}; uint8x8_t v_fc = {0}; @@ -31743,17 +31743,17 @@ wuffs_png__decoder__filter_4_distance_3_arm_neon( uint8x8_t v_pickb = {0}; { - wuffs_base__slice_u8 i_slice_c = a_curr; - v_c.ptr = i_slice_c.ptr; - wuffs_base__slice_u8 i_slice_p = a_prev; - v_p.ptr = i_slice_p.ptr; - i_slice_c.len = ((size_t)(wuffs_base__u64__min(i_slice_c.len, i_slice_p.len))); - v_c.len = 4; - v_p.len = 4; - uint8_t* i_end0_c = v_c.ptr + wuffs_base__iterate_total_advance((i_slice_c.len - (size_t)(v_c.ptr - i_slice_c.ptr)), 7, 6); - while (v_c.ptr < i_end0_c) { - v_fb = vreinterpret_u8_u32(vdup_n_u32(wuffs_base__peek_u32le__no_bounds_check(v_p.ptr))); - v_fx = vreinterpret_u8_u32(vdup_n_u32(wuffs_base__peek_u32le__no_bounds_check(v_c.ptr))); + wuffs_base__slice_u8 i_slice_curr = a_curr; + v_curr.ptr = i_slice_curr.ptr; + wuffs_base__slice_u8 i_slice_prev = a_prev; + v_prev.ptr = i_slice_prev.ptr; + i_slice_curr.len = ((size_t)(wuffs_base__u64__min(i_slice_curr.len, i_slice_prev.len))); + v_curr.len = 4; + v_prev.len = 4; + uint8_t* i_end0_curr = v_curr.ptr + wuffs_base__iterate_total_advance((i_slice_curr.len - (size_t)(v_curr.ptr - i_slice_curr.ptr)), 7, 6); + while (v_curr.ptr < i_end0_curr) { + v_fb = vreinterpret_u8_u32(vdup_n_u32(wuffs_base__peek_u32le__no_bounds_check(v_prev.ptr))); + v_fx = vreinterpret_u8_u32(vdup_n_u32(wuffs_base__peek_u32le__no_bounds_check(v_curr.ptr))); v_fafb = vreinterpretq_u8_u16(vaddl_u8(v_fa, v_fb)); v_fcfc = vreinterpretq_u8_u16(vaddl_u8(v_fc, v_fc)); v_pa = vreinterpretq_u8_u16(vabdl_u8(v_fb, v_fc)); @@ -31764,13 +31764,13 @@ wuffs_png__decoder__filter_4_distance_3_arm_neon( v_picka = vmovn_u16(vreinterpretq_u16_u8(vandq_u8(v_cmpab, v_cmpac))); v_pickb = vmovn_u16(vreinterpretq_u16_u8(vreinterpretq_u8_u16(vcleq_u16(vreinterpretq_u16_u8(v_pb), vreinterpretq_u16_u8(v_pc))))); v_fx = vadd_u8(v_fx, vbsl_u8(v_picka, v_fa, vbsl_u8(v_pickb, v_fb, v_fc))); - wuffs_base__poke_u24le__no_bounds_check(v_c.ptr, vget_lane_u32(vreinterpret_u32_u8(v_fx), 0)); + wuffs_base__poke_u24le__no_bounds_check(v_curr.ptr, vget_lane_u32(vreinterpret_u32_u8(v_fx), 0)); v_fc = v_fb; v_fa = v_fx; - v_c.ptr += 3; - v_p.ptr += 3; - v_fb = vreinterpret_u8_u32(vdup_n_u32(wuffs_base__peek_u32le__no_bounds_check(v_p.ptr))); - v_fx = vreinterpret_u8_u32(vdup_n_u32(wuffs_base__peek_u32le__no_bounds_check(v_c.ptr))); + v_curr.ptr += 3; + v_prev.ptr += 3; + v_fb = vreinterpret_u8_u32(vdup_n_u32(wuffs_base__peek_u32le__no_bounds_check(v_prev.ptr))); + v_fx = vreinterpret_u8_u32(vdup_n_u32(wuffs_base__peek_u32le__no_bounds_check(v_curr.ptr))); v_fafb = vreinterpretq_u8_u16(vaddl_u8(v_fa, v_fb)); v_fcfc = vreinterpretq_u8_u16(vaddl_u8(v_fc, v_fc)); v_pa = vreinterpretq_u8_u16(vabdl_u8(v_fb, v_fc)); @@ -31781,18 +31781,18 @@ wuffs_png__decoder__filter_4_distance_3_arm_neon( v_picka = vmovn_u16(vreinterpretq_u16_u8(vandq_u8(v_cmpab, v_cmpac))); v_pickb = vmovn_u16(vreinterpretq_u16_u8(vreinterpretq_u8_u16(vcleq_u16(vreinterpretq_u16_u8(v_pb), vreinterpretq_u16_u8(v_pc))))); v_fx = vadd_u8(v_fx, vbsl_u8(v_picka, v_fa, vbsl_u8(v_pickb, v_fb, v_fc))); - wuffs_base__poke_u24le__no_bounds_check(v_c.ptr, vget_lane_u32(vreinterpret_u32_u8(v_fx), 0)); + wuffs_base__poke_u24le__no_bounds_check(v_curr.ptr, vget_lane_u32(vreinterpret_u32_u8(v_fx), 0)); v_fc = v_fb; v_fa = v_fx; - v_c.ptr += 3; - v_p.ptr += 3; + v_curr.ptr += 3; + v_prev.ptr += 3; } - v_c.len = 4; - v_p.len = 4; - uint8_t* i_end1_c = v_c.ptr + wuffs_base__iterate_total_advance((i_slice_c.len - (size_t)(v_c.ptr - i_slice_c.ptr)), 4, 3); - while (v_c.ptr < i_end1_c) { - v_fb = vreinterpret_u8_u32(vdup_n_u32(wuffs_base__peek_u32le__no_bounds_check(v_p.ptr))); - v_fx = vreinterpret_u8_u32(vdup_n_u32(wuffs_base__peek_u32le__no_bounds_check(v_c.ptr))); + v_curr.len = 4; + v_prev.len = 4; + uint8_t* i_end1_curr = v_curr.ptr + wuffs_base__iterate_total_advance((i_slice_curr.len - (size_t)(v_curr.ptr - i_slice_curr.ptr)), 4, 3); + while (v_curr.ptr < i_end1_curr) { + v_fb = vreinterpret_u8_u32(vdup_n_u32(wuffs_base__peek_u32le__no_bounds_check(v_prev.ptr))); + v_fx = vreinterpret_u8_u32(vdup_n_u32(wuffs_base__peek_u32le__no_bounds_check(v_curr.ptr))); v_fafb = vreinterpretq_u8_u16(vaddl_u8(v_fa, v_fb)); v_fcfc = vreinterpretq_u8_u16(vaddl_u8(v_fc, v_fc)); v_pa = vreinterpretq_u8_u16(vabdl_u8(v_fb, v_fc)); @@ -31803,18 +31803,18 @@ wuffs_png__decoder__filter_4_distance_3_arm_neon( v_picka = vmovn_u16(vreinterpretq_u16_u8(vandq_u8(v_cmpab, v_cmpac))); v_pickb = vmovn_u16(vreinterpretq_u16_u8(vreinterpretq_u8_u16(vcleq_u16(vreinterpretq_u16_u8(v_pb), vreinterpretq_u16_u8(v_pc))))); v_fx = vadd_u8(v_fx, vbsl_u8(v_picka, v_fa, vbsl_u8(v_pickb, v_fb, v_fc))); - wuffs_base__poke_u24le__no_bounds_check(v_c.ptr, vget_lane_u32(vreinterpret_u32_u8(v_fx), 0)); + wuffs_base__poke_u24le__no_bounds_check(v_curr.ptr, vget_lane_u32(vreinterpret_u32_u8(v_fx), 0)); v_fc = v_fb; v_fa = v_fx; - v_c.ptr += 3; - v_p.ptr += 3; + v_curr.ptr += 3; + v_prev.ptr += 3; } - v_c.len = 3; - v_p.len = 3; - uint8_t* i_end2_c = v_c.ptr + (((i_slice_c.len - (size_t)(v_c.ptr - i_slice_c.ptr)) / 3) * 3); - while (v_c.ptr < i_end2_c) { - v_fb = vreinterpret_u8_u32(vdup_n_u32(wuffs_base__peek_u24le__no_bounds_check(v_p.ptr))); - v_fx = vreinterpret_u8_u32(vdup_n_u32(wuffs_base__peek_u24le__no_bounds_check(v_c.ptr))); + v_curr.len = 3; + v_prev.len = 3; + uint8_t* i_end2_curr = v_curr.ptr + (((i_slice_curr.len - (size_t)(v_curr.ptr - i_slice_curr.ptr)) / 3) * 3); + while (v_curr.ptr < i_end2_curr) { + v_fb = vreinterpret_u8_u32(vdup_n_u32(wuffs_base__peek_u24le__no_bounds_check(v_prev.ptr))); + v_fx = vreinterpret_u8_u32(vdup_n_u32(wuffs_base__peek_u24le__no_bounds_check(v_curr.ptr))); v_fafb = vreinterpretq_u8_u16(vaddl_u8(v_fa, v_fb)); v_fcfc = vreinterpretq_u8_u16(vaddl_u8(v_fc, v_fc)); v_pa = vreinterpretq_u8_u16(vabdl_u8(v_fb, v_fc)); @@ -31825,12 +31825,12 @@ wuffs_png__decoder__filter_4_distance_3_arm_neon( v_picka = vmovn_u16(vreinterpretq_u16_u8(vandq_u8(v_cmpab, v_cmpac))); v_pickb = vmovn_u16(vreinterpretq_u16_u8(vreinterpretq_u8_u16(vcleq_u16(vreinterpretq_u16_u8(v_pb), vreinterpretq_u16_u8(v_pc))))); v_fx = vadd_u8(v_fx, vbsl_u8(v_picka, v_fa, vbsl_u8(v_pickb, v_fb, v_fc))); - wuffs_base__poke_u24le__no_bounds_check(v_c.ptr, vget_lane_u32(vreinterpret_u32_u8(v_fx), 0)); - v_c.ptr += 3; - v_p.ptr += 3; + wuffs_base__poke_u24le__no_bounds_check(v_curr.ptr, vget_lane_u32(vreinterpret_u32_u8(v_fx), 0)); + v_curr.ptr += 3; + v_prev.ptr += 3; } - v_c.len = 0; - v_p.len = 0; + v_curr.len = 0; + v_prev.len = 0; } return wuffs_base__make_empty_struct(); } @@ -31844,8 +31844,8 @@ wuffs_png__decoder__filter_4_distance_4_arm_neon( wuffs_png__decoder* self, wuffs_base__slice_u8 a_curr, wuffs_base__slice_u8 a_prev) { - wuffs_base__slice_u8 v_c = {0}; - wuffs_base__slice_u8 v_p = {0}; + wuffs_base__slice_u8 v_curr = {0}; + wuffs_base__slice_u8 v_prev = {0}; uint8x8_t v_fa = {0}; uint8x8_t v_fb = {0}; uint8x8_t v_fc = {0}; @@ -31861,17 +31861,17 @@ wuffs_png__decoder__filter_4_distance_4_arm_neon( uint8x8_t v_pickb = {0}; { - wuffs_base__slice_u8 i_slice_c = a_curr; - v_c.ptr = i_slice_c.ptr; - wuffs_base__slice_u8 i_slice_p = a_prev; - v_p.ptr = i_slice_p.ptr; - i_slice_c.len = ((size_t)(wuffs_base__u64__min(i_slice_c.len, i_slice_p.len))); - v_c.len = 4; - v_p.len = 4; - uint8_t* i_end0_c = v_c.ptr + (((i_slice_c.len - (size_t)(v_c.ptr - i_slice_c.ptr)) / 8) * 8); - while (v_c.ptr < i_end0_c) { - v_fb = vreinterpret_u8_u32(vdup_n_u32(wuffs_base__peek_u32le__no_bounds_check(v_p.ptr))); - v_fx = vreinterpret_u8_u32(vdup_n_u32(wuffs_base__peek_u32le__no_bounds_check(v_c.ptr))); + wuffs_base__slice_u8 i_slice_curr = a_curr; + v_curr.ptr = i_slice_curr.ptr; + wuffs_base__slice_u8 i_slice_prev = a_prev; + v_prev.ptr = i_slice_prev.ptr; + i_slice_curr.len = ((size_t)(wuffs_base__u64__min(i_slice_curr.len, i_slice_prev.len))); + v_curr.len = 4; + v_prev.len = 4; + uint8_t* i_end0_curr = v_curr.ptr + (((i_slice_curr.len - (size_t)(v_curr.ptr - i_slice_curr.ptr)) / 8) * 8); + while (v_curr.ptr < i_end0_curr) { + v_fb = vreinterpret_u8_u32(vdup_n_u32(wuffs_base__peek_u32le__no_bounds_check(v_prev.ptr))); + v_fx = vreinterpret_u8_u32(vdup_n_u32(wuffs_base__peek_u32le__no_bounds_check(v_curr.ptr))); v_fafb = vreinterpretq_u8_u16(vaddl_u8(v_fa, v_fb)); v_fcfc = vreinterpretq_u8_u16(vaddl_u8(v_fc, v_fc)); v_pa = vreinterpretq_u8_u16(vabdl_u8(v_fb, v_fc)); @@ -31882,13 +31882,13 @@ wuffs_png__decoder__filter_4_distance_4_arm_neon( v_picka = vmovn_u16(vreinterpretq_u16_u8(vandq_u8(v_cmpab, v_cmpac))); v_pickb = vmovn_u16(vreinterpretq_u16_u8(vreinterpretq_u8_u16(vcleq_u16(vreinterpretq_u16_u8(v_pb), vreinterpretq_u16_u8(v_pc))))); v_fx = vadd_u8(v_fx, vbsl_u8(v_picka, v_fa, vbsl_u8(v_pickb, v_fb, v_fc))); - wuffs_base__poke_u32le__no_bounds_check(v_c.ptr, vget_lane_u32(vreinterpret_u32_u8(v_fx), 0)); + wuffs_base__poke_u32le__no_bounds_check(v_curr.ptr, vget_lane_u32(vreinterpret_u32_u8(v_fx), 0)); v_fc = v_fb; v_fa = v_fx; - v_c.ptr += 4; - v_p.ptr += 4; - v_fb = vreinterpret_u8_u32(vdup_n_u32(wuffs_base__peek_u32le__no_bounds_check(v_p.ptr))); - v_fx = vreinterpret_u8_u32(vdup_n_u32(wuffs_base__peek_u32le__no_bounds_check(v_c.ptr))); + v_curr.ptr += 4; + v_prev.ptr += 4; + v_fb = vreinterpret_u8_u32(vdup_n_u32(wuffs_base__peek_u32le__no_bounds_check(v_prev.ptr))); + v_fx = vreinterpret_u8_u32(vdup_n_u32(wuffs_base__peek_u32le__no_bounds_check(v_curr.ptr))); v_fafb = vreinterpretq_u8_u16(vaddl_u8(v_fa, v_fb)); v_fcfc = vreinterpretq_u8_u16(vaddl_u8(v_fc, v_fc)); v_pa = vreinterpretq_u8_u16(vabdl_u8(v_fb, v_fc)); @@ -31899,18 +31899,18 @@ wuffs_png__decoder__filter_4_distance_4_arm_neon( v_picka = vmovn_u16(vreinterpretq_u16_u8(vandq_u8(v_cmpab, v_cmpac))); v_pickb = vmovn_u16(vreinterpretq_u16_u8(vreinterpretq_u8_u16(vcleq_u16(vreinterpretq_u16_u8(v_pb), vreinterpretq_u16_u8(v_pc))))); v_fx = vadd_u8(v_fx, vbsl_u8(v_picka, v_fa, vbsl_u8(v_pickb, v_fb, v_fc))); - wuffs_base__poke_u32le__no_bounds_check(v_c.ptr, vget_lane_u32(vreinterpret_u32_u8(v_fx), 0)); + wuffs_base__poke_u32le__no_bounds_check(v_curr.ptr, vget_lane_u32(vreinterpret_u32_u8(v_fx), 0)); v_fc = v_fb; v_fa = v_fx; - v_c.ptr += 4; - v_p.ptr += 4; + v_curr.ptr += 4; + v_prev.ptr += 4; } - v_c.len = 4; - v_p.len = 4; - uint8_t* i_end1_c = v_c.ptr + (((i_slice_c.len - (size_t)(v_c.ptr - i_slice_c.ptr)) / 4) * 4); - while (v_c.ptr < i_end1_c) { - v_fb = vreinterpret_u8_u32(vdup_n_u32(wuffs_base__peek_u32le__no_bounds_check(v_p.ptr))); - v_fx = vreinterpret_u8_u32(vdup_n_u32(wuffs_base__peek_u32le__no_bounds_check(v_c.ptr))); + v_curr.len = 4; + v_prev.len = 4; + uint8_t* i_end1_curr = v_curr.ptr + (((i_slice_curr.len - (size_t)(v_curr.ptr - i_slice_curr.ptr)) / 4) * 4); + while (v_curr.ptr < i_end1_curr) { + v_fb = vreinterpret_u8_u32(vdup_n_u32(wuffs_base__peek_u32le__no_bounds_check(v_prev.ptr))); + v_fx = vreinterpret_u8_u32(vdup_n_u32(wuffs_base__peek_u32le__no_bounds_check(v_curr.ptr))); v_fafb = vreinterpretq_u8_u16(vaddl_u8(v_fa, v_fb)); v_fcfc = vreinterpretq_u8_u16(vaddl_u8(v_fc, v_fc)); v_pa = vreinterpretq_u8_u16(vabdl_u8(v_fb, v_fc)); @@ -31921,14 +31921,14 @@ wuffs_png__decoder__filter_4_distance_4_arm_neon( v_picka = vmovn_u16(vreinterpretq_u16_u8(vandq_u8(v_cmpab, v_cmpac))); v_pickb = vmovn_u16(vreinterpretq_u16_u8(vreinterpretq_u8_u16(vcleq_u16(vreinterpretq_u16_u8(v_pb), vreinterpretq_u16_u8(v_pc))))); v_fx = vadd_u8(v_fx, vbsl_u8(v_picka, v_fa, vbsl_u8(v_pickb, v_fb, v_fc))); - wuffs_base__poke_u32le__no_bounds_check(v_c.ptr, vget_lane_u32(vreinterpret_u32_u8(v_fx), 0)); + wuffs_base__poke_u32le__no_bounds_check(v_curr.ptr, vget_lane_u32(vreinterpret_u32_u8(v_fx), 0)); v_fc = v_fb; v_fa = v_fx; - v_c.ptr += 4; - v_p.ptr += 4; + v_curr.ptr += 4; + v_prev.ptr += 4; } - v_c.len = 0; - v_p.len = 0; + v_curr.len = 0; + v_prev.len = 0; } return wuffs_base__make_empty_struct(); } @@ -31973,44 +31973,44 @@ static wuffs_base__empty_struct wuffs_png__decoder__filter_1_distance_3_fallback( wuffs_png__decoder* self, wuffs_base__slice_u8 a_curr) { - wuffs_base__slice_u8 v_c = {0}; + wuffs_base__slice_u8 v_curr = {0}; uint8_t v_fa0 = 0; uint8_t v_fa1 = 0; uint8_t v_fa2 = 0; { - wuffs_base__slice_u8 i_slice_c = a_curr; - v_c.ptr = i_slice_c.ptr; - v_c.len = 3; - uint8_t* i_end0_c = v_c.ptr + (((i_slice_c.len - (size_t)(v_c.ptr - i_slice_c.ptr)) / 6) * 6); - while (v_c.ptr < i_end0_c) { - v_fa0 = ((uint8_t)(v_fa0 + v_c.ptr[0])); - v_c.ptr[0] = v_fa0; - v_fa1 = ((uint8_t)(v_fa1 + v_c.ptr[1])); - v_c.ptr[1] = v_fa1; - v_fa2 = ((uint8_t)(v_fa2 + v_c.ptr[2])); - v_c.ptr[2] = v_fa2; - v_c.ptr += 3; - v_fa0 = ((uint8_t)(v_fa0 + v_c.ptr[0])); - v_c.ptr[0] = v_fa0; - v_fa1 = ((uint8_t)(v_fa1 + v_c.ptr[1])); - v_c.ptr[1] = v_fa1; - v_fa2 = ((uint8_t)(v_fa2 + v_c.ptr[2])); - v_c.ptr[2] = v_fa2; - v_c.ptr += 3; + wuffs_base__slice_u8 i_slice_curr = a_curr; + v_curr.ptr = i_slice_curr.ptr; + v_curr.len = 3; + uint8_t* i_end0_curr = v_curr.ptr + (((i_slice_curr.len - (size_t)(v_curr.ptr - i_slice_curr.ptr)) / 6) * 6); + while (v_curr.ptr < i_end0_curr) { + v_fa0 = ((uint8_t)(v_fa0 + v_curr.ptr[0])); + v_curr.ptr[0] = v_fa0; + v_fa1 = ((uint8_t)(v_fa1 + v_curr.ptr[1])); + v_curr.ptr[1] = v_fa1; + v_fa2 = ((uint8_t)(v_fa2 + v_curr.ptr[2])); + v_curr.ptr[2] = v_fa2; + v_curr.ptr += 3; + v_fa0 = ((uint8_t)(v_fa0 + v_curr.ptr[0])); + v_curr.ptr[0] = v_fa0; + v_fa1 = ((uint8_t)(v_fa1 + v_curr.ptr[1])); + v_curr.ptr[1] = v_fa1; + v_fa2 = ((uint8_t)(v_fa2 + v_curr.ptr[2])); + v_curr.ptr[2] = v_fa2; + v_curr.ptr += 3; } - v_c.len = 3; - uint8_t* i_end1_c = v_c.ptr + (((i_slice_c.len - (size_t)(v_c.ptr - i_slice_c.ptr)) / 3) * 3); - while (v_c.ptr < i_end1_c) { - v_fa0 = ((uint8_t)(v_fa0 + v_c.ptr[0])); - v_c.ptr[0] = v_fa0; - v_fa1 = ((uint8_t)(v_fa1 + v_c.ptr[1])); - v_c.ptr[1] = v_fa1; - v_fa2 = ((uint8_t)(v_fa2 + v_c.ptr[2])); - v_c.ptr[2] = v_fa2; - v_c.ptr += 3; + v_curr.len = 3; + uint8_t* i_end1_curr = v_curr.ptr + (((i_slice_curr.len - (size_t)(v_curr.ptr - i_slice_curr.ptr)) / 3) * 3); + while (v_curr.ptr < i_end1_curr) { + v_fa0 = ((uint8_t)(v_fa0 + v_curr.ptr[0])); + v_curr.ptr[0] = v_fa0; + v_fa1 = ((uint8_t)(v_fa1 + v_curr.ptr[1])); + v_curr.ptr[1] = v_fa1; + v_fa2 = ((uint8_t)(v_fa2 + v_curr.ptr[2])); + v_curr.ptr[2] = v_fa2; + v_curr.ptr += 3; } - v_c.len = 0; + v_curr.len = 0; } return wuffs_base__make_empty_struct(); } @@ -32021,29 +32021,29 @@ static wuffs_base__empty_struct wuffs_png__decoder__filter_1_distance_4_fallback( wuffs_png__decoder* self, wuffs_base__slice_u8 a_curr) { - wuffs_base__slice_u8 v_c = {0}; + wuffs_base__slice_u8 v_curr = {0}; uint8_t v_fa0 = 0; uint8_t v_fa1 = 0; uint8_t v_fa2 = 0; uint8_t v_fa3 = 0; { - wuffs_base__slice_u8 i_slice_c = a_curr; - v_c.ptr = i_slice_c.ptr; - v_c.len = 4; - uint8_t* i_end0_c = v_c.ptr + (((i_slice_c.len - (size_t)(v_c.ptr - i_slice_c.ptr)) / 4) * 4); - while (v_c.ptr < i_end0_c) { - v_fa0 = ((uint8_t)(v_fa0 + v_c.ptr[0])); - v_c.ptr[0] = v_fa0; - v_fa1 = ((uint8_t)(v_fa1 + v_c.ptr[1])); - v_c.ptr[1] = v_fa1; - v_fa2 = ((uint8_t)(v_fa2 + v_c.ptr[2])); - v_c.ptr[2] = v_fa2; - v_fa3 = ((uint8_t)(v_fa3 + v_c.ptr[3])); - v_c.ptr[3] = v_fa3; - v_c.ptr += 4; + wuffs_base__slice_u8 i_slice_curr = a_curr; + v_curr.ptr = i_slice_curr.ptr; + v_curr.len = 4; + uint8_t* i_end0_curr = v_curr.ptr + (((i_slice_curr.len - (size_t)(v_curr.ptr - i_slice_curr.ptr)) / 4) * 4); + while (v_curr.ptr < i_end0_curr) { + v_fa0 = ((uint8_t)(v_fa0 + v_curr.ptr[0])); + v_curr.ptr[0] = v_fa0; + v_fa1 = ((uint8_t)(v_fa1 + v_curr.ptr[1])); + v_curr.ptr[1] = v_fa1; + v_fa2 = ((uint8_t)(v_fa2 + v_curr.ptr[2])); + v_curr.ptr[2] = v_fa2; + v_fa3 = ((uint8_t)(v_fa3 + v_curr.ptr[3])); + v_curr.ptr[3] = v_fa3; + v_curr.ptr += 4; } - v_c.len = 0; + v_curr.len = 0; } return wuffs_base__make_empty_struct(); } @@ -32116,90 +32116,90 @@ wuffs_png__decoder__filter_3_distance_3_fallback( wuffs_png__decoder* self, wuffs_base__slice_u8 a_curr, wuffs_base__slice_u8 a_prev) { - wuffs_base__slice_u8 v_c = {0}; - wuffs_base__slice_u8 v_p = {0}; + wuffs_base__slice_u8 v_curr = {0}; + wuffs_base__slice_u8 v_prev = {0}; uint8_t v_fa0 = 0; uint8_t v_fa1 = 0; uint8_t v_fa2 = 0; if (((uint64_t)(a_prev.len)) == 0) { { - wuffs_base__slice_u8 i_slice_c = a_curr; - v_c.ptr = i_slice_c.ptr; - v_c.len = 3; - uint8_t* i_end0_c = v_c.ptr + (((i_slice_c.len - (size_t)(v_c.ptr - i_slice_c.ptr)) / 6) * 6); - while (v_c.ptr < i_end0_c) { - v_fa0 = ((uint8_t)((v_fa0 / 2) + v_c.ptr[0])); - v_c.ptr[0] = v_fa0; - v_fa1 = ((uint8_t)((v_fa1 / 2) + v_c.ptr[1])); - v_c.ptr[1] = v_fa1; - v_fa2 = ((uint8_t)((v_fa2 / 2) + v_c.ptr[2])); - v_c.ptr[2] = v_fa2; - v_c.ptr += 3; - v_fa0 = ((uint8_t)((v_fa0 / 2) + v_c.ptr[0])); - v_c.ptr[0] = v_fa0; - v_fa1 = ((uint8_t)((v_fa1 / 2) + v_c.ptr[1])); - v_c.ptr[1] = v_fa1; - v_fa2 = ((uint8_t)((v_fa2 / 2) + v_c.ptr[2])); - v_c.ptr[2] = v_fa2; - v_c.ptr += 3; + wuffs_base__slice_u8 i_slice_curr = a_curr; + v_curr.ptr = i_slice_curr.ptr; + v_curr.len = 3; + uint8_t* i_end0_curr = v_curr.ptr + (((i_slice_curr.len - (size_t)(v_curr.ptr - i_slice_curr.ptr)) / 6) * 6); + while (v_curr.ptr < i_end0_curr) { + v_fa0 = ((uint8_t)((v_fa0 / 2) + v_curr.ptr[0])); + v_curr.ptr[0] = v_fa0; + v_fa1 = ((uint8_t)((v_fa1 / 2) + v_curr.ptr[1])); + v_curr.ptr[1] = v_fa1; + v_fa2 = ((uint8_t)((v_fa2 / 2) + v_curr.ptr[2])); + v_curr.ptr[2] = v_fa2; + v_curr.ptr += 3; + v_fa0 = ((uint8_t)((v_fa0 / 2) + v_curr.ptr[0])); + v_curr.ptr[0] = v_fa0; + v_fa1 = ((uint8_t)((v_fa1 / 2) + v_curr.ptr[1])); + v_curr.ptr[1] = v_fa1; + v_fa2 = ((uint8_t)((v_fa2 / 2) + v_curr.ptr[2])); + v_curr.ptr[2] = v_fa2; + v_curr.ptr += 3; } - v_c.len = 3; - uint8_t* i_end1_c = v_c.ptr + (((i_slice_c.len - (size_t)(v_c.ptr - i_slice_c.ptr)) / 3) * 3); - while (v_c.ptr < i_end1_c) { - v_fa0 = ((uint8_t)((v_fa0 / 2) + v_c.ptr[0])); - v_c.ptr[0] = v_fa0; - v_fa1 = ((uint8_t)((v_fa1 / 2) + v_c.ptr[1])); - v_c.ptr[1] = v_fa1; - v_fa2 = ((uint8_t)((v_fa2 / 2) + v_c.ptr[2])); - v_c.ptr[2] = v_fa2; - v_c.ptr += 3; + v_curr.len = 3; + uint8_t* i_end1_curr = v_curr.ptr + (((i_slice_curr.len - (size_t)(v_curr.ptr - i_slice_curr.ptr)) / 3) * 3); + while (v_curr.ptr < i_end1_curr) { + v_fa0 = ((uint8_t)((v_fa0 / 2) + v_curr.ptr[0])); + v_curr.ptr[0] = v_fa0; + v_fa1 = ((uint8_t)((v_fa1 / 2) + v_curr.ptr[1])); + v_curr.ptr[1] = v_fa1; + v_fa2 = ((uint8_t)((v_fa2 / 2) + v_curr.ptr[2])); + v_curr.ptr[2] = v_fa2; + v_curr.ptr += 3; } - v_c.len = 0; + v_curr.len = 0; } } else { { - wuffs_base__slice_u8 i_slice_c = a_curr; - v_c.ptr = i_slice_c.ptr; - wuffs_base__slice_u8 i_slice_p = a_prev; - v_p.ptr = i_slice_p.ptr; - i_slice_c.len = ((size_t)(wuffs_base__u64__min(i_slice_c.len, i_slice_p.len))); - v_c.len = 3; - v_p.len = 3; - uint8_t* i_end0_c = v_c.ptr + (((i_slice_c.len - (size_t)(v_c.ptr - i_slice_c.ptr)) / 6) * 6); - while (v_c.ptr < i_end0_c) { - v_fa0 = ((uint8_t)(((uint8_t)(((((uint32_t)(v_fa0)) + ((uint32_t)(v_p.ptr[0]))) / 2))) + v_c.ptr[0])); - v_c.ptr[0] = v_fa0; - v_fa1 = ((uint8_t)(((uint8_t)(((((uint32_t)(v_fa1)) + ((uint32_t)(v_p.ptr[1]))) / 2))) + v_c.ptr[1])); - v_c.ptr[1] = v_fa1; - v_fa2 = ((uint8_t)(((uint8_t)(((((uint32_t)(v_fa2)) + ((uint32_t)(v_p.ptr[2]))) / 2))) + v_c.ptr[2])); - v_c.ptr[2] = v_fa2; - v_c.ptr += 3; - v_p.ptr += 3; - v_fa0 = ((uint8_t)(((uint8_t)(((((uint32_t)(v_fa0)) + ((uint32_t)(v_p.ptr[0]))) / 2))) + v_c.ptr[0])); - v_c.ptr[0] = v_fa0; - v_fa1 = ((uint8_t)(((uint8_t)(((((uint32_t)(v_fa1)) + ((uint32_t)(v_p.ptr[1]))) / 2))) + v_c.ptr[1])); - v_c.ptr[1] = v_fa1; - v_fa2 = ((uint8_t)(((uint8_t)(((((uint32_t)(v_fa2)) + ((uint32_t)(v_p.ptr[2]))) / 2))) + v_c.ptr[2])); - v_c.ptr[2] = v_fa2; - v_c.ptr += 3; - v_p.ptr += 3; + wuffs_base__slice_u8 i_slice_curr = a_curr; + v_curr.ptr = i_slice_curr.ptr; + wuffs_base__slice_u8 i_slice_prev = a_prev; + v_prev.ptr = i_slice_prev.ptr; + i_slice_curr.len = ((size_t)(wuffs_base__u64__min(i_slice_curr.len, i_slice_prev.len))); + v_curr.len = 3; + v_prev.len = 3; + uint8_t* i_end0_curr = v_curr.ptr + (((i_slice_curr.len - (size_t)(v_curr.ptr - i_slice_curr.ptr)) / 6) * 6); + while (v_curr.ptr < i_end0_curr) { + v_fa0 = ((uint8_t)(((uint8_t)(((((uint32_t)(v_fa0)) + ((uint32_t)(v_prev.ptr[0]))) / 2))) + v_curr.ptr[0])); + v_curr.ptr[0] = v_fa0; + v_fa1 = ((uint8_t)(((uint8_t)(((((uint32_t)(v_fa1)) + ((uint32_t)(v_prev.ptr[1]))) / 2))) + v_curr.ptr[1])); + v_curr.ptr[1] = v_fa1; + v_fa2 = ((uint8_t)(((uint8_t)(((((uint32_t)(v_fa2)) + ((uint32_t)(v_prev.ptr[2]))) / 2))) + v_curr.ptr[2])); + v_curr.ptr[2] = v_fa2; + v_curr.ptr += 3; + v_prev.ptr += 3; + v_fa0 = ((uint8_t)(((uint8_t)(((((uint32_t)(v_fa0)) + ((uint32_t)(v_prev.ptr[0]))) / 2))) + v_curr.ptr[0])); + v_curr.ptr[0] = v_fa0; + v_fa1 = ((uint8_t)(((uint8_t)(((((uint32_t)(v_fa1)) + ((uint32_t)(v_prev.ptr[1]))) / 2))) + v_curr.ptr[1])); + v_curr.ptr[1] = v_fa1; + v_fa2 = ((uint8_t)(((uint8_t)(((((uint32_t)(v_fa2)) + ((uint32_t)(v_prev.ptr[2]))) / 2))) + v_curr.ptr[2])); + v_curr.ptr[2] = v_fa2; + v_curr.ptr += 3; + v_prev.ptr += 3; } - v_c.len = 3; - v_p.len = 3; - uint8_t* i_end1_c = v_c.ptr + (((i_slice_c.len - (size_t)(v_c.ptr - i_slice_c.ptr)) / 3) * 3); - while (v_c.ptr < i_end1_c) { - v_fa0 = ((uint8_t)(((uint8_t)(((((uint32_t)(v_fa0)) + ((uint32_t)(v_p.ptr[0]))) / 2))) + v_c.ptr[0])); - v_c.ptr[0] = v_fa0; - v_fa1 = ((uint8_t)(((uint8_t)(((((uint32_t)(v_fa1)) + ((uint32_t)(v_p.ptr[1]))) / 2))) + v_c.ptr[1])); - v_c.ptr[1] = v_fa1; - v_fa2 = ((uint8_t)(((uint8_t)(((((uint32_t)(v_fa2)) + ((uint32_t)(v_p.ptr[2]))) / 2))) + v_c.ptr[2])); - v_c.ptr[2] = v_fa2; - v_c.ptr += 3; - v_p.ptr += 3; + v_curr.len = 3; + v_prev.len = 3; + uint8_t* i_end1_curr = v_curr.ptr + (((i_slice_curr.len - (size_t)(v_curr.ptr - i_slice_curr.ptr)) / 3) * 3); + while (v_curr.ptr < i_end1_curr) { + v_fa0 = ((uint8_t)(((uint8_t)(((((uint32_t)(v_fa0)) + ((uint32_t)(v_prev.ptr[0]))) / 2))) + v_curr.ptr[0])); + v_curr.ptr[0] = v_fa0; + v_fa1 = ((uint8_t)(((uint8_t)(((((uint32_t)(v_fa1)) + ((uint32_t)(v_prev.ptr[1]))) / 2))) + v_curr.ptr[1])); + v_curr.ptr[1] = v_fa1; + v_fa2 = ((uint8_t)(((uint8_t)(((((uint32_t)(v_fa2)) + ((uint32_t)(v_prev.ptr[2]))) / 2))) + v_curr.ptr[2])); + v_curr.ptr[2] = v_fa2; + v_curr.ptr += 3; + v_prev.ptr += 3; } - v_c.len = 0; - v_p.len = 0; + v_curr.len = 0; + v_prev.len = 0; } } return wuffs_base__make_empty_struct(); @@ -32212,8 +32212,8 @@ wuffs_png__decoder__filter_3_distance_4_fallback( wuffs_png__decoder* self, wuffs_base__slice_u8 a_curr, wuffs_base__slice_u8 a_prev) { - wuffs_base__slice_u8 v_c = {0}; - wuffs_base__slice_u8 v_p = {0}; + wuffs_base__slice_u8 v_curr = {0}; + wuffs_base__slice_u8 v_prev = {0}; uint8_t v_fa0 = 0; uint8_t v_fa1 = 0; uint8_t v_fa2 = 0; @@ -32221,47 +32221,47 @@ wuffs_png__decoder__filter_3_distance_4_fallback( if (((uint64_t)(a_prev.len)) == 0) { { - wuffs_base__slice_u8 i_slice_c = a_curr; - v_c.ptr = i_slice_c.ptr; - v_c.len = 4; - uint8_t* i_end0_c = v_c.ptr + (((i_slice_c.len - (size_t)(v_c.ptr - i_slice_c.ptr)) / 4) * 4); - while (v_c.ptr < i_end0_c) { - v_fa0 = ((uint8_t)((v_fa0 / 2) + v_c.ptr[0])); - v_c.ptr[0] = v_fa0; - v_fa1 = ((uint8_t)((v_fa1 / 2) + v_c.ptr[1])); - v_c.ptr[1] = v_fa1; - v_fa2 = ((uint8_t)((v_fa2 / 2) + v_c.ptr[2])); - v_c.ptr[2] = v_fa2; - v_fa3 = ((uint8_t)((v_fa3 / 2) + v_c.ptr[3])); - v_c.ptr[3] = v_fa3; - v_c.ptr += 4; + wuffs_base__slice_u8 i_slice_curr = a_curr; + v_curr.ptr = i_slice_curr.ptr; + v_curr.len = 4; + uint8_t* i_end0_curr = v_curr.ptr + (((i_slice_curr.len - (size_t)(v_curr.ptr - i_slice_curr.ptr)) / 4) * 4); + while (v_curr.ptr < i_end0_curr) { + v_fa0 = ((uint8_t)((v_fa0 / 2) + v_curr.ptr[0])); + v_curr.ptr[0] = v_fa0; + v_fa1 = ((uint8_t)((v_fa1 / 2) + v_curr.ptr[1])); + v_curr.ptr[1] = v_fa1; + v_fa2 = ((uint8_t)((v_fa2 / 2) + v_curr.ptr[2])); + v_curr.ptr[2] = v_fa2; + v_fa3 = ((uint8_t)((v_fa3 / 2) + v_curr.ptr[3])); + v_curr.ptr[3] = v_fa3; + v_curr.ptr += 4; } - v_c.len = 0; + v_curr.len = 0; } } else { { - wuffs_base__slice_u8 i_slice_c = a_curr; - v_c.ptr = i_slice_c.ptr; - wuffs_base__slice_u8 i_slice_p = a_prev; - v_p.ptr = i_slice_p.ptr; - i_slice_c.len = ((size_t)(wuffs_base__u64__min(i_slice_c.len, i_slice_p.len))); - v_c.len = 4; - v_p.len = 4; - uint8_t* i_end0_c = v_c.ptr + (((i_slice_c.len - (size_t)(v_c.ptr - i_slice_c.ptr)) / 4) * 4); - while (v_c.ptr < i_end0_c) { - v_fa0 = ((uint8_t)(((uint8_t)(((((uint32_t)(v_fa0)) + ((uint32_t)(v_p.ptr[0]))) / 2))) + v_c.ptr[0])); - v_c.ptr[0] = v_fa0; - v_fa1 = ((uint8_t)(((uint8_t)(((((uint32_t)(v_fa1)) + ((uint32_t)(v_p.ptr[1]))) / 2))) + v_c.ptr[1])); - v_c.ptr[1] = v_fa1; - v_fa2 = ((uint8_t)(((uint8_t)(((((uint32_t)(v_fa2)) + ((uint32_t)(v_p.ptr[2]))) / 2))) + v_c.ptr[2])); - v_c.ptr[2] = v_fa2; - v_fa3 = ((uint8_t)(((uint8_t)(((((uint32_t)(v_fa3)) + ((uint32_t)(v_p.ptr[3]))) / 2))) + v_c.ptr[3])); - v_c.ptr[3] = v_fa3; - v_c.ptr += 4; - v_p.ptr += 4; + wuffs_base__slice_u8 i_slice_curr = a_curr; + v_curr.ptr = i_slice_curr.ptr; + wuffs_base__slice_u8 i_slice_prev = a_prev; + v_prev.ptr = i_slice_prev.ptr; + i_slice_curr.len = ((size_t)(wuffs_base__u64__min(i_slice_curr.len, i_slice_prev.len))); + v_curr.len = 4; + v_prev.len = 4; + uint8_t* i_end0_curr = v_curr.ptr + (((i_slice_curr.len - (size_t)(v_curr.ptr - i_slice_curr.ptr)) / 4) * 4); + while (v_curr.ptr < i_end0_curr) { + v_fa0 = ((uint8_t)(((uint8_t)(((((uint32_t)(v_fa0)) + ((uint32_t)(v_prev.ptr[0]))) / 2))) + v_curr.ptr[0])); + v_curr.ptr[0] = v_fa0; + v_fa1 = ((uint8_t)(((uint8_t)(((((uint32_t)(v_fa1)) + ((uint32_t)(v_prev.ptr[1]))) / 2))) + v_curr.ptr[1])); + v_curr.ptr[1] = v_fa1; + v_fa2 = ((uint8_t)(((uint8_t)(((((uint32_t)(v_fa2)) + ((uint32_t)(v_prev.ptr[2]))) / 2))) + v_curr.ptr[2])); + v_curr.ptr[2] = v_fa2; + v_fa3 = ((uint8_t)(((uint8_t)(((((uint32_t)(v_fa3)) + ((uint32_t)(v_prev.ptr[3]))) / 2))) + v_curr.ptr[3])); + v_curr.ptr[3] = v_fa3; + v_curr.ptr += 4; + v_prev.ptr += 4; } - v_c.len = 0; - v_p.len = 0; + v_curr.len = 0; + v_prev.len = 0; } } return wuffs_base__make_empty_struct(); @@ -32337,8 +32337,8 @@ wuffs_png__decoder__filter_4_distance_3_fallback( wuffs_png__decoder* self, wuffs_base__slice_u8 a_curr, wuffs_base__slice_u8 a_prev) { - wuffs_base__slice_u8 v_c = {0}; - wuffs_base__slice_u8 v_p = {0}; + wuffs_base__slice_u8 v_curr = {0}; + wuffs_base__slice_u8 v_prev = {0}; uint32_t v_fa0 = 0; uint32_t v_fa1 = 0; uint32_t v_fa2 = 0; @@ -32362,16 +32362,16 @@ wuffs_png__decoder__filter_4_distance_3_fallback( uint32_t v_pc2 = 0; { - wuffs_base__slice_u8 i_slice_c = a_curr; - v_c.ptr = i_slice_c.ptr; - wuffs_base__slice_u8 i_slice_p = a_prev; - v_p.ptr = i_slice_p.ptr; - i_slice_c.len = ((size_t)(wuffs_base__u64__min(i_slice_c.len, i_slice_p.len))); - v_c.len = 3; - v_p.len = 3; - uint8_t* i_end0_c = v_c.ptr + (((i_slice_c.len - (size_t)(v_c.ptr - i_slice_c.ptr)) / 3) * 3); - while (v_c.ptr < i_end0_c) { - v_fb0 = ((uint32_t)(v_p.ptr[0])); + wuffs_base__slice_u8 i_slice_curr = a_curr; + v_curr.ptr = i_slice_curr.ptr; + wuffs_base__slice_u8 i_slice_prev = a_prev; + v_prev.ptr = i_slice_prev.ptr; + i_slice_curr.len = ((size_t)(wuffs_base__u64__min(i_slice_curr.len, i_slice_prev.len))); + v_curr.len = 3; + v_prev.len = 3; + uint8_t* i_end0_curr = v_curr.ptr + (((i_slice_curr.len - (size_t)(v_curr.ptr - i_slice_curr.ptr)) / 3) * 3); + while (v_curr.ptr < i_end0_curr) { + v_fb0 = ((uint32_t)(v_prev.ptr[0])); v_pp0 = ((uint32_t)(((uint32_t)(v_fa0 + v_fb0)) - v_fc0)); v_pa0 = ((uint32_t)(v_pp0 - v_fa0)); if (v_pa0 >= 2147483648) { @@ -32391,10 +32391,10 @@ wuffs_png__decoder__filter_4_distance_3_fallback( } else { v_fa0 = v_fc0; } - v_c.ptr[0] = ((uint8_t)(v_c.ptr[0] + ((uint8_t)((v_fa0 & 255))))); - v_fa0 = ((uint32_t)(v_c.ptr[0])); + v_curr.ptr[0] = ((uint8_t)(v_curr.ptr[0] + ((uint8_t)((v_fa0 & 255))))); + v_fa0 = ((uint32_t)(v_curr.ptr[0])); v_fc0 = v_fb0; - v_fb1 = ((uint32_t)(v_p.ptr[1])); + v_fb1 = ((uint32_t)(v_prev.ptr[1])); v_pp1 = ((uint32_t)(((uint32_t)(v_fa1 + v_fb1)) - v_fc1)); v_pa1 = ((uint32_t)(v_pp1 - v_fa1)); if (v_pa1 >= 2147483648) { @@ -32414,10 +32414,10 @@ wuffs_png__decoder__filter_4_distance_3_fallback( } else { v_fa1 = v_fc1; } - v_c.ptr[1] = ((uint8_t)(v_c.ptr[1] + ((uint8_t)((v_fa1 & 255))))); - v_fa1 = ((uint32_t)(v_c.ptr[1])); + v_curr.ptr[1] = ((uint8_t)(v_curr.ptr[1] + ((uint8_t)((v_fa1 & 255))))); + v_fa1 = ((uint32_t)(v_curr.ptr[1])); v_fc1 = v_fb1; - v_fb2 = ((uint32_t)(v_p.ptr[2])); + v_fb2 = ((uint32_t)(v_prev.ptr[2])); v_pp2 = ((uint32_t)(((uint32_t)(v_fa2 + v_fb2)) - v_fc2)); v_pa2 = ((uint32_t)(v_pp2 - v_fa2)); if (v_pa2 >= 2147483648) { @@ -32437,14 +32437,14 @@ wuffs_png__decoder__filter_4_distance_3_fallback( } else { v_fa2 = v_fc2; } - v_c.ptr[2] = ((uint8_t)(v_c.ptr[2] + ((uint8_t)((v_fa2 & 255))))); - v_fa2 = ((uint32_t)(v_c.ptr[2])); + v_curr.ptr[2] = ((uint8_t)(v_curr.ptr[2] + ((uint8_t)((v_fa2 & 255))))); + v_fa2 = ((uint32_t)(v_curr.ptr[2])); v_fc2 = v_fb2; - v_c.ptr += 3; - v_p.ptr += 3; + v_curr.ptr += 3; + v_prev.ptr += 3; } - v_c.len = 0; - v_p.len = 0; + v_curr.len = 0; + v_prev.len = 0; } return wuffs_base__make_empty_struct(); } @@ -32456,8 +32456,8 @@ wuffs_png__decoder__filter_4_distance_4_fallback( wuffs_png__decoder* self, wuffs_base__slice_u8 a_curr, wuffs_base__slice_u8 a_prev) { - wuffs_base__slice_u8 v_c = {0}; - wuffs_base__slice_u8 v_p = {0}; + wuffs_base__slice_u8 v_curr = {0}; + wuffs_base__slice_u8 v_prev = {0}; uint32_t v_fa0 = 0; uint32_t v_fa1 = 0; uint32_t v_fa2 = 0; @@ -32488,16 +32488,16 @@ wuffs_png__decoder__filter_4_distance_4_fallback( uint32_t v_pc3 = 0; { - wuffs_base__slice_u8 i_slice_c = a_curr; - v_c.ptr = i_slice_c.ptr; - wuffs_base__slice_u8 i_slice_p = a_prev; - v_p.ptr = i_slice_p.ptr; - i_slice_c.len = ((size_t)(wuffs_base__u64__min(i_slice_c.len, i_slice_p.len))); - v_c.len = 4; - v_p.len = 4; - uint8_t* i_end0_c = v_c.ptr + (((i_slice_c.len - (size_t)(v_c.ptr - i_slice_c.ptr)) / 4) * 4); - while (v_c.ptr < i_end0_c) { - v_fb0 = ((uint32_t)(v_p.ptr[0])); + wuffs_base__slice_u8 i_slice_curr = a_curr; + v_curr.ptr = i_slice_curr.ptr; + wuffs_base__slice_u8 i_slice_prev = a_prev; + v_prev.ptr = i_slice_prev.ptr; + i_slice_curr.len = ((size_t)(wuffs_base__u64__min(i_slice_curr.len, i_slice_prev.len))); + v_curr.len = 4; + v_prev.len = 4; + uint8_t* i_end0_curr = v_curr.ptr + (((i_slice_curr.len - (size_t)(v_curr.ptr - i_slice_curr.ptr)) / 4) * 4); + while (v_curr.ptr < i_end0_curr) { + v_fb0 = ((uint32_t)(v_prev.ptr[0])); v_pp0 = ((uint32_t)(((uint32_t)(v_fa0 + v_fb0)) - v_fc0)); v_pa0 = ((uint32_t)(v_pp0 - v_fa0)); if (v_pa0 >= 2147483648) { @@ -32517,10 +32517,10 @@ wuffs_png__decoder__filter_4_distance_4_fallback( } else { v_fa0 = v_fc0; } - v_c.ptr[0] = ((uint8_t)(v_c.ptr[0] + ((uint8_t)((v_fa0 & 255))))); - v_fa0 = ((uint32_t)(v_c.ptr[0])); + v_curr.ptr[0] = ((uint8_t)(v_curr.ptr[0] + ((uint8_t)((v_fa0 & 255))))); + v_fa0 = ((uint32_t)(v_curr.ptr[0])); v_fc0 = v_fb0; - v_fb1 = ((uint32_t)(v_p.ptr[1])); + v_fb1 = ((uint32_t)(v_prev.ptr[1])); v_pp1 = ((uint32_t)(((uint32_t)(v_fa1 + v_fb1)) - v_fc1)); v_pa1 = ((uint32_t)(v_pp1 - v_fa1)); if (v_pa1 >= 2147483648) { @@ -32540,10 +32540,10 @@ wuffs_png__decoder__filter_4_distance_4_fallback( } else { v_fa1 = v_fc1; } - v_c.ptr[1] = ((uint8_t)(v_c.ptr[1] + ((uint8_t)((v_fa1 & 255))))); - v_fa1 = ((uint32_t)(v_c.ptr[1])); + v_curr.ptr[1] = ((uint8_t)(v_curr.ptr[1] + ((uint8_t)((v_fa1 & 255))))); + v_fa1 = ((uint32_t)(v_curr.ptr[1])); v_fc1 = v_fb1; - v_fb2 = ((uint32_t)(v_p.ptr[2])); + v_fb2 = ((uint32_t)(v_prev.ptr[2])); v_pp2 = ((uint32_t)(((uint32_t)(v_fa2 + v_fb2)) - v_fc2)); v_pa2 = ((uint32_t)(v_pp2 - v_fa2)); if (v_pa2 >= 2147483648) { @@ -32563,10 +32563,10 @@ wuffs_png__decoder__filter_4_distance_4_fallback( } else { v_fa2 = v_fc2; } - v_c.ptr[2] = ((uint8_t)(v_c.ptr[2] + ((uint8_t)((v_fa2 & 255))))); - v_fa2 = ((uint32_t)(v_c.ptr[2])); + v_curr.ptr[2] = ((uint8_t)(v_curr.ptr[2] + ((uint8_t)((v_fa2 & 255))))); + v_fa2 = ((uint32_t)(v_curr.ptr[2])); v_fc2 = v_fb2; - v_fb3 = ((uint32_t)(v_p.ptr[3])); + v_fb3 = ((uint32_t)(v_prev.ptr[3])); v_pp3 = ((uint32_t)(((uint32_t)(v_fa3 + v_fb3)) - v_fc3)); v_pa3 = ((uint32_t)(v_pp3 - v_fa3)); if (v_pa3 >= 2147483648) { @@ -32586,14 +32586,14 @@ wuffs_png__decoder__filter_4_distance_4_fallback( } else { v_fa3 = v_fc3; } - v_c.ptr[3] = ((uint8_t)(v_c.ptr[3] + ((uint8_t)((v_fa3 & 255))))); - v_fa3 = ((uint32_t)(v_c.ptr[3])); + v_curr.ptr[3] = ((uint8_t)(v_curr.ptr[3] + ((uint8_t)((v_fa3 & 255))))); + v_fa3 = ((uint32_t)(v_curr.ptr[3])); v_fc3 = v_fb3; - v_c.ptr += 4; - v_p.ptr += 4; + v_curr.ptr += 4; + v_prev.ptr += 4; } - v_c.len = 0; - v_p.len = 0; + v_curr.len = 0; + v_prev.len = 0; } return wuffs_base__make_empty_struct(); } @@ -32608,37 +32608,37 @@ static wuffs_base__empty_struct wuffs_png__decoder__filter_1_distance_4_x86_sse42( wuffs_png__decoder* self, wuffs_base__slice_u8 a_curr) { - wuffs_base__slice_u8 v_c = {0}; + wuffs_base__slice_u8 v_curr = {0}; __m128i v_x128 = {0}; __m128i v_a128 = {0}; { - wuffs_base__slice_u8 i_slice_c = a_curr; - v_c.ptr = i_slice_c.ptr; - v_c.len = 4; - uint8_t* i_end0_c = v_c.ptr + (((i_slice_c.len - (size_t)(v_c.ptr - i_slice_c.ptr)) / 8) * 8); - while (v_c.ptr < i_end0_c) { - v_x128 = _mm_cvtsi32_si128((int32_t)(wuffs_base__peek_u32le__no_bounds_check(v_c.ptr))); + wuffs_base__slice_u8 i_slice_curr = a_curr; + v_curr.ptr = i_slice_curr.ptr; + v_curr.len = 4; + uint8_t* i_end0_curr = v_curr.ptr + (((i_slice_curr.len - (size_t)(v_curr.ptr - i_slice_curr.ptr)) / 8) * 8); + while (v_curr.ptr < i_end0_curr) { + v_x128 = _mm_cvtsi32_si128((int32_t)(wuffs_base__peek_u32le__no_bounds_check(v_curr.ptr))); v_x128 = _mm_add_epi8(v_x128, v_a128); v_a128 = v_x128; - wuffs_base__poke_u32le__no_bounds_check(v_c.ptr, ((uint32_t)(_mm_cvtsi128_si32(v_x128)))); - v_c.ptr += 4; - v_x128 = _mm_cvtsi32_si128((int32_t)(wuffs_base__peek_u32le__no_bounds_check(v_c.ptr))); + wuffs_base__poke_u32le__no_bounds_check(v_curr.ptr, ((uint32_t)(_mm_cvtsi128_si32(v_x128)))); + v_curr.ptr += 4; + v_x128 = _mm_cvtsi32_si128((int32_t)(wuffs_base__peek_u32le__no_bounds_check(v_curr.ptr))); v_x128 = _mm_add_epi8(v_x128, v_a128); v_a128 = v_x128; - wuffs_base__poke_u32le__no_bounds_check(v_c.ptr, ((uint32_t)(_mm_cvtsi128_si32(v_x128)))); - v_c.ptr += 4; + wuffs_base__poke_u32le__no_bounds_check(v_curr.ptr, ((uint32_t)(_mm_cvtsi128_si32(v_x128)))); + v_curr.ptr += 4; } - v_c.len = 4; - uint8_t* i_end1_c = v_c.ptr + (((i_slice_c.len - (size_t)(v_c.ptr - i_slice_c.ptr)) / 4) * 4); - while (v_c.ptr < i_end1_c) { - v_x128 = _mm_cvtsi32_si128((int32_t)(wuffs_base__peek_u32le__no_bounds_check(v_c.ptr))); + v_curr.len = 4; + uint8_t* i_end1_curr = v_curr.ptr + (((i_slice_curr.len - (size_t)(v_curr.ptr - i_slice_curr.ptr)) / 4) * 4); + while (v_curr.ptr < i_end1_curr) { + v_x128 = _mm_cvtsi32_si128((int32_t)(wuffs_base__peek_u32le__no_bounds_check(v_curr.ptr))); v_x128 = _mm_add_epi8(v_x128, v_a128); v_a128 = v_x128; - wuffs_base__poke_u32le__no_bounds_check(v_c.ptr, ((uint32_t)(_mm_cvtsi128_si32(v_x128)))); - v_c.ptr += 4; + wuffs_base__poke_u32le__no_bounds_check(v_curr.ptr, ((uint32_t)(_mm_cvtsi128_si32(v_x128)))); + v_curr.ptr += 4; } - v_c.len = 0; + v_curr.len = 0; } return wuffs_base__make_empty_struct(); } @@ -32655,8 +32655,8 @@ wuffs_png__decoder__filter_3_distance_4_x86_sse42( wuffs_png__decoder* self, wuffs_base__slice_u8 a_curr, wuffs_base__slice_u8 a_prev) { - wuffs_base__slice_u8 v_c = {0}; - wuffs_base__slice_u8 v_p = {0}; + wuffs_base__slice_u8 v_curr = {0}; + wuffs_base__slice_u8 v_prev = {0}; __m128i v_x128 = {0}; __m128i v_a128 = {0}; __m128i v_b128 = {0}; @@ -32666,83 +32666,83 @@ wuffs_png__decoder__filter_3_distance_4_x86_sse42( if (((uint64_t)(a_prev.len)) == 0) { v_k128 = _mm_set1_epi8((int8_t)(254)); { - wuffs_base__slice_u8 i_slice_c = a_curr; - v_c.ptr = i_slice_c.ptr; - v_c.len = 4; - uint8_t* i_end0_c = v_c.ptr + (((i_slice_c.len - (size_t)(v_c.ptr - i_slice_c.ptr)) / 8) * 8); - while (v_c.ptr < i_end0_c) { + wuffs_base__slice_u8 i_slice_curr = a_curr; + v_curr.ptr = i_slice_curr.ptr; + v_curr.len = 4; + uint8_t* i_end0_curr = v_curr.ptr + (((i_slice_curr.len - (size_t)(v_curr.ptr - i_slice_curr.ptr)) / 8) * 8); + while (v_curr.ptr < i_end0_curr) { v_p128 = _mm_avg_epu8(_mm_and_si128(v_a128, v_k128), v_b128); - v_x128 = _mm_cvtsi32_si128((int32_t)(wuffs_base__peek_u32le__no_bounds_check(v_c.ptr))); + v_x128 = _mm_cvtsi32_si128((int32_t)(wuffs_base__peek_u32le__no_bounds_check(v_curr.ptr))); v_x128 = _mm_add_epi8(v_x128, v_p128); v_a128 = v_x128; - wuffs_base__poke_u32le__no_bounds_check(v_c.ptr, ((uint32_t)(_mm_cvtsi128_si32(v_x128)))); - v_c.ptr += 4; + wuffs_base__poke_u32le__no_bounds_check(v_curr.ptr, ((uint32_t)(_mm_cvtsi128_si32(v_x128)))); + v_curr.ptr += 4; v_p128 = _mm_avg_epu8(_mm_and_si128(v_a128, v_k128), v_b128); - v_x128 = _mm_cvtsi32_si128((int32_t)(wuffs_base__peek_u32le__no_bounds_check(v_c.ptr))); + v_x128 = _mm_cvtsi32_si128((int32_t)(wuffs_base__peek_u32le__no_bounds_check(v_curr.ptr))); v_x128 = _mm_add_epi8(v_x128, v_p128); v_a128 = v_x128; - wuffs_base__poke_u32le__no_bounds_check(v_c.ptr, ((uint32_t)(_mm_cvtsi128_si32(v_x128)))); - v_c.ptr += 4; + wuffs_base__poke_u32le__no_bounds_check(v_curr.ptr, ((uint32_t)(_mm_cvtsi128_si32(v_x128)))); + v_curr.ptr += 4; } - v_c.len = 4; - uint8_t* i_end1_c = v_c.ptr + (((i_slice_c.len - (size_t)(v_c.ptr - i_slice_c.ptr)) / 4) * 4); - while (v_c.ptr < i_end1_c) { + v_curr.len = 4; + uint8_t* i_end1_curr = v_curr.ptr + (((i_slice_curr.len - (size_t)(v_curr.ptr - i_slice_curr.ptr)) / 4) * 4); + while (v_curr.ptr < i_end1_curr) { v_p128 = _mm_avg_epu8(_mm_and_si128(v_a128, v_k128), v_b128); - v_x128 = _mm_cvtsi32_si128((int32_t)(wuffs_base__peek_u32le__no_bounds_check(v_c.ptr))); + v_x128 = _mm_cvtsi32_si128((int32_t)(wuffs_base__peek_u32le__no_bounds_check(v_curr.ptr))); v_x128 = _mm_add_epi8(v_x128, v_p128); v_a128 = v_x128; - wuffs_base__poke_u32le__no_bounds_check(v_c.ptr, ((uint32_t)(_mm_cvtsi128_si32(v_x128)))); - v_c.ptr += 4; + wuffs_base__poke_u32le__no_bounds_check(v_curr.ptr, ((uint32_t)(_mm_cvtsi128_si32(v_x128)))); + v_curr.ptr += 4; } - v_c.len = 0; + v_curr.len = 0; } } else { v_k128 = _mm_set1_epi8((int8_t)(1)); { - wuffs_base__slice_u8 i_slice_c = a_curr; - v_c.ptr = i_slice_c.ptr; - wuffs_base__slice_u8 i_slice_p = a_prev; - v_p.ptr = i_slice_p.ptr; - i_slice_c.len = ((size_t)(wuffs_base__u64__min(i_slice_c.len, i_slice_p.len))); - v_c.len = 4; - v_p.len = 4; - uint8_t* i_end0_c = v_c.ptr + (((i_slice_c.len - (size_t)(v_c.ptr - i_slice_c.ptr)) / 8) * 8); - while (v_c.ptr < i_end0_c) { - v_b128 = _mm_cvtsi32_si128((int32_t)(wuffs_base__peek_u32le__no_bounds_check(v_p.ptr))); + wuffs_base__slice_u8 i_slice_curr = a_curr; + v_curr.ptr = i_slice_curr.ptr; + wuffs_base__slice_u8 i_slice_prev = a_prev; + v_prev.ptr = i_slice_prev.ptr; + i_slice_curr.len = ((size_t)(wuffs_base__u64__min(i_slice_curr.len, i_slice_prev.len))); + v_curr.len = 4; + v_prev.len = 4; + uint8_t* i_end0_curr = v_curr.ptr + (((i_slice_curr.len - (size_t)(v_curr.ptr - i_slice_curr.ptr)) / 8) * 8); + while (v_curr.ptr < i_end0_curr) { + v_b128 = _mm_cvtsi32_si128((int32_t)(wuffs_base__peek_u32le__no_bounds_check(v_prev.ptr))); v_p128 = _mm_avg_epu8(v_a128, v_b128); v_p128 = _mm_sub_epi8(v_p128, _mm_and_si128(v_k128, _mm_xor_si128(v_a128, v_b128))); - v_x128 = _mm_cvtsi32_si128((int32_t)(wuffs_base__peek_u32le__no_bounds_check(v_c.ptr))); + v_x128 = _mm_cvtsi32_si128((int32_t)(wuffs_base__peek_u32le__no_bounds_check(v_curr.ptr))); v_x128 = _mm_add_epi8(v_x128, v_p128); v_a128 = v_x128; - wuffs_base__poke_u32le__no_bounds_check(v_c.ptr, ((uint32_t)(_mm_cvtsi128_si32(v_x128)))); - v_c.ptr += 4; - v_p.ptr += 4; - v_b128 = _mm_cvtsi32_si128((int32_t)(wuffs_base__peek_u32le__no_bounds_check(v_p.ptr))); + wuffs_base__poke_u32le__no_bounds_check(v_curr.ptr, ((uint32_t)(_mm_cvtsi128_si32(v_x128)))); + v_curr.ptr += 4; + v_prev.ptr += 4; + v_b128 = _mm_cvtsi32_si128((int32_t)(wuffs_base__peek_u32le__no_bounds_check(v_prev.ptr))); v_p128 = _mm_avg_epu8(v_a128, v_b128); v_p128 = _mm_sub_epi8(v_p128, _mm_and_si128(v_k128, _mm_xor_si128(v_a128, v_b128))); - v_x128 = _mm_cvtsi32_si128((int32_t)(wuffs_base__peek_u32le__no_bounds_check(v_c.ptr))); + v_x128 = _mm_cvtsi32_si128((int32_t)(wuffs_base__peek_u32le__no_bounds_check(v_curr.ptr))); v_x128 = _mm_add_epi8(v_x128, v_p128); v_a128 = v_x128; - wuffs_base__poke_u32le__no_bounds_check(v_c.ptr, ((uint32_t)(_mm_cvtsi128_si32(v_x128)))); - v_c.ptr += 4; - v_p.ptr += 4; + wuffs_base__poke_u32le__no_bounds_check(v_curr.ptr, ((uint32_t)(_mm_cvtsi128_si32(v_x128)))); + v_curr.ptr += 4; + v_prev.ptr += 4; } - v_c.len = 4; - v_p.len = 4; - uint8_t* i_end1_c = v_c.ptr + (((i_slice_c.len - (size_t)(v_c.ptr - i_slice_c.ptr)) / 4) * 4); - while (v_c.ptr < i_end1_c) { - v_b128 = _mm_cvtsi32_si128((int32_t)(wuffs_base__peek_u32le__no_bounds_check(v_p.ptr))); + v_curr.len = 4; + v_prev.len = 4; + uint8_t* i_end1_curr = v_curr.ptr + (((i_slice_curr.len - (size_t)(v_curr.ptr - i_slice_curr.ptr)) / 4) * 4); + while (v_curr.ptr < i_end1_curr) { + v_b128 = _mm_cvtsi32_si128((int32_t)(wuffs_base__peek_u32le__no_bounds_check(v_prev.ptr))); v_p128 = _mm_avg_epu8(v_a128, v_b128); v_p128 = _mm_sub_epi8(v_p128, _mm_and_si128(v_k128, _mm_xor_si128(v_a128, v_b128))); - v_x128 = _mm_cvtsi32_si128((int32_t)(wuffs_base__peek_u32le__no_bounds_check(v_c.ptr))); + v_x128 = _mm_cvtsi32_si128((int32_t)(wuffs_base__peek_u32le__no_bounds_check(v_curr.ptr))); v_x128 = _mm_add_epi8(v_x128, v_p128); v_a128 = v_x128; - wuffs_base__poke_u32le__no_bounds_check(v_c.ptr, ((uint32_t)(_mm_cvtsi128_si32(v_x128)))); - v_c.ptr += 4; - v_p.ptr += 4; + wuffs_base__poke_u32le__no_bounds_check(v_curr.ptr, ((uint32_t)(_mm_cvtsi128_si32(v_x128)))); + v_curr.ptr += 4; + v_prev.ptr += 4; } - v_c.len = 0; - v_p.len = 0; + v_curr.len = 0; + v_prev.len = 0; } } return wuffs_base__make_empty_struct(); @@ -32760,8 +32760,8 @@ wuffs_png__decoder__filter_4_distance_3_x86_sse42( wuffs_png__decoder* self, wuffs_base__slice_u8 a_curr, wuffs_base__slice_u8 a_prev) { - wuffs_base__slice_u8 v_c = {0}; - wuffs_base__slice_u8 v_p = {0}; + wuffs_base__slice_u8 v_curr = {0}; + wuffs_base__slice_u8 v_prev = {0}; __m128i v_x128 = {0}; __m128i v_a128 = {0}; __m128i v_b128 = {0}; @@ -32774,16 +32774,16 @@ wuffs_png__decoder__filter_4_distance_3_x86_sse42( __m128i v_z128 = {0}; { - wuffs_base__slice_u8 i_slice_c = a_curr; - v_c.ptr = i_slice_c.ptr; - wuffs_base__slice_u8 i_slice_p = a_prev; - v_p.ptr = i_slice_p.ptr; - i_slice_c.len = ((size_t)(wuffs_base__u64__min(i_slice_c.len, i_slice_p.len))); - v_c.len = 4; - v_p.len = 4; - uint8_t* i_end0_c = v_c.ptr + wuffs_base__iterate_total_advance((i_slice_c.len - (size_t)(v_c.ptr - i_slice_c.ptr)), 7, 6); - while (v_c.ptr < i_end0_c) { - v_b128 = _mm_cvtsi32_si128((int32_t)(wuffs_base__peek_u32le__no_bounds_check(v_p.ptr))); + wuffs_base__slice_u8 i_slice_curr = a_curr; + v_curr.ptr = i_slice_curr.ptr; + wuffs_base__slice_u8 i_slice_prev = a_prev; + v_prev.ptr = i_slice_prev.ptr; + i_slice_curr.len = ((size_t)(wuffs_base__u64__min(i_slice_curr.len, i_slice_prev.len))); + v_curr.len = 4; + v_prev.len = 4; + uint8_t* i_end0_curr = v_curr.ptr + wuffs_base__iterate_total_advance((i_slice_curr.len - (size_t)(v_curr.ptr - i_slice_curr.ptr)), 7, 6); + while (v_curr.ptr < i_end0_curr) { + v_b128 = _mm_cvtsi32_si128((int32_t)(wuffs_base__peek_u32le__no_bounds_check(v_prev.ptr))); v_b128 = _mm_unpacklo_epi8(v_b128, v_z128); v_pa128 = _mm_sub_epi16(v_b128, v_c128); v_pb128 = _mm_sub_epi16(v_a128, v_c128); @@ -32793,16 +32793,16 @@ wuffs_png__decoder__filter_4_distance_3_x86_sse42( v_pc128 = _mm_abs_epi16(v_pc128); v_smallest128 = _mm_min_epi16(v_pc128, _mm_min_epi16(v_pb128, v_pa128)); v_p128 = _mm_blendv_epi8(_mm_blendv_epi8(v_c128, v_b128, _mm_cmpeq_epi16(v_smallest128, v_pb128)), v_a128, _mm_cmpeq_epi16(v_smallest128, v_pa128)); - v_x128 = _mm_cvtsi32_si128((int32_t)(wuffs_base__peek_u32le__no_bounds_check(v_c.ptr))); + v_x128 = _mm_cvtsi32_si128((int32_t)(wuffs_base__peek_u32le__no_bounds_check(v_curr.ptr))); v_x128 = _mm_unpacklo_epi8(v_x128, v_z128); v_x128 = _mm_add_epi8(v_x128, v_p128); v_a128 = v_x128; v_c128 = v_b128; v_x128 = _mm_packus_epi16(v_x128, v_x128); - wuffs_base__poke_u24le__no_bounds_check(v_c.ptr, ((uint32_t)(_mm_cvtsi128_si32(v_x128)))); - v_c.ptr += 3; - v_p.ptr += 3; - v_b128 = _mm_cvtsi32_si128((int32_t)(wuffs_base__peek_u32le__no_bounds_check(v_p.ptr))); + wuffs_base__poke_u24le__no_bounds_check(v_curr.ptr, ((uint32_t)(_mm_cvtsi128_si32(v_x128)))); + v_curr.ptr += 3; + v_prev.ptr += 3; + v_b128 = _mm_cvtsi32_si128((int32_t)(wuffs_base__peek_u32le__no_bounds_check(v_prev.ptr))); v_b128 = _mm_unpacklo_epi8(v_b128, v_z128); v_pa128 = _mm_sub_epi16(v_b128, v_c128); v_pb128 = _mm_sub_epi16(v_a128, v_c128); @@ -32812,21 +32812,21 @@ wuffs_png__decoder__filter_4_distance_3_x86_sse42( v_pc128 = _mm_abs_epi16(v_pc128); v_smallest128 = _mm_min_epi16(v_pc128, _mm_min_epi16(v_pb128, v_pa128)); v_p128 = _mm_blendv_epi8(_mm_blendv_epi8(v_c128, v_b128, _mm_cmpeq_epi16(v_smallest128, v_pb128)), v_a128, _mm_cmpeq_epi16(v_smallest128, v_pa128)); - v_x128 = _mm_cvtsi32_si128((int32_t)(wuffs_base__peek_u32le__no_bounds_check(v_c.ptr))); + v_x128 = _mm_cvtsi32_si128((int32_t)(wuffs_base__peek_u32le__no_bounds_check(v_curr.ptr))); v_x128 = _mm_unpacklo_epi8(v_x128, v_z128); v_x128 = _mm_add_epi8(v_x128, v_p128); v_a128 = v_x128; v_c128 = v_b128; v_x128 = _mm_packus_epi16(v_x128, v_x128); - wuffs_base__poke_u24le__no_bounds_check(v_c.ptr, ((uint32_t)(_mm_cvtsi128_si32(v_x128)))); - v_c.ptr += 3; - v_p.ptr += 3; + wuffs_base__poke_u24le__no_bounds_check(v_curr.ptr, ((uint32_t)(_mm_cvtsi128_si32(v_x128)))); + v_curr.ptr += 3; + v_prev.ptr += 3; } - v_c.len = 4; - v_p.len = 4; - uint8_t* i_end1_c = v_c.ptr + wuffs_base__iterate_total_advance((i_slice_c.len - (size_t)(v_c.ptr - i_slice_c.ptr)), 4, 3); - while (v_c.ptr < i_end1_c) { - v_b128 = _mm_cvtsi32_si128((int32_t)(wuffs_base__peek_u32le__no_bounds_check(v_p.ptr))); + v_curr.len = 4; + v_prev.len = 4; + uint8_t* i_end1_curr = v_curr.ptr + wuffs_base__iterate_total_advance((i_slice_curr.len - (size_t)(v_curr.ptr - i_slice_curr.ptr)), 4, 3); + while (v_curr.ptr < i_end1_curr) { + v_b128 = _mm_cvtsi32_si128((int32_t)(wuffs_base__peek_u32le__no_bounds_check(v_prev.ptr))); v_b128 = _mm_unpacklo_epi8(v_b128, v_z128); v_pa128 = _mm_sub_epi16(v_b128, v_c128); v_pb128 = _mm_sub_epi16(v_a128, v_c128); @@ -32836,21 +32836,21 @@ wuffs_png__decoder__filter_4_distance_3_x86_sse42( v_pc128 = _mm_abs_epi16(v_pc128); v_smallest128 = _mm_min_epi16(v_pc128, _mm_min_epi16(v_pb128, v_pa128)); v_p128 = _mm_blendv_epi8(_mm_blendv_epi8(v_c128, v_b128, _mm_cmpeq_epi16(v_smallest128, v_pb128)), v_a128, _mm_cmpeq_epi16(v_smallest128, v_pa128)); - v_x128 = _mm_cvtsi32_si128((int32_t)(wuffs_base__peek_u32le__no_bounds_check(v_c.ptr))); + v_x128 = _mm_cvtsi32_si128((int32_t)(wuffs_base__peek_u32le__no_bounds_check(v_curr.ptr))); v_x128 = _mm_unpacklo_epi8(v_x128, v_z128); v_x128 = _mm_add_epi8(v_x128, v_p128); v_a128 = v_x128; v_c128 = v_b128; v_x128 = _mm_packus_epi16(v_x128, v_x128); - wuffs_base__poke_u24le__no_bounds_check(v_c.ptr, ((uint32_t)(_mm_cvtsi128_si32(v_x128)))); - v_c.ptr += 3; - v_p.ptr += 3; + wuffs_base__poke_u24le__no_bounds_check(v_curr.ptr, ((uint32_t)(_mm_cvtsi128_si32(v_x128)))); + v_curr.ptr += 3; + v_prev.ptr += 3; } - v_c.len = 3; - v_p.len = 3; - uint8_t* i_end2_c = v_c.ptr + (((i_slice_c.len - (size_t)(v_c.ptr - i_slice_c.ptr)) / 3) * 3); - while (v_c.ptr < i_end2_c) { - v_b128 = _mm_cvtsi32_si128((int32_t)(wuffs_base__peek_u24le__no_bounds_check(v_p.ptr))); + v_curr.len = 3; + v_prev.len = 3; + uint8_t* i_end2_curr = v_curr.ptr + (((i_slice_curr.len - (size_t)(v_curr.ptr - i_slice_curr.ptr)) / 3) * 3); + while (v_curr.ptr < i_end2_curr) { + v_b128 = _mm_cvtsi32_si128((int32_t)(wuffs_base__peek_u24le__no_bounds_check(v_prev.ptr))); v_b128 = _mm_unpacklo_epi8(v_b128, v_z128); v_pa128 = _mm_sub_epi16(v_b128, v_c128); v_pb128 = _mm_sub_epi16(v_a128, v_c128); @@ -32860,16 +32860,16 @@ wuffs_png__decoder__filter_4_distance_3_x86_sse42( v_pc128 = _mm_abs_epi16(v_pc128); v_smallest128 = _mm_min_epi16(v_pc128, _mm_min_epi16(v_pb128, v_pa128)); v_p128 = _mm_blendv_epi8(_mm_blendv_epi8(v_c128, v_b128, _mm_cmpeq_epi16(v_smallest128, v_pb128)), v_a128, _mm_cmpeq_epi16(v_smallest128, v_pa128)); - v_x128 = _mm_cvtsi32_si128((int32_t)(wuffs_base__peek_u24le__no_bounds_check(v_c.ptr))); + v_x128 = _mm_cvtsi32_si128((int32_t)(wuffs_base__peek_u24le__no_bounds_check(v_curr.ptr))); v_x128 = _mm_unpacklo_epi8(v_x128, v_z128); v_x128 = _mm_add_epi8(v_x128, v_p128); v_x128 = _mm_packus_epi16(v_x128, v_x128); - wuffs_base__poke_u24le__no_bounds_check(v_c.ptr, ((uint32_t)(_mm_cvtsi128_si32(v_x128)))); - v_c.ptr += 3; - v_p.ptr += 3; + wuffs_base__poke_u24le__no_bounds_check(v_curr.ptr, ((uint32_t)(_mm_cvtsi128_si32(v_x128)))); + v_curr.ptr += 3; + v_prev.ptr += 3; } - v_c.len = 0; - v_p.len = 0; + v_curr.len = 0; + v_prev.len = 0; } return wuffs_base__make_empty_struct(); } @@ -32886,8 +32886,8 @@ wuffs_png__decoder__filter_4_distance_4_x86_sse42( wuffs_png__decoder* self, wuffs_base__slice_u8 a_curr, wuffs_base__slice_u8 a_prev) { - wuffs_base__slice_u8 v_c = {0}; - wuffs_base__slice_u8 v_p = {0}; + wuffs_base__slice_u8 v_curr = {0}; + wuffs_base__slice_u8 v_prev = {0}; __m128i v_x128 = {0}; __m128i v_a128 = {0}; __m128i v_b128 = {0}; @@ -32900,16 +32900,16 @@ wuffs_png__decoder__filter_4_distance_4_x86_sse42( __m128i v_z128 = {0}; { - wuffs_base__slice_u8 i_slice_c = a_curr; - v_c.ptr = i_slice_c.ptr; - wuffs_base__slice_u8 i_slice_p = a_prev; - v_p.ptr = i_slice_p.ptr; - i_slice_c.len = ((size_t)(wuffs_base__u64__min(i_slice_c.len, i_slice_p.len))); - v_c.len = 4; - v_p.len = 4; - uint8_t* i_end0_c = v_c.ptr + (((i_slice_c.len - (size_t)(v_c.ptr - i_slice_c.ptr)) / 8) * 8); - while (v_c.ptr < i_end0_c) { - v_b128 = _mm_cvtsi32_si128((int32_t)(wuffs_base__peek_u32le__no_bounds_check(v_p.ptr))); + wuffs_base__slice_u8 i_slice_curr = a_curr; + v_curr.ptr = i_slice_curr.ptr; + wuffs_base__slice_u8 i_slice_prev = a_prev; + v_prev.ptr = i_slice_prev.ptr; + i_slice_curr.len = ((size_t)(wuffs_base__u64__min(i_slice_curr.len, i_slice_prev.len))); + v_curr.len = 4; + v_prev.len = 4; + uint8_t* i_end0_curr = v_curr.ptr + (((i_slice_curr.len - (size_t)(v_curr.ptr - i_slice_curr.ptr)) / 8) * 8); + while (v_curr.ptr < i_end0_curr) { + v_b128 = _mm_cvtsi32_si128((int32_t)(wuffs_base__peek_u32le__no_bounds_check(v_prev.ptr))); v_b128 = _mm_unpacklo_epi8(v_b128, v_z128); v_pa128 = _mm_sub_epi16(v_b128, v_c128); v_pb128 = _mm_sub_epi16(v_a128, v_c128); @@ -32919,16 +32919,16 @@ wuffs_png__decoder__filter_4_distance_4_x86_sse42( v_pc128 = _mm_abs_epi16(v_pc128); v_smallest128 = _mm_min_epi16(v_pc128, _mm_min_epi16(v_pb128, v_pa128)); v_p128 = _mm_blendv_epi8(_mm_blendv_epi8(v_c128, v_b128, _mm_cmpeq_epi16(v_smallest128, v_pb128)), v_a128, _mm_cmpeq_epi16(v_smallest128, v_pa128)); - v_x128 = _mm_cvtsi32_si128((int32_t)(wuffs_base__peek_u32le__no_bounds_check(v_c.ptr))); + v_x128 = _mm_cvtsi32_si128((int32_t)(wuffs_base__peek_u32le__no_bounds_check(v_curr.ptr))); v_x128 = _mm_unpacklo_epi8(v_x128, v_z128); v_x128 = _mm_add_epi8(v_x128, v_p128); v_a128 = v_x128; v_c128 = v_b128; v_x128 = _mm_packus_epi16(v_x128, v_x128); - wuffs_base__poke_u32le__no_bounds_check(v_c.ptr, ((uint32_t)(_mm_cvtsi128_si32(v_x128)))); - v_c.ptr += 4; - v_p.ptr += 4; - v_b128 = _mm_cvtsi32_si128((int32_t)(wuffs_base__peek_u32le__no_bounds_check(v_p.ptr))); + wuffs_base__poke_u32le__no_bounds_check(v_curr.ptr, ((uint32_t)(_mm_cvtsi128_si32(v_x128)))); + v_curr.ptr += 4; + v_prev.ptr += 4; + v_b128 = _mm_cvtsi32_si128((int32_t)(wuffs_base__peek_u32le__no_bounds_check(v_prev.ptr))); v_b128 = _mm_unpacklo_epi8(v_b128, v_z128); v_pa128 = _mm_sub_epi16(v_b128, v_c128); v_pb128 = _mm_sub_epi16(v_a128, v_c128); @@ -32938,21 +32938,21 @@ wuffs_png__decoder__filter_4_distance_4_x86_sse42( v_pc128 = _mm_abs_epi16(v_pc128); v_smallest128 = _mm_min_epi16(v_pc128, _mm_min_epi16(v_pb128, v_pa128)); v_p128 = _mm_blendv_epi8(_mm_blendv_epi8(v_c128, v_b128, _mm_cmpeq_epi16(v_smallest128, v_pb128)), v_a128, _mm_cmpeq_epi16(v_smallest128, v_pa128)); - v_x128 = _mm_cvtsi32_si128((int32_t)(wuffs_base__peek_u32le__no_bounds_check(v_c.ptr))); + v_x128 = _mm_cvtsi32_si128((int32_t)(wuffs_base__peek_u32le__no_bounds_check(v_curr.ptr))); v_x128 = _mm_unpacklo_epi8(v_x128, v_z128); v_x128 = _mm_add_epi8(v_x128, v_p128); v_a128 = v_x128; v_c128 = v_b128; v_x128 = _mm_packus_epi16(v_x128, v_x128); - wuffs_base__poke_u32le__no_bounds_check(v_c.ptr, ((uint32_t)(_mm_cvtsi128_si32(v_x128)))); - v_c.ptr += 4; - v_p.ptr += 4; + wuffs_base__poke_u32le__no_bounds_check(v_curr.ptr, ((uint32_t)(_mm_cvtsi128_si32(v_x128)))); + v_curr.ptr += 4; + v_prev.ptr += 4; } - v_c.len = 4; - v_p.len = 4; - uint8_t* i_end1_c = v_c.ptr + (((i_slice_c.len - (size_t)(v_c.ptr - i_slice_c.ptr)) / 4) * 4); - while (v_c.ptr < i_end1_c) { - v_b128 = _mm_cvtsi32_si128((int32_t)(wuffs_base__peek_u32le__no_bounds_check(v_p.ptr))); + v_curr.len = 4; + v_prev.len = 4; + uint8_t* i_end1_curr = v_curr.ptr + (((i_slice_curr.len - (size_t)(v_curr.ptr - i_slice_curr.ptr)) / 4) * 4); + while (v_curr.ptr < i_end1_curr) { + v_b128 = _mm_cvtsi32_si128((int32_t)(wuffs_base__peek_u32le__no_bounds_check(v_prev.ptr))); v_b128 = _mm_unpacklo_epi8(v_b128, v_z128); v_pa128 = _mm_sub_epi16(v_b128, v_c128); v_pb128 = _mm_sub_epi16(v_a128, v_c128); @@ -32962,18 +32962,18 @@ wuffs_png__decoder__filter_4_distance_4_x86_sse42( v_pc128 = _mm_abs_epi16(v_pc128); v_smallest128 = _mm_min_epi16(v_pc128, _mm_min_epi16(v_pb128, v_pa128)); v_p128 = _mm_blendv_epi8(_mm_blendv_epi8(v_c128, v_b128, _mm_cmpeq_epi16(v_smallest128, v_pb128)), v_a128, _mm_cmpeq_epi16(v_smallest128, v_pa128)); - v_x128 = _mm_cvtsi32_si128((int32_t)(wuffs_base__peek_u32le__no_bounds_check(v_c.ptr))); + v_x128 = _mm_cvtsi32_si128((int32_t)(wuffs_base__peek_u32le__no_bounds_check(v_curr.ptr))); v_x128 = _mm_unpacklo_epi8(v_x128, v_z128); v_x128 = _mm_add_epi8(v_x128, v_p128); v_a128 = v_x128; v_c128 = v_b128; v_x128 = _mm_packus_epi16(v_x128, v_x128); - wuffs_base__poke_u32le__no_bounds_check(v_c.ptr, ((uint32_t)(_mm_cvtsi128_si32(v_x128)))); - v_c.ptr += 4; - v_p.ptr += 4; + wuffs_base__poke_u32le__no_bounds_check(v_curr.ptr, ((uint32_t)(_mm_cvtsi128_si32(v_x128)))); + v_curr.ptr += 4; + v_prev.ptr += 4; } - v_c.len = 0; - v_p.len = 0; + v_curr.len = 0; + v_prev.len = 0; } return wuffs_base__make_empty_struct(); } diff --git a/std/png/decode_filter_arm_neon.wuffs b/std/png/decode_filter_arm_neon.wuffs index 3ee7e90c..4fe7e07f 100644 --- a/std/png/decode_filter_arm_neon.wuffs +++ b/std/png/decode_filter_arm_neon.wuffs @@ -31,14 +31,15 @@ pri func decoder.filter_1_distance_4_arm_neon!(curr: slice base.u8), choose cpu_arch >= arm_neon, { - var c : slice base.u8 + var curr : slice base.u8 + var fa : base.arm_neon_64 var fx : base.arm_neon_64 - iterate (c = args.curr)(length: 4, advance: 4, unroll: 2) { - fx = fx.create_vdup_n_u32(a: c.peek_u32le()) + iterate (curr = args.curr)(length: 4, advance: 4, unroll: 2) { + fx = fx.create_vdup_n_u32(a: curr.peek_u32le()) fx = fx.vadd_u8(b: fa) - c.poke_u32le!(a: fx.vget_lane_u32(b: 0)) + curr.poke_u32le!(a: fx.vget_lane_u32(b: 0)) fa = fx } } @@ -50,26 +51,27 @@ pri func decoder.filter_1_distance_4_arm_neon!(curr: slice base.u8), pri func decoder.filter_3_distance_4_arm_neon!(curr: slice base.u8, prev: slice base.u8), choose cpu_arch >= arm_neon, { - var c : slice base.u8 - var p : slice base.u8 + var curr : slice base.u8 + var prev : slice base.u8 + var fa : base.arm_neon_64 var fb : base.arm_neon_64 var fx : base.arm_neon_64 if args.prev.length() == 0 { - iterate (c = args.curr)(length: 4, advance: 4, unroll: 2) { - fx = fx.create_vdup_n_u32(a: c.peek_u32le()) + iterate (curr = args.curr)(length: 4, advance: 4, unroll: 2) { + fx = fx.create_vdup_n_u32(a: curr.peek_u32le()) fx = fx.vadd_u8(b: fa.vhadd_u8(b: fb)) - c.poke_u32le!(a: fx.vget_lane_u32(b: 0)) + curr.poke_u32le!(a: fx.vget_lane_u32(b: 0)) fa = fx } } else { - iterate (c = args.curr, p = args.prev)(length: 4, advance: 4, unroll: 2) { - fb = fb.create_vdup_n_u32(a: p.peek_u32le()) - fx = fx.create_vdup_n_u32(a: c.peek_u32le()) + iterate (curr = args.curr, prev = args.prev)(length: 4, advance: 4, unroll: 2) { + fb = fb.create_vdup_n_u32(a: prev.peek_u32le()) + fx = fx.create_vdup_n_u32(a: curr.peek_u32le()) fx = fx.vadd_u8(b: fa.vhadd_u8(b: fb)) - c.poke_u32le!(a: fx.vget_lane_u32(b: 0)) + curr.poke_u32le!(a: fx.vget_lane_u32(b: 0)) fa = fx } } @@ -89,8 +91,9 @@ pri func decoder.filter_4_distance_3_arm_neon!(curr: slice base.u8, prev: slice // // Differences between that function and this one are marked with a §. - var c : slice base.u8 - var p : slice base.u8 + var curr : slice base.u8 + var prev : slice base.u8 + var fa : base.arm_neon_64 var fb : base.arm_neon_64 var fc : base.arm_neon_64 @@ -106,9 +109,9 @@ pri func decoder.filter_4_distance_3_arm_neon!(curr: slice base.u8, prev: slice var pickb : base.arm_neon_64 // § The advance is 3, not 4. - iterate (c = args.curr, p = args.prev)(length: 4, advance: 3, unroll: 2) { - fb = fb.create_vdup_n_u32(a: p.peek_u32le()) - fx = fx.create_vdup_n_u32(a: c.peek_u32le()) + iterate (curr = args.curr, prev = args.prev)(length: 4, advance: 3, unroll: 2) { + fb = fb.create_vdup_n_u32(a: prev.peek_u32le()) + fx = fx.create_vdup_n_u32(a: curr.peek_u32le()) fafb = fa.vaddl_u8(b: fb) fcfc = fc.vaddl_u8(b: fc) pa = fb.vabdl_u8(b: fc) @@ -122,14 +125,14 @@ pri func decoder.filter_4_distance_3_arm_neon!(curr: slice base.u8, prev: slice b: picka.vbsl_u8(b: fa, c: pickb.vbsl_u8(b: fb, c: fc))) // § poke_u24le replaces poke_u32le. - c.poke_u24le!(a: fx.vget_lane_u32(b: 0)) + curr.poke_u24le!(a: fx.vget_lane_u32(b: 0)) fc = fb fa = fx } else (length: 3, advance: 3, unroll: 1) { // § peek_u24le_as_u32 replaces peek_u32le. - fb = fb.create_vdup_n_u32(a: p.peek_u24le_as_u32()) + fb = fb.create_vdup_n_u32(a: prev.peek_u24le_as_u32()) // § peek_u24le_as_u32 replaces peek_u32le. - fx = fx.create_vdup_n_u32(a: c.peek_u24le_as_u32()) + fx = fx.create_vdup_n_u32(a: curr.peek_u24le_as_u32()) fafb = fa.vaddl_u8(b: fb) fcfc = fc.vaddl_u8(b: fc) pa = fb.vabdl_u8(b: fc) @@ -143,7 +146,7 @@ pri func decoder.filter_4_distance_3_arm_neon!(curr: slice base.u8, prev: slice b: picka.vbsl_u8(b: fa, c: pickb.vbsl_u8(b: fb, c: fc))) // § poke_u24le replaces poke_u32le. - c.poke_u24le!(a: fx.vget_lane_u32(b: 0)) + curr.poke_u24le!(a: fx.vget_lane_u32(b: 0)) // § These assignments are unnecessary; this is the last iteration. // fc = fb // fa = fx @@ -153,8 +156,9 @@ pri func decoder.filter_4_distance_3_arm_neon!(curr: slice base.u8, prev: slice pri func decoder.filter_4_distance_4_arm_neon!(curr: slice base.u8, prev: slice base.u8), choose cpu_arch >= arm_neon, { - var c : slice base.u8 - var p : slice base.u8 + var curr : slice base.u8 + var prev : slice base.u8 + var fa : base.arm_neon_64 var fb : base.arm_neon_64 var fc : base.arm_neon_64 @@ -169,9 +173,9 @@ pri func decoder.filter_4_distance_4_arm_neon!(curr: slice base.u8, prev: slice var picka : base.arm_neon_64 var pickb : base.arm_neon_64 - iterate (c = args.curr, p = args.prev)(length: 4, advance: 4, unroll: 2) { - fb = fb.create_vdup_n_u32(a: p.peek_u32le()) - fx = fx.create_vdup_n_u32(a: c.peek_u32le()) + iterate (curr = args.curr, prev = args.prev)(length: 4, advance: 4, unroll: 2) { + fb = fb.create_vdup_n_u32(a: prev.peek_u32le()) + fx = fx.create_vdup_n_u32(a: curr.peek_u32le()) fafb = fa.vaddl_u8(b: fb) // fafb = (fa + fb) fcfc = fc.vaddl_u8(b: fc) // fcfc = (fc + fc) @@ -191,7 +195,7 @@ pri func decoder.filter_4_distance_4_arm_neon!(curr: slice base.u8, prev: slice b: picka.vbsl_u8(b: fa, c: pickb.vbsl_u8(b: fb, c: fc))) - c.poke_u32le!(a: fx.vget_lane_u32(b: 0)) + curr.poke_u32le!(a: fx.vget_lane_u32(b: 0)) fc = fb fa = fx } diff --git a/std/png/decode_filter_fallback.wuffs b/std/png/decode_filter_fallback.wuffs index f8500ee8..3c544ee5 100644 --- a/std/png/decode_filter_fallback.wuffs +++ b/std/png/decode_filter_fallback.wuffs @@ -42,37 +42,39 @@ pri func decoder.filter_1!(curr: slice base.u8), } pri func decoder.filter_1_distance_3_fallback!(curr: slice base.u8) { - var c : slice base.u8 + var curr : slice base.u8 + var fa0 : base.u8 var fa1 : base.u8 var fa2 : base.u8 - iterate (c = args.curr)(length: 3, advance: 3, unroll: 2) { - fa0 = fa0 ~mod+ c[0] - c[0] = fa0 - fa1 = fa1 ~mod+ c[1] - c[1] = fa1 - fa2 = fa2 ~mod+ c[2] - c[2] = fa2 + iterate (curr = args.curr)(length: 3, advance: 3, unroll: 2) { + fa0 = fa0 ~mod+ curr[0] + curr[0] = fa0 + fa1 = fa1 ~mod+ curr[1] + curr[1] = fa1 + fa2 = fa2 ~mod+ curr[2] + curr[2] = fa2 } } pri func decoder.filter_1_distance_4_fallback!(curr: slice base.u8) { - var c : slice base.u8 + var curr : slice base.u8 + var fa0 : base.u8 var fa1 : base.u8 var fa2 : base.u8 var fa3 : base.u8 - iterate (c = args.curr)(length: 4, advance: 4, unroll: 1) { - fa0 = fa0 ~mod+ c[0] - c[0] = fa0 - fa1 = fa1 ~mod+ c[1] - c[1] = fa1 - fa2 = fa2 ~mod+ c[2] - c[2] = fa2 - fa3 = fa3 ~mod+ c[3] - c[3] = fa3 + iterate (curr = args.curr)(length: 4, advance: 4, unroll: 1) { + fa0 = fa0 ~mod+ curr[0] + curr[0] = fa0 + fa1 = fa1 ~mod+ curr[1] + curr[1] = fa1 + fa2 = fa2 ~mod+ curr[2] + curr[2] = fa2 + fa3 = fa3 ~mod+ curr[3] + curr[3] = fa3 } } @@ -154,64 +156,66 @@ pri func decoder.filter_3!(curr: slice base.u8, prev: slice base.u8), } pri func decoder.filter_3_distance_3_fallback!(curr: slice base.u8, prev: slice base.u8) { - var c : slice base.u8 - var p : slice base.u8 + var curr : slice base.u8 + var prev : slice base.u8 + var fa0 : base.u8 var fa1 : base.u8 var fa2 : base.u8 if args.prev.length() == 0 { - iterate (c = args.curr)(length: 3, advance: 3, unroll: 2) { - fa0 = (fa0 / 2) ~mod+ c[0] - c[0] = fa0 - fa1 = (fa1 / 2) ~mod+ c[1] - c[1] = fa1 - fa2 = (fa2 / 2) ~mod+ c[2] - c[2] = fa2 + iterate (curr = args.curr)(length: 3, advance: 3, unroll: 2) { + fa0 = (fa0 / 2) ~mod+ curr[0] + curr[0] = fa0 + fa1 = (fa1 / 2) ~mod+ curr[1] + curr[1] = fa1 + fa2 = (fa2 / 2) ~mod+ curr[2] + curr[2] = fa2 } } else { - iterate (c = args.curr, p = args.prev)(length: 3, advance: 3, unroll: 2) { - fa0 = ((((fa0 as base.u32) + (p[0] as base.u32)) / 2) as base.u8) ~mod+ c[0] - c[0] = fa0 - fa1 = ((((fa1 as base.u32) + (p[1] as base.u32)) / 2) as base.u8) ~mod+ c[1] - c[1] = fa1 - fa2 = ((((fa2 as base.u32) + (p[2] as base.u32)) / 2) as base.u8) ~mod+ c[2] - c[2] = fa2 + iterate (curr = args.curr, prev = args.prev)(length: 3, advance: 3, unroll: 2) { + fa0 = ((((fa0 as base.u32) + (prev[0] as base.u32)) / 2) as base.u8) ~mod+ curr[0] + curr[0] = fa0 + fa1 = ((((fa1 as base.u32) + (prev[1] as base.u32)) / 2) as base.u8) ~mod+ curr[1] + curr[1] = fa1 + fa2 = ((((fa2 as base.u32) + (prev[2] as base.u32)) / 2) as base.u8) ~mod+ curr[2] + curr[2] = fa2 } } } pri func decoder.filter_3_distance_4_fallback!(curr: slice base.u8, prev: slice base.u8) { - var c : slice base.u8 - var p : slice base.u8 + var curr : slice base.u8 + var prev : slice base.u8 + var fa0 : base.u8 var fa1 : base.u8 var fa2 : base.u8 var fa3 : base.u8 if args.prev.length() == 0 { - iterate (c = args.curr)(length: 4, advance: 4, unroll: 1) { - fa0 = (fa0 / 2) ~mod+ c[0] - c[0] = fa0 - fa1 = (fa1 / 2) ~mod+ c[1] - c[1] = fa1 - fa2 = (fa2 / 2) ~mod+ c[2] - c[2] = fa2 - fa3 = (fa3 / 2) ~mod+ c[3] - c[3] = fa3 + iterate (curr = args.curr)(length: 4, advance: 4, unroll: 1) { + fa0 = (fa0 / 2) ~mod+ curr[0] + curr[0] = fa0 + fa1 = (fa1 / 2) ~mod+ curr[1] + curr[1] = fa1 + fa2 = (fa2 / 2) ~mod+ curr[2] + curr[2] = fa2 + fa3 = (fa3 / 2) ~mod+ curr[3] + curr[3] = fa3 } } else { - iterate (c = args.curr, p = args.prev)(length: 4, advance: 4, unroll: 1) { - fa0 = ((((fa0 as base.u32) + (p[0] as base.u32)) / 2) as base.u8) ~mod+ c[0] - c[0] = fa0 - fa1 = ((((fa1 as base.u32) + (p[1] as base.u32)) / 2) as base.u8) ~mod+ c[1] - c[1] = fa1 - fa2 = ((((fa2 as base.u32) + (p[2] as base.u32)) / 2) as base.u8) ~mod+ c[2] - c[2] = fa2 - fa3 = ((((fa3 as base.u32) + (p[3] as base.u32)) / 2) as base.u8) ~mod+ c[3] - c[3] = fa3 + iterate (curr = args.curr, prev = args.prev)(length: 4, advance: 4, unroll: 1) { + fa0 = ((((fa0 as base.u32) + (prev[0] as base.u32)) / 2) as base.u8) ~mod+ curr[0] + curr[0] = fa0 + fa1 = ((((fa1 as base.u32) + (prev[1] as base.u32)) / 2) as base.u8) ~mod+ curr[1] + curr[1] = fa1 + fa2 = ((((fa2 as base.u32) + (prev[2] as base.u32)) / 2) as base.u8) ~mod+ curr[2] + curr[2] = fa2 + fa3 = ((((fa3 as base.u32) + (prev[3] as base.u32)) / 2) as base.u8) ~mod+ curr[3] + curr[3] = fa3 } } } @@ -289,8 +293,9 @@ pri func decoder.filter_4!(curr: slice base.u8, prev: slice base.u8), } pri func decoder.filter_4_distance_3_fallback!(curr: slice base.u8, prev: slice base.u8) { - var c : slice base.u8 - var p : slice base.u8 + var curr : slice base.u8 + var prev : slice base.u8 + var fa0 : base.u32 var fa1 : base.u32 var fa2 : base.u32 @@ -313,8 +318,8 @@ pri func decoder.filter_4_distance_3_fallback!(curr: slice base.u8, prev: slice var pc1 : base.u32 var pc2 : base.u32 - iterate (c = args.curr, p = args.prev)(length: 3, advance: 3, unroll: 1) { - fb0 = p[0] as base.u32 + iterate (curr = args.curr, prev = args.prev)(length: 3, advance: 3, unroll: 1) { + fb0 = prev[0] as base.u32 pp0 = (fa0 ~mod+ fb0) ~mod- fc0 pa0 = pp0 ~mod- fa0 if pa0 >= 0x8000_0000 { @@ -335,11 +340,11 @@ pri func decoder.filter_4_distance_3_fallback!(curr: slice base.u8, prev: slice } else { fa0 = fc0 } - c[0] = c[0] ~mod+ ((fa0 & 0xFF) as base.u8) - fa0 = c[0] as base.u32 + curr[0] = curr[0] ~mod+ ((fa0 & 0xFF) as base.u8) + fa0 = curr[0] as base.u32 fc0 = fb0 - fb1 = p[1] as base.u32 + fb1 = prev[1] as base.u32 pp1 = (fa1 ~mod+ fb1) ~mod- fc1 pa1 = pp1 ~mod- fa1 if pa1 >= 0x8000_0000 { @@ -360,11 +365,11 @@ pri func decoder.filter_4_distance_3_fallback!(curr: slice base.u8, prev: slice } else { fa1 = fc1 } - c[1] = c[1] ~mod+ ((fa1 & 0xFF) as base.u8) - fa1 = c[1] as base.u32 + curr[1] = curr[1] ~mod+ ((fa1 & 0xFF) as base.u8) + fa1 = curr[1] as base.u32 fc1 = fb1 - fb2 = p[2] as base.u32 + fb2 = prev[2] as base.u32 pp2 = (fa2 ~mod+ fb2) ~mod- fc2 pa2 = pp2 ~mod- fa2 if pa2 >= 0x8000_0000 { @@ -385,15 +390,16 @@ pri func decoder.filter_4_distance_3_fallback!(curr: slice base.u8, prev: slice } else { fa2 = fc2 } - c[2] = c[2] ~mod+ ((fa2 & 0xFF) as base.u8) - fa2 = c[2] as base.u32 + curr[2] = curr[2] ~mod+ ((fa2 & 0xFF) as base.u8) + fa2 = curr[2] as base.u32 fc2 = fb2 } } pri func decoder.filter_4_distance_4_fallback!(curr: slice base.u8, prev: slice base.u8) { - var c : slice base.u8 - var p : slice base.u8 + var curr : slice base.u8 + var prev : slice base.u8 + var fa0 : base.u32 var fa1 : base.u32 var fa2 : base.u32 @@ -423,8 +429,8 @@ pri func decoder.filter_4_distance_4_fallback!(curr: slice base.u8, prev: slice var pc2 : base.u32 var pc3 : base.u32 - iterate (c = args.curr, p = args.prev)(length: 4, advance: 4, unroll: 1) { - fb0 = p[0] as base.u32 + iterate (curr = args.curr, prev = args.prev)(length: 4, advance: 4, unroll: 1) { + fb0 = prev[0] as base.u32 pp0 = (fa0 ~mod+ fb0) ~mod- fc0 pa0 = pp0 ~mod- fa0 if pa0 >= 0x8000_0000 { @@ -445,11 +451,11 @@ pri func decoder.filter_4_distance_4_fallback!(curr: slice base.u8, prev: slice } else { fa0 = fc0 } - c[0] = c[0] ~mod+ ((fa0 & 0xFF) as base.u8) - fa0 = c[0] as base.u32 + curr[0] = curr[0] ~mod+ ((fa0 & 0xFF) as base.u8) + fa0 = curr[0] as base.u32 fc0 = fb0 - fb1 = p[1] as base.u32 + fb1 = prev[1] as base.u32 pp1 = (fa1 ~mod+ fb1) ~mod- fc1 pa1 = pp1 ~mod- fa1 if pa1 >= 0x8000_0000 { @@ -470,11 +476,11 @@ pri func decoder.filter_4_distance_4_fallback!(curr: slice base.u8, prev: slice } else { fa1 = fc1 } - c[1] = c[1] ~mod+ ((fa1 & 0xFF) as base.u8) - fa1 = c[1] as base.u32 + curr[1] = curr[1] ~mod+ ((fa1 & 0xFF) as base.u8) + fa1 = curr[1] as base.u32 fc1 = fb1 - fb2 = p[2] as base.u32 + fb2 = prev[2] as base.u32 pp2 = (fa2 ~mod+ fb2) ~mod- fc2 pa2 = pp2 ~mod- fa2 if pa2 >= 0x8000_0000 { @@ -495,11 +501,11 @@ pri func decoder.filter_4_distance_4_fallback!(curr: slice base.u8, prev: slice } else { fa2 = fc2 } - c[2] = c[2] ~mod+ ((fa2 & 0xFF) as base.u8) - fa2 = c[2] as base.u32 + curr[2] = curr[2] ~mod+ ((fa2 & 0xFF) as base.u8) + fa2 = curr[2] as base.u32 fc2 = fb2 - fb3 = p[3] as base.u32 + fb3 = prev[3] as base.u32 pp3 = (fa3 ~mod+ fb3) ~mod- fc3 pa3 = pp3 ~mod- fa3 if pa3 >= 0x8000_0000 { @@ -520,8 +526,8 @@ pri func decoder.filter_4_distance_4_fallback!(curr: slice base.u8, prev: slice } else { fa3 = fc3 } - c[3] = c[3] ~mod+ ((fa3 & 0xFF) as base.u8) - fa3 = c[3] as base.u32 + curr[3] = curr[3] ~mod+ ((fa3 & 0xFF) as base.u8) + fa3 = curr[3] as base.u32 fc3 = fb3 } } diff --git a/std/png/decode_filter_x86_sse42.wuffs b/std/png/decode_filter_x86_sse42.wuffs index 7068e80a..8c53ee1d 100644 --- a/std/png/decode_filter_x86_sse42.wuffs +++ b/std/png/decode_filter_x86_sse42.wuffs @@ -22,19 +22,20 @@ // pri func decoder.filter_1_distance_3_x86_sse42!(curr: slice base.u8), // choose cpu_arch >= x86_sse42, // { -// var c : slice base.u8 +// var curr : slice base.u8 +// // var x128 : base.x86_m128i // var a128 : base.x86_m128i // -// iterate (c = args.curr)(length: 4, advance: 3, unroll: 2) { -// x128.load_u32!(a: c.peek_u32le()) +// iterate (curr = args.curr)(length: 4, advance: 3, unroll: 2) { +// x128.load_u32!(a: curr.peek_u32le()) // x128 = x128._mm_add_epi8(b: a128) // a128 = x128 -// c.poke_u24le!(a: x128.truncate_u32()) +// curr.poke_u24le!(a: x128.truncate_u32()) // } else (length: 3, advance: 3, unroll: 1) { -// x128.load_u32!(a: c.peek_u24le_as_u32()) +// x128.load_u32!(a: curr.peek_u24le_as_u32()) // x128 = x128._mm_add_epi8(b: a128) -// c.poke_u24le!(a: x128.truncate_u32()) +// curr.poke_u24le!(a: x128.truncate_u32()) // } // } // @@ -45,15 +46,16 @@ pri func decoder.filter_1_distance_4_x86_sse42!(curr: slice base.u8), choose cpu_arch >= x86_sse42, { - var c : slice base.u8 + var curr : slice base.u8 + var x128 : base.x86_m128i var a128 : base.x86_m128i - iterate (c = args.curr)(length: 4, advance: 4, unroll: 2) { - x128.load_u32!(a: c.peek_u32le()) + iterate (curr = args.curr)(length: 4, advance: 4, unroll: 2) { + x128.load_u32!(a: curr.peek_u32le()) x128 = x128._mm_add_epi8(b: a128) a128 = x128 - c.poke_u32le!(a: x128.truncate_u32()) + curr.poke_u32le!(a: x128.truncate_u32()) } } @@ -73,8 +75,9 @@ pri func decoder.filter_1_distance_4_x86_sse42!(curr: slice base.u8), pri func decoder.filter_3_distance_4_x86_sse42!(curr: slice base.u8, prev: slice base.u8), choose cpu_arch >= x86_sse42, { - var c : slice base.u8 - var p : slice base.u8 + var curr : slice base.u8 + var prev : slice base.u8 + var x128 : base.x86_m128i var a128 : base.x86_m128i var b128 : base.x86_m128i @@ -83,7 +86,7 @@ pri func decoder.filter_3_distance_4_x86_sse42!(curr: slice base.u8, prev: slice if args.prev.length() == 0 { k128 = k128.create_mm_set1_epi8(a: 0xFE) - iterate (c = args.curr)(length: 4, advance: 4, unroll: 2) { + iterate (curr = args.curr)(length: 4, advance: 4, unroll: 2) { // The predictor, p128, is just half (rounded down) of the previous // pixel, a128. In this branch, b128 stays zero so the average of // a128 and b128 is just half of a128. _mm_avg_epu8 rounds up, but @@ -92,17 +95,17 @@ pri func decoder.filter_3_distance_4_x86_sse42!(curr: slice base.u8, prev: slice // Add the predictor to the residual and, for the next iteration, // set its previous pixel, a128, to this one, x128. - x128.load_u32!(a: c.peek_u32le()) + x128.load_u32!(a: curr.peek_u32le()) x128 = x128._mm_add_epi8(b: p128) a128 = x128 - c.poke_u32le!(a: x128.truncate_u32()) + curr.poke_u32le!(a: x128.truncate_u32()) } } else { k128 = k128.create_mm_set1_epi8(a: 0x01) - iterate (c = args.curr, p = args.prev)(length: 4, advance: 4, unroll: 2) { + iterate (curr = args.curr, prev = args.prev)(length: 4, advance: 4, unroll: 2) { // Load the pixel from the row above. - b128.load_u32!(a: p.peek_u32le()) + b128.load_u32!(a: prev.peek_u32le()) // The predictor, p128, is the average (rounded down) of the // previous pixel, a128, and the pixel above, b128. @@ -115,10 +118,10 @@ pri func decoder.filter_3_distance_4_x86_sse42!(curr: slice base.u8, prev: slice // Add the predictor to the residual and, for the next iteration, // set its previous pixel, a128, to this one, x128. - x128.load_u32!(a: c.peek_u32le()) + x128.load_u32!(a: curr.peek_u32le()) x128 = x128._mm_add_epi8(b: p128) a128 = x128 - c.poke_u32le!(a: x128.truncate_u32()) + curr.poke_u32le!(a: x128.truncate_u32()) } } } @@ -137,8 +140,9 @@ pri func decoder.filter_4_distance_3_x86_sse42!(curr: slice base.u8, prev: slice // // Differences between that function and this one are marked with a §. - var c : slice base.u8 - var p : slice base.u8 + var curr : slice base.u8 + var prev : slice base.u8 + var x128 : base.x86_m128i var a128 : base.x86_m128i var b128 : base.x86_m128i @@ -151,8 +155,8 @@ pri func decoder.filter_4_distance_3_x86_sse42!(curr: slice base.u8, prev: slice var z128 : base.x86_m128i // § The advance is 3, not 4. - iterate (c = args.curr, p = args.prev)(length: 4, advance: 3, unroll: 2) { - b128.load_u32!(a: p.peek_u32le()) + iterate (curr = args.curr, prev = args.prev)(length: 4, advance: 3, unroll: 2) { + b128.load_u32!(a: prev.peek_u32le()) b128 = b128._mm_unpacklo_epi8(b: z128) pa128 = b128._mm_sub_epi16(b: c128) pb128 = a128._mm_sub_epi16(b: c128) @@ -166,19 +170,19 @@ pri func decoder.filter_4_distance_3_x86_sse42!(curr: slice base.u8, prev: slice mask: smallest128._mm_cmpeq_epi16(b: pb128))._mm_blendv_epi8( b: a128, mask: smallest128._mm_cmpeq_epi16(b: pa128)) - x128.load_u32!(a: c.peek_u32le()) + x128.load_u32!(a: curr.peek_u32le()) x128 = x128._mm_unpacklo_epi8(b: z128) x128 = x128._mm_add_epi8(b: p128) a128 = x128 c128 = b128 x128 = x128._mm_packus_epi16(b: x128) // § poke_u24le replaces poke_u32le. - c.poke_u24le!(a: x128.truncate_u32()) + curr.poke_u24le!(a: x128.truncate_u32()) // § The length and advance are both 3, not 4. } else (length: 3, advance: 3, unroll: 1) { // § peek_u24le_as_u32 replaces peek_u32le. - b128.load_u32!(a: p.peek_u24le_as_u32()) + b128.load_u32!(a: prev.peek_u24le_as_u32()) b128 = b128._mm_unpacklo_epi8(b: z128) pa128 = b128._mm_sub_epi16(b: c128) pb128 = a128._mm_sub_epi16(b: c128) @@ -193,7 +197,7 @@ pri func decoder.filter_4_distance_3_x86_sse42!(curr: slice base.u8, prev: slice b: a128, mask: smallest128._mm_cmpeq_epi16(b: pa128)) // § peek_u24le_as_u32 replaces peek_u32le. - x128.load_u32!(a: c.peek_u24le_as_u32()) + x128.load_u32!(a: curr.peek_u24le_as_u32()) x128 = x128._mm_unpacklo_epi8(b: z128) x128 = x128._mm_add_epi8(b: p128) // § These assignments are unnecessary; this is the last iteration. @@ -201,15 +205,16 @@ pri func decoder.filter_4_distance_3_x86_sse42!(curr: slice base.u8, prev: slice // c128 = b128 x128 = x128._mm_packus_epi16(b: x128) // § poke_u24le replaces poke_u32le. - c.poke_u24le!(a: x128.truncate_u32()) + curr.poke_u24le!(a: x128.truncate_u32()) } } pri func decoder.filter_4_distance_4_x86_sse42!(curr: slice base.u8, prev: slice base.u8), choose cpu_arch >= x86_sse42, { - var c : slice base.u8 - var p : slice base.u8 + var curr : slice base.u8 + var prev : slice base.u8 + var x128 : base.x86_m128i var a128 : base.x86_m128i var b128 : base.x86_m128i @@ -221,9 +226,9 @@ pri func decoder.filter_4_distance_4_x86_sse42!(curr: slice base.u8, prev: slice var smallest128 : base.x86_m128i var z128 : base.x86_m128i - iterate (c = args.curr, p = args.prev)(length: 4, advance: 4, unroll: 2) { + iterate (curr = args.curr, prev = args.prev)(length: 4, advance: 4, unroll: 2) { // Load the pixel from the row above. - b128.load_u32!(a: p.peek_u32le()) + b128.load_u32!(a: prev.peek_u32le()) // Convert from u8 to i16 by unpacking it with zeroes. b128 = b128._mm_unpacklo_epi8(b: z128) @@ -258,12 +263,12 @@ pri func decoder.filter_4_distance_4_x86_sse42!(curr: slice base.u8, prev: slice // its previous pixels, a128 and c128, to x128 and b128. // // Again, convert between u8 and i16 by unpacking and re-packing. - x128.load_u32!(a: c.peek_u32le()) + x128.load_u32!(a: curr.peek_u32le()) x128 = x128._mm_unpacklo_epi8(b: z128) x128 = x128._mm_add_epi8(b: p128) a128 = x128 c128 = b128 x128 = x128._mm_packus_epi16(b: x128) - c.poke_u32le!(a: x128.truncate_u32()) + curr.poke_u32le!(a: x128.truncate_u32()) } }